id
stringlengths 27
33
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 13
1.81M
|
|---|---|---|---|
warning/0001/nlin0001008.html | ar5iv | text | # Relaxation oscillations and negative strain rate sensitivity in the Portevin - Le Chatelier effect
## I Introduction
The Portevin - Le Chatelier effect or the jerky flow has been an object of continued interest in materials science for quite some time. The phenomenon refers to an instability seen in the form of repeated stress drops followed by periods of reloading observed when tensile specimens are deformed in a certain range of strain rates and temperatures . The effect is seen in many interstitial and substitutional metallic alloys (commercial aluminium, brass, alloys of aluminium and magnesium ). Each of the load drops is related to the formation and propagation of dislocation bands . The traditional picture of the instability is that it stems from dynamic interaction of mobile dislocations with solute atoms and is called as dynamic strain ageing . It is this that is expected to lead to negative strain rate sensitivity (SRS) of the flow stress .
Plastic flow is intrinsically nonlinear and therefore methods of nonlinear dynamics have a natural role to play in understanding plastic instabilities . Use of these new techniques have led to insights which were hitherto not possible. The first attempt to look at the phenomenon from a nonlinear dynamical angle was taken by Ananthakrishna and coworkers, which offers a natural basis for the description of the time dependent aspects of the PLC effect which were ignored in the earlier theories . Their model allows for explicit inclusion and interplay of different time scales inherent in the dynamics of dislocations. These authors show that the occurrence of the instability is a consequence of Hopf bifurcation as a function of the applied strain rate. Many known features of the PLC effect such as the existence of a window of strain rates and temperatures within which it occurs, etc., were correctly reproduced. Most importantly, and for the first time, the negative SRS was shown to emerge naturally in the model, as a result of nonlinear interaction of the participating defects. It also predicts the existence of chaotic stress drops in a range of strain rates, which has been recently verified . Even the number of degrees of freedom estimated turn out to be the same as in the model offering justification for ignoring spatial degrees of freedom (See Ref. 16 also). Further dynamical analysis of the model for the creep case has shown that the temperature dependence of the strain bursts is consistent with experimental findings.
The study of the PLC effect from a dynamical angle has been useful in elucidating several features; but it has also brought certain other issues into sharp focus which were hitherto not investigated in depth. This paper is intended to address one such issue related to the time scales relevant to the dynamics of the PLC effect within the framework of the above model. This is reflected in the two well known attributes of the PLC effect, namely, the negative strain rate behavior of the flow stress and the stick-slip or relaxational nature of the dynamics reflected in the stress time series. In order to motivate, we will present arguments showing that conflicting conclusions can be arrived at when one analyses this question starting from these two angles.
We start with a discussion of the well accepted physical picture of the PLC effect namely the dynamic strain ageing. At a qualitative level, theories of strain ageing already have an implicit suggestion that the occurrence of the negative SRS is related to the competition of diffusive time scale and the waiting time of dislocations at obstacles , even though, there is no dynamics involved in these theories. The physical picture of strain ageing is as follows. At small velocities, solute atoms have enough time to diffuse to the temporarily arrested dislocations thus providing additional pinning thereby impeding their breakaway from localized obstacles. Due to the constant applied strain rate, the overall stress to keep the dislocations moving increases bringing the stress to a threshold level beyond which dislocations break away. At high velocities, such additional pinning due to solute atoms is not possible since the waiting time of dislocations at obstacles is too short for the diffusion to occur. A schematic diagram of the SRS is shown in Fig. 1. In the language of the stick-slip dynamics, the branch $`A^{}B^{}`$ corresponds to the stick-state and $`B^{}C^{}`$ to the slip-state (Fig. 1). The slope of stress verses velocity (’friction coefficient’) at low velocities is much higher than that corresponding to high velocities since in the former case, solute atoms have to be dragged along with the dislocations, while in the latter case there is no solute atmosphere. Based on physical considerations, these two stable branches are assumed to be separated by an unstable branch with a negative slope to reflect the nonaccessible nature.
The occurrence of the negative flow rate characteristic is not just limited to the PLC effect. With particular reference to the conceptual aspects of the negative branch, we cite two other mechanical systems, namely, the peeling of an adhesive tape and frictional sliding of a block of material over another which shows the inaccessible nature of the negative slope branch. However, in the case of the PLC effect, there has been attempts to obtain experimental points in this domain of strain rates which has lead to some confusion about the measurability of the negative slope branch of the SRS which will be discussed later (Section 5). Therefore, it is important to understand the meaning of the negative branch of SRS from a dynamical point of view with reference to the PLC effect which hopefully will lead to a better understanding of other stick-slip phenomenon.
Returning to the PLC effect, Penning was the first to recognize that the negative SRS could be used to explain the strain rate jumps observed in experiments. Subsequently, the negative SRS feature has been used as an input into several theories . In particular, it has helped to successfully explain the nature of yield drops occurring in different regimes of strain rate and temperature . Pertinent to the our discussion of time scales inherent to the PLC effect, we note that in such theories, two slow time scales corresponding to the two dissipative branches, $`A^{}B^{}`$ and $`C^{}D^{}`$, show up along with two fast time scales corresponding to the jumps ( $`B^{}C^{}`$ and $`D^{}A^{}`$) in the strain rates. A more direct reflection of the time scales inherent to the dynamics of the PLC effect can be deduced from stress-strain curve.
To facilitate discussion of time scales involved in an experimental stress-time series, consider the so called machine equation written as
$$\dot{\sigma }_a=\kappa [\dot{ϵ}_a\dot{ϵ}_p]$$
(1)
where $`\sigma _a,\dot{ϵ}_a,\dot{ϵ}_p`$ refer to the stress, applied strain rate and plastic strain rate respectively and $`\kappa `$ is the combined elastic constant of the machine and the sample. We note that only stress is monitored by the load sensing devise. Apart from this, it is possible to measure the plastic strain rate using strain gauges or using cinematographic techniques . Using Eq.(1), we can now identify different time scales in an experimental curve. Consider a typical stress-strain curve for an applied strain rate of 8.3 x $`10^5s^1`$ for the PLC effect in $`Cu10\%Al`$ is shown in Fig.2. (For our purpose, we will ignore the nonperiodic nature.) From the saw-tooth shape of the stress - strain series, two points emerge: (a) the positive slope of $`\sigma _aϵ_a`$ curve is close to the elastic loading rate ( $`\kappa \dot{ϵ}_a`$), and (b) the duration of each stress drop is very short. From Eq.(1), we see that the stress drop duration is the time interval during which $`\dot{ϵ_p}(t)`$ larger than $`\dot{ϵ_a}`$. We also note that the changes in slopes, when they occur, are abrupt (within the recording accuracy of 0.05$`s^1`$). Knowing that $`\dot{ϵ}_p`$ is proportional to the mobile dislocation density and using Eq. (1), we can see that the mobile dislocation density should be nearly constant in the rising part of $`\sigma ϵ_a`$ curve and therefore correspond to the stick-state. Further, the short duration of the stress drop should be a result of rapid multiplication of mobile dislocation and therefore corresponds to the slip-state. This must be followed by the process of immobilization of dislocations. However, the abrupt change in the slope ( from negative to positive) also implies that immobilization time scale is also fast. Indeed, as is clear from Fig. 2, it is not possible to separate out these two fast time scales. Thus, from the stress-strain curve, we see only one slow time scale and two fast time scales which is in apparent conflict with what was argued from the schematic diagram of the orbit $`A^{}B^{}C^{}D^{}`$ in Fig. 1. This discussion raises several questions relating the origin of these time scales causing jumps in dislocations densities which needs to understood if the above inconsistency has to be resolved. Specifically: (i) What is dynamical mechanism which keeps the mobile dislocation density constant and in low levels for long intervals of time? (ii) What are the mechanisms for rapid multiplication and immobilization of mobile dislocations ? As we shall see, resolving these issues will also help us to interpret the negative SRS in an appropriate way. Further, associating various time scales with different branches of the SRS provides a better insight into the stick-slip dynamics of the PLC effect.
Analysis of time scales can be best understood from a dynamical point of view. It is well known that relaxation oscillations are at the root of stick-slip behavior. One of the standard ways of understanding relaxation oscillations is by analyzing the slow manifold geometry of the underlying model. Following this, we shall attempt to understand the above issues from the point of view of relaxation oscillations. The paper is organized as follows. In section 2, we briefly introduce the model along with the known results. In section 3, we state some bifurcation features relevant for further discussion. In section 4, we show that the nature of relaxation oscillation in the model is atypical and is due to the atypical nature of bent nature of the slow manifold of the model. This analysis further helps us to understand the dynamical basis of different time scales relevant to the PLC effect. In section 5, we discuss the concept of negative SRS and its measurement in some detail to highlight the meaning of the negative branch. Using the geometry of the slow manifold, we calculate the dependence of stress on the plastic strain rate and show the connection between the various branches of the SRS and the time scales operating in different regions of slow manifold which in turn helps us to resolve the inconsistency of time scales. Section 6 is devoted to discussion and conclusions.
## II A Dynamical Model for Jerky Flow
The model consists of mobile dislocations and immobile dislocations and another type which mimics the Cottrell’s type, which are dislocations with clouds of solute atoms . Let the corresponding densities be $`N_m`$, $`N_{im}`$ and $`N_i`$, respectively. The rate equations for the densities of dislocations are:
$`\dot{N}_m`$ $`=`$ $`\theta V_mN_m\beta N_m^2\beta N_mN_{im}+\gamma N_{im}\alpha _mN_m,`$ (2)
$`\dot{N}_{im}`$ $`=`$ $`\beta N_m^2\beta N_{im}N_m\gamma N_{im}+\alpha _iN_i,`$ (3)
$`\dot{N}_i`$ $`=`$ $`\alpha _mN_m\alpha _iN_i.`$ (4)
The overdot, here, refers to the time derivative. The first term in Eq. (2) is the rate of production of dislocations due to cross glide with a rate constant $`\theta `$. $`V_m`$ is the velocity of the mobile dislocations which in general depends on some power of the applied stress $`\sigma _a`$. The second term refers to the annihilation or immobilization of two mobile dislocations. The third term also represents the annihilation of a mobile dislocation with an immobile one. The fourth term represents the remobilization of the immobile dislocations due to stress or thermal activation ( see $`\gamma N_{im}`$ in Eq. 3). The last term represents the immobilization of mobile dislocations either due to solute atoms or due to other pinning centers. $`\alpha _m`$ refers to the concentration of the solute atoms which participate in slowing down the mobile dislocations. Once a mobile dislocation starts acquiring solute atoms we regard it as a new type of dislocation, namely the Cottrell’s type $`N_i`$, i.e, the incoming term in Eq. (4). As they acquire more and more solute atoms they slow down and eventually stop the dislocation entirely. At this point, they are considered to have transformed to $`N_{im}`$ (loss term in Eq. (4) and a gain term in Eq. (3)). Indeed, the whole process can be mathematically represented by defining $`N_i=_{\mathrm{}}^tK(tt^{})N_m(t^{})𝑑t^{}=\alpha _m_{\mathrm{}}^texp(\alpha _i(tt^{}))N_m(t^{})𝑑t^{}`$ which represents the entire process of slowing down of $`N_m`$ in an exponential fashion with a time constant $`\alpha _i`$. (The choice of $`K`$ as having a exponential form is obviously a simplification of the actual process.)
These equations should be dynamically coupled to the machine equation which now takes the form
$`\dot{\sigma _a}=\kappa (\dot{ϵ_a}B_0N_mV_m),`$ (5)
where $`V_m`$ is the velocity of mobile dislocations and $`B_0`$ is the Burgers vector. A power law dependence of $`V_m=V_0\left(\sigma _a/\sigma _0\right)^m`$ is used. These equations can be cast into a dimensionless form by using scaled variables $`x=N_m\left(\beta /\gamma \right)`$, $`y=N_{im}\left(\beta /\theta V_0\right)`$, $`z=N_i\left(\beta \alpha _i/\gamma \alpha _m\right)`$, $`\tau =\theta V_0t`$ and $`\varphi =\sigma _a/\sigma _0`$.
$`\dot{x}`$ $`=`$ $`\varphi ^mxaxb_0x^2xy+y,`$ (6)
$`\dot{y}`$ $`=`$ $`b_0\left(b_0x^2xyy+az\right),`$ (7)
$`\dot{z}`$ $`=`$ $`c(xz),`$ (8)
$`\dot{\varphi }`$ $`=`$ $`d\left(e\varphi ^mx\right),`$ (9)
Here $`a=\alpha _m/\theta V_0,b_0=\gamma /\theta V_0,c=\alpha _i/\theta V_0`$, $`\kappa =(\theta \beta \sigma _0d/\gamma B_0)`$ and $`e=(\dot{ϵ_a}\beta /B_0V_0\gamma )`$. For these set of equations there is only one steady state which is stable. There is a range of the parameters $`a,b,c,d,m`$ and $`e`$ for which the linearized equations are unstable. In this range $`x,y,z`$ and $`\varphi `$ are oscillatory.
Among these physically relevant parameters, we study the behavior of the model as a function of most important parameters namely the applied strain rate $`e`$ and the velocity exponent $`m`$. The values of other parameters are kept fixed at $`a=0.7,b_0=0.002,c=0.008,`$ and $`d=0.0001`$. As can be verified these equations exhibit a strong volume contraction in the four dimensional phase space. We note that there are widely differing time scales corresponding to $`a,b_0,c`$ and $`d`$ (in the decreasing order) in the dynamics of the model. For this reason, the equations are stiff and the numerical integration routines were designed specifically to solve this set of equations. We have used a variable order Taylor series expansion method as the basic integration technique where the coefficients are determined using a recursive algorithm.
## III Summary of bifurcation exhibited the model
The model exhibits a rich variety of dynamics such as period bubbling, period doubling, and complex bifurcation sequences referred to as mixed mode oscillations(MMOs) in literature. Here, we will briefly recall only those aspects of the bifurcation diagram relevant for the discussion of relaxation oscillations. The gross features of the phase diagram in the $`(m,e)`$ plane are shown in Fig. 3. In our discussion, we use $`e`$ as primary control parameter and $`m`$ as the unfolding parameter. For values of $`m>m_d6.8`$, the equilibrium fixed point of the system of equations, denoted by ($`x_0,y_0,z_0,\varphi _0`$), is stable. Both $`x_0`$ and $`z_0`$ are $`e/2`$ and $`y_0`$ and $`\varphi _0`$ are independent of $`e`$. At $`m=m_d`$, we have a degenerate Hopf bifurcation as a function of $`e`$. For values less than $`m_d`$, we have a back-to-back Hopf bifurcation, the first occurring at $`e=e_{c_1}`$ and the reverse at $`e=e_{c_2}`$. The periodic orbit connecting these back-to-back Hopf bifurcations is referred to as principal periodic orbit. The dynamics of the system is essentially bounded by these two Hopf bifurcations. In Fig. 3, the broken line represents the Hopf bifurcation and the dotted lines correspond to the locus of first three successive period doubling bifurcations. The inner, continuous lines represent the locus of saddle node bifurcations corresponding to period 3, 4 and 5 which are the first three dominant periodic windows in the alter nating periodic chaotic bifurcation sequence. Complex bifurcation sequences, characterized by alternate periodic-chaotic sequences are seen in the hatched region of the parameter space. A codimension two bifurcation point in the form of a cusp at $`(e_c,m_c)`$ formed by the merging of the locus of two saddle node bifurcations of the principal periodic orbit (represented by bold lines) is shown as filled diamond in Fig. 2. Phase plots and other diagrams have been obtained by plotting the maxima of any one of the variables $`x`$,$`y`$,$`z`$ or $`\varphi `$ as a function of the control parameters ($`e,m`$).
## IV Mechanism of relaxation oscillations
One characteristic feature of the dynamics of the system is its strong relaxational nature. This feature persists even in regions of the $`(m,e)`$ plane wherein complex periodic-chaotic oscillations are seen (hatched region in Fig. 3). The presence of relaxations oscillations and complex periodic chaotic oscillations are interrelated and are a result of the geometry of the slow manifold. (For details see Ref. .) Relaxation oscillations that manifest in the model is a type of relaxation oscillation wherein the fast variable takes on large values for a short time after which it assumes small values of the same order of magnitude as that of the slow variables. The time spent by the fast variable in the part of phase space where the amplitude is small is a substantial portion of the period of the orbit. A typical plot of $`x(t)`$ ( continuous line) and $`z(t)`$ ( dotted line) are shown in the inset of Fig. 4 for $`e=200.0`$ and $`m=1.2`$.
To understand the nature of the relaxation oscillations, we first study the structure of the slow manifold ($`S`$) and the behavior of the trajectories visiting different regions of $`S`$. The slow manifold of a multiple timescale dynamical system is given by the surface spanning the time invariant solutions of the fast variable. In our case, it is given by
$`\dot{x}=g(x,y,\varphi )`$ $`=`$ $`b_0x^2+x\delta +y=0`$ (10)
with $`\delta =\varphi ^mya`$. Here, the slow variables $`y`$ and $`\varphi `$ (and therefore $`\delta `$) are regarded as parameters. Further, as we will see below, it is simpler to deal with the structure of the slow manifold in terms of the $`\delta `$ instead of both $`y`$ and $`\varphi `$. Then, the physically allowed solution of the above equation is
$$x=\frac{\delta +\sqrt{\delta ^2+4b_0y}}{2b_0}$$
(11)
where $`\delta `$ can take on both positive and negative values. Noting that $`b_0`$ is small and therefore $`\delta ^24b_0y`$, two distinct cases arise corresponding to $`\delta >0`$ and $`\delta <0`$ for which $`x\delta /b_0`$ and $`xy/\delta `$ respectively. Further, since the slow variable $`\varphi `$ and $`y`$ take on values of the order of unity, the range of $`\delta =\delta (y,\varphi )`$ is of the same order as that of $`\varphi `$ and $`y`$ (as is evident from Figs. 4 and 5). Thus, we see that $`xy/\delta `$ is small and $`x\delta /b_0`$ is large. For values around $`\delta =0`$ and positive, we get $`x\left(y/b_0\right)^{1/2}`$.
The bent-slow manifold structure along with the two portions of the slow manifold, namely, $`S_1`$ ($`\delta >0`$) and $`S_2`$ ($`\delta <0`$) are shown by bold lines in the $`(x,\delta )`$ plane in Fig. 4. We have also shown a trajectory corresponding to a mono-periodic relaxation oscillation ($`m=1.2`$ and $`e=200.0`$) by a thin line. As can be seen, the trajectory spends most of the time on $`S_1`$ and $`S_2`$. A local stability analysis for points on $`S_1`$ and $`S_2`$ shows that $`g/x=\delta 2b_0x`$ is negative implying that the rate of growth of $`x`$ is damped. Hence these regions, $`S_1`$ and $`S_2`$ will be referred to as attracting or ”stable”. For points below the line $`2b_0x=\delta `$ ($`\delta >0`$), $`g/x>0`$ and hence we call this region as ’unstable’ (shaded region of Fig. 4). Even then, the trajectory starting on $`S_2`$ does continue in the direction of increasing $`\delta `$ beyond $`\delta =0`$. We note that this region is not a part of the slow manifold. Once the trajectory is in this region, it moves up rapidly in the $`x`$ direction (due to the ‘unstable’ nature) until it reaches $`x=\delta /2b_0`$ line, thereafter, the trajectory quickly settles down on to the $`S_1`$ part of the slow manifold as $`g/x`$ becomes negative. As the trajectory descends on $`S_1`$ approaching $`S_2`$, we see that the trajectory deviates away from $`S_1`$. This happens when the value of $`x`$ is such that $`2b_0x<\delta `$, i.e., $`g/x>0`$. Thus, points on $`S_1`$ satisfying this condition are locally unstable. Thus, the trajectory makes a jump from $`S_1`$ to $`S_2`$ in a short time. This roughly explains the origin of the relaxation oscillation in terms of the reduced variables $`\delta `$ and $`x`$.
The actual dynamics is in a higher dimensional space and a proper understanding will involve analysis of the movement of the trajectory in the appropriate space. Moreover, unlike the standard $`S`$ shaped manifold with upper and lower attracting pleats with the repulsive (unstable) branch, in our model, both branches of the bent-slow manifold are connected, and there is no repulsive branch of the slow manifold. Thus, the mechanism of jumping of the orbit from $`S_2`$ to $`S_1`$ is not clear. In order to understand this, consider a 3-d plot of the trajectory shown in Fig. 5. The region $`S_2`$ corresponding to small values of $`x`$ lies more or less on the $`y\varphi `$ plane and the region $`S_1`$ corresponding to large values of $`x`$ is nearly normal to the $`y\varphi `$ plane due to the large $`b_0^1`$ factor. Regions $`S_1`$ and $`S_2`$ are demarcated by the ‘fold curve’ given by $`\delta =\varphi ^mya=0`$ which dominantly lies in the $`y\varphi `$ plane. The rapidly growing nature of the trajectory lying to right of the ‘fold curve’ is due to $`g/x>0`$.
The principal features of the relaxation oscillations that we need to explain are: a) very slow time scale for evolution on $`S_2`$, b) fast transition from $`S_2`$ to $`S_1`$ and c) evolution on $`S_1`$. In order to understand this, it is necessary to establish how the trajectory (viz, $`x`$,$`y`$,$`z`$ and $`\varphi `$) visits various regions of the slow manifold in a sequential way. However, our emphasis is more on those aspects of relaxation oscillation pertaining to the issue of time scales raised in the Introduction, i.e., the timescales involved in the stress - time curve (Fig. 1). (A detailed investigation on the behavior of trajectories on this slow manifold has also been carried out. See Ref. .) However, to understand the dependence of stress $`\varphi (t)`$, we would also require information of $`y`$ which in turn depends on $`z`$. In order to understand this, we shall analyze Eqs. (7) and (9) by recasting them in terms of $`\delta `$ in various regions of $`S`$. This will help to understand the general features of the flow viz., on $`S_2`$, just outside $`S_2`$ and on $`S_1`$. In the whole analysis, it would be helpful to keep in mind the range of values of $`x,y,z`$ and $`\varphi `$, shown in Figs. 4 and 5, in particular their values as the trajectory enters and leaves $`S_1`$.
First, consider rewriting Eq. (7) valid on the slow manifold $`S`$ in terms of $`\delta `$:
$$\dot{y}=b_0(x\delta xy+az).$$
(12)
The presence of the $`z`$ variable in Eq. (12) poses some problems. Using detailed arguments based on the knowledge of the magnitude of $`x`$ and $`z`$ just inside, on and outside $`S_2`$, it can be shown that the trajectory enters $`S_2`$ at small values of $`y`$ and leaves $`S_2`$ at relatively larger values. Further one can show that there is a turning point for $`y`$ on $`S_2`$ (see Fig. 5). For details, see Ref.
With this information on the evolution of $`y`$ on $`S_2`$, we now consider the changes in $`\varphi `$ as the trajectory enters and leaves $`S_2`$. From Eq.(9), it is clear that an yield drop starts when $`x`$ is large (i.e., when $`x\delta /b_0`$ on $`S_1`$) and ends when $`x`$ is close to minimum, when the trajectory is on $`S_2`$, which implies that $`\varphi `$ is small when the trajectory enters $`S_2`$. Using the value of $`x=y/|\delta |`$ on $`S_2`$ in Eq. (9), we find that $`e\varphi ^my/\left|\delta \right|`$, since $`y`$ is near its minimum value as the trajectory enters $`S_2`$. Thus, $`\varphi `$ increases linearly from small values of $`\varphi `$ at a rate close to $`de1`$. We recall that the loading rate in the experimental stress-strain curve was $`\kappa \dot{ϵ_a}(de`$ in scaled variables), which now can be understood as due to the structure of the slow manifold. This is a direct consequence of the fact that the magnitude of $`x`$ remains constant as $`\dot{x}0`$ for the entire interval the trajectory on $`S_2`$. This is consistent with what we argued from the stress - time plot (Fig. 2), namely, the mobile dislocation density should be constant during the loading period. As the trajectory moves into $`S_2`$, $`y`$ goes through a maximum whereas $`\varphi `$ continues to increase since $`xy/|\delta |`$ remains small. However, as the trajectory is just outside $`S_2`$ for which $`x\left(y/b_0\right)^{1/2}`$ for $`\delta >0`$ and small, $`\varphi ^m\left(y/b_0\right)^{1/2}e`$, since $`\varphi `$ and $`y`$ are relatively large which implies that $`\varphi `$ is about to decrease. The above discussion on $`\dot{y}`$ and $`\dot{\varphi }`$ for region just outside and inside the fold curve also gives us the direction of movement of the trajectory in this region, namely, it enters $`S_2`$ in the region corresponding to small values of $`y`$ and $`\varphi `$, and makes an exit for relatively larger values of $`\varphi `$ and $`y`$ (compared to their values as the trajectory enters $`S_2`$). Further, as $`\dot{x}0`$, we see that the dynamics on $`S_2`$ is controlled by the slow variables.
Finally, just to the right of $`\delta =0`$ line, $`\dot{x}x\delta `$, with $`\delta `$ very small, which suggests that the time constant is small. Thus, the growth of $`x`$ is slow in the neighbourhood of $`\delta =0`$, and is tangential to the $`S_2`$ plane even in the ‘unstable’ region. However, once the trajectory moves away from $`\delta =0`$, the growth of the trajectory is controlled by $`g/x`$ and hence the time scale of growth of $`x`$ is of the order of $`\delta ^1`$ which is of the order of unity explaining the short time span of the stress drop seen in Fig. 2. This also explains why the trajectory tends to leave stable portion of the slow manifold $`S_2`$ and move into the ‘unstable’ region.
Once in the ‘unstable’ region, the value of $`x`$ continues to grow in this region of the phase space as can be seen from Eq. (9) until the value of $`x`$ is such that $`\varphi ^mx=e`$ is satisfied. Beyond this value of $`\varphi `$, $`\dot{\varphi }`$ is negative. Thus, the trajectory leaving $`S_2`$ eventually falls onto the $`S_1`$ part of the slow manifold. We can again evaluate $`\dot{y}`$ and $`\dot{\varphi }`$ just as the trajectory reaches $`S_1`$. Using $`x\delta /b_0`$ in Eq. (7), it can be shown that $`y`$ decreases. Now, consider the equation for $`\varphi `$. Using $`x\delta /b_0`$ on $`S_1`$, we see that $`\varphi ^m\delta /b_0>e`$. Thus $`\dot{\varphi }<0`$ when the trajectory reaches $`S_1`$ with a time constant $`d/b_0`$ which are relatively fast. (These statements are true only as the trajectory hits $`S_1`$.) We recall here that in the experimental time series, the stress drops from a peak value to its minimum in a very short time span. Further, we have argued that this should be the sum of contributions arising from fast multiplication of dislocations (which we have already argued has a time scale of $`\delta ^1`$) and subsequent immobilization. The latter is reflected in another rapid time scale $`d/b_0`$. This explains the difficulty in separating the contributions arising from the two processes in the experimental time series. Moreover, since $`x`$ is a fast variable, the changes in $`x`$ component dominates the descent of the trajectory. Finally, as the trajectory approaches $`S_2`$, $`g/x`$ becomes positive and the trajectory jumps from $`S_1`$ to $`S_2`$. Combining these results, we see that the trajectory moves towards the region of smaller values of $`y`$ and $`\varphi `$ entering $`S_2`$ in a region of small values of $`y`$ and $`\varphi `$.
In summary, the sequential way the orbit visits various parts of the phase space is as follows. The trajectory enters $`S_2`$ part of the slow manifold in regions of small $`y`$ and $`\varphi `$ making an exit along $`S_2`$ for relatively large $`\varphi `$ and $`y`$. Thereafter, the trajectory moves through the ‘unstable’ part of the phase space before falling onto the $`S_1`$ and quickly descends on $`S_1`$. This completes the cyclic movement of the trajectory and explains the geometrical feature of the trajectory shuttling between these two parts of the manifold and the associated time scales.
Now, the question that remains to be answered is $``$ do the trajectories always visit both $`S_1`$ and $`S_2`$ or is there a possibility that the trajectory remains confined to $`S_1`$ ? It is clear that if the former is true, relaxation oscillations with large amplitude will occur and if the latter is true, these are likely to be nearly sinusoidal small amplitude oscillations. Here, we recall that the coordinates of the saddle focus fixed point are $`x_0=z_0e/2`$ which is much larger than the value of $`xy/\left|\delta \right|`$ on $`S_2`$. Thus, the fixed point located on the $`S_1`$ will be close to the ‘fold’ at the first Hopf bifurcation which occurs at small values of $`e=e_{c_1}5`$. Due to the unstable nature of the fixed point, the trajectories spiralling out are forced onto the $`S_2`$ part of the manifold resulting in relaxation oscillation. This point has been illustrated by considering the example of a period eleven orbit for $`m=1.2`$ and $`e=267.0`$ shown in Fig. 6. As is clear from this diagram, the small amplitude oscillations are located on the $`S_1`$. As $`e`$ is further varied, the small amplitude oscillations grow with $`e`$, but the relaxation nature does not manifest until the orbit crosses over to $`S_2`$. To the best of the authors knowledge the mechanism suggested here for pulsed type relaxation oscillations is new.
As we will see, the analysis of the slow manifold and the time scales operating in different parts of the phase space will be useful in providing an appropriate interpretation of the various branches of the SRS.
## V Negative strain rate sensitivity
At the outset, we stress that it has been recognized that the negative unstable branch is not accessible to the dynamics of the PLC effect. Even so, early formulations and the way experimental measurements have been carried out has given rise to considerable confusion. The purpose of the material presented below is to briefly discuss the concept of negative SRS and working methods adopted in the literature, and also clear some misconceptions.
Theories of dynamic strain ageing (DSA) assume that the interaction of dislocations with solute atoms when averaged over the specimen dimensions can be represented by a constitutive relation connecting stress, strain and strain rate which is conventionally written as
$$\sigma =hϵ+F(\dot{ϵ}).$$
(13)
The basic assumption inherent in Eq. (13) is that stress can be split into a function of $`ϵ`$ and another of $`\dot{ϵ}`$ alone. Then, the SRS is defined as
$$𝒮=\frac{\sigma }{ln\dot{ϵ}}|_ϵ=\dot{ϵ}\frac{d\sigma }{d\dot{ϵ}}$$
(14)
Clearly, this definition uses $`ϵ`$ as a state variable. This unfortunately is not correct since strain is history dependent. Inspite of this, conventionally, strain is fixed at a small nominal value and the flow stress at that value is used to obtain the SRS. It is interesting to note that the existence of critical strain for the onset of the PLC effect implies that when the nominal strain value is lower than $`ϵ_c`$, there are no serrations even when the applied strain rate value is in the domain of the PLC effect ( $`e_{c_1}<e<e_{c_2}`$). Yet, the onset of serrations for higher strains is somehow reflected in the measured nonmonotonic behavior of the flow stress. In experiments, by fixing $`ϵ`$ at some nominal value less than $`ϵ_c`$, the flow stress ( at the fixed strain) is found to increase as a function of applied strain rate $`e`$ for $`e<e_{c_1}`$, shows a decreasing trend for $`e_{c_1}<e<e_{c_2}`$ and again reverts to an increasing trend for $`e>e_{c_2}`$ . Thus, the flow stress has the form shown in Fig. 1. No explanation has been offered in the literature as to why this nonmonotonic behavior should be seen. However, explanation from the dynamical point of view is fairly straightforward and is as follows. We recall here that the model predicts the existence the critical strain $`ϵ_c`$, as also the existence of window of strain rates $`e_{c1}<e<e_{c2}`$ within which serration can occur. Thus, from Eq. (9) it is easy to understand the increasing order in which the stress-strain curves are placed for increasing values of $`e`$ when $`e<e_{c_1}`$ and $`e>e_{c_2}`$. In this range of $`e`$, the fixed point is stable and thus all trajectories converge to the fixed point. However, for $`e_{c_1}<e<e_{c_2}`$, we note that serrations result only for large enough strains, i.e., once the time of deformation is such that strain crosses $`ϵ_c`$. In our theory, serrations are equated with the existence of periodic (or aperiodic) solutions when $`e_{c_1}<e<e_{c_2}`$. These steady state solutions are usually reached only after transients die down. Thus, low value of nominal strain implies short evolution time which in turn implies that the stress is being monitored at a transient state. Thus, the decreasing trend of the flow stress for $`e_{c_1}<e<e_{c_2}`$ is a reflection of the impending periodic (aperiodic) steady state that will be reached eventually. Indeed, this was the method followed in our earlier calculation since the procedure was easy to implement numerically. However, in many experimental situations, it is not possible to choose a nominal strain value low enough that it is less than $`ϵ_c`$ for the entire range of strain rate values. In such a case, since the stress-strain curves are serrated, there is an ambiguity in the value of stress to be used. A working method adopted is to use a stress value as the mean value of the upper and lower stress values . Then, the flow stress appears to decrease for the domain of applied strain rate values where the PLC effect manifests. Thus, this method gives the impression of actually measuring the unstable branch.
The above methods are not suitable for adoption since they do not permit the use of the knowledge of the slow manifold. There is an alternate method which uses the relaxation oscillations inherent to the dynamics of the PLC effect. In this method, by analogy with electrical analogues, one assumes that there exists a family of curves $`F(\dot{ϵ}_p)`$ for each $`ϵ`$ of the form shown in Fig.1 which trigger relaxation oscillations in the form of plastic strain rate bursts and stress drops. By comparing the measured stress drops and strain bursts, one concludes the existence of the unstable branch, but one never records any points in this region. This method is suitable for our study since we will use relaxation oscillations arising in the model.
In the following we shall argue that the two slow time scales in the dynamics actually translate into the two stable dissipative branches of the SRS and the two fast time scales to jumps in plastic strain rate across the stable branches of the negative SRS. Since SRS represents $`\varphi `$ as a function of the plastic strain rate $`\dot{ϵ}_p=\varphi ^mx`$, in Fig. 7, we have shown a projection of the phase space trajectory on the $`\varphi \dot{ϵ}_p`$ plane (instead of $`\varphi x`$) corresponding to a monoperiodic relaxation oscillation. ( Here, we have retained the same notation for the scaled plastic strain rate as for the unscaled one.) The unstable fixed point is also shown. Starting from any initial value around the unstable focus, trajectories spirals out converging onto the limit cycle. In Fig. 7, we have identified different regions of the phase space with different regions of the slow manifold, $`S_1`$ and $`S_2`$. We first note that there is a considerable similarity between Fig. 7 and the schematic representation of the relaxational oscillation obtained using the negative SRS shown in Fig. 1. Note also that in contrast to the artificial flat parts $`B^{}C^{}`$ and $`C^{}D^{}`$ of Fig. 1, the equivalent parts in Fig. 7 have a finite negative slope. Lastly, as in experiments, the strain rate jump from B to C is over two orders of magnitude.
Here, we set up a correspondence between the dynamics in the phase space (Fig. 7) and the slow manifold (Figs. 4 and 5). From our earlier discussion, we know that when the trajectory is on $`S_2`$, $`x`$ is constant and small in magnitude. Consequently, according to Eq. (9), $`\varphi `$ should increases linearly and hence this corresponds the rising branch AB in Fig. 7. Further, noting that $`\dot{x}0`$ for the entire interval of time spent by the trajectory on $`S_2`$ (see Fig. 4), the branch AB of Fig. 7 corresponds to the pinned state of dislocations. For this branch, one can easily see that the (mean) value of $`𝒮3.5`$ using Eq. (14). Further, as we move up on this branch towards B (Fig. 4), the value of $`\delta `$ approaches zero, and $`\varphi `$ reaches its maximum value. Once $`\delta `$ becomes positive, the trajectory leaves $`S_2`$, and thus, the strain rate jump from B to C in Fig. 7, corresponds to the trajectory jumping from $`S_2`$ to $`S_1`$ in Figs. 4 and 5. Note that the slope $`\varphi /\dot{ϵ}_p`$ for this portion of the orbit is quite small and negative unlike the zero value for the equivalent part in Fig. 1. Further, we know from Fig. 4, once the trajectory reaches $`S_1`$, the value of $`x`$ decreases rapidly resulting in the decrease of $`\dot{ϵ}_p`$. Thus, the region CD in Fig. 7 corresponds to the movement of the trajectory on $`S_1`$ ( in Fig. 4 for which $`\delta >0`$). For this branch, one can quickly check that the strain rate sensitivity $`𝒮`$ is positive having a mean value ( $`1.5`$) which is a factor of two less than that for the branch AB, implying that the nature of dissipation is quite different from that operating on AB. This is consistent with known facts about the two branches as mentioned in the introduction. Combining this with the fact that $`\dot{x}`$ is decreasing, the branch CD in Fig. 7 mimics the equivalent branch $`C^{}D^{}`$ in Fig. 1, which is identified with the slowing down of the mobile dislocations without solute atmosphere.
We recall that the stress drop duration has contributions from two fast processes, namely, dislocation multiplication and it’s subsequent immobilization. But, these two time scales could not be separated in the stress strain curve. However, in the present phase plot representation (Fig. 7), we see that the fast multiplication of dislocations correspond to BC and that of immobilization to CD. This correspondence has been possible due to the mapping of the relevant time scales in the dynamics of the dislocations obtained from the analysis of the slow manifold to the various regions in the phase plot thereby allowing us to identify the individual contributions. (Note also that in Fig. 7, we have plotted points of the trajectory at equal intervals of time which shows that the time interval corresponding to BD is small.) From Figs. 4 and 5, we see that, as the trajectory descends on $`S_1`$ part of the slow manifold and gets close to $`S_2`$, it leaves $`S_1`$, since $`g/x`$ becomes positive ($`x50`$). Further, the strain rate sensitivity parameter $`𝒮`$ changes sign at D. For the corresponding DA part in Fig. 7, the slope is small and negative as for the part BC. Noting that B and D are the points at which $`𝒮`$ turns negative, and noting that the fixed point is unstable, the so called ”unstable branch” of the SRS, not accessible to the dynamics, can be inferred by drawing a (dotted) line connecting the maximum and the minimum of the stress and passing it through the unstable fixed point (Fig. 7).
We will now attempt to use the results of our analysis of time scales in different regions of the slow manifold to obtain the dependence of $`\dot{ϵ}_p`$ as a function of $`\varphi `$. The equation for $`\dot{ϵ}_p`$ is
$$\frac{d\dot{ϵ}_p}{dt}=xm\varphi ^{m1}\dot{\varphi }+\varphi ^m\dot{x}$$
(15)
which on using Eq.(6) and (9) gives
$$\frac{d\dot{ϵ}_p}{d\varphi }=\frac{\dot{ϵ}_p(\frac{mde}{\varphi }+\delta )\dot{ϵ}_p^2(\frac{md}{\varphi }+\frac{b}{\varphi ^m})+y\varphi ^m}{d(e\dot{\dot{ϵ}_p})}$$
(16)
Here we note that in the slow manifold description, all slow variables appear as parameters. However, since SRS describes the dependence of the slow variable $`\varphi `$ as a function of the (derived) fast variable $`\dot{ϵ}_p`$, we will consider the other two variables $`y`$ or $`\delta `$ or both as parameters. Numerical solution of Eq.(16) has been attempted using $`y`$ and $`\delta `$ as parameters. Good numerical approximation is obtained by noting that $`y`$ and hence $`\delta `$ is periodic. Thus, any reasonable approximation for the periodicity of $`y`$, for example, sine function with a proper amplitude and phase gives a good fit with the phase plot. However, our interest here is to obtain approximate expressions for $`\dot{ϵ}_p(\varphi )`$ on different branches. For this reason, we will use typical values of $`\delta `$ and $`y`$ for the interval under question. From section 4, the trajectory has different dynamics in different regions of the slow manifold. These are : (1) on $`S_2`$ where $`\dot{x}`$ is nearly zero for the entire time spent by the trajectory on $`S_2`$, (2) just outside $`S_2`$ where $`\dot{x}x\delta `$,(3) on $`S_1`$ where $`x\delta /b_0`$ for $`\dot{ϵ_p}>e`$ and (4) when the trajectory jumps from $`S_1`$ to $`S_2`$. Approximate solutions obtained for these cases are shown in the phase plot by solid lines. Details are given in the Appendix. It is clear that these solid lines reproduce the general features of the phase plot quite well. We stress here that these lines correspond to the simplest approximation.
The above analysis refers to a fixed value of $`e`$. As a function of $`e`$, we find that the magnitude of the stress drops increases initially and then decreases. This feature is a direct result of the existence of back-to-back Hopf bifurcations in the model. On the other hand, experimentally one sees only a decreasing trend. While the decreasing trend is consistent, the increasing trend seen in the model for low strain values can be traced to the effect of another crucial parameter in the model, namely $`b_0`$. We recall that this parameter corresponds to the remobilization of immobile dislocations. For the value of $`b_0`$ used in the present calculation, the bifurcation from the steady state is a mildly subcritical Hopf bifurcation, i.e., across the transition the amplitude of the stress change is abrupt but the magnitude is small. However, for smaller values of $`b_0`$, this jump can be made sufficiently large in which case the amplitude of the stress drops can be made to decrease with $`e`$ right from the onset of the PLC effect.
## VI Discussion and Conclusions
The study of the relaxation oscillations in the model was motivated by the need to explain the apparent inconsistency between the time scales observed in experimental stress-time series and those that could be argued on the basis of the negative SRS feature commonly used in the literature. The study of relaxation oscillations using the geometry of the slow manifold has helped us to identify different time scales operating in different regions of the phase space, apart from showing that the nature of the relaxation in the model is due to the atypical bent geometry of the slow manifold. This geometry is very different from the standard $`S`$ shaped manifold and hence the relaxation oscillations seen here differs qualitatively from that seen in systems with $`S`$shaped slow manifold. Some comparative comments between these two types of manifolds may be in order here. As in the $`S`$shaped manifold, there are two attracting branches in our case also, namely $`S_1`$ and $`S_2`$. The dynamics on $`S_2`$ is slow as it is controlled by the slow variables $`y`$ and $`\varphi `$. On the other hand, on $`S_1`$, the time dependence of the trajectory is largely controlled by the fast variable $`x`$. In this sense, the dynamics on $`S_2`$ is slow and that on $`S_1`$ is fast. with a large magnitude and with a much smaller magnitude from $`S_1`$ to $`S_2`$. Though there are two fast jumps as in the $`S`$shaped manifold, in our case, there is no equivalent unstable part of the slow manifold which causes these jumps.
The analysis of the time scales controlling the relaxation oscillations has been directly used to reconstruct the relaxation oscillation in the $`\varphi \dot{ϵ}_p`$ plane which bears a strong resemblance to the the relaxation oscillations resulting from the assumed form of the negative SRS. The information on different time scales operating in different regions of the slow manifold has been used to calculate the dependence of $`\varphi `$ on $`\dot{ϵ}_p`$ for the two dissipative branches and the associated strain rate jumps between them. This has helped to identify the various regions of the slow manifold with the stick state and the slip state of dislocations. It has also helped us to clarify the inconsistency in the time scales of the dynamics. Further, several important features of the SRS derived from the model compare well with that reported in the literature. In particular, we note that the slope of the first dissipative branch AB is larger than that of the second branch CD (Fig. 7). Further, we recall that $`\delta =\varphi ^m+y+a`$ which is positive for AB, gradually approaches zero as B is reached followed by strain rate jump. Similarly, for the branch CD, $`\delta `$ approaches zero as we approach D followed by a jump in the strain rate. Thus, vanishing of $`\delta `$ is indicative of strain rate jumps just as the strain rate sensitive parameter $`S`$ also vanishes. Noting that $`y`$ is the immobile dislocation density, it is tempting to interpret $`\delta `$ as being related to some kind of effective stress. (Recall that the effective stress is $`\sigma ^{}=\sigma _aHN_{im}^{1/2}`$, where H is the work hardening coefficient.) Thus, the points at which strain rate jumps occur correspond to points at which the effective stress vanishes which is very much like the classical explanation. Since the definition of strain rate sensitivity assumes strain as a state variable which is not true, $`\delta `$ may be an effective alternate parameter for defining strain rate sensitivity. Thus, it is nice to see that we can attribute a physical meaning to this parameter.
The analysis has also helped us to provide a dynamical interpretation of the negative SRS. The analysis also shows that the large jumps in the strain rate across the stable branches are due to the relaxational nature of the dynamics which in turn is a result of bent nature of the slow manifold and the fact that the bifurcation is of the Hopf type. Using this, we have inferred the existence of the unstable branch as containing the two points, B and D, where strain rate jumps (where $`\delta `$ and $`S`$ are zero) and the unstable fixed point. In this sense, Hopf bifurcation is at the root of the ‘negative’ SRS. Similar features of SRS were found to operate in a model designed to mimic stick-slip dynamics of tectonic faults . There are other studies on stick-slip dynamics, both experimental and theoretical , which support the view that Hopf bifurcation was found to be responsible for the instability. Thus, it is likely that Hopf bifurcation is relevant to situations where stick-slip dynamics operates and wherever one measures the two stable branches and the jumps across the branches.
The relaxation oscillations in the model are reminiscent of the canard type of oscillations in multiple type scale dynamical systems . The latter type of oscillations result from ‘sticking’ of the trajectory to the repelling part of the $`S`$shaped slow manifold before jumping to the attracting pleat of the slow manifold. In our case, although these oscillations have a similarity with canard type of solutions, the repelling part of slow manifold does not exist. Instead, the trajectories stick to the ‘unstable’ part of the phase space where the dynamics is accelerated once the trajectory moves well into this region. This aspect coupled to the fact that there is no inherent constraint on the manifold structure leads to oscillations of all sizes. It is clear that such oscillations result from the trajectory ‘sticking’ to the direction of the $`S_2`$ plane and moving into the ‘unstable’ part of the phase space by varying amounts each time the trajectory visits $`S_2`$. These jumps translate into stress drops of varying sizes which are generally seen in experimental time series (Fig. 2). This also means that $`\dot{ϵ_p}\varphi `$ is not a simple limit cycle, and the simplistic approach of inferring the ‘negative’ SRS should be given up. The present analysis stresses the importance of using sound dynamical tools such as the slow manifold as the basis for studying more complex oscillations rather than phenomenological concepts such as the negative SRS.
## A
Here, we obtain approximate analytical expressions for $`\dot{ϵ}_p(\varphi )`$ for different regions of $`\varphi \dot{ϵ_p}`$ phase plot (Fig. 7) using the knowledge of time scales obtained from the analysis of relaxation oscillations. For the numerical evaluation, the values of control parameters have been chosen as $`e=200`$, $`m=1.2`$, $`b_0=0.002`$ and $`d=0.0001`$.
Region AB : When the trajectory is on $`S_2`$, $`\dot{x}0`$, for the entire interval of time. Using $`x=y/\delta `$ in Eq.(15), we get
$$\frac{d\dot{ϵ}_p}{d\varphi }=\frac{\varphi ^{m1}my}{\delta }.$$
(A1)
Noting that $`\delta =\varphi ^mya`$, this equation can be integrated thereby reducing the number of parameters to one, namely $`y`$. Integrating, we get
$$\dot{ϵ}_p=y\mathrm{𝑙𝑛}\left(\frac{\varphi ^m(y+a)}{\varphi ^m(0)(y+a)}\right)+\dot{ϵ_p}(0)$$
(A2)
where $`\dot{ϵ}_p(0)`$ and $`\varphi (0)`$ refer to their respective values as the trajectory enters $`S_2`$. In Fig. 7, we have used $`\dot{ϵ}_p(0)=4.7`$, $`\varphi (0)=2.2`$ and $`y=6.15`$.
Region BC : This region corresponds to the jump from $`S_2`$ to $`S_1`$. This happens when the trajectory is just outside $`S_2`$. For this region, $`\varphi `$ is near $`\varphi _{max}`$, and since $`x\left(y/b_0\right)^{1/2}`$, the evolution of $`x`$ is well described by $`\dot{x}x\delta `$, implying that the time of evolution is very short. Thus, we can regard the evolution of $`\varphi `$ as being mainly determined by that of $`x`$. (This region also corresponds to $`\delta >0`$ and small $`0.2`$.) Thus, we use $`\varphi =\varphi _{max}`$ on the RHS of Eq. (9). Then,
$$\frac{d\varphi }{dt}=d(e\varphi _{max}^mx_{s_2}e^{\delta t})$$
(A3)
where $`x_{s_2}`$ is the value of $`x`$ at the time of leaving $`S_2`$. Integrating this equation with initial conditions at $`t=0`$, $`\varphi =\varphi (0)=\varphi _{max}`$, we get
$$e^{\delta t}=\frac{te\delta }{x_{s_2}\varphi _{max}^m}\frac{(\varphi (t)\varphi _{max})\delta }{x_{s_2}\varphi _{max}^md}+1$$
(A4)
Clearly, the first term is small since the time span of evolution that we are interested is also $`\delta `$. Now consider Eq. (15). Using $`\dot{x}x\delta `$, we get
$$\frac{d\dot{ϵ}_p}{dt}=\dot{ϵ}_p\left(\frac{med}{\varphi _{max}}+\delta \right)\frac{m\dot{ϵ}_p^2d}{\varphi _{max}}$$
(A5)
Since $`med/\varphi _{max}<<\delta `$, we drop the first term. Integrating the above equation with the initial value $`\dot{ϵ_p}(0)=e`$, leads to
$$\dot{ϵ}_p=\frac{\varphi _{max}e(\varphi _{max}\varphi (t))\delta }{md^2\varphi _{max}^mx_{s_2}(\varphi _{max}\delta /mde)(\varphi (t)\varphi _{max})emd}$$
(A6)
In Fig. 7, we have used the values $`\delta =0.021`$, $`\varphi _{max}=4.98`$ and $`x_{s_2}=1.7`$.
Region CD : Consider the trajectory on $`S_1`$ with $`x\delta /b_0`$ and $`\dot{ϵ}_pe`$. Then, Eq. (16) reads,
$$\frac{d\dot{ϵ}_p}{d\varphi }=\frac{\dot{ϵ}_p(\frac{med}{\varphi }+\delta )\dot{ϵ}_p^2(\frac{md}{\varphi }+\frac{b_0}{\varphi ^m})+y\varphi ^m}{d\dot{ϵ_p}(1\frac{e}{\dot{ϵ}_p})}$$
(A7)
Since $`e/\dot{ϵ}_p<1.0`$, we expand the denominator and retain terms upto $`\dot{ϵ}_p^1`$. We note here that on $`S_1`$, $`x`$ is rapidly decreasing and therefore, using slow manifold values is not a good approximation. Even so, as a simplest approximation we use $`x\delta /b_0`$. Then, we get
$$\frac{d\dot{ϵ}_p}{d\varphi }=\frac{me^2b_0}{\varphi ^{m+1}\delta }\frac{eb_0^2y}{\varphi ^md\delta ^2}+\frac{m\varphi ^{m1}\delta }{b_0}\frac{b_0y}{d\delta }$$
(A8)
In this equation, both $`y`$ and $`\delta `$ appear as parameters whose values are chosen appropriate to this region. Integrating the above equation with the initial values of $`\varphi (0)`$ and $`\dot{ϵ_p}(0)`$, we get
$`\dot{ϵ}_p`$ $`=`$ $`{\displaystyle \frac{eb_0^2y}{(m1)d\delta ^2}}(\varphi ^{1m}\varphi ^{1m}(0))+{\displaystyle \frac{e^2b_0}{\delta }}(\varphi ^m\varphi ^m(0))`$ (A10)
$`+{\displaystyle \frac{\delta }{b_0}}(\varphi ^m\varphi ^m(0)){\displaystyle \frac{b_0y}{d\delta }}(\varphi \varphi (0)))+\dot{ϵ_p}(0)`$
In Fig. 7, we have used the values of $`\delta =3.57`$, $`y=1.0`$ with $`\varphi (0)=4.3`$ and $`\dot{ϵ_p}(0)=7460.0`$.
Region DA : This region again corresponds to the the trajectory descending on $`S_1`$ but $`\dot{ϵ}_p<e`$. Then, Eq.(16) can be written in the form
$$\frac{d\dot{ϵ_p}}{d\varphi }=\frac{\dot{ϵ_p}(\frac{med}{\varphi }+\delta )\dot{ϵ_p}^2(\frac{md}{\varphi }+\frac{b_0}{\varphi ^m})+y\varphi ^m}{ed(1\frac{\dot{ϵ_p}}{e})}$$
(A11)
In this region, the value of $`\varphi `$ is slowly varying with its value near the minimum for which $`\delta 0.15`$. Since $`\varphi `$ is near $`\varphi _{min}`$, we use $`\varphi =\varphi _{min}`$ and regard the variation as largely arising due to the changes in $`\dot{ϵ}_p`$. Using $`(1\frac{\dot{ϵ}_p}{e})^1(1+\frac{\dot{ϵ}_p}{e})`$, we have
$$\frac{d\dot{ϵ_p}}{d\varphi }=(A_1\dot{ϵ_p}+B_1\dot{ϵ_p}^2C_1\dot{ϵ_p}^3)$$
(A12)
where $`A_1=\frac{m}{\varphi _{min}}\frac{|\delta |}{ed}`$ and $`B_1=\frac{|\delta |}{e^2d}\frac{b_0}{e\varphi _{min}^md}`$ and $`C_1=\frac{m}{\varphi _{min}e^2}+\frac{b_0}{e^2d\varphi _{min}^m}`$. Since $`C_1A_1`$ and $`B_1`$, we drop the last term. Integrating the above equation with the initial conditions, $`\varphi (0)=\varphi _{min}`$ and $`\dot{ϵ_p}=e`$, we get
$$\dot{ϵ_p}=\frac{eA_1e^{A_1(\varphi \varphi _{min})}}{A_1+eB_1(1e^{A_1(\varphi \varphi _{min})})}$$
(A13)
In Fig. 7, we have used $`\varphi _{min}=1.55`$. |
warning/0001/hep-th0001219.html | ar5iv | text | # Effects of Trilinear Term in Softly Broken 𝑁=1 Supersymmetric QCD
## Abstract
Softly broken dual magnetic theory of $`N=1`$ supersymmetric $`SU(N_c)`$ QCD with $`N_f`$ flavours is investigated with the inclusion of trilinear coupling term of scalar fields in the case of $`N_f>N_c+1`$. It is found that the the trilinear soft supersymmetric breaking term greatly change the phase and the vacuum structure.
There has been a big progress in understanding strongly coupled $`N=1`$ supersymmetric Yang-Mills theory in the last few years . A complete phase diagram of the theory, the particle spectrum and the dynamical phenomenon in each phase have been quantitively or qualitatively figured out. In particular, it was found that $`N=1`$ supersymmetric QCD with gauge group $`SU(N_c)`$ and $`N_f`$ flavour quarks possesses a “conformal window” $`3N_c/2<N_f<3N_c`$ in the infrared region of the theory when $`N_f>N_c+1`$, where the theory can be described by a physically equivalent $`N=1`$ supersymmetric $`SU(N_fN_c)`$ theory with $`N_f`$ flavours and $`N_f^2`$ singlets but with the strong and weak coupling be exchanged and vice versa. This is exactly the realization of the old Montonen-Olive non-Abelian electric-magnetic duality conjecture in $`N=1`$ supersymmetric gauge theory, which exists only in $`N=4`$ and has an analogue in low-energy $`N=2`$ supersymmetric gauge theories.
It is natural to extend these non-perturbative analysis to the ordinary QCD since its non-perturbative aspect is not clear yet. To do this, the first step is breaking supersusymmetry. The most convenient breaking method is the the introduction of superparticle mass terms such as squarks and gaugino. However, when $`N_f>N_c+1`$ the dual magnetic theory has an additional flavour interaction superpotential. A trilinear soft supersymmetry breaking term composed of the scalar fields is thus allowed. In this talk we shall emphasize the effects of this trilinear term in determining the phase structure and the vacuum structure of the soft broken $`N=1`$ dual QCD.
$`N=1`$ supersymmetric QCD with gauge group $`SU(N_c)`$ and $`N_f`$ flavour quark chiral supermultiples $`Q_{ir}`$, $`\stackrel{~}{Q}_{ir}`$ $`i=1,\mathrm{},N_f`$, $`r=1,\mathrm{},N_c`$, has an anomaly free global symmetry
$`SU(N_f)_L\times SU(N_f)_R\times U(1)_B\times U(1)_R`$ (1)
At the low-energy the quark supermultiplets are confined, the dynamical degrees of freedom are the chiral supermultiplets, meson $`M_{ij}`$ and baryons $`B^{[i_1\mathrm{}i_{N_c}]}`$ and $`\stackrel{~}{B}^{[i_1\mathrm{}i_{N_c}]}`$. The low-energy effective theory still possesses the global symmetry (1). When $`N_f>N_c+1`$, the dynamics of $`M`$, $`B`$ and $`\stackrel{~}{B}`$ can be described $`N=1`$ supersymmetric $`SU(N_fN_c)`$ gauge theory with $`N_f`$ flavours of quark chiral supermultiplets $`q_{i\stackrel{~}{r}}`$, $`\stackrel{~}{q}_{i\stackrel{~}{r}}`$ and colour singet chiral fields $`_{ij}`$, $`i,j=1,\mathrm{},N_f`$, $`\stackrel{~}{r}=1,\mathrm{},N_fN_c`$, and an additional flavour interaction superpotential
$`W=\stackrel{~}{q}^{i\stackrel{~}{r}}_{ij}q^{jr}`$ (2)
This conjecture is supported by ’t Hooft anomaly matching.
The most important property in supersymmetric field theory is non-renormalization theorem, which determine the superpotential must be a holomorphic function of the chiral superfield. The holomorphy of superpotential and global symmetry (1) as well as the instanton calculation can give a series of exact result of low-energy supersymmetric QCD. Since supersymmetric QCD is very sensitive to the relative numbers of colour and flavours, so we list the non-perturbative dynamical phenomena according to the different ranges of the colour and flavour numbers.
When $`N_f<N_c`$ there will dynamically generate a superpotential, which eliminates all the supersymmetry vacua. In the case $`N_f=N_c`$, the non-perturbative quantum correction modify the classical moduli space constrained by $`detMB\stackrel{~}{B}=0`$ as the quantum one, $`detMB\stackrel{~}{B}=\mathrm{\Lambda }^{2N_c}`$. Depending on the vacuum choice in the moduli space, the theory can present various dynamical patterns. For examples, the vacuum $`M_j^i=\mathrm{\Lambda }^2\delta _j^i`$, $`B=\stackrel{~}{B}=0`$ leads to the chiral symmetry breaking and confinement; while the other vacuum choice $`M_j^i=0`$, $`B=\stackrel{~}{B}=\mathrm{\Lambda }^{N_c}`$ makes chiral symmetry unbroken and the baryon number violation. In the case $`N_f=N_c+1`$, the quantum moduli space is the same as classical moduli space. Consequently, the low-energy theory present confinement but no chiral symmetry breaking. When $`N_f>N_c+1`$, if $`N_f>3N_c`$, the theory at high energy level is not asymptotically free and hence the low-energy theory in a free electric phase, the coupling constant behaves as $`\alpha (R)1/\mathrm{ln}(R\mathrm{\Lambda })`$; The more interesting is the range $`3N_c/2<N_f<3N_c`$, here the theory can have an non-trivial IR fixed point, at which the low-energy theory becomes an interacting conformal field theory. Thus it is called Seiberg’s conformal window. The $`SU(N_fN_c)`$ theory describes the same physics as the high energy $`SU(N_c)`$ QCD, but with the strong and weak coupling exchanged and vice versa This is called the non-Abelian electro-magnetic duality..
When $`N_f<N_c`$ the quadratic soft supersymmetry breaking term at low-energy can be written out near the origin of the moduli space
$`_{\mathrm{sb}}`$ $`=`$ $`{\displaystyle }d^4\theta [B_TM_Q\text{Tr}(M^{}M)+B_BM_Q(B^{}B`$ (3)
$`+`$ $`\stackrel{~}{B}^{}\stackrel{~}{B})][{\displaystyle }d^2\theta M_gW^\alpha W_\alpha )+h.c.]`$ (4)
In the case $`N_fN_c+1`$, the composite superfields are equivalently replaced by the dual magnetic quarks and the soft breaking Lagrangian is
$`_{\mathrm{sb}}=B_Mm_M^2\text{Tr}(\varphi _M^{}\varphi _M)+B_qm_q^2(\varphi _q^{}\varphi _q+\varphi _{\stackrel{~}{q}}^{}\varphi _{\stackrel{~}{q}})`$ (5)
In the decoupling limit, $`m_g,m_Q\mathrm{}`$, the features of the ordinary QCD are expected.
The non-perturbative dynamical features in this soft broken supersymmetric QCD had been analyzed. The results show that when $`N_f<N_c`$, the standard $`SU_V(N_f)\times U(1)_B`$ QCD vacuum can arise, while in the case $`N_f=N_c`$, there emerges an exotic vacua with chiral symmetry and spontaneously breaking of the baryon number symmetry, i.e. the vacuum is invariant $`SU_L(N_f)\times SU_R(N_f)\times U(1)_R`$. When $`N_f>N_c`$, There is a vacuum state with unbroken chiral symmetry, but it is interesting that the non-Abelian electric-magnetic duality persists in the presence of soft supersymmetry breaking.
When $`N_fN_c+1`$, in addition to (5), the trilinear term
$`_{SB}^{}=h\varphi _{qi}\varphi _{Mj}^i\varphi _{\stackrel{~}{q}}^j`$ (6)
can also make the supersymmetry softly broken due to the superpotential (2), where $`h`$ is the trilinear coupling constant. With the inclusion of (6) the scalar potential becomes
$`V(\varphi _q,\varphi _{\stackrel{~}{q}},\varphi _M)={\displaystyle \frac{1}{k_T}}\text{Tr}\left(\varphi _q\varphi _q^{}\varphi _{\stackrel{~}{q}}^{}\varphi _{\stackrel{~}{q}}\right)`$ (7)
$`+{\displaystyle \frac{1}{k_T}}\text{Tr}\left(\varphi _q\varphi _M\varphi _M^{}\varphi _q^{}+\varphi _{\stackrel{~}{q}}^{}\varphi _M^{}\varphi _M\varphi _{\stackrel{~}{q}}\right)`$ (8)
$`+{\displaystyle \frac{\stackrel{~}{g}^2}{2}}\left(\text{Tr}\varphi _q^{}\stackrel{~}{T}^a\varphi _q\text{Tr}\varphi _{\stackrel{~}{q}}\stackrel{~}{T}^a\varphi _{\stackrel{~}{q}}^{}\right)^2`$ (9)
$`+m_q^2\text{Tr}(\varphi _q^{}\varphi _q)+m_{\stackrel{~}{q}}^2\text{Tr}(\varphi _{\stackrel{~}{q}}^{}\varphi _{\stackrel{~}{q}})+m_M^2\text{Tr}(\varphi _M^{}\varphi _M)`$ (10)
$`(h\text{Tr}\varphi _{qi}\varphi _{Mj}^i\varphi _{\stackrel{~}{q}}^j+h.c.)`$ (11)
where $`\stackrel{~}{T}^a`$ are the generators of magnetic gauge group $`SU(N_fN_c)`$, $`\stackrel{~}{g}`$ is the magnetic gauge coupling constant. The phase (or vacuum) structure can be revealed by analyzing the minima of (11). With assumption that $`h`$ is real, the minimum of potential can be obtained along diagonal direction
$`\varphi _{qi}^r=\{\begin{array}{cc}\varphi _{q(i)}\delta _i^r\hfill & i=1,\mathrm{},N_fN_c\hfill \\ 0\hfill & \text{otherwise}\hfill \end{array}`$ (14)
$`\varphi _{\stackrel{~}{q}i}^r=\{\begin{array}{cc}\varphi _{\stackrel{~}{q}(i)}\delta _i^r\hfill & i=1,\mathrm{},N_fN_c\hfill \\ 0\hfill & \text{otherwise}\hfill \end{array}`$ (17)
$`\varphi _{Mj}^i=\{\begin{array}{c}\varphi _{M(i)}\delta _j^i,i,j=1,\mathrm{},N_fN_c\hfill \\ 0,\text{otherwise}\hfill \end{array}`$ (20)
The analysis shows that in the direction $`\varphi _{q(i)}=q`$ and $`\varphi _{\stackrel{~}{q(i)}}=0`$ (or $`\varphi _{q(i)}=0`$ and $`\varphi _{\stackrel{~}{q}(i)}=q`$), the vacuum expectation value $`\varphi _{Mj}^i=0`$. If $`m_q^2>0`$ (or $`m_{\stackrel{~}{q}}^2>0`$), the scalar potential has the minimum $`V=0`$ at $`q=0`$, thus theory is in chiral symmetric phase. However, if $`m_q^2<0`$ (or $`m_{\stackrel{~}{q}}^2<0`$), the scalar potential unbounded from below and the theory becomes unphysical. In the flat direction of $`D`$-term, $`\varphi _{q(i)}=\varphi _{\stackrel{~}{q}(i)}=X_i`$, if soft breaking parameters satisfy $`h^22/k_q\left(m_q^2+m_{\stackrel{~}{q}}^2\right)`$, then we find
$`{\displaystyle \frac{h\sqrt{h^22(m_q^2+m_{\stackrel{~}{q}}^2)/k_q}}{2/k_q}}\varphi _{M(i)}`$ (21)
$`{\displaystyle \frac{h+\sqrt{h^22(m_q^2+m_{\stackrel{~}{q}}^2)/k_q}}{2/k_q}}`$ (22)
Thus chiral symmetry broken phase arises and the baryon number violation occurs. Furthermore, depending on the ratio $`\rho 2m_M^2/(m_q^2+m_{\stackrel{~}{q}}^2)k_q/k_M`$, the phase structure in $`D`$-flat directions presents various patterns. For examples, in the phase diagram labeled by $`(h^2,(m_q^2+m_{\stackrel{~}{q}}^2)/2)`$, when $`\rho =1`$, the theory only has one chiral symmetry broken phase and one unbroken phase iff all $`m_q^2`$, $`m_{\stackrel{~}{q}}^2`$ and $`m_M^2`$ are positive, whereas when $`\rho =20`$ theory has two unbroken phases and two chiral symmetry broken phases. If $`m_q^2`$, $`m_{\stackrel{~}{q}}^2`$ and $`m_M^2`$ are negative, scalar potential is unbounded from below and becomes unphysical. Furthermore, in chiral symmetric phase, the $`SU(N_f)^3`$ and $`SU(N_f)^2U(1)_B`$ ’t Hooft anomalies match. In the broken phase, the two softly broken dual theories also present same anomaly structure. This seems to suggest Seiberg’s duality remains after SUSY breaking, even in the chiral symmetry broken phase.
In the case $`N_f=N_c+1`$, the flavour symmetry (1) allows a trilinear soft breaking term $`h^{}\varphi _{Bi}\varphi _{Mj}^i\varphi _{\stackrel{~}{B}}^j`$. With the combination of the quadratic term and the terms from the effective potential $`𝒲=(B_iM_j^i\stackrel{~}{B}^jdetM)/\mathrm{\Lambda }^{2N_c1}`$, the whole scalar potential is
$`V`$ $`=`$ $`\lambda _M^2{\displaystyle \underset{i,j}{}}|\varphi _{Bi}\varphi _{\stackrel{~}{B}}^j(\stackrel{}{det}\varphi _M)_i^j|^2`$ (23)
$`+`$ $`\lambda _B^2\text{Tr}(|\varphi _{\stackrel{~}{B}}\varphi _M|^2+|\varphi _B\varphi _M|^2)`$ (24)
$`+`$ $`m_B^2\text{Tr}|\varphi _B|^2+m_{\stackrel{~}{B}}^2\text{Tr}|\varphi _{\stackrel{~}{B}}|^2m_M^2\text{Tr}|\varphi _M|^2`$ (25)
$``$ $`(h^{}\varphi _{Bi}\varphi _{Mj}^i\varphi _{\stackrel{~}{B}}^j+h.c.)`$ (26)
where $`\lambda _M=1/(k_M\mathrm{\Lambda }^{2N_c1})`$, $`\lambda _B=1/(k_B\mathrm{\Lambda }^{2N_c1})`$, and $`(det^{}M)_i^jdetM/M_j^i`$. The minimum of superpotential has been analyzed along the diagonal direction, $`M_j^i=M_i\delta _j^i`$. Choosing a special direction with $`B_i\stackrel{~}{B}^j(\text{det}^{}M)_i^j=0`$, considering a simple case where $`B_i=\stackrel{~}{B}^i=B`$, $`M_{(i)}=M`$, and assuming $`B`$ and $`M`$ are real, we write the scalar potential as
$`{\displaystyle \frac{V}{N_c+1}}`$ $`=`$ $`2\lambda _B^2\varphi _M^{N_c+2}2h^{}\varphi _M^{N_c+1}`$ (27)
$`+`$ $`(m_B^2+m_{\stackrel{~}{B}}^2)\varphi _M^{N_c}+m_M^2\varphi _M^2`$ (28)
This scalar potential yields that if $`h^{}`$ is sufficiently large compared with soft scalar masses, there will arise vacua with $`M0`$, $`B0`$,$`\stackrel{~}{B}0`$, thus we get the vacua with chiral symmetry breaking but baryon number symmetry also spontaneously breaking.
When $`N_fN_c`$, there is no possible trilinear term, but a new quadratic SUSY breaking term, $`h_BB\stackrel{~}{B}+h.c.`$ can be introduced. However, it produces no new effects compared with the case with only quadratic terms.
In summary, we studied the softly broken $`N=1`$ supersymmetric QCD with the inclusion of the trilinear terms and find that in comparison with only the quadratic soft breaking term, some remarkable effects on the phase structure can be produced. First, In case $`N_f>N_c+1`$, depending on trilinear coupling constant and other soft breaking parameters, we get vacua with both the chiral symmetry breaking and baryon number violation. Furthermore, with various choices of soft breaking parameters, we find different phase structures. Whereas in case with only quadratic breaking term, only exotic vacua with unbroken chiral symmetry arise. In addition, Seiberg’s duality seems to persist in both chiral symmetry broken and unbroken phases. Second when $`N_f=N_c+1`$, if the trilinear coupling is strong enough, there will arise vacua with spontaneously chiral symmetry breaking but baryon number violation. In the case with only quadratic terms, the origin of the moduli space is the only vacuum and is hence $`SU(N_f)_L\times SU(N_f)_R\times U(1)_B\times U(1)_R`$ i nvariant. The significance of this investigation is providing an enlightenment that there is a long way to go for us to understand the non-perturbative QCD starting from the soft breaking of supersymmetric QCD. |
warning/0001/hep-ph0001140.html | ar5iv | text | # 𝑍' Bosons and Kaluza-Klein Excitations at Muon Colliders 1footnote 11footnote 1To appear in the Proceedings of the 5^{th} International Conference on Physics Potential and Development of μ⁺μ⁻ Colliders, Fairmont Hotel, San Francisco, CA, 15-17 December 1999
## Search Reaches for Z’ Bosons
The indirect search reach for new gauge bosons at future colliders has been the subject of much investigation but with few new results in the past couple of years except for refinements of previously existing analyses. We refer the reader to the summaries provided in Ref.snow .
## KK Excitations of SM Gauge Bosons
In theories with extra dimensions, $`d1`$, the gauge fields of the Standard Model(SM) will have Kaluza-Klein(KK) excitations if they are allowed to propagate in the bulk of the extra dimensions. If such a scenario is realized then, level by level, the masses of the excited states of the photon, $`Z`$, $`W`$ and gluon would form highly degenerate towers. The possibility that the masses of the lowest lying of these states could be as low as $``$ a few TeV or less leads to a very rich and exciting phenomenology at future and, possibly, existing collidersold . For the case of one extra dimension compactified on $`S^1/Z_2`$ the spectrum of the excited states is given by $`M_n=n/R`$ and the couplings of the excited modes relative to the corresponding zero mode to states remaining on the wall at the orbifold fixed points, such as the SM fermions, is simply $`\sqrt{2}`$ for all $`n`$.
If such KK states exist what is the lower bound on their mass? We already know from direct $`Z^{}/W^{}`$ and dijet bump searches at the Tevatron from Run I that they must lie above $`0.85`$ TeV. A null result for a search made with data from Run II will push this limit to $`1.1`$ TeV. To do better than this at present we must rely on the indirect effects associated with KK tower exchange. Such limits rely upon a number of additional assumptions, in particular, that the effect of KK exchanges is the only new physics beyond the SM. The strongest and least model-dependent of these bounds arises from an analysis of charged current contact interactions at both HERA and the Tevatroncornet where one obtains a bound of $`R^1>3.4`$ TeV. Similar analyses have been carried out by a number of authorshost ; rw ; the best limit arises from an updated combined fit to the precision electroweak data as presented at the 1999 summer conferences and yieldskktest $`R^1>3.9`$ TeV for the case of one extra dimension. From the previous discussion we can also draw a further conclusion for the case $`d=1`$: the lower bound $`M_1>3.9`$ TeV is so strong that the second KK excitations, whose masses must now exceed 7.8 TeV due to the above scaling law, will be beyond the reach of the LHC and thus the LHC will at most only detect the first set of KK excitations for $`d=1`$.
In all analyses that obtain indirect limits on $`M_1`$, one is actually constraining a dimensionless quantity such as
$$V=\underset{𝐧=1}{\overset{\mathrm{}}{}}\frac{g_𝐧^2}{g_0^2}\frac{M_w^2}{M_𝐧^2},$$
(1)
where, generalizing the case to $`d`$ additional dimensions, $`g_𝐧`$ is the coupling and $`M_𝐧`$ the mass of the $`n^{th}`$ KK level labelled by the set of $`d`$ integers n and $`M_w`$ is the $`W`$ boson mass which we employ as a typical weak scale factor. For $`d=1`$ this sum is finite since $`M_n=n/R`$ and $`g_n/g_0=\sqrt{2}`$ for $`n>1`$; one immediately obtains $`V=\frac{\pi ^2}{3}(M_w/M_1)^2`$ with $`M_1`$ being the mass of the first KK excitation. From the precision data one obtains a bound on $`V`$ and then uses the above expression to obtain the corresponding bound on $`M_1`$. For $`d>1`$, however, independently of how the extra dimensions are compactified, the above sum in $`V`$ diverges and so it is not so straightforward to obtain a bound on $`M_1`$. We also recall that for $`d>1`$ the mass spectrum and the relative coupling strength of any particular KK excitation now become dependent upon how the additional dimensions are compactified.
There are several ways one can deal with this divergence: ($`i`$) The simplest approach is to argue that as the states being summed in $`V`$ get heavier they approach the mass of the string scale, $`M_s`$, above which we know little and some new theory presumably takes over. Thus we should just truncate the sum at some fixed maximum value $`n_{max}M_sR`$ so that masses KK masses above $`M_s`$ do not contribute. ($`ii`$) A second possibility is to note that the wall on which the SM fermions reside is not completely rigid having a finite tension. The authors in Ref.wow argue that this wall tension can act like an exponential suppression of the couplings of the higher KK states in the tower thus rendering the summation finite, i.e., $`g_𝐧^2g_𝐧^2e^{(M_n/M_1)^2/n_{max}^2}`$, where $`n_{max}`$ now parameterizes the strength of the exponential cut-off. Antoniadiskktest has argued that such an gaussian suppression can also arise from considerations of string scattering amplitudes at high energies. ($`iii`$) A last scenarioschm is to note the possibility that the SM wall fermions may have a finite size in the extra dimensions which smear out and soften the couplings appearing in the sum to yield a finite result. In this case the suppression is also of the Gaussian variety. Table I shows how the $`d=1`$ lower bound of 3.9 TeV for the mass of $`M_1`$ changes as we consider different compactifications for $`d>1`$. We see that in some cases the value of $`M_1`$ is so large it will be beyond the mass range accessible to the LHC as it is for all cases of the $`d=3`$ example.
## SM KK States Before the Muon Collider
Let us return to the $`d=1`$ case at the LHC where the degenerate KK states $`\gamma ^{(1)}`$, $`Z^{(1)}`$, $`W^{(1)}`$ and $`g^{(1)}`$ are potentially visible. It has been shown kktest that for masses in excess of $`4`$ TeV the $`g^{(1)}`$ resonance in dijets will be washed out due to its rather large width and the experimental jet energy resolution available at the LHC detectors. Furthermore, $`\gamma ^{(1)}`$ and $`Z^{(1)}`$ will appear as a single resonance in Drell-Yan that cannot be resolved and looking very much like a single $`Z^{}`$ as can be seen in Fig.1. Thus if we are lucky the LHC will observe what appears to be a degenerate $`Z^{}/W^{}`$. How can we identify these states as KK excitations when we remember that the rest of the members of the tower are too massive to be produced? We remind the reader that many extended electroweak models exist which predict a degenerate $`Z^{}/W^{}`$. Without further information, it would seem likely that this would become the most likely guess of what had been found. In the case of the 4 TeV resonance there is sufficient statistics that the KK mass will be well measured and one can also imagine measuring the forward-backward asymmetry, $`A_{FB}`$, if not the full angular distribution of the outgoing leptons, since the final state muon charges can be signed. However, for such a heavy resonance it is unlikely that much further information could be obtained about its couplings and other properties based on the conclusion of several years of $`Z^{}`$ analyses. Thus we will never know from LHC data alone whether the first KK resonance has been discovered or, instead, some extended gauge model scenario has been realized. To make further progress we need a lepton collider.
It is well-known that future $`e^+e^{}`$ linear colliders(LC) operating in the center of mass energy range $`\sqrt{s}=0.51.5`$ TeV will be sensitive to indirect effects arising from the exchange of new $`Z^{}`$ bosons with masses typically 6-7 times greater than $`\sqrt{s}`$snow . This sensitivity is even greater in the case of KK excitations since towers of both $`\gamma `$ and $`Z`$ exist all of which have couplings larger than their SM zero modes. Furthermore, analyses have shown that with enough statistics the couplings of the new $`Z^{}`$ to the SM fermions can be extractedsnow in a rather precise manner, especially when the $`Z^{}`$ mass is already approximately known from elsewhere, e.g., the LHC. In the present situation, we imagine that the LHC has discovered and determined the mass of a $`Z^{}`$-like resonance in the 4-6 TeV range. Can the LC tell us anything about the couplings of this object?
We find that it is sufficient for our arguments below to do this solely for the leptonic channels. The idea is the following: we measure the deviations in the differential cross sections and angular dependent Left-Right polarization asymmetry, $`A_{LR}^{\mathrm{}}`$, for the three lepton generations and combine those with $`\tau `$ polarization data. Assuming lepton universality(which would be observed in the LHC data anyway), that the resonance mass is well determined, and that the resonance is an ordinary $`Z^{}`$ we perform a fit to the hypothetical $`Z^{}`$ coupling to leptons. To be specific, let us consider the case of only one extra dimension with a 4 TeV KK excitation and employ a $`\sqrt{s}=500`$ GeV collider with an integrated luminosity of 200 $`fb^1`$. The result of performing this fit demonstrate, as shown in Ref.new , that the coupling values are ‘well determined’, i.e., the size of the $`95\%`$ CL allowed region we find is quite small as we would have expected from previous $`Z^{}`$ analyses.
The only problem with the fit is that the $`\chi ^2`$ is very large leading to a very small confidence level, i.e., $`\chi ^2/d.o.f.=95.06/58`$ or CL=$`1.55\times 10^3`$! For an ordinary $`Z^{}`$ it has been shown that fits of much higher quality, based on confidence level values, are obtained by this same procedure. Fig.2 shows the results for the CL following the above approach as we vary both the luminosity and the mass of the first KK excitation at a 500 GeV $`e^+e^{}`$ linear collider. From this analysis one finds that the resulting CL is below $`10^3`$ for a first KK excitation with a mass of 4(5,6) TeV when the integrated luminosity at the 500 GeV collider is 200(500,900)$`fb^1`$ whereas at a 1 TeV for excitation masses of 5(6,7) TeV we require luminosities of 150(300,500)$`fb^1`$ to realize this same CL. Barring some unknown systematic effect the only conclusion that one could draw from such bad fits is that the hypothesis of a single $`Z^{}`$, and the existence of no other new physics, is simply wrong. If no other exotic states are observed below the first KK mass at the LHC, this result would give very strong indirect evidence that something more unusual than a conventional $`Z^{}`$ had been found but it cannot prove that this is a KK state.
## SM KK States at Muon Colliders
In order to be completely sure of the nature of the first KK excitation, we must produce it directly at a higher energy lepton collider and sit on and near the peak of the KK resonance. To reach this mass range will most likely require a Muon Collider. Sitting on the resonance there are a very large number of quantities that can be measured: the mass and apparent total width, the peak cross section, various partial widths and asymmetries etc. From the $`Z`$-pole studies at SLC and LEP, we recall a few important tree-level results which we would expect to apply here as well provided our resonance is a simple $`Z^{}`$. First, we know that the value of $`A_{LR}`$, as measured on the $`Z`$ by SLD, does not depend on the fermion flavor of the final state and second, that the relationship $`A_{LR}A_{FB}^{pol}(f)=A_{FB}^f`$ holds, where $`A_{FB}^{pol}(f)`$ is the polarized Forward-Backward asymmetry as measured for the $`Z`$ at SLC and $`A_{FB}^f`$ is the usual Forward-Backward asymmetry. The above relation is seen to be trivially satisfied on the $`Z`$(or on a $`Z^{}`$) since $`A_{FB}^{pol}(f)=\frac{3}{4}A_f`$, $`A_{LR}=A_e`$, and $`A_{FB}^f=\frac{3}{4}A_eA_f`$. Both of these relations are easily shown to fail in the present case of a ‘dual’ resonance though they will hold if only one particle is resonating.
A short exercisekktest yields the results in Fig.2 explicitly showing the flavor dependence of $`A_{LR}`$. In principle, to be as model independent as possible in a numerical analysis, we should allow the individual widths $`\mathrm{\Gamma }_i`$ of the two resonances to be greater than or equal to their SM values as such heavy KK states may decay to SM SUSY partners as well as to presently unknown exotic states. Since the expressions above only depend upon the ratio of widths, we let $`R=\lambda R_0`$ where $`R_0`$ is the width ratio obtained assuming that the KK states have only SM decay modes. We then treat $`\lambda `$ as a free parameter in what follows and explore the range $`1/5\lambda 5`$. Once $`\lambda `$ is determined from the value of one observable all of the electroweak parameters of the dual resonance are completely fixed and can directly compared with data proving that a composite resonance corresponding to the first KK excitation has been discoverednew .
In Figs. 3a and 3b we show that although on-resonance measurements of the electroweak observables, being quadratic in the $`Z^{(1)}`$ and $`\gamma ^{(1)}`$ couplings, will not distinguish between the usual KK scenario and that of the Arkani-Hamed and Schmaltz(AS) (whose KK couplings to quarks are of opposite sign from the conventional assignments for odd KK levels since quarks and leptons are assumed to be separated by a distance $`D=\pi R`$ in their scenario) the data below the peak in the hadronic channel will easily allow such a separation. The cross section and asymmetries for $`\mu ^+\mu ^{}e^+e^{}`$ is, of course, the same in both cases. Such data can be collected by using radiative returns if sufficient luminosity is available. The combination of on and near resonance measurements will thus completely determine the nature of the resonance as well as the separation between various fermions on the wall. Fig.4 shows that with even larger energies muon colliders will be able to probe both the number of extra dimensions as well as the geometry of their compactification manifolds since these can be uniquely determined by the KK excitation spectrum. |
warning/0001/quant-ph0001003.html | ar5iv | text | # Photon-added one-photon and two-photon nonlinear coherent states
## Abstract
From the photon-added one-photon nonlinear coherent states $`a^m|\alpha ,f`$, we introduce a new type of nonlinear coherent states with negative values of $`m.`$ The nonlinear coherent states corresponding to the positive and negative values of $`m`$ are shown to be the result of nonunitarily deforming the number states $`|m`$ and $`|0`$, respectively. As an example, we study the sub-Poissonian statistics and squeezing effects of the photon-added geometric states with negative values of $`m`$ in detail. Finally we investigate the photon-added two-photon nonlinear coherent states and find they are still the two-photon nonlinear coherent states with certain nonlinear functions.
PACS: 42.50.Dv, 03.65.Fd
1. Introduction
Recently there has much interest in the study of nonlinear coherent states(NLCSs), which are right-hand eigenstates of the product of the boson annihilation operator $`a`$ and a nonlinear function $`f(N)`$ of the number operator $`N`$,
$$f(N)a|\alpha ,f=\alpha |\alpha ,f.$$
(1)
Here $`\alpha `$ is a complex eigenvalue. It has been shown that a class of NLCSs may appear as stationary states of the centre-of-mass motion of a trapped ion. These nonlinear coherent states exhibit nonclassical features like squeezing and self-splitting.
Another type of interesting nonclassical states consists of the photon-added states
$$|m,\psi =\frac{a^m|\psi }{\psi |a^ma^m|\psi },$$
(2)
where $`|\psi `$ may be an arbitrary quantum state, $`a^{}`$ is the boson creation operator, $`m`$ is a non-negative integer-the number of added quanta. For the first time these states were introduced by Agarwal and Tara as photon-added coherent states. The photon-added squeezed states ,even(odd) photon-added states and photon-added thermal state were also introduced and studied. The photon-added states can be produced in the interaction of a two-level atom with a cavity field initially prepared in the state $`|\psi `$.
Sivakumar showed that the photon-added coherent states are nonlinear coherent states. As a generalization we showed a general result that photon-added NLCSs(PANLCSs) are still NLCSs with different nonlinear functions. The PANLCSs are defined as
$$|m,\alpha ,f=\frac{a^m|\alpha ,f}{\alpha ,f|a^ma^m|\alpha ,f}.$$
(3)
They satisfy
$$f(Nm)[1m/(N+1)]a|m,\alpha ,f=\alpha |m,\alpha ,f.$$
(4)
As seen from Eq.(4), the PANLCS is an NLCS with the nonlinear function $`f(Nm)[1m/(N+1)].`$ Naturally Eq.(4) reduces to Eq.(1) when $`m=0.`$ The well-known geometric states(GSs) and negative binomial states(NBSs) are NLCSs. Therefore, the photon-added GSs and photon-added NBSs are still NLCSs and are special cases of the PANLCSs.
In the present paper we show that the PANLCSs are the result of nonunitarily deforming the number state $`|m.`$ We introduce the PANLCS with negative values of $`m`$ , which are the result of nonunitairily deforming the vacuum state $`|0`$. As an example, we study the sub-Poissonian statistics and squeezing effects of the photon-added geometric states with negative values of $`m`$ in detail. We also investigate the photon-added two-photon nonlinear coherent states.
2.The PANLCS as deformed number state $`|m`$
In this section we show that the PANLCS can be written as a nonunitarily deformed number state. This is achieved by the method given by Shanta et al. Here we give a brief review of the method.
Consider an annihilation operator $`A`$ which annihilates a set of number states $`|n_i,i=1,2,\mathrm{}k.`$ Then we can construct a sector $`S_i`$ by repeatedly applying $`A^{},`$ the adjoint of $`A,`$ on the number state $`|n_i.`$ Thus we have $`k`$ sectors corresponding to the states that are annihilated by $`A.`$ A given sector may turn out to be either of finite or infinite dimension. If a sector, say $`S_j,`$ is of infinite dimension then we can construct an operator $`G_j^{}`$ such that $`[A,G_j^{}]=1`$ holds in that sector. Then the eigenstates of $`A`$ can be written as $`\mathrm{exp}(\alpha G_j^{})|n_j.`$ If an operator $`A`$ is of the form $`f(N)a^p,`$ where $`p`$ is non-negative integer, such that it annihilates the number state $`|j`$ then $`G_j^{}`$ is constructed as
$$G_j^{}=\frac{1}{p}A^{}\frac{1}{AA^{}}(a^{}a+pj).$$
(5)
It is interesting that the operator $`f(Nm)[1m/(N+1)]a`$ in Eq.(4) annihilates both the vacuum state $`|0`$ and Fock state $`|`$$`m.`$ The states between the vacuum state and Fock state $`|`$$`m`$ are not annihilated. To discuss the case of the PANLCS $`|m,\alpha ,f`$ let
$$A=f(Nm)[1m/(N+1)]a,A^{}=a^{}f(Nm)[1m/(N+1)]$$
(6)
We construct sector $`S_0`$ by repeated applying $`A^{}`$ on the vacuum state. $`S_0`$ is the set $`|i,i=0,1,2,\mathrm{},m1`$ and it is of finite dimension. The sector $`S_m,`$ built by the repeated application of $`A^{}`$ on $`|m,`$ is the set $`|i,i=m,m+1,\mathrm{}`$ and it is of infinite dimension. Hence we can construct an operator $`G^{}`$ such that $`[A,G^{}]=1`$ holds in $`S_m.`$ To construct $`G^{},`$ we set $`p=1`$ and $`j=m`$ in Eq.(5) and this yields
$$G^{}=a^{}\frac{1}{f(Nm)}$$
(7)
In fact, by direct verification, we have
$$[f(Nm)[1m/(N+1)]a,a^{}\frac{1}{f(Nm)}]=1.$$
(8)
Therefore the PANLCS can be written as
$$|m,\alpha ,f=\mathrm{exp}(G^{})|m=\mathrm{exp}[a^{}\frac{1}{f(Nm)}]|m$$
(9)
up to a normalization constant. From the above equation it is shown that the PANLCS can be viewed as nonunitarily deformed Fock(number) state $`|m.`$
3.The PANLCS with negative $`m`$
The form of $`A,`$ given by Eq.(6), suggests that it is a well-defined operator-valued function also for negative values of $`m`$ on the Fock space. In this section the PANLCS with negative $`m`$ is constructed. Denoting the the PANLCS with negtive $`m`$ by $`|m,\alpha ,f,`$ the equation to determine them are
$$f(N+m)[1+m/(N+1)]a|m,\alpha ,f=\alpha |m,\alpha ,f.$$
(10)
The operator $`A=f(N+m)[1+m/(N+1)]a`$ only annihilates the vacuum state. When $`f(N)1,`$ the state $`|m,\alpha ,f`$ reduces to that studied in Ref.. The sector $`S_0,`$ built by the repeated application of $`A^{}=`$ $`a^{}f(N+m)[1+m/(N+1)]`$ on $`|0,`$ is the set $`|i,i=0,1,\mathrm{}`$ and it is just the infinite dimensional Fock space. To construct $`G^{},`$ corresponding to the operator $`A=f(N+m)[1+m/(N+1)]a`$ , we set $`p=1`$ and $`j=0`$ in Eq.(5) and this yields
$$G^{}=a^{}\frac{N+1}{f(N+m)(N+m+1)}$$
(11)
Thus the PANLCS with negative $`m`$ can be written as
$$|m,\alpha ,f=\mathrm{exp}(G^{})|0=\mathrm{exp}[a^{}\frac{N+1}{f(N+m)(N+m+1)}]|0$$
(12)
up to a normalization constant. The state $`|m,\alpha ,f`$ is obtained by nonunitarily deforming the vacuum state $`|0`$ while the state $`|m,\alpha ,f`$ is obtained by nonunitarily deforming the Fock state $`|m.`$
The PANLCS is obtained by the action of $`a^m`$ on the NLCS $`|\alpha ,f`$. The state $`|m,\alpha ,f`$ can be written in a similar form using the inverse operators $`a^1`$ and $`a^1`$. These operators are defined in terms of their actions on the number state $`|n`$ as follows
$`a^1|n`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{n+1}}}|n+1,`$ (13)
$`a^1|n`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{n}}}|n1,`$ (14)
$`a^1|0`$ $`=`$ $`0.`$ (15)
Using these inverse operators the state $`|m,\alpha ,f`$ can be rewritten as $`|m,\alpha ,f=a^ma^m|\alpha ,f^{}`$ up to a normalization constant. Here $`|\alpha ,f^{}`$ is the NLCS with the nonlinear function $`f^{}(N)=f(N+m).`$ The state $`|m,\alpha ,f`$ is obtained by the action of the operator $`a^ma^m`$ on the NLCS $`|\alpha ,f^{}`$ while the state $`|m,\alpha ,f`$ is obtained by the action of the operator $`a^m`$ on the NLCS $`|\alpha ,f.`$
From Eq.(12) the number state expansion of the PANLCS with negative $`m`$ can be easily obtained as
$$|m,\alpha ,f=\underset{n=0}{\overset{\mathrm{}}{}}\frac{\alpha ^n\sqrt{n!}}{f(n+m1)\mathrm{}f(0)(n+m)!}|n$$
(16)
up to a normalization constant. The expansion is useful in the following discussions.
4.Photon-added geometric state with negative $`m`$
In this section we consider a special example of the PANLCS with negative $`m,`$ the photon-added geometric state with negative $`m.`$
The geometric state is defined as
$$|\eta =\eta ^{1/2}\underset{n=0}{\overset{\mathrm{}}{}}(1\eta )^{n/2}|n,\text{ }0<\eta <1,$$
(17)
It satisfies
$$\frac{1}{\sqrt{N+1}}a|\eta =\sqrt{1\eta }|\eta .$$
(18)
In comparison with Eq.(1), we see that the geometric state is an NLCS with the nonlinear function $`1/\sqrt{N+1}.`$ The photon-added geometric state is defined as
$`|m,\eta `$ $`=`$ $`{\displaystyle \frac{a^m|\eta }{\eta |a^ma^m|\eta }}`$ (19)
$`=`$ $`\eta ^{(m+1)/2}{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}\left({\displaystyle \genfrac{}{}{0pt}{}{m+n}{n}}\right)^{n/2}(1\eta )^{n/2}|n,`$ (20)
which is just the negative binomial state introduced by Barnett. We have studied the statistical properties and algebraic characteristics of the photon-added geometric state in detail. From Eqs.(4) and (16) we get
$$\frac{\sqrt{Nm+1}}{N+1}a|m,\eta =\sqrt{1\eta }|m,\eta .$$
(21)
The state $`|`$$`m,\eta `$ is an NLCS with the nonlinear function $`f(N)=\sqrt{Nm+1}/(N+1).`$ When $`m=0,`$ Eq.(19) reduces to Eq.(16) as we expected.
We would like to study the state $`|m,\eta ,`$ the photon-added geometric state with negative values of $`m,`$ which satisfies
$$\frac{\sqrt{N+m+1}}{N+1}a|m,\eta =\sqrt{1\eta }|m,\eta .$$
(22)
From Eq.(14), the number state expansion of the state $`|m,\eta `$ is given by
$$|m,\eta =\text{ }\sqrt{\frac{m!}{{}_{2}{}^{}F_{1}^{}(1,1;m+1;1\eta )}}\underset{n=0}{\overset{\mathrm{}}{}}(1\eta )^{n/2}\sqrt{\frac{n!}{(n+m)!}}|n,$$
(23)
where $`{}_{2}{}^{}F_{1}^{}(1,1;m+1;1\eta )`$ is the hypergeometric function.
The photon statistics of a quantum state can be conveniently studied by Mandel’s $`Q`$-parameter
$$Q=\frac{N^2N^2N}{N}$$
(24)
A negative $`Q`$ indicates that the photon number distribution is sub-Poissonian and it is a nonclassical feature. A positive $`Q`$ indicates the super-Poissonian distribution and $`Q`$=0 indicates Poissonian distribution. The photon-added geometric state $`|m,\eta `$ can be sub-Poissoian depending on the parameter $`\eta `$. For the state $`|m,\eta (`$Eq.(21)$`),`$ the mean value of $`N^k`$ is easily obtained as
$$N^k=\frac{m!}{{}_{2}{}^{}F_{1}^{}(1,1;m+1;1\eta )}\underset{n=0}{\overset{\mathrm{}}{}}n^k(1\eta )^n\frac{n!}{(n+m)!}.$$
(25)
In Fig.1 the $`Q`$-parameter, calculated using Eqs.(22) and (23), for the state $`|m,\eta `$ is shown as a function of $`\eta .`$ The $`Q`$-parameter is always greater than zero indicating that they are super-Poissonian. For larger values of $`\eta ,`$ the $`Q`$-parameter is close to zero since the state $`|m,\eta `$ reduces to $`|0`$ in the limit $`\eta 1`$.
5.Squeezing in $`|m,\eta `$
Define the quadrature operators $`X`$(coordinate) and $`Y`$ (momentum) by
$$X=\frac{1}{2}(a+a^{}),Y=\frac{1}{2i}(aa^{}).$$
(26)
Then their variances
$$Var(X)=X^2X^2,Var(Y)=Y^2Y^2$$
(27)
obey the Heisenberg’s uncertainty relation
$$Var(X)Var(Y)\frac{1}{16}.$$
(28)
If one of the variances is less than 1/4, the squeezing occurs. In the present case, $`a`$ and $`a^2`$ are real. Thus, the variances of $`X`$ and $`Y`$ can be written as
$`Var(X)`$ $`=`$ $`{\displaystyle \frac{1}{4}}+{\displaystyle \frac{1}{2}}(a^{}a+a^22a^2),`$ (29)
$`Var(Y)`$ $`=`$ $`{\displaystyle \frac{1}{4}}+{\displaystyle \frac{1}{2}}(a^{}aa^2).`$ (30)
From Eq.(21), the expectation value $`m,\eta |a^k|m,\eta `$ is directly obtained as
$`m,\eta |a^k|m,\eta `$ $`=`$ $`{\displaystyle \frac{m!}{{}_{2}{}^{}F_{1}^{}(1,1;m+1;1\eta )}}`$ (32)
$`{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}(1\eta )^{n+k/2}{\displaystyle \frac{(n+k)!}{\sqrt{(n+m)!(n+m+k)!}}}`$
The variances of $`X`$ and $`Y`$ can be calculated from Eqs.(27), (28) and (29). In Fig.2 we show the variances of the quadrature operators $`X`$ and $`Y`$ as a function of $`\eta `$ for different values of $`m.`$ The squeezing exists in the quadrature $`Y.`$ For the quadrature $`Y,`$ the degree of the squeezing becomes deep with the increase of the parameter $`m.`$ In the limit $`\eta 1,`$ the variances are all equal to $`1/4.`$ This is because the state $`|m,\eta `$ reduces the vacuum state $`|0`$ in this limit.
5. Photon-added two-photon nonlinear coherent states
In this section, we investigate the photon-added two-photon nonlinear coherent states. The two-photon nonlinear coherent state is defined as
$$F(N)a^2|\alpha ,F=\alpha |\alpha ,F,$$
(33)
and the corresponding photon-added two-photon nonlinear coherent state is
$$|\alpha ,F,m=a^m|\alpha ,F$$
(34)
up to a normailization constant.
Acting the operator $`a^2a^m`$ on Eq.(33) from the left, we obtain
$$F(Nm+2)a^2a^ma^2|\alpha ,F=\alpha (N+1)(N+2)a^{(m2)}|\alpha ,F.$$
(35)
Since
$$a^2a^ma^2=(N+4m)(N+3m)a^2a^{(m2)},$$
(36)
we obtain
$$F(Nm+2)(N+4m)(N+3m)a^2a^{(m2)}|\alpha ,F=\alpha (N+1)(N+2)a^{(m2)}|\alpha ,F.$$
(37)
Let $`m2m`$ in the above equation and note that the operator $`(N+1)(N+2)`$ is positive in the whole Fock space, we get
$$F(Nm)(1\frac{m}{N+2})(1\frac{m}{N+1})a^2|\alpha ,F,m=\alpha |\alpha ,F,m.$$
(38)
This shows that the photon-added nonlinear coherent states $`|\alpha ,F,m`$ are still nonlinear coherent states with the nonlinear function
$$F(Nm)(1\frac{m}{N+2})(1\frac{m}{N+1}).$$
(39)
Since the squeezed vaccum state and squeezed first Fock state are two-photon nonlinear coherent state, we conclude that the photon-added squeezed vaccum state and photon-added squeezed first Fock state are also two-photon nonlinear coherent states as discussed in Ref.. We can also introduce the photon-added two-photon nonlinear coherent states with negative $`m`$ and make a similar discussion as one-photon case. We will not explicitly present them here.
6.Conclusions
In conclusion, we have studied a special NLCSs, the PANLCSs. From the PANLCS we introduce a new type of quantum state, the PANLCS with negative values of $`m`$. The states corresponding to the positive and negative values of $`m`$ are shown to be the result of nonunitarily deforming the number states $`|m`$ and $`|0`$, respectively. As a example, we study the sub-Poissonian statistics and squeezing effects in the photon-added geometric state with negative values of $`m`$ in detail. The results shows that photon-added geometric state with negative values of $`m`$ are always super-Poissonian and the state can be squeezed in the quadrature $`Y.`$ We also consider the photon-added two-photon nonlinear coherent states and find a similar concusion as one-photon case, i.e, the photon-added two-photon nonlinear coherent states are still two-photon nonlinear coherent states with certain nonlinear functions.
Acknowledgment: The author thanks for the discussions with Prof. Hong-Chen Fu and help of Prof.Chang-Pu Sun, Shao-Hua Pan, Guo-Zhen Yang. The work is partially supported by the National Science Foundation of China with grant number:19875008.
Figure Captions:
Figure1, Mandel’s Q parameter as a function of $`\eta `$ for different values of $`m.`$
Figure2, Variances of the quadrature operators $`X`$ and $`Y`$ as a function of $`\eta `$ for different values of $`m.`$ |
warning/0001/quant-ph0001041.html | ar5iv | text | # Untitled Document
CSL Collapse Model And Spontaneous Radiation: An Update
Philip Pearle and James Ring
Department of Physics, Hamilton College, Clinton, N.Y.,13323
Juan I. Collar
EP Division, CERN CH-1211
Frank T. Avignone III
Department of Physics and Astronomy, University of South Carolina, Columbia South Carolina 29208
Abstract
A brief review is given of the Continuous Spontaneous Localization (CSL) model in which a classical field interacts with quantized particles to cause dynamical wavefunction collapse. One of the model’s predictions is that particles “spontaneously” gain energy at a slow rate. When applied to the excitation of a nucleon in a Ge nucleus, it is shown how a limit on the relative collapse rates of neutron and proton could be obtained, and a rough estimate is made from data. When applied to the spontaneous excitation of 1s electrons in Ge, by a more detailed analysis of more accurate data than previously given, an updated limit is obtained on the relative collapse rates of the electron and proton, suggesting that the coupling of the field to electrons and nucleons is mass proportional.
1. Introduction
It is appropriate to discuss comparison of experiment to a theory with fundamental pretensions in a volume dedicated to Dan Greenberger whose own theoretical work of a fundamental nature has seldom been far from testability.
In standard quantum theory (SQT) the statevector evolves in two ways. One evolution procedes smoothly via Schrodinger’s equation. The other is the abrupt (and ill defined) “collapse” of the statevector. This is the replacement of a statevector equal to a sum of vectors (each describing a different outcome of an “experiment”) by one vector in the sum. One might guess that this dual evolution is indicative of a fundamental deficiency in present day physics. In the hope of finding new physics, one may begin by trying to modify Schrodinger’s equation so that the statevector undergoes only a smooth evolution, giving both the usual quantum behavior and the collapse behavior.
This program, begun three decades ago,<sup>1,2</sup> received a crucial impetus one decade ago from the work of Ghirardi, Rimini and Weber (GRW)<sup>3</sup> and has evolved into the Continuous Spontaneous Localization (CSL) model.<sup>4</sup> At present, this is the only fully developed nonrelativistic collapse model, with definite predictions applicable to any nonrelativistic experimental situation.
In the CSL model, a term which depends upon a randomly fluctuating field $`w(𝐱,t)`$ is added to Schrodinger’s equation. (The physical nature of this field is unspecified, but metric fluctuations,<sup>5</sup> possibly with a tachyonic spectrum,<sup>6</sup> have been suggested.) The probabilistic behavior of Nature is explained as due to our lack of control of the field $`w`$. When an experiment is under way,the particles in the system+apparatus interact with the particular sample field $`w`$ that is present, causing a rapid evolution of the statevector to one of the alternative outcomes of the experiment. A different sample field leads to a different outcome. CSL also specifies the probability that a particular sample field $`w(𝐱,t)`$ actually occurs. When all possible fields are taken into account, together with their probabilities, the result is that each outcome occurs with (essentially — see next paragraph) the probability predicted by SQT. Thus two relations, the Modified Schrodinger Equation and the Probability Rule constitute the CSL model.
As with any modification of SQT, one expects—and hopes—for certain specially designed experiments where SQT and CSL lead to different predictions, making tests possible. For example, one such test, presently not practicable, is a two slit interference experiment with a sufficiently large bound state object. Once the two wavepackets for the object leave the slits, SQT says that their amplitudes will never change so that interference is possible at any time. CSL says that the amplitudes will fluctuate and, after a long enough wait, eventually one of the packets will become negligible in amplitude, giving no interference pattern. Such an interference experiment with e.g., 90Å diameter drops of mercury<sup>7</sup> over a time interval of seconds could provide such a test. However, this is a difficult experiment. For example, it is hard to prevent the two packets from being put into different angular momentum eigenstates by interaction with the environment, and then they would not interfere for this reason.<sup>8</sup>
The tests that are most practicable at present stem from the consequence of CSL that the collapse process imparts energy to particles. (One may think of this energy as provided by the field $`w`$.) The reason is that collapse entails the narrowing of wavefunctions. By the uncertainty principle, this leads to an increased momentum spread, and thus to an increased energy. Thus, any bound ground state of, e.g., atoms or nucleii, will be excited by the collapse part of the Schrodinger equation.<sup>3,4,9</sup> The usual part of the Schrodinger equation will describe the radiation emitted as the system returns to the ground state. Also, a free charged particle is “shaken” by the field $`w`$, and so it will radiate<sup>10</sup>. Similarly, the quarks inside a proton should be excited, and the proton should radiate mesons.<sup>11</sup> Thus, a signature of CSL is that matter should emit “spontaneous radiation.” In section 3, various radiation rates are given.
It is interesting that, at this time, quite a number of low noise experiments are being undertaken to look for radiation appearing in an apparatus for a variety of reasons, e.g., because of collisions with purported Dark Matter. Some of these experiments are sensitive enough to provide useful constraints on the parameters of CSL.
Some experiments which look at radiation appearing in a slab of Germanium are described in section 4. In section 5 we show how data from one such experiment, “Rico Grande,” applied to the spontaneous excitation of a proton in a Ge nucleus, can provide a limit on the relative collapse rates of neutron and proton. Only a rough estimate is given because greater precision requires a more careful calculation of Ge nuclear dipole matrix elements than we are prepared to give here.
In section 6 we consider the spontaneous ionization rate of a 1s electron in a Ge atom. In a previous paper,<sup>12</sup> it was argued that the upper limit on this rate given by a single data point from the “TWIN” experiment<sup>13</sup> suggested that the coupling of the field $`w`$ to an electron or nucleon (here assumed to be the same for neutron and proton) is proportional to the particle’s mass, supporting previous proposals that there is a connection between gravity and collapse<sup>5,14,15</sup>. Unfortunately it subsequently turned out that the data point was inaccurate (see section 4). However, the more complete analysis on “COSME” data presented here gives essentially the previous result.
Eventually, from such experiments, one may hope that the collapse rate parameter of CSL will either be constrained to be so small that the model will be ruled out—or that spontaneous radiation from collapse will actually be observed!
2. CSL
Underlying CSL are two mechanisms. One is the Gambler’s Ruin mechanism. This explains how the random process embodied in the noise $`w`$ produces the probabilities of SQT for the collapsed states.<sup>16</sup> The other is the GRW “hitting” mechanism. It allows collapse to occur rapidly, for macroscopic objects, to states which we see around us (localized objects), while microscopic objects are scarcely affected.<sup>3</sup>
Here is the Gambler’s Ruin analogy. Suppose, at the beginning of a game, gambler 1 (2) starts with $`d_1(0)`$ ($`d_2(0)`$) dollars, and $`d_1(0)+d_2(0)=100`$. This is to be analogous to the initial statevector $`|\psi (0)>=a_1(0)|1>+a_2(0)|2>`$, with the correspondence $`d_i(0)/100|a_i(0)|^2`$. The gamblers toss a coin (analogous to the fluctuating field $`w`$) and, depending on the result, one gives a dollar to the other. As the game proceeds, the $`d_i(t)`$ fluctuate, just as do the squared amplitudes $`|a_i(t)|^2`$. One gambler finally wins all the money and the game stops, with e.g., gambler 2 winning with probability $`d_2(0)/100`$. Precisely analogously, collapse finally occurs, e.g., with $`|\psi (t)>0|1>+1|2>`$ with probability $`|a_2(0)|^2`$. This is, of course, the probability predicted by SQT of the outcome $`|2>`$ if $`|1>`$ and $`|2>`$ represent two states of an apparatus.
The GRW model postulates a physical process which produces a sudden random change (“hit”) of a many-particle wavefunction: the wavefunction is multiplied by a gaussian function $`\mathrm{exp}(𝐱_n𝐳)^2/2a^2`$, where $`𝐱_n`$ is the position coordinate of the nth particle. The center of the gaussian, $`𝐳`$, is chosen according to a Probability Rule which depends in a certain way upon the wavefunction, making it most likely that $`𝐳`$ is located where the wavefunction is largest. The effect of a hit is to narrow to width $`a`$ the part of the wavefunction which depends upon the nth particle: GRW chose the mesoscopic length $`a10^5`$ cm. A hit on one particle occurs at a slow rate $`\lambda `$: GRW chose $`\lambda 10^{16}`$sec$`{}_{}{}^{1}`$ once in 300 million years. But, each particle is equally likely to be hit, so a hit on an N particle object occurs rapidly, on average in $`1/\lambda N`$sec. The wavefunction of an $`N`$ particle object in a superposition of states, each describing the object in a different place, has the particles entangled in such a way that one hit on one such particle causes the wavefunction to collapse in $`1/\lambda N`$sec to one of the states in the superposition. This explains how macroscopic objects are always observed as localized. (A defect of this model is that the (anti) symmetry of the wavefunction is destroyed by the hitting process.)
CSL may be thought of as embodying a continuous hitting process. A hit occurs every $`\mathrm{\Delta }t`$ sec, but the hit wavefunction is multiplied by $`\mathrm{\Delta }t`$ and added to the original wavefunction. Thus the wavefunction shape fluctuates gradually (not suddenly as in GRW’s model), and the gambler’s ruin dynamics is obtained. The (anti) symmetry of the wavefunction is properly preserved in CSL.
The CSL modified Schrodinger equation is
$$\frac{d|\psi ,t>_w}{dt}=iH|\psi ,t>_w\frac{1}{4\lambda }𝑑𝐱[w(𝐱,t)2\lambda A(𝐱)]^2|\psi ,t>_w$$
$`(2.1)`$
Given an arbitrary field $`w(𝐱,t)`$, one may solve (2.1) to find how the statevector evolves under its influence. Operator A(x) is
$$A(𝐱)\underset{\alpha }{}g_\alpha \frac{1}{(\pi a^2)^{\frac{3}{4}}}𝑑𝐳N_\alpha (𝐳)e^{{\displaystyle \frac{(𝐱𝐳)^2}{2a^2}}}$$
$`(2.1a)`$
The integral in (2.1a) is essentially the number of particles in a sphere of diameter $`a`$ centered around $`𝐱`$. $`N_\alpha (𝐳)`$ is the number density operator for particles of type $`\alpha `$, e.g., electrons , protons and neutrons, and $`\lambda g_\alpha ^2`$ is their one-particle collapse rate (see Eq. (2.5) below).
Thus, the parameters which characterize CSL are $`a`$, $`\lambda `$ and the ratios of the $`g_\alpha `$’s, e.g., for ordinary matter, $`g_n/g_p`$ and $`g_e/g_p`$ where the subscripts $`p`$, $`n`$, and $`e`$ refer to the proton, neutron and electron. With no loss of generality we may take $`g_p=1`$, so $`\lambda `$ is an individual proton’s collapse rate.
CSL requires a second equation, giving the probability density $`P(w)`$ functional that the field $`w(𝐱,t)`$ occurs:
$$P(w)=_w<\psi ,t|\psi ,t>_w$$
$`(2.2)`$
Because the evolution (2.1) is nonunitary, the norm of the statevector $`|\psi ,t>_w`$ changes with time. Eq.(2.2) says that the most probable fields to occur are those which lead to statevectors of largest norm.
Eqs.(2.1) and (2.2) ensure that an initial statevector, which is in a superposition of states of different particle number density, evolves toward one of these states for each probable $`w`$—and that the ensemble of all evolutions is such that each final state occurs with (essentially) the SQT probability. However, we shall not discuss here how Eqs.(2.1), (2.2) lead to collapse for an individual statevector. Instead we shall go right to the appropriate object for discussing experimental predictions, the density matrix $`\rho `$. The evolution equation for the density matrix whch follows from Eqs.(2.1), (2.2) can be shown to be <sup>4</sup>
$$\begin{array}{cc}& \frac{<x|\rho (t)|x^{}>}{t}=i<x|[H,\rho (t)]|x^{}>\hfill \\ & \frac{\lambda }{2}\underset{j=1}{\overset{N}{}}\underset{k=1}{\overset{N}{}}g_{\alpha (j)}g_{\alpha (k)}[\mathrm{\Phi }(𝐱_j𝐱_k)+\mathrm{\Phi }(𝐱_j^{}𝐱_k^{})2\mathrm{\Phi }(𝐱_j𝐱_k^{})]<x|\rho (t)|x^{}>\hfill \end{array}$$
$`(2.3)`$
where $`|x>=|𝐱_1,𝐱_2,\mathrm{}>`$ is the position eigenstate for all particles and
$$\mathrm{\Phi }(𝐳)e^{{\displaystyle \frac{𝐳^2}{4a^2}}}$$
$`(2.4)`$
As a simple example of how Eq. (2.3) works, set $`H=0`$ so as to concentrate on the collapse dynamics alone, and consider a clump of particles of type $`\alpha `$ in a superposed state. That is, let the initial state be $`a_1(0)|1>+a_2(0)|1>`$, where $`|1>`$ and $`|2>`$ each describe N particles of type $`\alpha `$ in a localized state with dimensions $`<<a`$, but with centers of mass of the two states at a distance $`>>a`$ apart. Then $`\mathrm{\Phi }(𝐱_j𝐱_k)1`$ if $`𝐱_j`$, $`𝐱_k`$ are both located in region 1 (or both in 2) and $`\mathrm{\Phi }(𝐱_j𝐱_k^{})0`$ if $`𝐱_j`$, $`𝐱_k^{}`$ are located in regions 1 and 2 respectively. Therefore, Eq.(2.3) yields
$$\frac{<1|\rho (t)|2>}{t}=\frac{\lambda g_\alpha ^2}{2}\underset{j=1}{\overset{N}{}}\underset{k=1}{\overset{N}{}}(1+120)<1|\rho (t)|2>=\lambda g_\alpha ^2N^2<1|\rho (t)|2>$$
$`(2.5)`$
showing that the off-diagonal elements of $`\rho `$ decay at the rate $`\lambda N^2`$. This illustrates how the collapse rate is large for a superposition of states describing a large number of particles in different locations.
3. Excitation Rate Predictions
As mentioned in section 1, a byproduct of the collapse process is that particles gain energy. It is easy to show, using Eq.(2.3), that the average total energy $`\overline{H}(t)`$Tr$`H\rho (t)`$ ($`H=_{i=1}^N𝐩_i^2/2M_i+V(𝐱_1\mathrm{}𝐱_N)`$) increases according to
$$\frac{d\overline{H}(t)}{dt}=\frac{3\lambda }{2}\underset{j=1}{\overset{N}{}}g_{\alpha (j)}^2\frac{\mathrm{}^2}{2M_ja^2}$$
$`(3.1)`$
Assuming $`g_\alpha =1`$ for all particles, and using the GRW values for $`\lambda `$ and $`a`$, then $`10^{24}`$ nucleons gain $`.3`$ eV/sec and $`10^{24}`$ electrons gain $`600`$ eV/sec. This is quite small, corresponding to a temperature increase over the age of the universe of $`.001^{}`$K and $`2^{}`$K respectively. (The low particle density in the universe assures that the effect of this increased energy on the the cosmic radiation bath is negligible). If we take $`10^{24}`$ electrons as roughly the number in a cc. of condensed matter, this corresponds to an energy increase of about $`10^{15}`$ joules/sec, which is close to the experimentally detectable lower bound by present day bolometric measurements.<sup>17</sup> This is much less sensitive than the experiment discussed here.
However, while (3.1) gives the average behavior, there are infrequent but large energy fluctuations.
The energy increase in Eq.(3.1) is the sum of increased internal energy and of increased center of mass energy $`H_{cm}𝐏_{cm}^2/2M+V_{cm}(𝐐`$) ($`𝐏_{cm}_{j=1}^N𝐩_j`$, $`M_{j=1}^NM_j`$, $`𝐐_{j=1}^NM_j𝐱_j/M`$). From Eq.(2.3) we find
$$\begin{array}{cc}\hfill \frac{d\overline{H}_{cm}(t)}{dt}& =\frac{3\lambda \mathrm{}^2}{4a^2M}\underset{j=1}{\overset{N}{}}\underset{k=1}{\overset{N}{}}g_{\alpha (j)}g_{\alpha (k)}\text{Tr}\left\{\left[1\frac{(𝐱_j𝐱_k)^2}{6a^2}\right]e^{\frac{(𝐱_j𝐱_k)^2}{4a^2}}\rho (t)\right\}\hfill \\ & =\frac{3\lambda \mathrm{}^2}{4a^2M}\left[\underset{j=1}{\overset{N}{}}g_{\alpha (j)}\right]^2o(a^4)\hfill \end{array}$$
$`(3.1a,b)`$
where the first term in the expansion (3.1b) dominates if the system under consideration, like an atom or nucleus, has dimensions $`<<a`$. The condition for the total energy increase to be completely due to the center of mass energy increase to order $`a^2`$, i.e., for there to be no internal excitation to this order, is found by equating (3.1) to (3.1b):
$$0=\underset{j=1}{\overset{N}{}}g_{\alpha (j)}^2/M_j\left[\underset{j=1}{\overset{N}{}}g_{\alpha (j)}\right]^2/M=(2M)^1\underset{j=1}{\overset{N}{}}\underset{k=1}{\overset{N}{}}\left[g_{\alpha (j)}\sqrt{M_k/M_j}g_{\alpha (k)}\sqrt{M_j/M_k}\right]^2$$
i.e., if $`g_{\alpha (j)}=CM_j`$. Therefore, for such mass–proportionality of the coupling constants, the internal energy does not increase to order $`a^2`$. Moreover, the leading term in the internal energy increase, proportional to $`a^4`$, is compensated by an identical decrease in the center of mass energy since, by (3.1), the total energy increase vanishes to order $`a^4`$ and higher.
If the coupling constants are not mass–proportional, the internal excitation rate $`a^2`$ is found as follows. Consider a transition from a state $`|\psi >|\chi >`$ to a state $`|\varphi >|\chi ^{}>`$, where $`|\psi >`$ is an initial bound state, $`|\varphi >`$ is an orthogonal final state, $`|\chi >`$ is an initial state of the center of mass and $`|\chi ^{}>`$ is an arbitrary final state of the center of mass. The probability per second of a transition from $`|\psi >`$ to $`|\varphi >`$, regardless of the final center of mass state is $`\dot{P}_{|\chi ^{}>}<\chi ^{}|<\varphi |\dot{\rho }(0)|\varphi >|\chi ^{}>`$, where $`\rho (0)=|\chi >|\psi ><\psi |<\chi |`$. Expansion of Eq. (2.3) to first order in (dimension of system/$`a`$)<sup>2</sup> yields<sup>11</sup>
$$\dot{P}_1=\frac{\lambda }{2a^2}<\varphi |𝐑|\psi ><\psi |𝐑|\varphi >$$
$`(3.2)`$
where $`𝐑_{j=1}^Ng_{\alpha (j)}𝐑_j`$, $`𝐑_j𝐱_j𝐐`$.
It is the predictions of Eq.(3.2) that we shall test in this paper. It is worth remarking that the matrix element in Eq.(3.2) involving a charged particle’s $`𝐑_j`$ is the same as that involved in describing an electric dipole transition between $`|\psi >`$ and $`|\varphi >`$. Thus one can evaluate this contribution to (3.2) either by calculating the relevant matrix element or by expressing it in terms of measurable transition rates. Indeed, as we shall see in sections 5 and 6, using $`𝐑0`$ if $`g_\alpha M_\alpha `$, it is possible to express the matrix elements of one type of particles in terms of another type so one need only calculate or measure the matrix elements of the excited particle type to evaluate (3.2).
Incidentally, by summing Eq.(3.2) over all states $`|\varphi >`$ orthogonal to $`|\psi >`$, we obtain $`\dot{P}_1^T`$, the total probability/sec for excitation of $`|\psi >`$:
$$\dot{P}_1^T=\frac{\lambda }{2a^2}<\psi |[𝐑<\psi |𝐑|\psi >]^2|\psi >$$
$`(3.3)`$
Assume the GRW values for $`\lambda `$ and $`a`$. For an atomic electron undergoing spontaneous excitation from, e.g., the 1s state of an atom with atomic number $`Z`$ to a higher energy state, bound or free, the order of magnitude of $`\dot{P}_1`$ is $`g_e^210^{23}/Z^2`$sec<sup>-1</sup>. For, e.g., a proton in an outer shell of a nucleus of mass number $`A`$ it is $`g_p^210^{32}A^{2/3}`$sec<sup>-1</sup>. With such rates, the 1s electrons in $`10^{24}`$ such atoms would be expected to provide $`g_e^210/Z^2`$ photon pulses each second while each spontaneously excited proton in $`10^{24}`$ such nucleii would be expected to provide $`g_p^2.3A^{2/3}`$ gammas each year. This large difference in rates explains why we are able to obtain good experimental limits on the electron’s coupling constant $`g_e`$ in section 6, but not on the neutron’s coupling constant $`g_n`$ in section 5.
Although we shall only apply Eq. (3.2) in our data analysis, for completeness we include Eqs. (3.4), (3.5) below which could be appled if more accurate data becomes available. We note again that if $`g_\alpha `$ is mass–proportional (i.e., if for protons $`g_p=1`$, then for neutrons $`g_n1.001`$ and for electrons $`g_e.00054`$), then $`𝐑0`$, and therefore (3.2) vanishes. We should then need $`\dot{P}`$ to order $`a^4`$:
$$\dot{P}_2=\frac{\lambda }{16a^4}[|<\varphi |S|\psi >|^2+2\underset{m=1}{\overset{3}{}}\underset{n=1}{\overset{3}{}}|<\varphi |S^{mn}|\psi >|^2]$$
$`(3.4)`$
where $`S^{mn}_{j=1}^Ng_{\alpha (j)}(𝐑_j)^m(𝐑_j)^n`$, $`S_{n=1}^3S^{nn}`$. If $`𝐑_j`$ corresponds to a charged particle, its matrix elements here describe electric monopole or quadrupole transitions. These matrix elements are smaller than the corresponding matrix elements in Eq.(3.2) by the factor (size of bound state/$`a`$)<sup>2</sup>.
Fu<sup>10</sup> has considered the spontaneous electromagnetic radiation of a free charged particle in CSL, obtaining the probability/sec/energy of radiating a photon of energy $`E=\mathrm{}k`$:
$$\frac{d\dot{P}(E)}{dE}=g_\alpha ^2\frac{\lambda }{4\pi ^2}\frac{e^2}{\mathrm{}c}\left(\frac{\mathrm{}/M_\alpha c}{a}\right)^2\frac{1}{E}\frac{310^{31}}{E\text{ in keV}}\text{ counts/sec/keV}$$
$`(3.5a,b)`$
(the infrared divergence is treated as usual). Eq.(3.5b) gives the rate for an electron with $`g_e=1`$. This radiation rate (3.5) for free particles is smaller than the excitation rate (3.2) for bound particles by the factor $`e^2/\mathrm{}c1/137`$ and by the replacement of (bound state size/$`a`$)<sup>2</sup> by (Compton wavelength/$`a`$)<sup>2</sup>. We note, with mass-proportionality, that the spontaneous radiation rate for free electrons is the same as for free protons, on account of the factor $`(g_\alpha /M_\alpha )^2`$ in (3.5a).
This completes our collection of equations giving CSL spontaneous excitation rates.
We now consider the spontaneous excitation of valence nucleons in a Germanium nucleus and the spontaneous ionization of 1s electrons in a Germanium atom, using data from what are, at present, the lowest noise relevant experiments.
4. Experiments.
The data used in the analysis<sup>18</sup> of the atomic excitation comes from a small (253 g) p-type coaxial HPGe crystal, “COSME”, built specifically for a Dark Matter search<sup>19</sup>. It features, at only 1.6 keV, the (so-far) lowest energy threshold of any detector dedicated to such searches, and has as well an excellent resolution of 0.43 keV (FWHM) at 10.3 keV. Special low-radioactivity measures were taken, such as mounting the detector on a specially-designed electroformed copper cryostat, shielding of electronic components close to the crystal with 450-yr-old lead, and use of a 2000-yr-old roman lead layer in the innermost part of the shielding. Additional photon, neutron and vibrational shielding were used.
The detector set-up was installed in the Canfranc-1 underground laboratory in the Spanish Pyrenees, at a depth of 675 meters of water equivalent. The microphonic component characteristic of very low-threshold detectors, extending up to $`15`$keV, was filtered-out using specially-developed techniques <sup>20</sup>. While the low-energy background level was slightly higher than that in the “TWIN’ detectors <sup>13</sup>, the improved resolution —typically inversely proportional to the mass of the crystal— allows one to impose more stringent limits on a sharply defined signal that might be buried in an otherwise featureless background, as is the case for the emitted radiation in CSL.
At this point a remark is in order: unfortunately the TWIN data used to extract CSL limits on $`g_e`$ in reference 12 belonged to a preliminary set coming from un-amplified digitized pulses. Later comparison with the spectrum collected with a multichannel analyzer showed that this earlier data was corrupted at energies below $`200`$ keV, i.e., a large fraction of events were not recorded. This faulty set was not used in other TWIN results, namely for double-beta decay <sup>13</sup>, Dark Matter <sup>21</sup> or electron half-life <sup>22</sup>. In reference 12, only one data point was used, the (erroneous) rate .049 counts/keV/kg/day at 11 keV (where the highest predicted value of spontaneous radiation occurs) to set an upper limit on $`g_e`$. In the present work, it is necessary to perform a Chi<sup>2</sup> analysis of a fit to the whole predicted shape of the spontaneous radiation, in order to obtain a limit comparable to the (erroneous) limit of reference 12.
The data used in the analysis of the nuclear excitation comes from the “Rico Grande” crystals, which are part of the IGEX ensemble of large enriched germanium detectors<sup>23</sup>, dedicated to searching for neutrinoless double-beta decay. The two detectors from which these data were extracted have a fiducial mass of $``$2 kg each and are enriched to 86% Ge<sub>76</sub> and 14% Ge<sub>74</sub>. As a result, the prevailing source of background in the energy neighborhood of interest here is this two-neutrino double-beta decay from Ge<sub>76</sub> which, however, cuts off at an energy below the region employed in the analysis in section 5. At the time of collection of the present data, the Ricos were operated in the Homestake mine in similar conditions to TWIN.
5. Constraint on $`g_n`$/$`g_p`$?
We now apply Eq. (3.2) to the spontaneous excitation rate of a single proton or neutron in a Ge nucleus.
The Jparity for the ground state of Ge is 0+ for the even-even nuclides Ge<sub>70</sub> (20.6%), Ge<sub>72</sub> (27.4%), Ge<sub>74</sub> (36.7%) and Ge<sub>76</sub> (7.7%)), and it is 9/2+ for Ge<sub>73</sub> (7.7%). Spontaneous excitation is predicted (the matrix elements are nonvanishing) for a transition from the ground state to 1- states in the case of the even-even nuclides and to 11/2-, 9/2- or 7/2- states in Ge<sub>73</sub>. From the point of view of the shell model, among other possibilities, a proton or neutron can be excited from its ground state valence level to a higher energy state.
The return of a proton from the excited state to ground could be direct, via an electric dipole transition. It could also proceed indirectly, through intermediate states or internal conversion and, in the case of a neutron it must proceed indirectly, through magnetic transitions. In any case, the lifetimes are in most cases so short that a photon pulse would rapidly appear at the energy difference of the two states. Therefore, by looking at the data for a signature peak of instrumental resolution width at the expected energy, one may hope to observe the radiation resulting from these spontaneous transitions or at least get an upper limit on their rate.
Not only does the matrix element of $`𝐑_i`$ for the excited particle not vanish, but the matrix element of $`𝐑_i`$ for the other particles also does not vanish due to their dependence on the center of mass operator. We can relate the matrix element of the protons to that of the neutrons by using $`\mathrm{\Sigma }_jM_j𝐑_j0`$, which implies that
$$\mathrm{\Sigma }_{i=1}^{AZ}𝐑_{ni}=(M_p/M_n)\mathrm{\Sigma }_{i=1}^Z𝐑_{pi}$$
$`(5.1a)`$
(we neglect the electron contribution of o$`(M_e/(M_n)`$) so, setting $`M_p/M_n=1`$, the matrix element in (3.2) is
$$<\varphi |g_p\mathrm{\Sigma }_{i=1}^Z𝐑_{pi}+g_n\mathrm{\Sigma }_{i=1}^{AZ}𝐑_{ni}|\psi >=[1g_n]<\varphi |\mathrm{\Sigma }_{i=1}^Z𝐑_{pi}|\psi >$$
$`(5.1b)`$
(remembering $`g_p1`$). Thus, from Eqs. (3.2) and (5.1), the excitation rate $`\mathrm{\Gamma }`$ in sec<sup>-1</sup> of e.g., one of the four valence protons from the ground state $`|\psi >`$ to one of the 12 degenerate excited states $`|\varphi >`$ (any of the four protons may be excited, and the 1- state of this proton plus its subshell partner can have three possible orientations) can be expressed purely in terms of proton matrix elements:
$$\begin{array}{cc}\hfill \mathrm{\Gamma }=& \frac{\lambda }{2a^2}[1g_n]^212|<\varphi |\mathrm{\Sigma }_{i=1}^Z𝐑_{pi}|\psi >|^2\hfill \\ \hfill =& \frac{\lambda }{2a^2}[1g_n]^212\mathrm{\Sigma }_m\frac{4\pi }{3}|<\varphi |\mathrm{\Sigma }_{i=1}^{i=Z}R_{pi}Y_{1,m}(\theta _i,\varphi _i)|\psi >|^2\hfill \end{array}$$
$`(5.2a,b)`$
A similar expression may be written for the excitation rate of a neutron totally in terms of neutron matrix elements.
It is worth remarking that the expression for the lifetime $`\tau (\varphi \psi )`$ of one of the excited state $`|\varphi >`$ to decay by an electric dipole transition to the ground state $`|\psi >`$ can be written in terms of the same matrix elements as appear in (5.2)<sup>24</sup>:
$$\frac{1}{\tau (\varphi \psi )}=\frac{16\pi c}{9}\left(\frac{E}{\mathrm{}c}\right)^3\left(\frac{e^2}{\mathrm{}c}\right)\mathrm{\Sigma }_m|<\varphi |\mathrm{\Sigma }_{i=1}^{i=Z}R_{pi}Y_{1,m}(\theta _i,\varphi _i)|\psi >|^2$$
$`(5.3)`$
where $`E`$ is the energy difference of the two states. Thus we may express $`\mathrm{\Gamma }`$ in terms of this lifetime:
$$\mathrm{\Gamma }=\frac{\lambda }{2a^2}[1g_n]^2\frac{9}{c(e^2/\mathrm{}c)}\left(\frac{\mathrm{}c}{E}\right)^3\frac{1}{\tau (\varphi \psi )}$$
$`(5.4)`$
Eq. (5.4) would be useful if we have the experimental lifetimes and branching ratios of the state $`\varphi `$. Unfortunately, in this case we do not, so we are forced to estimate the matrix element in (5.2b).
In this paper we shall approximate the matrix element by using the same “very rough estimate” employed by Blatt and Weisskopf<sup>25</sup> for calculating the lifetime $`\tau `$. They assume the radial wavefunction of the proton in both states $`\varphi `$ and $`\psi `$ is $`\mathrm{\Theta }(R_0R)[3/R_0^3]^{1/2}`$ where $`\mathrm{\Theta }`$ is the step function and $`R_0=1.4\times 10^{13}A^{1/3}`$ is the nuclear radius. As they point out, the actual radial integral is expected to be “somewhat smaller” since radial wavefunctions oscillate: say, $`\beta `$ times smaller. We obtain from (5.2b) the result
$$\mathrm{\Gamma }=\frac{\lambda }{a^2}[1g_n]^2\beta ^2R_0^2$$
$`(5.5)`$
(a numerical factor 9/8 has been replace by 1).
The expected total count $`𝒞`$ for an experimental run of $`D`$ kg-days from a transition due to a nuclide which comprises the fraction $`X`$ of the $`8.3\times 10^{24}`$ atoms/kg in common Ge is found from (5.5), with the GRW parameters, to be
$$𝒞=\frac{\beta ^2}{4}[1g_n]^2XD$$
$`(5.6)`$
For example, consider a transition in Ge<sub>74</sub> to the 2165 keV 1- state. We shall use the data from the Rico Grande experiment<sup>22</sup>, similar to COSME but with $`D=1.135`$ kg-yrs=414.3 kg-days (COSME’s $`D=85.24`$ kg-days) with $`X=.14`$ (COSME’s $`X=.37`$). Denoting by $`𝒞_{\mathrm{expt}}`$ the upper limit on the number of observed counts, we obtain
$$1+.26\frac{𝒞_{\mathrm{expt}}^{\frac{1}{2}}}{\beta }g_n1.26\frac{𝒞_{\mathrm{expt}}^{\frac{1}{2}}}{\beta }$$
$`(5.7)`$
For this experiment, the upper limit (obtained from the counts under a Chi<sup>2</sup> fit to a background quadratic plus the expected experimental resolution shape centered on 2165 keV) is $`𝒞_{\mathrm{expt}}=.89`$ counts at the 68% confidence level (3.9 counts at the 95% confidence level). For $`\beta .3`$ one obtains $`1.8g_n.2`$. However, our choice of $`\beta `$ is just hypothetical, as we have not made the effort to seriously evaluate the matrix element<sup>26,27</sup>. The points to be made are that the range of $`g_n`$ is not so far from $`g_p=1`$ and that the various numbers involved in calculating (5.7) tantalizingly contrive to be on the edge of showing mass proportionality. Indeed, this would more easily be shown with a larger value of $`\lambda /a^2`$ than the GRW value: for instance, with $`\lambda /a^2=100\lambda /a_{\mathrm{GRW}}^2`$, the above inequality becomes $`1.1g_n.9`$. But, with the GRW parameters, it would require a long counting time and a proper calculation of matrix elements before one might say that $`g_n/g_p1`$.
6. Constraint on $`g_e`$/$`g_p`$.
Here we apply Eq. (3.2) to calculate the spontaneous ionization rate of the 1s electrons in a Ge atom.
If a 1s electron is spontaneously ionized, the remaining electrons in the atom rapidly cascade into the (singly ionized) ground state, emitting a photon pulse of 11.1 keV (the ionization energy of a 1s electron). The ionized electron also deposits its kinetic energy in the Ge sample, which augments the energy of the pulse. Thus the signature of these events is a distribution of photon pulses of energy $`E>11.1`$ keV.
In reference 12, a Hartree calculation of the matrix element in (3.2) for the electrons was numerically performed, where $`|\psi >`$ is the ground state of Ge and $`|\varphi >`$ is a state where a 1s electron is ionized. The result of the calculation may be expressed as a function $`C(E)`$ which gives the expected pulse counting rate if GRW parameters are assumed and if the electron is totally responsible for collapse (i.e., $`g_e=1`$, $`g_n=g_p=0`$). $`C(E)`$ is zero for $`E<11.1`$ keV, abruptly rises to 5370 counts/keV/kg/day at $`E=11.1`$ keV, and decays in roughly exponential fashion, with the value $`4000`$ counts/keV/kg/day at $`E=12`$ keV, and $`1500`$ counts/keV/kg/day at $`E=16`$ keV.
If we assume that $`g_n=g_p=1`$ (as we shall hereafter do) then, as in the preceding section’s Eq. (5.1), we can express the matrix element for the nucleons as $`M_e/M_p`$ times the matrix element for the electrons. Putting this into Eq. (3.2), the resulting rate $`\mathrm{\Gamma }`$ in counts/sec/kg/day may then be written as
$$\mathrm{\Gamma }=\frac{(\lambda /a^2)}{(\lambda /a^2)_{GRW}}[g_e\frac{M_e}{M_p}]^2C(E)$$
$`(6.1)`$
We note that the rate in (6.1) vanishes if there is mass proportionality ($`g_e/g_p=M_e/M_p`$).
Figure 1 shows a graph of the counts/.1keV/85.24kg-day from COSME in the energy range 5 to 17 keV. A recent paper<sup>21</sup> describes a search with the same apparatus in a similar energy range. These authors were looking at the TWIN data for a signature 11.1 keV peak resulting from a hypothesized violation of charge conservation (in which a K-shell electron decays to neutrals, and the other electrons in the atom readjust). They fit the data to three x-ray peaks (Cu, Zn and Ga) known to result from cosmogenic excitation of the Ge isotopes in the sample under observation, together with a background quadratic polynomial plus the hypothesized process, and they look at the difference between the fit and the data at the 68% and 90% confidence levels.
We employ here the same procedure. Fig. 1 (solid line) shows the best fit to the data by this method, without the hypothesized process, a multiple of $`C(E)`$. Superimposed upon this fit is a graph of $`10^3C(E)`$ (dash-dotted line), folded in with the detector resolution shape (a gaussian of standard deviation .18 keV). It is clear that this is by no means a good fit to the data, so the coefficient of $`C(E)`$ in Eq. (6.1) is considerably smaller than $`10^3`$.
Also shown in Fig. 1 is the fit with the hypothesized process at $`4.2\times 10^5C(E)`$ (dotted line) which corresponds to the 68% confidence level. Not shown is a similar shaped curve at $`9.7\times 10^5C(E)`$ which corresponds to the 95% confidence level.
If we assume that the parameters $`\lambda `$ and $`a`$ are the same as those given by GRW, then we conclude from Eq.(6.1) that, at the 68% confidence level, $`[g_e/g_pM_e/M_p]^2C(E)4.2\times 10^5C(E)`$ or
$$0\frac{g_e}{g_p}13\frac{M_e}{M_p}$$
$`(6.2)`$
Thus, according to CSL with the GRW parameters, nucleons are mostly responsible for collapse.
It is worth noting that it would take an improvement in the experimental limit by e.g., a factor of 1/300, which results in $`.3M_e/M_pg_e/g_p1.7M_e/M_p`$, to suggest that $`g_e/g_p=0`$ may be ruled out. However, it should also be noted that we need not be wedded to the GRW parameters. Thus an increase of $`\lambda /a^2`$ by a factor of 300 or more would have the same effect. On the other hand it would take a decrease in $`\lambda /a^2`$ by a factor of $`4\times 10^5`$ or more for the limit obtained in this experiment not to suggest that nucleons collapse more rapidly than electrons.
Acknowledgments
We would like to thank Brian Collett for computational help. One of us (P.P.) would like to thank the Institute for Advanced Studies of the Hebrew University in Jerusalem, where some of this work was done, for its support and hospitality, and Yakir Aharonov for setting up the workshop at the Hebrew University.
References
1. D. Bohm and J. Bub, Revs. Mod. Phys. 38, 453 (1966).
2. P. Pearle, Phys. Rev. D13, 857 (1976). For more on this pre-GRW work, see P. Pearle in Sixty-Two Years of Uncertainty, A. Miller ed. (Plenum, New York 1990), p. 193.
3. G. C. Ghirardi, A. Rimini and T. Weber, Physical Review D34, 470 (1986); Physical Review D36, 3287 (1987); Foundations of Physics 18, 1 (1988).
4. P. Pearle, Physical Review A39, 2277 (1989); G. C. Ghirardi, P. Pearle and A. Rimini, Physical Review A42, 78 (1990). For reviews see G. C. Ghirardi and A. Rimini in Sixty-Two Years of Uncertainty, edited by A. Miller (Plenum, New York 1990); G. C. Ghirardi and P. Pearle in Proceedings of the Philosophy of Science Foundation 1990, Volume 2, edited by A. Fine, M. Forbes and L. Wessels (PSA Association, Michigan 1992), p. 19 and p. 35.
5. F. Karolyhazy, Nuovo Cimento 42A, 1506 (1966); P. Pearle and E. Squires, Found. Phys. 26, 291 (1996).
6. P. Pearle, Relativistic Collapse Model with Tachyonic Features (Hamilton College preprint, 1997).
7. J. R. Clauser (private communication).
8. A. Zeilinger (private communication).
9. E. J. Squires, Phys. Lett. A158, 432 (1991).
10. Q. Fu, Phys. Rev.A56, 1806 (1997).
11. P. Pearle and E. J. Squires, Phys. Rev. Lett. 73, 1 (1994).
12. B. Collett, P. Pearle, F. Avignone and S. Nussinov, Found. Phys. 25, 1399 (1995).
13. H. S. Miley, F. T. Avignone III, R. L. Brodzinski, J. I. Collar and J. H. Reeves, Phys. Rev. Lett. 65, 3092 (1990).
14. R. Penrose in Quantum Concepts in Space and Time, edited by R. Penrose and C. J. Isham (Clarendon, Oxford 1986), p. 129; in The Emperor’s New Mind, (Oxford University Press, Oxford, 1992); in Shadows of the Mind, (Oxford University Press, Oxford, 1994).
15. L. Diosi, Phys. Rev. A40, 1165 (1989); G. C. Ghirardi, R. Grassi and A. Rimini, Phys. Rev. A42, 1057 (1990); ref. 6.
16. P. Pearle, Foundations of Physics 12, 249 (1982)
17. We would like to thank Leo Stodolsky for bringing such experiments to our attention, and for providing information about the experimental accuracy.
18. J. I. Collar, PhD diss. U of SC 1992.
19. E. Garcia et al Phys. Rev. D 51, 1458 (1995).
20. J. Morales et al, Nucl. Instr. Meth. A 321, 410 (1992).
21. A.K. Drukier et al, Nucl. Phys. B (Proc. Suppl.) 28A, 293 (1992).
22. Y. Aharonov et al Phys. Rev. D 52, 3785 (1995).
23. C. E. Aalseth et. al., Nuclear Phys. B (Proc. Suppl.) 48, 223 (1996).
24. J. M. Blatt and V. F. Weisskopf, Theoretical Nuclear Physics, (Wiley, New York 1960), p. 595.
25. Ibid, p.625 et.seq.
26. R. D. Lawson, Theory Of The Nuclear Shell Model, (Clarendon, Oxford 1980).
27. J. Joubert, F. J. W. Hahne, P. Navratil and H. B. Geyer, Phys. Rev. C 50, 177 (1994) and references therein.
Figure Captions
Figure. 1 A graph of the COSME data in the region 5-17 keV is shown along with the best fit to the three known X-ray peaks plus a quadratic polynomial background (solid curve). Two additional curves are shown, corresponding to predicted rates folded in with the experimental resolution (a gaussian of width .18keV). The dash-dotted curve corresponds to $`10^3C(E)`$ and the dotted curve to the 68% confidence level value of $`4.2\times 10^5C(E)`$. |
warning/0001/hep-th0001027.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Superstring theory is a leading candidate for the unification of fundamental interactions. As such it is expected to reconcile quantum mechanics and general relativity. As is well known, field theoretic approach to quantize gravity encounters serious difficulties at short distances. String theory contains the notion of the minimal length (string scale) which is expected to cure such difficulties.
Although the first quantization of string theory is well understood, it is ripe to obtain fully nonperturbative formulation of superstring theory. We have proposed such a formalism as a large $`N`$ reduced model with the maximal chiral SUSY . It is a finite theory which has a potential to predict the structure of space-time.
Another school of thought which also introduces the minimum length scale advocates to replace Riemannian geometry by noncommutative geometry. Furthermore string theory in noncommutative tori has been studied in .
Remarkably, these various schools of thought have converged recently. We have pointed out that noncommutative Yang-Mills theory is naturally obtained in IIB matrix model by expanding the theory around a noncommutative space-time. We have further studied the Wilson loops. These investigations have shown that the theory contains not only point like field theoretic degrees of freedom but also open string like extended objects . In terms of the matrix representation, the former is close to diagonal and the latter is far off-diagonal degrees of freedom. In string theory these backgrounds are interpreted as $`D`$-branes with constant $`b_{\mu \nu }`$ field background. These problems are further studied in .
In this paper, we continue our investigation of noncommutative field theory as twisted reduced models. We have mapped the twisted reduced model onto the noncommutative field theory by expanding the matrices by the momentum eigenstates. In this paper we propose a more natural decomposition rule for the matrices in terms of a bi-local basis. We show that they corresponds to ‘open strings’ and we can reproduce quenched reduced models with long ‘open strings’. We also show that the diagonal elements represent ordinary plane waves. Noncommutative field theory exhibits crossover at the noncommutative scale. At large momentum scale, it is identical to large $`N`$ field theory and the relevant degrees of freedom are described by ‘open strings’ longer than the noncommutative scale while it reduces to field theory in the opposite limit. We make these statements more precise in this paper. The equivalence to large $`N`$ field theory at large momentum scale is further supported since we find quenched reduced models in noncommutative field theory. In the low energy limit, we find long range interactions in noncommutative field theory which are absent in the ordinary field theory. We point out that this effect can be simply understood in terms of block-block interactions in the matrix model picture.
## 2 Noncommutative field theories as twisted reduced models
In this section, we briefly recapitulate our approach to noncommutative field theory as twisted reduced models. Reduced models are defined by the dimensional reduction of $`d`$ dimensional gauge theory down to zero dimension (a point). The application of reduced models to string theory is pioneered in . We consider $`d`$ dimensional $`U(n)`$ gauge theory coupled to adjoint matter as an example:
$$S=d^dx\frac{1}{g^2}Tr(\frac{1}{4}[D_\mu ,D_\nu ][D_\mu ,D_\nu ]+\frac{1}{2}\overline{\psi }\mathrm{\Gamma }_\mu [D_\mu ,\psi ]),$$
(2.1)
where $`\psi `$ is a Majorana spinor field. The corresponding reduced model is
$$S=\frac{1}{g^2}Tr(\frac{1}{4}[A_\mu ,A_\nu ][A_\mu ,A_\nu ]+\frac{1}{2}\overline{\psi }\mathrm{\Gamma }_\mu [A_\mu ,\psi ]).$$
(2.2)
Now $`A_\mu `$ and $`\psi `$ are $`n\times n`$ Hermitian matrices and each component of $`\psi `$ is $`d`$-dimensional Majorana-spinor. IIB matrix model is obtained when $`d=10`$ and $`\psi `$ is further assumed to be Weyl-spinor as well. Any noncommutative field theory is realized as well in terms of a twisted reduced model by the same mapping rules as below but here we explain the case of the above reduced models of gauge theory. We expand the theory around the following classical solution:
$$[\widehat{p}_\mu ,\widehat{p}_\nu ]=iB_{\mu \nu },$$
(2.3)
where $`B_{\mu \nu }`$ are $`c`$-numbers. We assume the rank of $`B_{\mu \nu }`$ to be $`\stackrel{~}{d}`$ and define its inverse $`C^{\mu \nu }`$ in $`\stackrel{~}{d}`$ dimensional subspace. The directions orthogonal to the subspace is called the transverse directions. $`\widehat{p}_\mu `$ satisfy the canonical commutation relations and they span the $`\stackrel{~}{d}`$ dimensional phase space. The semiclassical correspondence shows that the volume of the phase space is $`V_p=n(2\pi )^{\stackrel{~}{d}/2}\sqrt{detB}`$.
We expand $`A_\mu =\widehat{p}_\mu +\widehat{a}_\mu `$. We Fourier decompose $`\widehat{a}_\mu `$ and $`\widehat{\psi }`$ fields as
$`\widehat{a}`$ $`=`$ $`{\displaystyle \underset{k}{}}\stackrel{~}{a}(k)exp(iC^{\mu \nu }k_\mu \widehat{p}_\nu ),`$
$`\widehat{\psi }`$ $`=`$ $`{\displaystyle \underset{k}{}}\stackrel{~}{\psi }(k)exp(iC^{\mu \nu }k_\mu \widehat{p}_\nu ),`$ (2.4)
where $`exp(iC^{\mu \nu }k_\mu \widehat{p}_\nu )`$ is the eigenstate of adjoint $`P_\mu =[\widehat{p}_\mu ,]`$ with the eigenvalue $`k_\mu `$. The Hermiticity requires that $`\stackrel{~}{a}^{}(k)=\stackrel{~}{a}(k)`$ and $`\stackrel{~}{\psi }^{}(k)=\stackrel{~}{\psi }(k)`$. Let us consider the case that $`\widehat{p}_\mu `$ consist of $`\stackrel{~}{d}/2`$ canonical pairs $`\widehat{p}_i,\widehat{q}_i`$ which satisfy $`[\widehat{p}_i,\widehat{q}_j]=iB\delta _{ij}`$. We also assume that the solutions possess the discrete symmetry which exchanges canonical pairs and $`\widehat{p}_i\widehat{q}_i`$ in each canonical pair. We then find $`V_p=\mathrm{\Lambda }^{\stackrel{~}{d}}`$ where $`\mathrm{\Lambda }`$ is the extension of each $`\widehat{p}_\mu `$. The volume of the unit quantum in phase space is $`\mathrm{\Lambda }^{\stackrel{~}{d}}/n=\lambda ^{\stackrel{~}{d}}`$ where $`\lambda `$ is the spacing of the quanta. $`B`$ is related to $`\lambda `$ as $`B=\lambda ^2/(2\pi )`$. Let us assume the topology of the world sheet to be $`T^{\stackrel{~}{d}}`$ in order to determine the distributions of $`k_\mu `$. Then we can formally construct $`\widehat{p}_\mu `$ through unitary matrices as $`\gamma _\mu =exp(i2\pi \widehat{p}_\mu /\mathrm{\Lambda })`$. The polynomials of $`\gamma _\mu `$ are the basis of $`exp(iC^{\mu \nu }k_\mu \widehat{p}_\nu )`$. We can therefore assume that $`k_\mu `$ is quantized in the unit of $`|k^{min}|=\lambda /n^{1/\stackrel{~}{d}}`$. The eigenvalues of $`\widehat{p}_\mu `$ are quantized in the unit of $`\mathrm{\Lambda }/n^{2/\stackrel{~}{d}}=\lambda /n^{1/\stackrel{~}{d}}`$. Hence we restrict the range of $`k_\mu `$ as $`n^{1/\stackrel{~}{d}}\lambda /2<k_\mu <n^{1/\stackrel{~}{d}}\lambda /2`$. So $`_k`$ runs over $`n^2`$ degrees of freedom which coincide with those of $`n`$ dimensional Hermitian matrices.
We can construct a map from a matrix to a function as
$$\widehat{a}a(x)=\underset{k}{}\stackrel{~}{a}(k)exp(ikx),$$
(2.5)
where $`kx=k_\mu x^\mu `$. By this construction, we obtain the $``$ product
$`\widehat{a}\widehat{b}`$ $``$ $`a(x)b(x),`$
$`a(x)b(x)`$ $``$ $`exp({\displaystyle \frac{C^{\mu \nu }}{2i}}{\displaystyle \frac{^2}{\xi ^\mu \eta ^\nu }})a(x+\xi )b(x+\eta )|_{\xi =\eta =0}.`$ (2.6)
The operation $`Tr`$ over matrices can be exactly mapped onto the integration over functions as
$$Tr[\widehat{a}]=\sqrt{detB}(\frac{1}{2\pi })^{\frac{\stackrel{~}{d}}{2}}d^{\stackrel{~}{d}}xa(x).$$
(2.7)
The twisted reduced model can be shown to be equivalent to noncommutative Yang-Mills by the the following map from matrices onto functions
$`\widehat{a}`$ $``$ $`a(x),`$
$`\widehat{a}\widehat{b}`$ $``$ $`a(x)b(x),`$
$`Tr`$ $``$ $`\sqrt{detB}({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}x}.`$ (2.8)
The following commutator is mapped to the covariant derivative:
$$[\widehat{p}_\mu +\widehat{a}_\mu ,\widehat{o}]\frac{1}{i}_\mu o(x)+a_\mu (x)o(x)o(x)a_\mu (x)[D_\mu ,o(x)],$$
(2.9)
We may interpret the newly emerged coordinate space as the semiclassical limit of $`\widehat{x}^\mu =C^{\mu \nu }\widehat{p}_\nu `$. The space-time translation is realized by the following unitary operator:
$$exp(i\widehat{p}d)\widehat{x}^\mu exp(i\widehat{p}d)=\widehat{x}^\mu +d^\mu .$$
Applying the rule eq.(2.8), the bosonic action becomes
$`{\displaystyle \frac{1}{4g^2}}Tr[A_\mu ,A_\nu ][A_\mu ,A_\nu ]`$ (2.10)
$`=`$ $`{\displaystyle \frac{\stackrel{~}{d}nB^2}{4g^2}}\sqrt{detB}({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle }d^{\stackrel{~}{d}}x{\displaystyle \frac{1}{g^2}}({\displaystyle \frac{1}{4}}[D_\alpha ,D_\beta ][D_\alpha ,D_\beta ]`$
$`+{\displaystyle \frac{1}{2}}[D_\alpha ,\phi _\nu ][D_\alpha ,\phi _\nu ]+{\displaystyle \frac{1}{4}}[\phi _\nu ,\phi _\rho ][\phi _\nu ,\phi _\rho ])_{}.`$
In this expression, the indices $`\alpha ,\beta `$ run over $`\stackrel{~}{d}`$ dimensional world volume directions and $`\nu ,\rho `$ over the transverse directions. We have replaced $`a_\nu \phi _\nu `$ in the transverse directions. Inside $`()_{}`$, the products should be understood as $``$ products and hence commutators do not vanish. The fermionic action becomes
$`{\displaystyle \frac{1}{g^2}}Tr\overline{\psi }\mathrm{\Gamma }_\mu [A_\mu ,\psi ]`$ (2.11)
$`=`$ $`\sqrt{detB}({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}x\frac{1}{g^2}(\overline{\psi }\mathrm{\Gamma }_\alpha [D_\alpha ,\psi ]+\overline{\psi }\mathrm{\Gamma }_\nu [\phi _\nu ,\psi ])_{}}.`$
We therefore find noncommutative U(1) gauge theory.
In order to obtain noncommutative Yang-Mills theory with $`U(m)`$ gauge group, we consider new classical solutions which are obtained by replacing each element of $`\widehat{p}_\mu `$ by the $`m\times m`$ unit matrix:
$$\widehat{p}_\mu \widehat{p}_\mu \mathrm{𝟏}_m.$$
(2.12)
We require $`N=mn`$ dimensional matrices for this construction. The fluctuations around this background $`\widehat{a}`$ and $`\widehat{\psi }`$ can be Fourier decomposed in the analogous way as in eq.(2.4) with $`m`$ dimensional matrices $`\stackrel{~}{a}(k)`$ and $`\stackrel{~}{\psi }(k)`$ which satisfy $`\stackrel{~}{a}(k)=\stackrel{~}{a}^{}(k)`$ and $`\stackrel{~}{\psi }(k)=\stackrel{~}{\psi }^{}(k)`$. It is then clear that $`[\widehat{p}_\mu +\widehat{a}_\mu ,\widehat{o}]`$ can be mapped onto the nonabelian covariant derivative $`[D_\mu ,o(x)]`$ once we use $``$ product. Applying our rule (2.8) to the action in this case, we obtain
$`{\displaystyle \frac{\stackrel{~}{d}NB^2}{4g^2}}\sqrt{detB}({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle }d^{\stackrel{~}{d}}x{\displaystyle \frac{1}{g^2}}tr({\displaystyle \frac{1}{4}}[D_\alpha ,D_\beta ][D_\alpha ,D_\beta ]`$
$`+{\displaystyle \frac{1}{2}}[D_\alpha ,\phi _\nu ][D_\alpha ,\phi _\nu ]+{\displaystyle \frac{1}{4}}[\phi _\nu ,\phi _\rho ][\phi _\nu ,\phi _\rho ]`$
$`+{\displaystyle \frac{1}{2}}\overline{\psi }\mathrm{\Gamma }_\alpha [D_\alpha ,\psi ]+{\displaystyle \frac{1}{2}}\overline{\psi }\mathrm{\Gamma }_\nu [\phi _\nu ,\psi ])_{}.`$ (2.13)
where $`tr`$ denotes taking trace over $`m`$ dimensional subspace. The Yang-Mills coupling is found to be $`g_{NC}^2=(2\pi )^{\frac{\stackrel{~}{d}}{2}}g^2/B^{\stackrel{~}{d}/2}`$. Therefore it will decrease if the density of quanta in phase space decreases with fixed $`g^2`$.
The Hermitian models are invariant under the unitary transformation: $`A_\mu UA_\mu U^{},\psi U\psi U^{}`$. As we shall see, the gauge symmetry can be embedded in the $`U(N)`$ symmetry. We expand $`U=exp(i\widehat{\lambda })`$ and parameterize
$$\widehat{\lambda }=\underset{k}{}\stackrel{~}{\lambda }(k)exp(ik\widehat{x}).$$
(2.14)
Under the infinitesimal gauge transformation, we find the fluctuations around the fixed background transform as
$`\widehat{a}_\mu `$ $``$ $`\widehat{a}_\mu +i[\widehat{p}_\mu ,\widehat{\lambda }]i[\widehat{a}_\mu ,\widehat{\lambda }],`$
$`\widehat{\psi }`$ $``$ $`\widehat{\psi }i[\widehat{\psi },\widehat{\lambda }].`$ (2.15)
We can map these transformations onto the gauge transformation in noncommutative Yang-Mills by our rule eq.(2.8):
$`a_\alpha (x)a_\alpha (x)+{\displaystyle \frac{}{x^\alpha }}\lambda (x)ia_\alpha (x)\lambda (x)+i\lambda (x)a_\alpha (x),`$
$`\phi _\nu (x)\phi _\nu (x)i\phi _\nu (x)\lambda (x)+i\lambda (x)\phi _\nu (x),`$
$`\psi (x)\psi (x)i\psi (x)\lambda (x)+i\lambda (x)\psi (x).`$ (2.16)
## 3 Bi-local field representations
We have constructed our mapping rule eq.(2.8) by expanding matrices in terms of the wave functions $`exp(ik\widehat{x})`$. It is the eigenstate of the adjoint $`\widehat{P}_\mu `$ with the eigenvalue of $`k_\mu `$. With the $`n\times n`$ matrix regularization, the momenta $`k_\mu `$ are quantized with a unit $`k_{min}=\lambda n^{1/\stackrel{~}{d}}`$ and can take values $`|k|<\lambda n^{1/\stackrel{~}{d}}/2`$. These momentum eigenstates correspond to the ordinary plane waves when $`|k_\mu |<\lambda `$. In noncommutative space-time, it is not possible to consider states which are localized in the domain whose volume is smaller than the noncommutative scale. Therefore if we consider the states with larger momentum or smaller longitudinal length scale than the noncommutative scale, they must expand in the transverse directions. Recall that the both momentum space and coordinate space are embedded in the matrices of twisted reduced models. They are related by $`\widehat{x}^\mu =C^{\mu \nu }\widehat{p}_\nu `$. The corresponding eigenstates such as $`exp(ik^1\widehat{x})`$ and $`exp(ik^2\widehat{x})`$ are not commutative to each other if $`|k_\mu ^i|>\lambda `$, since $`exp(ik^1\widehat{x})exp(ik^2\widehat{x})=exp(ik^2\widehat{x})exp(ik^1\widehat{x})exp(iC^{\mu \nu }k_\mu ^1k_\nu ^2)`$. The momentum eigenstate $`exp(ik\widehat{x})`$ is also written as $`exp(id\widehat{p})`$ where $`d^\mu =C^{\mu \nu }k_\nu `$, and this implies that the eigenstate with $`|k_\mu |>\lambda `$ is more appropriately interpreted as noncommutative translation operators (2) rather than the ordinary plane waves. In other words, the eigenstate $`exp(ik\widehat{x})`$ with $`|k_\mu |>\lambda `$ may be interpreted as string like extended objects whose length is $`|C^{\mu \nu }k_\nu |`$.
### 3.1 Operator - bi-local field mapping
In order to make these statements more transparent, we consider another representation of matrices in this section. For simplicity we consider the two dimensional case first:
$$[\widehat{x},\widehat{y}]=iC$$
(3.1)
where $`C`$ is positive. This commutation relation is realized by the guiding center coordinates of the two dimensional system of electrons in magnetic field. The generalizations to arbitrary even $`\stackrel{~}{d}`$ dimensions are straightforward. We recall that we have $`n`$ quanta with $`n`$ dimensional matrices. Each quantum occupies the space-time volume of $`2\pi C`$. We may consider a square von Neumann lattice with the lattice spacing $`l_{NC}`$ where $`l_{NC}^2=2\pi C`$. This spacing $`l_{NC}`$ gives the noncommutative scale. Let us denote the most localized state centered at the origin by $`|0`$. It is annihilated by the operator $`\widehat{x}^{}=\widehat{x}i\widehat{y}`$. We construct states localized around each lattice site by utilizing translation operators $`|x_i=exp(ix_i\widehat{p})|0`$. They are the coherent states on a von Neumann lattice $`𝐱_i=l_{NC}(n_i𝐞^x+m_i𝐞^y)`$ where $`n,m𝐙.`$ They are complete but non-orthogonal. In the case of higher dimensions $`\stackrel{~}{d}`$, we introduce
$$\widehat{x}^{\pm a}=\widehat{x}^{2a1}\pm i\widehat{x}^{2a}$$
(3.2)
where $`a=1,\mathrm{},\stackrel{~}{d}/2`$ and define the states $`|𝐱_i`$ accordingly. In the following, we set $`C^{2a1,2a}=C`$ for simplicity.
The basic identity in this section is
$$0|exp(ik\widehat{x})|0=exp(\frac{Ck^2}{4}).$$
(3.3)
Using this identity, we indeed find
$$\rho _{ij}x_i|x_j=exp(\frac{i}{2}B_{\mu \nu }x_i^\mu x_j^\nu )exp(\frac{(x_ix_j)^2}{4C}).$$
(3.4)
Although $`|x_i`$ are non-orthogonal, $`x_i|x_j`$ exponentially vanishes when $`(x_ix_j)^2`$ gets large. We note that the following wave functions of the two dimensional system of free electrons in the lowest Landau level are exponentially localized around $`x_i`$:
$$c_{n_im_i}(x)=\frac{1}{l_{NC}}<x|x_i>.$$
(3.5)
Completeness of the basis leads to the resolution of unity
$$1=\underset{i,j}{}(\rho ^1)_{ij}|x_ix_j|$$
(3.6)
and the trace of an operator $`\widehat{𝒪}`$ is given by
$$Tr\widehat{𝒪}=\underset{i,j}{}(\rho ^1)_{ij}x_j|\widehat{𝒪}|x_i.$$
(3.7)
We also find
$$x_i|exp(ik\widehat{x})|x_j=exp(ik\frac{(x_i+x_j)}{2}+\frac{i}{2}B_{\mu \nu }x_i^\mu x_j^\nu )exp(\frac{(x_ix_jd)^2}{4C})$$
(3.8)
where $`d^\mu =C^{\mu \nu }k_\nu `$. This matrix element sharply peaks at $`x_ix_j=d`$. It supports our interpretation that the eigenstate $`exp(ik\widehat{x})`$ with $`|k_\mu |>\lambda `$ can be interpreted as string like extended objects whose length is $`|C^{\mu \nu }k_\nu |`$. When $`|k_\mu |<\lambda `$, on the other hand, this matrix becomes close to diagonal whose matrix elements go like
$$x_i|exp(ik\widehat{x})|x_jexp(ikx_i)x_i|x_j.$$
(3.9)
It again supports our interpretation that $`exp(ik\widehat{x})`$ correspond to the ordinary plane waves when $`|k_\mu |<\lambda `$. They are represented by the matrices which are close to diagonal.
We now propose the bi-local field representation of noncommutative field theories. We may expand matrices $`\widehat{\varphi }`$ in the twisted reduced model by the following bi-local basis as follows:
$$\widehat{\varphi }=\underset{i,j}{}\varphi (x_i,x_j)|x_ix_j|$$
(3.10)
where the Hermiticity of $`\widehat{\varphi }`$ implies $`\varphi ^{}(x_j,x_i)=\varphi (x_i,x_j)`$. The matrices $`\widehat{\varphi }`$ represent $`\widehat{a_\mu }`$ or $`\widehat{\psi }`$ in the super Yang-Mills case but the setting here is more generally applied to an arbitrary noncommutative field theory. The bi-local basis spans the whole $`n^2`$ degrees of freedom of matrices. The product of two operators is also given as
$$\widehat{\varphi }_1\widehat{\varphi }_2=\underset{i,j,k,l}{}\varphi _1(x_i,x_j)\rho _{jk}\varphi _2(x_k,x_l)|x_ix_l|$$
(3.11)
and therefore
$$(\varphi _1\varphi _2)(x_i,x_j)=\underset{k,l}{}\varphi _1(x_i,x_k)\rho _{kl}\varphi _2(x_l,x_j).$$
(3.12)
The trace of operators $`\widehat{𝒪}_1\mathrm{}\widehat{𝒪}_s`$ in the bi-local basis is
$$Tr\widehat{𝒪}_1\mathrm{}\widehat{𝒪}_s=\underset{i_1\mathrm{}i_{2s}}{}𝒪_1(x_{i_1},x_{i_2})\rho _{i_2i_3}\mathrm{}𝒪_s(x_{i_{2s1}},x_{i_{2s}})\rho _{i_{2s}i_1}.$$
(3.13)
Here we work out the translation rule between the momentum eigenstate representation $`\widehat{\varphi }=_k\stackrel{~}{\varphi }(k)exp(ik\widehat{x})`$ and the bi-local field representation of eq.(3.10):
$`\stackrel{~}{\varphi }(k)`$ $`=`$ $`{\displaystyle \frac{1}{n}}Tr(exp(ik\widehat{x})\widehat{\varphi })={\displaystyle \frac{1}{n}}{\displaystyle \underset{i,j}{}}x_i|exp(ik\widehat{x})|x_j\varphi (x_j,x_i)`$
$`=`$ $`{\displaystyle \frac{1}{n}}{\displaystyle \underset{x_c}{}}\varphi (x_c)exp(ikx_c),`$
$`\varphi (x_c)`$ $`=`$ $`{\displaystyle \underset{x_r}{}}exp({\displaystyle \frac{i}{2}}B_{\mu \nu }x_r^\mu x_c^\nu )exp({\displaystyle \frac{(x_rd)^2}{4C}})\varphi (x_j,x_i),`$ (3.14)
where $`x_c=(x_i+x_j)/2`$ and $`x_r=x_ix_j`$. From eq.(3.14), we observe that the slowly varying field with the momentum smaller than $`\lambda `$ consists of the almost diagonal components. Hence close to diagonal components of the bi-local field are identified with the ordinary slowly varying field $`\varphi (x_c)`$. On the other hand, rapidly oscillating fields are mapped to the off-diagonal open string states. A large momentum in $`\nu `$ direction $`|k_\nu |>\lambda `$ corresponds to a large distance in the $`\mu `$-th direction $`|d^\mu |=|C^{\mu \nu }k_\nu |>l_{NC}`$. We can decompose $`d`$ as $`d=d_0+\delta d`$ where $`d_0`$ is a vector which connects two points on the von Neumann lattice and $`|\delta d|<l_{NC}`$. Then the summation over $`x_r`$ in (3.14) is dominated at $`x_r=d_0`$. In this way the large momentum degrees of freedom are more naturally interpreted as extended open string-like fields. They are denoted by ‘open strings’ in this paper.
The adjoint $`P^2`$ acts on $`\widehat{\varphi }`$ as
$`P^2\widehat{\varphi }={\displaystyle \frac{B^2}{2}}X^2\widehat{\varphi }`$ (3.15)
$`=`$ $`{\displaystyle \frac{B^2}{2}}(\widehat{x}^{+a}\widehat{x}^a\widehat{\varphi }\widehat{x}^a\widehat{\varphi }\widehat{x}^{+a}\widehat{x}^{+a}\widehat{\varphi }\widehat{x}^a+\widehat{\varphi }\widehat{x}^{+a}\widehat{x}^a+\stackrel{~}{d}C\widehat{\varphi }).`$
The kinetic term of bosonic fields is of the form $`Tr((P_\mu \widehat{\varphi })^2)=Tr(\widehat{\varphi }P^2\widehat{\varphi })`$ and is evaluated in this basis as
$`{\displaystyle \frac{1}{2}}Tr\widehat{\varphi }P^2\widehat{\varphi }={\displaystyle \frac{B^2}{2}}Tr\widehat{\varphi }X^2\widehat{\varphi }`$ (3.16)
$`=`$ $`{\displaystyle \frac{B^2}{2}}{\displaystyle \underset{i,j,k,l}{}}\varphi (x_i,x_j)\varphi (x_k,x_l)\rho _{jk}\rho _{li}\{(x_l^{+a}x_j^{+a})(x_i^ax_k^a)+\stackrel{~}{d}C\},`$
where we have used the property of the coherent basis $`\widehat{x}^a|x_i>=x_i^a|x_i>`$.
Here we decompose $`\rho _{ij}`$ by inserting the identity constructed by the orthogonal basis:
$$1=\underset{k}{}|x_k\}\{x_k|.$$
(3.17)
An explicit construction of orthogonal localized wave functions $`w_{m_in_i}(x)=<x|x_i\}`$ is carried out in the two dimensional case. They have a nearly Gaussian shape up to the radius $`r<l_{NC}`$ from the center. Outside this region, it is small, oscillates, and falls off with $`r^2`$. The remarkable property of the asymptotic form is that it vanishes on the von Neumann lattice. It implies that $`<x_i|x_j\}`$ vanish for large $`|x_ix_j|`$ very rapidly. Then eq.(3.16) can be rewritten for large $`|x_ix_j|`$ as
$`{\displaystyle \frac{B^2}{2}}{\displaystyle \underset{i,j,k,l,n,m}{}}\varphi (x_i,x_j)\varphi (x_k,x_l)<x_j|x_m\}\{x_m|x_k><x_l|x_n\}\{x_n|x_i>`$ (3.18)
$`\times (x_l^{+a}x_j^{+a})(x_i^ax_k^a)`$
$`=`$ $`{\displaystyle \frac{B^2}{2}}{\displaystyle \underset{n,m}{}}\stackrel{ˇ}{\varphi }(x_n,x_m)\stackrel{ˇ}{\varphi }(x_m,x_n)(x_nx_m)^2,`$
where
$$\stackrel{ˇ}{\varphi }(x_i,x_j)=\underset{k,l}{}\{x_i|x_k>\varphi (x_k,x_l)<x_l|x_j\}=\{x_i|\widehat{\varphi }|x_j\}.$$
(3.19)
In this derivation, we have used the fact that $`\{x_n|x_i>`$ are supported only for small $`|x_nx_i|`$.
The kinetic term is diagonal for large $`|x_ix_j|`$ and we read off the propagator of the bi-local field
$$\stackrel{ˇ}{\varphi }(x_i,x_j)\stackrel{ˇ}{\varphi }(x_j,x_i)=\frac{C^2}{(x_ix_j)^2}.$$
(3.20)
Long bi-local fields are thus interpreted to be far off-shell. It may be reinterpreted as
$$\stackrel{ˇ}{\varphi }(p_i,p_j)\stackrel{ˇ}{\varphi }(p_j,p_i)=\frac{1}{(p_ip_j)^2},$$
(3.21)
where $`p_{i,\mu }=B_{\mu \nu }x_i^\nu `$. For small $`|x_ix_j|`$, it is more appropriate to use the momentum eigenstate representation and we can obtain the standard propagator in that way.
We next consider three point vertices. Let us first consider the simplest three point vertex:
$`Tr(\widehat{\varphi }^3)`$ $`=`$ $`{\displaystyle \underset{i,j,k,l,m,n}{}}\varphi (x_i,x_j)\rho _{jk}\varphi (x_k,x_l)\rho _{lm}\varphi (x_m,x_n)\rho _{ni}`$ (3.22)
$`=`$ $`{\displaystyle \underset{i,j,k}{}}\stackrel{ˇ}{\varphi }(x_i,x_j)\stackrel{ˇ}{\varphi }(x_j,x_k)\stackrel{ˇ}{\varphi }(x_k,x_i).`$
We can similarly obtain higher point vertices as well:
$`Tr(\widehat{\varphi }^s)`$ $`=`$ $`{\displaystyle \underset{i_1\mathrm{}i_{2s}}{}}\varphi (x_{i_1},x_{i_2})\rho _{i_2i_3}\mathrm{}\varphi (x_{i_{2s1}},x_{i_{2s}})\rho _{i_{2s}i_1}`$ (3.23)
$`=`$ $`{\displaystyle \underset{i_1\mathrm{}i_s}{}}\stackrel{ˇ}{\varphi }(x_{i_1},x_{i_2})\mathrm{}\stackrel{ˇ}{\varphi }(x_{i_s},x_{i_1}).`$
In this way, we make contact with the quenched reduced models in the large momentum region. In quenched reduced models, the eigenvalues of matrices are identified with momenta. Therefore we have found that the large momentum behavior of twisted reduced models is identical to quenched reduced models. In this correspondence, the relative distance of the two ends of ‘open string’ in twisted reduced model is related to the momentum in quenched reduced model by the relation $`\widehat{p}_\mu =B_{\mu \nu }\widehat{x}^\nu `$.
### 3.2 Perturbations
Perturbative behaviors of noncommutative field theories have been investigated . In particular, it was pointed out that the effective action exhibits infrared singular behaviors due to the nonplanar diagrams. We now look at the perturbation of the noncommutative field theories in the bi-local basis. Let us consider the $`\varphi ^3`$ theory as a simple example. The matrix model action is given by
$$S=Tr\left(\frac{1}{2}[\widehat{p}_\mu ,\widehat{\varphi }]^2+\frac{\lambda }{3}\widehat{\varphi }^3\right).$$
(3.24)
In the large momentum region, the propagator and vertex are given by (3.20) and (3.22). In the one-loop approximation, there are two types of diagrams.
Fig. 1 is planar and gives the correction to the propagator of the bi-local field $`\stackrel{ˇ}{\varphi }(x_i,x_j)`$:
$$\underset{i,j}{}\underset{k}{}\frac{C^4\stackrel{ˇ}{\varphi }(x_i,x_j)\stackrel{ˇ}{\varphi }(x_j,x_i)}{(x_ix_k)^2(x_jx_k)^2}.$$
(3.25)
This diagram corresponds to the following planar diagram amplitude in the conventional perturbative expansion of noncommutative field theories:
$$\underset{p,k}{}\frac{\stackrel{~}{\varphi }(p)\stackrel{~}{\varphi }(p)}{k^2(pk)^2}.$$
(3.26)
As originally proved in , noncommutative phases which can be assigned to the propagators are canceled in planar diagrams and the amplitudes of such graphs are the same as the commutative cases except for the phase factors which only depend on the external momenta. In $`\stackrel{~}{d}`$-dimensions, the integral (3.26) is UV divergent as $`\mathrm{\Lambda }^{\stackrel{~}{d}4}`$ where $`\mathrm{\Lambda }`$ is the ultraviolet cutoff. In the bi-local representation (3.25), the summation is also divergent but it originates from the infrared extension of the von Neumann lattice. It is reminiscent of the tadpole divergences in string theory. However this analogy requires more work to substantiate it. For example, we may be able to work out a direct world sheet interpretation of the bi-local propagators.
Fig. 2 is a nonplanar diagram and induces effective interactions for only the diagonal components $`\stackrel{ˇ}{\varphi }(x_i,x_i)`$ of the bi-local fields. Since the diagonal components are interpreted as ordinary (slowly varying ) local fields $`\varphi (x_i)`$, this diagram induces long range interactions:
$$S_{eff}=C^4\underset{i,j}{}\frac{\varphi (x_i)\varphi (x_j)}{(x_ix_j)^4}.$$
(3.27)
This type of interactions have been well known as the block-block interactions in the matrix models .
In the conventional picture of noncommutative $`\varphi ^3`$ theory, this diagram provides the following nonplanar one-loop correction to the propagator.
$$\underset{p,k}{}\frac{\stackrel{~}{\varphi }(p)\stackrel{~}{\varphi }(p)}{k^2(pk)^2}exp(iC_{\mu \nu }k^\mu p^\nu ).$$
(3.28)
It is very specific to noncommutative field theory since it exhibits infrared singular behaviors for small $`p`$.
A physically intuitive argument which relates the nonplanar amplitude in the bi-local basis (3.27) and the conventional nonplanar amplitude (3.28) is given as follows:
$`S_{eff}`$ $`=`$ $`C^4{\displaystyle \underset{p,q}{}}{\displaystyle \underset{i,j}{}}{\displaystyle \frac{\stackrel{~}{\varphi }(p)\stackrel{~}{\varphi }(q)}{(x_ix_j)^4}}e^{ipx_i}e^{iqx_j}`$ (3.29)
$`=`$ $`nC^4{\displaystyle \underset{p}{}}{\displaystyle \underset{d}{}}{\displaystyle \frac{\stackrel{~}{\varphi }(p)\stackrel{~}{\varphi }(p)}{d^4}}e^{ipd}.`$
If we rewrite $`k_\mu =B_{\mu \nu }d^\nu `$, this has the same expression as eq.(3.28). Here we also recall that the momentum lattice $`\{k_\mu \}`$ has finer resolutions than the von Neumann lattice $`\{d_\mu \}`$. Therefore $`_k=n_d`$ for large $`|d|`$. Note that the nonplanar phase is interpreted as the wave functions of plane waves here and the summation over $`k`$ in (3.28) amounts to performing the Fourier transformation. On the other hand, the summation over $`k`$ in (3.26) leads to the standard one loop integration in the planar contribution.
From these arguments, it is clear that the long range interaction $`1/d^4`$ is still induced even when the $`\varphi `$ field has a mass term $`m`$ as long as $`d>Cm`$. In the next section, we investigate these long range interactions of matrix models in more detail.
## 4 Interactions between diagonal blocks
In this section, we investigate the renormalization property of noncommutative field theory. After integrating long ‘open strings’, we obtain long range interactions. This phenomena appear as nonplanar contributions in noncommutative field theory. We show that it can be simply understood as block-block interactions in the matrix model picture.
### 4.1 Noncommutative $`\varphi ^3`$
We consider a noncommutative $`\varphi ^3`$ field theory as a simple example. The matrix model action is given by (3.24). Following the same procedure as in eqs. (2.5 \- 2.8), we can obtain a noncommutative $`U(m)`$ $`\varphi ^3`$ field theory:
$$S=d^{\stackrel{~}{d}}xtr\left(\frac{1}{2}(_\mu \varphi (x))^2+\frac{\lambda ^{}}{3}\varphi (x)^3\right)_{}.$$
(4.1)
Here $`tr`$ means a trace over $`(m\times m)`$ matrices and $`\lambda ^{}=\lambda (2\pi /B)^{\stackrel{~}{d}/4}`$.
We consider a group of backgrounds $`\varphi ^{(i)}`$. We assume that the corresponding functions $`\varphi ^{(i)}(x)`$ are localized and well separated. $`\varphi ^{(i)}`$ can be specified by its Fourier components $`\stackrel{~}{\varphi }^{(i)}(k)`$:
$$\widehat{\varphi }^{(i)}=\underset{k}{}\stackrel{~}{\varphi }^{(i)}(k)exp(ik\widehat{x}).$$
(4.2)
We further assume that $`\stackrel{~}{\varphi }^{(i)}(k)`$ is only supported for $`|k|<<\lambda `$ which implies the much larger localization scale than the noncommutativity scale. This condition means that the background in the bi-local representation $`\varphi ^{(i)}(x,y)`$ is almost diagonal. We hence expect that the following representation is accurate :
$$\widehat{\varphi }^{(i)}=\underset{j}{}\varphi ^{(i)}(x_j)|x_jx_j|.$$
(4.3)
The background is assumed to be localized around $`x=d^{(i)}`$. In this section we obtain the effective action for these backgrounds by integrating over the off-diagonal components of the bi-local field, that is, the higher momentum fields.
The assumption of no overlap imposes certain conditions on $`\stackrel{~}{\varphi }^{(i)}(k)`$. The commutators are mapped in our mapping rule as follows:
$$[\varphi ^{(i)},\varphi ^{(j)}]\varphi ^{(i)}(x)\varphi ^{(j)}(x)\varphi ^{(j)}(x)\varphi ^{(i)}(x).$$
(4.4)
Then we can see that $`\varphi ^{(i)}`$ and $`\varphi ^{(j)}`$ are commutative to each other from eq.(4.4). Therefore they can be represented by a block diagonal form as follows:
$$\widehat{\varphi }_{cl}=\widehat{\varphi }^{(i)}=\left(\begin{array}{cccc}\varphi ^{(1)}& & & \\ & \varphi ^{(2)}& & \\ & & \varphi ^{(3)}& \\ & & & \mathrm{}\end{array}\right),$$
(4.5)
where $`\varphi ^{(i)}`$ $`(i=1,2,\mathrm{})`$ is a $`n_i\times n_i`$ matrix.
$`[\widehat{p}_\mu ,\varphi ^{(i)}]`$ can be mapped onto $`i_\mu \varphi ^{(i)}(x)`$. Since $`_\mu \varphi ^{(i)}(x)\varphi ^{(j)}(x)\varphi ^{(j)}(x)_\mu \varphi ^{(i)}(x)`$ vanishes for well separated wave-packets, $`[\widehat{p}_\mu ,\varphi ^{(i)}]`$ are also of the block diagonal form as in eq.(4.5). This fact in turn implies that $`\widehat{p}_\mu `$ can be represented in the same block diagonal form when it acts on these backgrounds:
$$\widehat{p}_\mu =\left(\begin{array}{cccc}p_\mu ^{(1)}& & & \\ & p_\mu ^{(2)}& & \\ & & p_\mu ^{(3)}& \\ & & & \mathrm{}\end{array}\right),$$
(4.6)
Here $`p_\mu ^{(i)}`$ denotes the projection of $`\widehat{p}_\mu `$ onto the subspace specified by $`\varphi ^{(i)}`$. $`\widehat{p}_\mu `$ can be represented as the block diagonal form since we have projected it onto the subspace where the backgrounds $`\varphi ^{(i)}`$ are supported.
We now calculate the one-loop effective action between diagonal blocks in order to investigate the renormalization property of noncommutative field theory. We may further decompose $`p_\mu ^{(i)}`$ as
$`p_\mu ^{(i)}`$ $`=`$ $`B_{\mu \nu }d_\nu ^{(i)}1_{n_i}+\stackrel{~}{p}_\mu ^{(i)},`$
$`Tr\stackrel{~}{p}_\mu ^{(i)}`$ $`=`$ $`0,`$ (4.7)
where $`d_\nu ^{(i)}`$ is a $`c`$ number representing the center of mass coordinate of the $`i`$-th block. Here we assume that the blocks are separated far enough from each other, that is, for all $`i`$ and $`j`$’s, $`(d_\mu ^{(i)}d_\mu ^{(j)})^2`$’s are larger than the localization scale of each block.
We first recall some notations which are introduced in . We denote the $`(i,j)`$ block of a matrix $`X`$ as $`X^{(i,j)}`$. It is clear that $`P_\mu =[\widehat{p}_\mu ,]`$ operates on each $`X^{(i,j)}`$ independently. In fact we have
$$(P_\mu X)^{(i,j)}=B_{\mu \nu }(d_\nu ^{(i)}d_\nu ^{(j)})X^{(i,j)}+\stackrel{~}{p}_\mu ^{(i)}X^{(i,j)}X^{(i,j)}\stackrel{~}{p}_\mu ^{(j)}.$$
(4.8)
In the bi-local basis developed in the previous section, $`(i,j)`$ block of a matrix $`X^{(i,j)}`$ may correspond to a collections of bi-local fields which connect the $`i`$-th and $`j`$-th diagonal blocks. We further simplify this equation by introducing notations such as
$`d_\mu ^{(i,j)}X^{(i,j)}`$ $`=`$ $`(d_\mu ^{(i)}d_\mu ^{(j)})X^{(i,j)},`$
$`P_{L\mu }^{(i,j)}X^{(i,j)}`$ $`=`$ $`\stackrel{~}{p}_\mu ^{(i)}X^{(i,j)},`$
$`P_{R\mu }^{(i,j)}X^{(i,j)}`$ $`=`$ $`X^{(i,j)}\stackrel{~}{p}_\mu ^{(j)}.`$ (4.9)
Note that $`d_\mu ^{(i,j)}`$,$`P_{L\mu }^{(i,j)}`$ and $`P_{R\mu }^{(i,j)}`$ commute each other, and the operation of $`P_\mu `$ on $`X^{(i,j)}`$ is expressed as
$$(P_\mu X)^{(i,j)}=(B_{\mu \nu }d_\nu ^{(i,j)}+P_{L\mu }^{(i,j)}+P_{R\mu }^{(i,j)})X^{(i,j)}.$$
(4.10)
We can also decompose the action of $`\widehat{\varphi }_{cl}`$ onto $`X^{(i,j)}`$, which should be made symmetric between the right and the left multiplications for Hermiticity, in the same way as (4.10):
$$(\mathrm{\Phi }X)^{(i,j)}(\mathrm{\Phi }_L^{(i,j)}+\mathrm{\Phi }_R^{(i,j)})X^{(i,j)},$$
(4.11)
where
$`\mathrm{\Phi }_L^{(i,j)}X^{(i,j)}`$ $`=`$ $`\varphi ^{(i)}X^{(i,j)},`$
$`\mathrm{\Phi }_R^{(i,j)}X^{(i,j)}`$ $`=`$ $`X^{(i,j)}\varphi ^{(j)}.`$ (4.12)
Since the left and right multiplication are totally independent, we have
$`(OX)^{(i,j)}`$ $``$ $`O_L^{(i,j)}X^{(i,j)}O_R^{(i,j)},`$
$`𝒯rO`$ $`=`$ $`{\displaystyle \underset{i,j=1}{\overset{n}{}}}TrO_L^{(i,j)}TrO_R^{(i,j)},`$ (4.13)
for operators consisting of $`P_\mu `$ and $`\mathrm{\Phi }`$. Here $`𝒯r`$ denotes the trace of the operators which act on the matrices.
The one loop effective action is
$$W=\frac{1}{2}𝒯rlog(1\frac{1}{P_\mu ^2}\lambda \mathrm{\Phi })).$$
(4.14)
Now we expand the expression of the one-loop effective action with respect to the inverse power of $`d_\mu ^{(i,j)}`$’s. We drop the linear term in the background fields since we adopt the background field method which has no ambiguity in this case unlike the gauge fixing ambiguity in gauge theory. We have,
$`W`$ $`=`$ $`{\displaystyle \frac{1}{4}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }\right)`$ (4.15)
$`{\displaystyle \frac{1}{6}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }\right)`$
$`{\displaystyle \frac{1}{8}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }\right)+O((\mathrm{\Phi })^5).`$
Since as in (4.10) and (4.11) $`P_\mu `$ and $`\mathrm{\Phi }`$ act on the $`(i,j)`$ blocks independently, the one-loop effective action $`W`$ is expressed as the sum of contributions of the $`(i,j)`$ blocks $`W^{(i,j)}`$. Therefore we may consider $`W^{(i,j)}`$ as the interaction between the $`i`$-th and $`j`$-th blocks. Using (4.13) and (4.15) we can easily evaluate $`W^{(i,j)}`$ to the leading order of $`1/\sqrt{(d^{(i)}d^{(j)})^2}`$ as
$`W^{(i,j)}`$ $`=`$ $`{\displaystyle \frac{C^4}{(d^{(i)}d^{(j)})^4}}({\displaystyle \frac{\lambda ^2}{4}})𝒯r^{(i,j)}\left(\mathrm{\Phi }\mathrm{\Phi }\right)`$ (4.16)
$`+O((1/(d^{(i)}d^{(j)})^5)`$
$`=`$ $`{\displaystyle \frac{C^4}{(d^{(i)}d^{(j)})^4}}({\displaystyle \frac{\lambda ^2}{4}})(n_jTr(\varphi ^{(i)}\varphi ^{(i)})+n_iTr(\varphi ^{(j)}\varphi ^{(j)})+2Tr(\varphi ^{(i)})Tr(\varphi ^{(j)}))`$
$`+O((1/(d^{(i)}d^{(j)})^5).`$
The third term in the above expression can be interpreted as the signature for the existence of massless particles corresponding to a scalar in six dimensions.
A more conventional approach to calculate $`𝒯r`$ is to use the plane-wave basis $`exp(ik\widehat{x})`$. It corresponds to a standard one loop calculation in noncommutative $`\varphi ^3`$:
$`W`$ $`=`$ $`{\displaystyle \frac{1}{4}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }\right)`$ (4.17)
$`=`$ $`{\displaystyle \frac{1}{4n}}{\displaystyle \underset{k}{}}Tr(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }{\displaystyle \frac{1}{P^2}}\lambda \mathrm{\Phi }exp(ik\widehat{x}))`$
$`=`$ $`{\displaystyle \frac{\lambda ^2}{4}}{\displaystyle \underset{l}{}}\stackrel{~}{\varphi }_{cl}(l)\stackrel{~}{\varphi }_{cl}(l){\displaystyle \underset{k}{}}{\displaystyle \frac{2}{k^2(k+l)^2}}(1+exp(iC^{\mu \nu }k_\mu l_\nu )).`$
Here we find the both planar and nonplanar contributions. The latter contains the nontrivial phase factor $`exp(iC^{\mu \nu }k_\mu l_\nu )`$.
We evaluate the nonplanar contributions:
$`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k}{}}{\displaystyle \frac{1}{k^2(k+l)^2}}exp(iC^{\mu \nu }k_\mu l_\nu )`$ (4.18)
$`=`$ $`({\displaystyle \frac{1}{2\pi B}})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}k\frac{1}{k^2(k+l)^2}exp(iC^{\mu \nu }k_\mu l_\nu )}`$
$`=`$ $`({\displaystyle \frac{1}{2\pi B}})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle 𝑑\alpha _1𝑑\alpha _2(\frac{\pi }{\alpha _1+\alpha _2})^{\frac{\stackrel{~}{d}}{2}}exp(\frac{\alpha _1\alpha _2}{\alpha _1+\alpha _2}l^2\frac{(Cl)^2}{4(\alpha _1+\alpha _2)})}.`$
When $`\stackrel{~}{d}>4`$, the above integral is evaluated as
$$\mathrm{\Gamma }(\frac{\stackrel{~}{d}}{2}2)(\frac{C}{2})^{4\frac{\stackrel{~}{d}}{2}}(l^2)^{2\frac{\stackrel{~}{d}}{2}}.$$
(4.19)
We therefore find the nonplanar contribution:
$`n{\displaystyle \frac{\lambda ^2}{2}}{\displaystyle \underset{l}{}}\stackrel{~}{\varphi }(l)\stackrel{~}{\varphi }(l)({\displaystyle \frac{C}{2}})^{4\frac{\stackrel{~}{d}}{2}}(l^2)^{2\frac{\stackrel{~}{d}}{2}}\mathrm{\Gamma }({\displaystyle \frac{\stackrel{~}{d}}{2}}2)`$ (4.20)
$`=`$ $`n^2{\displaystyle \frac{\lambda ^2C^4}{2}}({\displaystyle \frac{1}{2}})^{4\frac{\stackrel{~}{d}}{2}}\mathrm{\Gamma }({\displaystyle \frac{\stackrel{~}{d}}{2}}2)({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}l\stackrel{~}{\varphi }(l)\stackrel{~}{\varphi }(l)(l^2)^{2\frac{\stackrel{~}{d}}{2}}}`$
$`=`$ $`{\displaystyle \frac{\lambda ^2C^4}{2}}({\displaystyle \frac{1}{2\pi C}})^{\stackrel{~}{d}}{\displaystyle d^{\stackrel{~}{d}}xd^{\stackrel{~}{d}}y\varphi (x)\frac{1}{(xy)^4}\varphi (y)}`$
where we have used the identity:
$$\frac{d^{\stackrel{~}{d}}k}{(2\pi )^{\stackrel{~}{d}}}exp(ikx)(\frac{1}{k^2})^{\frac{\stackrel{~}{d}}{2}2}=\frac{1}{2^{\stackrel{~}{d}4}}(\frac{1}{\pi })^{\frac{\stackrel{~}{d}}{2}}\frac{1}{\mathrm{\Gamma }(\frac{\stackrel{~}{d}}{2}2)}\frac{1}{|x|^4}.$$
(4.21)
We also note that our convention is
$$\stackrel{~}{\varphi }_{cl}(l)=\frac{1}{n(2\pi C)^{\frac{\stackrel{~}{d}}{2}}}d^{\stackrel{~}{d}}xexp(ilx)\varphi _{cl}(x).$$
(4.22)
We observe that eq.(4.20) can be identified with the last term of eq.(4.16) since it can be reexpressed as follows due to our mapping rule eq.(2.8):
$`{\displaystyle \underset{ij}{}}{\displaystyle \frac{C^4}{(d^{(i)}d^{(j)})^4}}({\displaystyle \frac{\lambda ^2}{2}})Tr(\varphi ^{(i)})Tr(\varphi ^{(j)})`$ (4.23)
$`=`$ $`{\displaystyle \frac{\lambda ^2C^4}{2}}({\displaystyle \frac{1}{2\pi C}})^{\stackrel{~}{d}}{\displaystyle d^{\stackrel{~}{d}}xd^{\stackrel{~}{d}}ytr\varphi _{cl}(x)\frac{1}{(xy)^4}tr\varphi _{cl}(y)}.`$
Here we have replaced the trace over the localized classical background $`\varphi _i`$ by the integration over the space-time coordinates in the neighborhood of the wave packet. By integrating over the entire space-time, we also sum over different localized blocks. Since we are only considering the interactions of the well separated backgrounds, the above formula is only valid in the low momentum region. What we have found here is that the nonplanar contributions which give rise to the long range interactions can be identified with those from off-diagonal components in matrix models. We can simply reproduce it by considering block-block interactions. These are distinguished features of noncommutative field theory which are absent in field theory.
### 4.2 Noncommutative $`\varphi ^4`$
We consider noncommutative $`\varphi ^4`$ theory next
$$S=Tr\left(\frac{1}{2}[\widehat{p}_\mu ,\widehat{\varphi }]^2+\frac{\lambda }{4}\widehat{\varphi }^4\right).$$
(4.24)
The one loop effective action is
$$W=\frac{1}{2}𝒯rlog(1\frac{1}{P_\mu ^2}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R)).$$
(4.25)
Now we expand the expression of the one-loop effective action with respect to the inverse power of $`d_\mu ^{(i,j)}`$’s. We have,
$`W`$ $`=`$ $`{\displaystyle \frac{1}{2}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R)\right)`$ (4.26)
$`{\displaystyle \frac{1}{4}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R){\displaystyle \frac{1}{P^2}}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R)\right)+O((\mathrm{\Phi })^5).`$
We can easily evaluate $`W^{(i,j)}`$ to the leading order of $`1/\sqrt{(d^{(i)}d^{(j)})^2}`$ as
$`W^{(i,j)}`$ $`=`$ $`{\displaystyle \frac{1}{(d^{(i)}d^{(j)})^2}}({\displaystyle \frac{\lambda }{2}})𝒯r^{(i,j)}\left(\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R\right)`$ (4.27)
$`+O((1/(d^{(i)}d(j))^3)`$
$`=`$ $`{\displaystyle \frac{1}{(d^{(i)}d^{(j)})^2}}({\displaystyle \frac{\lambda }{2}})(n_jTr(\varphi ^{(i)}\varphi ^{(i)})+n_iTr(\varphi ^{(j)}\varphi ^{(j)})+Tr(\varphi ^{(i)})Tr(\varphi ^{(j)}))`$
$`+O((1/(d^{(i)}d^{(j)})^3).`$
We find the exchanges of massless particles corresponding to a scalar in four dimensions.
We also evaluate the leading term of the effective action by using the plane-wave basis $`exp(ik\widehat{x})`$.
$`W`$ $`=`$ $`{\displaystyle \frac{1}{2}}𝒯r\left({\displaystyle \frac{1}{P^2}}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R)\right)`$ (4.28)
$`=`$ $`{\displaystyle \frac{1}{2n}}{\displaystyle \underset{k}{}}Tr(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}\lambda (\mathrm{\Phi }_L^2+\mathrm{\Phi }_R^2+\mathrm{\Phi }_L\mathrm{\Phi }_R)exp(ik\widehat{x}))`$
$`=`$ $`{\displaystyle \frac{\lambda }{2}}{\displaystyle \underset{l}{}}\stackrel{~}{\varphi }(l)\stackrel{~}{\varphi }(l){\displaystyle \underset{k}{}}{\displaystyle \frac{1}{k^2}}(2+exp(iC^{\mu \nu }k_\mu l_\nu )).`$
Here we also find the both planar and nonplanar contributions.
We evaluate the nonplanar contributions:
$`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k}{}}{\displaystyle \frac{1}{k^2}}exp(iC^{\mu \nu }k_\mu l_\nu )`$ (4.29)
$`=`$ $`({\displaystyle \frac{1}{2\pi B}})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}k\frac{1}{k^2}exp(iC^{\mu \nu }k_\mu l_\nu )}`$
$`=`$ $`({\displaystyle \frac{1}{2\pi B}})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle 𝑑\alpha (\frac{\pi }{\alpha })^{\frac{\stackrel{~}{d}}{2}}exp(\frac{(Cl)^2}{4\alpha })}.`$
When $`\stackrel{~}{d}>2`$, the above integral is evaluated as
$$\mathrm{\Gamma }(\frac{\stackrel{~}{d}}{2}1)(\frac{C}{2})^{2\frac{\stackrel{~}{d}}{2}}(l^2)^{1\frac{\stackrel{~}{d}}{2}}.$$
(4.30)
We therefore find the nonplanar contribution:
$`n{\displaystyle \frac{\lambda }{2}}{\displaystyle \underset{l}{}}\stackrel{~}{\varphi }(l)\stackrel{~}{\varphi }(l)({\displaystyle \frac{C}{2}})^{2\frac{\stackrel{~}{d}}{2}}(l^2)^{1\frac{\stackrel{~}{d}}{2}}\mathrm{\Gamma }({\displaystyle \frac{\stackrel{~}{d}}{2}}1)`$ (4.31)
$`=`$ $`n^2{\displaystyle \frac{\lambda C^2}{2}}({\displaystyle \frac{1}{2}})^{2\frac{\stackrel{~}{d}}{2}}\mathrm{\Gamma }({\displaystyle \frac{\stackrel{~}{d}}{2}}1)({\displaystyle \frac{1}{2\pi }})^{\frac{\stackrel{~}{d}}{2}}{\displaystyle d^{\stackrel{~}{d}}l\stackrel{~}{\varphi }(l)\stackrel{~}{\varphi }(l)(l^2)^{1\frac{\stackrel{~}{d}}{2}}}`$
$`=`$ $`{\displaystyle \frac{\lambda C^2}{2}}({\displaystyle \frac{1}{2\pi C}})^{\stackrel{~}{d}}{\displaystyle d^{\stackrel{~}{d}}xd^{\stackrel{~}{d}}y\varphi (x)\frac{1}{(xy)^2}\varphi (y)}`$
where we have used the identity:
$$\frac{d^{\stackrel{~}{d}}k}{(2\pi )^{\stackrel{~}{d}}}exp(ikx)(\frac{1}{k^2})^{\frac{\stackrel{~}{d}}{2}1}=\frac{1}{2^{\stackrel{~}{d}2}}(\frac{1}{\pi })^{\frac{\stackrel{~}{d}}{2}}\frac{1}{\mathrm{\Gamma }(\frac{\stackrel{~}{d}}{2}1)}\frac{1}{|x|^2}.$$
(4.32)
We observe that eq.(4.31) can be identified with the last term of eq.(4.27) since it can be reexpressed as follows due to our mapping rule:
$`{\displaystyle \underset{ij}{}}{\displaystyle \frac{C^2}{(d^{(i)}d^{(j)})^2}}({\displaystyle \frac{\lambda }{2}})Tr(\varphi ^{(i)})Tr(\varphi ^{(j)})`$ (4.33)
$`=`$ $`{\displaystyle \frac{\lambda C^2}{2}}({\displaystyle \frac{1}{2\pi C}})^{\stackrel{~}{d}}{\displaystyle d^{\stackrel{~}{d}}xd^{\stackrel{~}{d}}ytr\varphi _{cl}(x)\frac{1}{(xy)^2}tr\varphi _{cl}(y)}.`$
## 5 Block-block interactions in gauge theory
In the previous section we have seen that the long range block-block interactions are induced in the effective action for the slowly varying fields of noncommutative scalar theories after integrating out the off-diagonal components. They correspond to nonplanar diagrams which carry nontrivial phase factors.
These are the familiar interactions in matrix models and indeed in IIB matrix model they are interpreted as massless particle propagations . There is a conceptual difference, however, between scalar theory and gauge theory. In scalar theory we must introduce extra matrices $`\widehat{p}_\mu `$ to represent the noncommutative space-time. In reduced models of gauge theory, they are considered as special backgrounds of the dynamical variables ($`A_\mu =\widehat{p}_\mu +\widehat{a}_\mu `$). The block-block interactions in reduced models of gauge theory such as type IIB matrix model are universal and not restricted to specific backgrounds $`\widehat{p}_\mu `$ which define the twisted reduced models. Actually the original calculation of such interactions in does not assume any specific conditions for backgrounds. We have obtained the long range interactions which decay as $`1/r^8`$. It does not depend whether we assume uniform distributions of the matrix eigenvalues in ten dimensions or in lower dimensions such as four as is discussed in this section. These long range interactions are interpreted as propagations of the massless type IIB supergravity multiplets and this fact can be considered as one of the evidences that IIB matrix model contains gravity.
In this section we reexamine the block-block interactions of D-instantons in IIB matrix model. There we have considered the backgrounds which represent four dimensional noncommutative space-time. By expanding IIB matrix model around such backgrounds, we obtain four dimensional noncommutative Yang-Mills theory. There are nontrivial classical solutions in this theory which reduce to instantons in the large instanton size limit.
We have considered a classical solution of IIB matrix model which represents an instanton and an (anti)instanton. We can realize $`U(4)`$ gauge theory by considering four D3 branes. We embed an instanton into the first $`SU(2)`$ part and the other (anti)instanton into the remaining $`SU(2)`$ part. We separate them in the fifth dimension by the distance $`b`$:
$`A_0`$ $`=`$ $`\left(\begin{array}{cc}p_0+a_0& 0\\ 0& p_0+a_0^{}\end{array}\right)`$ (5.3)
$`A_1`$ $`=`$ $`\left(\begin{array}{cc}p_1+a_1& 0\\ 0& p_1+a_1^{}\end{array}\right)`$ (5.6)
$`A_2`$ $`=`$ $`\left(\begin{array}{cc}p_2+a_2& 0\\ 0& p_2+a_2^{}\end{array}\right)`$ (5.9)
$`A_3`$ $`=`$ $`\left(\begin{array}{cc}p_3+a_3& 0\\ 0& p_3+a_3^{}\end{array}\right)`$ (5.12)
$`A_4`$ $`=`$ $`\left(\begin{array}{cc}\frac{b}{2}& 0\\ 0& \frac{b}{2}\end{array}\right)`$ (5.15)
$`A_\rho `$ $`=`$ $`0`$ (5.16)
where $`\rho =5,\mathrm{},9`$.
While two instanton system receives no quantum corrections, the instanton - anti-instanton system receives quantum corrections since it is no longer BPS. We have evaluated the one loop effective potential due to an instanton and (anti)instanton. They are local excitations and couple to gravity. These solutions are characterized by the adjoint field strength $`F_{\mu \nu }`$ which does not vanish at the locations of the instantons. We assume that they are separated by a long distance compared to their sizes. We also assume that $`b>>l_{NC}`$. Then we can choose two disjoint blocks in each of which a large part of an (anti)instanton is contained. Let the location and the size of the two instantons $`(x_i,\rho _i)`$ and $`(x_j,\rho _j)`$.<sup>4</sup><sup>4</sup>4Here we have used the indices $`i`$ and $`j`$ to denote two instantons since they are also represented by the diagonal blocks in the matrix model. However we consider only two instantons in this section. The ten dimensional distance of them is $`r^2=(x_ix_j)^2+b^2`$. Here we have assumed that $`r>>\rho `$.
The one-loop effective action of IIB matrix model is
$$ReW=\frac{1}{2}𝒯r\mathrm{log}(P_\lambda ^2\delta _{\mu \nu }2iF_{\mu \nu })\frac{1}{4}𝒯r\mathrm{log}((P_\lambda ^2+\frac{i}{2}F_{\mu \nu }\mathrm{\Gamma }^{\mu \nu })(\frac{1+\mathrm{\Gamma }_{11}}{2}))𝒯r\mathrm{log}(P_\lambda ^2).$$
(5.17)
Here $`P_\mu `$ and $`F_{\mu \nu }`$ are operators acting on the space of matrices as
$`P_\mu X`$ $`=`$ $`[p_\mu ,X],`$
$`F_{\mu \nu }X`$ $`=`$ $`[f_{\mu \nu },X],`$ (5.18)
where $`f_{\mu \nu }=i[p_\mu ,p_\nu ]`$. Now we expand the general expression of the one-loop effective action (5.17) with respect to the inverse power of $`d_\mu ^{(i,j)}`$’s just like the preceding section. We can take traces of the $`\gamma `$ matrices after expanding the logarithm in (5.17). Due to the supersymmetry, contributions of bosons and fermions cancel each other to the third order in $`F_{\mu \nu }`$, and we have,
$`W`$ $`=`$ $`𝒯r\left({\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\nu \lambda }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\rho \mu }\right)`$ (5.19)
$`2𝒯r\left({\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\mu \rho }{\displaystyle \frac{1}{P^2}}F_{\lambda \nu }\right)`$
$`+{\displaystyle \frac{1}{2}}𝒯r\left({\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }\right)`$
$`+{\displaystyle \frac{1}{4}}𝒯r\left({\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }\right)+O((F_{\mu \nu })^5).`$
Since as in (4.10) and (4.11) $`P_\mu `$ and $`F_{\mu \nu }`$ act on the $`(i,j)`$ blocks independently, the one-loop effective action $`W`$ is expressed as the sum of contributions of the $`(i,j)`$ blocks $`W^{(i,j)}`$. Therefore we may consider $`W^{(i,j)}`$ as the interaction between the $`i`$-th and $`j`$-th blocks.
Using (4.13) and (5.19) we can easily evaluate $`W^{(i,j)}`$ to the leading order of $`1/\sqrt{(d^{(i)}d^{(j)})^2}`$ as
$`W^{(i,j)}`$ $`=`$ $`{\displaystyle \frac{1}{r^8}}(𝒯r^{(i,j)}(F_{\mu \nu }F_{\nu \lambda }F_{\lambda \rho }F_{\rho \mu })2𝒯r^{(i,j)}(F_{\mu \nu }F_{\lambda \rho }F_{\mu \rho }F_{\lambda \nu })`$ (5.20)
$`+{\displaystyle \frac{1}{2}}𝒯r^{(i,j)}(F_{\mu \nu }F_{\mu \nu }F_{\lambda \rho }F_{\lambda \rho })+{\displaystyle \frac{1}{4}}𝒯r^{(i,j)}(F_{\mu \nu }F_{\lambda \rho }F_{\mu \nu }F_{\lambda \rho })`$
$`=`$ $`{\displaystyle \frac{3}{2r^8}}(n_j\stackrel{~}{b}_8(f^{(i)})n_i\stackrel{~}{b}_8(f^{(j)})`$
$`8Tr(f_{\mu \nu }^{(i)}f_{\nu \sigma }^{(i)})Tr(f_{\mu \rho }^{(j)}f_{\rho sigma}^{(j)})+Tr(f_{\mu \nu }^{(i)}f_{\mu \nu }^{(i)})Tr(f_{\rho \sigma }^{(j)}f_{\rho \sigma }^{(j)})`$
$`+Tr(f_{\mu \nu }^{(i)}\stackrel{~}{f}_{\mu \nu }^{(i)})Tr(f_{\rho \sigma }^{(j)}\stackrel{~}{f}_{\rho \sigma }^{(j)}))`$
where $`\stackrel{~}{f}_{\mu \nu }=ϵ_{\mu \nu \rho \sigma }f_{\rho \sigma }/2`$ and
$$\stackrel{~}{b}_8(f)=\frac{2}{3}(Tr(f_{\mu \nu }f_{\nu \lambda }f_{\lambda \rho }f_{\rho \mu })+2Tr(f_{\mu \nu }f_{\lambda \rho }f_{\mu \rho }f_{\lambda \nu })\frac{1}{2}Tr(f_{\mu \nu }f_{\mu \nu }f_{\lambda \rho }f_{\lambda \rho })\frac{1}{4}Tr(f_{\mu \nu }f_{\lambda \rho }f_{\mu \nu }f_{\lambda \rho })).$$
(5.21)
In eq.(5.20), we have kept axion type interactions also. Note that the $`\stackrel{~}{b}_8(f)=0`$ for an (anti)instanton configuration. So the potential between an instanton and an (anti)instanton is
$$\frac{3}{2r^8}(8Tr(f_{\mu \nu }^{(i)}f_{\nu \sigma }^{(i)})Tr(f_{\mu \rho }^{(j)}f_{\rho \sigma }^{(j)})+Tr(f_{\mu \nu }^{(i)}f_{\mu \nu }^{(i)})Tr(f_{\rho \sigma }^{(j)}f_{\rho \sigma }^{(j)})+Tr(f_{\mu \nu }^{(i)}\stackrel{~}{f}_{\mu \nu }^{(i)})Tr(f_{\rho \sigma }^{(j)}\stackrel{~}{f}_{\rho \sigma }^{(j)})).$$
(5.22)
Here we can apply the low energy approximation such as
$`Tr(f_{\mu \nu }^{(i)}f_{\mu \nu }^{(i)}){\displaystyle \frac{C^4}{l_{NC}^4}}{\displaystyle d^4xtr([D_\mu ^i,D_\nu ^i][D_\mu ^i,D_\nu ^i])}={\displaystyle \frac{l_{NC}^4}{\pi ^2}},`$
$`Tr(f_{\mu \nu }^{(i)}\stackrel{~}{f}_{\mu \nu }^{(i)}){\displaystyle \frac{C^4}{2l_{NC}^4}}{\displaystyle d^4xtrϵ_{\mu \nu \rho \sigma }([D_\mu ^i,D_\nu ^i][D_\rho ^i,D_\sigma ^i])}=\pm {\displaystyle \frac{l_{NC}^4}{\pi ^2}},`$
$`Tr(f_{\mu \nu }^{(i)}f_{\nu \rho }^{(i)}){\displaystyle \frac{C^4}{l_{NC}^4}}{\displaystyle d^4xtr([D_\mu ^i,D_\nu ^i][D_\nu ^i,D_\rho ^i])}={\displaystyle \frac{l_{NC}^4}{4\pi ^2}}\delta _{\mu \rho }`$ (5.23)
where $`D_\mu ^i`$ denotes the covariant derivative of the instanton background which is localized at the $`i`$-th block. So the interactions eq.(5.22) can be interpreted due to the exchange of dilaton, axion and graviton. We have found that the potential between two instantons vanish due to their BPS nature. On the other hand, the following potential is found between an instanton and an anti-instanton
$$\frac{3}{\pi ^4}\frac{l_{NC}^8}{r^8}.$$
(5.24)
There is no reason to believe the above approximation when $`b<l_{NC}`$. In this case the interactions between an instanton and an anti-instanton is well described by the gauge fields which are low energy modes of IIB matrix model. They are close to diagonal degrees of freedom in IIB matrix model. Their contribution can be estimated by gauge theory. The one loop effective potential can be calculated by gauge theory certainly when $`b<<l_{NC}`$
$$\mathrm{\Gamma }=\frac{C^4}{2(4\pi )^2b^4}d^4xb_8$$
(5.25)
where we have assumed $`b\rho >>C`$. The above expression is estimated as follows:
$`\mathrm{\Gamma }`$ $`=`$ $`{\displaystyle \frac{144C^4}{\pi ^2b^4}}{\displaystyle d^4x\frac{\rho ^4}{((xy)^2+\rho ^2)^4}\frac{\rho ^4}{((xz)^2+\rho ^2)^4}}`$ (5.26)
$`{\displaystyle \frac{24\rho ^4C^4}{r^8b^4}}`$
where $`r=|yz|`$ is assumed to be much larger than $`\rho `$. We note that eq.(5.26) falls off with the identical power for large $`r`$ with eq.(5.24). On the other hand when $`b>>l_{NC}`$, the standard gauge theory description is no longer valid since we have to take account of the noncommutativity. In fact we have argued that the block-block interaction gives us the correct result. The purpose of this section is to underscore these arguments and explain that the contributions of nonplanar diagrams in noncommutative gauge theory indeed reproduce the block-block interactions. We also provide a more accurate estimate of the crossover scale.
For this purpose, we evaluate the leading term of the effective action by using the plane-wave basis $`exp(ik\widehat{x})`$ just like the preceding section.
$`W`$ $`=`$ $`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k}{}}Tr\left(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}_{\nu \lambda }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\rho \mu }exp(ik\widehat{x})\right)`$ (5.27)
$`{\displaystyle \frac{2}{n}}{\displaystyle \underset{k}{}}Tr\left(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\mu \rho }{\displaystyle \frac{1}{P^2}}F_{\lambda \nu }exp(ik\widehat{x})\right)`$
$`+{\displaystyle \frac{1}{2n}}{\displaystyle \underset{k}{}}Tr\left(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }exp(ik\widehat{x})\right)`$
$`+{\displaystyle \frac{1}{4n}}{\displaystyle \underset{k}{}}Tr\left(exp(ik\widehat{x}){\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }{\displaystyle \frac{1}{P^2}}F_{\mu \nu }{\displaystyle \frac{1}{P^2}}F_{\lambda \rho }exp(ik\widehat{x})\right).`$
The above expression is calculated as follows:
$`W`$ $`=`$ $`n({\displaystyle \frac{1}{2\pi B}})^2{\displaystyle d^4k(\frac{1}{k^2+(Bb)^2})^4exp(iC^{\mu \nu }k_\mu l_\nu )}`$ (5.28)
$`{\displaystyle \underset{p,p^{},q}{}}{\displaystyle \frac{3}{2}}(8tr(f_{\mu \nu }^{(i)}(p)f_{\nu \sigma }^{(i)}(p^{}))tr(f_{\mu \rho }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))+tr(f_{\mu \nu }^{(i)}(p)f_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))`$
$`+tr(f_{\mu \nu }^{(i)}(p)\stackrel{~}{f}_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)\stackrel{~}{f}_{\rho \sigma }^{(j)}(q^{})))`$
where $`l=p+p^{}=q+q^{}`$. Here we have assumed that the external momenta are small compared to the noncommutativity scale. Hence we have dropped the phase which only depends on the external momenta. Note that eq.(5.28) contains the phase factor $`exp(iC^{\mu \nu }k_\mu l_\nu )`$ due to noncommutativity.
This phase becomes non-trivial when $`|k|1/|l|l_{NC}^2`$. We can choose $`1/|l|\rho `$ in this case. Note that we have another scale $`|k|b/l_{NC}^2`$ which is associated with the propagator. When $`b>\rho >>l_{NC}`$, we show that eq.(5.28) can be simply understood in terms of the block-block interactions. It is evaluated as follows:
$`W`$ $`=`$ $`n({\displaystyle \frac{1}{2\pi B}})^2{\displaystyle \frac{1}{(Bb)^8}}{\displaystyle d^4kexp(iC^{\mu \nu }k_\mu l_\nu )}`$ (5.29)
$`{\displaystyle \underset{p,p^{},q}{}}{\displaystyle \frac{3}{2}}(8tr(f_{\mu \nu }^{(i)}(p)f_{\nu \sigma }^{(i)}(p^{}))tr(f_{\mu \rho }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))+tr(f_{\mu \nu }^{(i)}(p)f_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))`$
$`+tr(f_{\mu \nu }^{(i)}(p)\stackrel{~}{f}_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)\stackrel{~}{f}_{\rho \sigma }^{(j)}(q^{})))`$
$`=`$ $`n^2{\displaystyle \frac{1}{(Bb)^8}}`$
$`\times {\displaystyle \underset{p,q}{}}{\displaystyle \frac{3}{2}}(8tr(f_{\mu \nu }^{(i)}(p)f_{\nu \sigma }^{(i)}(p))tr(f_{\mu \rho }^{(j)}(q)f_{\rho \sigma }^{(j)}(q))+tr(f_{\mu \nu }^{(i)}(p)f_{\mu \nu }^{(i)}(p))tr(f_{\rho \sigma }^{(j)}(q)f_{\rho \sigma }^{(j)}(q))`$
$`+tr(f_{\mu \nu }^{(i)}(p)\stackrel{~}{f}_{\mu \nu }^{(i)}(p))tr(f_{\rho \sigma }^{(j)}(q)\stackrel{~}{f}_{\rho \sigma }^{(j)}(q)))`$
$`=`$ $`{\displaystyle \frac{1}{(Bb)^8}}`$
$`\times {\displaystyle \frac{3}{2}}(8({\displaystyle \frac{1}{2\pi C}})^2{\displaystyle }d^4xtr(f_{\mu \nu }^{(i)}(x)f_{\nu \sigma }^{(i)}(x))({\displaystyle \frac{1}{2\pi C}})^2{\displaystyle }d^4ytr(f_{\mu \rho }^{(j)}(y)f_{\rho \sigma }^{(j)}(y))`$
$`+({\displaystyle \frac{1}{2\pi C}})^2{\displaystyle d^4xtr(f_{\mu \nu }^{(i)}(x)f_{\mu \nu }^{(i)}(x))(\frac{1}{2\pi C})^2d^4ytr(f_{\rho \sigma }^{(j)}(y)f_{\rho \sigma }^{(j)}(y))}`$
$`+({\displaystyle \frac{1}{2\pi C}})^2{\displaystyle }d^4xtr(f_{\mu \nu }^{(i)}(x)\stackrel{~}{f}_{\mu \nu }^{(i)}(x))({\displaystyle \frac{1}{2\pi C}})^2{\displaystyle }d^4ytr(f_{\rho \sigma }^{(j)}(y)\stackrel{~}{f}_{\rho \sigma }^{(j)}(y))).`$
Using the large instanton size approximation eq.(5.23), we indeed reproduce the result of eq.(5.24).
On the other hand, we can neglect the phase factor $`exp(iC^{\mu \nu }k_\mu l_\nu )`$ when $`b<\rho `$. With this approximation, we obtain:
$`W`$ $`=`$ $`n({\displaystyle \frac{1}{2\pi B}})^2{\displaystyle d^4k(\frac{1}{k^2+(Bb)^2})^4}`$ (5.30)
$`{\displaystyle \underset{p,p^{},q}{}}{\displaystyle \frac{3}{2}}(8tr(f_{\mu \nu }^{(i)}(p)f_{\nu \sigma }^{(i)}(p^{}))tr(f_{\mu \rho }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))+tr(f_{\mu \nu }^{(i)}(p)f_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)f_{\rho \sigma }^{(j)}(q^{}))`$
$`+tr(f_{\mu \nu }^{(i)}(p)\stackrel{~}{f}_{\mu \nu }^{(i)}(p^{}))tr(f_{\rho \sigma }^{(j)}(q)\stackrel{~}{f}_{\rho \sigma }^{(j)}(q^{})))`$
$`=`$ $`{\displaystyle \frac{C^4}{2(4\pi )^2b^4}}{\displaystyle d^4xb_8}.`$
In this way, we reproduce the ordinary gauge theory result eq.(5.26).
## 6 Conclusions
In this paper we have proposed a bi-local field representation of noncommutative field theories. In the momentum eigenstate representation, the momenta of fields can become much larger than the noncommutative scale $`l_{NC}`$. However we can no longer regard these large momentum degrees of freedom as ordinary local fields since the wave functions become noncommutative. Due to the noncommutativity of space-time, the wave functions with large momenta are more naturally interpreted as translation operators and hence those degrees of freedom are interpreted as bi-local fields. In the $`n\times n`$ matrix reguralization of noncommutative field theories, there are $`n^2`$ degrees of freedom and this number is much larger than $`n`$ which is the degrees of freedom of ordinary local field with UV cutoff $`l_{NC}^1`$. This implies that noncommutative field theories contain infinitely many particles. It is interesting to investigate the possibility that these infinite degrees of freedom may pile up to strings. <sup>5</sup><sup>5</sup>5see also
We have also calculated long range interactions in noncommutative field theories. In the conventional approach, they manifest as IR singular behaviors of nonplanar diagrams. We have seen that this feature can be naturally understood in terms of block-block interactions in the matrix model picture. This type of interactions are characteristic to matrix models, not restricted to twisted reduced models which are directly related to noncommutative field theories. In the case of type IIB matrix model, we have shown the appearance of block-block interactions in generic backgrounds and interpreted them as propagations of massless supergravity multiplets. These interactions differ from those of the ordinary gauge theory such as $`D=4`$ $`𝒩=4`$ super Yang-Mills field theory as is explained in the preceding section. This point demonstrate the richness of matrix models over the corresponding field theories and it is one of advantages to consider IIB matrix model as a constructive formulation of superstring.
Acknowledgments
This work is supported in part by the Grant-in-Aid for Scientific Research from the Ministry of Education, Science and Culture of Japan. We would like to thank M. Hayakawa and T. Tada for discussions. |
warning/0001/math-ph0001035.html | ar5iv | text | # Constructive Fractional-Moment Criteria for Localization in Random Operators
## 1 Introduction
In the study of Anderson localization the analogy with the statistical mechanics of spin systems has often served as a source of insight . In this note we report on some recent results which were motivated in part by this analogy, and in part by the desire to develop elementary methods for the study of different aspects of the localization phenomena. Our goal is to present only an outline – the detailed statements and proofs are given in ref. .
The subject of our discussion are random operators acting on the Hilbert space $`\mathrm{}^2(\mathrm{}^d)`$ of square summable functions defined over the regular $`d`$-dimensional lattice $`\mathrm{}^d`$. A prototypical example is the discrete Schrödinger operator:
$$H_\omega =\mathrm{\Delta }+V_\omega (x),$$
(1)
where $`V_\omega (x)`$ is a random potential, and $`\mathrm{\Delta }`$ is the nearest neighbor difference operator (“discrete Laplacian”). The operator may be augmented by the addition of a magnetic field, and/or random off-diagonal hopping terms. However, for our results we assume translation invariance at least in the stochastic sense and up to gauge transformations (i.e., magnetic shifts). (The statements may also be adapted to periodic and quasi-periodic structures.) It is now well understood that such operators have energy regimes in which the spectrum consists of an infinite collection of eigenvalues associated to exponentially localized eigenfunctions. It is also expected, although the theory is sorely lacking, that in certain situations such operators also possess energy regions associated to extended states.
A very useful tool is provided by the Green function of $`H_\omega `$:
$$G_\omega (x,y;E+\mathrm{i}\eta ):=<x|\frac{1}{H_\omega E\mathrm{i}\eta }|y>.$$
(2)
To illustrate the relation with spin systems one may note the analogy between $`G_\omega `$ and the spin-spin correlation functions suggested by the functional integral expression for the former as a Gaussian integral:
$$G_\omega (x,y;E+\mathrm{i}\eta )=(\mathrm{i})\frac{[𝒟\mathrm{\Phi }]\mathrm{}^{\mathrm{i}\mathrm{\Phi },(H_\omega E\mathrm{i}\eta )\mathrm{\Phi }}\mathrm{\Phi }(x)\mathrm{\Phi }(y)}{[𝒟\mathrm{\Phi }]\mathrm{}^{\mathrm{i}\mathrm{\Phi },(H_\omega E\mathrm{i}\eta )\mathrm{\Phi }}}.$$
(3)
The relevant rigorous methods familiar from the study of spin systems do not apply here since the integral involves a complex action. Nevertheless, exponential bounds for $`G_\omega `$ in the limit $`\eta 0`$ are a “signature” of the localized regime, much as exponential decay of spin-spin correlations indicates the high-temperature regime.
Since the Green function depends on the disorder, it is tempting to consider the averaged function $`𝔼\left(|G_\omega (x,y;E+\mathrm{i}\eta )|\right)`$, where $`𝔼()`$ indicates the expectation with respect to the potential. However, for $`E`$ in the spectrum of $`H_\omega `$, this quantity may diverge as $`\eta 0`$. As was realized in ref. , we can avoid this problem by considering a “fractional moment” of the Green function: $`𝔼\left(|G_\omega (x,y;E+\mathrm{i}\eta )|^s\right)`$ for $`s<1`$.
Thus, a technically convenient “signature” of localization is the exponential decay of such fractional moments, at some suitable $`s(0,1)`$,
$$𝔼(|<x|\frac{1}{H_\omega E\mathrm{i}\eta }|y>|^s)A(s)\mathrm{}^{\mu (s)|xy|},$$
(4)
where the bound is satisfied uniformly in $`\eta \mathrm{}`$ for all energies in some range $`E(a,b)`$.
Before we turn to the new results, which offer constructive criteria for the validity of the fractional moment condition (4), let us list several known implications of this condition, each of which also presumes some mild regularity conditions on the distribution of the random potential:
* Spectral localization ( \- using ): The spectrum of $`H_\omega `$ within the interval $`(a,b)`$ is almost-surely of the pure-point type, and the corresponding eigenfunctions are exponentially localized.
* Dynamical localization (): Wave packets with energies in the specified range do not spread (and in particular the SULE condition of is met):
$$𝔼\left(\underset{t\mathrm{}}{sup}|<x|\mathrm{}^{\mathrm{i}tH}P_{H(a,b)}|y>|\right)\stackrel{~}{A}\mathrm{}^{\stackrel{~}{\mu }|xy|}.$$
(5)
* Absence of level repulsion (). Minami has shown that (4) implies that in the range $`(a,b)`$ the energy gaps have Poisson-type statistics.
* Exponential decay of the projection kernel ():
$$𝔼(|<x|P_{HE}|y>|)\widehat{A}\mathrm{}^{\widehat{\mu }|xy|}.$$
(6)
This condition plays an important role in the quantization of Hall conductance in the ground state of the two–dimensional electron gas with Fermi level $`E_F(a,b)`$ .
The fractional moment condition (4) has already been established for certain regimes: large disorder and extreme energies , and also at weak disorder but for energies far from the spectrum of $`\mathrm{\Delta }`$ . Our new results extend the reach of the fractional moment method by showing that the entire region in which (4) holds can, in principle, be determined by a sequence of finite calculations.
In their general appearance, these results may remind one of some of the constructive criteria for the high temperature phases in certain models of statistical mechanics, which are mentioned below. As in that case, the results also yield some conclusions about the critical behavior, which in the present context refers to the behaviour in the vicinity of the mobility edge – wherever such an edge occurs.
Before the introduction of the fractional-moment method, localization regimes have been established using the multiscale analysis of Fröhlich and Spencer which yields exponential bounds of the form:
$$|G_\omega (x,y;E)|A(\omega ,x)\mathrm{}^{\mu (\omega )|xy|},$$
(7)
with $`\mu (\omega )>0`$ and $`A(\omega ,x)<\mathrm{}`$ for almost every $`\omega `$ and all $`x\mathrm{}^d`$. The bounds which the multiscale analysis provides for the probability of the exceptional cases decay faster than any power of $`|xy|`$ but not exponentially fast. It is difficult to use such results to demonstrate exponential bounds on expectation values such as those seen in equations (6) and (5). Nevertheless we note that dynamical localization was recently established also by arguments starting from the bounds provided by the multiscale analysis , using methods not related to this work. We shall return to the relation between the fractional-moment method and the multiscale analysis in the final section of this note.
## 2 The finite-volume criteria
The results presented herein describe certain conditions which when satisfied by the operator $`H_{\mathrm{\Lambda };\omega }`$ obtained by restricting $`H_\omega `$ to some finite volume $`\mathrm{\Lambda }`$ are sufficient to deduce the fractional moment condition (4) for the full operator $`H_\omega `$. For simplicity we state these results only in the case of random Schrödinger operators. The reader is directed to for versions which apply to more general operators.
To guarantee that the fractional moments are finite, we require certain regularity of the joint probability distribution of the site potentials $`V(x)`$. An additional technical assumption related to the “decoupling lemmas” used in is also required. In their mildest form the conditions required are somewhat technical to state, so for the present note we shall call a probabilty distribution of the potential regular if the site values $`V(x)`$ are independent identically distributed random variables whose distribution has a bounded density with compact support. The interested reader may find the more general assumptions in .
In order to state our results, we must introduce some notation. Given a finite region $`\mathrm{\Lambda }\mathrm{}^d`$, we denote by $`\mathrm{\Gamma }(\mathrm{\Lambda })`$ the set of lattice bonds (nearest neighbor pairs) connecting sites in $`\mathrm{\Lambda }`$ with sites in $`\mathrm{}^d\mathrm{\Lambda }`$, and by $`\mathrm{\Lambda }^+`$ the region obtained from $`\mathrm{\Lambda }`$ by adding to it all of its nearest neighbors. The number of elements of a set $`W`$ is denoted $`|W|`$.
As mentioned above, $`𝔼()`$ indicates the expectation with respect to the random potential. We also let
$$𝔼_{+\mathrm{i0}(\mathrm{i0})}(G(E)):=\underset{\stackrel{\eta 0}{(\eta 0)}}{lim}𝔼(G(E+\mathrm{i}\eta )).$$
(8)
Following is the first of our results.
###### Theorem 1
Let $`H_\omega `$ be a random Schrödinger operator with a regular distribution of the random potential. Then for each $`s<1`$ there exists $`C_s<\mathrm{}`$ such that if for some $`E\mathrm{}`$ (in fact also $`E\mathrm{}`$) and some finite region $`\mathrm{\Lambda }\mathrm{}^d`$ which contains the origin $`O`$:
$$\left(1+\frac{C_s}{\lambda ^s}|\mathrm{\Gamma }(\mathrm{\Lambda })|\right)^2\underset{<u,u^{}>\mathrm{\Gamma }(\mathrm{\Lambda })}{}𝔼\left(|<O|\frac{1}{H_{\mathrm{\Lambda };\omega }E}|u>|^s\right)<1,$$
(9)
then $`H_\omega `$ satisfies the fractional-moment condition (4), and there exist $`\mu (s)>0,A(s)<\mathrm{}`$, which depend on $`E`$ only through the value of the LHS in eq. (9), so that for any region $`\mathrm{\Omega }\mathrm{}^d`$,
$$𝔼_{\pm \mathrm{i0}}\left(|<x|\frac{1}{H_{\mathrm{\Omega };\omega }E}|y>|^s\right)A(s)\mathrm{}^{\mu (s)\mathrm{dist}_\mathrm{\Omega }(x,y)},$$
(10)
with
$$\mathrm{dist}_\mathrm{\Omega }(x,y)=\mathrm{min}\{|xy|,[\mathrm{dist}(x,\mathrm{\Omega })+\mathrm{dist}(y,\mathrm{\Omega })]\}.$$
(11)
The modified metric, $`\mathrm{dist}_\mathrm{\Omega }(x,y)`$, is a distance function relative to which the entire boundary of $`\mathrm{\Omega }`$ is regarded as one point. It permits us to state that there is exponential decay in the “bulk” without ruling out the possible existence of extended boundary states in some geometry.
One may also formulate finite-volume criteria which rule out extended boundary states, that is which permit us to conclude exponential decay of the fractional moments in any region $`\mathrm{\Omega }`$. The trade-off is that the finite volume test, presented in the next result, is a bit more involved.
###### Theorem 2
Let $`H_\omega `$ be a random Schrödinger operator with a regular distribution of the random potential. Then for each $`s<1`$ there exists $`\stackrel{~}{C}_s>0`$ such that if for some $`E\mathrm{}`$ (alternatively, complex $`E`$) and some finite region $`O\mathrm{\Lambda }\mathrm{}^d`$:
$$\underset{W\mathrm{\Lambda }}{\mathrm{max}}\left\{|\mathrm{\Gamma }(\mathrm{\Lambda }^+)|\frac{\stackrel{~}{C}_s}{\lambda ^s}\underset{<u,u^{}>\mathrm{\Gamma }(\mathrm{\Lambda })}{}𝔼\left(|<O|\frac{1}{H_{W;\omega }E}|u>|^s\right)\right\}<1,$$
(12)
then there are $`\mu (s)>0`$ and $`A(s)<\mathrm{}`$ — which depend on the energy $`E`$ only through the value of the LHS of eq. (12) — such that for any region $`\mathrm{\Omega }\mathrm{}^d`$
$$𝔼_{\pm \mathrm{i0}}\left(|<x|\frac{1}{H_{\mathrm{\Omega };\omega }z}|y>|^s\right)A(s)\mathrm{}^{\mu (s)|xy|}.$$
(13)
It is rather obvious that the collection of finite-volume criteria provided in Theorem 2 covers the entire regime in which the conclusion, eq. (13), holds. The corresponding statement for Theorem 1 is a bit less immediate, but it is also true:
###### Theorem 3
Let $`H_\omega `$ be a random Schrödinger operator with a regular distribution of the random potential, and fix $`s<1`$. If at some energy $`E`$ (or $`E\mathrm{}`$) the localization condition (4) is satisfied, with some $`A<\mathrm{}`$ and $`\mu >0`$, then for all large enough (but finite) $`L`$ the condition (9) is met for $`\mathrm{\Lambda }=[L,L]^d`$.
## 3 Some Implications
We shall now mention a number of implications of the finite-volume criteria for fractional moment localization.
First, of course, are explicit bounds, and we already obtain such bounds with a single site estimate corresponding to $`\mathrm{\Lambda }=\{0\}`$. The test provided by Theorem 2 is met for all $`\lambda `$ and $`E`$ such that:
$$\frac{2d^2(2d+1)C_s}{\lambda ^s}𝔼(\frac{1}{|\lambda VE|^s})<1.$$
(14)
This implies localization for strong disorder, and at extremal energies, in the manner of ref. .
The above explicit criterion may now be systematically improved. However, since the calculations quickly become quite laborious, perhaps the main benefit are certain qualitative statements. Those bear some resemblance to results derived using the multiscale approach; however the conclusions drawn here go beyond the latter by yielding results on the exponential decay of the mean values.
### 3.1 Fast power decay $``$ exponential decay
An interesting and useful implication (as seen below) is that fast enough power law implies exponential decay. In this sense, random Schrödinger operators join other statistical mechanical models in which such principles have been previously recognized. The list includes the general Dobrushin-Shlosman results and the more specific two-point function bounds for: percolation , Ising ferromagnets , certain $`O(N)`$ models , and time-evolution models .
###### Theorem 4
Let $`H_\omega `$ be a random Schrödinger operator on $`\mathrm{}^2(\mathrm{}^d)`$ with a regular potential. Then there are $`L_o,B_1,B_2<\mathrm{}`$ such that if for some $`E\mathrm{}`$ and some finite $`LL_o`$, either
$$\underset{L/2yL}{sup}𝔼\left(|<O|\frac{1}{H_{\mathrm{\Lambda }_L,\omega }E}|y>|^s\right)B_1/L^{3(d1)},$$
(15)
or
$$\underset{L/2yL}{sup}𝔼\left(|<O|\frac{1}{H_\omega E}|y>|^s\right)B_2/L^{4(d1)},$$
(16)
where $`\mathrm{\Lambda }_L=[L,L]^d`$ and $`y_j|y_j|`$, then the exponential localization (4) holds for all energies in some open interval $`(a,b)`$ containing $`E`$.
### 3.2 Lower bounds for $`G_\omega (x,y;E_{\mathrm{edge}}+i0)`$ at mobility edges
Boundary points of the continuous spectrum are referred to as mobility edges. The random Schrödinger operators considered here are ergodic, hence the location of such points does not depend on the realization . However, except for the Bethe lattice , the proof of the occurrence of continuous spectrum is still an open problem. Nevertheless, it is interesting to note that Theorem 4 yields the following pair of lower bounds on the decay rate of the Green function at mobility edges, $`E_{\mathrm{edge}}`$, for a random Schrödinger operator with regular potential:
$$\underset{L/2yL}{sup}𝔼\left(|<O|\frac{1}{H_{[L,L]^d;\omega }E_{\mathrm{edge}}}|y>|^s\right)B_1L^{3(d1)},$$
(17)
and
$$\underset{L/2yL}{sup}𝔼\left(|<O|\frac{1}{H_\omega E_{\mathrm{edge}}}|y>|^s\right)B_2L^{4(d1)}.$$
(18)
We do not expect these bounds to be optimal.
Vaguely similar bounds are known for the critical two-point functions in the statistical mechanical models mentioned above.
### 3.3 Extending off the real axis
The following statement is of somewhat technical interest, but it has interesting implications, such as the decay of the projection kernel, for which it is useful to have bounds on the resolvent at $`E+i\eta `$ which are uniform in $`\eta `$. Such bounds permit integrating the resolvent estimates along contours which cut the real axis, as in the derivation of (eq. (6)) in ref. .
###### Theorem 5
Let $`H_\omega `$ be a random Schrödinger operator with a regular potential. Suppose that for some $`E\mathrm{}`$, and $`\mathrm{\Delta }E>0`$, the following bound holds uniformly for $`\xi [E\mathrm{\Delta }E,E+\mathrm{\Delta }E]`$:
$$𝔼\left(|<x|\frac{1}{H_\omega \xi i0}|y>|^s\right)Ae^{\mu |xy|}.$$
(19)
Then for all $`\eta \mathrm{}`$:
$$𝔼\left(|<x|\frac{1}{H_\omega Ei\eta }|y>|^s\right)\stackrel{~}{A}e^{\stackrel{~}{\mu }|xy|},$$
(20)
with some $`\stackrel{~}{A}<\mathrm{}`$ and $`\stackrel{~}{\mu }>0`$ – which depend on $`\mathrm{\Delta }E`$ and the bound (19).
### 3.4 Localization in spectral tails.
The finite volume criteria presented above allow us to conclude exponential localization from suitable bounds on the density of states of the operators in regions $`\mathrm{\Lambda }_L=[L,L]^d`$. The following statement will be useful for such a purpose.
###### Theorem 6
Let $`H_\omega `$ be a random Schrödinger operator on $`\mathrm{}^2(\mathrm{}^d)`$ with a regular potential. For each $`L>0`$ there exist $`\delta _L>0`$ and $`P_L>0`$ such that if
$$\mathrm{Prob}\left[\mathrm{dist}(\sigma (H_{\mathrm{\Lambda }_L;\omega }),E)\delta _L\right]<P_L,$$
(21)
then the exponential localization condition (4) holds in some open interval containing $`E`$. Furthermore, given $`\beta (0,1)`$ and $`\xi >3(d1)`$, it is possible to choose $`\delta _L`$ and $`p_L`$ such that $`\underset{¯}{\mathrm{lim}}L^\beta \delta _L>0`$ and $`\underset{¯}{\mathrm{lim}}L^\xi P_L>0`$.
Remarks: 1. It is of interest to combine the criterion presented above with Lifschitz tail estimates on the density of states at the bottom of the spectrum and at band edges. As an example, consider the bottom of the spectrum of $`H_\omega `$: $`E_0=\lambda V_0`$ where $`V_0`$ is the minimum value in the support of $`V`$. Using Lifschitz tail estimates, it is possible to show that :
$$\mathrm{Prob}\left[inf\sigma (H_{\mathrm{\Lambda }_L;\omega })E_0+\mathrm{\Delta }E\right]\mathrm{Const}.L^de^{\mathrm{\Delta }E^{d/2}}.$$
(22)
By choosing $`\mathrm{\Delta }EL^\beta `$ with $`\beta (0,1)`$ for large enough $`L`$, we infer fractional moment localization from this bound via Theorem 6. Previous results in this vein may be found in .
2. The input conditions (21) are similar to the input used in the multiscale analysis. In fact, there it is not important that $`\xi >3(d1)`$, and it suffices to assume the condition is met for some $`\xi >0`$. However, one may note that wherever the multiscale analysis applies, its conclusion allows to deduce the condition as stated here. Thus, the exponential localization in the stronger sense discussed in our work may be concluded also for the regime for which localization may be established through the multiscale analysis. |
warning/0001/gr-qc0001087.html | ar5iv | text | # Exotic spacetimes, superconducting strings with linear momentum, and (not quite) all that
## I Introduction
The possibility that the spacetime around a superconducting cosmic string with constant momentum is endowed with exotic properties has been indicated in a recent article by Thatcher and Morgan . In accordance with the analysis carried out in , in the resulting string spacetime, test particles are deflected as they approach the string, effectively isolating the defect from the outside universe. This, and other strange properties of the metric, such as the possibility of causal violations, would imply that the inclusion of gravity into models of charged strings and vortons may have significant consequences as regards their possible role in a cosmological context.
Because of the very peculiar nature of the results obtained in , and since the analysis is carried out, at least in part, through approximations or numerical integration of some complex system of equations, it seems appropriate to try, as a check, to reobtain those results through a different approach, and also to extend the analysis of the properties of the associated metrics, as regards their causal and other properties. Therefore, in this paper we start, in Section II, with a derivation of the general form of the vacuum metrics external to an infinite, non rotating, axisymmetric stationary cylinder. Depending on the choice of the integration constants that appear in solving Einstein’s equations, these turn out to be members of either two families of exact solutions of the form of the Lewis metrics, one of which can be interpreted as a “boosted” Levi-Civita metric (static metrics), and the other one is similar (but not quite equal) to the “exotic” metrics found in (stationary but not static metrics).
The “boosted” Levi-Civita metrics can be brought to a standard diagonal form by a coordinate transformation, so the spacetime they describe has ordinary properties. In the case of the “exotic” metrics this is not possible, and we find that non-spacelike geodesics are always trapped, i.e., their radial coordinate cannot take arbitrarily large (or small) values. Moreover, we show that although it is possible to assign a well defined time orientation, so that we may distinguish between future and past directed causal curves at any point, there are causal curves through every point in these spacetimes, whose extensions to the past and to the future eventually self-intersect, and that it is possible to construct non-spacelike curves joining any two points in that spacetime. In particular, there are closed timelike curves connecting any pair of points in the spacetime. Since the metrics are singular on the symmetry axis, we might expect that the singularity, as in the case of some of the Levi-Civita metrics, can be replaced by a cylindrical source, satisfying regularity and other physical requirements. We would therefore have a physically admissible source for an “exotic” spacetime. That such a non causal behavior results from a physical source cannot be dismissed in principle, since it is well known that closed timelike curves appear in spacetimes containing cosmic strings in relative motion. Unfortunately, as we show in Section III, the construction given in does not give a positive answer to the question of the existence of such physical sources for the exotic metric. As we indicate in that Section, the ansatz given in can be solved, once a simple coordinate transformation is introduced, by a diagonal metric, and therefore there is no exotic behavior. Basically, since the momentum is constant, one can always choose a new coordinate frame in which it is zero.
In Section IV we extend the results of , relating properties of the metric coefficients to those of the source, to the non diagonal case, the main result being that a source for the “exotic” metrics must violate the dominant energy condition (DEC) under quite general conditions. In Section V we exemplify these results by considering the possibility of having an “exotic” metric outside one or more concentric cylindrically symmetric shells with regular axis, and showing explicitly that this requires that the DEC is violated by the matter making up the shells.
We use geometrized units ($`G=c=1`$), the signature of the metric is $`(++++)`$, and (sometimes) we use abstract indexes.
## II Cylindrically symmetric, non rotating, stationary vacuum metrics
We consider first the vacuum spacetime external to a non rotating, cylindrically symmetric distribution of mass-energy, in a stationary state of motion along the symmetry axis. This may be described, in general, by a metric of the form
$$ds^2=dr^2+g_{\theta \theta }d\theta ^2+g_{zz}dz^2+2g_{tz}dzdtg_{tt}dt^2,$$
(1)
where $`g_{\theta \theta },g_{zz},g_{tt},g_{tz}`$ are functions of $`r`$. The general solution of the vacuum Einstein equations corresponding to (1) may be written in the form
$`g_{\theta \theta }`$ $`=`$ $`C_5(rr_0)^{2k1},`$ (2)
$`g_{zz}`$ $`=`$ $`(C_1)^2|rr_0|^{2k2}(C_3)^2|rr_0|^{2k3},`$ (3)
$`g_{tt}`$ $`=`$ $`(C_2)^2|rr_0|^{2k2}+(C_4)^2|rr_0|^{2k3},`$ (4)
$`g_{tz}`$ $`=`$ $`C_1C_2|rr_0|r^{2k2}C_3C_4|rr_0|^{2k3},`$ (5)
where $`r_0`$ is a real constant, and the constants $`k_i`$ ($`i=1\mathrm{}3`$) must satisfy
$$k_1+k_2+k_3=1,(k_1)^2+(k_2)^2+(k_3)^2=1.$$
(6)
We must also choose the constants $`C_i`$, such that
$$C_1C_4C_2C_3,$$
(7)
otherwise the metric would be degenerate. We shall assume that $`\theta `$ is restricted to $`0\theta 2\pi `$, with $`\theta =0`$ and $`\theta =2\pi `$ identified, consistent with the interpretation of cylindrical symmetry. In this context $`r`$ and $`z`$ would, in principle, be the remaining “cylindrical coordinates”, but, as we show below, this needs closer examination in general.
For any choice of real $`k_i`$, $`C_i`$ satisfying equations (6) and (7), we may parameterize the constants $`k_i`$ by a real parameter $`\mathrm{\Delta }`$,
$$k_1=\frac{2(\mathrm{\Delta }1)}{(\mathrm{\Delta }^2+3)},k_2=\frac{2(\mathrm{\Delta }+1)}{(\mathrm{\Delta }^2+3)},k_3=\frac{(\mathrm{\Delta }^21)}{(\mathrm{\Delta }^2+3)},$$
and introduce a linear transformation of the coordinates $`(z,t)`$ of the form
$$zC_1z+C_2t,tC_3z+C_4t,$$
(8)
that puts the metric (1) in the usual diagonal Levi-Civita form,
$$ds^2=dr^2+p_1(arr_0)^{4(\mathrm{\Delta }1)/(\mathrm{\Delta }^2+3)}d\theta ^2+p_2(arr_0)^{4(\mathrm{\Delta }+1)/(\mathrm{\Delta }^2+3)}dz^2p_3(arr_0)^{2(\mathrm{\Delta }^21)/(\mathrm{\Delta }^2+3)}dt^2,$$
with $`p_1`$, $`p_2`$, and $`p_3`$ arbitrary real and positive constants. Therefore, in this case the metric given by (2,3,4,5) is just a “boosted” form of the Levi-Civita metric.
We notice, however, that the Einstein equations are also satisfied if we choose $`k_1`$ real, and $`(k_2,k_3)`$ as a complex conjugate pair, such that Eqs. (6) are satisfied. We must also choose
$$C_3=i(C_1)^{},C_4=i(C_2)^{},$$
where a star indicates complex conjugation, so that the resulting metric is real. After some appropriate renaming of constants, the metric can be written in the form (1), with the coefficients given by
$`g_{\theta \theta }`$ $`=`$ $`c_1|rr_0|^{2q_1},`$ (9)
$`g_{zz}`$ $`=`$ $`c_2^2|rr_0|^{2q_2}\mathrm{cos}[2k\mathrm{ln}(|rr_0|)+2\varphi _1)],`$ (10)
$`g_{tt}`$ $`=`$ $`c_3^2|rr_0|^{2q_2}\mathrm{cos}\left[2k\mathrm{ln}(|rr_0|)+2\varphi _2\right],`$ (11)
$`g_{tz}`$ $`=`$ $`c_2c_3|rr_0|^{2q_2}\mathrm{cos}\left[2k\mathrm{ln}(|rr_0|)+\varphi _1+\varphi _2\right],`$ (12)
where $`\alpha `$ , $`k`$, $`c_1`$, $`c_2`$, $`c_3`$, $`\varphi _1`$, and $`\varphi _2`$ are arbitrary real constants, and
$$q_1=\frac{1}{3}[1+2s(1+3k^2)^{1/2}],q_2=\frac{1}{3}[1s(1+3k^2)^{1/2}],s=\pm 1.$$
The determinant of the metric is
$$c_1c_2c_3(rr_0)^2\left[\mathrm{cos}(2\varphi _12\varphi _2)1\right],$$
and, thus, the metric is nondegenerate if and only if (for $`rr_0`$)
$$c_1,c_2,c_30\text{and}\varphi _1\varphi _2n\pi (nZ).$$
(13)
It is static only when $`k=0`$, corresponding to a Levi-Civita spacetime with $`\mathrm{\Delta }=\pm 1`$ or $`\pm 3`$ if $`s=1`$ or $`s1`$, respectively (Levi-Civita metrics with opposite values of $`\mathrm{\Delta }`$ are isometric).
This type of metric (satisfying Eq. (13), and with $`k0`$), which, following , we shall call “exotic” hereafter, is of Lewis type , and has been usually analyzed in relation with rotating cylinders (i.e., the non vanishing crossed coefficient of the metric is $`g_{t\theta }`$) . The procedure here used to obtain these solutions is similar to that used to obtain the “windmill solutions” of Macintosh .
### A Some properties of the exotic metrics
To justify the name “exotic metric” we consider here (see also ) several classes of non spacelike curves, both geodesics and non geodesics. We first notice that by appropriate rescalings and linear transformations of coordinates we may write the metric in the form
$$ds^2=dr^2+r^{2q_1}d\theta ^2+r^{2q_2}\left[\mathrm{cos}(2k\mathrm{ln}r)dz^2\mathrm{cos}(2k\mathrm{ln}r)dt^2+2\mathrm{sin}(2k\mathrm{ln}r)dtdz\right],$$
(14)
and, therefore, the constants $`c_i`$ and $`\varphi _i`$ are irrelevant, as far as the geometrical properties of the metric are concerned. Note, from Eq. (14), that the sign of $`k`$ can be freely chosen; we will take advantage of this freedom later in Sections IV and V.
As already noticed in , for a metric of the form (14), the Killing vectors $`_t`$ and $`_z`$ change from spacelike to timelike and vice versa as one moves along the $`r`$ coordinate (for fixed $`\theta `$). In other words, if, for fixed $`\theta `$, we consider a fixed coordinate grid, where $`(r,t,z)`$ are taken as cartesian coordinates, the local light cones appear to “rotate” along the $`r`$-axis, while they have fixed directions in a given $`(t,z)`$ “plane”, (i.e. a constant $`(r,\theta )`$ surface). Nevertheless, the form (14) for the metric immediately shows that, since the light cones are well defined everywhere (except, of course, for $`r=0`$), we may, in spite of this “rotation” of the light cones, impose a definite time orientation on the spacetime, by simply defining a future direction at a given point, and then extending this definition by continuity to all other points.
A definite time orientation on the spacetime, however, does not preclude the possibility of the existence of closed causal (i.e. non spacelike) curves. Some peculiar behavior regarding null curves was already noticed in . We therefore start our analysis by considering geodesic curves. Cylindrical symmetry implies that there is a subfamily of geodesics with constant $`\theta `$. Restricting to this type of geodesics, we may write their tangent vector in the form
$$u^a=u^z(_z)^a+u^t(_t)^a+u^r(_r)^a.$$
Since $`(_t)^a`$ and $`(_z)^a`$ are Killing vectors, and $`u^a`$ is geodesic, we have
$$u_a(_z)^a=p,u_a(_t)^a=E,$$
where $`E`$ and $`p`$ are constants. From these relations, we find
$`u^z`$ $`=`$ $`\dot{z}=(E^2+p^2)^{1/2}r^{2q_2}\mathrm{cos}(2k\mathrm{ln}ra),`$ (15)
$`u^t`$ $`=`$ $`\dot{t}=(E^2+p^2)^{1/2}r^{2q_2}\mathrm{sin}(2k\mathrm{ln}ra),`$ (16)
$`(u^r)^2`$ $`=`$ $`\dot{r}^2=U(r),U(r)=s(E^2+p^2)r^{2q_2}\mathrm{cos}(2k\mathrm{ln}r2a+\pi ),`$ (17)
where $`a=\mathrm{arctan}(E/p)`$, $`s=0`$ ($`s=1`$) for null (timelike) geodesics, and a dot indicates derivation with respect to the affine parameter, e.g. $`\dot{z}=dz/d\tau `$.
For null geodesics, we may rescale $`\tau `$ to set $`(E^2+p^2)^{1/2}=1`$, and therefore, without loss of generality we may then write
$$\frac{dr}{d\tau }=\pm r^{q_2/2}\left[\mathrm{cos}(2k\mathrm{ln}r2a+\pi )\right]^{1/2},$$
(18)
where the plus/minus sign corresponds, respectively, to “outgoing” and “ingoing” geodesics. Notice that this implies that all null geodesics have “turning points” in $`r`$ (points where $`\mathrm{cos}(2k\mathrm{ln}r2a+\pi )=0`$), and therefore, the corresponding values of $`r`$ are restricted to a finite segment of the $`r`$-axis, where $`\mathrm{cos}(2k\mathrm{ln}r2a+\pi )0`$. Upon reaching a turning point the sign of $`dr/d\tau `$ is changed, and the sense of the motion along the $`r`$-axis is reversed, but there is no associated singularity in the metric. Turning points are properties of individual null geodesics, i.e., they change when we consider different null geodesics going through the same point, and therefore there are no horizons, or any other peculiar geometrical property associated to the turning points of a given null geodesic.
Since null geodesics going through a certain $`r`$ eventually come back to the same value of $`r`$, it is important to compute the corresponding change in $`t`$ and $`z`$. These may be obtained by eliminating the affine parameter $`\tau `$, and looking essentially at the corresponding trajectories in spacetime. From equations (18) and (15,1617), we have
$`{\displaystyle \frac{dz}{dr}}`$ $`=`$ $`\pm r^{q_2/2}{\displaystyle \frac{\mathrm{cos}(2k\mathrm{ln}ra)}{\sqrt{\mathrm{cos}(2k\mathrm{ln}r2a+\pi )}}},`$ (19)
$`{\displaystyle \frac{dt}{dr}}`$ $`=`$ $`\pm r^{q_2/2}{\displaystyle \frac{\mathrm{sin}(2k\mathrm{ln}ra)}{\sqrt{\mathrm{cos}(2k\mathrm{ln}r2a+\pi )}}},`$ (20)
where the plus/minus sign is determined from Eq. (18). After an integration by parts of the right hand sides of Eqs. (19) and (20), we obtain the following expressions for the changes in $`z`$ and $`t`$, corresponding to a null geodesic that goes from an initial point in $`r=r_0`$ to a turning point at $`r=r_1`$, and comes back to $`r=r_0`$,
$`\mathrm{\Delta }t`$ $`=`$ $`2[t(r_1)t(r_0)]=\pm \left[2\alpha 2k^1r_0h(r_0)\mathrm{cos}a\right],`$
$`\mathrm{\Delta }z`$ $`=`$ $`2[z(r_1)z(r_0)]=\pm \left[2\beta +2k^1r_0h(r_0)\mathrm{sin}a\right],`$
with
$$=_{r_0}^{r_1}h(r)dr,h(r)=r^{q_2/2}\mathrm{cos}^{1/2}(2k\mathrm{ln}r2a+\pi ),$$
and
$$\alpha =\mathrm{sin}a\mu \mathrm{cos}a,\beta =\mathrm{cos}a+\mu \mathrm{sin}a,\mu =\frac{1}{2k}(q_2+2).$$
To see if the end point can be in the causal past of the initial point, using the same arguments as those leading to the form (14) for the metric, we may choose, without loss of generality, $`r_0=1`$, and assume that a future directed vector at $`r=1`$ has $`\dot{t}>0`$. Then, the end point will be in either the causal past or future of the starting point if $`\mathrm{\Delta }l^2=\mathrm{\Delta }t^2+\mathrm{\Delta }z^20`$. But we have
$$\mathrm{\Delta }l^2=\frac{4\dot{r}_0^2}{k^2}\left[\mu k+h(r_0)\right]^2+4^2(3\mathrm{sin}^2a+\mathrm{cos}^2a)8||\dot{t}_0\mathrm{\Delta }t,$$
so that a necessary condition for $`\mathrm{\Delta }l^20`$, is $`\dot{t}_0\mathrm{\Delta }t>0`$, and therefore the end point is causally connected to the starting point only if $`\mathrm{\Delta }t>0`$. We conclude that a null geodesics cannot return to its causal past, and since $`\mathrm{\Delta }t=0`$ implies $`\mathrm{\Delta }l^2>0`$, null geodesics cannot self intersect.
We have seen that null geodesics necessarily have turning points. We may similarly show that timelike geodesics necessarily have turning points as well. Since these are not horizons, we may ask whether turning points necessarily appear also for general (i.e., non geodesic) causal curves. That this is not so can be illustrated with a simple example. Consider a curve parameterized so that its tangent vector satisfies
$$u_r=s,u_z=r^{B\mathrm{cos}b1}\mathrm{cos}[\mathrm{ln}(r^{B\mathrm{sin}b})b],u_t=r^{B\mathrm{cos}b1}\mathrm{sin}[\mathrm{ln}(r^{B\mathrm{sin}b})b],$$
where $`b`$ and $`B`$ are related to $`k`$ and $`q_2`$ in Eq. (14) by
$$s=\pm 1,k=B\mathrm{sin}b0,q_2=22B\mathrm{cos}b.$$
(21)
Without loss of generality we may choose $`B>0`$ and $`\mathrm{sin}b>0`$. We may easily check that, for either choice of $`s`$, this represents a null curve, well defined for all values of $`r`$, where $`r`$ monotonically increases ($`s=+1`$), or decreases ($`s=1`$), without turning points.
We may obtain the trajectory (worldline) corresponding to this curve integrating $`t`$ and $`z`$ as functions of $`r`$. The result is
$`t(r)`$ $`=`$ $`s\left\{r^{B\mathrm{cos}b}\mathrm{cos}\left[\mathrm{ln}(r^{B\mathrm{sin}b})b\right]+C_1\right\}B^1,`$
$`z(r)`$ $`=`$ $`s\left\{r^{B\mathrm{cos}b}\mathrm{sin}\left[\mathrm{ln}(r^{B\mathrm{sin}b})b\right]+C_2\right\}B^1,`$
where $`C_1`$, and $`C_2`$ are integrations constants. Again, with sufficient generality, we may choose these constants such that $`t=z=0`$ for $`r=1`$. This corresponds to $`C_1=\mathrm{cos}b`$, $`C_2=\mathrm{sin}b`$. As indicated, for fixed $`s`$ these curves have no turning points. We may, however, at any point $`r=r_1`$, match a portion of this curve with, say, $`s=+1`$, with a portion of a similar curve with $`s=1`$, the same values of $`B`$ and $`b`$, and where we choose $`C_1`$ and $`C_2`$ in such a way that $`r`$, $`t`$, $`z`$, $`u_t`$, and $`u_z`$ are continuous, and $`u_r`$ changes sign. Since this is still a null curve, for a photon this would correspond to “bouncing off a mirror”, whose normal points in the $`r`$ direction. More geometrically, we may always “round corners”, without changing the non spacelike nature of the curve , so that the sharp change in $`u_r`$ is used only to simplify the computations, but has no special significance.
We may compute now the changes in $`t`$, $`\mathrm{\Delta }t=t(r_1)t(1)`$, and $`z`$, $`\mathrm{\Delta }z=z(r_1)z(1)`$, in the “round trip” from $`r=1`$ to $`r=r_1`$, and back to $`r=1`$, along the resulting null curve. These are simply twice the changes in $`t`$ and $`z`$ in going from $`r=1`$ to $`r=r_1`$, i.e., $`\mathrm{\Delta }t=2[t(r_1)t(1)]`$ and $`\mathrm{\Delta }z=2[z(r_1)z(1)]`$. We define
$$y=\mathrm{ln}(r^{B\mathrm{sin}b})b,x=r^{B\mathrm{cos}b}.$$
We then have three possible cases, depending on the sign of $`\mathrm{cos}(b)`$:
* First, if $`\mathrm{cos}b>0`$, the function $`x(r)`$ is increasing with $`r`$, and, therefore, the equation $`z(r)=0`$ has solutions for sufficiently large $`r`$. Moreover, for sufficiently large $`r`$, these zeros of $`z`$ correspond to $`yn\pi `$, since $`x`$ is unbounded. Then, again for sufficiently large $`r`$, and for the same reason, the zeros of $`z`$ correspond alternatively to positive or negative values of $`t`$. So we may always choose $`r_1`$ such that $`z(r_1)=0`$ and $`t(r_1)<0`$. But then, the return point on $`r=1`$ is in the causal past of the initial point, and we may construct a closed, everywhere future directed, non spacelike curve, by simply joining the end point and the initial point with the timelike, future directed curve whose trajectory corresponds to $`r=1,z=0`$.
* In the case $`\mathrm{cos}(b)<0`$, the function $`x(r)`$ increases as $`r`$ decreases. The same type of reasoning indicates that in this case there are points $`r_1`$, (with $`r_1<1`$), such that $`z(r_1)=0`$, with $`t(r_1)<0`$, and therefore we may also construct closed, nonspacelike, everywhere future directed null curves.
* Finally, if $`\mathrm{cos}(b)=0`$, points with $`z(r_1)=0`$ correspond also to $`t(r_1)=0`$, and the return portion of the curve intersects the initial portion, resulting directly in a closed curve.
Although, for simplicity, we used the example of a null curve, since this is not geodesic , we conclude that there are closed, non spacelike (either null or timelike), everywhere future directed curves through every point of the manifold (see, e.g., ). This last result follows by taking into account that the metric at any point on the manifold can be put in the form (14), in a coordinate patch where the coordinates of the point are $`(r=1,t=0,z=0)`$.
A further consequence, which is not difficult to prove, is that any pair of points in the manifold can be joined by a future directed, nonspacelike curve, irrespective of the order of the points (note that, by a simple extension of the previous discussion, given any point, we may reach a point arbitrarily in its past by an appropriate choice of $`r_1`$).
### B A comment on the metric given by Thatcher and Morgan
For completeness, and to clarify a point related to the regularity of the metrics, let us now compare the metric (14) with the one given in :
$$ds^2=\mathrm{cos}\left[\frac{8m}{\delta }\mathrm{ln}\left(\frac{r}{\delta }\right)\right](dt^2+dz^2)+2\mathrm{sin}\left[\frac{8m}{\delta }\mathrm{ln}\left(\frac{r}{\delta }\right)\right]dtdz+dr^2+(18E)r^2d\theta ^2.$$
(22)
Contrary to what is stated in , this is not a vacuum solution (unless $`m=0`$, which gives a flat metric), because the Ricci tensor has a (unique) non vanishing component:
$$R_{rr}=32\frac{m^2}{\delta ^2r^2}.$$
The identification of (22) as a vacuum metric made in is based on the assumption that if, for a metric of the form (1) and a certain tensor $`𝒯_{ab}`$ having the same symmetries as $`g_{ab}`$ and whose divergence vanishes (i.e. $`^a𝒯_{ab}=0`$), the equations:
$$R_{tt}=S_{rr},R_{zz}=S_{zz},R_{tz}=S_{tz},R_{\theta \theta }=S_{\theta \theta }\left(\text{ with }S_{ab}=8\pi \left(𝒯_{ab}\frac{1}{2}g_{ab}𝒯\right)\right),$$
(23)
are satisfied, then we necessarily have $`R_{rr}=S_{rr}`$. In particular, this would imply that if Eqs. (23) are satisfied for $`𝒯_{ab}=0`$, then we necessarily have a vacuum solution. This, however, is incorrect. In fact, following Garfinkle , we may define the tensor
$$Q_{ab}=R_{ab}S_{ab}.$$
Then, equations (23) imply that $`Q_{ab}`$ has nonvanishing components only in $`rr`$, i.e.,
$$Q_{ab}=J_ar_br,$$
with $`J`$ a function of $`r`$. Using now the contracted Bianchi identities, $`^aR_{ab}=\frac{1}{2}_bR`$, together with the assumption $`^a𝒯_{ab}=0`$, we find
$$^aQ_{ab}=\frac{1}{2}_bQ(QQ_a^a).$$
For a metric of the form (1), this reduces to $`(Jg)^{}=0`$, where $`(g)`$ is the determinant of the metric, and, therefore, we have
$$J=k/g,$$
(24)
with $`k`$ a constant. Then, if the metric satisfies regularity conditions on the symmetry axis,
$$g_{tt}=1+O(r^2),g_{zz}=1+O(r^2),g_{\theta \theta }=r^2+O(r^4),g_{tz}=O(r^2),$$
(25)
taking the limit $`r0`$ in Eq. (24), we obtain $`k=0`$ (otherwise we would have $`J\mathrm{}`$, corresponding to a singular metric at the axis), i.e. $`J=0`$ and $`Q_{ab}=0`$; but $`J`$ need not vanish if the metric is not regular for $`r=0`$.
## III Superconducting strings with constant momentum
As indicated above, it was suggested in that exotic metrics of the type described in the previous Section may be associated with the spacetime external to a superconducting string with constant momentum. The construction given in proceeds as follows. The metric is assumed to have the form (1), the matter fields are written as $`\mathrm{\Phi }=Re^{i\psi }`$, and $`\sigma =Se^{i\varphi }`$, and their Lagrangean is
$`_{\text{matter}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}^aR_aR{\displaystyle \frac{1}{2}}^aS_aS{\displaystyle \frac{1}{2}}R^2(_a\psi +eA_a)(^a\psi +eA^a){\displaystyle \frac{1}{2}}S^2_a\varphi ^a\varphi `$ (27)
$`\lambda (R^2\eta ^2)^2{\displaystyle \frac{1}{16\pi }}F_{ab}F^{ab}fR^2S^2{\displaystyle \frac{1}{4}}\lambda _2S^4+{\displaystyle \frac{1}{2}}m^2S^2.`$
Then, the following ansatz for these fields is proposed in :
$$R=R(r),\psi =\theta ,A_a=\frac{1}{e}[P(r)1]_a\theta ,S=S(r),\varphi =kz\omega t.$$
The form assumed for $`\varphi `$ is consistent with the idea of endowing the string with a non vanishing momentum. With these assumptions the equations resulting from the Lagrangean (27) are
$`_a^aRR[4\lambda (R^2\eta ^2)+(_a\psi +eA_a)(^a\psi +eA^a))+2fS^2]`$ $`=`$ $`0,`$ (28)
$`_a^aSS\left[_a\varphi ^a\varphi +2fR^2+\lambda _2S^2m^2\right]`$ $`=`$ $`0,`$ (29)
$`_a\left[R^2(^a\psi +eA^a)\right]`$ $`=`$ $`0,`$ (30)
$`_a\left[S^2^a\varphi \right]`$ $`=`$ $`0,`$ (31)
$`^aF_{ab}4\pi eR^2(_b\psi +eA_b)`$ $`=`$ $`0.`$ (32)
Now suppose, as in , that $`_a\varphi `$ is timelike at the axis, i.e., $`\omega ^2k^2>0`$. Performing the change of variables
$$z(\omega ^2k^2)^{1/2}(\omega zkt),t(\omega ^2k^2)^{1/2}(kz\omega t),$$
the conditions at the axis remain unaltered but $`\varphi (\omega ^2k^2)^{1/2}t`$. Thus, without loss of generality we may (and we do) assume $`k=0`$.
Eqs. (30) and (31) are automatically satisfied and the other three are
$`\left(R^{}g^{1/2}\right)^{}g^{1/2}R\left[{\displaystyle \frac{1}{2}}\lambda (R^2\eta ^2)+{\displaystyle \frac{P^2}{g_{\theta \theta }}}+2fS^2\right]`$ $`=`$ $`0,`$ (33)
$`\left(S^{}g^{1/2}\right)^{}g^{1/2}S\left[{\displaystyle \frac{\omega ^2g_{\theta \theta }g_{zz}}{g}}+2fR^2+\lambda _2S^2m^2\right]`$ $`=`$ $`0,`$ (34)
$`\left({\displaystyle \frac{P^{}g^{1/2}}{g_{\theta \theta }}}\right)^{}g_{\theta \theta }g^{1/2}4\pi eR^2P=0.`$ (35)
Einstein equation, $`R_{ab}=8\pi (T_{ab}\frac{1}{2}Tg_{ab})`$, can be written as (for briefness, we do not write down the explicit expressions for the Ricci tensor)
$`R_{tt}`$ $`=`$ $`g_{tt}\left[{\displaystyle \frac{1}{e^2g_{\theta \theta }}}P^{}_{}{}^{}2\lambda \pi (R^2\eta ^2)^22\pi S^2\left(4fR^2+\lambda _2S^22m^2+{\displaystyle \frac{4\omega ^2}{g_{tt}}}\right)\right],`$ (36)
$`R_{zz}`$ $`=`$ $`g_{zz}\left[{\displaystyle \frac{1}{e^2g_{\theta \theta }}}P^{}_{}{}^{}2+\lambda \pi (R^2\eta ^2)^2+2\pi S^2(4fR^2+\lambda _2S^22m^2)\right],`$ (37)
$`R_{rr}`$ $`=`$ $`8\pi (R^{}_{}{}^{}2+S^{}_{}{}^{}2)+\lambda \pi (R^2\eta ^2)^2+2\pi S^2(4fR^2+\lambda _2S^22m^2)+{\displaystyle \frac{1}{e^2g_{\theta \theta }}}P^{}_{}{}^{}2,`$ (38)
$`R_{\theta \theta }`$ $`=`$ $`g_{\theta \theta }\left[{\displaystyle \frac{1}{g_{\theta \theta }}}\left({\displaystyle \frac{1}{e^2}}P^{}_{}{}^{}2+8\pi R^2P^2\right)+\lambda \pi (R^2\eta ^2)^2+2\pi S^2(4fR^2+\lambda _2S^22m^2)\right],`$ (39)
$`R_{tz}`$ $`=`$ $`g_{tz}\left[{\displaystyle \frac{1}{e^2g_{\theta \theta }}}P^{}_{}{}^{}2+\lambda \pi (R^2\eta ^2)^2+2\pi S^2(4fR^2+\lambda _2S^22m^2)\right].`$ (40)
Suppose that we have a solution of Eqs.(33, 34, 35 36,37,38,39,40) that satisfies the regularity conditions (25). We will now show that $`g_{tz}=0`$. For that purpose we first define
$$p=_0^r\left[\frac{h^{}g_{zz}g_{tt}}{2h^2}+\frac{g_{tz}}{g_{tz}^{}}\left(2t_{tz}+\frac{g_{tt}^{}g_{zz}^{}}{h}\right)\right],\text{and}q=g_{tz}^{}g_{\theta \theta }^{1/2}\mathrm{exp}\left(\frac{g_{zz}g_{tt}}{2h}\right),$$
with $`h`$ and $`t_{tz}`$ given by
$$h=g_{tt}g_{zz}g_{tz}^2<0,t_{tz}=\frac{1}{e^2g_{\theta \theta }}P^{}_{}{}^{}2+\lambda \pi (R^2\eta ^2)^2+2\pi S^2(4fR^2+\lambda _2S^22m^2).$$
The convergence for small $`r`$ of the integral that defines $`p`$ is guaranteed by the regularity of the metric (and its non degeneracy) and of the matter fields. With these definitions, Eq. (40) implies that, for $`r>0`$,
$$qe^p=C,$$
(41)
with $`C`$ a constant. Taking the limit $`r0`$ in the l.h.s. of Eq. (41) and using the conditions (25), we find that $`C=0`$, i.e. $`g_{tz}=0`$ $`r`$.
Thus, we have shown that the ansatz for the fields given in , plus regularity conditions on the axis $`r=0`$, imply that the metric can be made everywhere diagonal, and this excludes the possibility of “exotic” behavior. Therefore, we must conclude that the ansatz of does not lead to a source for the “exotic” metrics. The exotic behavior observed in would then have to be ascribed to some peculiarity in their numerical procedures, possibly leading to a failure in strictly satisfying the regularity conditions for $`r=0`$.
## IV Sources for the exotic metric
In this Section we consider some general properties of cylindrically symmetric and stationary regular sources that satisfy the DEC, and which are confined to a cylinder of arbitrary radius $`_1`$. We assume that the spacetime external to the source is vacuum, and obtain restrictions on the possible form of the metrics representing that part of the spacetime. We shall assume that in the source region, consistent with cylindrical symmetry, there exist two killing vector fields, one of them timelike and the other spacelike, which are not necessarily orthogonal to each other. We also require that these killing fields remain timelike and spacelike, respectively, i.e., that their norm does not vanish. Then, the metric can be written as
$$ds^2=e^Adt^2+e^Bdz^2+2D(r)e^{(A+B)/2}dtdz+dr^2+e^Cd\theta ^2.$$
(42)
with $`A,B,C,D`$ functions of $`r`$. We shall see that this kind of sources cannot give rise to exotic spacetimes. The first part of the proof consists in recalling certain inequalities that the DEC imposes on the eigenvalues of $`T_a^b`$. We then rewrite one of these inequalities as a differential one that can be integrated and seen not to be satisfied by the exotic metrics. Without loss of generality, we assume that the latter have $`k<0`$.
### A DEC, eigenvalues and eigenvectors
For metrics of the type (1), the non trivial equations for the eigenvalues and eigenvectors of $`T_a^b`$ correspond to the $`tz`$ sector. For the analysis of this subsection, it is convenient to choose an orthonormal basis $`(\widehat{t}^a,\widehat{z^a})`$ at the point where the analysis is carried out, i.e. $`g_{tz}=0`$, $`g_{tt}=1`$, and $`g_{zz}=1`$. Then, the eigenvalues are given by
$$\lambda _{(ϵ)}=\frac{1}{2}[(T_{zz}T_{tt})+ϵ\gamma ^{1/2}],\gamma =(T_{zz}+T_{tt})^24T_{tz}^2,ϵ=\pm 1.$$
The DEC states that, for any timelike and future directed $`\widehat{m}^a`$, $`n^a=T_{ab}\widehat{m}^a`$ is future directed and causal. It is easy to see, writing $`\widehat{m}^a=\widehat{t}^a\mathrm{cosh}\xi +\widehat{z}^a\mathrm{sinh}\xi `$, that, if the eigenvalues are complex ($`\gamma <0`$), $`n^a`$ is always spacelike if $`T_{zz}T_{tt}=0`$ , and it is spacelike for some $`\xi `$ if $`T_{zz}T_{tt}0`$.
On the other hand, $`n^a`$ is future directed $`\xi `$ if and only if
$$n^a\widehat{t}_a=(T_{zz}+T_{tt})\mathrm{sinh}^2\xi +T_{tt}0,$$
a condition that is satisfied $`\xi `$ if and only if $`T_{tt}0`$ and $`T_{zz}+T_{tt}0`$.
Suppose that the DEC is satisfied, then $`\gamma 0`$ and there are two real eigenvalues. The norms of their corresponding eigenvectors, $`e_{(ϵ)}^a`$, are then given by
$$e_{(ϵ)}^ae_{(ϵ)a}=\frac{ϵ}{2}\gamma ^{1/2}\left(T_{zz}+T_{tt}+\gamma ^{1/2}\right).$$
If $`\gamma =0`$, there is a unique (null) eigenvector, $`\widehat{t}^a+\widehat{z}^a`$, with multiplicity $`2`$, and the energy momentum tensor is of type II in the classification of . If $`\gamma >0`$, one of the eigenvectors is timelike and the other one spacelike. In this case the stress tensor is of type I in the classification of .
### B The proof
In a non orthogonal system of coordinates, such as that one of (42), the eigenvalues of $`T_a^b`$ are
$$\lambda =(2h)^1\left\{T_{tt}g_{zz}+T_{zz}g_{tt}2T_{tz}g_{tz}\pm \left[(T_{tt}g_{zz}T_{zz}g_{tt})^24(T_{zz}g_{tz}T_{tz}g_{zz})(T_{tz}g_{tt}T_{tt}g_{tz})\right]^{1/2}\right\}.$$
The reality of the eigenvalues is thus equivalent to
$$\alpha (T_{tt}g_{zz}T_{zz}g_{tt})^24(T_{zz}g_{tz}T_{tz}g_{zz})(T_{tz}g_{tt}T_{tt}g_{tz})0.$$
(43)
For the metric (1), we have
$$\alpha =\frac{1}{64\pi ^2}(\alpha _1^2\alpha _2^2),$$
with
$`\alpha _1`$ $`=`$ $`{\displaystyle \frac{1}{2}}(1+D^2)e^{(A+BC)/2}\left[(AB)^{}e^{(A+B+C)/2}(1+D^2)^{1/2}\right]^{},`$ (44)
$`\alpha _2`$ $`=`$ $`e^{(A+BC)/2}(1+D^2)^{1/2}\left[D^{}e^{(A+B+C)/2}(1+D^2)^{1/2}\right]^{}{\displaystyle \frac{D}{4}}e^{(A+B)}(AB)^2,`$ (45)
It can also be seen that
$$\alpha _1=8\pi (1+D^2)^{1/2}e^{(A+B)}(T_{ab}\widehat{t}^a\widehat{t}^b+T_{ab}\widehat{z}^a\widehat{z}^b),$$
(46)
so the WEC $`\alpha _10`$. Although it is not related to our proof, let us show something else. From Eqs. (44) and (46) we obtain
$$(AB)^{}e^{(A+B+C)/2}(1+D^2)^{1/2}=16\pi _0^re^{(A+B+C)/2}(1+D^2)^{1/2}(T_{ab}\widehat{t}^a\widehat{t}^b+T_{ab}\widehat{z}^a\widehat{z}^b),$$
and the WEC $`(AB)^{}0`$. Recalling that $`(AB)_{r=0}=0`$, we have $`|g_{tt}|g_{zz}r`$.
Returning to our proof, the point is that the nonnegativity of $`\alpha _1`$ implies that (43) can be written as $`\alpha _1|\alpha _2|0`$. But $`\alpha _1|\alpha _2|0\alpha _1\alpha _20`$. And we have
$$\alpha _1\alpha _2=\frac{1}{2}(1+D^2)e^{(A+BC)/2F}\left(qe^F\right)^{},$$
with
$$q=A^{}B^{}2D^{}(1+D^2)^{1/2},F=_0^r\frac{D\left[(AB)^{}(1+D^2)^{1/2}+2D^{}\right]}{2(1+D^2)e^{(A+B)/2}},$$
(convergence of the integral that defines $`F`$ is guaranteed by Eqs. (25)). Thus the DEC is equivalent to
$$\left(qe^F\right)^{}0.$$
(47)
Integrating Eq. (47) and using once again Eqs. (25), we have $`q0`$ $`r`$. In particular, $`q0`$ at the radius of matching, $`r=_1`$. By regularity, $`q`$, which is made up of the metric components and its first derivatives, is continuous , and, therefore, at $`r=_1`$ it can be evaluated using the exotic metrics, finding
$$q|_{r=_1}=\frac{2k\left[(1+\mathrm{sin}\alpha _1)\mathrm{cos}\alpha _2+(1\mathrm{sin}\alpha _2)\mathrm{cos}\alpha _1\right]}{(_1r_1)\mathrm{cos}\alpha _1\mathrm{cos}\alpha _2}0,$$
(48)
with
$$\alpha _12k\mathrm{ln}(_1r_1)+2\varphi _1,\alpha _22k\mathrm{ln}(_1r_1)+2\varphi _2\pi .$$
Let us now analyze the rhs of Eq. (48). In order for the exterior (exotic) metric to be regular, $`_1r_1>0`$. We also have that $`\mathrm{cos}\alpha _1>0`$ and $`\mathrm{cos}\alpha _2>0`$, since the metric coefficients at $`r=R_1`$ are
$$g_{zz}=c_2^2(_1r_1)^{2q_2}\mathrm{cos}\alpha _1>0,g_{tt}=c_3^2(_1r_1)^{2q_2}\mathrm{cos}\alpha _2<0.$$
Since $`k<0`$, the inequality (48) is equivalent to
$$\left[(1+\mathrm{sin}\alpha _1)\mathrm{cos}\alpha _2+(1\mathrm{sin}\alpha _2)\mathrm{cos}\alpha _1\right]0,$$
but this condition cannot be satisfied if $`\mathrm{cos}\alpha _1>0`$ and $`\mathrm{cos}\alpha _2>0`$.
## V Shells
In this section we construct and analyze the spacetimes corresponding to one or more concentric shells matched to an exotic exterior, and explicitly show that the DEC is violated. As we shall see, the interior metric can be diagonalized, so the corresponding hypothesis of the previous section trivially holds. However, the distribution of matter is not regular and, in particular, the function $`q`$ is not continuous at $`r=_1`$, as we have assumed in the previous section. Nevertheless, one can take into account the fact that we are dealing with shells, make minor changes on the proof above given, and see that the final result still holds.
For the discussions on energy conditions it is convenient to introduce a normalized base:
$$\widehat{r}^a=(_r)^a,\widehat{\theta }^a=g_{\theta \theta }^{1/2}(_\theta )^a,\widehat{z}^a=g_{zz}^{1/2}(_z)^a,\widehat{t}^a=|g_{tt}|^{1/2}(_t)^a.$$
As in the previous Section, we choose $`k<0`$.
### A One shell
In this subsection we analyze the case in which there is just one shell of radius, say, $`_0`$. The interior metric must be flat, otherwise it would be singular. It can be seen that, if the exterior metric is exotic, the shell must violate the DEC. We leave the proof of this statement for the following subsection. Thus, the exterior geometry of a physically reasonable shell is described by a Levi-Civita metric. The regularity at the axis plus the matching conditions at the shell imply that both the exterior and interior metric can be simultaneously diagonalized, we use this fact to write
$`ds^2`$ $`=`$ $`dr^2+r^2d\theta ^2+dz^2dt^2,\mathrm{\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}0}r_0,`$ (49)
$`ds^2`$ $`=`$ $`dr^2+R_0^2\left({\displaystyle \frac{rr_0}{R_0r_0}}\right)^{4(\mathrm{\Delta }1)/(\mathrm{\Delta }^2+3)}d\theta ^2+\left({\displaystyle \frac{rr_0}{R_0r_0}}\right)^{4(\mathrm{\Delta }+1)/(\mathrm{\Delta }^2+3)}dz^2+`$ (51)
$`\left({\displaystyle \frac{rr_0}{R_0r_0}}\right)^{2(\mathrm{\Delta }^21)/(\mathrm{\Delta }^2+3)}dt^2,_0r.`$
In the above expressions (and for the rest of the paper) we assume, for simplicity, $`_0>r_0`$; the analysis for $`_0<r_0`$ is similar, and one can see that the results that we are interested on also hold in that case. We now write down the components of $`T_{ab}`$ that are obtained when the metric defined by Eqs. (49,51) is used to evaluate the l.h.s. of Einstein equation, $`G_{ab}=8\pi T_{ab}`$. We shall need them in the next subsection. The results are
$`T_{ab}\widehat{\theta }^a\widehat{\theta }^b`$ $`=`$ $`{\displaystyle \frac{(\mathrm{\Delta }+1)^2}{8\pi (_0r_0)(\mathrm{\Delta }^2+3)}}\delta (r_0)`$ (52)
$`T_{ab}\widehat{z}^a\widehat{z}^b`$ $`=`$ $`{\displaystyle \frac{2_0(\mathrm{\Delta }+1)r_0(\mathrm{\Delta }^2+3)}{8\pi _0(_0r_0)(\mathrm{\Delta }^2+3)}}\delta (r_0),`$ (53)
$`T_{ab}\widehat{t}^a\widehat{t}^b`$ $`=`$ $`{\displaystyle \frac{_0(\mathrm{\Delta }^21)r_0(\mathrm{\Delta }^2+3)}{8\pi _0(_0r_0)(\mathrm{\Delta }^2+3)}}\delta (r_0),`$ (54)
and the trace of this stress - energy tensor is
$$T=\frac{r_0\delta (r_0)}{4\pi _0(_0r_0)},$$
so $`r_0=0T=0`$. That is, though $`r_0`$ locally corresponds to a simple shift in the $`r`$ coordinate, it provides non trivial information about the invariant $`T`$. Thus, one cannot choose $`r_0=0`$ without loss of generality. Something similar led to the belief that a rotating cylinder can only exist in general relativity for “incoherent (traceless) matter” (more on rotating cylindrical shells can be found in ).
The SEC, $`T_{ab}\widehat{m}^a\widehat{m}^b+\frac{1}{2}T0`$ for all unit timelike $`\widehat{m}^a`$, implies in this case (choosing $`\widehat{m}^a=\widehat{t}^a`$)
$$T_{ab}\widehat{t}^a\widehat{t}^b+\frac{1}{2}T=\frac{(\mathrm{\Delta }^21)}{8\pi (\mathrm{\Delta }^2+3)(_0r_0)}\delta (r_0)0,$$
and it is satisfied if and only if $`\mathrm{\Delta }^2>1`$ (the inequality must be reversed if $`_0<r_0`$). It is easy to see that this is a general property of Levi-Civita metrics, e.g., it does not depend on the fact that we are dealing with shells (see the last Section for further comments on the SEC).
Similarly, as a necessary condition for the WEC, we have
$$T_{ab}\widehat{t}^a\widehat{t}^b+T_{ab}\widehat{z}^a\widehat{z}^b=\frac{(\mathrm{\Delta }+1)(\mathrm{\Delta }3)}{8\pi (\mathrm{\Delta }^2+3)(_0r_0)}\delta (r_0)0.$$
(55)
In Sec. IV we saw that the WEC $`|g_{tt}|g_{zz}`$ under general conditions. Let us explicitly check this property: writing
$$g_{tt}+g_{zz}=\left(\frac{rr_0}{_0r_0}\right)^{2(\mathrm{\Delta }^21)/(\mathrm{\Delta }^2+3)}\left[1\left(\frac{rr_0}{_0r_0}\right)^{2(\mathrm{\Delta }+1)(\mathrm{\Delta }3)/(\mathrm{\Delta }^2+3)}\right],$$
we notice that Eq. (55) $`g_{tt}+g_{zz}0`$.
### B Concentric shells
We now consider a model consisting of two concentric shells. As discussed in the previous subsection, assuming that the interior of the innermost shell is empty and regular, and that it is made out of matter and (or) fields satisfying the DEC, the metric between the innermost and outermost shells must be taken to be of the Levi-Civita form. Our problem then is to analyze the matching of the Levi-Civita metric (51) with a metric of the form given by (9,10,11,12), through a singular shell at $`r=_1>_0`$. We obtain
$`g_{\theta \theta }`$ $`=`$ $`_0^2\left({\displaystyle \frac{_1r_0}{_0r_0}}\right)^{4(\mathrm{\Delta }1)/(\mathrm{\Delta }^2+3)}\left({\displaystyle \frac{rr_1}{_1r_1}}\right)^{2q_1},`$
$`g_{zz}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{cos}\alpha }}\left({\displaystyle \frac{_1r_0}{_0r_0}}\right)^{4(\mathrm{\Delta }+1)/(\mathrm{\Delta }^2+3)}\left({\displaystyle \frac{rr_1}{_1r_1}}\right)^{2q_2}\mathrm{cos}[2k\mathrm{ln}(rr_1)+2\varphi _1)],`$
$`g_{tt}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{cos}\alpha }}\left({\displaystyle \frac{_1r_0}{_0r_0}}\right)^{2(\mathrm{\Delta }^21)/(\mathrm{\Delta }^2+3)}\left({\displaystyle \frac{rr_1}{_1r_1}}\right)^{2q_2}\mathrm{cos}[2k\mathrm{ln}(rr_1)4k\mathrm{ln}(_1r_1)2\varphi _1)],`$
$`g_{tz}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{cos}\alpha }}\left({\displaystyle \frac{_1r_0}{_0r_0}}\right)^{(\mathrm{\Delta }+1)^2/(\mathrm{\Delta }^2+3)}\left({\displaystyle \frac{rr_1}{_1r_1}}\right)^{2q_2}\mathrm{sin}\left[2k\mathrm{ln}(rr_1)2k\mathrm{ln}(_1r_1)\right],`$
with $`\alpha =2k\mathrm{ln}(_1r_1)+2\varphi _1`$ such that $`\mathrm{cos}\alpha >0`$.
The components of the energy-momentum tensor for the innermost shell are given by (52,53,54), and for the outermost one by
$`T_{ab}\widehat{\theta }^a\widehat{\theta }^b`$ $`=`$ $`{\displaystyle \frac{1}{8\pi }}\left[{\displaystyle \frac{2(1s\sqrt{1+3k^2})}{3(_1r_1)}}{\displaystyle \frac{(\mathrm{\Delta }+1)^2}{(\mathrm{\Delta }^2+3)(_1r_0)}}\right]\delta (r_1),`$
$`T_{ab}\widehat{z}^a\widehat{z}^b`$ $`=`$ $`{\displaystyle \frac{1}{8\pi }}\left[{\displaystyle \frac{1}{3(_1r_1)}}\left(2+s\sqrt{1+3k^2}+3k\mathrm{tan}\alpha \right){\displaystyle \frac{(\mathrm{\Delta }1)^2}{(\mathrm{\Delta }^2+3)(_1r_0)}}\right]\delta (r_1),`$
$`T_{ab}\widehat{t}^a\widehat{t}^b`$ $`=`$ $`{\displaystyle \frac{1}{8\pi }}\left[{\displaystyle \frac{1}{3(_1r_1)}}\left(2s\sqrt{+3k^2}+3k\mathrm{tan}\alpha \right)+{\displaystyle \frac{4}{(\mathrm{\Delta }^2+3)(_1r_0)}}\right]\delta (r_1),`$
$`T_{ab}\widehat{t}^a\widehat{z}^b`$ $`=`$ $`{\displaystyle \frac{k}{8\pi (_1r_1)\mathrm{cos}\alpha }}\delta (r_1),`$
whereas its trace is
$$T=\frac{(r_1r_0)}{4\pi (_1r_1)(_1r_0)}\delta (r_1),$$
and, similar to what happens for a single shell, $`r_1=r_0T=0`$.
As a necessary condition for the WEC on the outermost shell,
$$T_{ab}\widehat{t}^a\widehat{t}^b+T_{ab}\widehat{z}^a\widehat{z}^b=\frac{1}{8\pi }\left[\frac{2k\mathrm{tan}\alpha }{(_1r_1)}\frac{(\mathrm{\Delta }+1)(\mathrm{\Delta }3)}{(\mathrm{\Delta }^2+3)(_1r_0)}\right]\delta (r_1)0.$$
(56)
Combining Eq. (56) with the analogous equation for the innermost shell, Eq. (55), we obtain
$$_1=_0^_1g^{1/2}(T_{ab}\widehat{t}^a\widehat{t}^b+T_{ab}\widehat{z}^a\widehat{z}^b)=\frac{_0(_1r_0)k\mathrm{tan}\alpha }{4\pi (_0r_0)(_1r_0)}0.$$
(57)
The other integral we need is
$$_2=_0^_1g^{1/2}T_{ab}\widehat{t}^a\widehat{z}^b=\frac{k_0(_1r_0)}{8\pi (_0r_0)(_1r_0)\mathrm{cos}\alpha },$$
(58)
which is also positive, not because of any energy condition, but rather due to the range of the constants that appear on it. Finally, the DEC implies that $`_12_2`$, an inequality that cannot be satisfied if $`k<0`$, and both (57) and (58) are positive. Therefore, as stated, the DEC is necessarily violated if the matching to the exotic metric is non trivial.
## VI Final comments
The possibility of the existence of an exotic metric associated with a physical source is of course very intriguing. In a general sense, it would amount to some form of “frame dragging”, that results from the presence of a momentum flux in the source, such that the stress - energy - momentum tensor cannot be diagonalized in general, in some way reminiscent of the frame dragging effect for source endowed with rotation. Unfortunately, in all the examples analyzed in this paper, we have not been able to construct such a source, if we also impose the dominant energy condition. We have also constructed a general proof of nonexistence of sources satisfying the usual physical requirements, but only under some restrictions. In particular, the main assumption of the proof given in Section IV is that there is a killing vector field that is everywhere spatial in the source region, and another one that is everywhere timelike, in the same region. It applies, in particular, to sources that have small relative (non trivial) flux of momentum, but it is not the general case. For example, the exotic metrics do not satisfy this condition (though, of course, they can not be used as sources because they are not regular at the axis). It would be interesting to have some result concerning sources that do not satisfy this hypotheses, either showing that they can generate exotic metrics, or extending the present proof to those cases.
The other important assumption in Sec. IV is that the matter satisfies the DEC. This is usually considered to be a physically reasonable assumption, satisfied by non tachyonic matter, in particular, by topological defects. We notice, however, that some of the latter violate another energy condition, the so called strong energy condition (SEC). There are different ways of realizing that sources for a given spacetime must violate it. For example (following the notation of (42) with $`D=0`$), if one looks at the geodesic equation, one notices that $`A^{}<0`$ corresponds to a “repulsive” gravitational field, a situation in which one would suspect that the SEC is being violated. This is indeed the case, the components of the Ricci tensor are
$`R_{ab}\widehat{t}^a\widehat{t}^b`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left[A^{}e^{(A+B+C)/2}\right]^{}e^{(A+B+C)/2},`$
$`R_{ab}\widehat{z}^a\widehat{z}^b`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left[B^{}e^{(A+B+C)/2}\right]^{}e^{(A+B+C)/2},`$
$`R_{\theta \theta }\widehat{\theta }^a\widehat{\theta }^b`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left[C^{}e^{(A+B+C)/2}\right]^{}e^{(A+B+C)/2},`$
$`R_{ab}\widehat{r}^a\widehat{r}^b`$ $`=`$ $`{\displaystyle \frac{1}{2}}(A+B+C)^{^{\prime \prime }}{\displaystyle \frac{1}{4}}(A^{}_{}{}^{}2+B^{}_{}{}^{}2+C^{}_{}{}^{}2),`$
and, thus, a regular solution of Einstein equations satisfies
$$A^{}=16\pi e^{(A+B+C)/2}_0^re^{(A+B+C)/2}𝒮𝒞,$$
(59)
where $`𝒮𝒞=T_{ab}\widehat{t}^a\widehat{t}^b+T/2`$. From Eq. (59) we can see that if $`A^{}<0`$, then $`𝒮𝒞<0`$ (at a set of finite measure), and the SEC is violated.
In general, topological defects that violate the SEC have a global symmetry. For example, following the notation of Section III, for a $`U(1)`$ global string we have
$$𝒮𝒞=\frac{1}{8}\lambda (R^2\eta ^2)^2,$$
which is manifestly negative, so the SEC is everywhere violated. Similarly, global monopoles and global vacuumless defects have repulsive gravitational fields that suggest that the SEC is violated. In the case of gauge defects, it is usually supposed that they satisfy the SEC. But, in such cases, we face the difficulty that $`𝒮𝒞`$ does not have definite sign, and, therefore, we do not know whether at a given point it is positive or not without knowledge of the solution of the field equations. Nevertheless, there might be an interesting exception to the general belief that gauge defects satisfy the SEC, an exception that seems not to have been noticed up to present. Amsterdamski and Laguna have numerically solved the equations describing a gravitating superconducting string and, interestingly, if one observes figure (7) of that paper, one notes that $`g_{tt}`$ clearly has a local minimum. According to the Eq. (59), this means that superconducting strings violate the SEC.
## Acknowledgments
This work was supported in part by funds of the University of Córdoba, and grants from CONICET and CONICOR (Argentina). R.J.G. is a member of CONICET. M.H.T. acknowledges financial support from CONICOR and CONICET. |
warning/0001/astro-ph0001387.html | ar5iv | text | # Interferometric Astrometry of the Detached White Dwarf - M Dwarf Binary Feige 24 Using Hubble Space Telescope Fine Guidance Sensor 3: White Dwarf Radius and Component Mass Estimates 1footnote 11footnote 1Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555
## 1 Introduction
Feige 24 ( = PG 0232+035 = HIP 12031) is DA white dwarf, red dwarf (M1-2V) (Liebert & Margon (1977)) binary (P = 4.23 days, Vennes & Thorstensen (1994)=VT94) that is described as the prototypical post-common envelope detached system with a low probability of becoming a Cataclysmic Variable (CV) within a Hubble time (King et al. (1994) and Marks, 1994). This object was selected for our HST parallax program because a directly measured distance could reduce the uncertainty of the radius of one of the hottest white dwarfs. Since the instigation of this program and the selection of targets over 15 years ago, at least two other groups have measured a parallax for Feige 24; USNO-Flagstaff (Dahn et al., (1988)), and HIPPARCOS (Perryman et al., (1997) and Vauclair et al., (1997)). We outlined the results of a preliminary analysis in Benedict et al., (2000). Here we discuss our analysis and final results in detail.
Provencal et al. (1998) presented radii derived from HIPPARCOS parallaxes for 21 white dwarfs. In most cases, the dominating error term for the white dwarf radii was the parallax uncertainty. Our parallax of Feige 24, while slow in coming, has provided a fractional parallax uncertainty, $`\frac{\mathrm{\Delta }\pi }{\pi }`$, similar to those in the Provencal et al. (1998) study, but for a much hotter, more distant object.
We time-tag our data with a modified Julian Date, $`MJD=JD2400000.5`$. We abbreviate millisecond of arc, mas; white dwarf, DA; and M dwarf, dM, throughout.
## 2 The Astrometry
Our astrometric observations were obtained with Fine Guidance Sensor 3 (FGS 3), a two-axis, white-light interferometer aboard HST. Bradley et al., (1991) provide an overview of the FGS 3 instrument and Benedict et al., (1999) describe the astrometric capabilities of FGS 3 and typical data acquisition and reduction strategies.
We use the term ‘pickle’ to describe the field of regard of the FGS. The instantaneous field of view of FGS 3 is a $`5\times 5`$ arcsec square aperture. Figure 1 shows a finding chart for Feige 24 and our astrometric reference stars in the FGS 3 pickle as observed on 08 Aug 1997. Note the less than ideal placement of the primary science target with respect to the reference frame. The placement of Feige 24, at one side of the distribution of reference stars, seems to have produced no adverse astrometric or photometric effects.
### 2.1 The Astrometric Reference Frame
Table 3 provides a list of the observation epochs. Our data reduction and calibration procedures are described in Benedict et al., (1999) and McArthur et al., (1999). We obtained a total of 71 successful measurements of our reference stars during eight ’observing runs’. For each of these eight observation sets we determine the scale and rotation relative to the sky, using a GaussFit (Jefferys et al., (1988)) model. The orientation of the observation sets is obtained from ground-based astrometry (USNO-A2.0, Monet (1998)) with uncertainties in the field orientation $`\pm 0\stackrel{}{\mathrm{.}}12`$.
Having only 8 observation sets and four reference stars precludes us from our usual practice (Benedict et al., (1999)) of constraining the proper motions and parallaxes to sum to zero ($`\mathrm{\Sigma }\mu =0`$ and $`\mathrm{\Sigma }\pi =0`$) for the entire reference frame. From a series of solutions we determined that only reference star ref-3 has a statistically significant proper motion and parallax. So, we constrain $`\mu =0`$ and $`\pi =0`$ for reference stars ref-2, -4 and -5.
We conclude from histograms (Figure 2) of the reference star residuals that we have obtained a per-observation precision of $`1`$ mas. The resulting reference frame ’catalog’ (Table 2) was determined with final errors $`<\sigma _\xi >=0.5`$ and $`<\sigma _\eta >=0.6`$ mas.
To determine if there might be unmodeled, but eventually correctable, systematic effects at the 1 mas level, we plotted the Feige 24 reference frame X and Y residuals against a number of spacecraft, instrumental, and astronomical parameters. These included X, Y position within the pickle; radial distance from the pickle center; reference star V magnitude and B-V color; and epoch of observation. We saw no trends, other than the expected increase in positional uncertainty with reference star magnitude.
### 2.2 Modeling the Parallax and Proper Motion of Feige 24
Spectroscopy of the reference frame stars obtained from the WIYN <sup>1</sup><sup>1</sup>1The WIYN Observatory is a joint facility of the University of Wisconsin-Madison, Indiana University, Yale University, and the National Optical Astronomy Observatories. and an estimate of color excess, E(B-V), from Burstein & Heiles (1982). (Table 2) shows that the colors of the reference stars and our science target differ, with $`\mathrm{\Delta }(\mathrm{B V})1`$. Therefore, we apply the differential correction for lateral color discussed in Benedict et al., (1999) to the Feige 24 observations and obtain a parallax relative to our reference frame, $`\pi _{rel}=13.8\pm 0.4`$ mas. The proper motion relative to the four astrometric reference stars is listed in Table 4.
Franz et al (1998) and Benedict et al., (1999) have demonstrated 1 mas astrometric precision for FGS 3. Table 3 presents our Feige 24 astrometric residuals obtained from the parallax and proper motion model. Histograms of these residuals are characterized by $`\sigma _x=1.0`$ and $`\sigma _y=1.2`$ mas. This was slightly larger than expected. To investigate whether or not the larger residuals could be attributed to Feige 24, Figure 3 presents the residuals phased to the VT94 orbital period, $`P=4.23160^d`$, with $`T_0=HJD2448578.3973`$. We find no significant trends in the astrometric residuals. In particular, there is no correlation with the two distinct HST orientations required by the pointing constraints discussed in Benedict et al., (1999). With any reasonable masses for the DA and dM components, a binary system at this distance, having this period, could exhibit maximum reflex motion at the 0.5 mas level. This null detection does not place very useful upper limits on the component masses.
Because our parallax for Feige 24 is determined with respect to the reference frame stars which have their own parallaxes, we must apply a correction from relative to absolute parallax. The WIYN spectroscopy and the estimated color excess (See Table 2) indicate a reference frame with an average parallax of $`<\pi >_{ref}=0.9\pm 0.4`$ mas. where the error is based on the dispersion of the individual spectrophotometric parallaxes. To check our correction to absolute, we compare it to those used in the Yale Parallax Catalog (YPC95, van Altena, Hoffleit, & Lee, Section 3.2). From YPC95, Fig. 2, the Feige 24 galactic latitude, $`b=50\stackrel{}{\mathrm{.}}3`$ and average magnitude for the reference frame, $`<V_{ref}>=13.4`$, we obtain a correction to absolute of 1.9 mas. Rather than use a galactic model-dependent correction, we adopt the spectroscopically derived $`<\pi >_{ref}=0.9\pm 0.4`$ mas. Applying this correction results in an absolute parallax of $`\pi _{abs}=+14.7\pm 0.6`$ mas, where the error has equal contributions from the HST FGS observations and the correction to absolute parallax. Finally, we note that our proper motion is smaller than either the HIPPARCOS or USNO values, where the HIPPARCOS value is an absolute proper motion while the USNO and the HST values are relative to their respective reference frame proper motions. If our reference stars are a representative statistical sample of the parent population, then based on the data in Table III in van Altena (1974), we expect a statistical uncertainty in the mean value of the correction to absolute proper motion (not applied here) of $`\pm 6`$ mas y<sup>-1</sup>.
We compare our absolute parallax to previous work in Table 4 and in Figure 4. We adopt for the remainder of this paper the weighted average absolute parallax, $`<\pi _{abs}>=14.6\pm 0.4`$ mas, shown as a horizontal dashed line in Figure 4. Weights used are $`1/\sigma ^2`$.
Lutz & Kelker (1973) show that for a uniform distribution of stars, the measured trigonometric parallaxes are strongly biased towards the observer (i.e., too large), rendering inferred distances and luminosities too small. This bias is proportional to $`(\sigma _\pi /\pi )^2`$. Using a space density determined for the CV RW Tri (McArthur et al. 1999), and presuming that Feige 24 is a member of that same class of object (binaries containing white dwarfs), we determine an LK correction of $`0.01\pm 0.01`$ magnitudes. Correcting our distance modulus, we obtain $`mM=4.17\pm 0.11`$.
### 2.3 Kinematic Age of the Feige 24 System
From the VT94 systemic radial velocity and either our proper motions or those from HIPPARCOS (Table 4) we derive the space velocity of Feige 24, 67 $`\pm `$ 1 km s<sup>-1</sup>. The velocity component perpendicular to the galactic plane, W, is -37 km s<sup>-1</sup>. Our new parallax places the star 53 parsecs below the Sun or 61 parsecs below the galactic plane. An object this far below the galactic plane and continuing to move further away from the plane so swiftly is more characteristic of a ’thick disk’ than a thin disk object (c.f. Thejll et al., (1997)). Feige 24, if truly a Pop I object, has a space velocity 3.5 times the young disk velocity dispersion. These data suggest that Feige 24 formed prior to the formation of the galactic disk, although subsequent evolution of the DA component is likely quite recent. This may be an instance of past mass transfer in an intermediate Population II object.
## 3 Astrophysics of the Feige 24 System
We discuss the consequences of a more precisely determined parallax, calculating some astrophysically relevant parameters for the DA and dM components. These are collected in Table 5. Our goals are the radius and mass of the DA component. We first calculate a radius, then estimate the time since the DA formation event. Component masses have been estimated by VT94. We will revisit this issue later. That we do not substantially improve the mass uncertainty motivates a future direct measurement of the component separation. This one measurement would yield precise masses. A series of measurements would provide individual orbits, possibly illuminating past and future component interactions.
### 3.1 Estimating the DA Radius
To estimate the DA radius we require an intrinsic luminosity. From Landolt (1983) we obtain a system total magnitude, $`V_{tot}=12.41\pm 0.01`$. The magnitude of the white dwarf is critical and difficult to obtain, because the M dwarf always contributes flux. Holberg et al., (1986) derive $`V_{DA}=12.56\pm 0.05`$ using IUE spectra. They ratio Feige 24 with other hot DA; G191 B2B, GD246 and HZ43. From the DA magnitude and total magnitude we obtain $`V_{dM}=14.63\pm 0.05`$ and $`\mathrm{\Delta }V=2.07`$. We assume an $`A_V=0`$ for Feige 24 at d = 69 pc, consistent with our adopted $`A_V=0.09`$ for the reference frame at an average distance d = 1600 pc (Table 2). The LK bias-corrected distance modulus ($`mM=4.17\pm 0.11`$) then yields absolute magnitudes $`M_V=10.46\pm 0.12`$ for the red dwarf companion and $`M_V=8.39\pm 0.12`$ for the DA.
A recently determined temperature of the Feige 24 DA, taking into account non-LTE and heavy element effects (Barstow Hubeny & Holberg (1998)), is $`T_{eff}^{DA}=56,370\pm 1,000K`$. This temperature yields a radius via differential comparison with the sun. This procedure requires a bolometric magnitude, hence, a bolometric correction. We could adopt the bolometric correction, B.C. = -4.88, generated by Bergeron et al. (1995) from a pure Hydrogen, $`logg=8`$ DA model convolved with a V bandpass. But, Feige 24 is neither $`logg=8`$ nor pure H.
Flower (1996) provides bolometric corrections for normal stars up to $`T_{eff}54,000K`$. From Flower (1996), figure 4, the relationship between $`logT_{eff}`$ and B.C. is linear for $`T_{eff}>25,000K`$. Hotter stars lie on the Rayleigh-Jeans tail of the blackbody curve, where flux is roughly proportional to $`T_{eff}`$, not $`T_{eff}^4`$. A small linear extrapolation yields B.C. = -4.82$`\pm 0.06`$ for the Feige 24 DA. The B.C. error comes from the uncertainty in $`T_{eff}^{DA}`$. Because a DA with some heavy elements in its atmosphere radiates more like a hot normal star than a pure H DA, we choose the Flower correction rather than the model correction. We are also encouraged by the near equality of the B.C. from observation and theory.
We obtain a DA bolometric luminosity $`M_{bol}^{DA}=M_V+B.C.=3.57\pm 0.13`$. $`R_{DA}`$ follows from the expression
$$M_{bol}^{\mathrm{}}M_{bol}^{DA}=10log(T_{eff}^{DA}/T_{eff}^{\mathrm{}})+5log(R_{DA}/R_{\mathrm{}})$$
(1)
where we assume for the Sun $`M_{bol}^{\mathrm{}}=+4.75`$ and $`T_{eff}^{\mathrm{}}=5800K`$. We find $`R_{DA}=0.0180\pm 0.0013R_{\mathrm{}}`$, following the error analysis of Provencal et al. (1998). The primary sources of error for this radius are the bolometric correction and the $`T_{eff}^{DA}`$.
A second approach to deriving $`R_{DA}`$ involves the V-band average flux, $`H_V`$, discussed in Bergeron et al. (1995). They list $`H_V^{DA}`$ as a function of temperature for, again, the pure Hydrogen, $`logg=8`$ model. If we can determine an $`H_V^{\mathrm{}}`$, we can derive $`R_{DA}`$ from
$$R_{DA}=(H_V^{\mathrm{}}/H_V^{DA})10^{0.4(M_V^{DA}M_V^{\mathrm{}}})$$
(2)
where $`M_V^{DA}=8.39\pm 0.12`$ comes from our parallax and $`M_V^{\mathrm{}}=4.82`$ is assumed. We obtain $`H_V^{\mathrm{}}`$ by convolving the Bessel (1990) V band response with the solar spectral distribution listed in Allen (1973). We calculate $`H_V^{\mathrm{}}=6.771\times 10^5`$ ergs cm<sup>-2</sup> s<sup>-1</sup> Å<sup>-1</sup> str<sup>-1</sup>. We obtain for $`T_{eff}=56,370K`$ an $`R_{DA}=0.0188\pm 0.0010`$.
A weighted average of the two independent determinations provides $`R_{DA}=0.0185\pm 0.0008R_{\mathrm{}}`$, where the error is certainly underestimated due to unknown systematic effects. Parallax is no longer a significant source of error for the radius determination. Comparing with the results presented in Provencal et al. (1998), figure 7, we find Feige 24 to have a radius larger than any other white dwarf.
With a temperature $`T_{eff}56,000`$, the time since the DA formation event is unlikely to be longer than 1.5 My. This conclusion is drawn from the DA cooling tracks as function of mass calculated by M. Wood, detailed in Sion (1999), fig. 7. These models also indicate that the DA mass must satisfy $`_{DA}0.4_{\mathrm{}}`$ to remain near this lofty $`T_{eff}`$ for longer than $`3\times 10^5`$ y.
### 3.2 Estimating The White Dwarf Mass
Before estimating $`_{DA}`$ we review the VT94 minimum component masses from their radial velocities and the Kepler relation for total system mass, separation, and period. Then, we estimate the DA mass using two different approaches. We first attempt to determine the most likely dM mass. The VT94 radial velocity amplitude ratio then provides the DA mass. The second, independent mass estimate follows from our derived radius along with the DA atmospheric parameter, $`logg`$, obtained through spectroscopy. Our DA mass estimate will differ little from VT94, and, if better, is so only by virtue of more recent dM models and DA atmospheric parameters.
#### 3.2.1 Minimum Component Masses from Binary Radial Velocities
The system total lower mass limit can be set by the VT94 radial velocities and the Kepler relation for mass, separation, and period. VT 94 give us the velocities along each component orbit, the fact that each orbit is circular (from the pure sine wave fits to the velocity curves), and the period, the time it takes to travel around each orbit. Assuming an edge-on system ($`i=90\mathrm{deg}`$), one that can produce the full vector amount of radial velocity amplitude measured by VT94, the minimum system mass is $`_{tot}=0.73_{\mathrm{}}`$. From the VT94 mass ratio, $`_{dM}/_{DA}=0.63\pm 0.04`$, we obtain the DA mass limit, $`_{DAKepler}0.44_{\mathrm{}}`$, and the dM mass limit, $`_{dMKepler}0.26_{\mathrm{}}`$. No smaller masses can produce the observed radial velocities for orbits of these known sizes. At d = 68.5 pc an edge-on system with minimum mass would separate the components by 672 microarcsec, or $`9.9R_{\mathrm{}}`$.
#### 3.2.2 Inclination from the Light Curve
VT 94 find H$`\alpha `$ equivalent width variations that phase with the orbital period. These show a maximum at $`\varphi =0.5`$. Photometric variations of Feige 24 might be detectable, because the photometric capabilities of FGS 3 approach a precision of 0.002 magnitude (Benedict et al., (1998)). Figure 5 shows the flat-fielded counts and the corresponding differential instrumental magnitudes as well as a sin wave fit with amplitude and phase as free parameters. There is a clear photometric signature with a peak-to-peak amplitude 0.028 magnitude, showing maximum system brightness at phase $`\varphi =0.58\pm 0.09`$. Given the sparse coverage, this phase at maximum is not surprisingly different from the H$`\alpha `$ equivalent width maximum seen at $`\varphi =0.5`$.
A likely mechanism for producing the single-peaked orbital light curve is heating of the dM star by the white dwarf (the reflection effect). As the dM star orbits the white dwarf, its heated face is alternately more or less visible, increasing and decreasing the observed flux from Feige 24 once per orbit. To test this hypothesis we calculated model light curves using an updated version of the light curve synthesis program described by Zhang et al. (1986). We initially adopted $`T_{eff}=56,370`$ K and $`R=0.0185R_{}`$ for the white dwarf, $`T_{eff}=3800`$ K and $`R=0.52R_{}`$ for the M1-2V star, and $`4.8\times 10^2`$ AU ($`10.3R_{\mathrm{}}`$) for the separation of their centers of mass, and then adjusted the temperature of the dM star so that it contributed 13.5% of the V flux from the system. The peak-to-peak amplitudes of the resulting model light curves are a function of orbital inclination, topping out at $`0.025`$ mag for $`i=90^{}`$, and can easily be made to agree in amplitude and shape with the observed light curve.
This photometric behavior is entirely consistent with reflection effects (c.f. Robinson et al. 2000). We find that the quality of the observed light curve is, however, inadequate to improve the parameters of the system, particularly the inclination. We have not sufficiently sampled the expected flat section of the light curve (near $`\varphi =0`$). Nevertheless, these results do provide quantitative evidence that (1) the orbital light curve is caused by heating and (2) the heating is consistent with the radius and temperature we have derived for the white dwarf – a useful external check on our results.
#### 3.2.3 DA Mass from the M Dwarf
The dM absolute magnitude ($`M_V=10.46\pm 0.12`$ ) implies a spectral type M2V (Henry, Kirkpatrick, & Simons (1994)), consistent with Liebert & Margon (1977). The absolute magnitude of an M dwarf star depends not only on mass, but also on age (evolutionary stage) and chemical composition. Baraffe et al., (1998) have produced a grid of models, varying metallicity, \[M/H\] and helium abundance, Y. We plot in Figure 6 their Mass-Luminosity curves for dwarfs with ages 10My and 10Gy, with \[M/H\] = 0 and 10Gy with \[M/H\] = -0.5, all with solar helium abundance. The complete grid of Baraffe et al. models shows that M dwarfs in the mass range $`0.175_{dM}0.43_{\mathrm{}}`$ with -0.5 $`<`$ \[M/H\] $`<`$ 0 have $`M_V=10.46`$ at some time in their evolution from 10My to 10Gy.
The dM mass now depends on metallicity and how quickly an M dwarf of a given mass decreases in brightness. Figure 7 shows the dependence of brightness on mass, age, and metallicity. These Baraffe et al. models indicate that solar metallicity stars with higher mass remain near $`M_V=10.46`$ far longer than low mass stars. However, kinematically, Feige 24 is more likely to be old and of lower than solar metallicity than young and of normal metallicity. First adopting the 10Gy model, \[M/H\] = 0, and calculated absolute magnitude, we estimate the dM star mass $`_{dM}=0.43\pm 0.08_{\mathrm{}}`$, because that mass remains at $`M_V=10.46`$ for a larger fraction of the total lifetime than any other. However, if we accept the kinematical suggestion of allegiance to a thick disk population, then \[M/H\] $`<`$ 0 is more likely. Assuming \[M/H\] = -0.5 results in a dM star mass $`_{dM}=0.185\pm 0.08_{\mathrm{}}`$.
Radial velocities from VT94 (dM from Kitt Peak, DA from IUE) provide the velocity amplitude ratio, $`K_{DA}/K_{dM}=0.63\pm 0.04=_{dM}/_{DA}`$. From the total possible dM mass range, $`0.185<_{dM}<0.43_{\mathrm{}}`$, and the mass ratio we derive a DA mass range, $`0.29<_{DA}<0.68_{\mathrm{}}`$. Applying the limit, $`_{DAKepler}0.44_{\mathrm{}}`$, we obtain $`0.44<_{DA}<0.68_{\mathrm{}}`$. Keplerian lower limits argue for a dM star mass $`0.26<_{dM}<0.43_{\mathrm{}}`$, a range consistent with a metallicity slightly less than solar and an age in excess of 0.3 Gy (Figure 7).
#### 3.2.4 DA Mass from Atmospheric Parameters
The dM star does not provide a particularly precise DA mass estimate. If one knows the surface gravity, $`g`$, and the radius, $`R`$, the mass can be obtained through
$$M=gR^2/G$$
(3)
where G is the gravitational constant. The quantity $`logg`$ comes from analysis of the line profiles in spectra. Recent determinations include: Marsh et al. (1997), $`logg=7.53\pm 0.09`$; Kidder (1991), $`logg=7.45\pm 0.51`$; Vennes et al., (1997), $`logg=7.2\pm 0.07`$; Finley Koester & Basri, (1997), $`logg=7.17\pm 0.15`$; and Barstow Hubeny & Holberg (1998), $`logg=7.36\pm 0.12`$. The full range of the measures and Equation 3 yield the range of mass values $`0.21_{DA}0.47_{\mathrm{}}`$. Applying the limit, $`_{DAKepler}0.44_{\mathrm{}}`$ eliminates nearly all of these mass determinations. In this case our radius and the Kepler limit indicate that $`logg`$ should be at the high end of these measures.
### 3.3 The White Dwarf Composition
We next place Feige 24 on a white dwarf mass-radius diagram (Figure 8). We plot our two independently determined mass ranges against our adopted radius, $`R_{DA}=0.0185\pm 0.0008R_{\mathrm{}}`$. We represent the radius error by the two horizontal long-short dashed lines. The top thick horizontal bar shows the $`_{DA}`$ determined from atmospheric parameters. Only the largest $`logg`$ at the largest radius produces masses in excess of the Keplerian limit. The thick bar at $`R_{DA}=0.0185R_{\mathrm{}}`$ indicates the $`_{DA}`$ range derived through the dM mass estimates. For this determination the mass error bars indicate the range of ages and \[M/H\] discussed in section 3.2.2. For any dM older than 1-2 Gy the lower masses are associated with lower metallicity. The vertical bold dotted line shows the lowest possible $`_{DA}`$ that can produce the observed VT 94 radial velocity amplitudes for an edge-on orientation of this binary system. We also plot several values of $`logg`$ (dashed) and $`\gamma _{grav}`$ (thin solid). The wide grey curves in Figure 8 are C and He DA models from Vennes et al., (1995). While uncertain, a carbon core DA seems more likely than a pure He core DA.
## 4 Discussion
While our estimated dM and DA masses differ little from VT94, our DA radius differs substantially. VT94 note the difference between their minimum radius, $`R_{DA}=0.028R_{\mathrm{}}`$, and that predicted by the Dahn et al., (1988) parallax. This discrepancy is exacerbated by the two new parallax determinations (HST and HIPPARCOS), folded into our weighted average parallax.
VT94 derive a DA gravitational redshift, $`\gamma _{grav}=8.7\pm 2kms^1`$ from the measured mean velocities for the dM and DA. Combined with our $`R_{DA}=0.0185R_{\mathrm{}}`$, this $`\gamma _{grav}`$ suggests a forbidden DA mass, $`_{DA}0.3_{\mathrm{}}`$. Reducing the mass of the DA component could reconcile the VT94 $`logg`$ and $`\gamma _{grav}`$ with our radius.
We speculate that a third component in the Feige 24 system, a low-mass companion to the DA star, could preserve the total system mass and lower the DA mass. If all components are coplanar, the VT94 DA radial velocities apply strict limits to this reconciliation, because too high a mass for component C would show up as large residuals. We estimate from the scatter that a radial velocity amplitude of $`\pm `$10 km s<sup>-1</sup>could ’hide’ in the VT94 DA radial velocity measurements. Stellar dynamics applies yet another constraint. Holman & Wiegert (1999) parameterize the stability of tertiary companions as a function of stellar component A and B mass function, $`\mu =_A/(_A+_B)`$, and AB binary orbit ellipticity, e. With e=0 and $`\mu =0.39`$ we find (from their table 3) that component C must have an orbital semi-major axis less than 0.3 times that of AB.
Insisting that $`_{DA}=0.30_{\mathrm{}}`$ (this mass - with our radius - would produce the upper limit VT94 $`\gamma _{grav}=10.7`$ km s<sup>-1</sup>) requires $`_C=0.14_{\mathrm{}}`$ ($`_A+_C=0.44_{\mathrm{}})`$. To hide the C component from the radial velocity technique requires a very low AC inclination, nearly face-on. However, non-coplanarity reduces the size of the stable AC semi-major axis even further (Weigert & Holman, 1997; Pendleton & Black (1983)). As an example suppose component C must have an orbital semi-major axis of 0.1 or less that of AB to insure stability. An AC period, P = 0.18<sup>d</sup> (4.3<sup>h</sup>), and i = 6° would produce a radial velocity signature of about $`\pm 10`$ km s<sup>-1</sup>. Finally, the Mass-Luminosity Relation of Henry et al. (1999) would predict $`M_V^C=14.0`$, hence, $`V_C17.2`$, likely undetectable in any of the spectra analyzed for radial velocities. Have we built a new CV, one that should evidence mass transfer and all the associated phenomena? A recent review of CVs (Beuermann, (1999)) indicates that the putative component C ($`_C=0.14_{\mathrm{}}`$) would have to orbit much closer (P$`1.5^h`$) to the DA primary before filling its Roche lobe and producing the characteristic signature of a CV.
Finally we note that our radius differs little from that derived by VT94 from the only trigonometric parallax then available (Dahn et al., (1988)). The unresolved inconsistency between radii (derived from direct parallaxes) and surface gravities (derived from minimum mass and those radii) illuminates the need for high angular resolution observations and direct mass determinations.
The Feige 24 DA mass will rest on an age- and metallicity-dependent lower main sequence Mass- Luminosity Relationship or still uncertain $`logg`$ measurements until the component separations are measured directly. Resolving the inconsistencies between the DA mass estimates (involving dM stellar models and uncertain temperatures, $`logg`$, and bolometric corrections) requires astrometry, both to further reduce the parallax uncertainty, and, more importantly, to spatially resolve this system. Astrometrically derived orbital parameters will provide unambiguous and precise mass determinations for both components. They may also offer insight regarding past and future component interactions.
This system and dozens more like it are ideal targets for the Space Interferometry Mission (http://sim.jpl.nasa.gov). Feige 24, at a distance of 69 pc with P = 4.23<sup>d</sup>, has a total component separation on order 700 microarcsec. The component orbits are much larger than the expected SIM measurement limits. Because shortward of 700 nm 70-80% of the system flux is contributed by the DA (Thorstensen et al. 1978), the wide SIM bandpass and spectral resolution should allow measurement of positions, magnitudes, and colors for both components, even with $`\mathrm{\Delta }V2`$.
Once launched SIM will provide crucial astrometry for this and similar systems at ten times the distance (determined by target magnitude, not astrometric precision). SIM measurements of this system along with many other binaries will provide data with which to create an age- and metallicity-dependent Mass-Luminosity Relationship of exquisite accuracy.
## 5 Conclusions
1. The weighted average of three independent parallax mesurements yields a distance to the dM + DA binary Feige 24 with $`\sigma _D/D`$ = 2.8%. D = $`68.4_{1.9}^{+2.0}`$ pc.
2. We estimate the radius of the DA component using two methods. The first requires either a model-dependent bolometric correction, or one that derives from hot, normal stars. The second utilizes a model-dependent V-band average flux, $`H_V`$. The two results agree within their errors and yield a weighted average $`R_{DA}=0.0185\pm 0.0008R_{\mathrm{}}`$, where the most significant contributions to the error are the uncertain $`T_{eff}^{DA}`$ and B.C.. This radius is larger than any of the WD discussed in Provencal et al. (1998).
3. FGS photometry provides quantitative evidence that the orbital light curve is caused by heating of the dM component by the DA. That signature is consistent with the assumed temperature and the radius we have derived for the white dwarf
4. The VT94 measured radial velocity amplitudes, amplitude ratios, and the assumption of Keplerian circular motion exclude $`_{DAKepler}<0.44_{\mathrm{}}`$ and $`_{dMKepler}<0.26_{\mathrm{}}`$.
5. We estimate the dM component mass, $`0.26<_{dM}<0.43_{\mathrm{}}`$, from the Baraffe et al., (1998) stellar evolution models, a lower limit from Keplerian circular orbits, and the VT 94 radial velocities. The upper range is due to unknown age and metallicity, \[M/H\]. A DA mass range ($`0.44<_{DA}<0.68_{\mathrm{}}`$) follows directly from the VT94 radial velocity amplitudes.
6. We determine $`_{DA}`$ from our $`R_{DA}`$ and a rather wide range of spectroscopically determined $`logg`$ values. This approach yields $`0.44_{DA}0.47_{\mathrm{}}`$, where again the lower limit is imposed by $`_{DAKepler}>0.44_{\mathrm{}}`$.
7. We plot these DA component mass ranges on the $`R`$ plane. With the assistance of the hard lower mass limit and C and He DA $`R`$ models from Vennes et al., (1995), we identify Feige 24 to have a carbon core. A pure He core DA seems less likely.
8. Noting that our radius and the minimum possible $`_{DA}`$ are inconsistent with the VT94 $`\gamma _{grav}`$, we explore the possibility of a tertiary component. A component C, orbiting a common center of mass with the DA, having a period in the range 1.5 $`<`$P $`<5^h`$ with the orbit plane nearly face-on, could reduce the DA mass to $`_{DA}=0.30_{\mathrm{}}`$ and not produce any observational evidence.
8. SIM will be able to measure the orbits of each known component and provide directly measured dynamical masses for both. Orbit size and precise shape may provide information on the nature of past and future interactions between the two components. SIM would also detect a tertiary, if present.
This research has made use of NASA’s Astrophysics Data System Abstract Service and the SIMBAD Stellar Database inquiry and retrieval system. We gratefully acknowledged web access to the astrometry and photometry in USNO-A2.0, provided by the United States Naval Observatory, Flagstaff Station. Support for this work was provided by NASA through grant GTO NAG5-1603 from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. We thank Don Winget and Sandi Catalán for discussions and draft paper reviews. Denise Taylor provided crucial scheduling assistance at the Space Telescope Science Institute. Travis Metcalfe kindly provided code for JD to HJD corrections. We thank an anonymous referee for suggestions that enhanced the clarity of our presentation, and for giving us the courage to speculate. |
warning/0001/hep-ph0001278.html | ar5iv | text | # Production of bound triplet 𝜇⁺𝜇⁻ system in collisions of electrons with atoms
## I Introduction
It is known that $`\mu ^+\mu ^{}`$ atoms exist in two spin states: paradimuonium (PM, singlet state $`n^1S_0`$) with lifetime in the ground state $`\tau _0=0.610^{12}`$ s and orthodimuonium (OM, triplet state $`n^3S_1`$) with lifetime in the ground state $`\tau _1=1.810^{12}`$ s. The dominant decay processes are
$$\mathrm{PM}\gamma \gamma ,\mathrm{OM}e^+e^{}.$$
(1)
The main physical motivation to study dimuonium production lies in the fact that dimuonium is one of the simplest hydrogenlike atom, that is very convenient for testing fundamental laws. Up to now there is a lot of theoretical predictions on dimuonium properties (for a review see, for example, proceedings ) but dimuonium has not been observed yet.
A promising method to create dimuonium — its production at relativistic heavy ion colliders — has been recently considered in Ref. . Another attractive possibility is the electroproduction of dimuonium on atoms. The PM electroproduction cross section was calculated in Ref. in the main logarithmic approximation, and discussed in more detail in Ref. . In these two papers a two-photon mechanism was considered where the incident electron and the nucleus emit each virtual photons which then collide to produce a C-even PM state.
In Ref. the electroproduction of C-odd OM state on atoms was assumed to proceed from the bremsstrahlung mechanism of Fig. 1. The obtained spectrum of OM production on nucleus of charge $`Ze`$ is
$$\frac{d\sigma _{\mathrm{br}}(x)}{dx}=\sigma _0f_{\mathrm{br}}(x),x=\frac{E_{\mu \mu }}{E_e},\sigma _0=\frac{\zeta (3)}{4}\frac{\alpha ^5(Z\alpha )^2}{m_\mu ^2},$$
(2)
$$f_{\mathrm{br}}(x)=\frac{x(1x)}{(1x+\epsilon )^2}\left(1x+\frac{1}{3}x^2\right)\left[\mathrm{ln}\frac{(1x)^2E_e^2}{(1x+\epsilon )m_\mu ^2}\mathrm{\hspace{0.17em}1}\right],\epsilon =\frac{m_e^2}{4m_\mu ^2}$$
where $`E_e`$ and $`E_{\mu \mu }`$ are the energies of electron and OM respectively, and $`m_e`$ and $`m_\mu `$ are the electron mass and the muon mass respectively; $`\alpha 1/137`$ and $`\zeta (3)=1.202\mathrm{}`$ . The obtained cross section has the form
$$\sigma _{\mathrm{br}}=7.16\sigma _0(L2.73),L=\mathrm{ln}\frac{E_e}{2m_\mu }$$
(3)
(note that this cross section is positive only for $`E_e>3.2`$ GeV).
However, the analysis of Ref. is incomplete since it did not take into account the important three-photon mechanism depicted in Fig. 2. Our aim in the present paper is to improve this previous analysis by estimating the production rate provided by the latter mechanism and by pointing out some of its main features. As we shall show, the three-photon process competes with the bremsstrahlung one and even predominates in the case of high-energy electron scattering by heavy atom target.
## II Calculation of three-photon orthodimuonium production
At high electron energy ($`E_em_\mu `$) the cross section $`d\sigma _{3\gamma }`$ corresponding to the diagram of Fig. 2 can be calculated using the equivalent-photon approximation. In this approximation the cross section $`d\sigma _{3\gamma }`$ is expressed as the product of the number of equivalent photons $`dn_\gamma (x,Q^2)`$ generated by the electron, by the cross section $`d\sigma _{\gamma Z}`$ for the real photoproduction of OM on the nucleus
$$d\sigma _{3\gamma }=dn_\gamma \sigma _{\gamma Z},dn_\gamma (x,Q^2)=\frac{\alpha }{\pi }\frac{dx}{x}\frac{dQ^2}{Q^2}\left[1x+\frac{1}{2}x^2(1x)\frac{Q_{\mathrm{min}}^2}{Q^2}\right],$$
$$Q^2=q^2,Q_{\mathrm{min}}^2=\frac{x^2m_e^2}{1x}.$$
(4)
After integration of $`dn_\gamma (x,Q^2)`$ over $`Q^2`$ from $`Q_{\mathrm{min}}^2`$ up to $`Q^2m_\mu ^2`$ we obtain the energy spectrum of equivalent photons
$$\frac{dn_\gamma (x)}{dx}=\frac{\alpha }{\pi }\frac{1}{x}\left[\left(1x+\frac{1}{2}x^2\right)\mathrm{ln}\frac{(1x)m_\mu ^2}{x^2m_e^2}\mathrm{\hspace{0.17em}1}+x\right].$$
(5)
The accuracy of this expression is a logarithmic one, i.e. the omitted items are of the order of $`1/l1/15`$ where $`l=\mathrm{ln}[(1x)m_\mu ^2/(x^2m_e^2)]`$.
The photoproduction cross section $`\sigma _{\gamma Z}`$ can be found in Ref. . At high photon energy ($`xE_em_\mu `$), the corresponding amplitude for photoproduction of OM in $`n^3S_1`$ state on nucleus of mass $`M_Z`$ takes the form
$$M_{\gamma Z}=4i\alpha ^4Z^2\frac{2xE_eM_Z}{n^{1/3}}\frac{F(𝐤_1^2)F(𝐤_2^2)}{𝐤_1^2𝐤_2^2}𝐞_\gamma .𝐞_{\mathrm{OM}}^{}\times $$
(6)
$$\left[\frac{4m_\mu ^2}{4m_\mu ^2+𝐩_{}^2}\frac{4m_\mu ^2}{4m_\mu ^2+(𝐩_{}2𝐤_1)^2}\right]\delta (𝐤_1+𝐤_2𝐩_{})d^2𝐤_1d^2𝐤_2$$
where $`F(𝐤_i^2)`$ is the nucleus form factor, $`𝐩`$ is the momentum of OM, $`𝐞_\gamma `$ and $`𝐞_{\mathrm{OM}}`$ are the polarization vectors for the equivalent photon and the final OM respectively. Because of the rapid decrease of the nuclear form factor at large transverse momenta, the main contribution to the amplitude comes from values of tranverse momenta such that
$$𝐤_i^2\stackrel{<}{}\frac{1}{r^2}$$
(7)
where $`r^2`$ is the mean square radius of the charge distribution of the nucleus. We use below the parameter $`\mathrm{\Lambda }`$ defined as
$$\frac{1}{6}r^2=\frac{1}{\mathrm{\Lambda }^2},\mathrm{\Lambda }=\frac{405}{A^{1/3}}\mathrm{MeV}$$
(8)
where $`A`$ is the atomic mass number, $`\mathrm{\Lambda }=70`$ MeV for Pb and $`\mathrm{\Lambda }=120`$ MeV for Ca. The cross section $`\sigma _{\gamma Z}`$ is found to be a constant at high photon energy $`xE_em_\mu `$. It was calculated in Ref. by numerical integration of Eq. (6), using a realistic nuclear form factor. After summing up over all $`n^3S_1`$ states of OM, the final result is
$$\sigma _{\gamma Z}=4\pi \alpha \sigma _0B\left(\frac{Z\mathrm{\Lambda }}{m_\mu }\right)^2,B=0.85.$$
(9)
As a result, the shape of the energy spectrum of OM is just that of the equivalent photon spectrum :
$$\frac{d\sigma _{3\gamma }(x)}{dx}=\sigma _0f_{3\gamma }(x),$$
(10)
$$f_{3\gamma }(x)=\left(\frac{Z\alpha \mathrm{\Lambda }}{m_\mu }\right)^2\frac{4B}{x}\left[\left(1x+\frac{1}{2}x^2\right)\mathrm{ln}\frac{(1x)m_\mu ^2}{x^2m_e^2}\mathrm{\hspace{0.17em}1}+x\right].$$
The total cross section is obtained after integration over $`x`$ (from $`x_{\mathrm{min}}=2m_\mu /E_e`$ up to $`x_{\mathrm{max}}=1`$)
$$\sigma _{3\gamma }=4B\sigma _0\left(\frac{Z\alpha \mathrm{\Lambda }}{m_\mu }\right)^2\left[L^2+\left(L\frac{3}{4}\right)\mathrm{ln}\frac{m_\mu ^2}{m_e^2}L\frac{\pi ^2}{6}\frac{1}{8}\right]=$$
$$=3.4\left(\frac{Z\alpha \mathrm{\Lambda }}{m_\mu }\right)^2\sigma _0\left[L^2+9.7(L1)\right].$$
(11)
## III Results and Conclusions
1. Let us compare the energy and angular distributions corresponding to the bremsstrahlung and to the three-photon production of OM respectively. The energy spectra are given by Eqs. (2), (10) or by the curves for $`f_{\mathrm{br}}(x)`$ and $`f_{3\gamma }(x)`$ presented in Fig. 3. The bremsstrahlung function $`f_{\mathrm{br}}(x)`$ has a peak at large $`x`$ ($`x1`$); it depends on the electron energy but not on the target properties. On the other hand, the three-photon function $`f_{3\gamma }(x)`$ has a peak at small $`x`$; it does not depend on the electron energy but it strongly depends on the type of nucleus.
The respective angular distributions are also different. The typical emission angle of OM in bremsstrahlung production was estimated in Ref. as
$$\theta _{\mathrm{br}}\frac{m_\mu }{E_e}$$
(12)
The analogous typical angle in three-photon production is of the order of (see Eq. (6))
$$\theta _{3\gamma }\frac{\mathrm{\Lambda }}{xE_e}$$
(13)
and is thus much larger than $`\theta _{\mathrm{br}}`$ in the range $`x1`$, where the three-photon process dominates.
Taking into account the quite different energy and angular distributions of OM in the two production mechanisms, we can conclude that the interference between bremsstrahlung and three-photon productions should be very small.
2. The total cross sections for the production mechanisms here discussed are given by Eqs. (3), (11) and are presented as functions of the electron energy in Figs. 4 and 5, for Pb target and Ca target respectively. It is clearly seen that the bremsstrahlung mechanism is the most important for electroproduction on light nuclei while the three-photon mechanism is dominant for electroproduction on heavy nuclei.
The knowledge of the total cross section gives us a possibility to estimate the production rate. A detailed procedure to obtain such a number can be found in Ref. where the rate for bremsstrahlung production of OM was estimated. For example, this rate in the case of Pb target is about 3 orthodimuonia per minute for an electron energy $`E_e=10`$ GeV and electron current of $`1`$ mA. According to Fig. 4 the rate provided by the three-photon orthodimuonium production for the same example is 2.5 times larger — about 7 orthodimuonia per minute.
3. Thorought this paper we have considered the electroproduction of OM on nuclei. Let us briefly discuss the electroproduction of OM on atoms where a possible atomic screening has to be taken into account. It is known that the atomic screening becomes important when the minimal momentum transfered to atom $`2m_\mu ^2/E_{\mu \mu }=2m_\mu ^2/(xE_e)`$ becomes comparable with the typical atomic momenta $`m_e\alpha Z^{1/3}`$. In other words, the atomic screening should be taken into account when the electron energy $`E_e`$ becomes of the order of or larger than the characteristic energy
$$\frac{2m_\mu ^2}{m_e\alpha Z^{1/3}}=\frac{6000}{Z^{1/3}}\mathrm{GeV}.$$
(14)
4. In the present paper we have calculated three-photon production of OM due to diagram of Fig. 2 with two photons being exchanged with the nucleus. But the C-odd $`\mu ^+\mu ^{}`$ bound system can also be produced in collision of a photon generated by the electron, with $`4,\mathrm{\hspace{0.33em}6},\mathrm{\hspace{0.33em}8},\mathrm{}`$ photons exchanged with the target. At first sight, this leads to an additional factor $`(Z\alpha )^{n2}`$ for $`n`$ exchanged photons. This is just the case for electroproduction of orthopositronium on nuclei. Since for heavy nuclei the parameter $`Z\alpha `$ is not small (for example, $`Z\alpha =0.6`$ for Pb), the whole series in $`(Z\alpha )^2`$ has to be summed. According to Ref. , the high-order $`(Z\alpha )^2`$ effects decrease the orthopositronium production cross section by about 40 % for the Pb target. However, for OM production we should take into account the restriction of the transverse momenta $`k_1,k_2,\mathrm{},k_n`$ due to nuclear form factors on the level given by Eq. (7). As a result, the effective parameter of the perturbation theory is not $`(Z\alpha )^2`$ but
$$\frac{(Z\alpha )^2}{r^2m_\mu ^2}<0.03$$
(15)
and, therefore, we can restrict ourselves to three-photon production only.
## Acknowledgments
The authors wish to thank Prof. I. F. Ginzburg for useful discussions.
V.G. Serbo is grateful to University Paris VI for a grant that allowed his stay in Paris where this work was completed, and to the Laboratoire de Physique Nucléaire et de Hautes Energies (LPNHE) for warm hospitality. |
warning/0001/physics0001003.html | ar5iv | text | # Tentative statistical interpretation of non-equilibrium entropy
We suggest a certain statistical interpretation for the entropy produced in driven thermodynamic processes. The exponential function of *half* irreversible entropy re-weights the probability of the standard Ornstein-Uhlenbeck-type thermodynamic fluctuations. (We add a proof of the standard Fluctuation Theorem which represents a more natural interpretation.)
In 1910 Einstein , paraphrasing Boltzmann’s lapidary formula $`S=\mathrm{log}W`$, expressed the probability distribution of thermodynamic variables $`x`$ through the entropy function $`S(x)`$:
$$W(x)e^{S(x)}.$$
(1)
This equation describes thermodynamic fluctuations in Gaussian approximation properly. Going beyond the stationary features, the time-dependence of fluctuations $`x_t`$ can be characterized by a certain probability functional $`W[x]`$ over complete paths $`\{x_t;t(\mathrm{},\mathrm{})\}`$. It turns out that, in driven thermodynamic processes, this probability is related to the irreversible entropy $`S_{irr}[x]`$. Symbolically, we can write the following relationship:
$$W[x]W_{OU}[x\overline{x}]e^{S_{irr}[x]/2},$$
(2)
where $`\overline{x}_t`$ is the ‘driving’ value of parameter $`x_t`$ and $`W_{OU}[z]`$ turns out to correspond to fluctuations $`z_t`$ of Ornstein-Uhlenbeck type. This relationship offers $`S_{irr}`$ a certain statistical interpretation, somehow resembling Einstein’s suggestion (1) for the equilibrium entropy $`S(x)`$. In this short note, Einstein’s approach to the thermodynamic fluctuations is outlined and standard equations of time-dependent fluctuations are invoked from irreversible thermodynamics. Then I give a precise form to the relationship (2) for driven thermodynamic processes.
The equilibrium conditions for isolated composite thermodynamic systems derive from the maximum entropy principle:
$$S(x)=max,$$
(3)
where $`S(x)`$ is the total entropy of the system in function of certain free thermodynamic parameters $`x`$ . If the function $`S(x)`$ is maximum at $`x=\overline{x}`$ then $`\overline{x}`$ is the equilibrium state. For example, $`x`$ may be the temperature $`T`$ of a small (yet macroscopic) subsystem in the large isolated system of temperature $`\overline{T}=\overline{x}`$. Then, the function $`S(x)`$ must be the total entropy of the isolated system, depending on the variation of the subsystem’s temperature around its equilibrium value. The equilibrium value $`\overline{x}`$ \[as well as $`S(x)`$ itself\] may vary with the deliberate alteration of the initial conditions. Surely, in our example the temperature $`\overline{T}`$ of the whole isolated system can always be controlled at will. For later convenience, especially in treating driven thermodynamic processes, we may prefer the explicit detailed notation $`S(x|\overline{x})`$ for $`S(x)`$. Though $`S(\overline{x})S(x)`$ might qualify the lack of equilibrium, nearby values $`x\overline{x}`$ have no interpretation in phenomenological thermodynamics. They only have it in the broader context of statistical physics. In finite thermodynamic systems there are fluctuations around the equilibrium state $`\overline{x}`$ and their probability follows Eq. (1):
$$W(x|\overline{x})dx=𝒩e^{S(x|\overline{x})S(\overline{x}|\overline{x})}dx.$$
(4)
Assume, for simplicity, that there is a single free variable $`x`$. The Taylor expansion of the entropy function yields Gaussian fluctuations:
$$W(x|\overline{x})=\frac{1}{\sqrt{2\pi \sigma ^2}}\mathrm{exp}(\frac{1}{2\sigma ^2}(x\overline{x})^2),$$
(5)
where
$$\frac{1}{\sigma ^2}=S^{\prime \prime }(\overline{x})\frac{^2S(x|\overline{x})}{x^2}|_{x=\overline{x}}.$$
(6)
In our concrete example $`\sigma ^2=T^2/C`$ where $`C`$ is the specific heat of the subsystem.
We are going to regard the time-dependence of the parameter $`x_t`$ fluctuating around $`\overline{x}`$, according to the standard irreversible thermodynamics . The time-dependent fluctuation $`z_tx_t\overline{x}`$ is an Ornstein-Uhlenbeck (OU) stochastic process of zero mean $`z_t0`$ and of correlation
$$z_tz_t^{}_{OU}=\sigma ^2e^{ł|tt^{}|}.$$
(7)
The relaxation rate $`ł`$ of fluctuations is related to the corresponding Onsager kinetic constant $`\gamma `$ by $`ł=\gamma /\sigma ^2`$. It can be shown that the probability distribution of $`x_t=z_t+\overline{x}`$ at any fixed time $`t`$ is the Gaussian distribution (5) as it must be. For the probability of the complete fluctuation path $`z_t`$, the zero mean and correlation (7) are equivalent with the following functional:
$$W_{OU}[z]𝒟z=\mathrm{exp}\left(\frac{1}{4\gamma }(\dot{z}_t^2+ł^2z_t^2)𝑑t\right)𝒟z,$$
(8)
where a possible constant of normalization has been absorbed into the functional measure $`𝒟z`$.
In order to construct and justify a relationship like (2) one needs to proceed to driven thermodynamic processes. In fact, we assume that we are varying the parameter $`\overline{x}`$ with small but finite velocity. Formally, the parameter $`\overline{x}`$ becomes time-dependent. For simplicity’s sake we assume that the coefficients $`\sigma ,\gamma `$ do not depend on $`\overline{x}`$ or, at least, that we can ignore their variation throughout the driven range of $`\overline{x}_t`$. We define the irreversible entropy production during the driven process as follows:
$$S_{irr}[x|\overline{x}]=\frac{1}{\sigma ^2}(\overline{x}_tx_t)𝑑x_t.$$
(9)
In our concrete example $`dS_{irr}=(C/T^2)(\overline{T}T)dTdQ(T^1\overline{T}^1)`$ which is indeed the entropy produced randomly by the heat transfer $`dQ`$ from the surrounding to the subsystem. By partial integration, Eq. (9) leads to an alternative form:
$$S_{irr}[x|\overline{x}]=\frac{1}{\sigma ^2}(\overline{x}_tx_t)𝑑\overline{x}_t+\frac{1}{\sigma ^2}(x_{\mathrm{}}\overline{x}_{\mathrm{}})^2\frac{1}{\sigma ^2}(x_{\mathrm{}}\overline{x}_{\mathrm{}})^2.$$
(10)
In relevant driven processes the entropy production is macroscopic, i.e., $`S_{irr}1`$ in $`k_B`$-units, hence it is dominated by the integral term above. I exploit this fact to replace expression (9) by
$$S_{irr}[x|\overline{x}]=\frac{1}{\sigma ^2}(\overline{x}_tx_t)𝑑\overline{x}_t$$
(11)
which vanishes for constant $`\overline{x}`$. In the sense of the guess (2), I suggest the following form for the probability distribution of the driven path:
$$W[x|\overline{x}]=𝒩[\overline{x}]W_{OU}[x\overline{x}]e^{S_{irr}[x|\overline{x}]/2}.$$
(12)
The non-trivial normalizing pre-factor is a consequence of $`\overline{x}`$’s time-dependence and will be derived below. Since the above distribution is a Gaussian functional and $`S_{irr}[x|\overline{x}]`$ is a linear functional (11) of $`x`$, we can easily calculate the expectation value of the irreversible entropy:
$$S_{irr}[\overline{x}]S_{irr}[x|\overline{x}]=\frac{1}{2\sigma ^2}\dot{\overline{x}}_t\dot{\overline{x}}_t^{}e^{ł|tt^{}|}𝑑t𝑑t^{}.$$
(13)
In case of moderate accelerations $`\ddot{\overline{x}}ł\dot{\overline{x}}`$, this expression reduces to the standard irreversible entropy $`\gamma ^1\dot{\overline{x}}_t^2𝑑t`$ of the phenomenological theory of driven processes . Coming back to the normalizing factor in Eq. (12), we can relate it to the mean entropy production (13): $`𝒩[\overline{x}]=exp(S_{irr}[\overline{x}]/4)`$. Hence, the ultimate form of Eq. (12) will be:
$$W[x|\overline{x}]=W_{OU}[x\overline{x}]e^{S_{irr}[x|\overline{x}]/2S_{irr}[\overline{x}]/4}.$$
(14)
This result gives the precise meaning to our symbolic relationship (2). If the entropy production $`S_{irr}`$ were negligible then the thermodynamic fluctuations $`x_t\overline{x}_t`$ would follow the OU statistics (7) like in case of a steady state $`\overline{x}_t=const`$. Even in slow irreversibly driven processes $`S_{irr}`$ may grow essential and $`exp[S_{irr}/2]`$ will re-weight the probability of OU fluctuations. The true stochastic expectation value of an arbitrary functional $`F[x]`$ can be expressed by the OU expectation values of the re-weighted functional:
$$F[x]=F[x]e^{S_{irr}[x|\overline{x}]/2S_{irr}[\overline{x}]/4}_{OU}.$$
(15)
I can verify the plausibility of Eq. (14) for the special case of small accelerations. Let us insert Eqs. (8,11) and also Eq. (13) while ignore $`\ddot{\overline{x}}`$ in comparison with $`ł\dot{\overline{x}}`$. We obtain:
$$W[x|\overline{x}]=W_{OU}[x\overline{x}+ł^1\dot{\overline{x}}].$$
(16)
Obviously, the fluctuations of the driven system are governed by the OU process $`z_t`$ (7) in the equilibrium case when $`\dot{\overline{x}}0`$. In driven process, when $`\dot{\overline{x}}0`$, there is only a simple change: The OU fluctuations happen around the retarded value $`\overline{x}_t\tau \dot{\overline{x}}\overline{x}_{t\tau }`$ of the driven parameter. The lag $`\tau `$ is equal to the thermodynamic relaxation time $`1/ł`$. Consequently, the driven random path takes the following form:
$$x_t=\overline{x}_{t\tau }+z_t,$$
(17)
where $`z_t`$ is the equilibrium OU process (7). This result implies, in particular, the equation $`x_t=\overline{x}_{t\tau }`$ which is just the retardation effect well-known in the thermodynamic theory of slightly irreversible driven processes. For example, in case of an irreversible heating process the subsystem’s average temperature will always be retarded by $`\tau \dot{\overline{T}}`$ with respect to the controlling temperature $`\overline{T}`$ .
Finally, let us summarize the basic features of Einstein’s formula (1) and of the present proposal (2). They characterize the quality of equilibrium in static and in driven steady states, respectively. They do it in terms of thermodynamic entropies while they refer to a statistical context lying outside both reversible and irreversible thermodynamics. Both formulae are only valid in the lowest non-trivial order and their correctness in higher orders is questionable . Contrary to their limited validity, they can no doubt give an insight into the role of thermodynamic entropy in statistical fluctuations around both equilibrium or non-equilibrium states.
Acknowledgments. I thank Bjarne Andresen, Karl Heinz Hoffmann, Attila Rácz, and Stan Sieniutycz for useful remarks regarding the problem in general. This work enjoyed the support of the EC Inco-Copernicus program Carnet 2.
Note added. The first version of this work proposed the relationship
$$WW_{OU}e^{S_{irr}},$$
the exponent was free from the funny factor $`1/2`$. The proof was wrong, of course. With the factor $`1/2`$, my statistical interpretation for $`S_{irr}`$ has become less attractive. I also realized that a more natural statistical interpretation was already discovered before. The present formalism offers the following convenient proof of the Fluctuation Theorem.
The true probability distribution of the slowly driven process is the Onsager–Machlup functional :
$$W_{OM}[x|\overline{x}]𝒟x=\mathrm{exp}\left(\frac{1}{4\gamma }[\dot{x}_t+ł(x_t\overline{x}_t)]^2𝑑t\right)𝒟x,$$
at fixed $`x_{\mathrm{}}`$. Let us imagine the probability distribution of the time-reversed process $`x_t^r=x_t`$ driven by the time-reversed surrounding $`\overline{x}_t^r=\overline{x}_t`$. Formally, we only have to change the sign of $`\dot{x}_t`$, yielding:
$$W_{OM}[x^r|\overline{x}^r]𝒟x=\mathrm{exp}\left(\frac{1}{4\gamma }[\dot{x}_tł(x_t\overline{x}_t)]^2𝑑t\right)𝒟x,$$
at fixed $`x_{\mathrm{}}^r`$. We can inspect that the above distributions of the true and the time-reversed processes, respectively, satisfy the following relationship:
$$\mathrm{log}W_{OM}[x|\overline{x}]\mathrm{log}W_{OM}[x^r|\overline{x}^r]=\frac{ł}{\gamma }(\overline{x}_tx_t)𝑑x_t.$$
Observe that the r.h.s. is the irreversible entropy production $`S_{irr}[x|\overline{x}]`$ of the driven process. This leads to the so-called Fluctuation Theorem:
$$W_{OM}[x^r|\overline{x}^r]=e^{S_{irr}[x|\overline{x}]}W_{OM}[x|\overline{x}].$$
The irreversible entropy turns out to be a concrete statistical measure of the time-reversal asymmetry. |
warning/0001/hep-th0001211.html | ar5iv | text | # Schwarzschild black hole in the dilatonic domain wall
## I Introduction
Recently there has been much interest in the Randall-Sundrum brane world. A key idea of this model is that our universe may be a brane embedded in the higher dimensional space. A concrete model is a single 3-brane embedded in the five-dimensional anti-de Sitter space ($`\mathrm{AdS}_5`$), which acts like a box of the size $`1/k`$. Randall and Sundrum have shown that a longitudinal part ($`h_{\mu \nu }`$) of the metric fluctuations satisfies the Schrödinger-like equation with an attractive delta-function. As a result, the massless KK modes which describe the localized gravity on the brane were found. Furthermore, the massive KK modes lead to corrections to the Newtonian potential like as $`V(r)=G_N\frac{m_1m_2}{r}(1+\frac{1}{k^2r^2})`$.
However, we would like to point out that this has been done in the 4D Minkowski space with the RS gauge<sup>*</sup><sup>*</sup>*In fact, this gauge for $`h_{MN}`$ is composed of Gaussian-Normal (GN) gauge ($`h_{44}=h_{4\mu }=0`$) and 4D transverse, tracefree (TTF) gauge($`^\mu h_{\mu \nu }=0`$, $`h_\mu ^\mu =0`$).. It seems that this gauge is so restrictive. In order to have well-defined theory on the brane, one has to include non-zero transverse parts of $`h_{4\mu }`$ and $`h_{44}`$ at the beginning. Ivanov and Volovich discussed along this direction by choosing the 5D de Donder gaugeThis corresponds to the 5D TTF gauge ($`_Mh^{MN}=0,h_M^M=0`$).. Also authors in Ref. studied the propagation of the metric including with non-zero transverse parts. It turned out that there are no massless scalar and vector propagations on the RS background. This implies that the RS gauge is a good choice for describing the brane world. Furthermore, it is shown that the RS Minkowski spacetime is stable only under the RS gauge.
If the Minkowski metric on the brane is replaced by any 4D Ricci-flat one, for example the Schwarzschild metric, the 5D metric still satisfies the Einstein equation with a negative cosmological constant. In this process we obtain a black string solution in an AdS<sub>5</sub> Precisely, this is not a 5D black hole solution to the RS model. At present, nobody knows that.. This is given by $`\overline{G}_{MN}^{\mathrm{BS}}=H^2(z)[\overline{g}_{\mu \nu }^S,1]`$ with the Schwarzschild metric $`\overline{g}_{\mu \nu }^S=\mathrm{diag}[(12M/r),(12M/r)^1,r^2,r^2\mathrm{sin}^2\theta ]`$. It is shown that this black string is unstable near the AdS<sub>5</sub> horizon of $`z=\mathrm{}`$ but it is stable far from the horizon. This is a result of the Gregory-Laflamme instability, which states that the black string has a tendency to fragment near the AdS<sub>5</sub>-horizon. A stable object left behind would resemble a black cigar, although we do not know its explicit metric. Hence we have the RS black hole-picture on the brane located at $`z=0`$. The horizon of black hole on the brane will be determined from the intersection of the black cigar with the brane. Here we assume that this is large such as $`r_{EH}=2M>1/k`$. In this case an observer on the brane perceives exactly the Schwarzschild solution, without any correction arising from the extra dimension. In other words, any massive KK mode are not excited in this circumstances . Then the zero modes of the bulk graviton can describe this situation very well. Hence the zero mode approach becomes a powerful technique in the study of black holes on the brane.
On the other hand, Youm showed that the RS solution can be found in the dilatonic domain wall. In order to study physics on the brane, we need its effective action. It is known that zero mode effective action takes a form of the Brans-Dicke model with a potential.
In this paper, we study stability of the large RS black hole with the $`z`$-independent perturbations such as $`h_{\mu \nu }(x)`$, $`h_{44}(x)`$. We find that these are massless graviton and scalar modes propagating in the RS black hole background. Here we do not require the 4D TTF gauge which is useful for the RS Minkowski space. Instead we choose the Regge-Wheeler (RW) gauge for our study of spherically symmetric background.
## II Randall-Sundrum solution and Brans-Dicke type model
We start with the 5D bulk action and the 4D domain wall action as
$$S=S_{\mathrm{bulk}}+S_{\mathrm{DW}}$$
(1)
with
$`S_{\mathrm{bulk}}`$ $`=`$ $`{\displaystyle \frac{1}{2\kappa _5^2}}{\displaystyle d^5x\sqrt{G}\left[R_5\frac{4}{3}_MD^MDe^{2aD}\mathrm{\Lambda }\right]},`$ (2)
$`S_{DW}`$ $`=`$ $`\sigma _{\mathrm{DW}}{\displaystyle d^4x\sqrt{\gamma }e^{aD}},`$ (3)
where $`\sigma _{\mathrm{DW}}`$ is the tension of the domain wall and $`\gamma `$ is the determinant of the induced metric $`\gamma _{\mu \nu }=_\mu X^M_\nu X^NG_{MN}`$ for the domain wall. Here $`M,N=0,1,2,3,4(x^4=z)`$ and $`\mu ,\nu =0,1,2,3(x^\mu =x)`$. “$`D`$” denotes the dilaton. We are interested in the second RS solution with<sup>§</sup><sup>§</sup>§With this ansatz, the dilatonic action Eq.(1) reduces to the second RS model exactly
$$\overline{G}_{MN}=H^2(z)\eta _{MN},\overline{D}=0,\mathrm{\Lambda }=12k^2,\sigma _{\mathrm{DW}}=6k/\kappa _5^2,a=0$$
(4)
with $`H=k|z|+1`$ and $`\eta _{MN}=\mathrm{diag}[++++]`$. Here overbar(<sup>-</sup>) means the background value. In this paper we follow the MTW conventions.
In order to obtain a 4D effective action, we introduce the metric $`\widehat{G}_{MN}`$ in $`G_{MN}=H^2(z)\widehat{G}_{MN}`$ which satisfies $`_z\widehat{G}_{MN}=0`$. Explicitly, the line element is given by
$`dS_5^2`$ $`=`$ $`G_{MN}dx^Mdx^N=H^2\widehat{G}_{MN}dx^Mdx^N`$ (5)
$`=`$ $`H^2\left[g_{\mu \nu }(x)dx^\mu dx^\nu +\mathrm{\Phi }^2(x)dz^2\right].`$ (6)
Off-diagonal elements are not allowed because if they exist, these violate the $`Z_2`$-symmetry argumentOne may introduce off-diagonal term of $`2A_\mu (x)dx^\mu dz`$. But this is not invariant under the reflection of $`zz`$ . Thus we must have $`A_\mu (x)=0`$.. At this stage we wish to remind the reader that $`\mathrm{\Phi }(x)`$ has nothing to do with the radion which is necessary for stabilizing the distance between two branes in the first RS model. This is because here we consider the second RS model. Substituting (6) with $`D=0`$ into (1) and integrating it over $`z`$ lead to the Brans-Dicke type model with a potential
$$S_{\mathrm{RS}}=\frac{1}{2\kappa _4^2}d^4x\sqrt{g}\left[\mathrm{\Phi }R+6k^2\left(\mathrm{\Phi }+\frac{1}{\mathrm{\Phi }}2\right)\right]$$
(7)
with $`\kappa _4^2=k\kappa _5^2`$. This is our key action which is suitable for the study of physics defined on the brane. Also it can be derived from the RS model directly. It is emphasized that the first term($`\mathrm{\Phi }R`$) is the BD term with $`\omega =0`$. The BD model with $`\omega =0`$ corresponds to the massless Kaluza-Klein model with $`g_{\mu \nu }`$, $`g_{44}(\mathrm{\Phi }^2)`$, and $`g_{\mu 4}(A_\mu )=0`$ This model is based on the action of $`S_{\mathrm{KK}}=\frac{1}{\kappa _5^2}d^5x\sqrt{G}R_5`$ with a factorizable geometry of $`H=1`$.. In this sense, we wish to call $`\mathrm{\Phi }`$ as the BD scalar. The second term arises from the fact that the 5D spacetime is an AdS<sub>5</sub> with a negative cosmological constant $`\mathrm{\Lambda }`$ and a domain wall located at $`z=0`$. Equivalently, this means that the RS solution gives us a non-factorizable geometry with $`H=k|z|+1`$. Hence this term accounts for the feature of the RS type solution and it plays the role of an effective potential. ¿From now on we wish to separate the pure BD model ($`_{\mathrm{BD}}=\mathrm{\Phi }R`$) from the RS model ($`_{\mathrm{RS}}=\mathrm{\Phi }R+6k^2(\mathrm{\Phi }+1/\mathrm{\Phi }2`$)) for comparison.
¿From (7) we derive the equations of motion ($`\delta _\mathrm{\Phi }S_{\mathrm{RS}}=0,\delta _{g_{\mu \nu }}S_{\mathrm{RS}}=0`$)
$`R+6k^2\left(1{\displaystyle \frac{1}{\mathrm{\Phi }^2}}\right)=0,`$ (8)
$`R_{\mu \nu }{\displaystyle \frac{1}{2}}g_{\mu \nu }R={\displaystyle \frac{1}{\mathrm{\Phi }}}\left\{_\mu _\nu \mathrm{\Phi }g_{\mu \nu }\mathrm{}\mathrm{\Phi }+3k^2\left(\mathrm{\Phi }+{\displaystyle \frac{1}{\mathrm{\Phi }}}2\right)g_{\mu \nu }\right\}.`$ (9)
Contracting Eq.(9) with $`g^{\mu \nu }`$ and using Eq.(8) leads to the RS scalar equation
$$\mathrm{}\mathrm{\Phi }+2k^2\left(\mathrm{\Phi }\frac{3}{\mathrm{\Phi }}+4\right)=0.$$
(10)
Also from Eq.(9), its contraction form, and (10) one finds the other Einstein equation
$$R_{\mu \nu }=\frac{1}{\mathrm{\Phi }}\left\{_\mu _\nu \mathrm{\Phi }+2k^2\left(\mathrm{\Phi }+1\right)g_{\mu \nu }\right\}.$$
(11)
Here one finds a solution which satisfies all of equations (8)-(11) simultaneously as
$$\overline{\mathrm{\Phi }}=1,\overline{R}=0,\overline{R}_{\mu \nu }=0.$$
(12)
This means that the Ricci-flat condition of $`\overline{R}_{\mu \nu }=0`$ with $`\overline{\mathrm{\Phi }}=1`$ describes the 4D vacuum configuration correctly.
As an example, we choose the Minkowski spacetime
$$\overline{g}_{\mu \nu }=\eta _{\mu \nu },\overline{\mathrm{\Phi }}=1.$$
(13)
To study the propagations on this background, we introduce the perturbation around the background (13) as
$$g_{\mu \nu }=\eta _{\mu \nu }+h_{\mu \nu },\mathrm{\Phi }=\overline{\mathrm{\Phi }}+\phi (g_{44}=\overline{g}_{44}+h_{44}).$$
(14)
Then the linearized equations to (10) and (11) are found to be
$`^2\phi +4k^2\phi =0,`$ (15)
$`\delta R_{\mu \nu }(h)_\mu _\nu \phi +2k^2\phi \eta _{\mu \nu }=0`$ (16)
with
$$\delta R_{\mu \nu }=\frac{1}{2}\left[^2h_{\mu \nu }+_\nu _\mu h_\rho ^\rho ^\rho _\mu h_{\nu \rho }^\rho _\nu h_{\mu \rho }\right].$$
(17)
Under the 4D TTF gauge, one finds that $`\delta R_{\mu \nu }=\frac{1}{2}^2h_{\mu \nu }`$ and $`\eta ^{\mu \nu }\delta R_{\mu \nu }=\frac{1}{2}^2h_\mu ^\mu =0.`$ Contracting Eq.(16) with $`\eta ^{\mu \nu }`$ leads to the other equation for $`\phi `$
$$^2\phi 8k^2\phi =0.$$
(18)
Eq.(15) allows a tachyonic solution because it has a negative potential term of $`4k^2`$. Fortunately we resolve this problem. We have two different equations (15) and (18) for the same field of $`\phi `$. Hence we require $`\phi =0`$ for consistency. This observation agrees with Refs.. Then Eq.(16) reduces to an equation for the massless graviton without a matter source on the brane
$$^2h_{\mu \nu }=0.$$
(19)
## III Schwarzschild black hole solutions
Introducing a spherically symmetric spacetime, one obtains the Schwarzschild black hole with $`\overline{\mathrm{\Phi }}=1`$ in the domain wall approach as
$`\overline{g}_{\mu \nu }=\mathrm{diag}[e^\nu ,e^\nu ,r^2,r^2\mathrm{sin}^2\theta ],`$ (20)
$`\overline{R}_{trtr}=2{\displaystyle \frac{M}{r^3}},\overline{R}_{t\theta t\theta }={\displaystyle \frac{(r2M)M}{r^2}},\overline{R}_{t\varphi t\varphi }=\mathrm{sin}^2\theta \overline{R}_{t\theta t\theta },`$ (21)
$`\overline{R}_{r\theta r\theta }={\displaystyle \frac{M}{r2M}},\overline{R}_{r\varphi r\varphi }=\mathrm{sin}^2\theta \overline{R}_{r\theta r\theta },\overline{R}_{\theta \varphi \theta \varphi }=2Mr\mathrm{sin}^2\theta ,`$ (22)
$`\overline{R}_{\mu \nu }=0,\overline{R}=0`$ (23)
with
$$e^\nu =1\frac{2M}{r}.$$
(24)
Here $`M=G_NM_4=\kappa _4^2M_4/8\pi =G_5kM_4`$ with $`G_N`$ (4D Newtonian constant) and $`G_5`$ (5D Newtonian constant). $`M_4`$ is the mass of a large black hole with $`M_4>1/(2k^2G_5)`$. The BD and RS black hole solutions are permitted because the Schwarzschild solution comes from the Ricci-flat condition. The BD (RS) black holes denote the Schwarzschild black hole with $`_{\mathrm{BD}}(_{\mathrm{RS}})`$, respectively. The BD black is introduced for reference. To study these black holes specifically, we introduce the perturbation
$$g_{\mu \nu }=\overline{g}_{\mu \nu }+h_{\mu \nu },\mathrm{\Phi }=\overline{\mathrm{\Phi }}+\phi .$$
(25)
Then the linearized equations to Eqs. (10) and (11) are found as
$`\overline{\mathrm{}}\phi +4k^2\phi =0,`$ (26)
$`\delta R_{\mu \nu }(h)\overline{}_\nu \overline{}_\mu \phi +2k^2\phi \overline{g}_{\mu \nu }=0`$ (27)
with the Lichnerowicz operator $`\delta R_{\mu \nu }(h)`$
$$\delta R_{\mu \nu }(h)=\frac{1}{2}\left[\overline{\mathrm{}}h_{\mu \nu }+\overline{}_\nu \overline{}_\mu h_\rho ^\rho \overline{}^\rho \overline{}_\mu h_{\nu \rho }\overline{}^\rho \overline{}_\nu h_{\mu \rho }\right].$$
(28)
We note that Eq. (27) is not a diagonalized form to obtain eigenmodes. If one introduces $`\widehat{h}_{\rho \nu }=h_{\rho \nu }+\phi \overline{g}_{\rho \nu }`$, then this leads to
$$\delta R_{\mu \nu }(\widehat{h})=0.$$
(29)
This is the perturbed equation of pure 4D gravity for $`\widehat{h}_{\mu \nu }`$<sup>\**</sup><sup>\**</sup>\**Its stability analysis was performed 30 years ago .. A way to analyze Eqs.(26) and (27) is known . For instance, it is possible if one uses the RW gauge instead of the 4D TTF gauge. The perturbation $`\widehat{h}_{\mu \nu }`$ falls into two distinct classes - odd and even parities with $`(1)^{l+1}`$ and $`(1)^l`$, respectively. $`l`$ denotes an angular quantum number on $`S^2`$: $`\overline{L}^2Y_{lm}(\theta ,\varphi )=l(l+1)Y_{lm}(\theta ,\varphi )`$. Among ten components in the axisymmetric perturbation, one can always choose six components by taking into account the general coordinate transformations : $`x_{}^{\mu }{}_{}{}^{}=x^\mu +ϵ\xi ^\mu `$. This is a choice of the RW gauge. And this is obvious here because we consider the propagation of gravitons on the brane. In the RW gauge we assign two components ($`h_0,h_1`$) for odd parity
$$\widehat{h}_{\mu \nu }^{\mathrm{odd}}=\left(\begin{array}{cccc}0& 0& 0& h_0(r)\\ 0& 0& 0& h_1(r)\\ 0& 0& 0& 0\\ h_0(r)& h_1(r)& 0& 0\end{array}\right)e^{i\omega t}\mathrm{sin}\theta \frac{dP_l(\theta )}{d\theta }$$
(30)
with Legendre polynomial $`P_l(\theta )`$. For even parity, we have $`\widehat{h}_{\mu \nu }^{even}=h_{\mu \nu }^{even}+\phi \overline{g}_{\mu \nu }`$, where $`h_{\mu \nu }^{even}`$ is composed of four components ($`H_0,H_1,H_2,K`$) as
$$h_{\mu \nu }^{\mathrm{even}}=\left(\begin{array}{cccc}H_0\left(1\frac{2M}{r}\right)& H_1& 0& 0\\ H_1& H_2\left(1\frac{2M}{r}\right)^1& 0& 0\\ 0& 0& r^2K(r)& 0\\ 0& 0& 0& r^2\mathrm{sin}^2\theta K(r)\end{array}\right)e^{i\omega t}P_l(\theta ).$$
(31)
These two cases are never mixed and thus they provide two degrees of freedom which is necessary for describing a massless spin-2 particle.
### A RS scalar perturbation
Let us first analyze the RS scalar perturbation using Eq.(26). Considering $`\phi \frac{\psi (r)}{r}Y_{lm}(\theta ,\varphi )e^{i\omega t}`$ and the background (20), one finds the Schrödinger-type equation
$$\frac{d^2\psi }{dr_{}^{}{}_{}{}^{2}}+\left(\omega ^2V_\psi ^{\mathrm{RS}}\right)\psi =0,$$
(32)
where the RS scalar potential is given by
$$V_\psi ^{\mathrm{RS}}(r)=\left(1\frac{2M}{r}\right)\left\{\frac{l(l+1)}{r^2}+\frac{2M}{r^3}4k^2\right\}$$
(33)
with the tortoise coordinate $`r^{}=r+2M\mathrm{ln}\left(\frac{r}{2M}1\right)`$. ¿From the analysis in ref., we find that for $`\omega =i\alpha ,0<\alpha <2k`$, the scalar perturbation has an exponentially growing mode of $`e^{\alpha t}`$. Therefore this system may be classically unstable. In other words, “$`4k^2`$” in the potential induces an instability of the asymptotically flat space of $`r\mathrm{}`$. Thus we call it as a potential instability. This may imply that the RS black hole solution is classically unstable. However, it is not true. Any exponentially growing mode is not allowed for the RS spherically symmetric background. An important point what we remind is that an AdS<sub>5</sub> acts like a box of size<sup>††</sup><sup>††</sup>††Although a conventional length scale of an AdS<sub>5</sub> is determined by $`d^5x\sqrt{G}R_5`$ in Eq.(2) as $`𝑑zH^3=1/k`$, we here choose a size of the AdS<sub>5</sub>-box as $`1/(2k)`$ for a definite calculation. This comes from :$`\mathrm{\Lambda }d^5x\sqrt{G}𝑑zH^5=1/(2k)`$. $`(2k)^1`$ . So it may allow an unstable perturbation of wavelength with $`\lambda (2k)^1`$ only. However, for this case, we cannot find any consistent (unstable) solution. In the unstable case of $`0<\alpha =\lambda ^1<2k`$, one finds a condition of $`\lambda >(2k)^1`$ which is forbidden inside an AdS<sub>5</sub>. The instability problem can be cured by considering the bulk spacetime. Hence we can include the RS scalar mode, as a physical field, which propagates in the RS black hole background.
### B Odd parity perturbation for $`\widehat{h}_{\mu \nu }`$
Now we discuss the odd parity perturbation for $`\widehat{h}_{\mu \nu }`$. ¿From Eq.(27), we have three equations :
$$\delta R_{03}=\delta R_{13}=\delta R_{23}=0.$$
(34)
Here we obtain the Regge-Wheeler equation using $`Q\frac{h_1}{r}\left(1\frac{2M}{r}\right)`$
$$\frac{d^2Q}{dr_{}^{}{}_{}{}^{2}}+\left(\omega ^2V_{\mathrm{RW}}(r)\right)Q=0,$$
(35)
where the Regge-Wheeler potential is given by
$$V_{\mathrm{RW}}(r)=\left(1\frac{2M}{r}\right)\left(\frac{l(l+1)}{r^2}\frac{6M}{r^3}\right).$$
(36)
Further $`h_0`$ is not an independent mode, it can be expressed in terms of $`Q`$ as $`h_0=\frac{i}{\omega }\frac{d}{dr^{}}(rQ)`$. We note that the BD black hole takes the same potential as in Eq.(36) for the odd parity perturbation. This case has already been analyzed by Vishveshwara in ref. and an allowed solution is a scattering state
$`Q_{\mathrm{}}`$ $`=`$ $`e^{i\omega r^{}}+A^{}e^{i\omega r^{}}(r^{}\mathrm{}),`$ (37)
$`Q_{2M}`$ $`=`$ $`B^{}e^{i\omega r^{}}(r^{}\mathrm{}).`$ (38)
### C Even parity perturbation for $`\widehat{h}_{\mu \nu }`$
In this case, from the remaining seven equations, we have the Zerilli equation in ref.
$$\frac{d^2\psi _\mathrm{Z}}{dr_{}^{}{}_{}{}^{2}}+\left(\omega ^2V_\mathrm{Z}(r)\right)\psi _\mathrm{Z}=0,$$
(39)
where the Zerilli potential is given by
$$V_\mathrm{Z}(r)=\left(1\frac{2M}{r}\right)\left\{\frac{2\lambda ^2(\lambda +1)r^3+6\lambda ^2Mr^2+18\lambda M^2r+18M^3}{r^3(\lambda r+3M)^2}\right\}$$
(40)
with $`\lambda =(l1)(l+2)/2`$. Here $`\psi _\mathrm{Z}(r)`$ is a gauge invariant combination of $`H_0,H_1,H_2,K,\psi /r`$. At this stage we would like to comment that the BD equation ($`\delta R_{\mu \nu }(h)_\nu _\mu \phi =0`$) in ref. takes the same equation as in Eq.(39). Also it is easily shown that an allowed solution is a plane wave like Eq.(38). This fact can be easily read off from the shape of potentials $`V_{\mathrm{RW}}`$ and $`V_\mathrm{Z}`$. Because these all belong to positive potential barrier for $`l2`$, there exist scattering states only. In other words, there are no bound state solutions. This means that one cannot find any exponentially growing mode in the graviton sector, even if $`k^2`$-term is involved in the RS equation (27).
## IV Discussions
We investigate the zero mode sector of the 5D dilatonic domain wall solution. This sector is very useful for describing the RS black hole on the brane. Assuming a spherically symmetric spacetime, one has a large black hole on the brane. We perform the analysis of stability to see whether or not the RS black hole truly exists. It is well-known that the BD black hole is stable. Here one finds an exponentially growing mode for the RS black hole because of a negative nature of its potential. If there is an exponentially growing mode, its black hole is classically unstable. ¿From the analysis of $`\phi `$ in the RS Minkowski spacetime, we have $`\phi =0`$ under the 4D TTF gauge. On the other hand, we cannot have $`\phi =0`$ in the spherically symmetric black hole spacetime with the RW gauge. Fortunately, we find that this instability is not allowed inside an AdS<sub>5</sub> whose size is $`(2k)^1`$. Hence the instability problem is cured. Actually BD scalar as well as gravitons can propagate in the black hole spacetime. Also this implies that the large RS black hole is stable.
For the metric perturbations, we don’t worry about their stability even for considering $`\phi 0`$. Choosing the RW gauge, the graviton sector including $`\phi `$ leads to the well-known two classes of odd and even-parities. Since this sector has always positive potential barriers for any $`l`$ with $`l2`$, there are no exponentially growing modes. Also this can be confirmed from the other side. If we introduce a new tensor $`\widehat{h}_{\mu \nu }`$, Eq. (27) reduces to Eq. (29). This is nothing but the perturbed equation for pure 4D gravity, which was proved to be stable thirty years ago.
In conclusion, the large black hole in the dilaton domain wall is stable. This can represent a stable form on the brane for the RS black cigar whose metric is not known up to now. Finally we comment that the RS black hole can be described by massless gravitons and a scalar mode with smaller wavelength than the size of an AdS<sub>5</sub>-box.
## Acknowledgement
We thank to J.Y. Kim and G. Kang for hepful discussions. This work was supported in part by the Brain Korea 21 Program of Ministry of Education, Project No. D-0025 and KOSEF, Project No. 2000-1-11200-001-3. |
warning/0001/quant-ph0001023.html | ar5iv | text | # Modification of relative entropy of Entanglement Supported by the National Natural Science Foundation of China under Grant No. 69773052 and the Fellowship of China Academy of Sciences
## Abstract
We present the modified relative entropy of entanglement (MRE) in order to both improve the computability for the relative entropy of entanglement and avoid the problem that the entanglement of formation seems to be greater than entanglement of distillation. For two qubit system we derive out an explicit and “weak” closed expression of MRE that depends on the pure state decompositions in the case of mixed states. For more qubit system, we obtain an algorithm to calculate MRE in principle. MRE significantly improves the computability of relative entropy of entanglement and decreases the dependence and sensitivity on the pure state decompositions. Moreover it is able to inherit most of the important physical features of the relative entropy of entanglement. In addition, a kind of states, as an extension of Werner’s states, is discussed constructively.
PACS: 03.65.Ud 03.67.-a
The entanglement is a vital feature of quantum information. It has important applications for quantum communication and quantum computation, for example, quantum teleportation , massive parallelism of quantum computation , decoherence in quantum computer and quantum cryptographic schemes . In the existing measures of entanglement, the entanglement of formation (EF) $`E_{EF}`$ and the relative entropy of entanglement (RE) $`E_{RE}`$ are defined by
$`E_{EF}(\rho _{AB})`$ $`=`$ $`\underset{\{p_i,\rho ^i\}𝒟}{\mathrm{min}}{\displaystyle \underset{i}{}}p_iS(\rho _B^i),`$ (1)
$`E_{RE}(\rho _{AB})`$ $`=`$ $`\underset{\rho _{AB}^R}{\mathrm{min}}S(\rho _{AB}||\rho _{AB}^\mathrm{R}),`$ (2)
where $`𝒟`$ in Eq.(1) is a set that includes all the possible decompositions of pure states $`\rho =_ip_i\rho ^i`$, and $``$ in Eq.(2) is a set that includes all the disentangled states. Note that $`\rho _B^i=\mathrm{Tr}_A\rho ^i`$ is the reduced density matrix of $`\rho ^i`$, $`S(\rho )`$ is von Neumann entropy of $`\rho `$, $`S(\rho ||\rho ^\mathrm{R})=\mathrm{Tr}(\rho \mathrm{log}\rho \rho \mathrm{log}\rho ^\mathrm{R})`$ is the quantum relative entropy and $`\rho ^\mathrm{R}`$ can be called the relative density matrix.
For a pure state in a bi-party system EF is an actually standard measure of entanglement. For an arbitrary state of two qubits, EF is also widely accepted . However, there is a surprised result recently, that is, it seems that EF is greater than the entanglement of distillation (DE) in the case of mixed state. If it is so, how to calculate RE will be important because RE is thought of an upper bound of DE in the case of mixed states . RE also appears promising by a series of the interesting results . However, its advantages suffers from the difficulty in computation. The set $``$ is so large that one can not sure when the minimumizing procedure is finished. In this letter, in order to avoid the problem that EF seems to be greater than DE and improve the computability of RE, we present a modified relative entropy of entanglement (MRE) as a possible upper bound of DE, through redefining a suitable relative density matrix first for the given pure states, and then for mixed states in terms of the physical idea to choose a minimum pure state decomposition such as EF. The physical intuition to propose MRE is original from organically combining the advantages of EF for the pure states and RE for the mixed states and avoiding their shortcomings as possibly.
How to calculate RE is also interesting even if in the simple enough case of two qubits except for one can prove that for an arbitrary state of two qubits EF calculated by Wootters’ method is not greater than DE. If this exception happens, RE is only important for a bigger system more than two qubits and then we have to study the relevant problem and further development based this letter. That is, we will extend MRE to multi-party systems naturally and overcome the difficulty that one only can discuss RE qualitatively for multi-party systems .
For our purpose, let’s first give out three lemmas.
Lemma one. For two qubits, the polarization vectors $`𝝃_A`$ and $`𝝃_B`$ corresponding to the reduced matrices $`\rho _{\{A,B\}}=\frac{1}{2}(\mathrm{𝟏}+𝝃_{\{A,B\}}𝝈)`$ read:
$$𝝃_A=\mathrm{Tr}(\rho 𝝈I),𝝃_B=\mathrm{Tr}(\rho I𝝈),$$
(3)
where $`𝝈`$ is the Pauli spin matrix.
Lemma Two. For the pure state of two qubits, there are the relations between $`𝝃_A`$ and $`𝝃_B`$
$$\xi _A^i=4\underset{j=1}{\overset{3}{}}a_{ij}\xi _B^j,4\underset{i=1}{\overset{3}{}}\xi _A^ia_{ij}=\xi _B^j,$$
(4)
where $`a_{ij}`$ is the expanding coefficients
$$\rho =\underset{\mu ,\nu =0}{\overset{3}{}}a_{\mu \nu }\sigma _\mu \sigma _\nu ,$$
(5)
in which $`\sigma _0`$ is the identity matrix. In the case of a pure state $`|\psi =a|00+b|01+c|10+d|11`$, it follows that
$$𝝃^2=𝝃_A^2=𝝃_B^2=14|adbc|^2.$$
(6)
Lemma three. If the relative density matrix in its eigenvector decomposition is:
$$\rho ^\mathrm{R}=\underset{\alpha }{}\lambda _\alpha \rho _\alpha ^\mathrm{R}=\underset{\alpha }{}\lambda _\alpha |v_\alpha ^\mathrm{R}v_\alpha ^\mathrm{R}|,$$
(7)
where $`\lambda _\alpha `$ is taken over all the eigenvalues and the eigen density matrices are assumed to be orthogonalized and idempotent without loss of generality, thus the relative entropy can be written as
$`S(\rho ||\rho ^\mathrm{R})`$ $`=`$ $`S(\rho ){\displaystyle \underset{\alpha }{}}\mathrm{log}\lambda _\alpha \mathrm{Tr}(\rho \rho _\alpha ^\mathrm{R})`$ (8)
$`=`$ $`S(\rho ){\displaystyle \underset{\alpha }{}}\mathrm{log}\lambda _\alpha v_\alpha ^\mathrm{R}|\rho |v_\alpha ^\mathrm{R}.`$ (9)
From the definition of the polarization vectors, Lemma one is easy to get. To prove Lemma two, it is used the fact that $`\rho ^2=\rho `$ for the pure state and Lemma one. It is easy to prove Lemma three by the simple computation in quantum mechanics. Lemma three implies that the key to calculate RE is to seek an appropriate relative density matrix $`\rho ^\mathrm{R}`$ and to find out its all the eigenvalues.
Now we can formulate the basic theorems of this letter.
Theorem one. In the case of the pure state $`\rho ^\mathrm{P}`$ of two qubits, the relative density matrix of RE can be taken as
$`R(\rho ^P)`$ $`=`$ $`{\displaystyle \frac{1+|𝝃|}{2}}{\displaystyle \frac{1}{2}}\left(I+{\displaystyle \frac{𝝃_A}{|𝝃|}}𝝈\right){\displaystyle \frac{1}{2}}\left(I+{\displaystyle \frac{𝝃_B}{|𝝃|}}𝝈\right)`$ (11)
$`+{\displaystyle \frac{1|𝝃|}{2}}{\displaystyle \frac{1}{2}}\left(I{\displaystyle \frac{𝝃_A}{|𝝃|}}𝝈\right){\displaystyle \frac{1}{2}}\left(I{\displaystyle \frac{𝝃_B}{|𝝃|}}𝝈\right).`$
For the maximum entanglement state
$$|\mathrm{\Phi }^\pm =\frac{1}{\sqrt{2}}(|00\pm |11,|\mathrm{\Psi }^\pm =\frac{1}{\sqrt{2}}(|01\pm |10,$$
(12)
because that $`|𝝃|=0`$, it should be
$`R(\mathrm{\Phi }^\pm )`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left({\displaystyle \frac{1}{2}}\left(I+\sigma _3\right){\displaystyle \frac{1}{2}}\left(I+\sigma _3\right)\right)`$ (14)
$`+{\displaystyle \frac{1}{2}}\left({\displaystyle \frac{1}{2}}\left(I\sigma _3\right){\displaystyle \frac{1}{2}}\left(I\sigma _3\right)\right),`$
$`R(\mathrm{\Psi }^\pm )`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left({\displaystyle \frac{1}{2}}\left(I+\sigma _3\right){\displaystyle \frac{1}{2}}\left(I\sigma _3\right)\right)`$ (16)
$`+{\displaystyle \frac{1}{2}}\left({\displaystyle \frac{1}{2}}\left(I\sigma _3\right){\displaystyle \frac{1}{2}}\left(I+\sigma _3\right)\right).`$
That RE are calculated in terms of $`R(\rho ^\mathrm{P})`$ for a pure state is equal to EF.
In order to prove this theorem, we, in Eq.(2), choose such a subset of $``$ that $`\rho ^\alpha `$ is purely separable as $`\rho ^\alpha =\rho _A^\alpha \rho _B^\alpha `$. For simplicity, only consider the case with two qubits. Because that the eigen density matrix is pure, $`\rho _A^\alpha `$ and $`\rho _B^\alpha `$ have to be pure. While the $`2\times 2`$ density matrix of the pure state can be written as $`\rho _{\{A,B\}}^\alpha =\frac{1}{2}(1+𝜼_{\{A,B\}}^\alpha 𝝈)`$, where $`|𝜼_{\{A,B\}}^\alpha |=1`$. Thus, from Lemma three it follows that
$`S(\rho ||\rho ^\mathrm{R})`$ $`=`$ $`S(\rho ){\displaystyle \underset{\alpha }{}}\mathrm{log}\lambda _\alpha {\displaystyle \underset{\mu ,\nu =0}{\overset{3}{}}}\eta _A^\alpha {}_{\mu }{}^{}a_{\mu \nu }^{}\eta _B^\alpha _\nu `$ (17)
$`=`$ $`S(\rho ){\displaystyle \underset{\alpha }{}}\omega ^\alpha \mathrm{log}\lambda _\alpha ,`$ (18)
where $`\omega ^\alpha ={\displaystyle \underset{\mu ,\nu =0}{\overset{3}{}}}\eta _A^\alpha {}_{\mu }{}^{}a_{mu\nu }^{}\eta _A^\alpha _\nu `$ and $`\eta _{\{A,B\}}=(1,𝜼_{\{A,B\}})`$. Because of the orthogonal property among the different $`\rho ^\alpha `$, we can choose $`𝜼_A^1=𝜼_A^3=𝒌`$, $`𝜼_A^2=𝜼_A^4=𝒎`$ and $`𝜼_B^1=𝜼_B^2=𝜼_B^3=𝜼_B^4=𝒏`$. In order to calculate the minimum value of the relative entropy, one has to find the partial derivatives of all the variables, set them to zero to form a equation system, and then solve this equation system. However, it doesn’t exist. So we only find the extreme surface fixing all the eigenvalues of $`\rho ^\mathrm{R}`$. It is easy to verify that $`\lambda _\alpha =\omega _\alpha `$ gives out the minimum surface. Thus, the minimum relative entropy is
$$S(\rho ||\rho ^\mathrm{R})=S(\rho )\underset{\alpha }{}\omega ^\alpha \mathrm{log}\omega _\alpha ,$$
(19)
Again substituting into the chosen $`R(\rho ^\mathrm{P})`$ in Theorem one, in terms of all of lemmas, it is obtained immediately
$$E_{RE}(\rho ^\mathrm{P})=S(\rho ^\mathrm{P}||R(\rho ^\mathrm{P}))=S(\rho _{\{A,B\}}^\mathrm{P})=E_{EF}(\rho ^\mathrm{P}).$$
(20)
For the mixed state, if theorem one is extended directly, we will find the result is not satisfied. So, in terms of the physical idea of EF to deal with the case of the mixed state, we obtain
Definition. For a pure state $`\rho ^\mathrm{P}`$ and a mixed state $`\rho ^\mathrm{M}`$, MRE is defined respectively as
$`E_{MRE}(\rho ^\mathrm{P})`$ $`=`$ $`S(\rho ^\mathrm{P}||R(\rho ^\mathrm{P}))=E_{EF}(\rho ^\mathrm{P}),`$ (21)
$`E_{MRE}(\rho ^\mathrm{M})`$ $`=`$ $`\underset{\{p_i,\rho ^i\}𝒟}{\mathrm{min}}S(\rho ^\mathrm{M}||{\displaystyle \underset{i}{}}p_iR(\rho ^i))`$ (22)
$`=`$ $`\underset{\{p_i,\rho ^i\}𝒟}{\mathrm{min}}S(\rho ^\mathrm{M}||R^\mathrm{M}),`$ (23)
where $`R(\rho ^\mathrm{P})`$ is such a relative density matrix corresponding to the pure state $`\rho ^\mathrm{P}`$ that Eq.(21) is satisfied and $`R(\rho ^\mathrm{P})`$ is an disentangled density matrix. In Eq.(22), the minimum is taken over the set $`𝒟`$ that includes all the possible decompositions of pure states $`\rho =_ip_i\rho ^i`$, and $`R^\mathrm{M}=_ip_iR(\rho ^i)`$ is the total relative density matrix for a mixed state, while $`R(\rho ^i)`$ is found out so that Eq.(21) is valid for the pure state $`\rho ^i`$. In particular, for two qubits, $`R(\rho ^i)`$ is chosen by Theorem one.
Theorem two Modified relative entropy of entanglement (MRE) always satisfies:
$$E_{MRE}(\rho )E_{EF}(\rho ).$$
(24)
When $`\rho `$ is a pure state, the equality is valid.
It is easy to prove it in terms of the joint convexity of the relative entropy
$$S(\underset{i}{}p_i\rho ^i||\underset{i}{}p_iR(\rho ^i))\underset{i}{}p_iS(\rho ^i||R(\rho ^i))$$
(25)
and the definition of $`E_{EF}`$ in Eq.(1). Obviously for a pure state, MRE is equal to RE and EF. The relative density matrix $`R`$ for MRE in a given pure state can be defined by this relation, that is solving $`S(\rho ^\mathrm{P}||R)=S(\rho _B^\mathrm{P})=E_{EF}(\rho ^\mathrm{P})`$ to find $`R`$. For the case of mixed state, one first finds the relative density matrix $`R(\rho _i)`$, in which $`\rho _i`$ belong to a pure state decomposition, by solving $`S(\rho ^i||R(\rho ^i))=S(\rho _B^i)=E_{EF}(\rho ^i)`$. Then, one can write the total relative density matrix for a mixed state as $`R^\mathrm{M}=_ip_iR(\rho ^i)`$ and calculate the corresponding relative entropy. The last, MRE is obtained by taking the minimum one among these relative entropies. Therefore we obtain a definite and constructive algorithm to calculate MRE as we have had in the calculation of EF. For two qubit systems, we have successfully obtained the explicit and general expression of the relative density matrix in a pure state or a mixed state with any given decomposition. MRE for two qubit systems can be easier calculated because the first step in our algorithm is largely simplified. For more than two qubits, we do not give clearly an explicit expression of the relative density matrix for a pure state in this paper, the first step needs more computations, but our algorithm still works in principle. This is because that from $`S(\rho _i^\mathrm{P}||R)=S(\rho _B^i)=E_{EF}(\rho ^i)`$ to find $`R(\rho ^i)`$ can be done within finite steps for a given pure state in general except for the solution $`R(\rho ^i)`$ does not exist. The exception is impossible because this implies that RE for the pure state $`\rho _i`$ has no a relative density matrix ($`S(\rho _B^i)=\mathrm{min}S(\rho _i||\rho ^R)=E_{RE}(\rho ^i)`$). It must be emphasized that our method is to calculate MRE but not EF, and the comparison between MRE and EF is given in their algorithm but not in their physical significance and requirement. In other words, EF and MRE can not be replaced each other, our method and the computational method of EF, for example Wootter’s method, can not be replaced each other. In above sense, MRE avoids the difficulty of RE to find the relative density matrix from an infinite large set of disentangled states and so improve the computability of RE. Moreover, in our recent paper , we have given an explicit expression of the relative density matrix for $`n`$-party systems (restricted to qubits).
In the case of mixed states, obviously RE$``$MRE$``$EF. Noting the fact that both RE and MRE are defined by the relative entropy and satisfy Theorem two, we think that MRE is able to inherit most of important physical features of RE if these features of RE are given and proved in terms of the fact stated above as well as some mathematical skills . For example, the properties under local general measurement (LGM) and classical communication (CC) can be proved by using of the similar methods in refs. at least for two qubit systems. We have seen that MRE is the function of the polarization vectors of the reduced density matrices of the decomposition density matrices for two qubits. Thus, EF and RE as well as MRE all belong to a kind of the generalized measures of entanglement proposed by , and the generalized measures of entanglement with all the known properties as a good measure are proved there. Two known main measures of entanglement are related together by the polarization vectors of the reduced density matrix. As to the properties in two qubit systems, such as its range is $`[0,1]`$, its maximum value $`1`$ corresponds to the maximally entangled states and its minimum value $`0`$ corresponds to the mixture of the disentangled states, can be directly and easily obtained from the definition of MRE. For two qubits, the relative density matrix is a function of the polarized vectors $`𝝃_A^i,𝝃_B^i`$, and $`𝝃_A^i,𝝃_B^i`$ are functions of the decomposition density matrices $`\rho _i`$. Thus, MRE is just a compound function of the decomposition density matrices $`\rho _i`$. Of course, it is not a good property that a measure of entanglement depends on the possible decompositions because it is not very easy to find all the elements of $`𝒟`$. But this exists in all the known measures of entanglement either. MRE has significantly improvement in this aspect for some kinds of states. For example, the state $`D`$ has two pure state decompositions
$`D`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(|0000|+|1111|\right)`$ (26)
$`=`$ $`{\displaystyle \frac{1}{2}}\left(|\mathrm{\Phi }^+\mathrm{\Phi }^+|+|\mathrm{\Phi }^{}\mathrm{\Phi }^{}|\right)`$ (27)
which respectively correspond to the minimum and maximum decompositions in the calculation of EF. But two decompositions have the same relative density matrices in the calculation of MRE. That is, both of them are the minimum for MRE and can be used to calculate MRE. This means that the minimum decomposition(s) to calculate MRE is (are) not the same as the minimum decomposition(s) to calculate EF in general. We can verify that this advantage is kept for Werner’s state
$`W`$ $`=`$ $`F|\mathrm{\Psi }^{}\mathrm{\Psi }^{}|+{\displaystyle \frac{1F}{3}}(|\mathrm{\Psi }^+\mathrm{\Psi }^+|`$ (29)
$`+|\mathrm{\Phi }^+\mathrm{\Phi }^+|+|\mathrm{\Phi }^{}\mathrm{\Phi }^{}|)`$
Its relative density matrix reads $`R(W)_{ij}(ij)=0,R(W)_{ii}=\{(1F)/3,(1+2F)/6,(1+2F)/6,(1F)/3\}`$. Thus, it is easy to get
$`S(W||R(W))`$ $`=`$ $`F\mathrm{log}F+{\displaystyle \frac{1F}{3}}\mathrm{log}\left({\displaystyle \frac{1F}{3}}\right)`$ (31)
$`{\displaystyle \frac{1+2F}{3}}\mathrm{log}\left({\displaystyle \frac{1+2F}{6}}\right),`$
It correctly gives out MRE of Werner state. In fact, from Eq.(31) it follows that when $`F=1/4`$ Werner’s state is disentangled, and when $`F=1`$ Werner’s state has the maximum entanglement. If we take
$`W`$ $`=`$ $`{\displaystyle \frac{4F1}{3}}|\mathrm{\Psi }^{}\mathrm{\Psi }^{}|+{\displaystyle \frac{1F}{3}}I\left(F{\displaystyle \frac{1}{4}}\right)`$ (32)
$`W`$ $`=`$ $`{\displaystyle \frac{14F}{3}}|\mathrm{\Psi }^+\mathrm{\Psi }^+|+{\displaystyle \frac{1F}{3}}(|0000|+|1111|)`$ (34)
$`+F(|0101|+|1010|)\left(F{\displaystyle \frac{1}{4}}\right),`$
the relative density matrix does not change and so the result is the same. This implies that MRE can decrease the dependence and sensitivity on the pure state decompositions at least for some interesting states. We can more easily find an adequate pure state decomposition in the calculation of MRE than do this in the calculation of EF.
If we extend Werner’s state to a new kind of states
$$W_\mathrm{E}=\underset{i=1}{\overset{4}{}}b_i|\mathrm{B}_\mathrm{i}\mathrm{B}_i|+\underset{i=1}{\overset{4}{}}c_i|ii|,$$
(35)
where $`|\mathrm{B}_\mathrm{i}`$ are four Bell states $`|\mathrm{\Phi }^\pm ,|\mathrm{\Psi }^\pm `$ and $`|i`$ are $`|00,|01,|10,|11`$ respectively. Note that $`_{i=1}^4(b_i+c_i)=1`$ and all of them are positive. We can find that MRE also depends on the pure state decompositions. For the simplicity, consider the state
$$\rho =\lambda |\mathrm{\Phi }^+\mathrm{\Phi }^+|+(1\lambda )|0000|.$$
(36)
Its eigen decomposition is
$`\rho `$ $`=`$ $`v_{}|V^{}V^{}|+v_+|V^+V^+|,`$ (37)
$`v_\pm `$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(1\pm \sqrt{12\lambda (1\lambda )}\right).`$ (38)
Two decompositions lead to the different the relative density matrices and relative entropies. Therefore, this implies that the minimum procedure is not unnecessary for the modified relative entropy of entanglement in general. It seems to us, it is interesting to give a good algorithm that can find all the elements of the set of the pure state decompositions $`𝒟`$. This is still an open question. But for a kind of mixed states with the form of the extension of Werner’s state (35), we can find that its definition is an adequate minimum decomposition (its reliability has not been strictly proved in mathematics, but we have not found a counterexample). This conclusion can be obtained perhaps because MRE of this kind of states is not very sensitive to the pure state decompositions. Our method is first to choose a minimum decomposition among all of decomposition that we can find, then check the result by the particular points (disentangled and maximum entangled) and repeat this process up to the case we can not continue it. Furthermore, we carry out some numerical checking. Thus, MRE of the extended Werner state can be written as
$`E_{MRE}(W_\mathrm{E})`$ $`=`$ $`{\displaystyle \underset{\alpha }{}}v_\alpha \mathrm{log}v_\alpha `$ (43)
$`{\displaystyle \frac{1}{2}}(b_1+b_2+2c_1)\mathrm{log}{\displaystyle \frac{1}{2}}(b_1+b_2+2c_1)`$
$`{\displaystyle \frac{1}{2}}(b_3+b_4+2c_2)\mathrm{log}{\displaystyle \frac{1}{2}}(b_3+b_4+2c_2)`$
$`{\displaystyle \frac{1}{2}}(b_3+b_3+2c_4)\mathrm{log}{\displaystyle \frac{1}{2}}(b_3+b_4+2c_3)`$
$`{\displaystyle \frac{1}{2}}(b_1+b_2+2c_4)\mathrm{log}{\displaystyle \frac{1}{2}}(b_1+b_2+2c_4),`$
where the eigenvalues of $`W_\mathrm{E}`$ are
$`v_1`$ $`=`$ $`{\displaystyle \frac{1}{2}}(b_1+b_2+c_1+c_4\sqrt{(b_1b_2)^2+(c_1c_4)^2}),`$ (44)
$`v_2`$ $`=`$ $`{\displaystyle \frac{1}{2}}(b_1+b_2+c_1+c_4+\sqrt{(b_1b_2)^2+(c_1c_4)^2}),`$ (45)
$`v_3`$ $`=`$ $`{\displaystyle \frac{1}{2}}(b_3+b_4+c_2+c_3\sqrt{(b_3b_4)^2+(c_2c_3)^2}),`$ (46)
$`v_4`$ $`=`$ $`{\displaystyle \frac{1}{2}}(b_3+b_4+c_2+c_3+\sqrt{(b_3b_4)^2+(c_2c_3)^2}).`$ (47)
Based on Peres’s condition, we can calculate the eigenvalues of partial transpose of the extended Werner state and obtain the condition that $`W_E`$ is separable
$`(b_1+b_2)^2`$ $``$ $`(b_3b_4)^24c_1c_4,`$ (48)
$`(b_3+b_4)^2`$ $``$ $`(b_1b_2)^24c_2c_3.`$ (49)
In conclusion, MRE can be useful based on four evidences. One is that MRE is a possible upper bound of entanglement of distillation such as RE, the second is MRE improves the compatibility of RE, the third is that MRE significantly decrease the dependence and sensitivity on the pure state decompositions at least for some interesting states, and the last is that MRE can be extended to multi-party systems naturally. This research is on progressing. |
warning/0001/cond-mat0001185.html | ar5iv | text | # Strong vortex pinning in the low–temperature superconducting phase of (U1-xThx)Be13
## 1 INTRODUCTION
Recent studies of vortex dynamics have uncovered a novel type of exceedingly strong vortex pinning not observed in any other hard type II superconductor .$`\mathrm{\backslash }sevenrm^\text{?}`$ In the low–temperature, low–field superconducting phase of UPt<sub>3</sub>, the so–called B–phase, metastable configurations of vortices do not creep towards equilibrium in a time scale of $`10^410^5`$ s. However, in the high temperature A–phase, vortex creep occurs with rates that increase rapidly as the temperature is increased. The transition from one creep regime to the other is very sharp and more than two orders of magnitude in size. The anomalous strong pinning detected only in the B-phase, indicates that this superconducting phase supports novel types of vortices and/or novel types of pinning structures.
The only other known example of a superconductor with more than one superconducting phase is thorium–doped UBe<sub>13</sub>. In the concentration range $`0.019x0.045`$, the heavy fermion (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> shows an additional second order transition below the onset of superconductivity as first seen in specific heat measurements .$`\mathrm{\backslash }sevenrm^\text{?}`$ Moreover, the lower critical field $`H_{c1}`$ shows a sudden break in slopes, indicating a clear increase in the superconducting condensation energy below the transition at $`T_{c2}`$ .$`\mathrm{\backslash }sevenrm^\text{?}`$ Muon spin relaxation data $`\mathrm{\backslash }sevenrm^\text{?}`$ reveal the existence of weak magnetic correlations in the low-T phase which are interpreted as evidence that this phase breaks time reversal symmetry.
Here we discuss the results of a recent investigation of vortex dynamics in a single crystal of (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with $`x=0.0275`$. Similar results as the ones presented here were obtained with a second single crystal from a different batch. We also compare the data on vortex creep with data on pure UBe<sub>13</sub> and UPt<sub>3</sub>.
## 2 EXPERIMENTAL ARRANGEMENT
The sample investigated consisted of a single crystal (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with $`x=0.0275`$ prepared at Los Alamos National Laboratory. It was cut in the form of a parallelepiped $`2.25\times 1.00\times 0.88`$ mm<sup>3</sup> in size. It has a transition temperature $`T_c`$ = 523 mK with a width $`\mathrm{\Delta }T_c=67`$ mK taken with the $`1090\%`$ criterion.
Vortex creep data was obtained from the relaxation of the remanent magnetization after cycling the zero-field cooled sample in high enough fields so that the critical state was established. The relaxation measurements were done from 1 s to $`10^410^5`$ s in the temperature range 5 mK$`TT_c`$. The experimental arrangement is described in reference 1.
## 3 RESULTS AND DISCUSSION
In Fig. 1 we show values of the remanent magnetization of (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> at two different times as function of temperature. The shadowed area indicates the amount of flux that leaves the specimen from the initial time of our measurement ($`t1`$ s) to $`t=10^4`$ s. As can be clearly seen, deep in the low-temperature superconducting phase, no flux leaves the sample on a time scale of $`10^4`$ s. On the other hand, in the high temperature phase, considerably amount of vortices leave the specimen in the time indicated. The initial creep rates from the data in Fig. 1 are plotted in Fig. 2. In this figure we also show the measured values of the lower critical field $`H_{c1}`$ for the same crystal as function of $`T^2`$. The sharp break of the $`H_{c1}`$ slope at $`T=350`$ mK occurs at the same temperature as the jump in the specific heat at $`T_{c2}`$ .$`\mathrm{\backslash }sevenrm^\text{?}`$ At this same temperature we observe a large transition from vortex creep rates $`\mathrm{ln}M/\mathrm{ln}t`$ of the order of $`3\times 10^2`$ to values as small as $`\mathrm{ln}M/\mathrm{ln}t10^5`$. This last figure reflects the limit of our sensitivity, determined mainly by the reproducibility of the background creep of the NbTi coil used to produce magnetic fields of the order of few hundred Oe. This coil is directly attached to the walls of the Epoxy mixing chamber.
In Fig. 3 a we compare the vortex transition in (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with similar transitions in UPt<sub>3</sub> $`\mathrm{\backslash }sevenrm^\text{?}`$ on a double logarithmic scale. Indeed, in both superconductors one detects ”ideal pinning” in their low temperature phases. However, although we do not observe creep on a scale of $`10^5`$ s at the lowest temperatures, we detect some sort of ”avalanche” or non logarithmic creep at times which become shorter and shorter as the temperature approaches $`T_{c2}`$. The data in Fig. 2 and Fig. 3 are based on the initial logarithmic slope calculated from the first couple of decades in time.
It is interesting to compare the vortex dynamics in pure UBe<sub>13</sub> with the dynamics of (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with $`x=0.0275`$ in view of some recent investigation by Kromer et al.$`\mathrm{\backslash }sevenrm^\text{?}`$ From thermal expansion measurements on samples of UBe<sub>13</sub> with different thorium concentrations, these authors concluded that the nature of the superconducting state in the critical concentration range ($`0.019x0.045`$) below $`T_{c1}`$ is not fundamentally different from that in the pure compound below $`T_c`$.
In Fig. 3 b we show that the vortex creep rate in a single crystal of pure UBe<sub>13</sub> $`\mathrm{\backslash }sevenrm^\text{?}`$ follows a well defined linear in $`T`$ dependence from $`T=5`$ mK up to $`TT_c`$ as expected from the Kim–Anderson theory of thermally activated creep. No indication of anomalous strong pinning is detected in this material. Indeed there is a fundamental difference between the low–$`T`$ superconducting phase of (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with $`x=0.0275`$ and pure UBe<sub>13</sub> concerning vortex dynamics.
At this point the physical origin of ”ideal pinning” is not known. However, by analogy with the superfluid phases of liquid <sup>3</sup>He, where broken symmetries are manifested in new physical properties of the quantized vortex lines under rotation, one has to expect that in superconductors with nonscalar order parameters new types of vortices can also lead to unusual behavior.
In conclusion, we have found in UPt<sub>3</sub> and (U<sub>1-x</sub>Th<sub>x</sub>)Be<sub>13</sub> with $`x=0.0275`$ sharp transitions of about three orders of magnitude in the vortex creep rates at the temperature marking the boundary between their two low-field superconducting phases. In both materials, deep in the low–$`T`$ phases no creep is observed on a scale of $`10^5`$ s. Theoretical input is needed to determine the physical origin of our observation.
## ACKNOWLEDGMENTS
We acknowledge valuable discussions with M. Sigrist and D. Agterberg. Part of this work was supported by the Swiss National Science Foundation. |
warning/0001/physics0001006.html | ar5iv | text | # Temporal correlations and neural spike train entropy
## Abstract
Sampling considerations limit the experimental conditions under which information theoretic analyses of neurophysiological data yield reliable results. We develop a procedure for computing the full temporal entropy and information of ensembles of neural spike trains, which performs reliably for limited samples of data. This approach also yields insight upon the role of correlations between spikes in temporal coding mechanisms. The method, when applied to recordings from complex cells of the monkey primary visual cortex, results in lower RMS error information estimates in comparison to a ‘brute force’ approach. PACS numbers: 87.19.Nn,87.19.La,89.70.+c,07.05.Kf
Cells in the central nervous system communicate by means of stereotypical electrical pulses called action potentials, or spikes . The Shannon information content of neural spike trains is fully described by the sequence of times of spike emission. In principle, the pattern of spike times provides a large capacity for conveying information beyond that due to the code commonly assumed by physiologists, the number of spikes fired . Reliable quantification of this spike timing information is made difficult by undersampling problems that scale with the number of possible spike patterns, and thus up to exponentially with the precision of spike observation (see Fig. 1). While advances have been made in experimental preparations where extensive sampling may be undertaken , our understanding of the temporal information properties of nerve cells from less accessible preparations such as the mammalian cerebral cortex is limited.
Any direct estimate of the complete spike train information is limited by sampling considerations to relatively small wordlengths, and therefore to the analysis of short time windows of data. However, it is possible to take advantage of this restriction itself to obtain estimators which have better sampling properties than a ‘brute force’ approach. In this Letter we present an approach based upon a Taylor series expansion of the entropy, to second order in the time window of observation . The analytical expression so derived allows the ensemble spike train entropy to be computed from limited data samples, and relates the entropy and information to the instantaneous probability of spike occurrence and the temporal correlations between spikes. Comparison with other procedures such as the ‘brute force’ approach indicates that our analytical expression gives substantially better performance for data sizes of the order typically obtained from mammalian neurophysiology experiments, as well as providing insight into potential coding mechanisms.
Consider a time period of duration $`T`$, associated with a dynamic or static sensory stimulus, during which the activity of $`C`$ cells is observed. The neuronal population response to the stimulus is described by the collection of spike arrival times $`\{t_i^a\}`$, $`t_i^a`$ being the time of the $`i`$-th spike emitted by the $`a`$-th neuron. The spike time is observed with finite precision $`\mathrm{\Delta }t`$, and this bin width is used to digitise the spike train (Fig. 1). For a given discretisation (temporal precision), the entropy of the spike train is a well defined quantity. The total entropy of the spike train ensemble is
$$H(\{t_i^a\})=\underset{\{t_i^a\}}{}P(\{t_i^a\})\mathrm{log}_2P(\{t_i^a\}),$$
(1)
where the summation is over all possible spike times within $`T`$ and over all possible total spike counts from the population of cells. This entropy quantifies the total variability of the spike train. Each different stimulus history (time course of characteristics within $`T`$) is denoted as $`s`$. The noise entropy, which quantifies the variability to repeated presentations of the same stimulus, is $`H^{\mathrm{noise}}=H(\{t_i^a\}|s)_s`$, where the angular brackets indicate the average over different stimuli, $`A(s)_s_{s𝒮}P(s)A(s)`$. The mutual information that the responses convey about which stimulus history invoked the spike train is the difference between these two quantities.
These entropies may be expanded as a Taylor series in the time window of measurement,
$$H=TH_t+\frac{T^2}{2}H_{tt}+O(T^3).$$
(2)
To compute the Taylor expansion, we made the following assumptions: (i) The time window is short enough, or the firing rate low enough, that there are few spikes per stimulus presentation. (ii) The entropy is analytic in $`T`$. (iii) Different trials are random realisations of the same process. We will use the bar notation for the average over trials at fixed stimulus, such that if $`r_a(t;s)=_i\delta _{t,t_i^a(s)}`$, the time-dependent instantaneous firing rate $`\overline{r}_a(t;s)`$ is its average over experimental trials. (iv) Spikes are not locked to each other with infinite precision; in other words, the conditional probability of a spike occuring at time $`\tau _j^b`$ given occurrence of a particular spike pattern $`\{t_i^a\}`$ scales for small $`\mathrm{\Delta }t`$ proportionally to $`\mathrm{\Delta }t`$ plus higher order terms, with no $`O(1)`$ terms: $`P(\tau _j^b|\{t_i^a\};s)\mathrm{\Delta }t+\mathrm{}`$ for each possible spike pattern $`\{t_i^a\}`$. The validity of these assumptions has been examined elsewhere .
The probability of observing a pattern with $`k`$ spikes can be expressed as a product of $`k`$ probabilities of each of the spikes given the presence of others. Thus from (iv), the probability of this pattern is proportional to $`\mathrm{\Delta }t^k`$, and the expansion is essentially in the total number of spikes emitted. This also implies that only the conditional probabilities between spike pairs are necessary for the 2nd order expansion. Parameterising the conditional probability between two spikes by the scaled correlation $`\gamma _{ab}(t_i^a,t_j^b;s)`$ , we can now write down the probabilities required by Eq. 1.
Denoting the no spikes event as 0 and the joint occurrence of a spike from cell $`a`$ at time $`t_1^a`$ and a spike from cell $`b`$ at time $`t_2^b`$ as $`t_1^at_2^b`$, the conditional response probabilities are, to second order:
$`P(\mathrm{𝟎}|s)`$ $`=`$ $`1{\displaystyle \underset{a=1}{\overset{C}{}}}{\displaystyle \underset{t_1^a}{}}\overline{r}_a(t_1^a;s)\mathrm{\Delta }t+{\displaystyle \frac{1}{2}}{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_1^a}{}}{\displaystyle \underset{t_2^b}{}}\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]\mathrm{\Delta }t^2`$ (3)
$`P(t_1^a|s)`$ $`=`$ $`\overline{r}_a(t_1^a;s)\mathrm{\Delta }t\overline{r}_a(t_1^a;s){\displaystyle \underset{b=1}{\overset{C}{}}}{\displaystyle \underset{t_2^b}{}}\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]\mathrm{\Delta }t^2a=1,\mathrm{},C`$ (4)
$`P(t_1^at_2^b|s)`$ $`=`$ $`\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]\mathrm{\Delta }t^2a=1,\mathrm{},C,b=1,\mathrm{},C.`$ (5)
The unconditional response probabilities are simply $`p(\{t_i^a\})=p(\{t_i^a\}|s)_s`$. Inserting $`p(\{t_i^a\})`$ into Eq. 1 and keeping only leading order terms yields for the first order total entropy
$$TH_t=\frac{1}{\mathrm{ln}2}\underset{a}{}\underset{t_1^a}{}\overline{r}_a(t_1^a;s)\mathrm{\Delta }t_s\underset{a}{}\underset{t_1^a}{}\overline{r}_a(t_1^a;s)\mathrm{\Delta }t_s\mathrm{log}_2\overline{r}_a(t_1^a;s)\mathrm{\Delta }t_s.$$
(6)
Inserting $`p(\{t_i^a\}|s)`$ instead yields a similar expression for the first order noise entropy $`TH_t^{\mathrm{noise}}`$, except with a single stimulus average $`_s`$ around the entire second term. Continuing the expansion, and noting that a factor of 1/2 is introduced to prevent overcounting of equivalent permutations, the additional terms up to second order are:
$`{\displaystyle \frac{T^2}{2}}H_{tt}`$ $`=`$ $`{\displaystyle \frac{1}{2\mathrm{ln}2}}{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_1^a}{}}{\displaystyle \underset{t_2^b}{}}\left\{\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]_s\overline{r}_a(t_1^a;s)_s\overline{r}_b(t_2^b;s)_s\right\}\mathrm{\Delta }t^2`$ (8)
$`+{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_t^a}{}}{\displaystyle \underset{t_2^b}{}}\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]\mathrm{\Delta }t^2_s\mathrm{log}_2{\displaystyle \frac{\overline{r}_a(t_1^a;s)_s}{\sqrt{\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]_s}}}`$
$`{\displaystyle \frac{T^2}{2}}H_{tt}^{\mathrm{noise}}`$ $`=`$ $`{\displaystyle \frac{1}{2\mathrm{ln}2}}{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_1^a}{}}{\displaystyle \underset{t_2^b}{}}\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\gamma _{ab}(t_1^a,t_2^b;s)_s\mathrm{\Delta }t^2`$ (10)
$`+{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_t^a}{}}{\displaystyle \underset{t_2^b}{}}\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]\mathrm{\Delta }t^2\mathrm{log}_2{\displaystyle \frac{\overline{r}_a(t_1^a;s)}{\sqrt{\overline{r}_a(t_1^a;s)\overline{r}_b(t_2^b;s)\left[1+\gamma _{ab}(t_1^a,t_2^b;s)\right]}}}_s.`$
The difference between the total and noise entropies gives the expression for the mutual information detailed in .
It has recently been found that correlations, even if independent of the stimulus identity, can increase the information present in a neural population . This applies both to cross-correlations between the spike trains from different neurons and to auto-correlations in the spike train from a single neuron . The equations derived above add something to the explanation of this phenomenon provided in . Observe that the second order total entropy can be rewritten in a form which shows that it depends only upon the grand mean firing rates across stimuli, and upon the correlation coefficient of the whole spike train, $`\mathrm{\Gamma }(t_i^a,t_j^b)`$ (defined across all trials rather than those with a given stimulus as for $`\gamma (t_i^a,t_j^b;s)`$). Thus,
$`{\displaystyle \frac{T^2}{2}}H_{tt}`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }t^2}{2\mathrm{ln}2}}{\displaystyle \underset{ab}{}}{\displaystyle \underset{t_1^a}{}}{\displaystyle \underset{t_2^b}{}}\overline{r}_a(t_1^a;s)_s\overline{r}_b(t_2^b;s)_s`$ (11)
$`\times `$ $`\left\{\mathrm{\Gamma }_{ab}(t_i^a,t_j^b)[1+\mathrm{\Gamma }_{ab}(t_i^a,t_j^b)]\mathrm{ln}[1+\mathrm{\Gamma }_{ab}(t_i^a,t_j^b)]\right\}.`$ (12)
It follows that the second order entropy is maximal when $`\mathrm{\Gamma }=0`$, and non-zero overall correlations in the spike trains (indicating statistical dependence) always decrease the total response entropy. $`\gamma (s)`$ acts on the noise entropy as $`\mathrm{\Gamma }`$ does upon the total entropy – it can only decrease the conditional entropy. The effect of $`\gamma (s)`$ on the total entropy is more complex, depending upon the correlation of the firing across stimuli. $`\gamma (s)`$ can be chosen so as to increase the total entropy (and thus the information, with the noise entropy fixed), and this increase will be maximal for the $`\gamma (s)`$ which lead exactly to $`\mathrm{\Gamma }=0`$. Neuronal or spike time interaction may therefore eliminate or reduce the effect of statistical dependencies introduced by other covariations.
The rate and correlation functions in practice must be estimated from a limited number of experimental trials, which leads to a bias in each of the entropy components. This bias was corrected for, as described in ; however, the sampling advantage that will be described was observed both with this correction, without bias correction, and with other bias correction approaches such as that used in .
To demonstrate its applicability, we applied the series entropy analysis to data recorded from the primary visual cortex (V1) of anaesthetised macaque monkeys . Fig. 2 examines, for a typical V1 complex cell, the dependence of the accuracy of the noise entropy estimate upon the number of experimental trials utilised. It is the noise entropy which is most affected by sampling constraints, so we shall concentrate upon this quantity here. The top panel shows the estimates before application of a bias removal procedure, using the series (our technique) and ‘brute force’ (simple application of Eqn. 1) approaches. The entropies are expressed as a fraction of the asymptotic entropy obtained by polynomial extrapolation . Reliable extrapolation to the asymptotic entropy was possible because of the large amount of data that happened to be available for this cell (which was chosen with that in mind; more usually between 20 and 100 trials were available). This allowed us to compare the performance of the methods on smaller subsets of the data against a known reference. The fact that series and brute-force estimators converged for this cell indicates that higher order correlations amongst spike times contributed little to the entropy.
The better performance of the series approach can be understood by considering that (at second order) it requires sampling from only the first two moments of the probability distribution, whereas the ‘brute force’ approach depends upon all moments. Higher moments have to be computed from events with lower and lower probability, as shown in Eqn. 4; estimation of these lower probability events is more error-prone, and leads to the larger bias of the ‘brute force’ approach.
Also shown in Fig. 2 is the Ma lower bound upon the entropy , which has been proposed as a useful bound which is relatively insensitive to sampling problems . The Ma bound is tight only when the probability distribution of words at fixed spike count is close to uniform. It can be seen that for the V1 complex cell data, the Ma bound is not tight at all. To understand the behaviour of the Ma bound for short time windows, we calculated series terms. The Ma entropy already differs from the true entropy at first order:
$`TH_t^{\mathrm{Ma}}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{ln}2}}{\displaystyle \underset{a}{}}{\displaystyle \underset{t_1^a}{}}\overline{r}_a(t_1^a;s)\mathrm{\Delta }t_s`$ (13)
$``$ $`{\displaystyle \underset{a;t_1^a}{}}\overline{r}_a(t_1^a;s)\mathrm{\Delta }t_s\mathrm{log}_2{\displaystyle \frac{\underset{a;t_1^a}{}\overline{r}_a(t_1^a;s)_s^2\mathrm{\Delta }t}{_{a;t_1^a}\overline{r}_a(t_1^a;s)_s}}`$ (14)
This coincides with Eqn. 5 only if there are no variations of rate across time and cells. If there were higher frequency rate variations, or more cells with different response profiles, the Ma bound would be still less useful.
Estimation quality depends upon not just sampling bias, but also variance; these can be summarised by the RMS error of the entropy estimate. We investigated the behaviour of the RMS error by fitting a Poisson model with matched time-dependent firing rate to the experimental data of Fig. 1. This model, although yielding a 5% lower noise entropy (because of correlations in the real data), predicted the ‘brute force’ sampling characteristics of Fig. 2 almost exactly. The model was used to generate a larger set of data (10,000 trials, or 160,000 stimulus presentations in total). This model yields worst-case sampling for the ‘brute force’ estimator; worst-case sampling for the series estimator would be achieved by even spread of probability throughout only the second order response space. The simulation serves to compare the estimators in a statistical regime similar to that of the typical cell of Fig. 2.
Fig. 3 shows the scaling of the RMS error before bias correction with data-size in this simulation. Scaling is qualitatively similar (but with a sharper decrease) after correction. The scaling behaviour resulting from the simulation predicts that with a ‘brute force’ approach, a RMS error of 2% of the entropy at a wordlength of 12 would require around 1400 trials with, and greater than 5000 trials without, application of the finite sampling correction. The series estimator reduces these requirements to approximately 50 and 400 trials respectively. These figures are dependent upon data statistics, and should be checked on a case by case basis; however, the dimensionality reduction with the series expansion provides a general improvement in the quality of entropy estimates for short time windows.
Some readers may wonder whether this new method amounts to computing the entropy with words with greater than 2 spikes thrown out. This is not the case: the proposed method considers pairwise interactions amongst all spikes in the word, no matter how many there are. It thus (unlike a truncated brute force approach) obtains the ability to take into account almost all of the entropy of longer words, while retaining the sampling benefits of being a second order method.
As neuroscience enters a quantitative phase, information theoretic techniques are being found useful for the analysis of data from physiological experiments. The methods developed here may broaden the scope of the study of neuronal information properties. In particular, they render feasible the information theoretic analysis of some recordings from anaesthetised and awake mammalian cerebral cortices.
SRS is supported by the HHMI, and SP by the Wellcome Trust. |
warning/0001/gr-qc0001056.html | ar5iv | text | # On Average Properties of Inhomogeneous Cosmologies
## 1 Motivation and Results in Newtonian Cosmology
### 1.1 Bridging the Gap
Does an inhomogeneous model of the Universe evolve on average like a homogeneous solution of Einstein‘s or Newton’s laws of gravity ? This question is not new, at least among relativists who think that the answer is certainly no, not only in view of the nonlinearity of the theories mentioned . The problem was and still is the notion of averaging whose specification and unambiguous definition turned out to be an endeavor of high magnitude, mainly because it is not straightforward to give a unique meaning to the averaging of tensors, e.g., a given metric of spacetime. This problem seems to lie in the backyard of relativists who, from time to time, add another effort towards a solution of this problem. On the other hand, the community of cosmologists “should” locate exactly this problem at the basis of their evolutionary models of the Universe. Although there are many exceptions (e.g. , , , , , , ; , , , a certainly incomplete list), most researchers in this field are drawn back to the historical development of cosmologies starting with Friedmann, Einstein and de–Sitter at the beginning of the last century. Despite the drastic changes of our picture of structures in the Universe on large scales, still, the cosmologist’s thinking rests on the hegemony of the so–called “standard model” (i.e. the family of FLRW models for homogeneous and isotropic matter distributions). This standard model, up to the present state of knowledge, explains (or better is employed to explain) a wide variety of orthogonal observations, and it is therefore hard to beat due to its (suggestively) established status of resistance against observational tests. Therefore, most discussions in this field are based on the vocabulary of the standard model, aiming to constrain its “cosmological parameters”, often on the basis of observations of structure in the regional Universe that is very different from homogeneous and isotropic.
Bridging the gap between an involved mathematical problem of general relativity and the practical modeling of cosmological dynamics is possible, if ambiguities of averagers could be removed, and results related to contemporary discussions in observational cosmology. In the sequel we shall follow a line of thought that will match this need and could intensify work directed to mastering an inhomogeneous spacetime.
### 1.2 Setting the Pace
Averaging procedures can be defined in a vast variety of ways (see, e.g., a recent summary by Stoeger et al. ). A smoothing–out operator can live on different foliations of spacetime; it can leave an averaged field space–dependent, thus operating on a given spatial scale; it can also smooth out all inhomogeneities. It could involve statistical ensemble averaging. It can act on scalars, vectors or tensors; it can smooth matter variables or, most elegantly, the geometry of spacetime itself , . It can average spacetime variables and not merely spatial variables, etc. To remove ambiguity here, it is best to work within the framework of Newtonian cosmology as a first step of understanding, since there the choice of foliation is not a problem and spatial averaging can be simply put into practice by Euclidean volume integration. As a second step, we confine ourselves to scalar variables. This allows us to get at least some corresponding answer in general relativity, since the averaging of scalars is, for a given foliation, a covariant operation. We therefore propose to look at the simplest (mass–conserving) averager as follows.
### 1.3 A Newtonian Averager
Consider any simply–connected spatial domain in Newtonian spacetime. With $`_𝒟`$ we denote spatial averaging in Eulerian space, e.g., for a spatial tensor field $`𝒜(𝐱,t)=\{A_{ij}(𝐱,t)\}`$ we simply have the Euclidean volume integral normalized by the volume of the domain:
$$𝒜_𝒟(t)=\frac{1}{V(t)}_𝒟\mathrm{d}^3x𝒜(𝐱,t).$$
(1)
(Here, $`𝐱`$ are non–rotating Eulerian coordinates.) Since, with this averager, the averaged field is only time–dependent, the space–dependence is only implicit by a functional dependence on the domain‘s morphology and position. We shall evolve the domain in time by preserving its mass content. This is a natural assumption, if we also want to extend the domain to the whole Universe.
Now, consider the volume of an Eulerian spatial domain $`𝒟`$ at a given time, $`V=_𝒟\mathrm{d}^3x`$, and follow the position vectors $`𝐱=𝐟(𝐗,t)`$ of all fluid elements (indexed by the Lagrangian coordinates $`𝐗`$) within the domain. Then, the volume elements are deformed according to $`\mathrm{d}^3x=J\mathrm{d}^3X`$, where $`J`$ is the Jacobian determinant of the transformation from Eulerian to Lagrangian coordinates. The total rate of change of the volume of the same collection of fluid elements may then be calculated as follows:
$$\frac{d_tV}{V}=\frac{1}{V}\mathrm{d}_t_{𝒟_𝐢}\mathrm{d}^3XJ=\frac{1}{V}_{𝒟_𝐢}\mathrm{d}^3Xd_tJ=\frac{1}{V}_{𝒟_𝐢}\mathrm{d}^3X\theta J=\theta _𝒟=:3H_𝒟,$$
(2)
where $`d_t`$ is the total (Lagrangian) time–derivative, $`\theta =𝐯=\frac{\dot{J}}{J}`$ the local expansion rate of a given velocity model $`𝐯(𝐱,t)`$; $`𝒟_𝐢`$ denotes the initial domain found by mapping back $`𝒟`$ with the help of $`𝐟^1`$ (provided $`𝐟^1`$ exists), and $`H_𝒟=\dot{a}_𝒟/a_𝒟`$ naturally defines an effective Hubble–parameter on the domain.
### 1.4 Commuting Averaging and Evolution
From the point of view of the standard model of cosmology it makes no difference, if we smooth the initial inhomogeneities (say, at a time in the matter dominated epoch) and then evolve the smoothed data with a FLRW solution, or if we evolve these inhomogeneities until present and then smooth the distribution. The conjecture is held that both ways produce the same values for the characteristic parameters of a FLRW cosmology.
However, averaging, as defined above, and evolving inhomogeneities are non–commuting operations. To see this we have to notice that the total (Lagrangian) time–derivative does not commute with spatial averaging in Eulerian space. For an arbitrary tensor field $`𝒜`$ we can readily derive, with the help of the above definitions, the following Commutation Rule :
$$\mathrm{d}_t𝒜_𝒟\mathrm{d}_t𝒜_𝒟=𝒜\theta _𝒟\theta _𝒟𝒜_𝒟.$$
(3)
This tells us that exchanging the operators for averaging and time–evolution produces a source due to the presence of inhomogeneities (that we will discuss to consist of positive–definite fluctuations). As an example consider the expansion rate itself. Setting $`𝒜=\theta `$, we get
$$\mathrm{d}_t\theta _𝒟\mathrm{d}_t\theta _𝒟=\theta ^2_𝒟\theta _𝒟^2=(\theta \theta _𝒟)^2_𝒟\mathrm{\hspace{0.33em}0},$$
(4)
i.e., the source is the averaged mean square fluctuation of the expansion rate. It vanishes for the case where the local rate equals the global one, which is true for homogeneous–isotropic matter distributions.
### 1.5 Formulating the General Expansion Law of Newtonian Cosmology
As suggested by Eq. (2) we may measure the effective expansion of a portion of the Universe with the help of the rate of volume change. Let us introduce an effective dimensionless scale–factor $`a_𝒟`$ via the domain’s volume $`V(t)=|𝒟|`$ and the initial volume $`V_𝐢=V(t_𝐢)=|𝒟_𝐢|`$ (compare Fig. 1):
$$a_𝒟(t)=\left(\frac{V(t)}{V_𝐢}\right)^{\frac{1}{3}},\mathrm{i}.\mathrm{e}.,H_𝒟=\frac{\dot{a}_𝒟}{a_𝒟}.$$
(5)
For domains $`𝒟`$ with constant mass $`M_𝒟`$, as for Lagrangian defined domains, the average mass density evolves as:
$$\varrho _𝒟=\frac{\varrho (t_𝐢)_{𝒟_𝐢}}{a_𝒟^3}=\frac{M_𝒟}{a_𝒟^3V_𝐢}.$$
(6)
Looking at the Commutation Rule, Eq. (4), we appreciate that this rule furnishes a purely kinematical relation provided we give a velocity model. Dynamics enters by saying how this velocity model is generated by gravity. In Eq. (4) we can express most of the terms already through the scale–factor and its time–derivatives. The only unknown is the evolution equation for the local expansion rate. This is furnished by Raychaudhuri’s equation (which employs the Newtonian field equation $`𝐠=\mathrm{\Lambda }4\pi G\varrho `$ for the gravitational field strength of dust matter $`𝐠=\dot{𝐯}`$).
Inserting Raychaudhuri’s equation,
$$\dot{\theta }=\mathrm{\Lambda }4\pi G\rho \frac{1}{3}\theta ^2+2(\omega ^2\sigma ^2);\sigma :=\sqrt{\frac{1}{2}\sigma _{ij}\sigma _{ij}},\omega :=\sqrt{\frac{1}{2}\omega _{ij}\omega _{1ij}},$$
(7)
into Eq. (4), we find that the scale–factor $`a_𝒟`$ in general obeys the dynamical expansion law :
$$3\frac{\ddot{a}_𝒟}{a_𝒟}+4\pi G\varrho _𝒟\mathrm{\Lambda }=Q_𝒟,$$
(8)
with Newton’s gravitational constant $`G`$, the cosmological constant $`\mathrm{\Lambda }`$, and the source term $`Q_𝒟`$, which we may call “backreaction term”, since it measures the departure from the standard model due to the influence of inhomogeneities.
$`Q_𝒟`$ depends on the kinematical scalars, the expansion rate $`\theta `$, the rate of shear $`\sigma `$, and the rate of vorticity $`\omega `$ featuring three positive–definite fluctuation terms:
$$Q_𝒟=\frac{2}{3}\left(\theta ^2_𝒟\theta _𝒟^2\right)+2\omega ^2_𝒟2\sigma ^2_𝒟.$$
(9)
### 1.6 The Quintessence of Inhomogeneities and Anisotropies
It is easily verified that $`Q_𝒟=0`$ is a necessary and sufficient condition for having $`a_𝒟=a(t)`$, with the global scale–factor $`a(t)`$ solving the standard Friedmann equations. Assuming for the moment that the averaged Universe would be described by the standard model, then the universal expansion rate would be determined by four components: the average mass density (commonly conceived to have baryonic and non–baryonic “dark” components), the value of the cosmological constant, and the global “curvature parameter” $`k`$, which arises as an integration constant by integrating the second–order Friedmann equation (Eq. (8) for $`Q_𝒟=0`$) in order to get the global expansion rate $`3H(t)=3\frac{\dot{a}}{a}`$.
Upon integrating the actual expansion law, Eq. (8), we instead obtain ( used a different sign convention for $`Q_𝒟`$):
$$3\frac{\dot{a}_𝒟^2}{a_𝒟^2}+3\frac{k_𝒟}{a_𝒟^2}8\pi G\varrho _𝒟\mathrm{\Lambda }=\frac{1}{a_𝒟^2}_{t_𝐢}^tdt^{}Q_𝒟\frac{\mathrm{d}}{\mathrm{d}t^{}}a_𝒟^2(t^{}),$$
(10)
where $`k_𝒟`$ enters as a domain–dependent integration constant. The effective Hubble–parameter $`H_𝒟=\frac{\dot{a}_𝒟}{a_𝒟}`$ is now determined by the previous (now domain–dependent) components, but also by a fifth component due to the presence of inhomogeneities.
Since there are many unresolved conundrums of the standard model (see, e.g., , ), cosmologists are searching for a fifth parameter that may resolve them (see the talk by Prof. Fujii on “Quintessence” in the present Proceedings). The use of the letter $`Q`$ for the “backreaction” is coincidential, but it may possibly be a working coincidence.
### 1.7 Cosmic Triangle or Cosmic Quartett ?
The vocabulary of the standard model may be condensed into the picture of a “cosmic triangle” , consisting of three cosmological parameters, which are constructed by normalizing the mass density, the cosmological constant, and the “curvature parameter” by the square of the Hubble–parameter.
As suggested by Eq. (10) we are led to define an additional dimensionless “kinematical backreaction parameter” through
$$\mathrm{\Omega }_Q^𝒟:=\frac{1}{3a_𝒟^2H_𝒟^2}_{t_𝐢}^tdt^{}Q_𝒟\frac{\mathrm{d}}{\mathrm{d}t^{}}a_𝒟^2(t^{}),$$
(11)
in addition to the common cosmological parameters:
$$\mathrm{\Omega }_m^𝒟:=\frac{8\pi G\varrho _𝒟}{3H_𝒟^2},\mathrm{\Omega }_\mathrm{\Lambda }^𝒟:=\frac{\mathrm{\Lambda }}{3H_𝒟^2},\mathrm{\Omega }_k^𝒟:=\frac{k_D}{a_𝒟^2H_𝒟^2}.$$
(12)
However, contrary to the standard model, all $`\mathrm{\Omega }^𝒟`$–parameters are now domain–dependent and transformed into fluctuating fields on the domain; for $`Q_𝒟=0`$ the “cosmic triangle” is undistorted and the parameters acquire their global standard values.
Comparing these definitions with Eq. (10) we have
$$\mathrm{\Omega }_m^𝒟+\mathrm{\Omega }_\mathrm{\Lambda }^𝒟+\mathrm{\Omega }_k^𝒟+\mathrm{\Omega }_Q^𝒟=\mathrm{\hspace{0.33em}1},$$
(13)
i.e., there are four players in the game (or five, respectively, if we split $`\mathrm{\Omega }_m^𝒟`$ into baryonic matter and non–baryonic “dark matter”).
In Friedmann–Lemaître cosmologies there is by definition no backreaction: $`\mathrm{\Omega }_Q^𝒟=0`$. In this case a universe model with $`\mathrm{\Omega }_m=1`$ conserves the matter parameter. Also, the initial value of the “curvature parameter” $`\mathrm{\Omega }_k=0`$ remains so during the entire evolution. This changes, if “backreaction” is taken into account. $`\mathrm{\Omega }_Q^𝒟`$ itself may act as a “kinematical dark matter”, or as a “kinematical cosmological term”, respectively, depending on the relative strength of shear–, epansion– and vorticity–fluctuations.
As we have learned and as we shall illustrate below, the parameters corresponding to the “cosmic triangle” can experience large changes, even in situations when the “backreaction parameter” is seemingly negligible, but non–zero.
### 1.8 Construction Principle for Evolution Models
In order to obtain quantitative results on the value and impact of the “backreaction parameter”, we have to employ evolution models for the inhomogeneities. Their implementation seems not to be straightforward, since cosmological evolution models (analytical or N–body simulations) are constructed in such a way that they evolve an initial power spectrum of density– and velocity–fluctuations on a given global background (a solution of the standard Friedmann equations), implying a vanishing “backreaction” by construction. However, the framework of Newtonian cosmology offers the possibility of making use of contemporary evolution schemes by assigning sense to a global background within the general expansion law of an averaged inhomogeneous model. We shall now elaborate on this fact by demonstrating the validity of the following Construction Principle:
The average expansion of a generic inhomogeneous matter distribution on (topologically) closed Newtonian space sections is given by the solution of the standard Friedmann equations.
This principle, although usually not explicitly stated, lies at the basis of any evolution model in Newtonian cosmology. It appears to be very restrictive in light of the general relativistic framework (see Subsection 2.5).
In order to proof this principle let us work with the invariants of the gradient of the velocity field<sup>2</sup><sup>2</sup>2A comma denotes partial derivative with respect to Eulerian coordinates $`/x_i,i`$.. They are expressible in terms of kinematical scalars and can be written as total divergences of vector fields, which has been used and discussed in the context of perturbation solutions (, ):
$$𝐈(v_{i,j})=𝐯=\theta ,\mathrm{𝐈𝐈}(v_{i,j})=\frac{1}{2}\left(𝐯(𝐯)(𝐯)𝐯\right)=\omega ^2\sigma ^2+\frac{1}{3}\theta ^2.$$
(14)
The “backreaction term” can be entirely expressed in terms of the first and second invariants:
$$Q_𝒟=2\mathrm{𝐈𝐈}(v_{i,j})_𝒟\frac{2}{3}𝐈(v_{i,j})_𝒟^2.$$
(15)
Now, let us formally assume that there exists a global, homogeneous and isotropic reference model with scale–factor $`a(t)`$. We introduce the following variables with respect to this reference background. With the global Hubble–parameter $`H=\dot{a}/a`$ we define comoving Eulerian coordinates $`𝐪:=𝐱/a`$ and peculiar–velocities $`𝐮:=𝐯H𝐱`$ as usual. Using the derivative $`_{q_j}u_iu_i/q_j`$ with respect to comoving coordinates we obtain for the first and second invariants:
$$𝐈(v_{i,j})=3H+\frac{1}{a}𝐈\left(_{q_j}u_i\right),\mathrm{𝐈𝐈}(v_{i,j})=3H^2+\frac{2H}{a}𝐈\left(_{q_j}u_i\right)+\frac{1}{a^2}\mathrm{𝐈𝐈}\left(_{q_j}u_i\right).$$
The “backreaction term” remains form–invariant (note: $`\frac{1}{a}_{q_j}u_i=u_{i,j}`$):
$$Q_𝒟=\frac{1}{a^2}\left(2\mathrm{𝐈𝐈}\left(_{q_j}u_i\right)_𝒟\frac{2}{3}𝐈\left(_{q_j}u_i\right)_𝒟^2\right).$$
(16)
Thus, all terms corresponding to the background flow cancel in this expression, and only inhomogeneities contribute to “backreaction”.
Using Eq. (14) we write the “backreaction term” as a volume–average over divergences. Hence, using the theorem of Gauss we obtain:
$$Q_𝒟=\frac{1}{a^2}\left[2\frac{1}{V_q}_{𝒟_q}d𝐒\left(𝐮(_q𝐮)(𝐮_q)𝐮\right)\frac{2}{3}\left(\frac{1}{V_q}_{𝒟_q}d𝐒𝐮\right)^2\right],$$
(17)
with the surface $`𝒟_q`$ bounding the comoving domain $`𝒟_q`$, the surface element $`\mathrm{d}𝐒`$, and the comoving differential operator $`_q`$.
From Eq. (17) we directly obtain $`Q_𝒟=0`$ for a domain with empty boundary, e.g., for toroidal space sections, or periodic peculiar–velocity fields, respectively. In turn, the assumed reference solution is a standard Hubble–flow and may serve as a global background. (For more details see .)
This establishes the Construction Principleq.e.d.
### 1.9 Generalization of the Top–Hat Model
The Construction Principle provides room for employing the language of standard theories of structure formation. As an example we apply the general expansion law to domains within a global Hubble–flow.
Assuming an Einstein–de–Sitter background for this example we subtract a standard Friedmann equation, $`3\frac{\ddot{a}}{a}+4\pi G\varrho _H=0`$, from Eq. (8), with the background density $`\varrho _H=\frac{3H^2}{8\pi G}`$, and obtain the following differential equation for $`a_𝒟(t)`$ :
$$3\left(\frac{\ddot{a}_𝒟}{a_𝒟}\frac{\ddot{a}}{a}\right)+\frac{3}{2}\left(\frac{\dot{a}}{a}\right)^2\delta _𝒟=Q_𝒟,$$
(18)
with $`\delta _𝒟=(\varrho _𝒟\varrho _H)/\varrho _H`$ specifying the averaged density contrast $`\delta `$ in $`𝒟`$. For $`Q_𝒟=0`$ and $`\delta _𝒟=0`$ this equation simply states that the time evolution of a domain follows the global expansion, $`a_𝒟(t)=a(t)`$. For $`Q_𝒟=0`$ and $`1+\delta _𝒟=\frac{\varrho _𝒟a^3}{\varrho _Ha_𝒟^3}`$ the evolution of $`a_𝒟`$ is still of Friedmann type, but with a mass different from the background mass. An important subcase with $`Q_𝒟=0`$ and $`\delta _𝒟0`$ is the well–known spherical top–hat model . In Eq. (18) there are two sources determining the deviations from the Friedmann acceleration, the over/under–density and the “backreaction term”. This shows that, in general, the evolution of a Newtonian portion of the Universe is triggered by an over/under–density and velocity–fluctuations.
### 1.10 An Averaged Lagrangian Perturbation Scheme
As a second example we apply the Construction Principle to the widely used and well–tested Lagrangian perturbation schemes (see: , for reviews, and for a tutorial).
The trajectory field in the mostly employed “Zel’dovich approximation” , (which can be derived as a subcase of the first–order Lagrangian perturbation scheme ), is given by:
$$𝐟^Z(𝐗,t)=a(t)\left(𝐗+\xi (t)_0\psi (𝐗)\right).$$
(19)
$`\psi (𝐗)`$ is the initial displacement field, $`_0`$ the gradient with respect to Lagrangian coordinates and $`\xi (t)`$ a global time–dependent function (given for all background models in ).
Given this trajectory field we can already approximate the rate of volume change by the volume deformation that is caused by the field. For the effective scale–factor we obtain :
$$(a_𝒟^{\mathrm{kin}})^3=a^3\left(1+\xi 𝐈_𝐢_{𝒟_𝐢}+\xi ^2\mathrm{𝐈𝐈}_𝐢_{𝒟_𝐢}+\xi ^3\mathrm{𝐈𝐈𝐈}_𝐢_{𝒟_𝐢}\right).$$
(20)
A better estimate is to calculate the “backreaction term” from the approximation of the velocity field $`𝐯^Z=\dot{𝐟}^Z`$ and solve the dynamical expansion law, Eq. (8), for $`a_𝒟`$. The “backreaction term” separates into its time–evolution given by $`\xi (t)`$ and the spatial dependence on the initial displacement field given by averages over the invariants of the displacement gradient $`𝐈_𝐢,\mathrm{𝐈𝐈}_𝐢`$, and $`\mathrm{𝐈𝐈𝐈}_𝐢`$:
$`Q_𝒟^Z={\displaystyle \frac{\dot{\xi }^2}{\left(1+\xi 𝐈_𝐢_{𝒟_𝐢}+\xi ^2\mathrm{𝐈𝐈}_𝐢_{𝒟_𝐢}+\xi ^3\mathrm{𝐈𝐈𝐈}_𝐢_{𝒟_𝐢}\right)^2}}\times `$ (21)
$`\left[\left(2\mathrm{𝐈𝐈}_𝐢_{𝒟_𝐢}{\displaystyle \frac{2}{3}}𝐈_𝐢_{𝒟_𝐢}^2\right)+\xi \left(6\mathrm{𝐈𝐈𝐈}_𝐢_{𝒟_𝐢}{\displaystyle \frac{2}{3}}𝐈_𝐢_{𝒟_𝐢}\mathrm{𝐈𝐈}_𝐢_{𝒟_𝐢}\right)+\xi ^2\left(2𝐈_𝐢_{𝒟_𝐢}\mathrm{𝐈𝐈𝐈}_𝐢_{𝒟_𝐢}{\displaystyle \frac{2}{3}}\mathrm{𝐈𝐈}_𝐢_{𝒟_𝐢}^2\right)\right].`$
The numerator of the first term is global and corresponds to the damping factor that also arises in the Eulerian linear theory of gravitational instability; in an Einstein–de–Sitter universe $`\dot{\xi }^2a^1`$. The denominator of the first term is the volume effect discussed above, whereas the second term in brackets features the initial “backreaction” as a leading term and higher–order terms.
A property of this approximation that renders it very useful as an averaged evolution model should be emphasized: it is exact for two orthogonal symmetry assumptions: for the evolution of plane–symmetric inhomogeneities, which is a consequence of the known properties of the first–order Lagrangian approximation , and for the evolution of spherically symmetric inhomogeneities, as will be shown below.
Employing the averaged Lagrangian scheme we can quantify the impact of backreaction on the domain–dependent “cosmological parameters”. For details the reader may look at a recent work . Here, I would like to mention the remarkable result that for an accelerating, i.e., under–dense region, $`\mathrm{\Omega }_Q^𝒟`$ may be numerically negligible as seen in (Fig. 2, upper plots), but dramatic changes in the other parameters are observed. For a collapsing domain with a present–day radius of 100Mpc the mass parameter of the domain may even differ by more than 100% from the global mass parameter (Fig. 2, lower plots).
### 1.11 Newton’s Iron Spheres
Looking at the general expansion law, Eq. (8), the careful reader may object that Friedmann’s equations also hold, if we study the motion of a spherically symmetric domain, although the matter distribution inside the sphere may be inhomogeneous. This fact is known as Newton’s “Iron Sphere Theorem”. Let us show that the general expansion law respects this theorem.
We note that, for a spherically symmetric distribution of matter within the domain, the invariants of the velocity gradient, averaged over a ball $`_R`$, obey relations resulting in :
$$\mathrm{𝐈𝐈}(v_{i,j})__R=\frac{1}{3}𝐈(v_{i,j})__R^2,\mathrm{and}\mathrm{𝐈𝐈𝐈}(v_{i,j})__R=\frac{1}{27}𝐈(v_{i,j})__R^3.$$
(22)
Inserting this into the “backreaction term”, Eq. (15), shows that $`Q__R^{\mathrm{spherical}}=0`$ in accordance with Newton’s theorem. The inhomogeneous model discussed in the last subsection also has this property: inserting the above expressions into Eq. (21) (using the proportionality of displacement gradient and initial velocity gradient), we also obtain $`Q__R^Z=0=Q__R^{\mathrm{spherical}}`$.
### 1.12 Averaging and the Evolution of Form
The expansion law, Eq. (8), is built on the rate of change of a simple morphological quantity, the volume content of a domain. Although functionally it depends on other morphological characteristics of a domain, it does not explicitly provide information on their evolution. An evolution equation for the “backreaction term” is missing. This fact touches on the problem of closing the hierarchy of dynamical evolution equations considered as a set of coupled ordinary differential equations in Lagrangian space. The problem of closing such a hierarchy of equations is often considered in the literature and various closure conditions are formulated (e.g., ), one of them being the “Silent Universe Model” in general relativity, which assumes a vanishing magnetic part of the Weyl tensor . Averaging such a hierarchy would result in evolution equations for the “backreaction term” and would, with some local closure condition, also close the system of averaged equations. Here, we will not pursue this problem further, but instead begin to develop the morphological point of view that eventually implies an alternative proposition of a morphological closure condition.
Let us focus our attention on the boundary of the spatial domain $`𝒟`$. A priori, the location of this boundary in space enjoys some freedom which we may constrain by saying that the boundary coincides with a velocity front of the fluid (hereby restricting attention to irrotational flows). This way we employ the Legendrian point of view of velocity fronts that is dual to the Lagrangian one of fluid trajectories. Let $`S(x,y,z,t)=s(t)`$ define a velocity front at Newtonian time $`t`$, $`𝐯=S`$. Below, we shall need that the three principal scalar invariants of the velocity gradient $`v_{i,j}=:S_{,ij}`$ can be transformed into divergences of vector fields as written explicitly in Eqs. (14,25).
Defining the unit normal vector $`𝐧`$ on the front, $`𝐧=\pm \frac{S}{|S|}`$ (the sign depends on the direction of its motion), the average expansion rate can be written as a flux integral using Gauss’ theorem:
$$\theta _𝒟=\frac{1}{V}_𝒟d^3x𝐯=\frac{1}{V}_𝒟\mathrm{𝐝𝐒}𝐯,$$
(23)
and, with the surface element $`d\sigma `$, $`\mathrm{𝐝𝐒}=𝐧d\sigma `$, we obtain the intuitive result that the average expansion rate is related to another morphological quantity of the domain, the total area of the enclosing surface:
$$\theta _𝒟=\frac{1}{V}_𝒟𝑑\sigma |S|.$$
(24)
Inserting the velocity potential also into the other invariants, Eq. (14) and
$$\mathrm{𝐈𝐈𝐈}(v_{i,j})=\frac{1}{3}\left(\frac{1}{2}\left(𝐯(𝐯)(𝐯)𝐯\right)𝐯\left(𝐯(𝐯)(𝐯)𝐯\right)𝐯\right),$$
(25)
and performing the spatial average, we obtain :
$$\mathrm{𝐈𝐈}_𝒟=\frac{1}{V}_𝒟d^3x\mathrm{𝐈𝐈}=_𝒟𝑑\sigma |S|^2𝐇;\mathrm{𝐈𝐈𝐈}_𝒟=\frac{1}{V}_𝒟d^3x\mathrm{𝐈𝐈𝐈}=_𝒟𝑑\sigma |S|^3𝐊,$$
(26)
where $`𝐇`$ is the mean curvature and $`𝐊`$ the Gaussian curvature of the $`2`$surface bounding the domain. $`|S|=\frac{ds}{dt}`$ equals $`1`$, if the instrinsic arc–length $`s`$ of the trajectories is used instead of the extrinsic Newtonian time $`t`$. The averaged invariants comprise, together with the volume a complete set of morphological characteristics related to the Minkowski functionals of a body:
$$𝒲_0(s):=_𝒟d^3x=V;𝒲_1(s):=\frac{1}{3}_𝒟𝑑\sigma ;𝒲_2(s):=\frac{1}{3}_𝒟𝑑\sigma 𝐇;𝒲_3(s):=\frac{1}{3}_𝒟𝑑\sigma 𝐊=\frac{4\pi }{3}\chi .$$
(27)
The Euler–characteristic $`\chi `$ determines the topology of the domain and is assumed to be an integral of motion ($`\chi =1`$), if the domain should remain simply–connected (a morphological closure condition).
Thus, we have gained a morphological interpretation of the “backreaction term”: it can be entirely expressed through three of the four Minkowski functionals:
$$Q_𝒟(s)=6\left(\frac{𝒲_2}{𝒲_0}\frac{𝒲_1^2}{𝒲_0^2}\right).$$
(28)
The $`𝒲_\alpha ;\alpha =0,1,2,3`$ have been introduced as “Minkowski functionals” into cosmology by Mecke et al. () in order to statistically assess morphological properties of cosmic structure. Minkowski functionals proved to be useful tools to also incorporate information from higher–order correlations, e.g., in the distribution of galaxies, galaxy clusters, density fields or cosmic microwave background temperature maps (, , , ; see the review by Kerscher and ref. therein). Related to the morphology of individual domains is the study of building blocks of large–scale cosmic structure , .
For a ball with radius $`R`$ we have for the Minkowski functionals:
$$𝒲_0^_R(s):=\frac{4\pi }{3}R^3;𝒲_1^_R(s):=\frac{4\pi }{3}R^2;𝒲_2^_R(s):=\frac{4\pi }{3}R;𝒲_3^_R(s):=\frac{4\pi }{3}.$$
(29)
Inserting these expressions into the “backreaction term”, Eq. (28), shows that $`Q_𝒟^_R(s)=0`$, and we have confirmed Newton‘s “Iron Sphere Theorem” once more. Moreover, we can understand now that the “backreaction term” encodes the deviations of the domain‘s morphology from that of a ball, a fact which we shall illustrate now with the help of Steiner‘s formula of integral geometry (see also ).
Let $`d\sigma _0`$ be the surface element on the unit sphere, then (according to the Gaussian map) $`d\sigma =R_1R_2d\sigma _0`$ is the surface element of a $`2`$surface with radii of curvature $`R_1`$ and $`R_2`$. Moving the surface a distance $`\epsilon `$ along its normal we get for the surface element of the parallel velocity front:
$$d\sigma _\epsilon =(R_1+\epsilon )(R_2+\epsilon )d\sigma _0=\frac{R_1R_2+\epsilon (R_1+R_2)+\epsilon ^2}{R_1R_2}d\sigma =(1+\epsilon 2𝐇+\epsilon ^2𝐊)d\sigma ,$$
(30)
where
$$𝐇=\frac{1}{2}\left(\frac{1}{R_1}+\frac{1}{R_2}\right),𝐊=\frac{1}{R_1R_2},$$
(31)
are the mean curvature and Gaussian curvature of the front as before.
Integrating Eq. (30) over the whole front we arrive at a relation between the total surface area $`A`$ of the front and $`A_\epsilon `$ of its parallel front. The gain in volume may then be expressed by an integral of the resulting relation with respect to $`\epsilon `$ (which is known as Steiner‘s formula defining the Minkowski functionals of a (convex) body in three spatial dimensions):
$$V_\epsilon =V+_0^\epsilon 𝑑\epsilon ^{}A_\epsilon ^{}=V+\epsilon A+\epsilon ^2_𝒟𝑑\sigma 𝐇+\epsilon ^3_𝒟𝑑\sigma 𝐊.$$
(32)
## 2 Average Models in General Relativity
### 2.1 An Averager for Scalars
The key to reducing ambiguity of choosing an averager in general relativity is to confine ourselves to scalars, since spatially averaging a scalar field $`\mathrm{\Psi }`$ is a covariant operation. However, we have to make sure that we understand the word “spatial” physically. Geometrically, we shall introduce a foliation of spacetime, but we shall encounter ambiguity in choosing it. Suppose for the moment that we insist on spatial averaging and consider for all what follows a foliation of spacetime as given. The simplest averager is then immediately written down as follows:
$$\mathrm{\Psi }(X^i,t)_𝒟:=\frac{1}{V}_𝒟Jd^3X\mathrm{\Psi }(X^i,t);V=_𝒟Jd^3X,$$
(33)
with $`J:=\sqrt{det(g_{ij})}`$, where $`g_{ij}`$ is the metric of the spatial hypersurfaces, and $`X^i`$ are coordinates in the hypersurfaces, conveniently chosen such that they are constant along flow lines.
### 2.2 Foliating Spacetime: Covariant Fluid Approach
We wish to conserve mass inside a spatial domain of spacetime. Therefore, let us first consider the (conserved) restmass flux vector<sup>3</sup><sup>3</sup>3Greek indices run through $`0\mathrm{}3`$, while latin indices run through $`1\mathrm{}3`$ as before; summation over repeated indices is understood. A semicolon will denote covariant derivative with respect to the 4–metric with signature $`(,+,+,+)`$; the units are such that $`c=1`$.
$$M^\mu :=\varrho u^\mu ;M_{;\mu }^\mu =0;\varrho >0,$$
(34)
where $`\varrho `$ is the restmass density and the flow lines are integral curves of the 4–velocity $`u^\mu `$. We shall confine ourselves to irrotational perfect fluids with energy density $`\epsilon `$ and pressure $`p`$, which allows us to simplify the splitting of spacetime, e.g. the spatial hypersurfaces can be chosen flow–orthogonal in this case, the unit normal on the hypersurfaces coincides with the 4–velocity and the shift vector vanishes. Irrotationality guarantees the existence of a scalar function $`S`$, such that
$$u^\mu =:\frac{^\mu S}{h},$$
(35)
In general, we identify the magnitude $`h`$ with the “injection energy per fluid element and unit restmass”,
$$h:=\frac{\epsilon +p}{\varrho },$$
(36)
which is related to the relativistic enthalpy $`\eta :=\frac{\epsilon +p}{n}`$ by $`h=\eta /m`$ with $`m`$ the unit restmass of a fluid element, and $`n`$ the baryon density. Note that $`d\epsilon =hd\varrho `$. For a barotropic fluid we can easily see that $`\epsilon `$ is a function of the restmass density only and, hence, $`h`$ is a function of $`\varrho `$. $`h`$ is identical to $`1`$ in the case of dust. The magnitude $`h`$ normalizes the 4–gradient $`^\mu S`$ so that $`u^\mu u_\mu =1`$,
$$h=\sqrt{^\alpha S_\alpha S}=u^\mu _\mu S=\dot{S}>\mathrm{\hspace{0.33em}0}.$$
(37)
The overdot stands for the material derivative operator along the flow lines of any tensor field $``$ as defined covariantly by $`\dot{}:=u^\mu _{;\mu }`$. It reduces in the case of a scalar function in a flow–orthogonal slicing (which we want to envisage) to the partial time–derivative multiplied by $`\frac{1}{N}`$, where $`N`$ is the lapse function (for more details see: ).
It can be shown that $`S`$ is spatially homogeneous. Thus, $`S(t)`$ and $`h(X^i,t)`$ play the role of “phase” and “amplitude” of the fluid’s wave fronts. Foliating the spacetime into hypersurfaces $`S(t)=const.`$ is a gauge condition that is naturally adapted to the fluid itself and, therefore, this foliation can be given a covariant meaning (, and ref. therein).
### 2.3 A Commutation Rule in General Relativity
The rate of change of the volume $`V(t)`$ in the hypersurfaces $`S(t)=const.`$ is evaluated by taking the partial time–derivative of the volume and dividing by the volume. Since $`_t`$ and $`d^3X`$ commute (but not $`\frac{d}{d\tau }:=\frac{_t}{N}`$ and $`d^3X`$ !) we obtain:
$$\frac{_tV}{V}=\frac{1}{V}_𝒟d^3X_tJ=\frac{1}{V}_𝒟d^3XN\dot{J}=\frac{1}{V}_𝒟d^3XN\theta J=N\theta _𝒟.$$
(38)
Introducing the scaled (t–)expansion $`\stackrel{~}{\theta }:=N\theta `$ we define an effective (t–)Hubble function in the hypersurfaces by
$$\stackrel{~}{\theta }_𝒟=\frac{_tV}{V}=3\frac{_ta_𝒟}{a_𝒟}=:3\stackrel{~}{H}_𝒟.$$
(39)
With the help of these definitions we readily derive the Commutation Rule:
$$_t\mathrm{\Psi }_𝒟_t\mathrm{\Psi }_𝒟=\mathrm{\Psi }\stackrel{~}{\theta }_𝒟\mathrm{\Psi }_𝒟\stackrel{~}{\theta }_𝒟.$$
(40)
The lapse–weighted quantities have to be introduced only in the case of non–vanishing pressure, since the pressure gradient induces deviations from a geodesic flow implying an inhomogeneous lapse function. In the much simpler case of a dust matter model, the lapse function is homogeneous and can be chosen equal to $`1`$, and the covariant fluid gauge is identical to the comoving and synchronous gauge, which makes the correspondence to the Newtonian investigation most transparent.
In the sequel we shall, for simplicity, discuss the case of a pressure–less fluid only and resume the discussion of the more general case thereafter.
### 2.4 Expansion Law of General Relativity: Dust Models
Employing the averaging procedure outlined above and following the line of thought of the Newtonian investigation in Section 1, we derive the general expansion law for dust matter in the synchronous and comoving gauge. The result will be covariant with respect to this foliation.
I here give the result derived in : the spatially averaged equations for the scale–factor $`a_𝒟`$, respecting mass conservation, read:
averaged Raychaudhuri equation:
$$3\frac{\ddot{a}_𝒟}{a_𝒟}+4\pi G\frac{M_𝒟}{V_𝐢a_𝒟^3}\mathrm{\Lambda }=Q_𝒟;$$
(41)
averaged Hamiltonian constraint:
$$\left(\frac{\dot{a}_𝒟}{a_𝒟}\right)^2\frac{8\pi G}{3}\frac{M_𝒟}{V_𝐢a_𝒟^3}+\frac{_𝒟}{6}\frac{\mathrm{\Lambda }}{3}=\frac{Q_𝒟}{6},$$
(42)
where the mass $`M_𝒟`$, the averaged spatial Ricci scalar $`_𝒟`$ and the “backreaction term” $`Q_𝒟`$ are domain–dependent and, except the mass, time–dependent functions. The backreaction source term is given by
$$Q_𝒟:=2\mathrm{𝐈𝐈}_𝒟\frac{2}{3}𝐈_𝒟^2=\frac{2}{3}\left(\theta \theta _𝒟\right)^2_𝒟2\sigma ^2_𝒟.$$
(43)
Here, $`𝐈`$ and $`\mathrm{𝐈𝐈}`$ denote the invariants of the extrinsic curvature tensor that correspond to the kinematical invariants we employed earlier. The same expression (except for the vorticity) as in Eq. (9) follows by introducing the split of the extrinsic curvature into the kinematical variables shear and expansion (second equality above).
We appreciate an intimate correspondence of the GR equations with their Newtonian counterparts (Eq. (8) and Eq. (10)). The first equation is formally identical to the Newtonian one, while the second delivers an additional relation between the averaged curvature and the “backreaction term” that has no Newtonian analogue. This implies an important difference that becomes manifest by looking at the time–derivative of Eq. (42). The integrability condition that this time–derivative agrees with Eq. (41) is non–trivial in the GR context and reads:
$$_tQ_𝒟+6\frac{\dot{a}_𝒟}{a_𝒟}Q_𝒟+_t_𝒟+2\frac{\dot{a}_𝒟}{a_𝒟}_𝒟=0.$$
(44)
The correspondence between the Newtonian $`k_𝒟`$–parameter and the averaged spatial Ricci curvature is more involved in the presence of a “backreaction term”:
$$\frac{k_𝒟}{a_𝒟^2}\frac{1}{3}a_𝒟^2_{t_i}^t𝑑t^{}Q_𝒟\frac{d}{dt^{}}a_𝒟^2(t^{})=\frac{1}{6}\left(_𝒟+Q_𝒟\right).$$
(45)
The time–derivative of Eq. (45) is equivalent to the integrability condition Eq. (44). Eq. (44) shows that averaged curvature and “backreaction term” are directly coupled unlike in the Newtonian case, where the domain–dependent $`k_𝒟`$–parameter is fixed by the initial conditions.
For dust models in general relativity we may therefore introduce dimensionless average characteristics that are slightly different from those in Newtonian cosmology, the difference being that the “backreaction parameter” is now directly expressible in terms of the “backreaction source term”, and not just as an integral expression:
$$\mathrm{\Omega }_m^𝒟:=\frac{8\pi GM_𝒟}{3V_{𝒟_𝐢}a_𝒟^3H_𝒟^2};\mathrm{\Omega }_\mathrm{\Lambda }^𝒟:=\frac{\mathrm{\Lambda }}{3H_𝒟^2};\mathrm{\Omega }_k^𝒟:=\frac{_𝒟}{6H_𝒟^2};\mathrm{\Omega }_Q^𝒟:=\frac{𝒬_𝒟}{6H_𝒟^2},$$
(46)
which also obey Eq. (13).
The evolution of these parameters is intimately related, unlike the situation in the standard model, as may be illustrated by the following equation :
$$\dot{\mathrm{\Omega }}_Q^𝒟+6H_𝒟\mathrm{\Omega }_Q^𝒟(1\mathrm{\Omega }_k^𝒟\mathrm{\Omega }_Q^𝒟)+\dot{\mathrm{\Omega }}_k^𝒟+2H_𝒟\mathrm{\Omega }_k^𝒟(1\mathrm{\Omega }_k^𝒟\mathrm{\Omega }_Q^𝒟)3H_𝒟(1\mathrm{\Omega }_\mathrm{\Lambda }^𝒟\mathrm{\Omega }_k^𝒟\mathrm{\Omega }_Q^𝒟)(\mathrm{\Omega }_k^𝒟+\mathrm{\Omega }_Q^𝒟)=\mathrm{\hspace{0.33em}0}.$$
(47)
It is, e.g., a good excercise to show with the help of this equation that an inhomogeneous model (including “backreaction”), in which the matter parameter stays $`1`$ like in the standard Einstein–de Sitter cosmos, does not exist.
### 2.5 Liberation from Strict Meter
Let us now hold in for a moment and sort out what the Newtonian and the relativistic expansion laws for dust matter distinguish. Their close correspondence bears the temptation of overlooking a crucial conceptual challenge for the modeling of inhomogeneities in general relativity.
Given the Newtonian average model, Eq. (8), and its quantitative consequences (elaborated in detail in ), we can draw the conclusion that we only have to consider spatial scales that are large enough to have a negligible influence from the “backreaction term”. This term, which brought the higher voltage of having mastered a generic inhomogeneous Newtonian cosmology, shows no global relevance, and it seems that we are drawn back to the previous state of low visibility of the standard cosmological models.
The key–reason for this outcome is the validity of the Construction Principle. We force a globally vanishing “backreaction” by the way we construct inhomogeneous evolution models. (Remember that the majority of cosmological N–body simulations and analytical approximations are Newtonian and assume periodic boundary conditions for inhomogeneities on a FLRW background.) The apparent global irrelevance of “backreaction” should not be a surprise, since our evaluations have taken place within a periodic cube into which the universe model was forced: the scale of this cube introduces a “strict meter” governed by the scale–factor of a standard FLRW universe, the action of fluctuations of inhomogeneities is confined to the interior of this cube.
The relativistic average model not only has a somewhat richer tone by linking the averaged spatial Ricci curvature to the “backreaction term” as a result of the Hamiltonian constraint, it also places an equal stress on both of them in the following sense. Suppose we let the self–gravitating fluid evolve freely and link its structure to the curvature of space sections through Einstein’s equations. Let now the inhomogeneities evolve out of an almost homogeneous and isotropic state on almost flat space sections. Naturally, the inhomogeneities grow in the coarse of structure formation. The average Ricci curvature of the space sections, however, is dictated by the fluid structure unlike in the standard model which would suggest an evolution inversely proportional to the “strict meter” $`a(t)`$. In contrast, the value of the averaged curvature is an outcome rather than a model assumption. Thus, a genuine property of the relativistic model, also globally, is a change of the average curvature that does not follow a predesigned global law; a priori it is not constrained on some large scale, because the link to the “backreaction term” makes sure that, at any time, the “interior” of a universe model (below this scale) also determines the global curvature. In other words: there is no analogue of the Construction Principle. This comment is not entirely correct, since further study may reveal that it could be possible to formulate an analoguous principle on the basis of topological constraints that may be imposed on the space sections. In any case, research should not lack commitment to the challenge of constructing inhomogeneous evolution models on curved space sections, since this is evidently the generic case. We note that existing work on general relativistic evolution models and their averages relies in most cases on assumptions that are closely designed after the Newtonian case in order to rescue standard procedures like periodic boundary conditions, decomposition of inhomogeneities into plane waves and set–ups of initial conditions on locally flat space sections (compare and comments in and ).
### 2.6 What Einstein Wanted
In the previous subsection we advocated a viewpoint that concentrates on the physics of fluids, which (actively) determines the geometry of spacetime, in particular its average properties. As a showcase we learned in the Newtonian framework that the morphological properties of spatial domains are determined by the averages over fluctuations in the kinematical variables (Subsection 1.12). A similar view may also apply to the relativistic context and, most interestingly, also to the global morphology of space sections.
Turning this around, we may alternatively take the (passive) viewpoint of understanding the global world structure as given in terms of geometrical (and topological) conjectures (which may themselves be subjected to observational falsification; an example is the possibility of falsifying possible topological spaceforms on the basis of microwave background observations (; see also and ref. therein).
Einstein’s vision of a globally static universe model stands out as a famous example. His “prejudice” that the Universe as a whole should be static and all evolution (assigning sense to the notion of time) should take place in the “interior” of this world model entails a strong determination of the model’s average properties. Only his invention of the cosmological term made this vision possible resulting in the beautiful structure of a spherical space in which, apart from topological degeneracies, many characteristics (like the radius of the universe model) are practically fixed. Hence, according to what we have learned about average models, the fluid’s fluctuations are “slaved” to the global predetermined structure. This global model even guarantees strong evolution, because fluctuations evolve exponentially in a static cosmos.
The problem that his model is unstable (within the class of FLRW models) applies also to the other expanding candidates, if the class is widened to include inhomogeneities. Einstein’s example so serves as a guide to think about the average model we have discussed in a geometrical way. The assumption of a static model (the general expansion law allows for a static cosmos even without the cosmological term) should be taken as an example for the possible slaving of fluctuations to global structural assumptions. Of course, other such assumptions are possible. This illustrates how global geometrical and topological constraints on spaceforms could provide “boundary conditions” for the evolution of fluctuations. Since, in Newtonian cosmology, we are always working on a toroidal space without further questioning this assumption, this point of view is not exotic, but plants the seed for possibly fruitful research directions. That these directions are mathematically very involved can be made obvious by pointing out the relation between the average curvature and its compatibility with topological spaceforms (a spaceform with globally negative spatial curvature is not compatible with simply–connected space sections). Due to the process of structure formation the boundary of the domain may break, or it may self–intersect and split into two domains (these events are topologically classified in the framework of Legendrian singularities, and ref. therein). Since the spatial metric is linked to the fluid, a singular dynamics could induce topology changes. Contrary to the existence of such natural metamorphoses, the averager we have defined relies on the assumption that the domain remains simply–connected (Fig. 3).
### 2.7 Expansion Law of General Relativity: Perfect Fluid Models
As an outlook I briefly give one of the results obtained by averaging a perfect fluid cosmology on hypersurfaces $`S(t)=const.`$ as explained above.
The averaged equations can be summarized in the following way :
Let us define effective densities as sources of an expansion law,
$$\epsilon _{\mathrm{eff}}:=\stackrel{~}{\epsilon }_𝒟\frac{\stackrel{~}{𝒬}_𝒟}{16\pi G},p_{\mathrm{eff}}:=\stackrel{~}{p}_𝒟\frac{\stackrel{~}{𝒬}_𝒟}{16\pi G}\frac{\stackrel{~}{𝒫}_𝒟}{12\pi G},$$
(48)
with the scaled matter sources $`\stackrel{~}{\epsilon }:=N^2\epsilon `$ and $`\stackrel{~}{p}:=N^2p`$. Then, the averaged equations can be cast into a form similar to the standard Friedmann equations:
$$3\frac{_t^2a_𝒟}{a_𝒟}+4\pi G\left(\epsilon _{\mathrm{eff}}+3p_{\mathrm{eff}}\right)=\mathrm{\hspace{0.33em}0};$$
(49)
$$6\stackrel{~}{H}_𝒟^2+\stackrel{~}{}_𝒟16\pi G\epsilon _{\mathrm{eff}}=\mathrm{\hspace{0.33em}0},$$
(50)
and the integrability condition of Eq. (49) to yield Eq. (50) has the form of a balance equation between the effective sources and the averaged spatial (t–)Ricci scalar $`\stackrel{~}{}:=N^2`$:
$$_t\epsilon _{\mathrm{eff}}+3\stackrel{~}{H}_𝒟\left(\epsilon _{\mathrm{eff}}+p_{\mathrm{eff}}\right)=\frac{1}{16\pi G}\left(_t\stackrel{~}{}_𝒟+2\stackrel{~}{H}_𝒟\stackrel{~}{}_𝒟\right).$$
(51)
The effective densities obey a conservation law, if the domains’ curvature evolves like in a “small” FLRW cosmology, $`\stackrel{~}{}_𝒟=0`$, or $`\stackrel{~}{}_𝒟a_𝒟^2`$, respectively.
The expressions $`\stackrel{~}{𝒬}_𝒟`$ and $`\stackrel{~}{𝒫}_𝒟`$ are given in . As can be seen above, $`\stackrel{~}{𝒬}_𝒟`$ (the “kinematical backreaction”) acts like a fluid component with “stiff” equation of state, similar to the action of a minimially coupled scalar field, while $`\stackrel{~}{𝒫}_𝒟`$ (the “dynamical backreaction”) is due to the non–vanishing pressure gradient in the hypersurfaces.
It is interesting that the “kinematical backreaction” of the inhomogeneities could play the role of a free scalar field component that is commonly introduced when early evolutionary stages of the Universe are studied. It is also interesting that the expansion law in the case of a minimally coupled scalar field source gets particularly simple. This case will be investigated in a forthcoming work .
## Acknowledgements
I would like to thank the organizers of this workshop for their invitation and support during a pleasant and inspiring stay at Hiroshima University. I also appreciate generous support and hospitality by the National Astronomical Observatory in Tokyo, as well as hospitality at Tohoku University in Sendai. Also, I would like to thank my collaborators with whome I share some of the reported results. |
warning/0001/hep-th0001053.html | ar5iv | text | # 1 Introduction
## 1 Introduction
In this paper we study the spectrum of critical bosonic string theory on $`AdS_3\times `$ with NS-NS backgrounds, where $``$ is a compact space. Understanding string theory on $`AdS_3`$ is interesting from the point of view of the $`AdS`$/CFT correspondence since it enables us to study the correspondence beyond the gravity approximation. Another motivation is to understand string theory on a curved space-time, where the timelike component $`g_{00}`$ of the metric is non-trivial.
This involves understanding the $`SL(2,R)`$ WZW model. In this paper, we always consider the case when the target space is the universal cover of the $`SL(2,R)`$ group manifold so that the timelike direction is non-compact. The states of the WZW model form representations of the current algebras $`\widehat{SL}(2,R)_L\times \widehat{SL}(2,R)_R`$. Once we know which representations of these algebras appear, we can find the physical states of a string in $`AdS_3`$ by imposing the Virasoro constraints on the representation spaces. The problem is to find the set of representations that one should consider. In WZW models for compact groups, the unitarity restricts the possible representations . Representations of $`\widehat{SL}(2,R)`$, on the other hand, are not unitary except for the trivial representation. Of course this is not a surprise; the physical requirement is that states should have non-negative norms only after we impose the Virasoro constraints. Previous work on the subject typically considered representations with $`L_0`$ bounded below and concluded that the physical spectrum does not contain negative norm states if there is the restriction $`0<j<k/2`$ on the $`SL(2,R)`$ spin $`j`$ of the representation; the spin of the $`SL(2,R)`$ is roughly the mass of the string state in $`AdS_3`$.
This restriction raises two puzzles. One is that it seems to imply an upper bound on the mass of the string states in $`AdS_3`$ so that the internal energy of the string could not be too high. For example, if the compact space $``$ has a nontrivial $`1`$-cycle, we find that there is an upper bound on the winding number on the cycle. This restriction, which is independent of the string coupling, looks very arbitrary and raises doubts about the consistency of the theory. The second puzzle is that, on physical grounds, we expect that the theory contains states corresponding to the long strings of . These are finite energy states where we have a long string stretched close to the boundary of $`AdS_3`$. These states are not found in any representation with $`L_0`$ bounded below. In this paper, we propose that the Hilbert space of the WZW model includes a new type of representations, and we show that this proposal resolves both the puzzles. In these new representations, $`L_0`$ is not bounded below. They are obtained by acting on the standard representations by elements of the loop group that are not continuously connected to the identity, through an operation called spectral flow. These representations in the $`SL(2,R)`$ WZW model have also been considered, with some minor variations, in . The authors of these papers were motivated by finding a modular invariant partition function. They were, however, considering the case when the target space is $`SL(2,R)`$ group manifold and not its universal cover.
Throughout this paper, we consider $`AdS_3`$ in global coordinates, which do not have a coordinate horizon. In these coordinates, the unitarity issue becomes clearer since strings cannot fall behind any horizon. The interested reader could refer to for studies involving $`AdS_3`$ in Poincare coordinates. From the point of view of the $`AdS`$/CFT correspondence, it is the spectrum of strings on $`AdS_3`$ in the global coordinates that determines the spectrum of conformal dimensions of operators in the boundary CFT, though in principle the same information could be extracted from the theory in Poincare coordinates.
In order to completely settle the question of consistency of the $`SL(2,R)`$ WZW model, one needs to show that the OPE of two elements of the set of representations that we consider contains only elements of this set. We plan to discuss this issue in our future publication.
The organization of this paper is as follows. In section 2, we study classical solutions of the $`SL(2,R)`$ WZW model and we show that the model has a spectral flow symmetry which relates various solutions. In section 3, we do a semi-classical analysis and have the first glimpse of what happens when we raises the internal excitation of the string beyond the upper bound implied by the restriction $`j<k/2`$. In section 4, we study the full quantum problem and we propose a set of representations that gives a spectrum for the model with the correct semi-classical limits. In section 5, we briefly discuss scattering amplitudes involving the long strings. We conclude the paper with a summary of our results in section 6. In appendix A, we extend the proof of the no-ghost theorem for the representations we introduced in section 4. In appendix B, we study the one-loop partition function in $`AdS_3`$ with the Lorentzian signature metric and show how the sum over spectral flow reproduces the result after taking an Euclidean signature metric, up to contact terms in the modular parameters of the worldsheet.
## 2 Classical solutions
We start by choosing a parameterization of the $`SL(2,R)`$ group element as
| $`g`$ | $`=e^{iu\sigma _2}e^{\rho \sigma _3}e^{iv\sigma _2}`$ |
| --- | --- |
| | $`=\left(\begin{array}{cc}\mathrm{cos}t\mathrm{cosh}\rho +\mathrm{cos}\varphi \mathrm{sinh}\rho \hfill & \mathrm{sin}t\mathrm{cosh}\rho \mathrm{sin}\varphi \mathrm{sinh}\rho \hfill \\ \mathrm{sin}t\mathrm{cosh}\rho \mathrm{sin}\varphi \mathrm{sinh}\rho \hfill & \mathrm{cos}t\mathrm{cosh}\rho \mathrm{cos}\varphi \mathrm{sinh}\rho \hfill \end{array}\right).`$ |
(1)
Here $`\sigma ^i`$ ($`i=1,2,3`$) are the Pauli matrices<sup>2</sup><sup>2</sup>2$`\sigma _1=\left(\begin{array}{cc}0\hfill & 1\hfill \\ 1\hfill & 0\hfill \end{array}\right)`$, $`\sigma _2=\left(\begin{array}{cc}0\hfill & i\hfill \\ i\hfill & 0\hfill \end{array}\right)`$ and $`\sigma _3=\left(\begin{array}{cc}1\hfill & 0\hfill \\ 0\hfill & 1\hfill \end{array}\right)`$. , and we set
$$u=\frac{1}{2}(t+\varphi ),v=\frac{1}{2}(t\varphi ).$$
(2)
Another useful parameterization of $`g`$ is
$$g=\left(\begin{array}{cc}X_1+X_1\hfill & X_0X_2\hfill \\ X_0X_2\hfill & X_1X_1\hfill \end{array}\right),$$
(3)
with
$$X_1^2+X_0^2X_1^2X_2^2=1.$$
(4)
This parameterization shows that the $`SL(2,R)`$ group manifold is a $`3`$-dimensional hyperboloid. The metric on $`AdS_3`$,
$$ds^2=dX_1^2dX_0^2+dX_1^2+dX_2^2,$$
is expressed in the global coordinates $`(t,\varphi ,\rho )`$ as
$$ds^2=\mathrm{cosh}^2\rho dt^2+d\rho ^2+\mathrm{sinh}^2\rho d\varphi ^2.$$
(5)
We will always work on the universal cover of the hyperboloid (4), and $`t`$ is non-compact.
Our theory has the WZW action
$$S=\frac{k}{8\pi \alpha ^{}}d^2\sigma \mathrm{Tr}\left(g^1gg^1g\right)+k\mathrm{\Gamma }_{WZ}$$
(6)
The level $`k`$ is not quantized since $`H^3`$ vanishes for $`SL(2,R)`$. The semi-classical limit corresponds to large $`k`$. We define the right and left moving coordinates on the worldsheet as,
$$x^\pm =\tau \pm \sigma ,$$
(7)
where $`\sigma `$ is periodic with the period $`2\pi `$. This action has a set of conserved right and left moving currents
$$J_R^a(x^+)=k\mathrm{Tr}\left(T^a_+gg^1\right),J_L^a(x^{})=k\mathrm{Tr}\left(T^ag^1_{}g\right)$$
(8)
where $`T^a`$ are a basis for the $`SL(2,R)`$ Lie algebra. It is convenient to take them as
$$T^3=\frac{i}{2}\sigma ^2,T^\pm =\frac{1}{2}(\sigma ^3\pm i\sigma _1).$$
In terms of our parameterization, the currents are expressed as
| $`J_R^3=`$ | $`k(_+u+\mathrm{cosh}2\rho _+v)`$ |
| --- | --- |
| $`J_R^\pm =`$ | $`k(_+\rho \pm i\mathrm{sinh}2\rho _+v)e^{i2u},`$ |
(9)
and
| $`J_L^3=`$ | $`k(_{}v+\mathrm{cosh}2\rho _{}u)`$ |
| --- | --- |
| $`J_L^\pm =`$ | $`k(_{}\rho \pm i\mathrm{sinh}2\rho _{}u)e^{i2v}.`$ |
(10)
The zero modes of $`J_{R,L}^3`$ are related to the energy $`E`$ and angular momentum $`\mathrm{}`$ in $`AdS_3`$ as
| $`J_0^3={\displaystyle _0^{2\pi }}{\displaystyle \frac{dx^+}{2\pi }}J_R^3={\displaystyle \frac{1}{2}}(E+\mathrm{})`$ |
| --- |
| $`\overline{J}_0^3={\displaystyle _0^{2\pi }}{\displaystyle \frac{dx^{}}{2\pi }}J_L^3={\displaystyle \frac{1}{2}}(E\mathrm{}).`$ |
(11)
The second Casimir of $`SL(2,R)`$ is
$$c_2=J^aJ^a=\frac{1}{2}\left(J^+J^{}+J^{}J^+\right)(J^3)^2.$$
(12)
The equations of motion derived from (6) is $`_{}(_+gg^1)=0`$, namely that the currents, $`J_R`$ and $`J_L`$, are purely right or left moving as indicated. A general solution of the equations of motion for $`SL(2,R)`$ is the product of two group elements each of which depends only on $`x^+`$ or $`x^{}`$ as
$$g=g_+(x^+)g_{}(x^{}).$$
(13)
Comparing (13) with (1) we can find the embedding of the worldsheet in $`AdS_3`$. The requirement that the string is closed under $`\sigma \sigma +2\pi `$ imposes the constraint,
$$g_+(x^++2\pi )=g_+(x^+)M,g_{}(x^{}2\pi )=M^1g_{}(x^{}),$$
(14)
with the same $`MSL(2,R)`$ for both $`g_+`$ and $`g_{}`$. The monodromy matrix $`M`$ is only defined up to a conjugation by $`SL(2,R)`$, and classical solutions of the WZW model are classified according to the conjugacy class of $`M`$.
For strings on $`AdS_3\times `$, we should impose the Virasoro constraints
$$T_{++}^{total}=T_{++}^{AdS}+T_{++}^{other}=0$$
(15)
and similarly $`T_{}^{total}=0`$, where
$$T_{++}^{AdS}=\frac{1}{k}J_R^aJ_R^a$$
is the energy-momentum tensor for the $`AdS_3`$ part<sup>3</sup><sup>3</sup>3In the quantum theory, we will have the same expression but with $`kk2`$. and $`T_{++}^{other}`$ represents the energy-momentum tensor for the sigma-model on $``$.
Let us analyze some simple classical solutions.
### 2.1 Geodesics in AdS<sub>3</sub>
Consider a solution
$$g_+=Ue^{iv_+(x^+)\sigma _2},g_{}=e^{iu_{}(x^{})\sigma _2}V,$$
(16)
where $`U`$ and $`V`$ are constant elements of $`SL(2,R)`$. The energy momentum tensor of this solution is
$$T_{++}^{AdS}=k(_+v_+)^2,T_{}^{AdS}=k(_{}u_{})^2.$$
(17)
Suppose we have some string excitation in the compact part $``$ of $`AdS_3\times `$, and set $`T_{\pm \pm }^{other}=h`$ for some constant $`h>0`$. We may regard $`h`$ as a conformal weight of the sigma-model on $``$. The Virasoro constraints $`T_{\pm \pm }^{total}=0`$ implies
$$(_+v_+)^2=(_{}u_{})^2=\frac{h}{k}.$$
Thus we can set $`v_+=\alpha x^+/2`$ and $`u_{}=\alpha x^{}/2`$ where $`\alpha =\pm \sqrt{4h/k}`$. Substituting this in (13), we obtain
$$g=U\left(\begin{array}{cc}\mathrm{cos}(\alpha \tau )\hfill & \mathrm{sin}(\alpha \tau )\hfill \\ \mathrm{sin}(\alpha \tau )\hfill & \mathrm{cos}(\alpha \tau )\hfill \end{array}\right)V.$$
(18)
Since the solution depends only on $`\tau `$ and not on $`\sigma `$, we interpret that the string is collapsed to a point which flows along the trajectory in $`AdS_3`$ parameterized by $`\tau `$. See Figure 1. If $`U=V=1`$, the solution (18) represents a particle sitting at the center of $`AdS_3`$,
$$t=\alpha \tau ,\rho =0.$$
(19)
A more general solution (18) is given by acting the $`SL(2,R)\times SL(2,R)`$ isometry on (19), and therefore it is a timelike geodesic<sup>4</sup><sup>4</sup>4In fact, any timelike geodesic can be expressed in the form (18). in $`AdS_3`$. For this solution, the currents are given by
$$J_R^aT^a=\frac{k}{2}\alpha UT^3U^1,$$
(20)
and similarly for $`J_L`$. The monodromy matrix $`M`$ defined by (14) is
$$M=\left(\begin{array}{cc}\mathrm{cos}(\alpha \pi )\hfill & \mathrm{sin}(\alpha \pi )\hfill \\ \mathrm{sin}(\alpha \pi )\hfill & \mathrm{cos}(\alpha \pi )\hfill \end{array}\right)$$
and belongs to the elliptic conjugacy class $`SL(2,R)`$.
A solution corresponding to a spacelike geodesic is
$$g=U\left(\begin{array}{cc}e^{\alpha \tau }\hfill & 0\hfill \\ 0\hfill & e^{\alpha \tau }\hfill \end{array}\right)V,$$
(21)
with $`U,VSL(2,R)`$. The energy-momentum tensor has a sign opposite of (17)
$$T_{\pm \pm }^{AdS}=\frac{1}{4}k\alpha ^2.$$
(22)
If we choose $`U=V=1`$, the solution is simply a straight line cutting the spacelike section $`t=0`$ of $`AdS_3`$ diagonally,
$$t=0,\rho e^{i\varphi }=\alpha \tau ,$$
(23)
See Figure 2 (A). A general solution (21) is given from this by the action of the isometry, and therefore is a spacelike geodesic. The currents for this solution are
$$J_R^aT^a=\frac{k}{2}\alpha UT^1U^1,$$
(24)
and the monodromy matrix is
$$M=\left(\begin{array}{cc}e^{\alpha \pi }\hfill & 0\hfill \\ 0\hfill & e^{\alpha \pi }\hfill \end{array}\right),$$
which belongs to the hyperbolic conjugacy class of $`SL(2,R)`$.
There is one more class of solutions whose monodromy matrices are in the parabolic conjugacy class of $`SL(2,R)`$. They correspond to null geodesics in $`AdS_3`$.
### 2.2 Spectral flow and strings with winding numbers
Given one classical solution $`g=\stackrel{~}{g}_+\stackrel{~}{g}_{}`$, we can generate new solutions by the following operation,
$$g_+=e^{i\frac{1}{2}w_Rx^+\sigma _2}\stackrel{~}{g}_+g_{}=\stackrel{~}{g}_{}e^{i\frac{1}{2}w_Lx^{}\sigma _2}.$$
(25)
Comparing this with the parameterization (1) of $`g=g_+g_{}`$, we see that this operation amounts to
| $`tt+{\displaystyle \frac{1}{2}}(w_R+w_L)\tau +{\displaystyle \frac{1}{2}}(w_Rw_L)\sigma `$ |
| --- |
| $`\varphi \varphi +{\displaystyle \frac{1}{2}}(w_R+w_L)\sigma +{\displaystyle \frac{1}{2}}(w_Rw_L)\tau .`$ |
(26)
The periodicity of the string worldsheet, under $`\sigma \sigma +2\pi `$, on the universal cover of $`SL(2,R)`$ requires<sup>5</sup><sup>5</sup>5 If the target space is the single cover of $`SL(2,R)`$, $`w_R`$ and $`w_L`$ can be different. In this case $`(w_Rw_L)`$ gives the winding number along the closed timelike curve on $`SL(2,R)`$. $`w_R=w_L=w`$ for some integer $`w`$.
One may regard (25) as an action by an element of the loop group $`\widehat{SL}(2,R)\times \widehat{SL}(2,R)`$ which is not continuously connected to the identity<sup>6</sup><sup>6</sup>6 The loop group $`\widehat{SL}(2,R)`$ has such an element since $`\pi _1(SL(2,R))=\text{}`$. Therefore, in the model whose the target space is the single cover of $`SL(2,R)`$, the full symmetry group of the model is the loop group of $`SL(2,R)\times SL(2,R)`$ and its connected components are parametrized by $`\text{}\times \text{}`$. In this paper, we are studying the model for the universal cover of $`SL(2,R)`$. In this case, some of these elements do not act properly on the field space, generating worldsheets which close only modulo time translation. However the ones parametrized by the diagonal are still symmetry of the model. The diagonal parameterizes the spectral flow operation performed simultaneously for both the left and right movers.. This particular symmetry of the theory will also be useful in our analysis of the Hilbert space. Here we see that it generates a new solution from an old solution. Furthermore, the currents (9) change in the following way
$$J_R^3=\stackrel{~}{J}_R^3+\frac{k}{2}w,J_R^\pm =\stackrel{~}{J}_R^\pm e^{iwx^+}$$
(27)
and a similar expression for $`J_L^a`$. Or, in terms of the Fourier modes,
$$J_n^3=\stackrel{~}{J}_n^3+\frac{k}{2}w\delta _{n,0},J_n^\pm =\stackrel{~}{J}_{nw}^\pm .$$
(28)
This means that the stress tensor will change to
$$T_{++}^{AdS}=\stackrel{~}{T}_{++}^{AdS}w\stackrel{~}{J}^3\frac{k}{4}w^2.$$
(29)
In the CFT literature, this operation is known as the spectral flow.
Let us study what happens if we act with this symmetry on the solutions corresponding to geodesics, (18) and (21). These solutions depend only on the worldsheet time coordinate $`\tau `$, and the spectral flow (26) with $`w=w_R=w_L`$ introduces $`\sigma `$ dependence as
| $`t`$ | $`=t_0(\tau )+w\tau `$ |
| --- | --- |
| $`\rho `$ | $`=\rho _0(\tau )`$ |
| $`\varphi `$ | $`=\varphi _0(\tau )+w\sigma .`$ |
(30)
Here $`(t_0,\rho _0,\varphi _0)`$ represents the original geodesic solution. So what the spectral flow does is to stretch the geodesic solution in the $`t`$-direction (by adding $`w\tau `$) and rotates it around $`w`$-times around the center $`\rho =0`$ of $`AdS_3`$ (by adding $`w\sigma `$). It is clear that the resulting solution describes a circular string, winding $`w`$-times around the center of $`AdS_3`$. Since the spectral flow changes the energy-momentum tensor, we need to impose the physical state condition $`T_{\pm \pm }^{AdS}+T_{\pm \pm }^{other}=0`$ with respect to the new energy-momentum tensor (29).
### 2.3 Short strings as spectral flow of timelike geodesics
A timelike geodesic in $`AdS_3`$ makes a periodic trajectory as shown in Figure 1, approaching the boundary of $`AdS_3`$, then coming back to the center and so on. In particular, when $`V=U^1`$ in (18), the geodesic periodically passes through the center $`\rho =0`$ of $`AdS_3`$, with the period $`2\pi `$ in the $`t`$-coordinate. The spectral flow,
$$tt+w\tau ,\varphi \varphi +w\sigma $$
stretches the geodesic in the time direction and rotate it around the center $`\rho =0`$; it is pictorially clear that the resulting solution describes a circular string which repeats expansion and contraction. This is shown shown in Figure 3 in the case of $`w=1`$. Assuming $`T_{\pm \pm }^{other}=h`$ as in the case of geodesics, the Virasoro constraint for the solution is
$$T_{++}^{total}=\stackrel{~}{T}_{++}^{AdS}w\stackrel{~}{J}^3\frac{k}{4}w^2+T_{++}^{other}=0.$$
(31)
Since
$$\stackrel{~}{T}_{++}^{AdS}=\frac{k}{4}\alpha ^2$$
for the timelike geodesic, we find
$$J_0^3=\stackrel{~}{J}_0^3+\frac{k}{2}w=\frac{k}{4}w+\frac{1}{w}\left(\frac{k}{4}\alpha ^2+h\right).$$
(32)
The spacetime energy $`E`$ of the string is given by $`E=2J_0^3`$, and is bounded above as
$$E=\frac{k}{2}w+\frac{1}{w}\left(k\alpha ^2+2h\right)<\frac{k}{2}w+\frac{2h}{w}.$$
(33)
It is not difficult to find an explicit form of the solution. When $`V=U^1`$ in (18), without loss of generality, we can set<sup>7</sup><sup>7</sup>7A different choice of $`U=V^1`$ simply results in shift of $`\varphi `$ in the solution. $`U=V^1=e^{\frac{1}{2}\rho _0\sigma _3}`$. The solution<sup>8</sup><sup>8</sup>8We have been informed that a similar classical solution has also been studied in . obtained by the spectral flow of (18) is then
| $`e^{i\varphi }\mathrm{sinh}\rho `$ | $`=ie^{iw\sigma }\mathrm{sinh}\rho _0\mathrm{sin}\alpha \tau `$ |
| --- | --- |
| $`\mathrm{tan}t`$ | $`={\displaystyle \frac{\mathrm{tan}w\tau +\mathrm{tan}\alpha \tau /\mathrm{cosh}\rho _0}{1\mathrm{tan}w\tau \mathrm{tan}\alpha \tau /\mathrm{cosh}\rho _0}}.`$ |
(34)
The currents of this solution are
| $`J_R^3`$ | $`={\displaystyle \frac{k}{2}}(\alpha \mathrm{cosh}\rho _0+w)`$ |
| --- | --- |
| $`J_R^\pm `$ | $`=\pm i{\displaystyle \frac{k}{2}}\alpha \mathrm{sinh}\rho _0e^{iwx^+},`$ |
(35)
and similarly for $`J_L`$. Comparing this with (32), we find
$$\alpha =\alpha _\pm =w\mathrm{cosh}\rho _0\pm \sqrt{w^2\mathrm{sinh}^2\rho _0+\frac{4h}{k}}.$$
(36)
If we choose the branch $`\alpha =\alpha _+`$, the spacetime energy $`E`$ of the solution is positive and is given by
$$E=2J_0^3=2\overline{J}_0^3=k\left(\mathrm{cosh}\rho _0\sqrt{\frac{4h}{k}+w^2\mathrm{sinh}^2\rho _0}w\mathrm{sinh}^2\rho _0\right).$$
(37)
There are several interesting features of this formula for the energy $`E`$. Except for the case of $`h=kw^2/4`$, the energy is a monotonically increasing function of $`\rho _0`$, which approaches $`Ekw/2+2h/w`$ as $`\rho _0\mathrm{}`$. One may view that the solution describe a bound state trapped inside of $`AdS_3`$. At the exceptional value of $`h=kw^2/4`$, we have $`\alpha _+=0`$ and the energy of the solution becomes $`E=kw`$, completely independent of the size $`\rho _0`$ of the string. The solution in this case is
$$\rho =\rho _0,t=w\tau ,\varphi =w\sigma ,$$
(38)
and represents a string staying at the fixed radius $`\rho =\rho _0`$, neither contracting nor expanding. The fact that we have such a solution at any radius $`\rho _0`$ means that the string becomes marginally unstable in $`AdS_3`$.
Now let us turn to the case when $`UV^1`$, or to be more precise, when $`UV`$ does not commute with $`T^3=\frac{i}{2}\sigma ^2`$. (When $`UV`$ commutes with $`T^3`$, one can shift the value of $`\tau `$ to set $`U=V^1`$.) In this case, the geodesic does not necessarily pass through the center of $`AdS_3`$. Therefore the circular string obtained by its spectral flow does not collapse to a point. Since
$$\stackrel{~}{J^a}_LT_{}^{}{}_{}{}^{a}=\frac{k}{2}\alpha UT^3U^1,\stackrel{~}{J^a}_RT^a=\frac{k}{2}\alpha V^1T^3V,$$
(39)
$`\stackrel{~}{J}_L^3\stackrel{~}{J}_R^3`$ unless $`UV`$ commutes with $`T^3=\frac{i}{2}\sigma ^2`$, and the spacetime angular momentum $`\mathrm{}=J_R^3J_L^3=\stackrel{~}{J}_R^3\stackrel{~}{J}_L^3`$ is nonzero. Thus one may view that the circular string is kept from completely collapsing by the centrifugal force. Since $`T_{++}^{AdS}T_{}^{AdS}=w(\stackrel{~}{J}_R^3\stackrel{~}{J}_L^3),`$ the Virasoro constraint $`T_{\pm \pm }^{total}=0`$ requires that the left and right conformal weights $`(h_L,h_R)`$ of the internal part should be different and that $`h_Rh_L=w\mathrm{}`$.
### 2.4 Long strings as spectral flow of spacelike geodesics
We have seen in (33) that the spacetime energy $`E`$ of the solution given by the spectral flow of the timelike geodesic is bounded above as $`E<kw/2+2h/w`$. What will happen if we raise the energy above this value? To understand this, let us look at the spectral flow of the spacelike geodesic. Since $`\stackrel{~}{T}_{++}^{AdS}=+k\alpha ^2/2`$ for the spacelike geodesic, the Virasoro constraint (31) gives
$$J_0^3=\stackrel{~}{J}_0^3+\frac{k}{2}w=\frac{k}{4}w+\frac{1}{w}\left(\frac{k}{2}\alpha ^2+h\right),$$
(40)
and the spacetime energy is now bounded below,
$$E=2J_0^3>\frac{k}{2}w+\frac{2h}{w}.$$
(41)
As an example, let us consider the straight line cutting the spacelike section $`t=0`$ diagonally (23). The spectral flow with $`w`$ of this solution is
$$t=w\tau ,\rho e^{i\varphi }=\alpha \tau e^{iw\sigma },$$
(42)
namely
$$\rho =\frac{\alpha }{w}|t|.$$
(43)
The solution starts in the infinite past $`t=\mathrm{}`$ as a circular string of an infinite radius located at the boundary of $`AdS_3`$. The string then collapse, shrinks to a point at $`t=0`$, and expand away toward the boundary of $`AdS_3`$ as $`t+\mathrm{}`$. More generally, if we choose $`U=V^1=e^{\frac{1}{2}\rho _0\sigma _1}`$, the spectral flow of the geodesic (21) gives
| $`e^{i\varphi }\mathrm{sinh}\rho `$ | $`=e^{iw\sigma }\mathrm{cosh}\rho _0\mathrm{sinh}\alpha \tau `$ |
| --- | --- |
| $`\mathrm{tan}t`$ | $`={\displaystyle \frac{\mathrm{tan}w\tau +\mathrm{tanh}\alpha \tau \mathrm{sinh}\rho _0}{1\mathrm{tan}w\tau \mathrm{tanh}\alpha \tau \mathrm{sinh}\rho _0}}.`$ |
(44)
This solution, which we call a long string, is depicted in Figure 4.
The Virasoro constraint $`T_{++}^{total}=0`$ for the long string (44) is
$$T_{++}^{AdS}+T_{++}^{other}=\frac{k}{4}\left(\alpha ^22\alpha w\mathrm{sinh}\rho _0w^2\right)+h=0,$$
(45)
with the solutions
$$\alpha =\alpha _\pm =w\mathrm{sinh}\rho _0\sqrt{w^2\mathrm{cosh}^2\rho _0\frac{4h}{k}}.$$
(46)
The spacetime energy $`E`$ of these solutions are
$$E=2J_0^2=2\overline{J}_0^3=k\left(w\mathrm{cosh}^2\rho _0\mathrm{sinh}\rho _0\sqrt{w^2\mathrm{cosh}^2\rho _0\frac{4h}{k}}\right).$$
(47)
At the critical value $`h=kw^2/4`$, we have $`\alpha _+=0`$ and the energy for this solution becomes $`E=kw`$. At this point, the long string solution (44) coincides with (38). Thus we see that, as we increase the value of $`h`$ to $`h=kw^2/4`$, the short string solution (34) can turn into the long string solution (44) and escape to infinity.
As explained in , a string that winds in $`AdS_3`$ close to the boundary has finite energy because there is a balance between two large forces. One is the string tension that wants to make the string contract and the other is the NS-NS $`B`$ field which wants to make the string expand. These forces cancel almost precisely near the boundary and only a finite energy piece is left. The threshold energy for the long string computed in is $`kw/4`$, in agreement with (41) when $`h=0`$. These strings can have some momentum in the radial direction and that is a degree of freedom $`\alpha `$ that we saw explicitly above. One may view the long string as a scattering state, while the previous solution (34) is like a bound state trapped inside of $`AdS_3`$.
In general, if $`UV`$ commutes with $`T^3=\frac{i}{2}\sigma ^2`$, the long string collapses to a point once in its lifetime. If $`UV`$ does not commute with $`T^3`$, the angular momentum $`\mathrm{}=J_R^3J_L^3`$ of the solution does not vanish and the centrifugal force keeps the string from collapsing completely. In this case, the Virasoro constraint $`T_{\pm \pm }^{total}=0`$ requires $`h_Rh_L=w\mathrm{}`$ for the conformal weights of the internal sector.
For the long strings, one can define a notion of the S-matrix. In the infinite past, the size of the long string is infinite but its energy is finite. Therefore the interactions between them are expected to be negligible, and one can define asymptotic states consisting of long strings. The strings then approach the center of $`AdS_3`$ and are scattered back to the boundary. In this process, the winding number could in principle change.
## 3 Semi-classical analysis
In studying the classical solutions, we were naively identifying the winding number $`w`$ as associated to the cycle $`\varphi \varphi +2\pi `$. But since this cycle is contractible in $`AdS_3`$, we should be careful about what we mean by the integer $`w`$. The winding number is well-defined when the string is close to the boundary, so we expect that long strings close to the boundary have definite winding numbers. On the other hand, when the string collapses to a point, as shown in Figures 3 and 4, the winding number is not well-defined. Therefore, if we quantize the string, it is possible to have a process in which the winding number changes. There is however a sense in which string states are characterized by some integer $`w`$.
In order to clarify the meaning of $`w`$ when the string can collapse, let us look at the Nambu action
$$S=𝑑t\frac{d\sigma }{2\pi }\left[\sqrt{detg_{ind}}B_{t\varphi }_\sigma \varphi \right]$$
(48)
where $`g_{ind}`$ is the induced metric on the worldsheet, and $`B_{t\varphi }`$ is the NS-NS $`B`$-field. We have chosen the static gauge in the time direction $`t=\tau `$. We assume that initially we have a state with $`\rho =0`$, and we want to analyze small perturbations. Since the coordinate $`\varphi `$ is not well-defined, it is more convenient to use
$$X^1+iX^2=\rho e^{i\varphi }.$$
(49)
Let us compute the components of the induced metric $`g_{ind}`$. To be specific, we consider the case when the target space $`AdS_3\times S^3\times T^4`$, and consider a string winding around a cycle on $`T^4`$. By expanding in the quadratic order in $`\rho `$, we find
| $`g_{ind,00}`$ | $`=k\left[(1+\rho ^2)+_0X^a_0X^a\right]+_0Y^i_0Y^i`$ |
| --- | --- |
| $`g_{ind,01}`$ | $`=k_0X^a_1X^a+_0Y^i_1Y^i`$ |
| $`g_{ind,11}`$ | $`=k_1X^a_1X^a+_1Y^i_1Y^i,(a=1,2)`$ |
(50)
where $`Y^i`$’s are coordinates on $`T^4`$. For simplicity, we consider purely winding modes on $`T^4`$, so that only $`_1Y^i`$ is nonzero. For these states, the conformal weight $`h`$ is given by<sup>9</sup><sup>9</sup>9One factor of 2 comes from the fact that this includes left and right movers and the other from the fact that the expression for the energy involves $`1/2Y_{}^{}{}_{}{}^{2}`$.
$$4h=\frac{d\sigma }{2\pi }G_{ij}_1Y^i_1Y^j$$
(51)
Substituting (50) and (51) into the action and expanding to the quadratic order in $`\rho `$, we find
| $`S`$ | $`=\sqrt{4kh}{\displaystyle 𝑑\sigma ^2\left[1\frac{1}{2}(_0X^a)^2+\frac{1}{2}\frac{k}{4h}\left(_1X^a+ϵ_{ab}\sqrt{\frac{4h}{k}}X^b\right)^2+\mathrm{}\right]}`$ |
| --- | --- |
| | $`=\sqrt{4kh}{\displaystyle 𝑑\sigma ^2\left[1\frac{1}{2}|_0\mathrm{\Phi }|^2+\frac{1}{2}\frac{k}{4h}\left|\left(_1i\sqrt{\frac{4h}{k}}\right)\mathrm{\Phi }\right|^2+\mathrm{}\right]},`$ |
(52)
where $`\mathrm{\Phi }=X^1+iX^2`$.
The action (52) is the one for a massless charged scalar field on $`\text{}\times S^1`$ coupled to a constant gauge field $`A=\sqrt{4h/k}`$ around $`S^1`$. As we vary $`A`$, we observe the well-know phenomenon of the spectral asymmetry. Let us first assume that $`A`$ is not an integer. A general solution to the equation of motion derived from (52), requiring the periodicity in $`\sigma `$, is
$$\mathrm{\Phi }\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\left(a_n^{}e^{i(nA)(\stackrel{~}{\tau }+\sigma )}+b_ne^{i(nA)(\stackrel{~}{\tau }\sigma )}\right)\frac{e^{iA\sigma }}{nA},$$
(53)
where $`A=\sqrt{4h/k}`$ and $`\stackrel{~}{\tau }=\tau /A`$. Upon quantization, the commutation relations are given (modulo a positive constant factor) by
$$[a_n,a_m^{}]=(nA)\delta _{n,m},[b_n,b_m^{}]=(nA)\delta _{n,m}.$$
(54)
Notice that the sign in the right hand side of (54) determines whether $`a_n`$ or $`a_n^{}`$ should be regarded as the annihilation operator. Thus, assuming that the Hilbert space is positive definite, the vacuum state is defined by
| $`a_n|0=b_n|0=0,(n>A)`$ |
| --- |
| $`a_n^{}|0=b_n^{}|0=0,(n<A).`$ |
(55)
For $`\mathrm{\Phi }=\rho e^{i\varphi }`$ given by (53) and $`t=A\stackrel{~}{\tau }`$, we find
| $`J_R^+`$ | $`=k\left(e^{it}_+\mathrm{\Phi }^{}\mathrm{\Phi }^{}_+e^{it}\right)ik{\displaystyle \underset{n}{}}a_ne^{in(\stackrel{~}{\tau }+\sigma )}`$ |
| --- | --- |
| $`J_R^{}`$ | $`=k\left(e^{it}_+\mathrm{\Phi }\mathrm{\Phi }_+e^{it}\right)ik{\displaystyle \underset{n}{}}a_n^{}e^{in(\stackrel{~}{\tau }+\sigma )},`$ |
(56)
and similarly for $`J_L^\pm `$. Therefore $`J_n^+=ika_n`$ and $`J_n^{}=ika_n^{}`$. The vacuum state $`|0`$ defined by (55) then obeys
$$J_n^+|0=0(n>A),J_n^{}|0=0(n>A).$$
(57)
Thus the vacuum state $`|0`$ is not in a regular highest weight representation of the current algebra $`\widehat{SL}(2,R)`$. If we set
$$J_n^\pm =\stackrel{~}{J}_{nw}^\pm $$
(58)
with the integer $`w`$ defined by
$$w<A<w+1,$$
(59)
then $`|0`$ obeys the regular highest weight condition with respect to $`\stackrel{~}{J}_n^\pm `$,
$$\stackrel{~}{J}_n^+|0=0(n1),\stackrel{~}{J}_n^{}|0=0(n0)$$
(60)
The change of the basis (58) is nothing but the spectral flow (28) discussed earlier, so we can identify $`w`$ as the amount of spectral flow needed to transform the string state into a string state which obeys the regular conditions (60). We have found that, for a given value of $`h`$, there is a unique integer of $`w`$ associated to the string state. As we vary the conformal weight $`h`$, $`A=\sqrt{4h/k}`$ will become an integer. At that point, one of the modes of the field $`\mathrm{\Phi }`$ will have a vanishing potential. In fact we can check that classically this potential is completely flat. Giving an expectation value to that mode, we find configurations as in (38). Corresponding to various values of its momentum in the radial direction, we have a continuum of states. So, at this value of $`h`$, we do not have a normalizable ground state; instead we have a continuum of states which are $`\delta `$-function normalizable. If we continue to increase $`h`$, we find again normalizable states, but they are labeled by a new integer $`(w+1)`$. Notice that $`w`$ is not directly related to the physical winding of the string. In fact by exciting a coherent state of the oscillators $`a_n`$ or $`b_n`$ we can find string states that look like expanding and collapsing strings with winding number $`n`$ around the origin.
One of the puzzles we raised in the introduction was what happens when we increase the internal conformal weight $`h`$ of the string beyond the upper bound implied by the restriction $`j<k/2`$ on the $`SL(2,R)`$ spin $`j`$ due to the no-ghost theorem. In this section, we saw a semi-classical version of the puzzle and its resolution. When $`h`$ reaches the bound, we find that the state can become a long string with no cost in energy. Above the bound, we should consider a Fock space with a different bose sea level. In the fully quantum description of the model given below, we will find a similar situation but with minor corrections.
## 4 Quantum string in AdS<sub>3</sub>
The Hilbert space of the WZW model is a sum of products of representations of the left and the right-moving current algebras generated by
$$J_L^a=\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}J_n^ae^{inx^{}},J_R^a=\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\overline{J}_n^ae^{inx^+},$$
(61)
with $`a=3,\pm `$, obeying the commutation relations
| $`[J_n^3,J_m^3]={\displaystyle \frac{k}{2}}n\delta _{n+m,0}`$ |
| --- |
| $`[J_n^3,J_m^\pm ]=\pm J_{n+m}^\pm `$ |
| $`[J_n^+,J_m^{}]=2J_{n+m}^3+kn\delta _{n+m,0},`$ |
(62)
and the same for $`\overline{J}_n^a`$. We denote the current algebra by $`\widehat{SL}_k(2,R)`$. The Virasoro generator $`L_n`$ are defined by
| $`L_0={\displaystyle \frac{1}{k2}}\left[{\displaystyle \frac{1}{2}}(J_0^+J_0^{}+J_0^{}J_0^+)(J_0^3)^2+{\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}(J_m^+J_m^{}+J_m^{}J_m^+2J_m^3J_m^3)\right]`$ |
| --- |
| $`L_{n0}={\displaystyle \frac{1}{k2}}{\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}(J_{nm}^+J_m^{}+J_{nm}^{}J_m^+2J_{nm}^3J_m^3)`$ |
(63)
and obey the commutation relation
$$[L_n,L_m]=(nm)L_{n+m}+\frac{c}{12}(n^3n)\delta _{n+m,0},$$
(64)
where the central charge $`c`$ is given by
$$c=\frac{3k}{k2}.$$
(65)
We will find that the Hilbert space of the WZW model consists of sub-sectors parameterized by integer $`w`$, labeling the amount of spectral flow in a sense to be made precise below. We then formulate our proposal on how the complete Hilbert space of the WZW model is decomposed into representations of the current algebras and provide evidences for the proposal.
States in a representation of the current algebra are labeled by eigenvalues of $`L_0`$ and $`J_0^3`$. Since the kinetic term of the WZW model based on $`SL(2,R)`$ has an indefinite signature, it is possible that the Hilbert space of the model contains states with negative eigenvalues of $`L_0`$ as well as states with negative norms, and indeed both types of states appear as we will see below. For the moment, we will consider a representation in which eigenvalues of $`L_0`$ is bounded below. We call them positive energy representations, or unflowed representations. Since the action of $`J_n^{3,\pm }`$ with $`n1`$ on a state lowers the eigenvalue of $`L_0`$ by $`n`$, there has to be a set of states which are annihilated by them. We will call such states the primary states of the positive energy representation. All other states in the representation are obtained by acting $`J_n^{3,\pm }`$ ($`n1`$) on the primary states. The ground states make a representation of $`SL(2,R)`$ generated by $`J_0^{3,\pm }`$. So let us review irreducible representations of $`SL(2,R)`$.
### 4.1 Representations of the zero modes
We expect that physical states of a string in $`AdS_3`$ have positive norms. Since $`J_0^{3,\pm }`$ commute with the Virasoro constraints, physical spectrum of the string must be in unitary representations of $`SL(2,R)`$. Most of the mathematical references on representation theory of $`SL(2,R)`$ deal with the case with compact time<sup>10</sup><sup>10</sup>10For a review of representations of $`SL(2,R)`$, see for example .; we are however interested in the case with non-compact time. A clear analysis from the algebraic point of view is presented in , which we now summarize with some minor changes is notation.
There are the following five types of unitary representations. All the representations are parameterized by $`j`$, which is related to the second Casimir $`c_2=\frac{1}{2}(J_0^+J_0^{}+J_0^{}J_0^+)(J_0^3)^2`$ as $`c_2=j(j1)`$.
(1) Principal discrete representations (lowest weight):
A representation of this type is realized in the Hilbert space
$$𝒟_j^+=\{|j;m:m=j,j+1,j+2,\mathrm{}\},$$
where $`|j;j`$ is annihilated by $`J_0^{}`$ and $`|j;m`$ is an eigenstate of $`J_0^3`$ with $`J_0^3=m`$. The representation is unitarity if $`j`$ is real and $`j>0`$. For representations of the group $`SL(2,R)`$, $`j`$ is restricted to be a half of integer. Since we are considering the universal cover of $`SL(2,R)`$, $`j`$ can be any positive real number.
(2) Principal discrete representations (highest weight):
A charge conjugation of (1). A representation of this type is realized in the Hilbert space
$$𝒟_j^{}=\{|j;m:m=j,j1,j2,\mathrm{}\},$$
where $`|j;j`$ is annihilated by $`J_0^+`$ and $`|j;m`$ is an eigenstate of $`J_0^3`$ with $`J_0^3=m`$. The representation is unitary if $`j`$ is real and $`j>0`$.
(3) Principal continuous representations:
A representation of this type is realized in the Hilbert space of
$$𝒞_j^\alpha =\{|j,\alpha ;m:m=\alpha ,\alpha \pm 1,\alpha \pm 2,\mathrm{}\},$$
where $`|j,\alpha ;m`$ is an eigenstate of $`J_0^3`$ with $`J_0^3=m`$. Without loss of generality, we can restrict $`0\alpha <1`$. The representation is unitary if $`j=1/2+is`$ and $`s`$ is real<sup>11</sup><sup>11</sup>11Strictly speaking the representation with $`j=1/2,\alpha =1/2`$ is reducible as the sum of a highest weight and a lowest weight representation with $`j=1/2`$..
(4) Complementary representations:
A representation of this type is realized in the Hilbert space of
$$_j^\alpha =\{|j,\alpha ;m:m=\alpha ,\alpha \pm 1,\alpha \pm 2,\mathrm{}\},$$
where $`|j,\alpha ;m`$ is an eigenstate of $`J_0^3`$ with $`J_0^3=m`$. Without loss of generality, we can restrict $`0\alpha <1`$. The representation is unitary if $`j`$ is real, with $`1/2<j<1`$ and $`j1/2<|\alpha 1/2|`$.
(5) Identity representation:
This is the trivial representation with $`j=0`$.
The analysis that led to the above representation was completely algebraic and in a particular physical system we can have only a subset of all possible representations. Which of these representations appear in the Hilbert space of the WZW model? As the first approximation, let us consider the $`k\mathrm{}`$ limit. If we expand around a short string solutions, $`i.e.`$ oscillations near geodesics in $`AdS_3`$, the WZW model in this limit reduces to the quantum mechanics on $`AdS_3`$. The Hilbert space of the quantum mechanical model is the space of square-integrable<sup>12</sup><sup>12</sup>12Since $`AdS_3`$ is non-compact, we consider square-integrability in the delta-function sense. functions $`^2(AdS_3)`$ on $`AdS_3`$. The isometry of $`AdS_3`$ is $`SL(2,R)\times SL(2,R)`$, and one can decompose $`^2(AdS_3)`$ into its unitary representations. It is convenient to choose the basis of the Hilbert space in the following way. For each representation $``$, one can define a function on $`AdS_3`$ by $`F_{m,\overline{m}}(g)=m|g|\overline{m}`$ where $`gAdS_3`$, $`i.e.`$ universal cover of $`SL(2,R)`$, and $`|m`$ is an eigenstate of $`J_0^3`$ with $`J_0^3=m`$. Thus, for a given representation $``$ of $`SL(2,R)`$, the function $`F_{m,\overline{m}}(g)`$ on $`AdS_3`$ is in the tensor product of the representations $`\times `$ for the isometry group $`SL(2,R)\times SL(2,R)`$.
For a discrete representation $`𝒟_j^\pm `$, the wave-function $`f(\rho )`$ behaves as $`f(\rho )e^{2j\rho }`$ for large $`\rho `$. Thus $`\varphi ^2(AdS_3)`$ if $`j>1/2`$. Notice that in the range $`0<j<1`$ we have two representations with the same value of the Casimir but only one is in $`^2(AdS_3)`$, the one with $`1/2<j<1`$. As explained in , one could modify the norm so that the second solution with $`0<j<1/2`$ becomes normalizable. This modification of the norm is $`j`$-dependent. Similarly, supplementary series representations need a $`j`$-dependent modification to the norm to render them normalizable . Therefore these representations would appear in non-standard quantizations of geodesics, quantizations which do not use the $`^2`$ norm on $`AdS_3`$. In this paper, we will only consider the standard quantization using the $`^2`$ norm for the zero modes<sup>13</sup><sup>13</sup>13 Notice however, that even if the primary states have $`j>1/2`$, we could have states with smaller values of $`j_0`$ for the zero mode $`SL(2,R)`$ among the descendents, for example $`J_1^{}|j`$ with $`1<j<3/2`$, has $`j_0=j1<1/2`$.. Wave-functions in $`𝒞_{j=1/2+is}^\alpha `$ are also delta-function normalizable with respect to the $`^2`$ norm. It is known that $`𝒞_{j=1/2+is}^\alpha \times 𝒞_{j=1/2+is}^\alpha `$ and $`𝒟_j^\pm \times 𝒟_j^\pm `$ with $`j>1/2`$ form the complete basis of $`^2(AdS_3)`$.
For discrete lowest weight representations, the second Casimir is bounded above as $`c_2=j(j1)1/4`$. This corresponds to the well-known Breitenlohner-Freedman bound (mass)$`{}_{}{}^{2}1/4`$ for the Klein-Gordon equation. For the principal continuous representation $`𝒞_j^\alpha `$ with $`j=1/2+is`$, the second Casimir is $`c_2=1/4+s^2`$. Therefore an existence of such a particle would violate the Breitenlohner-Freedman bound. In the bosonic string theory, the only physical state of this type is the tachyon. In a perturbatively stable string theory, such particle states should be excluded from its physical spectrum. On the other hand, the continuous representations appear in $`^2(AdS_3)`$ and they are expected to be part of the Hilbert space of the WZW model before the Virasoro constraint is imposed.
### 4.2 Representations of the current algebra and no-ghost theorem
Given a unitary representation $``$ of $`SL(2,R)`$, one can construct a representation of $`\widehat{SL}(2,R)`$ by regarding $``$ as its primary states annihilated by $`J_{n1}^{3,\pm }`$. The full representation space is generated by acting $`J_{n1}^{3,\pm }`$ on $``$. Following the discussion in the previous subsection, we consider the cases when $`=𝒞_{j=1/2+is}^\alpha `$ and $`𝒟_j^\pm `$ with $`j>1/2`$. We denote by $`\widehat{𝒟}_j^\pm `$ and $`\widehat{𝒞}_j^\alpha `$ the representations of the full current algebra built on the corresponding representations of the zero modes. In Figure 5, we have shown the weight diagram of the positive energy representation $`\widehat{𝒟}_j^+`$.
A representation of $`\widehat{SL}_k(2,R)`$ in general contains states with negative norms. In order for a string theory on $`AdS_3`$ to be consistent, one should be able to remove these negative norm states by imposing the Virasoro constraint,
$$(L_n+_n\delta _{n,0})|\mathrm{physical}=0,n0,$$
(66)
on the Hilbert space for a single string state, where $`L_n`$ is the Virasoro generator of the $`SL(2,R)`$ WZW model and $`_n`$ for the sigma-model on $``$. It has been shown that this no-ghost theorem holds for states in $`\widehat{𝒞}_{j=1/2+is}^\alpha `$ or $`\widehat{𝒟}_j^\pm `$ with $`0<j<k/2`$ .
The no-ghost theorem is proved by first showing that all the solutions to the Virasoro constraint (66) can be expressed, modulo null states, as states in the coset $`SL(2,R)/U(1)`$ obeying
$$J_n^3|\psi =0,n1.$$
(67)
This statement is true for $`\widehat{𝒞}_{1/2+is}^\alpha `$ and $`\widehat{𝒟}_j^\pm `$ with $`0<j<k/2`$, if the total central charge of the Virasoro generator $`L_n+_n`$ is $`26`$ <sup>14</sup><sup>14</sup>14We also assume $`k>2`$.. We review the proof of this statement in the appendix A.1. The second step is to show that the condition (67) removes all negative norm states. This was shown in for the same class of representations.
The no-ghost theorem suggests that the spectrum of discrete representations has to be truncated for $`j<k/2`$. As we will see, this truncation is closely related to the existence of the long string states.
### 4.3 Spectral flow and the long string
The classical and semi-classical results discussed above indicate that, beyond positive energy representations that we have discussed so far, we have to include others related by spectral flow. To define a quantum version of the spectral flow, we note that, for any integer $`w`$, the transformation $`J_n^{3,\pm }\stackrel{~}{J}_n^{3,\pm }`$ given by
$$\stackrel{~}{J}_n^3=J_n^3\frac{k}{2}w\delta _{n,0},\stackrel{~}{J}_n^+=J_{n+w}^+,\stackrel{~}{J}_n^{}=J_{nw}^{},$$
(68)
preserves the commutation relations (62). The Virasoro generators $`\stackrel{~}{L}_n`$, which have the standard Sugawara form in terms of $`\stackrel{~}{J}_n^a`$, are different from $`L_n`$. They are given by
$$\stackrel{~}{L}_n=L_n+wJ_n^3\frac{k}{4}w^2\delta _{n,0}.$$
(69)
Of course, they obey the Virasoro algebra with the same central charge $`c`$. This is the same formula as saw in the classical counterpart (29) of the spectral flow.
The change of the basis (68) maps one representation into another, and this is called the spectral flow. In the case of a compact group such as $`SU(2)`$, the spectral flow maps a positive energy representation of the current algebra into another positive energy representation. An analogous transformation in the case of the $`N=2`$ superconformal algebra in two dimensions has been used to construct the spacetime supercharges for superstring.
In the case of $`SL(2,R)`$, the spectral flow generates a new class of representations. As shown in Figure 6, the spectral flow with $`w=1`$ maps the lowest weight representation $`\widehat{𝒟}_{\stackrel{~}{j}}^+`$ to a representation in which $`L_0`$ is not bounded below. The appearance of negative energy states is not too surprising since the kinetic term of the $`SL(2,R)`$ model is not positive definite. In general, a spectral flow of $`\widehat{𝒟}_{\stackrel{~}{j}}^+`$ with $`w1`$ or $`w2`$ gives a new representation in which $`L_0`$ is not bounded below. Similarly, the spectral flow of $`\widehat{𝒞}_{j=1/2+is}^\alpha `$ with $`w0`$ gives a representation in which $`L_0`$ is not bounded below. We denote the resulting representations by $`\widehat{𝒟}_{\stackrel{~}{j}}^{\pm ,w}`$ and $`\widehat{𝒞}_{\stackrel{~}{j}}^{\alpha ,w}`$, where $`\stackrel{~}{j}`$ labels the $`SL(2,R)`$ spin before the spectral flow.
These representations obtained by the spectral flow also contain negative norm states. In Appendix A.2, we generalize the proof of the no-ghost theorem and show that the Virasoro constraints indeed remove all negative norm states in the representations $`\widehat{𝒞}_{j=1/2+is}^{\alpha ,w}`$ and $`\widehat{𝒟}_{\stackrel{~}{j}}^{\pm ,w}`$ with $`\stackrel{~}{j}<k/2`$, for any integer $`w`$.
The only case where we get a representation with $`L_0`$ bounded below by the spectral flow is $`\widehat{𝒟}_j^\pm `$ with $`w=1`$. In this case, the representation is mapped to another positive energy representation $`\widehat{𝒟}_{\stackrel{~}{j}}^{\pm ,w=1}=\widehat{𝒟}_{k/2\stackrel{~}{j}}^{}`$. Note that, if we start with the representation with $`\stackrel{~}{j}>1/2`$, the representation one gets after the spectral flow satisfies $`j=k/2\stackrel{~}{j}<(k1)/2`$. Conversely, if there were a representation $`\widehat{𝒟}_j^\pm `$ with $`j>(k1)/2`$ in the Hilbert space, the spectral flow would generate a representation $`\widehat{𝒟}_j^{}`$ with $`j<1/2`$, in contradiction with the standard harmonic analysis of the zero modes in section 4.1. Therefore, if we assume that the spectral flow is a symmetry of the WZW model, the discrete representations $`\widehat{𝒟}_j^\pm `$ appearing in the Hilbert space are automatically restricted to be in $`1/2<j<(k1)/2`$. In particular, the spectrum of $`j`$ is truncated below the unitarity bound $`j<k/2`$ required by the no-ghost theorem. This further restriction on $`j`$ was discussed in a related context by .
### 4.4 Physical spectrum
Let us consider first the spectrum for strings with $`w=0`$. This is fairly standard. We start from an arbitrary descendent at level $`N`$ in the current algebra and some operator of the internal CFT with conformal weight $`h`$. The $`L_0`$ constraint reads
$$(L_01)|j,m,N,h=0\frac{j(j1)}{k2}+N+h1=0$$
(70)
If we demand that $`1/2j(k1)/2`$, this equation will have a solution as long as $`N+h`$ is within the range
$$0N+h1+\frac{1}{4(k2)}\frac{(k2)}{4}$$
(71)
If we allow $`j`$ to go all the way to $`k/2`$ we get $`k/4`$ on the right hand side of (71).
To analyze physical states of strings with $`w0`$, we start with a positive energy representation $`\widehat{𝒟}_{\stackrel{~}{j}}^+`$. After the spectral flow (68), a primary state $`|\stackrel{~}{j},\stackrel{~}{m}`$ of $`\widehat{𝒟}_{\stackrel{~}{j}}^+`$, as a state of $`\widehat{𝒟}_{\stackrel{~}{j}}^{+,w}`$, obeys
| $`J_{n+w}^+|\stackrel{~}{j},\stackrel{~}{m}=0,J_{nw}^{}|\stackrel{~}{j},\stackrel{~}{m}=0,J_n^3|\stackrel{~}{j},\stackrel{~}{m}=0,n1`$ |
| --- |
| $`J_0^3|\stackrel{~}{j},\stackrel{~}{m}=\left({\displaystyle \frac{k}{2}}w+\stackrel{~}{m}\right)|\stackrel{~}{j},\stackrel{~}{m}.`$ |
(72)
Let us look for physical states with respect to the Virasoro generator $`L_n`$. From (72), we find the Virasoro constraints are
| $`(L_01)|\stackrel{~}{j},\stackrel{~}{m}=\left({\displaystyle \frac{\stackrel{~}{j}(\stackrel{~}{j}1)}{k2}}w\stackrel{~}{m}{\displaystyle \frac{k}{4}}w^2+\stackrel{~}{N}+h1\right)|\stackrel{~}{j},\stackrel{~}{m},\stackrel{~}{N},h=0`$ |
| --- |
| $`L_n|\stackrel{~}{j},\stackrel{~}{m}=(\stackrel{~}{L}_nw\stackrel{~}{J}_n^3)|\stackrel{~}{j},\stackrel{~}{m}=0,n1.`$ |
(73)
where $`h`$ is the contribution to the conformal weight from the internal CFT and $`\stackrel{~}{N}`$ is the level inside the current algebra before we take the spectral flow. The state obeys the physical state conditions provided
$$\stackrel{~}{m}=\frac{k}{4}w+\frac{1}{w}\left(\frac{\stackrel{~}{j}(\stackrel{~}{j}1)}{(k2)}+\stackrel{~}{N}+h1\right).$$
(74)
The spacetime energy of this state measured by $`J_0^3`$ is then
$$J_0^3=\stackrel{~}{m}+\frac{k}{2}w=\frac{k}{4}w+\frac{1}{w}\left(\frac{\stackrel{~}{j}(\stackrel{~}{j}1)}{(k2)}+\stackrel{~}{N}+h1\right).$$
(75)
This is the quantum version of the classical formula (32), with the replacement
$$\frac{k}{4}\alpha ^2\frac{\stackrel{~}{j}(\stackrel{~}{j}1)}{k2}+1.$$
Notice that $`\stackrel{~}{m}=\stackrel{~}{j}+q`$ where $`q`$ is some integer, which could be negative<sup>15</sup><sup>15</sup>15$`\stackrel{~}{m}`$ is the total $`\stackrel{~}{J}^3`$ eigenvalue of the state so it can be lowered by applying $`J_n^{}`$ to the highest weight state. So we have the constraint $`q\stackrel{~}{N}`$.. Therefore the physical state condition becomes
$$\stackrel{~}{j}=\frac{1}{2}\frac{k2}{2}w+\sqrt{\frac{1}{4}+(k2)\left(h1+N_w\frac{1}{2}w(w+1)\right)}.$$
(76)
Here
$$N_w=\stackrel{~}{N}wq$$
(77)
is the level of the current algebra after the spectral flow by the amount $`w`$. Notice that the equation for $`\stackrel{~}{j}`$ is invariant under $`\stackrel{~}{N}\stackrel{~}{N}\pm w`$, $`qq\pm 1`$. This is reflecting the fact that $`J_0^\pm =\stackrel{~}{J}_w^\pm `$ commute with the Virasoro constraints and generate the spacetime $`SL(2,R)`$ multiplets. In particular, we see that the spacetime $`SL(2,R)`$ representations that we get are lowest energy representations, since repeated action of $`J_0^{}=\stackrel{~}{J}_w^{}`$ will eventually annihilate the state. In fact, it is shown in Appendix A.2 that the only physical state with zero spacetime energy, $`J_0^3=0`$, is the state $`J_1^{}|j=1`$, and its complex conjugate. This physical state corresponds to the dilaton field in $`AdS_3`$, which played an important role in the analysis of the spacetime Virasoro algebra in . All other states (except the tachyon with $`w=0`$) have nonzero energy, and form highest/lowest weight representations of $`SL(2,R)`$ spacetime algebra. The negative energy ones are the complex conjugates of the positive energy ones.
By solving the on-shell condition (76) for $`\stackrel{~}{j}>0`$ and substituting it into (75), one finds that the spacetime energy of the string is given by
$$\frac{E+\mathrm{}}{2}=J_0^3=q+w+\frac{1}{2}+\sqrt{\frac{1}{4}+(k2)\left(h1+N_w\frac{1}{2}w(w+1)\right)}.$$
(78)
Since both $`N_w`$ and $`q`$ are integers, the energy spectrum is discrete. This is reasonable since we are considering the string trapped inside of $`AdS_3`$. The constraint $`1/2<\stackrel{~}{j}<(k1)/2`$ translates into the inequality
$$\frac{k}{4}w^2+\frac{w}{2}<N_w+h1+\frac{1}{4(k2)}<\frac{k}{4}(w+1)^2\frac{w+1}{2}.$$
(79)
This is the quantum version of the semi-classical formula (59). In fact, if we take $`k,h\stackrel{~}{N},q,w`$, (79) reduces to (59). As in the semi-classical discussion, $`w`$ is not necessarily related to the physical winding number of the string. It is just an integer labeling the type of representation that the string state is in.
The analysis for the representations coming from the continuous representations for the zero modes is similar. If we do not spectral flow, the only state in the continuous representation is the tachyon. If we do spectral flow, we get the equation (74), which can be conveniently rewritten as
$$J_0^3=\stackrel{~}{m}+\frac{wk}{2}=\frac{kw}{4}+\frac{1}{w}\left(\frac{\frac{1}{4}+s^2}{k2}+\stackrel{~}{N}+h1\right)$$
(80)
For continuous representations $`w`$ is labeling the physical winding of the string when it approaches the boundary of $`AdS`$. In this case we do not get an equation like (76) since, for continuous representations, $`\stackrel{~}{m}`$ is not related to $`j`$. Comparing with the classical formula (40), we identify $`s`$ as the momentum $`\alpha /k`$ of the long string along the radial direction of $`AdS_3`$. We clearly see that the energy of this state is above the threshold to produce an excitation that will approach the boundary as a $`w`$-times wound string.
We can see that, whenever the value of $`h`$ is such that it saturates the range (79), we have a continuous representation with the same energy. This is clear for the lower bound in the case of $`w=0`$ since, for each state in the discrete representation with $`j=1/2`$, there is one in the continuous representation with the same values of $`L_0`$ and $`J_0^3`$. By the spectral flow, we see that the same is true for the lower bound in (79) for any $`w`$. Indeed we can check explicitly that a state in the discrete representation with parameters ($`h,w,q,\stackrel{~}{N}`$) saturating the lower bound in (79) has the same spacetime energy as a state in the continuous representation with parameters ($`h,w,s=0,\stackrel{~}{N}`$). (The parameter $`\alpha `$ in the continuous representation is fixed by the value of $`J_0^3`$ in (80).) Similarly, if we have a state in a discrete representation saturating the upper bound in (79), it has the same spacetime energy as a state in the continuous representation with parameters ($`h,w+1,s=0,\stackrel{~}{N}^{}=\stackrel{~}{N}+q`$). Note that, since $`q\stackrel{~}{N}`$ (see the footnote in the previous page), we have $`\stackrel{~}{N}^{}0`$. In this case, to show that the two states have the same energy, it is useful to identify the state in $`𝒟_{\stackrel{~}{j}=\stackrel{~}{j}}^{+,w}`$ as a state in $`𝒟_{\widehat{j}=k/2\stackrel{~}{j}}^{,w+1}`$. Since $`\stackrel{~}{j}(k1)/2`$ corresponds to $`\widehat{j}1/2`$ under this identification, we can apply the above argument for the lower bound to show that we will find a state in the continuous representation. The shift $`\stackrel{~}{N}^{}=\stackrel{~}{N}+q`$ comes from the fact that the identification $`𝒟_{\stackrel{~}{j}=\stackrel{~}{j}}^{+,w}=𝒟_{\widehat{j}=k/2\stackrel{~}{j}}^{,w+1}`$ involves spectral flow one more time.
The above paragraph explains what happens as we change $`\stackrel{~}{j}`$ in a discrete representation and we make it equal to the upper or lower bound: a continuous representation appears. Another question that one could ask is the following. Given a value of $`h`$, what is the state with the lowest value of $`J_0^3`$ that satisfies the physical state conditions? Let us first look for the lowest energy state in the discrete representations obeying the bound (79). Within this bound, one can show that $`J_0^3(h,w,q,\stackrel{~}{N})/q0`$ and $`J_0^3(h,w,q=\stackrel{~}{N},\stackrel{~}{N})/\stackrel{~}{N}0`$. Therefore, if we can set $`q=\stackrel{~}{N}=0`$, it will give the lowest energy state in the discrete representations. This is possible if $`h`$ is within the range,
$$\frac{k}{4}w^2+\frac{w}{2}<h1+\frac{1}{4(k2)}<\frac{k}{4}(w+1)^2\frac{w+1}{2}.$$
(81)
With some more work, one can show that, for $`h`$ in this range, there isn’t any state in a continuous representation whose energy is lower than that of the discrete representation state with $`\stackrel{~}{N}=q=0`$. As we saw in the above paragraph, at the upper or lower bound of (81), the energy of the discrete state $`(q=0,\stackrel{~}{N}=0)`$ coincides with that of the continuous state with $`(s=0,\stackrel{~}{N}=0)`$. Outside this range (81), it is not possible to set $`\stackrel{~}{N}=q=0`$, and the lowest energy state will be in a continuous representation. In our semi-classical discussion in the last section, we found that the discrete representation can decay into the continuous representation at $`h=kw^2/4`$. Now we see that, in the fully quantum description, the range over which a continuous representation has lower energy has expanded from the point $`h=kw^2/4`$ to a strip of width $`w`$:
$$\frac{k}{4}w^2\frac{w}{2}<h1+\frac{1}{4(k2)}<\frac{k}{4}w^2+\frac{w}{2}.$$
(82)
So far we have restricted our attention to right-moving sectors of the Hilbert space. Let us now discuss how the left and right movers are combined together. For the classical solution of the long string, the worldsheet periodicity requires that the spectral flow has to be done simultaneously on both the left and right movers with the same amount. If $`AdS_3`$ were not the universal cover of $`SL(2,R)`$ but its single cover, different amounts of the left and the right spectral flows would have been allowed since the resulting solution is periodic modulo the closed timelike curve of $`SL(2,R)`$. It is straightforward to identify the corresponding constraint in the quantum theory. Suppose we perform the spectral flows by the amount $`w_L`$ and $`w_R`$ on the left and the right-movers. A state with conformal weights $`(h_L,h_R)`$ and the $`J_0^3`$ charge $`(\stackrel{~}{m}_L,\stackrel{~}{m}_R)`$ is mapped by this transformation to a state with conformal weights $`(h_Lw_L\stackrel{~}{m}_L\frac{k}{4}w_L^2,h_Rw_R\stackrel{~}{m}_W\frac{k}{4}w_R^2)`$, according to (69). The worldsheet locality, which is the quantum counterpart of the periodicity of the classical solution, requires that the conformal weights $`h_L`$ and $`h_R`$ differ only by an integer. If this is the case before spectral flow, the same requirement after the flow implies
$$w_L\stackrel{~}{m}_L+\frac{k}{4}w_L^2=w_R\stackrel{~}{m}_R+\frac{k}{4}w_R^2(\mathrm{mod}\mathrm{integer}).$$
(83)
For generic values of $`(\stackrel{~}{m}_L,\stackrel{~}{m}_R)`$, the only solution to this constraint is $`w_L=w_R`$. In this paper, we are considering only the universal cover of $`SL(2,R)`$ as the target space of the model. In this case, the spectrum of $`(\stackrel{~}{m}_L,\stackrel{~}{m}_R)`$ is continuous, and only the left-right symmetric spectral flow $`w_L=w_R`$ is allowed.
Summary:
We propose that the spectrum of the $`SL(2,R)`$ WZW model (for the universal cover of $`SL(2,R)`$) contains the following two types of representations. First the spectral flow of the continous representations, with the same amount of spectral flow on the left and right, $`\widehat{𝒞}_{1/2+is,L}^{\alpha ,w}\times \widehat{𝒞}_{1/2+is,R}^{\alpha ,w}`$. Then the discrete representations $`\widehat{𝒟}_{\stackrel{~}{j},L}^{+,w}\times \widehat{𝒟}_{\stackrel{~}{j},R}^{+,w}`$ with the same amount of spectral flow on the left and right and the same value of $`\stackrel{~}{j}`$, with $`1/2<\stackrel{~}{j}<(k1)/2`$. In the string theory, these representations should be tensored with the states of the internal CFT, and the Virasoro constraints should be imposed.
## 5 Scattering of long string
When a long string comes in from the boundary of $`AdS_3`$ to the center, it will scatter back to the boundary. In this process the winding number could in principle change. In order to study the S-matrix between incoming and outgoing long strings, it is convenient to perform the rotations to Euclidean signature spaces, both on the worldsheet and in spacetime. Following the standard procedure, we define the hermiticity as is natural in the Lorentzian theory. For this reason we still have the $`SL(2,R)_L\times SL(2,R)_R`$ currents in the Euclidean theory. The relevant conformal field theory, whose target space is the 3-dimensional hyperbolic space $`H_3=SL(2,C)/SU(2)`$ has been studied in .
### 5.1 Vertex operators
To compute the scattering amplitudes, we would like to find vertex operators for all representations considered above. Spectral flow is realized in the vertex operator formalism in the following standard fashion . We bosonize the $`J^3`$ currents, introducing left and right moving chiral bosons<sup>16</sup><sup>16</sup>16 Reflecting the hermiticity of the $`SL(2,R)`$ model, the scalar field $`\varphi `$ is hermitian, but with a wrong sign for the two-point function $`\varphi (z)\varphi (z^{})=\mathrm{log}(zz^{})`$. through
$$J_R^3=i\sqrt{\frac{k}{2}}\varphi (z)J_L^3=i\sqrt{\frac{k}{2}}\overline{}\varphi (\overline{z})$$
(84)
A state with charge $`m`$ under $`J_R^3`$ contains an exponential in $`\varphi (z)`$ of the form $`e^{im\sqrt{\frac{2}{k}}\varphi (z)}`$. The other two currents therefore can be expressed as
$$J_R^+=\psi e^{i\sqrt{\frac{2}{k}}\varphi (z)},J_R^{}=\psi ^{}e^{i\sqrt{\frac{2}{k}}\varphi (z),}$$
(85)
and similarly for $`J_L^\pm `$. A primary field $`\mathrm{\Phi }_{jm\overline{m}}(z,\overline{z})`$ of the current algebra can be expressed as
$$\mathrm{\Phi }_{jm\overline{m}}=e^{im\sqrt{\frac{2}{k}}\varphi (z)+i\overline{m}\sqrt{\frac{2}{k}}\varphi (\overline{z})}\mathrm{\Psi }_{jm\overline{m}},$$
(86)
where $`\mathrm{\Psi }_{jm\overline{m}}`$ carries no charges with respect to $`J_{R,L}^3`$. In the case of the $`SU(2)`$ model, the field corresponding to $`\mathrm{\Psi }`$ is known as a parafermion. The parafermion for the $`SL(2,R)`$ model was studied in . The conformal weights of the parafermion field $`\mathrm{\Psi }_{jm\overline{m}}`$ is
| $`h_{\mathrm{\Psi };jm\overline{m}}={\displaystyle \frac{j(j1)}{k2}}+{\displaystyle \frac{m^2}{k}}`$ |
| --- |
| $`\overline{h}_{\mathrm{\Psi };jm\overline{m}}={\displaystyle \frac{j(j1)}{k2}}+{\displaystyle \frac{\overline{m}^2}{k}}.`$ |
(87)
In the discrete lowest weight representation, $`m,\overline{m}=j,j+1,j+2,\mathrm{}`$. In particular, when $`j=k/2`$, the field $`\mathrm{\Psi }_{j=k/2,m=\overline{m}=k/2}`$ has conformal weights $`h=\overline{h}=0`$. Since the parafermion field lives in the unitary conformal field theory it is natural to assume that it is the identity operator<sup>17</sup><sup>17</sup>17Recently we have learned that a similar argument has appeared in unpublished notes by A. B. Zamolodchikov. We thank him for having his note available to us .. Here we simply note that the operator
$$e^{i\sqrt{\frac{k}{2}}(\varphi (z)+\varphi (\overline{z}))}$$
has the correct OPE for the primary field of spin $`j=k/2`$ with the $`SL(2,R)`$ currents.
Using the parafermion notation, the operator obtained by the spectral flow by $`w`$ units is expressed as
$$\mathrm{\Phi }^w=e^{i(\stackrel{~}{m}+wk/2)\sqrt{\frac{2}{k}}\varphi (z)+i(\stackrel{~}{\overline{m}}+wk/2)\sqrt{\frac{2}{k}}\varphi (\overline{z})}\mathrm{\Psi }_{j\stackrel{~}{m}\stackrel{~}{\overline{m}}}$$
(88)
It is easy to see that the conformal weight is given by
$$L_0=\frac{j(j1)}{k2}mw+kw^2/2$$
(89)
### 5.2 Reflection coefficient
We will compute the amplitude, using the formulae obtained in , in the case that the winding number does not change.
The long string states are in the spectral flow of the continuum representation. The corresponding vertex operators are
| $`\mathrm{\Phi }_{m\overline{m}}^j=`$ | $`e^{m\varphi (z)+\overline{m}\varphi (\overline{z})}\mathrm{\Psi }_{\stackrel{~}{m}\stackrel{~}{\overline{m}}}^jV_{h\overline{h}}(z,\overline{z}),`$ |
| --- | --- |
| $`\stackrel{~}{m}=mwk/2,`$ | $`\stackrel{~}{\overline{m}}=\overline{m}wk/2,j={\displaystyle \frac{1}{2}}+is`$ |
(90)
where $`V_{h\overline{h}}`$ is an operator in the internal part with conformal weights $`(h,\overline{h})`$. The physical energy $`E`$ and angular momentum $`\mathrm{}`$ of a state in $`AdS_3`$ are given by
$$m=\frac{1}{2}(E+\mathrm{}),\overline{m}=\frac{1}{2}(E\mathrm{}).$$
(91)
The physical state constraint is (80) with $`\stackrel{~}{N}=0`$. This implies that
$$\stackrel{~}{m}=wk/4+\frac{1}{w}\left[\frac{1/4+s^2}{k2}+h1\right]$$
(92)
Now we can now consider the two point function
| $`\mathrm{\Phi }_{m\overline{m}}^j(z,\overline{z})\mathrm{\Phi }_{m^{}\overline{m}^{}}^j^{}(z^{},\overline{z}^{})=`$ | $`{\displaystyle \frac{\mathrm{\Gamma }(1/2+is\stackrel{~}{m})\mathrm{\Gamma }(1/2+is+\stackrel{~}{\overline{m}})\mathrm{\Gamma }(2is)\mathrm{\Gamma }(\frac{2is}{k2})}{\mathrm{\Gamma }(1/2is\stackrel{~}{m})\mathrm{\Gamma }(1/2is+\stackrel{~}{\overline{m}})\mathrm{\Gamma }(2is)\mathrm{\Gamma }(\frac{2is}{k2})}}`$ |
| --- | --- |
| | $`\times \delta (ss^{})\delta _{N+N^{}}\delta (E+E^{})`$ |
(93)
The $`z`$ dependence is just $`1/|zz^{}|^4`$ coming from the fact that the two operators have weight (1,1). This is the reflection amplitude and the values of $`\stackrel{~}{m},\stackrel{~}{\overline{m}}`$ are determined by (92)(notice that $`m`$ is the physical energy, not $`\stackrel{~}{m}`$.).
As explained in in this context, in string theory we have to integrate over $`z`$ and divide by the volume of $`SL(2,C)`$. We can use $`SL(2,C)`$ invariance to put $`z=0`$, $`z^{}=\mathrm{}`$ in the correlator. The volume of the rest of $`SL(2,C)`$ then gives $`\frac{d^2z}{|z|^2}`$, which cancels one of the delta-functions in (93). Notice that $`\delta (ss^{})\delta (E+E^{})=\delta (ss^{})\delta (0)`$, the volume of $`SL(2,C)`$ cancels the $`\delta (0)`$ piece.
Now if we study the poles of (93), we find that they are located at $`1/2+is\stackrel{~}{m}=q`$ with $`q=0,1,2,\mathrm{}`$. They come from the first Gamma-function. Taking this condition together with (92) we find that
$$1/2+is+q=\stackrel{~}{m}=wk/4+\frac{1}{w}\left[\frac{1/4+s^2}{k2}+h1\right]$$
(94)
and this equation is precisely the same as the usual mass shell equation for discrete states if we take $`\stackrel{~}{j}=1/2+is`$. There are similar poles from the second Gamma-function. There are no poles coming from the third factor since they cancel extra poles appearing in the other factors. Notice that the poles appearing in (94) satisfy precisely the equation (76) for bound states in the representation $`\widehat{𝒟}_{\stackrel{~}{j}}^{+,w}`$ (with $`\stackrel{~}{N}=0`$). There is however an important difference. In (76) the value of $`\stackrel{~}{j}`$ obeyed the condition
$$\frac{1}{2}<\stackrel{~}{j}<\frac{k1}{2}$$
(95)
while we do not have such a condition in (94). It is interesting to note that if $`\stackrel{~}{j}`$ satisfies (95) then the residue at the pole has the proper sign to be interpreted as coming from a bound state. When $`\stackrel{~}{j}=(k1)/2`$, $`i.e`$ at the upper bound of (95), we find that there is no pole. Moreover, immediately above that value, we have the wrong sign for the pole residue. This might make us worry that the amplitude is not having the right analytic structure. However, in order to have a one-to-one correspondence between poles of the scattering amplitude and bound states, the potential has to decrease sufficiently rapidly at the infinity , a condition that is not met in our case. In such a situation, it is possible to have extra poles that do not correspond to physical states. We plan to analyze the poles and their implications for physical states in a future publication.
### 5.3 Relation to the scattering off the two-dimensional black hole
The coset of the $`SL(2,R)`$ WZW by the $`U(1)`$ generated by $`J^3`$ gives a sigma-model whose target space is the two-dimensional black hole with the Euclidean signature metric . The geometry of the black hole is like a semi-infinite cigar with an asymptotic region in the form of the cylinder $`\text{}\times S^1`$. The dilaton field grows as one approaches the center of the black hole, but it remains finite since the geometry is terminated at the tip of the cigar. The string theory on $`SL(2,R)/U(1)\times `$(time)$`\times `$ is closely related to the string theory on $`AdS_3\times `$ since the physical state conditions for the latter implies $`J_n^3|\mathrm{physical}=0`$ for $`n1`$, as we show in Appendix A. Similarly the superstring theory on $`AdS_3\times `$ is related to the Kazama-Suzuki coset $`SL(2,R)/U(1)`$.
There is however difference between the zero mode sectors of the theories on $`AdS_3`$ and on the two-dimensional black hole. In order to construct representations for $`SL(2)/U(1)`$, we can start from the representations of $`\widehat{SL}(2,R)`$ that we described above and impose the condition that $`J_{n>0}^3`$ annihilate the state and that the total $`AdS_3`$ energy vanishes, $`J_{0,R}^3+J_{0,L}^3=m+\overline{m}=0`$. In terms of the parafermion $`\mathrm{\Psi }_{j\stackrel{~}{m}\stackrel{~}{\overline{m}}}`$ given in (86) and (88), the condition is $`\stackrel{~}{m}+\stackrel{~}{\overline{m}}=wk`$. The locality condition $`m\overline{m}=n`$ where $`n`$ is an integer implies that $`\stackrel{~}{m}\stackrel{~}{\overline{m}}=n`$. These two quantization conditions are the ones in (see equation 3.6 of that paper). The $`SL(2)/U(1)`$ theory has been studied recently in connection with “little” string theories in .
## 6 Conclusion
In this paper, we studied the physical spectrum of bosonic string theory in $`AdS_3`$. We proposed that the complete Hilbert space of the $`SL(2,R)`$ WZW model consists of the continuous representations and their spectral flow $`\widehat{𝒞}_{j=1/2+is}^{\alpha ,w}\times \widehat{𝒞}_{j=1/2+is}^{\alpha ,w}`$, and the discrete representations and their spectral flow $`\widehat{𝒟}_j^{\pm ,w}\times \widehat{𝒟}_j^{\pm ,w}`$ with the constraint $`1/2<j<(k1)/2`$. The sum over the spectral flow is required if we assume that the Hilbert space realizes the full loop group of $`SL(2,R)`$, including its topologically non-trivial elements. We found that this proposal leads to the physical spectrum of the string theory with the correct semi-classical limits.
In particular, we have solved the two puzzles which we mentioned in the introduction. The no-ghost theorem for $`\widehat{𝒟}_j^\pm `$ requires the constraint $`0<j<k/2`$. If we only had the unflowed sector (with $`w=0`$), it would imply the upper bound on allowed mass of string states, which appears artificial. This was one of the puzzles. We have resolved this puzzle by showing that the upper bound on the mass is removed if we include all the spectral flowed sectors in the Hilbert space. Moreover we showed that the consistency with the spectral flow and the standard harmonic analysis of the zero modes requires the constraint $`1/2<j<(k1)/2`$, more stringent than the one required by the no-ghost theorem. The constraint $`1/2<j<(k1)/2`$ is found to be consistent with the locations of the poles in the reflection coefficient (with the correct sign for the pole residues; see also ) and the modular invariance of the partition function.
Another puzzle was to identify states in the Hilbert space corresponding to the long strings. We found that these states are in the spectral flow of the continuous representations, $`\widehat{𝒞}_{j=1/2+is}^{\alpha ,w}\times \widehat{𝒞}_{j=1/2+is}^{\alpha ,w}`$. The integer $`w`$, which parametrized the amount of the spectral flow, is identified with the winding number of the long string stretched closed the boundary of $`AdS_3`$. The physical spectrum of the long strings obtained from these representations agrees with the expectations from the semi-classical analysis in .
The resolutions of these puzzles removes the longstanding doubts about the consistency of the model. Moreover it appears that the $`SL(2,R)`$ WZW model is exactly solvable, just as WZW models for compact groups, although its Hilbert space structure is significantly different from those of the compact cases. We hope that further study of the model will provide us more useful insignts into the $`AdS`$/CFT correspondence and strings in curved spaces in general.
## Acknowledgements
We would like to thank A. Zamolodchikov for discussions and for giving us a copy of his unpublished notes. We also thank N. Seiberg, C. Vafa and E. Witten for discussions. We would like to thank S. Hwang for useful comments on the earlier version of this paper.
H.O. would like to thank J. Schwarz and the theory group at Caltech for the kind hospitality while the bulk of this work was carried out. H.O. also thanks the hospitality of the theory group at Harvard University, where this work was initiated, ICTP, Trieste and ITP, Santa Barbara, where parts of this work were done.
The research of J.M. was supported in part by DOE grant DE-FGO2-91ER40654, NSF grant PHY-9513835, the Sloan Foundation and the David and Lucile Packard Foundations. The research of H.O. was supported in part by NSF grant PHY-95-14797, DOE grant DE-AC03-76SF00098, and the Caltech Discovery Fund.
## Appendix A No-ghost theorems
In this appendix we would like to extend the proof of the no-ghost theorem to all the representations considered above. We assume $`k>2`$. The proof of the no-ghost theorem for the standard lowest energy representations involves two parts. Part I consists of showing that a physical state can be chosen, up to a null state, to be such that $`J_n^3|\psi >=0`$, for $`n1`$. This first part uses $`0<j<k/2`$ for the $`𝒟_j^\pm `$ representations as well as $`c=26`$ and the mass shell condition. This was shown in . Part II consists in showing that any state that is annihilated by $`J_{n>0}^3`$ has non negative norm. This step also uses $`0<j<k/2`$ for the $`𝒟_j^\pm `$ representations. This was done in . Here we will use the same strategy and prove Part I for the all our representations. The no-ghost theorem then follows from Part II.
We first review the proof of Part I for the representations with $`w=0`$ and then we do Part I for the $`w0`$ representations.
### A.1 Proof of Part I for unflowed representations
Here we follow the proof in . It has essentially three steps.
Step 1: The first step of the proof is to show that states of the form
| $`L_{n_1}L_{n_2}\mathrm{}L_{n_N}J_{m_1}^3J_{m_2}^3\mathrm{}J_{m_M}^3|f`$ |
| --- |
| $`n_1n_2\mathrm{}n_N,m_1m_2\mathrm{}m_M,`$ |
| $`\mathrm{with}L_n|f=J_n^3|f=0\mathrm{for}n1,`$ |
(96)
are linearly independent and that they form a complete basis of the Hilbert space.
The states $`|f`$ are constructed from states in the current algebra times some states in an internal conformal field theory. This internal piece is assumed to be unitary. This step involved separating the piece of $`L_n`$ involving $`L^{(3)}=:J^3J^3:`$, defining $`\widehat{L}_n=L_nL_n^{(3)}`$. One can show that the states (96) are in one to one correspondence with states of the form
| $`L_{n_1}L_{n_2}\mathrm{}L_{n_N}J_{m_1}^3J_{m_2}^3\mathrm{}J_{m_M}^3|f`$ |
| --- |
| $`n_1n_2\mathrm{}n_N,m_1m_2\mathrm{}m_M,`$ |
(97)
Notice that the conditions (96) on $`|f`$ are the same as $`\widehat{L}_{n>0}|f=J_{n>0}^3|f=0`$. It is easier to show that (97) is a basis since now we can think of the CFT as a product of a $`U(1)`$ factor with the rest. The rest is a CFT with $`c=25`$ and therefore the fact that (97) is a basis reduces to showing that there are no null states in the Virasoro descendents on a primary field. This will be true if the conformal weight of the rest is positive. This reduces to showing that $`c_2/(k2)+m^2/k+M>0`$, where $`M`$ is the grade in the $`SL(2,R)`$ piece. For the continuous representations, this is obvious since $`c_2>0`$. For lowest weight representations, this inequality can be shown by rewriting it as
$$\frac{2j\left(k/2j\right)}{k(k2)}+\frac{2M}{k}\left(\frac{k}{2}j\right)+\frac{2j}{k}(j+m+M)+\frac{1}{k}(jm)^2>0$$
(98)
We to use $`0<j<k/2`$ and also the fact that $`mjM`$, which is true in general. Notice that the $`m`$ that appears here is the total $`J_0^3`$ value, after we applied $`J_n^\pm `$ any number of times. Notice that in this step we did not use that the states were obeying the mass shell condition, but we used $`0<j<k/2`$ and that $`c=26`$.
Step 2: Here we show that a physical state can be chosen so that it involves no $`L_n`$ when written as (96).
A physical state can be written as a state with no $`L_n`$ plus a spurious state. A spurious state is a state with at least one $`L_n`$. Then we use the fact that, when $`c=26`$, $`L_n`$ ($`n1`$) acting on a spurious state which satisfies the $`L_0=1`$ condition leaves it as a spurious state . If $`L_{n>0}`$ acts on a state of the form (96) with no $`L_n`$ then it will not produce any $`L_n`$. Together with the fact that (96) is a basis this implies that the part of the state with no $`L_n`$ satisfies the physical state condition on its own, and therefore the rest is a null state (a spurious physical state).
Step 3: We show that if the physical state $`|\psi `$ involves no $`L_n`$ when written as in (96) then $`J_n^3|\psi =0`$.
Since there are no $`L_n`$’s in the physical state $`\psi `$ this implies that $`L_n^{(3)}\psi =0`$ for $`n1`$. Then we try to show that the only states satisfying this will be states with $`J_n^3\psi =0`$ for $`n1`$. This would be true if there are no null states in the $`L^{(3)}`$ Virasoro descendents of the states $`|f`$ we considered above. If $`m0`$ then one can show that there is no null state in the Virasoro descendents in the $`L^{(3)}`$ Virasoro descendents. There are two states with $`m=0`$ one is in the continuous representation, but the mass shell condition automatically implies that $`N=0`$ (there are no $`J_n^a`$ in this state) and therefore the state has positive norm. The other is the state in the lowest weight representation
$$J_1^{}|j=1$$
(99)
(and of course its complex conjugate in the highest weight representation). This state has positive norm. Note that $`m`$ is the physical energy in $`AdS_3`$ of the state in question. Zero energy states, therefore imply that we have a normalizable zero mode. This is the state corresponding to the identity operator in the spacetime boundary conformal field theory, the state $`\overline{J}J\mathrm{\Phi }_1`$ of which played an important role in the computation of the spacetime Virasoro algebra.
One can show, using the mass shell condition, that all other states have $`m0`$. The mass shell condition is
$$\frac{j(j1)}{k2}+N+h^{}1=0$$
(100)
where $`N`$ is the grade in the $`SL(2,R)`$ part and $`h^{}`$ is the conformal weight of the rest, $`h^{}0`$. If $`0<j<1`$ then $`m`$ is nonzero because it can only change by an integer by the action of the $`J_n^\pm `$ currents. If $`j=1`$ with $`N=1`$ and $`h^{}=0`$ we find (99)and states with positive $`m`$.
Consider now $`j>1`$. If we had $`m=0`$ then we also need $`Nj`$, $`j2`$ (since $`m=0`$ only if $`j`$ is integer) and furthermore
$$\frac{j(j1)}{k2}+N1\frac{(j1)(k2j)}{k2}>0$$
(101)
provided $`jk/2`$. Since $`j`$ has to be at least 2, then $`k>4`$ and therefore $`k2k/2>0`$. Thus we conclude that (100) would not be obeyed if $`m=0`$.
### A.2 Proof of Part I for flowed representations
Now we would like to generalize the above discussion to the spectral flowed representations that we called $`\widehat{𝒞}_{1/2+is}^{\alpha ,w}`$ and $`\widehat{𝒟}_{\stackrel{~}{j}}^{+,w}`$. In the case of discrete representations we want to show that the no ghost theorem holds for $`0<\stackrel{~}{j}<k/2`$ where $`\stackrel{~}{j}`$ labels the representation before we perform the spectral flow operation, i.e. it labels a representation of the current algebra with $`\stackrel{~}{L}_0`$ bounded below. So we consider the same representations we had above but we modify the physical state conditions. This is equivalent to imposing the usual conditions on the flowed representations. We would like to prove that, given any state built on a lowest weight or continuous representation with respect to $`\stackrel{~}{J}_n`$, the physical state condition $`(L_n\delta _{n,0})|\psi =0`$ $`n0`$ with respect to $`L_n`$ removes non-negative norm states. We only consider spectral flow with $`w>1`$ on continuous or lowest weight representations $`\widehat{𝒟}_{\stackrel{~}{j}}^+`$. These and their complex conjugates cover all the representations we needed to consider. We reproduce now the steps in A.1.
Step 1: In (96) we need to show that they form a basis with $`L_n=\stackrel{~}{L}_nw\stackrel{~}{J}_n^3`$. We know that they would form a basis if we had an expression like (96)with $`L_n\stackrel{~}{L}_n`$. Fortunately there is an invertible one to one map between these two sets of states, so that they form a basis.
Step 2: It is the same since only $`c=26`$ is used.
Step 3: If we write a physical state, $`|\psi `$, as a state with no $`L_n`$ then $`L_n^{(3)}`$ with $`n1`$ annihilates it. Again we will try to show that $`m=\stackrel{~}{m}+kw/2`$ is nonzero and that will imply that $`J_{n>0}^3|\psi =0`$. For this we need to use the new mass shell condition
$$\frac{\stackrel{~}{c}_2}{k2}+\stackrel{~}{N}+h^{}wm+\frac{kw^2}{4}=1$$
(102)
where $`\stackrel{~}{N}`$ is the level inside the current algebra before the spectral flow, $`\stackrel{~}{c}_2`$ is the second casimir in terms of $`\stackrel{~}{j}`$ and $`h^{}`$ is the conformal weight of state in the internal conformal theory (the internal piece needs not be a primary state, and we only require that the whole combined state needs to be primary). We can assume with no loss of generality that $`w1`$. Let us start with the spectral flow of a continuous representation, (102)implies that if $`m=0`$ then $`\stackrel{~}{N}=0`$ and and there are no negative norm states. (The only solution with $`m=0`$ is in the case of $`k=3`$ and $`\stackrel{~}{j}=1/2`$).
Let us turn to lowest weight representations. Thanks to the restriction $`0<\stackrel{~}{j}<k/2`$, we have $`\stackrel{~}{c}_2/(k2)>k/4`$. Therefore, if $`m=0`$, the left hand side of (102) is larger than $`k/4(w^21)`$. If $`w2`$, (102) cannot be obeyed. If $`w=1`$, $`m=0`$ implies $`\stackrel{~}{m}=k/2`$ and $`\stackrel{~}{N}`$ in (102) has to be at least $`\stackrel{~}{N}\stackrel{~}{j}+k/2`$. However, in this case we find $`\stackrel{~}{c}_2/(k2)+\stackrel{~}{N}+k/4k/2+\stackrel{~}{j}>1`$ (here we used $`k>2`$) and again (102) is not satisfied.
So we conclude that all states can be mapped into states obeying $`J_{n>0}^3|\psi =0`$.
## Appendix B Partition function
In this Appendix, we discuss the partition function of the $`SL(2,R)`$ WZW model and its modular invariance.
### B.1 Partition function of the SU(2) model
Before we begin discussing the modular invariance of the $`SL(2,R)`$ theory, let us review the case of $`SU(2)`$.
The characters $`\chi _l^k(\tau ,\theta )`$ ($`l=0,\frac{1}{2},1,\mathrm{}\frac{k}{2}`$) of the irreducible representations of the $`SU(2)_k`$ affine algebra transform under the modular transformation as
$$\chi _l^k(1/\tau ,\theta /\tau )=\mathrm{exp}\left(2\pi i\frac{k}{4}\frac{\theta ^2}{\tau }\right)\underset{l^{}}{}S_{ll^{}}\chi _l^{}^k(\tau ,\theta ),$$
(103)
where $`S_{ll^{}}`$ is some orthonormal $`(k+1)\times (k+1)`$ matrix. The diagonal (so-called $`A_k`$-type) modular invariant combination is therefore
$$e^{2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}\underset{l}{}|\chi _l(\tau ,\theta )|^2.$$
(104)
The exponential factor $`e^{2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}`$ is there to cancel the exponential factor in (103) as
$$\frac{[Im(\theta /\tau )]^2}{Im(1/\tau )}=\frac{(Im\theta )^2}{Im\tau }+i\frac{\theta ^2}{2\tau }i\frac{\overline{\theta }^2}{2\overline{\tau }}.$$
(105)
It is known that the exponential factor in (104) is a consequence of the chiral anomaly and therefore of the OPE singularity,
$$J^3(z)J^3(w)\frac{k/2}{(zw)^2}.$$
(106)
### B.2 Partition function of the SL(2,C)/SU(2) model
In string theory, one-loop computations are done after performing the Euclidean rotation on both the target space and the worldsheet (or stay in the Lorentzian signature space and use the $`iϵ`$ prescription). The modular invariance of the partition function is imposed on the Euclidean worldsheet. In our case, the Euclidean rotation of the target space means $`SL(2,R)H_3=SL(2,C)/SU(2)`$. The partition function of the $`SL(2,C)/SU(2)`$ model has been evaluated in as
$$Z_{SL(2,C)/SU(2)}\frac{1}{\sqrt{Im\tau }e^{\frac{2\pi (Im\theta )^2}{Im\tau }}|\vartheta _1(\tau ,\theta )|^2}.$$
(107)
Note that our definition of the partition function differs from that in by the factor $`e^{2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}`$. It apears that, without this factor, the partition function is not modular invariant<sup>18</sup><sup>18</sup>18The puzzle about the apparent lack of the the modular invariance was recently resolved in .. One may expect that this partition function is related to the one for the $`SL(2,R)`$ model by the Euclidean rotation. In the discussion below, we first evaluate the $`SL(2,R)`$ partition function on the Lorentzian torus, and therefore take $`\tau ,\overline{\tau },\theta ,\overline{\theta }`$ to be independent real variables. We then analytically continue them to complex values so that $`(\tau ,\theta )`$ are complex conjugate of $`(\overline{\tau },\overline{\theta })`$. We will find that, by doing this analytic continuation, and ignoring contact terms, the $`SL(2,R)`$ partition function turns into the $`SL(2,C)/SU(2)`$ partition function (107), provided we impose the constraint $`1/2<j<(k1)/2`$ on the discrete representations.
### B.3 Discrete representations of SL(2,R)
The character of the discrete representation $`D_j^+`$ is
| $`\chi _j^+(\tau ,\theta )=`$ | $`\mathrm{𝑇𝑟}(e^{2\pi i\tau (L_0\frac{k}{8(k2)})}e^{2\pi i\theta J_0^3})`$ |
| --- | --- |
| $`=`$ | $`{\displaystyle \frac{\mathrm{exp}\left[2\pi i\tau \left(\frac{j(j1)}{k2}\frac{k}{8(k2)}\right)+2\pi i\theta j\right]}{(1e^{2\pi i\theta })_{n=1}^{\mathrm{}}(1e^{2\pi in\tau })(1e^{2\pi in\tau }e^{2\pi i\theta })(1e^{2\pi in\tau }e^{2\pi i\theta })}}`$ |
| $`=`$ | $`{\displaystyle \frac{\mathrm{exp}\left[\frac{2\pi i\tau }{k2}(j\frac{1}{2})^2+2\pi i\theta (j\frac{1}{2})\right]}{i\vartheta _1(\tau ,\theta )}}.`$ |
(108)
where $`\vartheta _1(\tau ,\theta )`$ is the elliptic theta-function
$$\vartheta _1(\tau ,\theta )=i\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}(1)^n\mathrm{exp}\left[\pi i\tau \left(n\frac{1}{2}\right)^2+2\pi i\theta \left(n\frac{1}{2}\right)\right].$$
(109)
The spectral flow
$$\stackrel{~}{L}_0=L_0+wJ_0^3\frac{k}{4}w^2,\stackrel{~}{J}_0^3=J_0^3\frac{k}{2}w,(w=0,\pm 1,\pm 2,\mathrm{}),$$
(110)
transforms the character $`\chi _j^+`$ as
| | $`\mathrm{𝑇𝑟}(e^{2\pi i\tau (\stackrel{~}{L}_0\frac{k}{8(k2)})}e^{2\pi i\theta \stackrel{~}{J}_0^3})`$ |
| --- | --- |
| $`=`$ | $`\mathrm{𝑇𝑟}(e^{2\pi i\tau (L_0+wJ_0^3\frac{k}{4}w^2\frac{k}{8(k2)})}e^{2\pi i\theta (J_0^3\frac{k}{2}w)})`$ |
| $`=`$ | $`{\displaystyle \frac{\mathrm{exp}\left[2\pi i\tau \left(\frac{(j\frac{1}{2})^2}{k2}w(j\frac{1}{2})+\frac{k}{4}w^2\right)+2\pi i\theta (j\frac{1}{2}\frac{k}{2}w)\right]}{i\vartheta _1(\tau ,\theta +w\tau )}}`$ |
| $`=`$ | $`(1)^w{\displaystyle \frac{\mathrm{exp}\left[\frac{2\pi i\tau }{k2}(j\frac{1}{2}\frac{k2}{2}w)^2+2\pi i\theta (j\frac{1}{2}\frac{k2}{2}w)\right]}{i\vartheta _1(\tau ,\theta )}},`$ |
(111)
where we used
$$\vartheta _1(\tau ,\theta +w\tau )=(1)^w\mathrm{exp}\left(\pi i\tau w^22\pi i\theta w\right)\vartheta _1(\tau ,\theta ).$$
(112)
We have also performed an analytic continuation such as
$$\underset{n=0}{\overset{\mathrm{}}{}}q^n=\underset{n=1}{\overset{\mathrm{}}{}}q^n,$$
ignoring terms like $`_{n=\mathrm{}}^{\mathrm{}}q^n\delta (\tau )`$. From here on, we allow $`(\tau ,\theta )`$ to take complex values and $`(\overline{\tau },\overline{\theta })`$ to be their complex conjugates.
Let us sum over allowed representation. According to our proposal about the Hilbert space of the WZW model, all the representations in the allowed range $`1/2<j<(k1)/2`$ should appear. We also require that the spectrum to be invariant under the spectral flow (110), so we need to sum over $`w`$. The part of the partition function made by discrete representations is then
| $`e^{+2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}{\displaystyle \underset{w=\mathrm{}}{\overset{\mathrm{}}{}}}{\displaystyle _{1/2}^{(k1)/2}}𝑑j{\displaystyle \frac{\mathrm{exp}\left[\frac{4\pi Im\tau }{k2}(j\frac{1}{2}\frac{k2}{2}w)^24\pi Im\theta (j\frac{1}{2}\frac{k2}{2}w)\right]}{|\vartheta _1(\tau ,\theta )|^2}}`$ |
| --- |
| $`=e^{+2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑t{\displaystyle \frac{\mathrm{exp}\left[\frac{4\pi Im\tau }{k2}t^24\pi Im\theta t\right]}{|\vartheta _1(\tau ,\theta )|^2}}`$ |
| $`{\displaystyle \frac{1}{\sqrt{Im\tau }e^{2\pi \frac{(Im\theta )^2}{Im\tau }}|\vartheta _1(\tau ,\theta )|^2}}.`$ |
(113)
It is interesting to note that the $`j`$-integral over the range $`1/2<j<(k1)/2`$ and the sum over $`w`$ fit together to give the $`t`$-integral over $`\mathrm{}<t<\mathrm{}`$. Since the spectral flow with $`w=1`$ maps $`D_j^+`$ to $`D_{k/2j}^{}`$, we do not have to consider the orbit of $`D_j^{}`$ separately. The exponential factor $`e^{+2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}`$ is due to the chiral anomaly, as in the $`SU(2)`$ case. The sign in the exponent is opposite here since the sign of the OPE of $`J^3`$ is opposite in the $`SL(2,R)`$ case.
The partition function computed in (113) is manifestly modular invariant. In fact, it is identical to (107) computed for the $`SL(2,C)/SU(2)`$ model. This gives an additional support for our claim that the Hilbert space of the $`SL(2,R)`$ model contains the discrete representations of $`1/2<j<(k1)/2`$ and their spectral flow.
The construction of the partition function here is closely related to the one given in . There, instead of the integral over $`j`$ in (113), the partition function was given by a sum over integral values of $`j`$. This is because they considered the string theory on the single cover of the $`SL(2,R)`$ group manifold with the closed timelike curve. The resulting partition function, after analytic continuation, is also modular invariant and appears to be a correct one for such a model. It is, however, different from the partition function (107) of the $`SL(2,C)/SU(2)`$ model, as it should since the Euclidean rotation of the $`SL(2,C)/SU(2)`$ model is naturally related to the model on the universal cover of $`SL(2,R)`$ rather than on its single cover.
### B.4 Continuous representations
It is curious that the sum over the discrete representations and their spectral flow alone reproduces the partition function of the $`SL(2,C)/SU(2)`$ model. In fact, the sum over the continuous representations and their spectral flow, although formally modular invariant by itself, does not contribute to the partition function if we assume the analytic continuation in $`\tau ,\overline{\tau },\theta ,\overline{\theta }`$ and ignore contact terms.
The character of the continuous representation is parametrized by a pair of real numbers $`(s,\alpha )`$ with $`0\alpha <1`$ and $`s`$ arbitrary. The character is given by
$$\chi _{j=1/2+is,\alpha }=\eta ^3e^{2\pi i\frac{s^2}{k2}\tau }e^{i\alpha \theta }\underset{n}{}e^{2\pi in\theta }.$$
(114)
As before, we regard the worldsheet metric to be of the Minkowski signature, and $`\theta `$ is real. So the sum $`_n`$ in the definition of $`\chi _{j,\alpha }`$ gives the periodic delta-function
$$\underset{n}{}e^{2\pi in\theta }=2\pi \underset{m}{}\delta (\theta +m).$$
(115)
After the spectral flow (110), the character becomes
$$\chi _{j=1/2+is,\alpha ;w}=\eta ^3e^{2\pi i\left(\frac{s^2}{k2}+\frac{k}{4}w^2\right)\tau }2\pi \underset{m}{}e^{2\pi im(\alpha \frac{k}{2}w)}\delta (\theta +w\tau +m).$$
(116)
Now let us take $`|\chi _{1/2+is,\alpha ;w}|^2`$ and integrate over $`s`$ and $`\alpha `$. The integral over $`\alpha `$ forces $`m_L=m_R`$ in the summation in (116). The integral over $`s`$ gives the factor $`1/\sqrt{Im\tau }`$. So we have
$$_{\mathrm{}}^{\mathrm{}}𝑑s_0^1𝑑\alpha |\chi _{1/2+is,\alpha ;w}|^2=e^{4\pi Im\tau \frac{k}{4}w^2}\frac{1}{\sqrt{Im\tau }|\eta |^6}\underset{m}{}\delta ^{(2)}(\theta +w\tau +m).$$
(117)
Let us sum this over $`w`$. We get a non-zero result only when there is some integer $`w`$ such that
$$w=\frac{Im\theta }{Im\tau }.$$
(118)
Therefore
| $`e^{+2\pi \frac{k}{2}\frac{(Im\theta )^2}{Im\tau }}{\displaystyle \underset{w}{}}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑s{\displaystyle _0^1}𝑑\alpha |\chi _{1/2+is,\alpha ;w}|^2`$ |
| --- |
| $`={\displaystyle \frac{1}{\sqrt{Im\tau }|\eta |^6}}{\displaystyle \underset{w,m}{}}\delta ^{(2)}(\theta +w\tau +m).`$ |
(119)
This expression is formally modular invariant since $`_{w,m}`$ sums over the modular orbit of the delta-function and $`1/|\eta |^4`$ cancels its modular weight. If we assume the analytic continuation, terms of this form are all set equal to zero. So, in this sense, the continuous representation does not contribute to the partition function of the $`SL(2,C)/SU(2)`$ theory after the Euclidean rotation. |
warning/0001/cond-mat0001439.html | ar5iv | text | # Metal-Kondo insulating transitions and transport in one dimension
## I Introduction
The interplay between magnetic impurities and itinerant electrons gives rise to fascinating situations, in link with the so called Kondo effect. Although the physics of one impurity in a metal is well understood, when interactions are included or the number of impurities increased, the problem remains largely open, solutions existing however through Gutzwiller variational approximations, mean-field methods, or still infinite-dimension treatments.
The one-dimensional (1D) models are usually much easier to handle than their counterpart in higher dimensions. They can even prove to be exactly solvable, as is the case for the Kondo model in a free electron gas, or still the 1D Hubbard model. Even for more complicated models, very powerful techniques, such as bosonization or renormalization calculations, are still applicable and generally give the correct physical results: For instance, these have allowed to predict the generic Luttinger liquid concept, induced at low energy by weak electron-electron interactions. Such one-dimensional physics has received a consequent attention recently, due to for examples, advances in nanofabrication, the existence of edge states in the fractional quantum Hall effect and the discovery of novel 1D materials such as carbon nanotubes. Finally, low dimensional models can still provide valuable information on the role of correlation effects in higher dimensions, e.g., on the physics of correlated fermions in two dimensions (in link with high-$`T_c`$ materials) or, still in our context, on the phase diagram of the two-impurity Kondo model in a three dimensional Landau-Fermi liquid.
Since the discovery of Anderson localization, non-magnetic impurity effects in these Luttinger liquids (LL’s) have always been a fascinating subject. The two extreme situations, respectively of one or two impurities and of a finite density of scatterers is now quite well understood. In this paper, we rather ponder the role of magnetic impurities in the transport of LL’s, using bosonization techniques. Precisely, we start with a conduction band very close to *half-filling* and a *perfect lattice* of quantum impurities, coupled through an antiferromagnetic Kondo exchange $`J_K`$: This produces a Kondo lattice model (KLM). As in higher-dimensions, this results in a Kondo insulating phase. Then, we may investigate two different classes of metal-Kondo insulating transitions occurring in low-dimensional KLM’s.
First, we study the commensurate-incommensurate transition arising in the strong Kondo coupling limit $`J_K/t1`$$`t`$ is the hopping amplitude of electrons — by analogy to Mott insulators. We show how the Kondo coupling plays the role of a strong umklapp process for conduction electrons.
Second, we stay at half-filling and study the weak coupling limit $`J_K/t1`$. Here, metal-insulating transitions arise rather by the strong renormalization of the backward Kondo exchange. The central idea of this part is therefore to show how the interplay between electron-electron interactions in the LL and backward Kondo scattering creates noteworthy phenomena in transport properties, in the delocalized regime. For weak attractive interactions, as a precursor of superconductivity, this will produce both a large jump in the Drude weight and a maximum in the d.c. conductivity, in the entrance of the Kondo insulating regime. For repulsive ones, the LL yields prominent spin-fluctuations, that can be easily pinned by local moments: Then the d.c. conductivity decreases *monotonically* with temperature, even at high temperatures. There is no remnant of the original umklapp process — driven by the Hubbard term.
We also make comparisons with the case of randomly distributed magnetic defects: This could lead to predictions on persistent currents of mesoscopic rings with prominent quantum defects.
The precise plan of the paper is organized as follows. In section II, we consider the pure KLM in the strong Kondo coupling limit, and show how a 1D Kondo insulator can be understood as an umklapp becoming relevant. Implications on the resulting commensurate-incommensurate transition are then considered. In section III, we present several Kondo insulators occurring in the weak Kondo coupling limit, dependently on the interaction between local moments and their spins. We also draw the phase diagram as a function of electron-electron interactions. In the central section IV, we investigate transport properties (a.c. and d.c. conductivities, Drude weight) in the high-temperature delocalized regime, and make substantial links with non-magnetic and magnetic Gaussian-correlated disorders. Finally, in section V, we link the two extreme cases, single- and many quantum impurities.
## II KLM in Strong coupling, and doping
Let us start with the pure KLM in the strong Kondo coupling limit $`J_Kt`$, at half-filling. The weak original electron-electron interaction can be here neglected. Moreover, in such limit, including a direct exchange J between local moments does not affect the results on charge properties at the commensurate-incommensurate transition. So, we ignore it as well and for simplicity *in this section*, we consider local spins with S=1/2. For an explicit description of such phase, consult ref..
One of the remarkable properties of the present spin-liquid phase is its different energy scales in spin and charge degrees of freedom. Most clearly, it is characterized by a ratio of the charge- and spin gaps, $`\mathrm{\Delta }_c/\mathrm{\Delta }_S`$ which is not equal to unity. The charge gap is larger by $`50\%`$ ($`\mathrm{\Delta }_S=J_K`$ and $`\mathrm{\Delta }_c=3J_K/2`$).
Of course, spin properties in the 1D Kondo lattice model (KLM) are completely different from those in the large $`U`$ Hubbard model. This is rather described by a quasi long-range Resonating Valence Bond wave function due to the absence of a spin gap in the spectrum. However, the origin of the charge gap is that when another electron is added to a local singlet, it costs a large energy of the order of $`J_K`$ by breaking the singlet. This process works as a strong effective on-site repulsion between original electrons of the order of $`J_K`$. The mechanism for opening the charge gap is completely identical to that in a Mott-Hubbard solid. Integrating out spin degrees of freedom (see Fig.1), a Kondo insulator occurs only as a result of commensurability: *In 1D, this may manifest in umklapp scattering becoming relevant*. In the following, we explain precisely why a 1D Kondo insulator can be built identically to a Mott insulator.
### A Mott insulator versus Kondo insulator
In a 1D Mott insulator, as a result of commensurability, the charge Hamiltonian is slightly more complicated than the so-called Luttinger Hamiltonian
$$_c=\frac{u}{2\pi }𝑑x\frac{1}{K}:(_x\mathrm{\Phi }_c)^2:+K:(\mathrm{\Pi }_c)^2:,$$
(1)
and contains, in addition to the quadratic part, a Sine-Gordon (SG) umklapp term
$$_{um}=\frac{2g_3}{(2\pi a)^2}𝑑x\mathrm{cos}\sqrt{8\pi }\mathrm{\Phi }_c(x).$$
(2)
The displacement field $`\mathrm{\Phi }_c`$ and phase field field $`\mathrm{\Theta }_c`$ satisfy the usual commutation rules. All interaction effects are taken into account through the velocity $`u`$ and the Luttinger liquid parameter K (LLP). This obeys $`K=1`$ in the absence of interactions, $`K>1`$ for attractive interactions and $`K<1`$ for repulsive ones. It is convenient to perform the following canonical transformation: $`\mathrm{\Phi }_c\sqrt{K}\mathrm{\Phi }_c`$, $`\mathrm{\Pi }_c\mathrm{\Pi }_c/\sqrt{K}`$. Then, $`_{um}`$ reads:
$$_{um}=\frac{2g_3}{(2\pi a)^2}𝑑x\mathrm{cos}\sqrt{8\pi K}\mathrm{\Phi }_c(x).$$
(3)
The interaction is then absorbed in the argument of the cosine term. For repulsive interactions, this produces a quasiparticle gap $`\mathrm{\Delta }g_{3}^{}{}_{}{}^{1/(2\frac{n^2}{4\pi })}`$, with $`n^2=8\pi K`$.
*A strong (on-site) repulsion between electrons must inevitably result in $`K=1/2`$ close to half-filling*. Then, we start with such bare value of the LLP. Charge degrees of freedom (holons) should behave as *spinless fermions*. Through the Jordan-Wigner transformation in 1D (Appendix A), one can formally rewrite bosons operators in terms of spinless fermions $`(q=\pm )`$:
$$\psi _q(x)=\frac{\eta _q}{\sqrt{2\pi a}}:\mathrm{exp}i\sqrt{\pi }(\mathrm{\Theta }_c+q\mathrm{\Phi }_c)(x):.$$
(4)
This procedure is called refermionization. Klein factors $`\eta _q`$ are built to fullfill anticommutation rules between left $`()`$ and right $`(+)`$ movers. More crucially, for $`K=1/2`$, the SG-umklapp definitely creates fermionic kinks
$`{\displaystyle \frac{g_3}{(2\pi a)^2}}\mathrm{cos}\sqrt{4\pi }\mathrm{\Phi }_c`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }}{2\pi a}}\mathrm{cos}\sqrt{4\pi }\mathrm{\Phi }_c`$ (5)
$`=`$ $`i\mathrm{\Delta }\text{(}\psi _+^{}\psi _{}\psi _{}^{}\psi _+\text{)}.`$ (6)
Klein factors have been chosen as $`\eta _+\eta _{}=+i`$. The mass term always favors the pinning of the $`4k_F`$ charge-density wave (CDW), producing in our case a Kondo insulator, which is rather characterized by an even number of electrons per unit cell when adding the local moments. The holons disappear from the Kondo ground state. An excited holon from the valence band carries a wave vector $`q=2k_F`$. A pair holon-(anti)holon produces a kink of $`2\pi `$ in the superfluid phase $`\sqrt{\pi }\mathrm{\Theta }_c`$ at $`x`$, i.e. a kink of 2 in the charge current, that definitely corresponds to a pair of empty and doubly occupied sites with fermionic statistics. Fourier transforming, one finds:
$$_o=\underset{k}{}(uk)qc_{q,k}^{}c_{q,k}+\frac{g_3}{2\pi a}\text{(}c_{+,k}^{}c_{,k}+h.c.\text{)}$$
(7)
Using a so-called Bogoliubov transformation, this gives us the new energy spectrum:
$$E_\pm =\pm \sqrt{u^2k^2+\mathrm{\Delta }^2}.$$
(8)
At half-filling, the umklapp term gives us a semi-conductor picture of two bands. Therefore, to recover the physics of the KLM in the limit $`J_K/t1`$, we have only to take the quasiparticle gap:
$$\mathrm{\Delta }=g_3/2\pi a=3J_K/4.$$
(9)
Remarkably, charge properties of half-filled KLM’s remain qualitatively unchanged tuning the Kondo interaction from the strong-coupling to the weak-coupling limit (Section III and especially Appendix B).
*Here, at low temperatures, the Kondo interaction can be exactly rewritten as an effective umklapp.* However, $`g_3`$ (typically the quasiparticle gap) has then a less explicit dependence in $`J_K`$, due to the consequent renormalization of the backward Kondo coupling at half-filling.
### B Consequences of doping
To study the influence of hole doping on transport properties, we now shift the chemical potential at the top of the lower subband $`(\mu _{ch}=\mathrm{\Delta })`$. We get:
$`_o`$ $`=`$ $`{\displaystyle \underset{k}{}}u_ck(a_{+,k}^{}a_{+,k}a_{,k}^{}a_{,k})`$ (10)
$`u_c`$ $`=`$ $`{\displaystyle \frac{E}{k}}={\displaystyle \frac{u^2k_c}{\sqrt{(uk_c)^2+\mathrm{\Delta }^2}}}`$ (11)
The fermions (holons) $`a_{\pm ,k}`$ now refer to the partially occupied subband and linearizing the band structure near the new Fermi points produces $`k_c=\pi \delta `$ \[$`\delta =1n`$, being the hole doping\].
By doping such a semiconductor, low-energy properties should be well described through a Luttinger liquid Hamiltonian with a velocity $`u_c\delta /\mathrm{\Delta }`$ and $`K1/2`$. In the 1D KLM below half-filling, such a LL behavior has been checked numerically, for instance, in ref.. This leads to important consequences.
This produces a metallic phase with finite compressibility $`\kappa K/u_c`$ and Drude weight $`𝒟u_cK`$ at zero frequency. Very close to half-filling $`(u_c\delta 0)`$, then the Drude weight should vanish continuously (with a dynamical exponent $`z=2`$) and $`\kappa `$ diverges as $`\delta ^1`$. A recent Density Matrix Renormalization-Group (DMRG) study has confirmed for large $`J_K`$ parameters and $`T=0`$, that $`\kappa =\chi _c`$ (charge susceptibility) diverges as $`1/\delta `$ near half-filling. Moreover, at low temperatures, one then expects an exponential behavior of the d.c. conductivity.
The regular part of a.c. conductivity must have two distinct regimes: an $`w^3`$ absorption at small frequency, and a tail $`w^3`$ at large frequency.
To summarize, the commensurate-incommensurate transition of the 1D KLM belongs to the so called Prokovsky-Talapov class of universality. Now, let us push forward the duality with the Hubbard model in the *strong* $`U`$-limit, *below* half-filling.
The model can be mapped onto a Luttinger Hamiltonian as long as $`n1`$. Moreover, there is a remarkable duality by exchanging the two parameters $`n\delta `$. Indeed, the parameters $`K`$ and $`u_c`$ which determine the behavior of the system are known to remain unchanged under the exchange $`n\delta `$. In particular $`u_c=2t\mathrm{sin}(\pi n)=2t\mathrm{sin}(\pi \delta )`$ is the exact velocity of the charge excitations for infinite repulsion. Actually, this implies that there are two equivalent holon representations in the 1D KLM, when $`J_K+\mathrm{}`$. First, treating the introduced empty sites as holons leads to a gas of noninteracting spinless fermions in low density $`\delta `$. As shown precedingly, this picture is particularly accurate to capture transport properties at the commensurate-incommensurate transition. Second, one could also identify the Kondo singlets as holons (single-occupied states would be then localized moments). This produces a gas of noninteracting spinless fermions in large density $`n`$. This picture has been preferred to show the occurrence of ferromagnetism between unscreened local moments, with an exchange coupling $`t^2/J_K`$.
In the pure KLM, the unscreened localized spins polarize completely in order to lower the kinetic energy of mobile holons. Such a case is a concrete application of the Nagaoka’s theorem in 1D. The physics behind this is the finite extension of the screening cloud, which leads to the stability of the ferromagnetic state at finite densities not restricted to the single-electron case.
Note that including an antiferromagnetic Heisenberg exchange $`J`$ between local moments $`Jt^2/J_K`$ produces a so-called t-J model and then rather a paramagnet.
### C Interacting Kinks
Decreasing $`J_K`$ from infinity, results found precedingly on the commensurate-incommensurate transition become now available only for *very low* doping: Kinks become interacting below half-filling.
Most clearly, a repulsive interaction between the neighboring spinless fermions is introduced in the order of $`𝒪(t^2/J_K)`$. The situation becomes then similar to the infinite $`U`$ Hubbard model with nearest neighbor repulsion whose $`K`$ becomes smaller than 1/2. More importantly, the ferromagnetic phase of the pure KLM is very unstable decreasing $`J_K`$ due to the long-range overlap between Kondo singlets resulting in a certain magnetic disorder (see next section). Precisely, a first order phase transition in the intermediate coupling limit from the ferromagnetic state to a paramagnetic one has been pointed out recently. The latter is accompanied by a jump both in the magnetization curve and the value of $`K`$ (which curiously sharply decreases). This cannot be explained using the preceding analysis.
To conclude: the dependence of $`K`$ as a function of $`J_K`$ for weaker Kondo couplings remains difficult to handle *below* half-filling. A general trend, however, is that $`K`$ is always smaller than 1/3 producing a dominance of $`4k_F`$ charge oscillations, and then a so called Wigner crystal.
## III Weak-coupling Kondo transitions
Next, we subsequently stay at half-filling. We explore the perturbative limit, $`U/t1`$ and $`J_K/t1`$. Here, bosonization tricks allow us to explain the occurrence of Kondo insulators in commensurate systems when $`J_K/t1`$, in a simple way. The small-$`J_K`$ regime is of most interest, since correlation effects (for instance correlations between local moments, or between electrons) are most important there. For this regime, the expansion in $`t/J_K`$ is, of course, not convergent while variational methods are not well controlled. This is mainly because $`J_K=0`$ is a singular limit leading to a metallic behavior. To bosonize the Kondo interaction we need the bosonic representations for the conduction electron- and the localized spin operators. *Again, we limit the following study to the case of a single-channel conduction band*. In a very general form, that can be used whatever the strength of the Heisenberg interaction J between local moments or their spins S, these are given by:
$`\mathrm{\Psi }_\alpha ^{}`$ $`(x)`$ $`{\displaystyle \frac{\sigma _{\alpha \beta }}{2}}\mathrm{\Psi }_\beta (x)=(\stackrel{}{J}_R+\stackrel{}{J}_L)`$ (12)
$`+`$ $`{\displaystyle \frac{e^{2ik_Fx}}{2\pi a}}\text{Tr}\{\stackrel{}{\sigma }(\mathrm{\Phi }^{(1/2)}+\mathrm{\Phi }^{(1/2)})\}\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c`$ (13)
$`\stackrel{}{S}(x)`$ $`=`$ $`S\text{(}a\stackrel{}{L}(x)+(1)^x\sqrt{1(aL)^2}\stackrel{}{N}(x)\text{)},`$ (14)
where $`|\stackrel{}{L}|a1`$ is the quickly varying ferromagnetic component of the local magnetization, and:
$$\stackrel{}{L}(x)\stackrel{}{N}(x)=0.$$
(15)
In the weak-coupling limit, fluctuations in the $`2k_F`$ electronic spin operator are produced by the presence both, of doubly occupied and empty sites \[for small $`U`$, holons are described in Appendix B\] and of so-called domain walls \[spinons of the Heisenberg chain are introduced in Appendix C\]. The Hamiltonian for electrons will be also decomposed into a holon part \[the same as Eq.(1) with $`K=1Ua/\pi v_F`$\] and the usual spinon part. Moreover, away from half-filling, the continuum limit of the Hamiltonian only contains a marginal Kondo-coupling of the currents:
$$_m=\lambda _2[\stackrel{}{J}_R+\stackrel{}{J}_L]\stackrel{}{L}(x).$$
(16)
At half-filling, the $`2k_F`$ oscillation becomes commensurate with the alternating localized spin operator and the most prevalent contribution reads:
$$_{hf}\lambda _3\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c\text{Tr}\{\stackrel{}{\sigma }(\mathrm{\Phi }^{(1/2)}+\mathrm{\Phi }^{(1/2)})\}\stackrel{}{N}(x),$$
(17)
$`(\lambda _2,\lambda _3)J_K`$ being the usual forward and backward Kondo scattering processes. Notice the close analogy with the two-chain spin system.
### A Renormalization-group- and exact treatments
First, it is convenient to use the renormalization group method, expanding in powers of the coupling constant $`\lambda _3`$. The perturbation is divergent, and one can derive renormalization equations upon rescaling of the short distance cut-off $`aae^l`$:
$$\frac{d\lambda _3}{dl}=\text{(}\frac{3}{2}\frac{K}{2}\mathrm{\Delta }_N\text{)}\lambda _3,$$
(18)
*$`\mathrm{\Delta }_N1/2`$, being in the following the scaling dimension of the localized spin operator (see below, for specific cases).*
At half-filling, we expect $`_{hf}`$ to produce a gap for charge excitations \[whatever J or the spin S of local moments\]. The charge gap should follow the quasiparticle gap at the Fermi level
$$\mathrm{\Delta }\lambda _3^{\frac{2}{3K2\mathrm{\Delta }_N}},$$
(19)
for either sign of $`\lambda _3`$. In the weak coupling limit, a Kondo insulator is then formed due to the strong renormalization of the Kondo exchange $`\lambda _3`$.
It should be noted that the exchange coupling $`\lambda _3`$ and the usual umklapp $`g_3U`$ — driven by the on-site Hubbard term — affect the Luttinger parameter $`K`$ in a symmetric way:
$$\frac{dK}{dl}=uK^2\text{(}C_0g_3^2+C_1\lambda _3^2\text{)}J_o(\delta (l)a).$$
(20)
$`J_o`$ is the Bessel function, and $`C_0`$, $`C_1`$ are constants. At half-filling we must take $`J_o(0)=1`$.
When $`\lambda _3=0`$, the usual umklapp process is known to produce a Mott transition at half-filling. It occurs for a critical value of K which is equal to one, i.e. at the noninteracting point. As soon as $`\lambda _30`$, the physics becomes ruled by the Kondo coupling, producing a Kondo transition for a critical Kondo coupling $`J_K^c=0`$. The usual umklapp process can be neglected for small $`U`$ because it (only) scales marginally to strong couplings.
Second, at low temperatures $`T\mathrm{\Delta }`$ the holons, becoming massive, also change of statistics (semions $``$ fermions). Charge carriers behave as fermionic kinks due to the strong renormalization of $`J_K`$, and as said before the backward Kondo interaction can be exactly rewritten as an “effective” umklapp process (see Appendix B) — remarkably, whatever $`\mathrm{\Delta }_N1/2`$. The charge gap will have, of course, dramatic consequences for the physical properties. First, this implies a long-range order in the $`\mathrm{\Phi }_c`$-field. Indeed, we have to take formally, $`K^{}=0`$, for $`T\mathrm{\Delta }`$. Then, at the Kondo transition, there is a finite jump in the compressibility and Drude weight. This will be analyzed in more details in the next section.
### B Spin properties for several $`\mathrm{\Delta }_N`$’s
Before investigating transport properties in great details, it is maybe appropriate to review several interesting Kondo spin-liquid phases occurring in the weak coupling limit, for particular values of $`\mathrm{\Delta }_N`$. Details of the technique can be found in Appendix C.
#### 1 $`\mathrm{\Delta }_N=1/2`$: Heisenberg-Kondo lattice for S=1/2
Let us start with the so-called Heisenberg-Kondo lattice, where the spins of the array are coupled through a consequent antiferromagnetic coupling exchange J of the order of t. Here, one can introduce spinon-pairs:
$$\stackrel{}{N}=\text{Tr}[(\mathrm{\Theta }^{(1/2)}+\mathrm{\Theta }^{(1/2)})\stackrel{}{\sigma }].$$
(21)
For weak Hubbard interactions (for a parallel, see formula B3), one gets also: $`J_K<\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c(x)>=\mathrm{\Delta }`$. Therefore, the backward Kondo interaction is transformed into:
$$_{hf}=\mathrm{\Delta }\text{Tr}𝚽^{(\mathrm{𝟏})}3\mathrm{\Delta }ϵ.$$
(22)
By analogy to usual spin ladder systems, the spin spectrum is composed of massive triplet excitations \[described by $`\text{Tr}𝚽^{(\mathrm{𝟏})}`$\] with a mass $`m_t=\mathrm{\Delta }`$, and of a high-energy singlet branch at $`m_s=3\mathrm{\Delta }`$. We expect the spin gap to decrease for any appreciable difference between J and t.
#### 2 $`\mathrm{\Delta }_N=0`$: Weakly coupled S=1/2-local moments
Precisely, let us now study the opposite limit where the local moments are weakly coupled i.e. $`JJ_K`$. At short distances, the RKKY interaction — that displays a very small decay in 1D — produces a perfectly static staggered potential with:
$$\stackrel{}{N}=\text{constant}.$$
(23)
The electrons are then subject to Bragg scattering. This opens a quasiparticle gap (linear in $`J_K`$). On the other hand, these screen away the internal field before a true magnetic transition takes place.
Then, there are still some triplet states in the quasiparticle gap because the SU(2)-spin symmetry is restored at long distances. The resulting spin gap becomes considerably smaller than the charge gap.
First, if J is not so far from $`J_K`$, these spin degrees of freedom are described in terms of an O(3) non-linear sigma model without the topological term. The implicit breaking of conformal invariance rescales the spin gap as:
$$\mathrm{\Delta }_S\mathrm{\Delta }\mathrm{exp}\pi \text{(}\frac{|Jt|}{t}\text{)}$$
(24)
Single electron excitations in the interval $`\mathrm{\Delta }_S`$ and $`\mathrm{\Delta }`$ can be viewed precisely as polarons — i.e. bound states of an electron with a *kink* of the vector $`\stackrel{}{N}`$.
Second, if J tends to 0, the condition for the presence of kink in the local staggered magnetization operator now breaks down. Spin degrees of freedom rather behave as follows. At not too long distances, half of the (electronic) spinon field begins to fluctuate due to the restoration of the SU(2)-symmetry, contributing to a new chiral phase of the same universality class as the 2D Ising model. At long distances, due to strong fluctuations in the spin array, the (electronic) spinon field cannot be pinned anymore by the backward Kondo coupling — i.e. $`|\stackrel{}{N}|=0`$. Then, spin flips should contribute. The system may scale to a perfect singlet ground state at a very low energy scale:
$$\mathrm{\Delta }_S\mathrm{\Delta }\mathrm{exp}2\pi v_F/\lambda _2\mathrm{\Delta },$$
(25)
typically the single-site Kondo temperature, in agreement with numerical calculations.
To conclude this part: For spin-1/2 local moments, the ground state is always a spin singlet. Furthermore, for small $`J_K`$’s, the antiferromagnetic correlation length is quite large producing disordered quantum fluids. It is already increased by an appreciable difference $`(Jt)`$, and becomes huge for J=0.
#### 3 $`\mathrm{\Delta }_N=3/8`$: Underscreened S=1-chain
Now, we consider another generic case where the local spins form a so-called S=1 Takhtajan-Babujan chain. Details of the calculations can be found in ref.. These can be extended to the case of a spin S-integrable chain $`(|12S|0)`$ — with $`C_v/L=2ST/(1+S)`$.
In the sense of critical theories, this model can be parametrized by a conformal anomaly $`C=3/2`$. In the continuum limit, this spin-1 chain has then only zero sound triplet excitations — described by three massless Majorana fermions — or rather magnons. The staggered magnetization $`\stackrel{}{N}`$ has now dimension $`\mathrm{\Delta }_N=3/8`$. Consult Appendix C3. The Kondo interaction grows to strong coupling, producing a quasiparticle gap $`\mathrm{\Delta }J_{K}^{}{}_{}{}^{8/5}`$.
Then, a new phenomenon arises: The occurrence of singlet bound states can be achieved only due to the “fractionalization” of each spin triplet excitation onto two spins-1/2. On each rung, (only) one is strongly coupled to the conduction band. This produces the same contribution as in Eq.(22), and then *optical magnons* in the system. From a general point of view, Luttinger liquids coupled to an active insulating environment via a Kondo-like coupling yield a spin gap. But, deconfined spinon-pairs still subsist in the ground state because the local moments are underscreened. This leads to an anomalous optical conductivity at low frequencies, and then a pseudogap phase. Starting with small J’s (the Haldane gap is irrelevant), predictions from the non-linear sigma model for S=1 give the same conclusion.
### C Role of electron-electron interactions
The present study shows that the smallest amount of Kondo potential $`\lambda _3`$ produces a Kondo insulator at least for sufficiently large length scales. Then, the backward Kondo scattering can be naturally rewritten as an effective *strong* umklapp process, such that the original umklapp term — again, driven by the Hubbard interaction — can be neglected.
One can check that the charge gap increases with (enhancing) $`U`$ — or decreasing the LLP. If the Hubbard repulsion is very large $`U/t1`$ one can expect to see a crossover from a Kondo insulator to a usual spin ladder system. Here, the insulating transition should be rather driven by U only, and then we have to take formally $`<\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c(x)>=\sqrt{U}=\text{constant}`$ and $`K=0`$ in the spin gap equation. For instance, for $`\mathrm{\Delta }_N=1/2`$, we recover previous results and for instance the spin gap $`\mathrm{\Delta }_sJ_K`$ (see Appendix C). *Another interesting feature is that the Kondo insulating state should even persist in presence of attractive interactions*.
A physical mechanism for the generation of attractive interactions is electron-phonon interaction. A renormalization group treatment of the important spin-backscattering term $`g_1`$ (given in Appendix D) predicts, for $`J_K=0`$, the existence of a spin gap
$$\mathrm{\Delta }_{ss}\frac{v_F}{a}\mathrm{exp}\text{(}\frac{\pi v_F}{\left|U\right|a}\text{)}$$
(26)
Remarkably, localization by magnetic impurities remain the most prominent phenomenon as long as $`K<K_c=32\mathrm{\Delta }_N`$. When $`K=K_c`$, the Kondo gap tends to zero and then, a superconducting-like ground state takes place. When $`KK_c`$, correlation functions for the singlet superconducting pairing field are considerably increased in the low-temperature phase $`(T\mathrm{\Delta }_{ss})`$ and the charge sector remains critical, leading to a perfect superfluid. This produces a coexistence between a Luther-Emery liquid and an insulating part characterized by strong antiferromagnetic fluctuations. The pairing also promotes $`2k_F`$ CDW fluctuations that are not sensitive to magnetic impurities.
To conclude: Close to the superconducting transition $`(K<K_c)`$, as a remnant of s-wave superconductivity (and CDW fluctuations) one expects that the electron liquid is less localized than for repulsive interactions. Let us now investigate transport properties and comment this point in more details.
## IV Transport properties
Transport in such commensurate systems is very interesting because the Kondo process $`\lambda _3`$ provides an important relaxation mechanism for electrons. Next, we will assume that the coupling $`\lambda _3`$ is sufficiently *weak* that some perturbative calculation of the conductivity as a function of $`\lambda _3`$ can be performed. The conductivity itself does not have a regular expansion in $`\lambda _3`$. A way out is provided by the memory function formalism.
If the system is a normal metal with finite dc-conductivity, one can define
$$\sigma (w)=\frac{2iKu}{\pi }\frac{1}{w+M(w)}$$
(27)
and $`M(w)`$ is the meromorphic memory function. This formula is well-suited for an “infinite” system: *Next, we are not interested in reservoir effects*.
### A Memory function approximation
The calculation of this function can be carried out perturbatively to give at the lowest order
$$M(w)=\frac{(F;F_wF;F_{w=0})/w}{\chi (0)}$$
(28)
At zero frequency, one gets the retarded current-current correlation function:
$$\chi (0)=2uK/\pi .$$
(29)
The factor 2 comes from the two colors of spin which both contribute to transport properties. The symbol $`;`$ designs a retarded correlator computed in the absence of magnetic impurities<sup>*</sup><sup>*</sup>*As long as $`\lambda _3`$ remains small, one can neglect all effects of self-adjustments of the ground state to the Kondo potential. and $`F=[J_c,H]`$, $`J_c`$ is the charge current. The F operators take into account the fact that the charge current is not a conserved quantity.
For frequencies $`wM(w)`$ and $`T0`$, one gets:
$$\sigma _{ac}(w)=\frac{i\chi (0)}{w}[1\frac{M(w)}{w}]$$
(30)
The regular part in the delocalized phase reads:
$$\sigma _{reg}(w)M(w)w^2.$$
(31)
The expression (28) does not necessarily remain valid at very low frequencies, even for finite temperatures. Its validity at low frequencies implicitly assumes in a self-consistent way that the d.c. conductivity behaves as:
$$\sigma _{dc}(T)=\frac{i\chi (0)}{M(T)},$$
(32)
and $`M(T)`$ is related to some relaxation time, i.e.
$$M(T)\tau _{rel}^1.$$
(33)
In the macroscopic pure system, one gets $`\sigma _{dc}(T)+\mathrm{}`$. Here, $`\tau _{rel}`$ should correspond to the effective elastic time between two magnetic diffusions. Of course, it will be highly non-universal and temperature-dependent (since we are in a ballistic transport). This should legitimize the memory function approximation, which gives in general good results as long as we stay far away from an insulating region. This approach, of course, will break down when $`T=\mathrm{\Delta }`$ i.e. when the effect of $`\lambda _3`$ on the ground state becomes strong: it opens a spin- and charge gap changing drastically the nature of the elementary excitations. At low temperatures, the source of large resistance in such magnetic systems comes from prominent quantum scattering which merely destroys the coherence or the propagation of the accelerated preexisting excitations, producing Kondo localization.
Interestingly, here the quantum correlation between spins of the array — driven by $`\mathrm{\Delta }_N`$, that plays a crucial role in the classification of the Kondo localizations (e.g. gap equation, spin spectrum), will explicitly appear in the computation of the high-temperature d.c. conductivity. The bosonization scheme gives:
$$J_c\sqrt{2}uK\mathrm{\Pi }_c(x)𝑑x.$$
(34)
Using the definition of $`_{hf}`$, the behavior of the memory function as a function of temperature can be easily obtained. We find:
$$T_\tau F(\tau )F(0)dx\lambda _{3}^{}{}_{}{}^{2}\text{(}\frac{1}{x^2+(u\tau )^2}\text{)}^{\sqrt{1+K+2\mathrm{\Delta }_N}}$$
(35)
The function in the integral describes the amplitude that a spinon and a holon recombine to give an electron at the point $`x`$ and time $`\tau `$, that is immediately diffused by an impurity. Fourier transforming, one finds:
$$M(w)w^{(K2+2\mathrm{\Delta }_N)},$$
(36)
and the same behavior as a function of temperature. In the delocalized regime $`T\mathrm{\Delta }`$ or $`w\mathrm{\Delta }_c=2\mathrm{\Delta }`$, the temperature and frequency dependent conductivity is
$$\sigma _{dc}(T)T^{2\mu },\sigma _{ac}(w)w^{\mu 4},\mu =K(T)+2\mathrm{\Delta }_N.$$
(37)
For more details concerning the method, consult ref..
Away from the transition, one may consider the parameter $`K(T)`$ as a constant $`K`$. We stress on the fact that dressing of the Kondo exchange by the other interactions results in a nonuniversal power-law dependence in the delocalized regime. Again, the whole perturbative scheme breaks down when $`\lambda _31`$, at a length scale which corresponds to the localization length of the system. A reasonable guess of the temperature dependence below this length scale is an exponentially activated conductivity because current-current correlation functions now decrease exponentially in time.
As can been seen from Fig.4, there is a maximum in $`\sigma _{dc}`$ as a function of T for weak attractive interactions $`(22\mathrm{\Delta }_N<K<K_c=32\mathrm{\Delta }_N)`$, occurring when the thermal coherence length is of the order in magnitude of the localization length $`T\mathrm{\Delta }`$. This maximum can be well understood as a remnant of s-wave superconductivity in the localized phase that cannot be easily pinned by magnetic impurities. Remind that the system is a very good conductor until one reaches the localization temperature $`\mathrm{\Delta }`$.
A similar phenomenon has been predicted in a disordered two-leg Hubbard ladder system with attractive interactions. In that case, ‘$`2k_F`$’ charge density fluctuations are forbidden due to the spin gap, and ‘$`4k_F`$’ charge density fluctuations, that aim to arise for repulsive interactions, become very small for attractive ones: this is an important consequence of the s-wave scenario in two-leg ladder systems. The nonmagnetic disorder can become efficient only at very low energy. For a comparison with the single Hubbard chain problem, see next subsection.
For repulsive interactions, we can check that there is no remnant of the original umklapp term — again, driven by the Hubbard interaction — which is known to produce $`\sigma _{dc}T^1`$ (with $`K1`$) in the delocalized regime, and then a maximum in $`\sigma _{dc}`$ in the entrance of the localized phase. In the case of electron-electron interactions, the relaxation time should result from the coupling of electrons with a thermal bath which, strictly speaking, is needed for the system to reach thermal equilibrium. In such case, the relaxation time is rather associated to phase-breaking or inelastic processes. In particular, note that $`\sigma _{dc}\tau _{in}`$ (see Appendix D). Furthermore, as long as a spinon is not scattered (as it is the case in the absence of impurities), it may combine locally with another holon to form an electron and the material is still a very good conductor. This is a consequence of spin-charge separation.
Most clearly, for repulsive interactions, the ground state of the Hubbard chain is almost a spin-density wave (SDW) (with a power-law decay of the correlation functions) whose charge density is uniform. Such a ground state couples very strongly to magnetic impurities leading to prominent elastic processes in the electron liquid. It is therefore not surprising that for repulsive interactions the main source of resistance for the electron liquid is due to the prominent scattering of a pair spinon-holon (electron) by a magnetic impurity.
Furthermore, the conductivity is found to decrease monotonically with temperature, even at high temperature. This noteworthy renormalization of exponents and the faster decay of the conductivity for repulsive interactions is a perfect signature of Kondo localization. Again, note the nonuniversality of exponents due to the factor $`\mathrm{\Delta }_N`$ which depends on the spin S of local moments and on the strength of the Heisenberg interaction between them. To summarize the main point: Now, we are really able to allege that for repulsive interactions the electron liquid is more localized by the magnetic impurities than for attractive interactions, both because of the scale of localization and because of the d.c. conductivity temperature dependence.
### B Duality to an effective Gaussian disorder
It should be noted that these results can be reproduced defining the (Kubo) d.c. conductivity as:
$$\sigma _{dc}(T)T^1/D_m(T)\text{and}\text{ }D_m(T)=\lambda _3(T)^2.$$
(38)
$`n(T)=1/T`$ denotes the electronic density at a temperature T. We must stop renormalization procedure at the thermal length $`L=l_{in}=v_F/T`$ at which inelastic and decoherent effects take place. By comparison with the case of a nonmagnetic Gaussian disorder,
$$D_m(T)=l_{in}/l_eT^{(3\mu )}.$$
(39)
The exponent $`\mu `$ has been defined in Eq.(37). In KLM’s, $`D_m`$ will be identified as the magnetic disorder parameter. For more details, consult Appendix D. This definition traduces that at high temperatures we have a perfect duality to a model of a 1D electron liquid submitted to a Gaussian correlated spin disorder, with an effective exponent $`\mu =K(T)+2\mathrm{\Delta }_N`$ for the electronic spin density-spin density correlation function. In the case of a Gaussian correlated nonmagnetic disorder, for SU(2) symmetry and repulsive interactions one rather expects:
$$\sigma _{dc}(T)T^{2\widehat{\mu }}\text{with}\text{ }\widehat{\mu }=1+K(T).$$
(40)
Far away from the localization energy scale i.e. for very small disorder, one can take $`\widehat{\mu }=1+K`$. Note that to couple to nonmagnetic disorder one has to distort the SDW and make a fluctuation of the charge density, a process which costs an energy of order U. Therefore, for repulsive interactions, the non-magnetic disorder effect must be inevitably weaker than for attractive interactions where $`2k_F`$ CDW fluctuations can get very easily pinned by nonmagnetic impurities. Starting with the Hubbard model and *small U*, the d.c. conductivity found above is then not completely correct. The reason is that the backscattering spin term $`g_1`$ couples to the non-magnetic disorder, and then it cannot be ignored (see Appendix D). At intermediate length scales, this results rather in:
$$\sigma _{dc}(T)T^{Ua/\pi v_F}.$$
(41)
Now, the maximum in the d.c. conductivity in the entrance of the Anderson glass phase occurs for repulsive interactions. The situation becomes opposite to that found with an array of magnetic impurities.
In general, for repulsive interactions one gets $`\mu <\widehat{\mu }`$, revealing that Kondo localizations may still take place in weakly disordered 1D Kondo lattice models. Especially when $`\mathrm{\Delta }_N=0`$, we have a constant uniaxial Kondo potential at high temperatures
$$\overline{V(x)}=\lambda _3,\overline{V(x)V(0)}=D_m,$$
(42)
and $`\mu =K(T)\widehat{\mu }`$. There is no quantum fluctuation. The long-range coherence of the spin array produces inevitably Kondo localization. On the other hand, adding explicitly a dominant antiferromagnetic direct exchange between local moments should produce a stronger competition between Anderson- and Kondo localizations, even if the ground state is always of Kondo type for prominent repulsive interactions. This comes from the facts that $`2k_F`$ CDW fluctuations are not completely forbidden at high temperatures for weak on-site repulsive interactions, and that the Kondo potential follows:
$$\overline{V(x)}=0,\overline{V(x)V(0)}=D_m/x^{2\mathrm{\Delta }_N}.$$
(43)
It is now short-range correlated, and we have $`\mathrm{\Delta }_N=3/8`$ (Takhtajan-Babujan S=1 chain), or $`\mathrm{\Delta }_N=1/2`$ (Heisenberg S=1/2 chain). The average is explicitly performed on quantum fluctuations (after normal ordering). Below, we will investigate the case of a Gaussian correlated magnetic disorder (subsection E): again it will lead to the same conclusion. Using Appendix D, one can easily check that repulsive interactions tend to make the nonmagnetic disorder less relevant and decrease localization. This has dramatic consequences on the effects of interactions on persistent currents. Comparisons with the renormalization of charge stiffness in 1D KLM’s will be made later in subsection D.
### C Charge incompressibility with $`\sigma (w,k=\pi )0`$
In the following, we present a new class of Kondo insulators which is (charge) incompressible, but yields only a pseudo gap in the (charge) optical conductivity at a large wave-vector $`k=\pi `$.
On the one hand, the Kondo transition at half-filling produces an incompressible system because $`\kappa K^{}=0`$. On the other hand, ordinary Kondo insulators \[where local moments are characterized by $`S=1/2`$\] have no charge conductivity for low frequencies. First, the charge current $`J_c`$ given by the expression (34) vanishes. Second, the presence of a spin gap in the system produces optical magnons, that doesnot allow any anomalous charge current at large wave-vectors (see Appendix C).
Now, let us investigate the case where local moments form a Takhtajan-Babujan S=1 chain. As shown precedingly, free spinons remain in the ground state of the system due to the ‘underscreening’ of the local impurities. This implies that the optical charge-conductivity would reveal no gap at low frequency and $`k=\pi `$.
Spinons are governed by an extra term in the action \[we use the duality: $`\mathrm{\Phi }^{(1/2)}=i\stackrel{}{\sigma }𝚽`$\],
$$\mathrm{\Gamma }(i\stackrel{}{\sigma }𝚽)=\frac{i}{8\pi }𝑑x𝑑\tau ϵ_{\mu \nu }\text{(}𝚽[_\mu 𝚽\times _\nu 𝚽]\text{)}=i\pi Q$$
(44)
where $`Q=1=𝑑xq(x)`$ is the associated topological charge. From this (topological) charge, one can define an anomalous current density,
$$\overline{\psi }(i\sigma ^z)\psi =𝒥_S=ϵ_{\mu \nu }\text{(}𝚽[_\mu 𝚽\times _\nu 𝚽]\text{)},$$
(45)
which has the scaling dimension 5/2. In this way, the charge- and the spin sector are related to each other. Then, the pair correlation function of $`J_S`$ induces an unusual feature in the optical conductivity at a large wave-vector $`k=\pi `$. The anomalous current is nothing but the ordinary current at $`k=\pi `$. We find a retarded current-current correlation function $`\chi (w)iw^3`$ and then an optical conductivity,
$$\sigma (w,k=\pi )=\frac{i}{w}\chi (w)w^2.$$
(46)
We obtain a new class of incompressible Kondo systems, with no gap in the optical conductivity (Fig.5). Similar conclusions can be reached in the case where the direct exchange between local moments is very weak.
It should be noticed that a new phase of disordered commensurate insulators yielding “similar” features, has been pointed out recently. The corresponding phase with $`\kappa =0`$ and $`\sigma (w,k=0)w^2`$ results from a strong competition between Mott- and Anderson localization which leads to unusual excitonic effects. This requires some interactions of finite extent.
### D Renormalization of the Drude peak
In addition to the regular frequency dependence of the a.c. conductivity found precedingly, one can compute the charge stiffness $`𝒟`$ of the macroscopic system, which measures the strength of the Drude peak (occurring at zero frequency):
$$\sigma _{ac}(w)=𝒟\delta (w)+\sigma _{reg}(w).$$
(47)
This comes from the first term in Eq.(30). For fermions with spins, the conductivity stiffness at length $`l`$ can be obtained using:
$$𝒟(l)=2u(l)K(l),$$
(48)
with $`𝒟(l=0)=2uK=2v_F`$. Again the factor of two comes from the fact that there are twice degrees of freedom compared to the spinless fermionic case. The velocity $`u`$ obeys a recursion law of the type:
$$\frac{du}{dl}=C_2u^2KD_m(l).$$
(49)
$`C_2`$ is a constant. From Eqs.(20),(49) one can easily obtain the renormalization group equation for $`𝒟`$:
$$\frac{d𝒟}{dl}=CD_m(l),$$
(50)
and C is a constant. Using the flow of $`D_m(l)`$ given explicitly in Appendix D, one obtains:
$$\frac{d𝒟}{dl}=CD_m(0)e^{(3\mu )l}.$$
(51)
By stopping the renormalization procedure at a length $`L=e^l=v_F/T`$, it follows:
$$𝒟(L)𝒟(1)=C\frac{D_m(1)}{(3\mu )}\text{[}L^{(3\mu )}1\text{]}.$$
(52)
We can notice that for repulsive interactions the Drude peak fastly decreases with $`L`$, that is also typical of a Kondo localization. For attractive interactions, the Drude peak decreases more slowly and the renormalization of $`𝒟`$ stops completely as soon as $`\mu =3`$ that coincides with $`K=K_c`$. This is in agreement with the fact that for $`KK_c`$ the system behaves as a perfect superfluid at low energy. In the localized phase $`(T\mathrm{\Delta })`$, the opening of the gap produces formally $`K^{}=0`$ and then the Drude weight vanishes. Using Eq.(52) for $`L\mathrm{\Delta }^1`$, one can therefore deduce that there is (also) a jump in the Drude weight at the transition. Using ref. and replacing $`\delta `$ by $`(T\mathrm{\Delta })`$, one then finds that the dynamical exponent is $`z=1`$, that is specific of a Kosterlitz-Thouless transition. Again note the close parallel with the calculation of the charge stiffness in presence of a Gaussian disorder:
$$𝒟(L)𝒟(1)=\widehat{C}\frac{D(1)}{(3\widehat{\mu })}\text{[}L^{(3\widehat{\mu })}1\text{]}.$$
(53)
Nonetheless, physical consequences are very different. For small U one gets:
$$(3\widehat{\mu })=1\frac{Ua}{\pi v_F}$$
(54)
For finite non-magnetic disorder, one finds that the charge stiffness of the system for repulsive interactions is larger than the one of the system for attractive interactions. Again, SDW fluctuations that are prominent for repulsive interactions tend to make the quenched disorder less relevant and decrease localization. The charge density tends to be rather homogeneous. Conversely, for attractive interactions, $`2k_F`$ CDW fluctuations can get very easily immobilized by non-magnetic impurities.
### E Randomly distributed magnetic defects
For completeness, we now make links with the case where magnetic impurities would be randomly distributed. To simplify at maximum the discussion, we consider here that these exercise a unixial random magnetic field along the z-axis. One gets the Hamiltonian (see Appendix D):
$$H_{m.dis}=𝑑xV(x)\rho _s^z(x),$$
(55)
$`V(x)`$ (complex) is assumed to be Gaussian correlated, resulting in
$$\overline{V(x)}=0,\overline{V^{}(x)V(0)}=D_m\delta (x).$$
(56)
The forward scattering potential, that does not affect transport properties, will be omitted. Now, we apply standard renormalization-group methods for one-dimensional disordered systems. In the Abelian representation, the random Kondo potential reads:
$$H_{m.dis}=𝑑xV(x)\mathrm{sin}\sqrt{2\pi }\mathrm{\Phi }_s(x)\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c(x).$$
(57)
Here, random magnetic impurities are not explicitly linked to each other (the Kondo potential is ‘on-site’ correlated). Now, we show that preceding conclusions remain unchanged: The electron liquid is more localized by magnetic impurities for repulsive interactions than for attractive ones. We follow exactly the same scheme as in Appendix A of ref.. From a perturbative expansion in the magnetic disorder parameter $`D_m`$ and the spin backscattering $`g_1`$, the result is:
$$\frac{dD_m}{d\mathrm{ln}T}=(3K_sK+\frac{g_1}{\pi v_F})D_m.$$
(58)
The main difference with non-magnetic disorder comes from the sign (plus and not minus) of the last term, coupling interactions and Kondo potential. This is because electronic spin degrees of freedom are linked to the random field via a sinus term (and not a cosine). For a simple sight, consult Appendix E. As a consequence, we still have a strong competition between localization of the electron liquid by (disordered) magnetic potential and superconductivity: For $`g_1\mathrm{}`$, the flow of $`D_m`$ scales to zero. For small U, one finds:
$$\frac{dD_m}{d\mathrm{ln}L}=(1+\frac{Ua}{\pi v_F})D_m.$$
(59)
and,
$`{\displaystyle \frac{dK}{dl}}`$ $`=`$ $`C_1uK^2D_m,`$ (60)
$`{\displaystyle \frac{dK_s}{dl}}`$ $`=`$ $`C_3K_{s}^{}{}_{}{}^{2}D_m.`$ (61)
$`C_3`$ is a constant. In the infra-red region, quantum coherence between impurities definitely appears and then $`K^{}`$ and $`K_s^{}`$ approach zero. The spin backscattering term $`g_1`$ remains small. Furthermore, the $`2k_F`$ electronic spin density wave will be weakly pinned by the random magnetic field — induced by local moments — resulting in a glassy- and presumably a disordered Ising phase.
Far in the delocalized regime, one can equate the exponent $`\mu `$ to $`1+K`$. The result is then:
$$\sigma _{dc}T^{Ua/\pi v_F}.$$
(62)
At short length scales, the impurities can be considered as *independent* (see next section). One can also notice the close correspondence with the so-called Heisenberg-Kondo lattice model with $`\mathrm{\Delta }_N=1/2`$. This fact certainly traduces that in the delocalized regime there is no fundamental difference between on-site- and (very) short-range correlated Kondo potentials. Likewise, a maximum in the d.c. conductivity curve again should take place in the entrance of the Kondo (glass) phase for attractive interactions. For an explicit comparison with non-magnetic impurities, consult Eq.(41). Finally, one can also check that repulsive interactions help in reducing the conductivity stiffness, while attractive interactions reduce the decrease of conductivity stiffness by magnetic disorder
$$𝒟(L)𝒟(1)=C\frac{D_m(1)}{(3\mu )}\text{[}L^{(1+\frac{Ua}{\pi v_F})}1\text{]}.$$
(63)
On the other hand, the stiffness $`𝒟`$ generally provides a nice measure of the persistent current in a mesoscopic ring induced by a small magnetic flux. Therefore, the above results could have some applications in nearly one-dimensional mesoscopic rings (of size L: The electron has to stay coherent along the whole ring) with *prominent* magnetic defects: for example, repulsive interactions should decrease persistent currents. Extensions of such model will be considered elsewhere.
## V Links with the single impurity case
Let us now compare with the case of a single magnetic impurity in a Luttinger liquid. At half-filling, here usual umklapps produce an Heisenberg chain. The resulting Kondo effect has been studied in ref.. Away from half-filling, the Kondo interaction gives the same two contributions as in Eqs.(16),(17) \[forward and backward scatterings\]. Since scattering occurs only on a finite part of the sample, the conductance is the most appropriate way to describe transport.
### A New renormalization flow
For weak $`J_K`$’s, one can use the same renormalization method expanding in the interactions $`\lambda _2`$ and $`\lambda _3`$. One obtains the RG equations
$`{\displaystyle \frac{dK}{dl}}`$ $`=`$ $`0`$ (64)
$`{\displaystyle \frac{d\lambda _2}{dl}}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi v_F}}\lambda _{2}^{}{}_{}{}^{2}+{\displaystyle \frac{K}{2\pi v_F}}\lambda _{3}^{}{}_{}{}^{2}`$ (65)
$`{\displaystyle \frac{d\lambda _3}{dl}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}(1K)\lambda _3+{\displaystyle \frac{1}{\pi v_F}}\lambda _2\lambda _3.`$ (66)
First, note that the first term in the third equation is seemingly different from the one for the lattice case: it comes from the fact that the impurity only acts at $`x=0`$, leaving only a double integral over time and producing $`\mathrm{\Delta }_N=0`$. The single-impurity gap (or the single-site Kondo temperature) now scales as
$$\mathrm{\Delta }\lambda _{3}^{}{}_{}{}^{\frac{2}{1K}}\lambda _{3}^{}{}_{}{}^{\frac{2\pi v_F}{U}}.$$
(67)
For $`U0`$, marginal operators produce the usual Kondo temperature \[identical to Eq.(25)\]. The second difference compared to the finite density of impurities is the absence of renormalization of the bulk exponent $`K`$, which clearly displays that there is no source of Drude resistivity (nor scattering effects) far away from the impurity spin. Thus, the two $`\beta `$-functions (for $`\lambda _2`$ and $`\lambda _3`$, respectively) become strongly coupled due to the presence of the extra term $`𝒪(\lambda _{3}^{}{}_{}{}^{2})`$ in the second equation. To show the existence of such a term, one can for instance utilize current algebera techniques for the operators. The charge field satisfies the operator product expansion (OPE) of a U(1) Gaussian model,
$$:\mathrm{exp}(i\sqrt{2\pi }\mathrm{\Phi }_c(\tau ))::\mathrm{exp}(i\sqrt{2\pi }\mathrm{\Phi }_c(0)):\frac{1}{u\left|\tau \right|^K}$$
(68)
One can also use the OPE of spinon pairs:
$$\text{Tr}[\mathrm{\Phi }^{(1/2)}(\tau )\sigma ^a]\text{Tr}[\mathrm{\Phi }^{(1/2)}(0)\sigma ^b]iϵ^{abc}[J_L^c+J_R^c].$$
(69)
All these operators act on $`x=0`$, and $`\tau `$ is close to 0. The local impurity spin obeys a usual SU(2) Lie algebra:
$$[S^a,S^b]=iϵ^{abc}S^c.$$
(70)
Exploiting Eqs.(68),(69) and (70) and the fact that $`\left|\tau \right|^Kdl`$, one inevitably finds the extra term in the second equation. We have also applied the equality $`uK=v_F`$, from the LL theory in presence of Galilean invariance (Appendix A). Let us insist on the fact that in the KLM at half-filling the exponent $`K^{}`$ tends to zero at low temperatures so that forward and backward Kondo exchanges are independent also in the infra-red region. It is not the case in the single impurity case; the low-temperature behavior is well controlled by $`\lambda _2=\lambda _3+\mathrm{}`$.
### B Transport and fixed points
At high temperatures, the only effect of the interaction $`\lambda _3`$ is therefore to change the Born amplitude of magnetic scattering. The conductance is given by the effective scattering at the scale $`L=\mathrm{exp}l1/T`$. Integrating out (64), one gets a Landauer-type conductance:
$$G_0G(T)=\delta G(T)(T)=D_m^oT^{\stackrel{~}{\mu }2},$$
(71)
with:
$$\stackrel{~}{\mu }=K+1.$$
(72)
$`G_0=2e^2K/h`$ is the conductance of the pure wire (again, without any reservoir-effect). Applying a generalized Fermi Golden’s rule at $`T0`$, one can check that $`(T)D_m^o(T)=\lambda _3(T)^2`$ is the resistance produced by a sole impurity. It obeys:
$$\frac{dD_m^o}{dl}=(2\stackrel{~}{\mu })D_m^o.$$
(73)
This produces the same thermal laws as a pointlike nonmagnetic defect at $`x=0`$ (see, for instance and references therein), i.e.
$$G_0G(T)T^{K1}.$$
(74)
On the one hand, attractive interactions $`(K>1)`$ favor the formation of singlet superconducting pairs that can completely suppress the back flow at zero temperature. Here, the impurity becomes unscreened for “vanishing” attractive interactions or at least as soon as $`U>J_K`$. In comparison, remember that the coherence between localized spins in the lattice problem should prevent a superconducting-like ground state until attractive enough electron-electron interactions, i.e. $`KK_c=32\mathrm{\Delta }_N`$. Furthermore, the Drude weight or superfluid stiffness is already large in the weak attractive patch $`(2v_F)`$.
One the other hand, if the backward and forward Kondo couplings are relevant i.e. for $`K<1`$, the weak coupling renormalization group scheme ceases to be valid as soon as the Born amplitude $`\lambda _{3}^{}{}_{}{}^{2}`$ is of order one (i.e. for $`T\mathrm{\Delta }`$). From Eq.(67) one can notice that the single-site Kondo temperature is substantially enhanced by the strong repulsive interactions in the 1D LL: Again, repulsive interactions promote SDW fluctuations in the electron liquid that interact prominently with the localized impurity. The screening of the impurity becomes prominent and we have the formation of a strong nonmagnetic barrier at $`x=0`$. The fact that $`\lambda _3`$ flows to strong couplings produces a new boundary condition at the origin:
$$<\mathrm{\Phi }_c(0)>=\sqrt{\frac{\pi }{2}}<\chi _{}^{}(0)\chi _+(0)>=\sqrt{\mathrm{\Delta }},$$
(75)
that pins the charge cosine potential. The impurity is screened by a physical electron and then holons now escape away from the impurity site (see Appendix B): no kink in $`\mathrm{\Phi }_c`$ is then possible at the origin. By symmetry, two different fixed points are authorized.
First, one can consider weak tunneling of electrons through the impurity site. This fixed point has been analyzed in refs. and still gives a power law for the conductance, but with a different exponent than in the weak coupling limit:
$$G(T)T^{1/K1}.$$
(76)
Of course, it vanishes at zero temperature. The exponent $`\stackrel{~}{\mu }`$ is renormalized as:
$$\stackrel{~}{\mu }=1+1/K.$$
(77)
In such case, backscattering off magnetic impurity will tend to reduce considerably the current in the wire. From this point of view, we obtain a perfect insulator at $`T=0`$: However, in the single impurity case the d.c. conductivity still grows at low temperatures \[see Eqs.(34),(76)\]. Since $`K`$ does not change with $`l`$, the transition between zero/finite conductance here occurs at $`\stackrel{~}{\mu }=2`$, i.e. in the neighborood of the noninteracting fixed point $`K_c=1`$. This situation would require that the screening characteristic time is very large compared to the tunneling one.
Second, one could rather investigate the opposite limit \[i.e. screening time $``$ tunneling time\]. By extension of the noninteracting case, one could then expect a nonperturbative ground state of Fermi liquid type. Using boundary conformal field theory, we have proved that electron-electron interactions could also produce another ground state with typically the same thermodynamic properties as a 1D Fermi liquid, but with “magnon” spin quasiparticles and a nonuniversal Wilson ratio.
Numerical works are up to now not very numerous: A recent DMRG study and a quantum Monte-Carlo procedure have both checked that the impurity is well screened at low temperatures and found for very large Hubbard interactions, a susceptibility law $`\chi (T)1/\mathrm{\Delta }`$, typical of a heavy Fermi liquid. In such limit, charge degrees of freedom are completely frozen and the unusual power law behavior predicted in ref. cannot arise anymore. This scenario maybe could help in describing the recent heavy-fermion behavior reported in the strongly correlated (two-dimensional) material $`Nd_{2x}Ce_xCuO_4`$. Furthermore, computing the susceptibility curve at low temperatures, the authors of ref. have reported a non-Fermi liquid behavior with an anomalous exponent $`\eta =1/K`$ for weak interactions, and a Fermi-liquid exponent $`\eta =2`$ for vanishing interactions.
The fact that a simultaneous existence of two leading irrelevant operators, one describing Fermi-liquid behavior and the other describing Furusaki-Nagaosa anomalous behavior \[implying: screening time $``$ tunneling time\], should not be possible for weak interactions would deserve further numerical works.
A summary of the various equations and transport properties for the Kondo lattice(s) and the single impurity case is given in table 1.
### C Array of independent impurities
To achieve this part, let us briefly compare with the case of $`N_i=/a`$ independent magnetic impurities in a wire of length $`L=`$ (i.e. the density of magnetic impurities is fixed to $`n_i=1`$). The total resistance of the wire reads
$$_{tot}(T)=\underset{N_i}{}D_m^o(T)=\frac{}{a}D_m^o(T)D_m(T).$$
(78)
Now, using the precious link between conductance and conductivity in 1D for a metallic system of size $``$,
$$G_{tot}=^1\sigma _{dc},$$
(79)
one finds the high-temperature law:
$$\sigma _{dc}(T)_{tot}^1=a[D_m^o(T)]^1\frac{1}{D_m^o}T^{1K}.$$
(80)
In the preceding studies of the Kondo lattices, we have obtained the same formula replacing $`\stackrel{~}{\mu }`$ by $`\mu `$. Since $`\mu \stackrel{~}{\mu }`$ \[because $`\mathrm{\Delta }_N1/2`$ and especially because $`K(T)`$ becomes smaller and smaller decreasing the temperature in the real lattice problem\], this leads to a faster decay of conductivity/conductance when the temperature decreases, compared to that of independent magnetic impurities. Coherent effects between magnetic impurities are included through the scaling dimension parameter $`\mathrm{\Delta }_N`$, and through the quantum renormalization of the exponent $`K`$, as well. These definitely enhance Kondo localization. Likewise, note that when the impurities are supposed to be completely independent, localization should not occur for attractive interactions \[see Eq.(73)\].
## VI Conclusion
To summarize, we have studied (two) metal-Kondo insulating transitions possibly occurring in one dimensional quantum wires. These result from a consequent antiferromagnetic (Kondo) coupling between conduction electrons very close to half-filling and a perfect lattice of local moments — producing a usual KLM.
As a result of commensurability, we have shown how a one-dimensional Kondo insulator can be understood as an effective (strong) umklapp process becoming relevant. Therefore, studying the pure KLM in the strong Kondo coupling limit $`J_Kt`$, results in a commensurate-incommensurate transition of Pokrovsky-Talapov type at zero temperature. For instance, the charge susceptibility should diverge as the inverse of the doping near half-filling, that is in good agreement with recent numerical datas.
In the weak coupling region $`(J_K,U)t`$, Kondo transitions can be classified as a function of the coherence-parameter in the spin array i.e. the scaling dimension of the staggered magnetization operator, namely $`\mathrm{\Delta }_N`$, which is always smaller than 1/2. These arise due to the important renormalization of the backward Kondo potential. In such regime, the interplay between electron-electron interactions and backward Kondo scattering is predicted to produce exotic and fascinating features in transport properties, in the high-temperature (ballistic) regime — whatever $`\mathrm{\Delta }_N1/2`$.
For repulsive interactions, the conductivity is found to decrease monotonically with temperature — even at high temperature: There is no remnant of the original umklapp process driven by the on-site Hubbard interaction. The resulting Kondo insulating phases are still stable in presence of weak non-magnetic disorder. These are perfect signatures of Kondo localizations.
In contrast, for weak attractive interactions, the electron liquid is less localized by magnetic impurities, and then the d.c. conductivity yields a maximum in the entrance of the localized phase. This is a precursor of the s-wave superconducting phase occurring for stronger attractive interactions.
The Kondo localization should even persist when the impurities are supposed to be randomly distributed. In particular, repulsive interactions also help in reducing the conductivity stiffness at intermediate length scales. As a consequence, persistent currents on thin mesoscopic rings with prominent magnetic defects should then *decrease* for repulsive interactions.
Finally, for completeness, consequences of the backward Kondo scattering on transport of LL’s have been summarized in Table 1, both in the one impurity case and that of a perfect lattice of magnetic impurities.
###### Acknowledgements.
I thank T. Giamarchi and T. Maurice Rice for relevant comments throughout this work. I also benefit from discussions with P. Prelovšek and A. Rosch on transport in low-dimensional systems.
## A Duality spinless fermion-boson in one dimension
We use a continuum version of the Jordan-Wigner scheme to rewrite the right and left electron fields $`\mathrm{\Psi }_q(x)`$ (with $`q=\pm `$) in a bosonic basis. We define
$$\mathrm{\Psi }(x)=\mathrm{\Psi }_+(x)+\mathrm{\Psi }_{}(x).$$
(A1)
and,
$$\psi _j\mathrm{\Psi }(x)\sqrt{a}$$
(A2)
$`a`$ being the lattice spacing. Each fermion operator can be re-expressed as,
$$\mathrm{\Psi }_q(x)=\frac{1}{\sqrt{2}}𝒪_q(x)(x)=\frac{1}{\sqrt{2}}\mathrm{exp}[i\pi q\underset{y<x}{}n(y)](x),$$
(A3)
where $`n(x)=_q\mathrm{\Psi }_q^{}(x)\mathrm{\Psi }_q(x)`$ is the total number operator, and $`(x)`$ is in principle an explicit hard core boson operator. The extra factor $`1/\sqrt{2}`$ allows to fullfill the Jordan-Wigner constraint,
$$n(x)=^{}(x)(x).$$
(A4)
By construction of spinless fermions, we have: $`^2(x)=0`$. On the other hand, as long as $`\left|xy\right|1`$ (1 being now the lattice spacing), we can check that Bose operators always commute. The trick is then to come back to a free boson system using the argument that in very low density there is no real difference between free- and hard core bosons. To perform this explicitly, we rewrite
$$n(x)=\rho _o+b^{}(x)b(x),\rho _o=\frac{1}{2}=\frac{k_F}{\pi }$$
(A5)
Considering a gas of spinless fermions near half-filling $`[n(x)1/2]`$ results in:
$$\frac{\delta \rho }{\pi }=b^{}(x)b(x)0,$$
(A6)
and approximately,
$$b(x)=(\frac{\delta \rho }{\pi })^{1/2}\mathrm{exp}i\sqrt{\pi }\mathrm{\Theta }_c.$$
(A7)
$`\delta \rho `$ measures density fluctuations and $`\mathrm{\Theta }_c`$ is the associated superfluid phase. This corresponds to the usual phase-amplitude decomposition. It is convenient to introduce a phonon-like displacement field:
$$\delta \rho (x)=\sqrt{\pi }_x\mathrm{\Phi }_c(x).$$
(A8)
As usual, the density and phase are canonically conjugate quantum variables taken to satisfy:
$$[\mathrm{\Theta }_c(x),_y\mathrm{\Phi }_c(y)]=i\delta (xy).$$
(A9)
This choice of wave-function is particularly judicious because it satisfies the (last) condition that $`\mathrm{\Psi }_q(x)`$ must remove a unit charge e at the coordinate x. To see this, one can rewrite:
$$e^{i\sqrt{\pi }\mathrm{\Theta }_c}=e^{i\sqrt{\pi }_{\mathrm{}}^x𝑑x^{}\mathrm{\Pi }_c(x^{})},$$
(A10)
and
$$\mathrm{\Pi }_c=_x\mathrm{\Theta }_c,$$
(A11)
is the momentum conjugate to $`\mathrm{\Phi }_c`$. Since $`\mathrm{\Pi }_c`$ is the generator of translations in $`\mathrm{\Phi }_c`$, this creates a kink in $`\mathrm{\Phi }_c`$ of height $`\sqrt{\pi }`$ centered at $`x`$, which corresponds to a localized unit of charge using the definition of $`b^{}(x)b(x)`$. Moreover, the Jordan-Wigner string can be rewritten:
$`𝒪_q`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}\mathrm{exp}iq{\displaystyle ^x}𝑑x^{}(\delta \rho (x^{})+k_F)`$ (A12)
$`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}\mathrm{exp}iq(k_Fx+\sqrt{\pi }\mathrm{\Phi }_c(x)).`$ (A13)
We have thereby identified the correct Bosonized form for the (continuum) electron operators:
$$\mathrm{\Psi }(x)=\psi _R(x)e^{ik_Fx}+\psi _L(x)e^{ik_Fx},$$
(A14)
with:
$$\psi _q=\frac{\eta _q}{\sqrt{2\pi a}}:\mathrm{exp}i\sqrt{\pi }(\mathrm{\Theta }_c+q\mathrm{\Phi }_c):.$$
(A15)
We have replaced the lattice step $`a`$ taken to satisfy Eq.(A3). This construction has been built originally by Haldane. The expression (A13) is called normal ordered (: :) meaning that all $`\mathrm{}`$ constants have been substracted from the fermionic ground state. In particular, we have:
$$\underset{q}{}\psi _q^{}(x)\psi _q(x)_x\mathrm{\Phi }_c(x)0.$$
(A16)
We have added Klein factors defined by $`\eta _{}^{}{}_{}{}^{2}=\eta _{+}^{}{}_{}{}^{2}=1`$ and $`\eta _+\eta _{}+\eta _{}\eta _+=0`$, to perfectly fullfill anticommutation rules between left and right movers. Moreover, the density fluctuation field $`_x\mathrm{\Phi }_c`$ creates propagating particles. This is just a consequence of the fact that in 1D electron-hole pairs are exactly coherent because they propagate with the same velocity. The bosonization scheme just reflects this important point.
Then, one can easily prove that the so-called Dirac Hamiltonian $`H=𝑑x`$ with Hamiltonian density
$$=iv_F[\psi _+^{}_x\psi _+\psi _{}^{}_x\psi _{}],$$
(A17)
can be exactly rewritten as a wave propagating at velocity $`u=v_F`$ (and $`K=1`$):
$$=\frac{u}{2\pi }[\frac{1}{K}:(_x\mathrm{\Phi }_c)^2:+K:(\mathrm{\Pi }_c)^2:].$$
(A18)
Remarkably, adding a weak Hubbard interaction between electrons results only in a renormalization of $`K`$ and $`u`$ as
$$uK=v_F\text{and}K=1\frac{Ua}{\pi v_F}$$
(A19)
We obtain the general Luttinger liquid Hamiltonian.
## B Weak coupling and fermionic kinks
In the weak coupling limit $`J_K/t1`$ and $`U/t1`$ and at half-filling, one must start with $`K1`$. The density of holons at $`q=4k_F0`$ is described by:
$$\chi _{}^{}(x)\chi _+(x)\mathrm{exp}[i\sqrt{2\pi }\mathrm{\Phi }_c(x)].$$
(B1)
This has a scaling dimension close to 1/2. We immediately deduce that the scaling dimension of the holon field is 1/4. Then, for the electron gas, the holon (like the spinon, see Appendix C) behaves as a semion or half of an electron. This is equivalent to say that the free electron decomposes itself into spinon and holon. At low temperatures and half-filling, holons acquire a mass due to the strong renormalization of the backward Kondo exchange, and charge carriers rather behave as fermionic kinks: The Kondo coupling is equivalent to an umklapp process. To show that in a quite exact manner, we adopt the following scheme. First, it is convenient to perform an average on the spin sector:
$$<\text{Tr}\stackrel{}{\sigma }\mathrm{\Phi }^{(1/2)}\stackrel{}{N}>=m^{\frac{\lambda ^2}{4\pi }}:\text{Tr}\stackrel{}{\sigma }\mathrm{\Phi }^{(1/2)}\stackrel{}{N}:,$$
(B2)
with $`m=\mathrm{\Delta }`$ (the spin gap at the Fermi level) and:
$$\lambda ^2=2\pi (1+2\mathrm{\Delta }_N).$$
(B3)
At the crossover temperature, we have:
$$:\text{Tr}\stackrel{}{\sigma }\mathrm{\Phi }^{(1/2)}\stackrel{}{N}:=\text{constant}.$$
(B4)
The Kondo exchange turns the spinons of the 1D LL into optical magnons. Using the gap-equation (19), this results in
$$J_K<\text{Tr}\stackrel{}{\sigma }\mathrm{\Phi }^{(1/2)}\stackrel{}{N}>\mathrm{\Delta }^{2\frac{K}{2}}.$$
(B5)
The Kondo interaction then reads:
$$_{hf,c}=\frac{\mathrm{\Delta }^{2\frac{K}{2}}}{2\pi a}\mathrm{cos}\sqrt{2\pi K}\mathrm{\Phi }_c.$$
(B6)
Second, when the bare LLP is (nearly) 1, one gets:
$$_{hf,c}=\frac{\mathrm{\Delta }^{3/2}}{2\pi a}\mathrm{cos}\sqrt{2\pi }\mathrm{\Phi }_c=\frac{\mathrm{\Delta }}{2\pi a}\mathrm{cos}\sqrt{4\pi }\mathrm{\Phi }_c.$$
(B7)
Using the definition (4), one definitely obtains:
$$_{hf,c}=i\mathrm{\Delta }(\psi _+^{}\psi _{}\psi _{}^{}\psi _+),$$
(B8)
that typically leads to the same spectrum given in Eq.(5). Again, we have chosen Klein factors such as $`\eta _+\eta _{}=+i`$. As soon as $`J_K0`$, we obtain a Kondo insulator: Pairs of holons cannot be excited at zero temperature and typically the LLP tends to zero (exactly at half-filling). Note that other excitation branches (‘breathers’), which correspond to kink-antikink bound states, cannot appear because the effective dimension of the SG-operator is 1. *The backward Kondo coupling is then equivalent to an umklapp process with $`K=1/2`$ and $`g_3=\mathrm{\Delta }`$ at low temperatures and low freqencies.* However, conversely to the strong coupling limit $`J_K/t1`$, this picture is here only appropriate at low energy. Charge excitations at high temperatures are rather semions. This produces differences e.g. in transport features at high frequency.
## C Field theories for spin degrees of freedom
Considering low-dimensional spin problems with SU(2) symmetry, we obtain a particular class of conformally invariant theories which has an Hamiltonian density quadratic in the currents $`J^a(x)`$ $`(a=x,y,z)`$:
$$_s(x)=\frac{1}{2+k}:\stackrel{}{J}(x)\stackrel{}{J}(x):.$$
(C1)
The velocity of spin excitations has been fixed to 1. Here $`k`$ is the Kac-Moody level which must be a positive integer. It follows that the Sugawara density Hamiltonian obeys the so-called Virasoro algebra with a conformal anomaly parameter: $`C=3k/(2+k)`$.
### 1 Heisenberg model
The low-temperature behavior of a single Heisenberg chain is described by a Sugawara Hamiltonian with $`k=1`$.
The physical particles (pairs of spin-1/2 excitations or spinons), are included through the primary fields $`\mathrm{\Phi }^{(1/2)}`$ and $`\mathrm{\Phi }^{(1/2)}`$ from the representation of the $`SU(2)`$ group. The action taking into account the dynamics of the spinon pairs is given by:
$`S_{WZW}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi }}{\displaystyle d^2x\text{Tr}\text{(}_\mu \mathrm{\Phi }^{(1/2)}_\mu \mathrm{\Phi }^{(1/2)}\text{)}}`$ (C2)
$`+`$ $`{\displaystyle \frac{i}{24\pi }}{\displaystyle _0^{\mathrm{}}}𝑑ϵ{\displaystyle d^2xϵ^{\alpha \beta \gamma }\text{Tr}\text{(}𝒜_\alpha 𝒜_\beta 𝒜_\gamma \text{)}},`$ (C3)
and,
$$𝒜_\alpha =\mathrm{\Phi }^{(1/2)}_\alpha \mathrm{\Phi }^{(1/2)}.$$
(C5)
The last term is from topological origin. The $`2k_F`$ SDW operator (i.e. the $`2k_F`$ spinon density) can be identified as:
$$\stackrel{}{N}(x)=e^{i2k_Fx}\text{Tr}(\mathrm{\Phi }^{(1/2)}+\text{Tr}\mathrm{\Phi }^{(1/2)})\stackrel{}{\sigma }.$$
(C6)
and has a scaling dimension 1/2. We immediately deduce that a single spinon at $`k=k_F`$ has a scaling dimension 1/4 and behaves as a semion .
### 2 Usual Spin ladder
Here, it is natural to rewrite the theory in terms of the total spin current:
$$\stackrel{}{R}_{L/R}(x)=\stackrel{}{J}_{L/R}(x)+\stackrel{}{I}_{L/R}(x).$$
(C7)
These obey the Kac-Moody algebra with $`k=2`$. The associated Hamiltonian, constructed from $`\stackrel{}{R}(x)`$, $`_s`$, has C=3/2. A crucial point is that $`_{sJ}+_{sI}`$ can be written as a sum of two commuting pieces, $`_s`$ and a remainder. The value of the conformal anomaly for this remainder Hamiltonian is $`C=23/2=1/2`$. There is a unique unitary conformal theory with this value of C, namely the Ising model: This is an example of Goddard-Kent-Olive coset construction. The eigenstates in the $`SU(2)_{2L}\times SU(2)_{2R}`$ sector appear in conformal towers labeled by spin quantum numbers $`j=0,1/2,1`$. The corresponding primary fields are: the identity 1, the fundamental field $`\gamma ^{(1/2)}`$, and the triplet operator (a $`3\times 3`$ matrix) $`𝚽^{(\mathrm{𝟏})}=_{i,j}\mathrm{\Phi }_{Lj}\mathrm{\Phi }_{Rj}`$. Their scaling dimension is given by $`0,3/8,1`$ respectively. Similarly, there are three primary fields in the Ising sector: the identity operator $`\mathrm{𝟏}`$, the Ising order parameter $`\sigma `$ and the energy operator $`ϵ`$ with scaling dimensions $`0,1/8,1`$ respectively.
The spin ladder system is then described by the effective Hamiltonian:
$$H=H_s^{(k=2)}+H^{(Ising)}+𝑑x[\mathrm{\Delta }\text{Tr}𝚽^{(\mathrm{𝟏})}3\mathrm{\Delta }ϵ].$$
(C8)
$`\mathrm{\Delta }`$ is typically the interchain coupling.
Notice that starting with two different intrachain Heisenberg coupling constants produces an extra term in the action, namely:
$$\delta S\frac{|J_1J_2|}{max(J_1,J_2)}d^2x\text{Tr}\text{(}_\mu \mathrm{\Phi }^{(1/2)}_\mu \mathrm{\Phi }^{(1/2)}\text{)}.$$
(C9)
Redefining $`\mathrm{\Phi }^{(1/2)}=i\stackrel{}{\sigma }𝚽`$, the result is an O(3) nonlinear sigma model built out of the field $`𝚽`$. Since the topological term has here no contribution, then the gapless ordered state of the isotropic sigma model is unstable. The spin gap should be rescaled as:
$$\mathrm{\Delta }_S\mathrm{\Delta }\mathrm{exp}\text{(}\frac{|J_1J_2|}{max(J_1,J_2)}\text{)}$$
(C10)
The distinction between $`\mathrm{\Delta }`$ and $`\mathrm{\Delta }_S`$ can be done only for appreciable differences between $`J_1`$ and $`J_2`$.
### 3 Takhtajan-Babujan chain
Now, we start with the integrable Takhtajan-Babujan S=1 chain on a lattice, described by:
$$H=J_{}\underset{j}{}(𝐒_j𝐒_{j+1})\beta (𝐒_j𝐒_{j+1})^2,$$
(C11)
where $`\beta =1`$. The unusual term quartic in spin can be generated, for example, by phonons. Solving Bethe Ansatz equations, the only elementary excitation is known to be a doublet of gapless spin-1/2 spin waves, with total spin 0 or 1. The specific heat behaves as: $`C_v/L=2ST/(1+S)+𝒪(T^3)`$, where S=1. Then, in the sense of critical theories, such a model could be parametrized by a conformal anomaly $`C=3S/(S+1)=3/2`$. This fact, together with numerical checks, leads to the conclusion that the criticality of this model is governed by operators satisfying a critical Sugawara model with $`k=2S=2`$. In particular, the staggered magnetization $`\stackrel{}{N}`$ can be simply expressed as:
$$\stackrel{}{N}(x)=\text{Tr}\{(\gamma ^{(1/2)}+\gamma ^{(1/2)})\stackrel{}{\sigma }\},$$
(C12)
with scaling dimension 3/8. The Hamiltonian is in turn equivalent to three Majorana fermions (or triplet excitations, and not doublets).
## D Links with non-magnetic Gaussian disorder
Let us now make comparisons with transport properties of a LL with many randomly distributed non-magnetic impurities.
For not too strong disorder, one usually approximates the disorder by a random potential,
$$H_{dis}=𝑑xV(x)\rho ^{(2k_F)}(x).$$
(D1)
$`\rho ^{(2k_F)}`$ is the $`2k_F`$ charge density operator. As usual, we can omit forward scattering because it can be incorporated into the shift of the chemical potential and does not affect the fixed point properties. $`V(x)`$ is generally Gaussian correlated i.e.:
$$\overline{V(x)}=0,\overline{V^{}(x)V(x^{})}=D\delta (xx^{}).$$
(D2)
One also assumes that the concentration of impurities becomes infinite $`n_i+\mathrm{}`$, but that the scattering potential of each impurity becomes weak $`V0`$ so that the product:
$$D=n_iV^2,$$
(D3)
remains a constant. Note the manifest difference with the Kondo lattice at half-filling. For $`\mathrm{\Delta }_N=0`$ and high temperatures, the RKKY interaction produces a constant (flat) uniaxial potential:
$$\overline{V(x)}=\lambda _3,\overline{V(x)V(0)}=D_m=\lambda _{3}^{}{}_{}{}^{2}.$$
(D4)
Adding an explicit antiferromagnetic exchange between local moments creates rather an isotropic potential; each component satisfies (i=x,y,z):
$`\overline{V(x)}`$ $`=`$ $`\lambda _3<\text{Tr}\text{(}\mathrm{\Theta }^{(1/2)}(x)\sigma ^i\text{)}>=0,`$ (D5)
$`\overline{V(x)V(0)}`$ $`=`$ $`D_m<N^i(x)N^i(0)>=D_m/x^{2\mathrm{\Delta }_N}.`$ (D6)
(The two-point correlation functions are here computed for equal times). For simplicity, in the following, we consider the bare lattice step equal to one. For spins S=1/2 and a resulting Heisenberg exchange, we have $`\mathrm{\Delta }_N=1/2`$. For an S=1 Takhtajan-Babujan chain, the result is $`\mathrm{\Delta }_N=3/8`$. The averages are essentially performed on quantum fluctuations. By analogy to non-magnetic disorder, we have defined:
$$H_{mag}=𝑑x\stackrel{}{V}(x)\stackrel{}{\rho _s}^{(2k_F)}(x),$$
(D7)
with:
$$V_i(x)=V(x)=\lambda _3N^i(x).$$
(D8)
In KLM’s with $`n_i=1`$, the magnetic potential has a Fourier component at $`q=\pi `$, and then it is only relevant at half-filling.
In the interplay between (Gaussian correlated) nonmagnetic disorder and interactions, the memory function approximation does not give the correct result. An extra term occurs coupling disorder and spin backscattering. We must proceed as follows.
In the pure system, one has to include the spin interaction:
$$_{bs}=\frac{2g_1}{(2\pi )^2}𝑑x\mathrm{cos}\sqrt{8\pi }\mathrm{\Phi }_s(x).$$
(D9)
written in the Abelian formalism. For small U, the spin Luttinger parameter reads
$$K_s=1+\frac{g_1}{\pi v_F}=1+\frac{U}{\pi v_F}$$
(D10)
For repulsive interactions, the fixed point is described by $`K_s^{}=1`$ and $`g_1^{}0`$. For attractive interactions, $`g_1`$ flows to $`\mathrm{}`$ and opens a superconducting gap. But, in both cases, $`g_1`$ affects much transport properties at intermediate length scales. In the Born (or still Hartree-Fock) approximation, one can write:
$$\sigma _{dc}(T)=T^1/D(T)=l_e(T).$$
(D11)
The conductivity, that is a physical quantity, is of course not changed under renormalization. The (dimensionless) non-magnetic disorder obeys:
$$D=l_{in}/l_e,$$
(D12)
where $`l_e`$ is the “effective” elastic mean-free path and $`l_{in}=v_F/T`$, the thermal length at which renormalization procedure must stop due to the generation of inelastic processes. We make the approximation that the equation flows are not modified up to this length scale. The obtained dc-conductivity corresponds to the one of the infinite system.
The delocalized regime is inevitably driven by quantum effects. The Anderson- (or still Kondo) localization length $`Ł_{loc}\mathrm{\Delta }^1`$ is very short in 1D: There is no diffusive or Boltzmann region, and then the mean-free path is not a universal quantity (the Einstein diffusion constant). Using the recursion law found in ref.:
$$\frac{dD}{d\mathrm{ln}T}=(3K_s(T)K\frac{g_1}{\pi v_F})D,$$
(D13)
one gets:
$$\sigma _{dc}(T)=T^{2\widehat{\mu }}\widehat{\mu }K_s(T)+K+\frac{g_1}{\pi v_F}$$
(D14)
This is the precise exponent of the d.c. conductivity at finite length scales. For small $`U`$, one obtains:
$$\sigma _{dc}(T)T^{U/\pi v_F}.$$
(D15)
Conversely to the case of magnetic impurities, repulsive interactions tend to make the disorder less relevant and decrease localization.
For sufficiently strong attractive interactions, one gets:
$$\sigma _{dc}(T)=T^{2K}.$$
(D16)
(One has to exclude the term in $`g_1D`$ because it becomes nonperturbative: its effect is anyway to create a gap resulting in $`K_s^{}0`$ and a quite small localization length).
For the 1D KLM’s with $`n_i=1`$, one can also write \[from Eq.(18)\]:
$$\frac{dD_m}{d\mathrm{ln}T}=(3\mu )D_m$$
(D17)
so that:
$$\sigma _{dc}(T)=T^1/D_m(T).$$
(D18)
This implies that at high temperatures we have a perfect duality to a model, where the distribution of the magnetic potential is rather Gaussian correlated and the effective exponent for the spin density-spin density correlation function in the electron liquid is $`\mu =K(T)+2\mathrm{\Delta }_N`$. It becomes then simple to investigate the competition between Anderson- and Kondo localization(s). For small $`U`$, one gets:
$`\mu `$ $`=`$ $`1{\displaystyle \frac{U}{\pi v_F}}+2\mathrm{\Delta }_N2{\displaystyle \frac{U}{\pi v_F}},`$ (D19)
$`\widehat{\mu }`$ $`=`$ $`2+{\displaystyle \frac{U}{\pi v_F}}`$ (D20)
For repulsive interactions, one finds $`\mu <\widehat{\mu }`$, producing Kondo localization whereas attractive interactions favor the Anderson glass to arise.
## E Third order terms for Gaussian magnetic randomness
We follow exactly the same scheme as in Appendix A of ref.. At lowest order, like for non-magnetic randomness, one obtains:
$`{\displaystyle \frac{dD_m}{d\mathrm{ln}L}}`$ $`=`$ $`(3K_sK)D_m,`$ (E1)
$`{\displaystyle \frac{dg_1}{d\mathrm{ln}L}}`$ $`=`$ $`2(1K_s)g_1.`$ (E2)
Here, $`D_m`$ and $`g_1`$ are dimensionless parameters. On the other hand, a third order term is also involved in the computation of 2-point correlation function $`R_\nu `$ $`(\nu =s,c)`$, which is:
$`R_\nu ^{\text{(III)}}(r_1r_2)=\text{const.}{\displaystyle \frac{g_1}{a^2}}{\displaystyle \frac{D_m}{a^2}}{\displaystyle \underset{ϵ_6=\pm }{}}{\displaystyle [\mathrm{}]}`$ (E3)
$`\times `$ $`<T_\tau e^{i\sqrt{2\pi }\mathrm{\Phi }_\nu (r_1)}e^{i\sqrt{2\pi }\mathrm{\Phi }_\nu (r_2)}\mathrm{cos}\sqrt{8\pi }\mathrm{\Phi }_s(r_3)`$ (E4)
$`\times `$ $`\mathrm{sin}\sqrt{2\pi }\mathrm{\Phi }_s(r_4)\mathrm{sin}\sqrt{2\pi }\mathrm{\Phi }_s(r_5)e^{iϵ_6[\sqrt{2\pi }\mathrm{\Phi }_c(r_4)\sqrt{2\pi }\mathrm{\Phi }_c(r_5)]}>`$ (E5)
with,
$$[\mathrm{}]=dx_3d\tau _3dx_4d\tau _4dx_5d\tau _5\delta (x_4x_5).$$
(E6)
If the points $`(x_4,\tau _4)`$ and $`(x_5,\tau _5)`$ are close together in a ring of inner radius a and width $`da`$, then the element of integration becomes:
$$[\mathrm{}]=\text{const.}da𝑑x_3𝑑\tau _3𝑑x_4𝑑\tau _4.$$
(E7)
When contracted and summed over $`ϵ_6`$ the
$$<e^{iϵ_6[\sqrt{2\pi }\mathrm{\Phi }_c(r_4)\sqrt{2\pi }\mathrm{\Phi }_c(r_5)]}>=\mathrm{exp}\text{(}F_c[0,\frac{a}{v_F}]\text{)}$$
(E8)
part gives a constant factor (that is independent of the lattice cut-off), and from the eliminated degrees of freedom we obtain:
$`+\text{const.}da{\displaystyle \frac{g_1}{a^2}}{\displaystyle \frac{D_m}{a^2}}{\displaystyle 𝑑x_3𝑑\tau _3𝑑x_4𝑑\tau _4}<T_\tau `$ (E9)
$`e^{i\sqrt{2\pi }\mathrm{\Phi }_\nu (r_1)}e^{i\sqrt{2\pi }\mathrm{\Phi }_\nu (r_2)}\mathrm{cos}\sqrt{8\pi }\mathrm{\Phi }_s(r_3)\mathrm{cos}\sqrt{8\pi }\mathrm{\Phi }_s(r_4)>`$ (E10)
which is identical to a $`g_1^2`$ term and contributes only to $`R_s`$. In particular, redefining $`D_maD_m`$ (the dimensionless disorder) one gets:
$$g_1^2(l+dl)=g_1^2(l)+\text{const}.dlD_mg_1(l).$$
(E11)
As usual, one puts: $`a=e^l`$ and $`\frac{da}{a}=dl`$. The recursion law for the spin backscattering becomes exactly:
$$\frac{dg_1}{d\mathrm{ln}L}=2(1K_s)g_1+D_m.$$
(E12)
If the points $`(x_3,\tau _3)`$ and $`(x_4,\tau _4)`$ \[or $`(x_5,\tau _5)`$\] are close together, one obtains an extra renormalization of $`D_m`$,
$$D_m(l+dl)=D_m(l)+\frac{g_1}{\pi v_F}dlD_m(l),$$
(E13)
and finally:
$$\frac{dD_m}{d\mathrm{ln}L}=(3K_sK+\frac{g_1}{\pi v_F})D_m.$$
(E14)
The sign of the last term is opposite to the one in the non-magnetic disorder problem. This reveals a strong competition between Kondo localization and superconductivity for attractive enough electron-electron interactions.
| Impurities | $`D_m`$ | $`dK/dl`$ | $`dD_m/dl`$ | exponent | $`\sigma _{dc}`$ or $`\delta G`$ | transition |
| --- | --- | --- | --- | --- | --- | --- |
| lattice | $`\lambda _{3}^{}{}_{}{}^{2}`$ | $`K^2D_m(l)`$ | $`[3\mu (l)]D_m`$ | $`\mu =K(T)+2\mathrm{\Delta }_N`$ | $`\sigma T^{2\mu [T]}`$ | Kondo insulator(s) |
| single | $`\lambda _{3}^{}{}_{}{}^{2}`$ | 0 | $`[2\stackrel{~}{\mu }(l)]D_m`$ | $`\stackrel{~}{\mu }=K+1`$ | $`\delta GT^{\stackrel{~}{\mu }2}`$ | Boundary effects |
Table 1: Weak coupling.— Renormalization of the various parameters, physical properties and transport properties both in the case of Kondo lattices and of a single magnetic impurity. In this Table, for simplicity we nominate $`D_m`$, the magnetic disorder parameters in both limits. In the case of randomly distributed magnetic defects, we have found $`\sigma _{dc}(T)T^{Ua/\pi v_F}`$ and a glassy phase which should correspond to a disordered Ising phase. For a comparison with non-magnetic impurities, see for instance ref.. |
warning/0001/math0001143.html | ar5iv | text | # On a conjecture of Shokurov: Characterization of toric varieties
## 1. Introduction
The aim of this note is to discuss the birational characterization of toric varieties. Let $`X`$ be a normal projective toric variety and let $`D=_{i=1}^rD_i`$ be the sum of invariant divisors. It is well known that the pair $`(X,D)`$ has only log canonical singularities (see e. g. \[3, 3.7\]), $`K_X+D`$ is linearly trivial and $`r=\mathrm{rk}(\mathrm{Weil}(X)/)+dim(X)`$, where $`\mathrm{Weil}(X)`$ is the group of Weil divisors and $``$ is the algebraic equivalence.
Shokurov observed that this property can characterize toric varieties:
###### Conjecture 1.1 ().
Let $`(X,D=d_iD_i)`$ be a projective log variety such that $`(X,D)`$ has only log canonical singularities and numerically trivial. Then $`d_i\mathrm{rk}(\mathrm{Weil}(X)/)+dim(X)`$. Moreover, if the equality holds, then $`(X,D)`$ is a toric pair.
Shokurov also conjectured the relative version of 1.1 (cf. Theorem 2.3) and he expects that one can replace numerical triviality of $`K_X+D`$ with nefness of $`(K_X+D)`$. We do not discuss these details here.
Conjecture 1.1 was proved in dimension two in (see also \[9, Sect. 8\] and Proposition 2.1 below). Our main result is the following partial answer on Conjecture 1.1 in dimension three:
###### Theorem 1.2.
4 Let $`(X,D=d_iD_i)`$ be a three-dimensional projective variety over $``$ such that $`K_X+D0`$ and $`(X,D)`$ has only purely log terminal singularities. Then
(1.3)
$$d_i\mathrm{rk}(\mathrm{Weil}(X)/)+3.$$
Moreover, if the equality holds, then up to isomorphisms one of the following holds:
1. $`X^3`$, $`D=0`$ or $`D=^2`$;
2. $`X^1\times ^2`$, $`D=0`$ or $`D=\{\mathrm{pt}\}\times ^2`$ or $`D=^1\times \{\mathrm{line}\}`$;
3. $`X\left(𝒪_^1𝒪_^1𝒪_^1(d)\right)`$, $`d1`$, $`D`$ is the section correspondong to the surjection $`𝒪_^1𝒪_^1𝒪_^1(d)𝒪_^1(d)`$;
4. $`X^1\times ^1\times ^1`$, $`D=0`$ or $`D=\{\mathrm{pt}\}\times ^1\times ^1`$ or $`D=\{\mathrm{pt}_1,\mathrm{pt}_2\}\times ^1\times ^1`$;
5. $`X\left(𝒪_^2𝒪_^2(d)\right)`$, $`d1`$, $`D`$ is the negative section, or a disjoint union of two sections, one of them is negative;
6. $`X\left(𝒪_{^1\times ^1}\right)`$, $`\mathrm{Pic}(^1\times ^1)`$, $`D`$ is the negative section, or a disjoint union of two sections, one of them is negative.
In all cases $`(X,D)`$ is toric.
Clearly, our theorem is not a characterization of toric varieties, but we hope that Conjecture 1.1 can be proved in a similar way.
This paper was subject of my talk given at Waseda seminar on January 18, 2000. I am grateful to the participants of this seminar, especially Professor S. Ishii, for attention and valuable discussions. I also would like to thank the Department of Mathematics of Tokyo Institute of Technology for hospitality during my stay on 1999–2000. This work was partially supported by the grant INTAS-OPEN-97-2072.
## 2. Preliminaries
### Notation
All varieties are defined over $``$. Basically we employ the standard notation of the Minimal Model Program (MMP, for short). Throughout this paper $`\rho (X)`$ is the Picard number and $`\overline{NE}(X)`$ is the Mori cone of $`X`$. We call a pair $`(X,D)`$ consisting of a normal algebraic variety $`X`$ and a boundary $`D`$ on $`X`$ a log variety or a log pair. Here a boundary is a $``$-Weil divisor $`D=d_iD_i`$ such that $`0d_i1`$ for all $`i`$. A contraction is a projective morphism $`\phi :XZ`$ of normal varieties such that $`\phi _{}𝒪_X=𝒪_Z`$. Abbreviations klt, plt, lc are reserved for Kawamata log terminal, purely log terminal and log canonical, respectively (refer to , and for the definitions). Let $`(X,D)`$ be a log pair and let $`S:=D`$. For simplicity, assume that $`(X,D)`$ is lc in codimension two. The Adjunction Formula proposed by Shokurov \[11, Sect. 3\] states $`(K_X+D)|_S=K_S+\mathrm{Diff}_S(DS)`$, where $`\mathrm{Diff}_S(DS)`$ is a naturally defined effective $``$-Weil divisor on $`S`$, so-called different. Moreover, $`K_X+D`$ is plt near $`S`$ if and only if $`S`$ is normal and $`K_S+\mathrm{Diff}_S(DS)`$ is klt \[4, 17.6\]. $`LCS(X,D)`$ denotes the locus of log canonical singularities of $`(X,D)`$ that is the set of all points where $`(X,D)`$ is not klt . Let $`\phi :XZ`$ be any fiber type contraction and let $`D=d_iD_i`$ be a $``$-divisor on $`X`$. We will write $`D=_{\mathrm{ver}}d_iD_i+_{\mathrm{hor}}d_iD_i=D^{\mathrm{ver}}+D^{\mathrm{hor}}`$, where $`_{\mathrm{ver}}`$ (respectively $`_{\mathrm{hor}}`$) runs through all components $`D_i`$ such that $`dim\phi (D_i)<dim(D_i)`$ (respectively $`\phi (D_i)=Z`$). We will frequently use the above notation without reference.
In dimension two Conjecture 1.1 is much easier than higher dimensional one. We need only the following weaker version:
###### Proposition 2.1.
Let $`(X,D=d_iD_i)`$ be a projective log surface such that $`(K_X+D)`$ is nef and $`(X,D)`$ is lc. Then $`d_i\rho (X)+2`$. Moreover, if the equality holds and $`(X,D)`$ is klt, then $`X^2`$, or $`X^1\times ^1`$.
For the general statement we refer to , see also .
###### Proof.
Assume that $`d_i\rho (X)20`$ and run $`K_X`$-MMP. According to , Log MMP works even in the category of log canonical pairs. On each step, $`d_i\rho (X)2`$ does not decrease and all assumptions are preserved (see \[4, 2.28\]). At the end we get one of the following:
### Case 1
$`\rho (X)=1`$. Then $`d_i\rho (X)20`$ by \[4, 18.24\].
### Case 2
There is an extremal contraction onto a curve $`\phi :XZ`$ (in particular, $`\rho (X)=2`$). Let $`\mathrm{}`$ be a general fiber. Then
(2.2)
$$2=K_X\mathrm{}D\mathrm{}\underset{\mathrm{hor}}{}d_i.$$
Hence $`_{\mathrm{ver}}d_i2`$ and $`K_X+D^{\mathrm{hor}}`$ is not nef. Let $`\varphi :XW`$ is a contraction of $`(K+D^{\mathrm{hor}})`$-negative extremal ray. If $`\varphi `$ is birational, we replace $`X`$ with $`W`$ and obtain Case 1 above. Thus we may assume that $`W`$ is a curve, so $`\phi `$ and $`\varphi `$ are symmetric. As above, $`_{\mathrm{ver}}d_i2`$, so $`d_i=4`$.
Now assume that $`(X,D)`$ is klt and $`d_i\rho (X)2=0`$. Then after each divisorial contraction $`d_i\rho (X)2`$ increases. Hence we have only cases 1 or 2 above. In Case 1, $`X^2`$ by Lemma 3.1 below. In Case 2 we have the equality in (2.2), so $`D_i\varphi ^1(w)=1`$ for any component of $`D^{\mathrm{ver}}`$ and a general fiber $`\varphi ^1(w)`$. Therefore $`D_i`$ is not a multiple fiber of $`\phi `$ and $`X`$ is smooth along $`D_i`$. Similarly, if $`D_j`$ is a component of $`D^{\mathrm{hor}}`$, then $`\varphi (D_j)=\mathrm{pt}`$, $`X`$ is smooth along $`D_j`$ and $`D_iD_j=1`$. Finally, $`\phi \times \varphi :XZ\times W=^1\times ^1`$ is a finite morphism of degree $`\phi ^1(\mathrm{pt})\times \varphi ^1(\mathrm{pt})=D_iD_j=1`$. ∎
The local version of Conjecture 1.1 was proved in \[4, 18.22\]:
###### Theorem 2.3.
Let $`(X,D=d_iD_i)`$ be a log pair which is log canonical at a point $`PD_i`$. Assume that $`K_X`$ and all $`D_i`$’s are $``$-Cartier at $`P`$. Then $`d_idim(X)`$. Moreover, if the equality holds, then $`(XP,D)`$ is an abelian quotient of a smooth point and $`(X,D)`$ is not plt at $`P`$.
Recall that for any plt pair $`(X,D)`$ of dimension $`3`$ there is a small birational contraction $`q:X^qX`$ such that $`X^q`$ is $``$-factorial and $`(X^q,D^q:=q_{}^1D)`$ is plt (see \[4, 6.11.1\], \[4, 17.10\]). Such $`q`$ is called a $``$-factorialization of $`(X,D)`$. Applying a $``$-factorialization in our situation and taking into account that $`\mathrm{rk}(\mathrm{Weil}(X)/)=(\mathrm{rk}\mathrm{Weil}(X^q)/)\rho (X^q)`$ we obtain that for Theorem 1.2 it is sufficient to prove the following
###### Proposition 2.4.
Let $`(X,D=d_iD_i)`$ be a three-dimensional projective plt pair such that $`K_X+D0`$ and $`X`$ is $``$-factorial. Then
(2.5)
$$d_i\rho (X)+3.$$
Moreover, if the equality holds, then for $`(X,D)`$ there are only possibilities (i)–(iv) of Theorem 1.2.
## 3. Lemmas
In this section we prove several facts related to Conjecture 1.1.
###### Lemma 3.1 (cf. \[4, 18.24\], ).
Let $`(X,D=_{i=1}^rd_iD_i)`$ be a projective $`n`$-dimensional log pair such that all $`D_i`$’s are $``$-Cartier, $`\rho (X)=1`$, $`(X,D)`$ is plt and $`(K_X+D)`$ is nef. Then
(3.2)
$$d_in+1.$$
Moreover, if the equality holds, then $`X^n`$ and $`D_1,\mathrm{},D_r`$ are hyperplanes.
Note that in the two-dimensional case any plt pair is automatically $``$-factorial.
###### Proof.
We will prove this lemma in the case when $`D=0`$ (i.e. $`K_X+D`$ is klt). The case when $`D`$ is non-trivial (and irreducible) can be treated in a similar way. The inequality (3.2) was proved in \[4, 18.24\], so we prove the second part of our lemma. Since $`K_X`$ is ample, $`\mathrm{Pic}(X)`$ (see e.g. \[8, 2.1.2\]). Let $`H`$ be an ample generator of $`\mathrm{Pic}(X)`$ and let $`D_ia_iH`$, $`a_i`$, $`a_i0`$. Assume that $`a_i<a_j`$ for $`ij`$ or $`K_X+D0`$. For $`0<\epsilon 1`$, consider
$$D^{(\epsilon )}:=\epsilon D_i+D\epsilon D_j.$$
Then $`K_X+D^{(\epsilon )}`$ is again klt (because the klt property is an open condition) and $`(K_X+D^{(\epsilon )})`$ is ample. Take $`N`$ so that $`N(K_X+D^{(\epsilon )})`$ is integral and very ample and let $`M|N(K_X+D^{(\epsilon )})|`$ be a general member. By Bertini type theorem \[3, Sect. 4\], $`(X,D^{(\epsilon )}+\frac{1}{N}M)`$ is klt (and numerically trivial). Moreover, the sum of coefficients of $`D^{(\epsilon )}+\frac{1}{N}M`$ is equal to $`n+1+\frac{1}{N}`$. This contradicts (3.2). Hence $`K_X+D0`$ and $`D_iD_j`$ for all $`i,j`$. Thus, for any pair $`D_{i,}D_j`$ there exists $`n_{i,j}`$ such that $`n_{i,j}(D_iD_j)0`$. By taking repeated cyclic covers (which are étale in codimension one) $`\pi :X^{}\mathrm{}X`$, we obtain a new plt pair $`(X^{},D^{}=_{i=1}^rd_iD_i^{})`$ \[4, 20.4\] such that $`D_i^{}D_j^{}`$, where $`D_i^{}=\pi ^{}D_i`$. On this step, we do not assume that $`D_i^{}`$ is irreducible. Then $`D_1^{},\mathrm{},D_r^{}`$ generate a linear system $``$ of Weil divisors. If $`\mathrm{Bs}()`$ is not empty, then we pick a point $`P^{}D_1^{}\mathrm{}D_r^{}`$. By construction, $`(X^{},D^{})`$ is klt at $`P^{}`$ and $`_{i=1}^rd_in+1`$, a contradiction with Theorem 2.3. Therefore $`\mathrm{Bs}()=\mathrm{}`$. In particular, $`D_1^{},\mathrm{},D_r^{}`$ are ample Cartier divisors and $`K_X^{}D^{}`$ is ample (i.e. $`X^{}`$ is a log Fano variety). This also shows that the Fano index of $`X^{}`$ is $`r(X^{})_{i=1}^rd_in+1`$. It is well known (see e.g. \[8, 3.1.14\]) that in this case we have $`r(X^{})=_{i=1}^rd_i=n+1`$, $`X^{}^n`$ and $`D_1^{},\mathrm{},D_r^{}`$ are hyperplanes. Since $`\pi :X^{}X`$ is étale outside of $`\mathrm{Sing}(X)`$ and $`X^{}`$ is smooth, the restriction $`X^{}\backslash \pi ^1\left(\mathrm{Sing}(X)\right)X\backslash \mathrm{Sing}(X)`$ is the universal covering. This gives us that $`\pi :X^{}X`$ is Galois. Hence $`X=^n/G`$, where $`GPGL_{n+1}`$ is a finite subgroup. Further, the group $`G`$ does not permute $`D_1^{},\mathrm{},D_r^{}`$. Thus $`G`$ has $`r>n+1`$ invariant hyperplanes $`D_1^{},\mathrm{},D_r^{}`$ in $`^n`$. Finally, the lemma follows by the following simple fact which can be proved by induction on $`n`$. ∎
###### Sublemma.
Let $`GPGL_{n+1}`$ be a finite subgroup acting on $`^n`$ free in codimension one. Assume that there are $`n+2`$ invariant hyperplanes $`H_1,\mathrm{},H_{n+2}^n`$. Then $`G=\{1\}`$.
###### Lemma 3.3.
Let $`\phi :XZo`$ be a three-dimensional flipping contraction and let $`D=d_iD_i`$ be a boundary on $`X`$ such that $`(X,D)`$ is plt, $`\rho (X/Z)=1`$, $`(K_X+D)`$ is $`\phi `$-nef and all $`D_i`$’s are $`\phi `$-ample. Assume that $`X`$ is $``$-factorial. Then $`d_i<2`$.
###### Proof.
Let $`\chi :X\stackrel{\phi }{}Z\stackrel{\phi ^+}{}X^+`$ be the flip with respect to $`K_X`$ and let $`D^+=d_iD_i^+`$ be the proper transform of $`D`$. Then all $`D_i^+`$’s are anti-ample over $`Z`$. Hence $`\phi ^{+1}(o)`$ is contained in $`D_i^+`$. Consider a general hyperplane section $`HX^+`$. Then $`(H,D|_H)`$ is plt \[3, Sect. 4\]. Applying Theorem 2.3 to $`H`$ we obtain $`d_i<2`$. ∎
###### Lemma 3.4.
Let $`\phi :XZ`$ be a contraction from a projective $``$-factorial threefold onto a surface such that $`\rho (X/Z)=1`$. Let $`D=d_iD_i`$ be a boundary on $`X`$ such that $`(X,D)`$ is lc, $`(X,DD)`$ is klt and $`(K_X+D)`$ is nef. Assume that $`D`$ has a component $`S`$ which is generically section of $`\phi `$. Then $`_id_i\rho (X)+3`$. Moreover, if the equality holds and $`(X,D)`$ is plt, then $`X`$ is smooth, $`\phi `$ is a $`^1`$-bundle, $`\phi |_S:SZ`$ is an isomorphism and $`Z^2`$ or $`Z^1\times ^1`$.
###### Proof.
Assume that $`_id_i\rho (X)+3`$. Since $`K_X`$ is $`\phi `$-ample, a general fiber $`\mathrm{}`$ of $`\phi `$ is isomorphic to $`^1`$. We have
(3.5)
$$2=K_X\mathrm{}=D^{\mathrm{hor}}\mathrm{}\underset{\mathrm{hor}}{}d_i,\underset{\mathrm{ver}}{}d_i\rho (X)+1=\rho (Z)+2.$$
Let $`\mu :=\phi |_S`$. Write $`\mathrm{Diff}_S(DS)==_i\beta _i\mathrm{\Theta }_i`$. Then
(3.6)
$$\beta _i=1\frac{1}{m_i}+\frac{1}{m_i}\underset{j𝔐_i}{}d_jk_{i,j},$$
where $`m_i\{\mathrm{}\}`$, $`k_{i,j}`$ and the sum runs through the set $`𝔐_i`$ of all components $`D_j`$ containing $`\mathrm{\Theta }_i`$ (see \[11, 3.10\]). Here $`m_i=\mathrm{}`$ when $`(X,D)`$ is not plt along $`\mathrm{\Theta }_i`$. It is easy to see that $`\beta _i_{j𝔐_i}d_j`$. Put $`\mathrm{\Xi }:=\mu _{}\mathrm{Diff}_S(DS)`$ and let $`\mathrm{\Xi }=\gamma _i\mathrm{\Xi }_i`$. Since $`(K_S+\mathrm{\Theta })`$ is nef, $`(Z,\mathrm{\Xi })`$ is lc \[4, 2.28\]. For any component $`D_i`$ of $`D^{\mathrm{ver}}`$ we have at least one component $`\mathrm{\Theta }_jD_iS`$ such that $`\mu (\mathrm{\Theta }_j)\mathrm{pt}`$. This yields
$$\underset{i}{}\gamma _i\underset{\mu (\mathrm{\Theta }_i)\mathrm{pt}}{}\beta _i\underset{\mathrm{ver}}{}d_j\rho (Z)+2.$$
Applying Proposition 2.1 to $`(Z,\mathrm{\Xi })`$, we obtain equalities
(3.7)
$$\underset{i}{}\gamma _i=\underset{\mu (\mathrm{\Theta }_i)\mathrm{pt}}{}\beta _i=\underset{\mathrm{ver}}{}d_j=\rho (Z)+2.$$
Hence $`_{\mathrm{hor}}d_i=2`$ and $`_id_i=\rho (X)+3`$. This shows the first part of the lemma.
Now assume that $`(X,D)`$ is plt. By Adjunction \[4, 17.6\], $`(S,\mathrm{Diff}_S(DS))`$ is klt and so is $`(Z,\mathrm{\Xi })`$. Again, by Proposition 2.1 we have either $`Z^2`$ or $`Z^1\times ^1`$. There exists a standard form of $`\phi `$ (see ), i. e. the commutative diagram
$$\begin{array}{ccc}\stackrel{~}{X}& & X\\ & & \\ \stackrel{~}{Z}& \stackrel{\sigma }{}& Z\end{array}$$
where $`\sigma :\stackrel{~}{Z}Z`$ is a birational morphism of smooth surfaces, $`\stackrel{~}{X}X`$ is a birational map and $`\stackrel{~}{\phi }:\stackrel{~}{X}\stackrel{~}{Z}`$ is a standard conic bundle (in particular, $`\stackrel{~}{X}`$ is smooth and $`\rho \left(\stackrel{~}{X}/X\right)=1`$). Take the proper transform $`\stackrel{~}{S}`$ of $`S`$ on $`\stackrel{~}{X}`$. For a general fiber $`\stackrel{~}{\mathrm{}}`$ of $`\stackrel{~}{\phi }`$ we have $`\stackrel{~}{S}\stackrel{~}{\mathrm{}}=1`$. Since $`\rho (\stackrel{~}{X}/\stackrel{~}{Z})=1`$, $`\stackrel{~}{S}`$ is $`\stackrel{~}{\phi }`$-ample. It gives us that each fiber of $`\stackrel{~}{\phi }`$ is reduced and irreducible, i. e. the morphism $`\stackrel{~}{\phi }`$ is smooth. By , there exists a standard conic bundle $`\widehat{\phi }:\widehat{X}Z`$ and a birational map $`\widehat{X}X`$ over $`Z`$. This map indices an isomorphism $`(\widehat{X}/\widehat{\phi }^1(𝔐))(X/\phi ^1(𝔐))`$, where $`𝔐Z`$ is a finite number of points. Since both $`\phi `$, $`\widehat{\phi }`$ are projective and $`\rho (X/Z)=\rho (\widehat{X}/Z)=1`$, we have $`\widehat{X}X`$. But then $`\phi :XZ`$ is smooth, i.e. $`\phi `$ is a $`^1`$-bundle.
Now we claim that $`\mu `$ is an isomorphism. Indeed, otherwise $`S`$ contains a fiber, say $`\mathrm{}_0`$. Then $`S`$ intersects all irreducible components of $`D^{\mathrm{hor}}S`$. If some component $`D_k`$ of $`D^{\mathrm{hor}}S`$ does not contain $`\mathrm{}_0`$, then $`\phi (SD_k)`$ is a component of $`\mathrm{\Xi }`$. By (3.7) we have
$$\rho (Z)+2=\underset{i}{}\gamma _i=\underset{\mu (\mathrm{\Theta }_i)\mathrm{pt}}{}\beta _id_k+\underset{\mathrm{ver}}{}d_j>\rho (Z)+2,$$
which is impossible. Therefore all components of $`D^{\mathrm{hor}}`$ contain $`\mathrm{}_0`$. Taking a general hyperplane section as in the proof of Lemma 3.3, we derive a contradiction. ∎
###### Corollary 3.7.1.
$`S`$ does not intersects $`\mathrm{Supp}(D^{\mathrm{hor}}S)`$ and all components of $`D^{\mathrm{hor}}S`$ are sections of $`\phi `$.
###### Lemma 3.8.
Let $`\phi :XZ`$ be a contraction from a $``$-factorial threefold onto a curve and let $`D=d_iD_i`$ be a boundary on $`X`$ such that $`(X,D)`$ is lc, $`(X,DD)`$ is klt. Let $`F`$ be a general fiber. Assume that $`(K_X+D)`$ is $`\phi `$-nef and $`\rho (X/Z)=1`$. Then $`_{\mathrm{hor}}d_i3`$. Moreover, if the equality holds and $`(X,D)`$ is plt, then $`F^2`$ and for any component $`D_i`$ of $`D^{\mathrm{hor}}`$ the scheme-theoretic restriction $`D_i|_F`$ is a line.
###### Proof.
Put $`\mathrm{\Delta }:=D|_F`$. Then $`(F,\mathrm{\Delta })`$ is lc, $`(F,\mathrm{\Delta }\mathrm{\Delta })`$ is klt (see \[3, Sect. 4\]) and $`(K_F+\mathrm{\Delta })`$ is nef. Moreover, if $`(X,D)`$ is plt, then so is $`(F,\mathrm{\Delta })`$. Write $`\mathrm{\Delta }=\delta _i\mathrm{\Delta }_i`$, where all $`\mathrm{\Delta }_i`$’s are irreducible curves on $`F`$. Clearly $`D^{\mathrm{ver}}|_F=0`$ and $`\delta _i_{\mathrm{hor}}d_i`$. If $`\rho (F)=1`$, then the assertion of 3.8 follows by Proposition 2.1. Assume that $`\rho (F)>1`$. Let $`C`$ be an extremal $`K_F`$-negative curve on $`F`$ (note that $`K_F`$ is not nef). Then $`C`$ intersects all components of $`\mathrm{\Delta }`$ (because $`\rho (X/Z)=1`$). Let $`\upsilon :FF^{}`$ be the contraction of $`C`$. If $`F^{}`$ is a curve, then we take $`C`$ to be a general fiber of $`\upsilon `$. By Adjunction, $`2=\mathrm{deg}K_C\mathrm{deg}\mathrm{\Delta }|_C`$. This gives us $`2\delta _i_{\mathrm{hor}}d_i`$. If $`\upsilon `$ is birational, then $`(F^{},\upsilon (\mathrm{\Delta }))`$ is lc and all components of $`\upsilon (\mathrm{\Delta })`$ pass through the point $`\upsilon (C)`$. By Theorem 2.3, the sum of coefficients of $`\upsilon (\mathrm{\Delta })`$ is $`2`$. Hence $`_{\mathrm{hor}}d_i\delta _i3`$. If $`(F,\mathrm{\Delta })`$ is klt, then so is $`(F^{},\upsilon (\mathrm{\Delta }))`$ and the inequality above is strict. Finally, if $`(F,\mathrm{\Delta })`$ is plt and $`\mathrm{\Delta }0`$, then we take $`C`$ to be $`(K_F+\mathrm{\Delta }\mathrm{\Delta })`$-negative extremal curve. Then $`C`$ is not a component of $`\mathrm{\Delta }`$. By \[3, 3.10\], $`(F^{},\upsilon (\mathrm{\Delta }))`$ is plt. Again, by Theorem 2.3 the sum of coefficients of $`\upsilon (\mathrm{\Delta })`$ is strictly less than $`2`$. So, $`_{\mathrm{hor}}d_i\delta _i<3`$. This proves Lemma 3.8. ∎
###### Corollary 3.8.1.
Notation as in Lemma 3.8. Assume additionally that $`X`$ is projective, $`(K_X+D)`$ is nef (not only over $`Z`$), $`_{\mathrm{hor}}d_i=3`$, $`_{\mathrm{ver}}d_i=2`$ and $`(X,D)`$ is plt. If $`D^{\mathrm{hor}}\mathrm{}`$, then $`D^{\mathrm{hor}}=D^1\times ^1`$ and $`X`$ is smooth along $`D`$. In particular, $`X`$ has at most isolated singularities.
###### Proof.
Put $`S:=D`$. By \[4, 17.5\], $`S`$ is normal. Since $`\rho (X/Z)=1`$, $`S`$ is irreducible and all components of $`DS`$ meet $`S`$. Let $`\mathrm{Diff}_S(DS)=\beta _i\mathrm{\Theta }_i`$. Clearly, $`(K_S+\mathrm{Diff}_S(DS))`$ is nef. By \[4, 17.6\], $`(S,\mathrm{\Theta })`$ is klt. As in the proof of Lemma 3.4 we see $`\beta _id_i14`$. If $`\rho (S)=2`$, then equalities $`\beta _i=d_i1=4`$ and Proposition 2.1 give us the assertion. Assume that $`\rho (S)>2`$. Then some fiber of $`\phi |_S:SZ`$ is not irreducible. Let $`\mathrm{\Gamma }`$ be its irreducible component and let $`\upsilon :SS^{}`$ be the contraction of $`\mathrm{\Gamma }`$. Taking into account that $`\mathrm{\Gamma }`$ intersects all components of $`D^{\mathrm{hor}}`$. As in Lemma 3.4 we get a contradiction. ∎
###### Lemma 3.9 (cf. \[11, 6.9\]).
Let $`\phi :XZo`$ be a $`K_X`$-negative contraction from a $``$-factorial variety $`X`$ such that $`\rho (X/Z)=1`$ and every fiber has dimension one. Let $`D`$ be a boundary on $`X`$ such that $`(X,DD)`$ is klt and $`K_X+D`$ is $`\phi `$-numerically trivial. Assume that $`D`$ is disconnected near $`\phi ^1(o)`$. Then $`K_X+D`$ is plt near $`\phi ^1(o)`$.
###### Proof.
Regard $`\phi :XZo`$ as a germ near $`\phi ^1(o)`$. Put $`S:=D`$. Clearly, for a general fiber $`\mathrm{}`$ of $`\phi `$ we have $`K_X\mathrm{}=D\mathrm{}=2`$. If $`S^{}`$ is an irreducible component of $`S`$ such that $`S^{}\mathrm{}=0`$, then $`S^{}=\phi ^1(C)`$ for a curve $`CZ`$. In this case, $`S^{}`$ contains $`\phi ^1(o)`$ and $`S`$ is connected near $`\phi ^1(o)`$. Therefore $`S`$ has exactly two connected components $`S_1`$, $`S_2`$, they are irreducible and $`S_1\mathrm{}=S_2\mathrm{}=1`$. Then $`S_i`$, $`i=1,2`$ meets all components of $`\phi ^1(o)`$. Hence $`S_i\phi ^1(o)`$ is $`0`$-dimensional. Since $`Z`$ is normal and $`\phi |_{S_i}:S_iZ`$ is birational, $`S_iZ`$ and $`S_i\phi ^1(o)`$ is a single point. In particular, $`\phi ^1(o)`$ is irreducible. Clearly, $`LCS(X,D)S=S_1S_2`$. Assume that $`(X,D)`$ is not plt. Then there is a divisor $`E`$ of the function field $`K(X)`$ with discrepancy $`a(E,D)<1`$. Let $`VX`$ be its center. Then $`VS`$ and we may assume that $`VS_1`$ (and $`VS_1`$). Let $`LZ`$ be any effective prime divisor containing $`\phi (V)`$ and let $`F:=\phi ^1(L)`$. Clearly, $`(X,D+F)`$ is not lc near $`V`$. For sufficiently small positive $`\epsilon `$ the log pair $`(X,D+F\epsilon S_1)`$ is not lc near $`V`$ and not klt near $`S_2`$. This contradicts Connectedness Lemma \[4, 17.4\]
## 4. Proof of Theorem 1.2
In this section we prove Proposition 2.4.
### 4.1. Inductive hypothesis
Notation and assumption as in Proposition 2.4. Our proof is by induction on $`\rho (X)`$. In the case $`\rho (X)=1`$, the assertion is a consequence of Lemma 3.1. To prove 2.4 for $`\rho (X)2`$ we fix $`\rho `$, $`\rho >1`$. Assume that inequality (2.5) holds if $`\rho (X)<\rho `$ and for $`\rho (X)=\rho `$ we have
(4.2)
$$d_i\rho (X)30.$$
### 4.3.
If $`(X,D)`$ is klt, then we run $`K_X`$-MMP. On each step $`KD`$ cannot be nef. Obviously, all steps preserve our assumptions (see \[4, 2.28\]) and the left hand side of (4.2) does not decrease. Moreover, by our assumptions we have no divisorial contractions on $`X`$ (because after any divisorial contraction the left hand side of (4.2) decreases). Therefore after a number of flips, we obtain a fiber type contraction $`\phi :XZ`$. Since $`\rho (X)=\rho 2`$, $`dim(Z)=1`$ or $`2`$. Note that all varieties from Theorem 1.2 have no small contractions. Thus, it is sufficient to prove Proposition 2.4 on our new model $`(X,D)`$.
This procedure does not work if $`(X,D)`$ is not klt. The difference is that contractions of components of $`D`$ do not contradict the inductive hypothesis. If $`(X,D)`$ is not klt, then we run $`(K_X+DD)`$-MMP. Note that $`D`$ is normal and irreducible \[4, 17.5\]. For every extremal ray $`R`$ we have $`DR>0`$, so we cannot contract an irreducible component of $`D`$. Therefore after every divisorial contraction $`d_i\rho (X)`$ decrease, a contradiction with our assumption. Thus, all steps of the MMP are flips. By \[4, 2.28\], they preserve the plt property of $`K+D`$. At the end we get a fiber type contraction $`\phi :XZ`$, where $`dim(Z)<3`$ and $`D`$ is $`\phi `$-ample (i.e. $`D^{\mathrm{hor}}0`$). Since $`D^{\mathrm{hor}}`$ has a component which intersects all components of $`D^{\mathrm{ver}}`$, $`D^{\mathrm{ver}}=0`$.
### 4.4. Case: $`dim(Z)=1`$
Then $`\rho (X)=2`$. By Lemma 3.8 and our assumption (4.2), we have $`_{\mathrm{hor}}d_i3`$ and $`_{\mathrm{ver}}d_i2`$. In particular, $`D^{\mathrm{ver}}0`$. Components of $`D^{\mathrm{ver}}`$ are fibers of $`\phi `$, so they are numerically proportional. Clearly, the log divisor $`K_X+D^{\mathrm{hor}}D^{\mathrm{ver}}`$ is not nef and curves in fibers of $`\phi `$ are trivial with respect to it. Let $`Q`$ be the extremal $`(K_X+D^{\mathrm{hor}})`$-negative ray of $`\overline{NE}(X)^2`$ and let $`\varphi :XW`$ be its contraction. It follows by Lemma 3.3 that $`\varphi `$ cannot be a flipping contraction. Let $`\mathrm{}`$ be a general curve such that $`\varphi (\mathrm{})=\mathrm{pt}`$. Then $`\mathrm{}`$ dominates $`Z`$ and $`\mathrm{}^1`$. Hence $`Z^1`$.
#### 4.4.1. Subcase $`D=0`$
We will prove that $`X^2\times ^1`$. By our inductive hypothesis, $`\varphi `$ cannot be divisorial. Therefore $`dim(W)=2`$. Further, $`D^{\mathrm{ver}}\mathrm{}D\mathrm{}=K_X\mathrm{}=2`$. Since $`\mathrm{}`$ intersects all components of $`D^{\mathrm{ver}}`$, $`_{\mathrm{ver}}d_i2`$. This yields $`_{\mathrm{ver}}d_i=2`$ and $`_{\mathrm{hor}}d_i=3`$. In particular, this proves inequality (2.5). Moreover, $`\mathrm{}D^{\mathrm{ver}}=2`$, $`\mathrm{}D^{\mathrm{hor}}=0`$ and for any component $`D_i`$ of $`D^{\mathrm{ver}}`$ we have $`D_i\mathrm{}=1`$. Fix two components of $`D^{\mathrm{ver}}`$, say $`D_0`$ and $`D_1`$. Then $`K_X+D_0+D_1+D^{\mathrm{hor}}K_X+D0`$, so $`(X,D_0+D_1+D^{\mathrm{hor}})`$ is plt by Lemma 3.9. Applying Lemma 3.4 we obtain $`D_0D_1Z^2`$, $`X`$ is smooth and $`\varphi `$ is a $`^1`$-bundle. By \[6, 3.5\], $`\phi `$ is a $`^2`$-bundle. We have a finite morphism $`\phi \times \varphi :XZ\times W=^1\times ^2`$. Clearly, $`\mathrm{deg}(\phi \times \varphi )=\phi ^1(\mathrm{pt})\mathrm{}=1`$. Hence $`\phi \times \varphi `$ is an isomorphism.
#### 4.4.2. Subcase: $`D0`$
Since $`\rho (X/Z)=1`$, $`D`$ is irreducible. Put $`S:=D`$. Let $`F`$ be a general fiber of $`\phi `$. By construction, $`K_X`$ is $`\phi `$-ample. First, assume that $`dim(W)=2`$. Then $`D^{\mathrm{ver}}\mathrm{}D\mathrm{}=K_X\mathrm{}=2`$. Since $`\mathrm{}`$ intersects all components of $`D^{\mathrm{ver}}`$, $`_{\mathrm{ver}}d_i2`$. This yields $`_{\mathrm{ver}}d_i=2`$, $`_{\mathrm{hor}}d_i=3`$ and $`d_i=5`$. Moreover, $`D^{\mathrm{hor}}\mathrm{}=0`$. By Lemma 3.8, $`F^2`$, $`X`$ is smooth along $`F`$ and for any component $`D_i`$ of $`D^{\mathrm{hor}}`$ the scheme-theoretic restriction $`D_i|_F`$ is a line. Hence components of $`D^{\mathrm{hor}}`$ are numerically equivalent. Let $`D_1`$ be a component of $`D^{\mathrm{hor}}S`$. Consider the new boundary $`D^{}:=D+\epsilon D_0\epsilon S`$. If $`0<\epsilon 1`$, then $`(X,D^{})`$ is klt and $`K_X+D^{}0`$. Applying Case 4.4.1 we get $`W^2`$ and $`X^2\times ^1`$.
Now assume that $`\varphi `$ is divisorial. By the inductive hypothesis, $`\varphi `$ contract $`S`$. Since the contraction is extremal, $`\varphi (S)`$ is a curve (otherwise curves $`S\phi ^1(\mathrm{pt})`$ is contracted by $`\varphi `$ and $`\phi `$). All components of $`\varphi (D^{\mathrm{ver}})`$ pass through $`\varphi (S)`$. By taking a general hyperplane section as in the proof of Lemma 3.3, we obtain $`_{\mathrm{ver}}d_i2`$. By Corollary 3.8.1, we obtain that $`S^1\times ^1`$, $`X`$ has only isolated singularities and $`X`$ is smooth along $`S`$. By Lemma 3.8, $`F^2`$ and $`X`$ is smooth along $`F`$. The curve $`FS`$ is ample on $`F`$, so it is connected and smooth by the Bertini theorem. Therefore $`FS`$ is a is a generator of $`S=^1\times ^1`$. Since $`\phi |_S`$ is flat, he same holds for arbitrary fiber $`F_0`$. Hence all fibers of $`\phi `$ are numerically equivalent and any fiber $`F_0`$ contains an ample smooth rational curve. Moreover, this also means $`F_0`$ is not multiple. Thus it is a normal surface. Now as in Case 4.4.1, $`K_X+F_0+F_1+D^{\mathrm{hor}}0`$ and by Lemma 3.9, $`(X,F_0+F_1+D^{\mathrm{hor}})`$ is plt for any fibers $`F_0`$, $`F_1`$. By Adjunction, $`(F_0,D^{\mathrm{hor}}|_{F_0})`$ is klt. Clearly, $`K_{F_0}K_X|_{F_0}`$ and $`S|_{F_0}`$ are numerically proportional. Hence $`F_0`$ is a log del Pezzo surface of Fano index $`>1`$. Since $`\phi `$ is flat, $`\left(K_{F_0}\right)^2=\left(K_F\right)^2=9`$. Therefore, $`F_0^2`$ and $`X`$ is smooth. By \[6, 3.5\], $`\phi `$ is a $`^2`$-bundle so, $`X_^1()`$, where $`=𝒪_^1𝒪_^1(a)𝒪_^1(b)`$, $`0ab`$. The Grothendiek tautological bundle $`𝒪_{()}(1)`$ is generated by global sections and not ample. Therefore $`𝒪_{()}(1)`$ gives us a supporting function for the extremal ray $`Q`$. Since $`\varphi `$ is birational, $`𝒪_{()}(1)^3=a+b>0`$. Finally, $`X`$ contains $`S=^1\times ^1`$. Hence $`a=0`$. This proves Proposition 2.4 in the case when $`Z`$ is a curve.
### 4.5. Case: $`dim(Z)=2`$
Note that $`Z`$ has only log terminal singularities (see e. g. \[4, 15.11\]). Since $`K_X`$ is $`\phi `$-ample, a general fiber $`\mathrm{}`$ of $`\phi `$ is $`^1`$. Hence $`2=K_X\mathrm{}=D\mathrm{}=D^{\mathrm{hor}}\mathrm{}_{\mathrm{hor}}d_i`$. By our assumption, $`_{\mathrm{ver}}d_i\rho (X)+1`$. If $`(X,D)`$ is not plt, then $`D`$ is $`\phi `$-ample. Clearly, $`D^{\mathrm{ver}}=0`$.
###### Claim 4.5.1.
Notation as above. Then $`K_Z`$ is not nef.
###### Proof.
Run $`(K_X+D^{\mathrm{ver}})`$-MMP. After a number of flips we get either a divisorial contraction (of the proper transform of a component of $`D`$), or a fiber type contraction. In both cases $`Z`$ is dominated by a family of rational curves \[2, 5-1-4, 5-1-8\]. Therefore $`K_Z`$ is not nef by . ∎
###### Claim 4.5.2.
Notation as in 4.5. Then $`Z`$ contains no contractible curves. In particular, $`\rho (Z)2`$.
###### Proof.
Assume the converse. i. e. there is an irreducible curve $`\mathrm{\Gamma }Z`$ and a birational contraction $`\mu :ZZ^{\prime \prime }`$ such that $`\mu (\mathrm{\Gamma })=\mathrm{pt}`$ and $`\rho (Z/Z^{\prime \prime })=1`$. Denote $`F:=\phi ^1(\mathrm{\Gamma })`$. Since $`FD`$, $`(X,D+\epsilon F)`$ is plt for $`0<\epsilon 1`$ \[4, 2.17\]. First we assume that $`\mathrm{\Gamma }`$ is contractible: Run $`(K+D+\epsilon F)`$-MMP over $`Z^{\prime \prime }`$. By our inductive hypothesis, there are no divisorial contractions (because such a contraction must contract $`F`$). At the end we cannot get a fiber type contraction (because $`K+D+\epsilon F\epsilon F`$ cannot be anti-ample over a lower-dimensional variety). Thus after a number of flips $`XX^{}`$, we get a model $`X^{}`$ over $`Z^{\prime \prime }`$ such that $`K_X^{}+D^{}+\epsilon F^{}\epsilon F^{}`$ is nef over $`Z^{\prime \prime }`$, where $`D^{}`$ and $`F^{}`$ are proper transforms of $`D`$ and $`F`$, respectively. Then $`F^{}0`$ (because $`F0`$). Let $`\mathrm{}^{}`$ be the proper transform of a general fiber of $`\phi `$. Since $`F^{}`$ is nef over $`Z^{\prime \prime }`$, $`F^{}\mathrm{}^{}=0`$ and $`\rho (X^{}/Z^{\prime \prime })=2`$, we obtain that $`\mathrm{}^{}`$ generates an extremal ray of $`\overline{NE}(X^{}/Z^{\prime \prime })`$. Let $`\phi ^{}:X^{}Z^{}`$ be its contraction and let $`\mu ^{}:Z^{}Z^{\prime \prime }`$ be the natural map. Then $`dim(Z^{})=2`$, $`\mathrm{\Gamma }^{}:=\phi ^{}(F^{})`$ is a curve and $`\mu ^{}(\mathrm{\Gamma }^{})=\mu (\phi (F))=\mu (\mathrm{\Gamma })=\mathrm{pt}`$. Therefore $`\left(\mathrm{\Gamma }^{}\right)^2<0`$. On the other hand, $`\left(\mathrm{\Gamma }^{}\right)^20`$, which is a contradiction. Indeed, Let $`C^{}F^{}`$ be any curve such that $`\phi ^{}(C^{})=\mathrm{\Gamma }^{}`$. Then $`C^{}F^{}0`$. By the projection formula, $`\left(\mathrm{\Gamma }^{}\right)^20`$.
Since $`K_Z`$ is not nef, there is an extremal contraction $`\psi :ZV`$. By the above it is not birational. Therefore $`dim(V)=1`$ and $`\rho (Z)=2`$. ∎
###### Corollary 4.5.3.
Notation as in 4.5. One of the following holds:
1. $`\rho (Z)=1`$ and $`K_Z`$ is ample;
2. $`\rho (Z)=2`$ and there is a $`K_Z`$-negative extremal contraction $`\psi :ZV`$ onto a curve.
#### 4.5.4. Subcase $`D=0`$
Let $`D_i`$ be a component of $`D^{\mathrm{ver}}`$. Run $`(K+Dd_iD_i)`$-MMP:
$$\chi _{(i)}:XX^{(i)}.$$
As above we get a fiber type contraction $`\phi _{(i)}:X^{(i)}Z^{(i)}`$. Notations $`D^{\mathrm{ver}}`$ and $`D^{\mathrm{hor}}`$ will be fixed with respect to our original $`\phi `$. If $`dim(Z^{(i)})=1`$, then replacing $`X`$ with $`X^{(i)}`$ we get the case $`dim(Z)=1`$ above. Thus we can assume that $`dim(Z^{(i)})=2`$ for any chose of $`D_i`$. Let $`\mathrm{}^{(i)}X^{(i)}`$ be a general fiber of $`\phi _{(i)}`$ and let $`L^{(i)}X`$ be its proper transform. Clearly, $`\chi _{(i)}`$ is an isomorphism along $`L^{(i)}`$. Hence $`K_XL^{(i)}=2`$, $`L^{(i)}`$ is nef and $`D_iL^{(i)}>0`$. For $`i=1,\mathrm{},r`$ we get rational curves $`L^{(1)},\mathrm{},L^{(r)}`$. We shift indexing so that $`X=X^{(0)}`$ and put $`Z=Z^{(0)}`$ and $`\phi =\phi _{(0)}`$.
Up to permutations we can take $`L^{(0)},\mathrm{},L^{(s)}`$, $`s+1r`$ to be linearly independent in $`N_1(X)`$. Then for any $`D_i`$ there exists $`L^{(j)}`$ such that $`D_iL^{(j)}>0`$. Thus we have
$$\begin{array}{c}2(s+1)=K_X\underset{j=0}{\overset{s}{}}L^{(j)}=D\underset{j=0}{\overset{s}{}}L^{(j)}=\hfill \\ \hfill \underset{i=1}{\overset{r}{}}d_i\left(D_i\underset{j=0}{\overset{s}{}}L^{(j)}\right)\underset{i=1}{\overset{r}{}}d_i\rho (X)+3.\end{array}$$
Since $`3\rho (Z)+1=\rho (X)s+1`$, this yields $`\rho (X)=s+1=3`$. Thus,
(4.6)
$$D_i\underset{j=0}{\overset{2}{}}L^{(j)}=1$$
holds for all $`i`$. Moreover, $`L^{(0)},L^{(1)},L^{(2)}`$ generate $`N_1(X)`$ and components of $`D`$ generate $`N^1(X)`$.
Taking into account that $`2=K_XL^{(j)}=DL^{(j)}`$, we decompose $`D`$ into the sum $`D=D^{(0)}+D^{(1)}+D^{(2)}`$ of effective $``$-divisors without common components so that
(4.7)
$$D^{(i)}L^{(j)}=\{\begin{array}{ccc}0\hfill & \text{if}\hfill & ij\hfill \\ 2\hfill & \text{otherwise}\hfill & \end{array}$$
Then $`D^{(i)}=\phi ^{}\mathrm{\Delta }^{(i)}`$ for $`i=1,2`$, where $`\mathrm{\Delta }^{(1)}`$, $`\mathrm{\Delta }^{(2)}`$ are effective $``$-divisors on $`Z`$. Put $`C^{(i)}:=\phi (L^{(i)})`$, $`i=1,2`$. Since families $`L^{(j)}`$ are dense on $`X`$, $`C^{(j)}`$ are nef and $`0`$. By the projection formula,
$$\mathrm{\Delta }^{(i)}C^{(j)}\{\begin{array}{ccc}=0\hfill & \text{if}\hfill & 1ij2\hfill \\ >0\hfill & \text{if}\hfill & 1i=j2\hfill \end{array}$$
Hence $`\mathrm{\Delta }^{(1)}`$ and $`\mathrm{\Delta }^{(2)}`$ generate extremal rays of $`\overline{NE}(Z)^2`$. By (4.7), these $``$-divisors have more than one component, so they are nef and $`\left(\mathrm{\Delta }^{(1)}\right)^2=\left(\mathrm{\Delta }^{(2)}\right)^2=0`$. This gives us that $`C^{(1)}`$ and $`C^{(2)}`$ also generate extremal rays. Therefore $`C^{(i)}`$ and $`\mathrm{\Delta }^{(j)}`$ are numerically proportional whenever $`ij`$ and $`\left(C^{(1)}\right)^2=\left(C^{(2)}\right)^2=0`$. In particular, $`C^{(i)}`$, $`i=1,2`$ generate an one-dimensional base point free linear system which defines a contraction $`Z^1`$. This also shows that $`D^{(i)}=\phi ^{}\mathrm{\Delta }^{(i)}`$, $`i=1,2`$ are nef on $`X`$.
Now we claim that $`D^{(0)}`$ is nef. Assume the opposite. Then for small $`\epsilon >0`$, $`(X,D+\epsilon D^{(0)})`$ is klt \[4, 2.17\]. There is a $`(K_X+D+\epsilon D^{(0)})`$-negative extremal ray, say $`R`$. By out inductive hypothesis, the contraction of $`R`$ must be of flipping type. Since $`\mathrm{\Delta }^{(1)}`$, $`\mathrm{\Delta }^{(2)}`$ generate $`N^1(Z)`$, we have $`D^{(i)}R>0`$ for $`i=1`$ or $`2`$. By (4.7), $`_j^{(i)}d_j=2`$, where $`_j^{(i)}`$ runs through all components $`D_j`$ of $`D^{(i)}`$. Since components of $`D^{(i)}=\phi ^{}(\mathrm{\Delta }^{(i)})`$ are numerically proportional, this contradicts Lemma 3.3. Therefore $`D^{(i)}`$ are nef for $`i=0,1,2`$.
We claim that $`L^{(i)}`$, $`i=0,1,2`$ generate $`\overline{NE}(X)`$. Indeed, let $`z\overline{NE}(X)`$ be any element. Then $`z\alpha _i[L^{(i)}]`$ for $`\alpha _i`$. By (4.7), $`0D^{(j)}z=\alpha _j`$. This shows that $`L^{(i)}`$ generate $`\overline{NE}(X)`$. From (4.6) we see that components of $`D^{(0)}`$ are numerically equivalent.
Fix two components $`D^{}`$ and $`D^{\prime \prime }`$ of $`D^{(0)}`$. Then $`K_X+D^{}+D^{\prime \prime }+D^{(1)}+D^{(2)}0`$. By Lemma 3.9, $`(X,D^{}+D^{\prime \prime }+D^{(1)}+D^{(2)})`$ is plt. Lemma 3.4 implies that $`D^{}D^{\prime \prime }S^1\times ^1`$, $`X`$ is smooth and $`\phi `$ is a $`^1`$-bundle. As in the case $`dim(Z)=1`$, we have $`X^1\times ^1\times ^1`$.
#### 4.7.5. Subcase $`D0`$
Let $`S`$ be a component of $`D`$. Clearly, $`S\mathrm{}2`$. If $`S`$ is generically a section of $`\phi `$, then by Lemma 3.4, $`X`$ is smooth, $`\phi `$ is a $`^1`$-bundle and $`S^2`$ or $`^1\times ^1`$. Therefore $`X()`$, where $``$ is a rank two vector bundle on $`Z`$. Since $`\phi `$ has disjoint sections, $``$ is decomposable. So we may assume that $`=𝒪_Z+`$, where $``$ is a line bundle. By the projection formula, all components of $`D^{\mathrm{ver}}`$ are nef. Let $`R`$ be a $`(K_X+D^{\mathrm{hor}})`$-negative extremal curve and let $`\varphi :XW`$ be its contraction. Assume that $`\varphi `$ is of flipping type. By , $`K_XR0`$. Hence $`D^{\mathrm{hor}}R<0`$, so $`R`$ is contained in a section of $`\phi `$. But all curves on $`^2`$ and $`^1\times ^1`$ are movable, a contradiction. If $`\varphi `$ is of fiber type, then as in the case $`D=0`$ we get $`XZ\times ^1`$. Assume that $`\varphi `$ is of divisorial type. By inductive hypothesis, $`\varphi `$ contracts a component of $`D`$.
Finally, consider the case when $`\phi |_S:SZ`$ is generically finite of degree $`2`$. Obviously, $`D^{\mathrm{hor}}=S`$. If $`\rho (Z)=1`$, then all components of $`D^{\mathrm{ver}}`$ are numerically proportional and $`_{\mathrm{ver}}d_i4`$. If $`dim(W)=2`$, then $`D^{\mathrm{ver}}\varphi ^1(w)2`$ for general $`wW`$. Hence $`_{\mathrm{ver}}d_i4`$, a contradiction. Then by lemmas 3.3 and 3.8, $`\varphi `$ is divisorial and $`\varphi `$ must contract $`S`$. We derive a contradiction with Theorem 2.3 for $`(W,\varphi (D^{\mathrm{ver}}))`$ near $`\varphi (S)`$.
Therefore $`\rho (Z)=2`$ and there is a $`K_Z`$-negative extremal contraction $`\psi :ZV`$ onto a curve (see Corollary4.5.3). Let $`\pi :X\stackrel{\phi }{}Z\stackrel{\psi }{}V`$ be the composition map. Clearly, all fibers of $`\pi `$ are irreducible. Write $`D=^{}d_iD_i+^{\prime \prime }d_iD_i=D^{}+D^{\prime \prime }`$, where $`^{}`$ (respectively $`^{\prime \prime }`$) runs through all components $`D_i`$ such that $`\pi (D_i)=\mathrm{pt}`$ (respectively $`\varphi (D_i)=V`$). Thus, $`S`$ is a component of $`D^{\prime \prime }`$ and components of $`D^{}`$ are numerical proportional. Let $`F`$ be a general fiber. Then $`\rho (F)=2`$. Consider the contraction $`\phi |_F:F\phi (F)`$ and denote $`D^{\prime \prime }|_F=D|_F`$ by $`\mathrm{\Phi }=\alpha _i\mathrm{\Phi }_i`$. Then $`(F,\mathrm{\Phi })`$ is plt and $`K_F+\mathrm{\Phi }0`$. Clearly, the curve $`S|_F=\mathrm{\Phi }`$ is a $`2`$-section and components of $`\mathrm{\Phi }\mathrm{\Phi }`$ are fibers of $`\phi |_F`$. As in the proof of Lemma 3.8 using the fact that $`S|_F`$ intersects components of $`\mathrm{\Phi }\mathrm{\Phi }`$ twice, one can check $`\alpha _i<3`$. This yields $`^{\prime \prime }d_i<3`$ and $`^{}d_i>2`$. Let $`R`$ be a $`(K_X+D^{\prime \prime })`$-negative extremal ray. Since $`^{}d_i>2`$ and $`\rho (X)>2`$, $`R`$ cannot be fiber type. According to Lemma 3.3 $`R`$ also cannot be flipping type. Therefore $`R`$ is divisorial and contract $`S`$ to a point. Since $`S`$ intersects all components of $`D^{\mathrm{ver}}`$, this contradicts Theorem local. The proof of Proposition 2.4 is finished.
#### Concluding remark
In contrast with the purely log terminal case we have no complete results in the log canonical case. The reason is that the steps of MMP are not so simple. In particular, we can have divisorial contractions which contract components of $`D`$. |
warning/0001/cond-mat0001153.html | ar5iv | text | # A hybrid functional for the exchange-correlation kernel in time-dependent density functional theory
## I Introduction
Ground-state density functional theory is well-established as an inexpensive alternative to traditional ab initio quantum chemical methods. Now time-dependent density functional theory (TDDFT) is rapidly emerging as an inexpensive accurate method for the calculation of electronic excitation energies in quantum chemistry. Calculation of dynamic response properties using TDDFT has a long history, since the pioneering work of Zangwill and Soven. It is only relatively recently that attention has been focussed on the direct extraction of excitation energies. Already, this method has been implemented in several quantum chemistry packages, such as deMon, Turbomole , ADF, and QCHEM. Important calculations include the calculation of excited-state crossings in formaldehyde, excitations with significant doubly-excited character, the photospectrum of chlorophyll A, and the response of 2-D quantum strips.
How are excitation energies calculated using TDDFT? First, a self-consistent ground-state Kohn-Sham calculation is performed, using some approximation for the exchange-correlation energy $`E_{\mathrm{XC}}`$, such as B3LYP or PBE. This yields a set of Kohn-Sham eigenvalues $`ϵ_i`$ and orbitals $`\varphi _i`$. Even with the exact ground-state energy functional and potential, these eigenvalues are in general not the true excitations of the system, but are closely related. In a second step, the central equation of TDDFT response theory is solved, which extracts the true linear response function from its Kohn-Sham counterpart. This equation includes a second unknown functional, the exchange-correlation kernel $`f_{\mathrm{XC}}(𝐫,𝐫^{};\omega )`$, which is the Fourier transform of the functional derivative of the time-dependent exchange-correlation potential. The poles of the exact response function are shifted from those of the KS function, and occur at the true excitations of the system. These steps are typically repeated for several nuclear positions.
The success of any density functional method, however, depends on the quality of the approximate functionals employed. The above calculation requires two distinct density functional approximations: one for the ground-state energy, which implies a corresponding approximation for the exchange-correlation potential $`v_{\mathrm{XC}}(𝐫)=\delta E_{\mathrm{XC}}/\delta \rho (𝐫)`$, and a second for the exchange-correlation kernel. Most calculations now appearing in the chemical literature use the adiabatic local density approximation (ALDA) for $`f_{\mathrm{XC}}`$. Adiabatic implies that the frequency-dependence of $`f_{\mathrm{XC}}`$ is ignored, and its $`\omega =0`$ value used, and LDA implies $`f_{\mathrm{XC}}(𝐫,𝐫^{})=\delta v_{\mathrm{XC}}(𝐫)/\delta \rho (𝐫^{})`$, where $`v_{\mathrm{XC}}`$ is the LDA potential from ground-state LDA calculations.
The more drastic of the two approximations is that for the ground-state. Very often, functionals which yield accurate ground-state energies have very poor-looking potentials. (How this can happen can be understood by considering the virial theorem, which relates energies to potentials). These poor potentials then have in turn badly behaved virtual orbitals. In particular, local and semi-local approximations, i.e., generalized gradient approximations, fail to capture the correct asymptotic behavior of the potential, and many virtual states are not even bound. Even hybrid functionals, such as B3LYP and PBE0, do not much improve the asymptotic behavior, since they only mix a fraction of exact exchange. This restricts calculations within these approximations to only low-lying excitations. This difficulty is most pronounced in atoms, becomes smaller for bigger molecules, and is irrelevant for bulk solids. Recently, we have shown that even with approximations which are free from self-interaction error, such as exact exchange, or self-interaction corrected LDA, and which therefore reproduce the dominant part of the asymptotic decay of the potential, inaccuracies in the Kohn-Sham eigenvalues (mostly due to the incorrect position of the highest occupied level) dominate over any errors introduced in the second step, namely the correction of KS levels to the true levels.
The present work studies the accuracy of approximations to $`f_{\mathrm{XC}}`$ alone. This is because, for low-lying states, errors in $`f_{\mathrm{XC}}`$ must be disentangled from errors in the KS energy levels, and also because many people are working to improve approximations to the exact ground-state potential, which we hope will ultimately reduce those errors discussed above. Hence all our calculations are performed using the exact Kohn-Sham potentials of the He and Be atoms, for which we thank Cyrus Umrigar. We study a variety of approximations to $`f_{\mathrm{XC}}`$, all adiabatic. We focus especially on the spin-decomposition of such approximations, which determines the relative positions of singlet and triplet excitations in TDDFT. We use the single-pole approximation for the susceptibility to directly relate errors in $`f_{\mathrm{XC}}`$ to errors in excitation energies. We find that exact exchange works well for parallel spins, which determines the mean energy of the singlet and triplet, while the antiparallel contribution to $`f_{\mathrm{XC}}`$ determines their splitting, which is well-approximated in ALDA. With this insight, we construct a hybrid of exact exchange and ALDA, which greatly improves results for He, and moderately improves them for Be. We use atomic units ($`e^2=\mathrm{}=m_e=1`$) throughout, except in Fig. 1.
## II Methodology
The basic response equation of TDDFT has the same form as that of time-dependent Hartree-Fock theory, or of the Random Phase Approximation, i.e.,
$`\chi ^{\sigma \sigma ^{}}(𝐫,𝐫^{};\omega )`$ $`=`$ $`\chi _\mathrm{S}^{\sigma \sigma ^{}}(𝐫,𝐫^{};\omega )`$ (1)
$`+`$ $`{\displaystyle \underset{\sigma ^{\prime \prime }\sigma ^{\prime \prime \prime }}{}}{\displaystyle d^3r^{\prime \prime }d^3r^{\prime \prime \prime }\chi _\mathrm{S}^{\sigma \sigma ^{\prime \prime }}(𝐫,𝐫^{\prime \prime };\omega )}`$ (3)
$`f_{\mathrm{HXC}}^{\sigma ^{\prime \prime }\sigma ^{\prime \prime \prime }}(𝐫^{\prime \prime },𝐫^{\prime \prime \prime };\omega )\chi ^{\sigma ^{\prime \prime \prime }\sigma ^{}}(𝐫^{\prime \prime \prime },𝐫^{};\omega )`$
where $`\chi `$ is the exact frequency-dependent susceptibility of the system, while $`\chi _\mathrm{S}`$ is its Kohn-Sham analog, and $`f_{\mathrm{HXC}}^{\sigma \sigma ^{}}(𝐫,𝐫^{})=1/|𝐫𝐫^{}|+f_{\mathrm{XC}}^{\sigma \sigma ^{}}(𝐫,𝐫^{})`$. The $`\sigma `$ indices denote spin, i.e., $`\sigma =`$ or $``$. By various means, the poles of $`\chi `$ as a function of $`\omega `$ can be found. Our method for finding these poles is to consider only the discrete poles, i.e., those corresponding to bound states. In the particular case of a frequency-independent model for $`f_{\mathrm{XC}}`$, we can show that these poles occur at the eigenvalues of the matrix
$`M_{qq^{}}^{\sigma \sigma ^{}}`$ $`=`$ $`\delta _{qq^{}}^{\sigma \sigma ^{}}\omega _{q\sigma }+\alpha _{q^{}\sigma ^{}}{\displaystyle }d^3r{\displaystyle }d^3r^{}\mathrm{\Phi }_{q\sigma }^{}(𝐫)\times `$ (5)
$`f_{\mathrm{HXC}}^{\sigma \sigma ^{}}(𝐫,𝐫^{})\mathrm{\Phi }_{q^{}\sigma ^{}}(𝐫^{}).`$
For notational brevity, we have used double indices $`q(j,k)`$ to characterize the excitation energy $`\omega _{q\sigma }ϵ_{j\sigma }ϵ_{k\sigma }`$ of the single-particle transition $`(k\sigma j\sigma )`$. Consequently, we set $`\alpha _{q\sigma }:=f_{k\sigma }f_{j\sigma }`$, where $`f_{k\sigma }`$ is the occupation number of that orbital, and $`\mathrm{\Phi }_{q\sigma }(𝐫)=\phi _{k\sigma }^{}(𝐫)\phi _{j\sigma }(𝐫)`$. While we do include sufficient bound-state poles to converge to an accurate result, our method does neglect continuum contributions, and this effect will be discussed in the next section.
All approximations we study for $`f_{\mathrm{XC}}`$ are adiabatic. The most ubiquitous is ALDA (or more precisley, the adiabatic local spin density approximation) in which
$$f_{\mathrm{XC}}^{\sigma \sigma ^{}\mathrm{ALDA}}(𝐫,𝐫^{})=\delta (𝐫𝐫^{})\frac{^2e_{\mathrm{XC}}^{\mathrm{unif}}(n_{},n_{})}{n_\sigma n_\sigma ^{}}|_{n_{}(𝐫),n_{}(𝐫^{})}.$$
(6)
Note that this leads to a completely short-ranged approximation to $`f_{\mathrm{XC}}`$. Similarly, any adiabatic GGA approximation leads to an approximate $`f_{\mathrm{XC}}`$ which is almost as short-ranged.
A second distinct approximation to $`f_{\mathrm{XC}}`$ is in terms of its (usually) dominant exchange contribution. A highly accurate approximation to the exact exchange-only equations of ground-state density functional theory (the optimized effective potential equations) was introduced by Krieger, Li, and Iafrate. This approximation has been extended to the time-dependent case:
$$f_\mathrm{X}^{\sigma \sigma ^{}}(𝐫,𝐫^{})=\delta _{\sigma \sigma ^{}}\frac{\left|\underset{k}{}f_{k\sigma }\phi _{k\sigma }(𝐫)\phi _{k\sigma }^{}(𝐫^{})\right|^2}{n_\sigma (𝐫)|𝐫𝐫^{}|n_\sigma (𝐫^{})}.$$
(7)
This is exact for one electron, and for (spin-unpolarized) two-electron exchange. Note that this approximation has a long-ranged contribution, which can cancel exactly the direct hartree contribution to the matrix $`𝐌`$ in Eq. (5).
A third approximation which we tried is the self-interaction corrected (SIC) ALDA, which is simply the second functional derivative of the SIC-LDA energy:
$$E_{\mathrm{XC}}^{\mathrm{SIC}\mathrm{LDA}}=E_{\mathrm{XC}}^{\mathrm{LDA}}+\underset{i\sigma }{}(E_\mathrm{X}[n_{i\sigma }]E_{\mathrm{XC}}^{\mathrm{LDA}}[n_{i\sigma }]),$$
(8)
where $`n_{i\sigma }=|\phi _{i\sigma }|^2`$ is the density of a single-orbital, and $`E_\mathrm{X}[n_{i\sigma }]`$ is the exact Hartree self-interaction energy of that orbital. This approximation should improve over ALDA, in avoiding spurious self-interaction errors, and over just exchange, by including some correlation.
## III Data
In this section, we report calculations for the He and Be atoms using the exact ground-state Kohn-Sham potentials, the three approximations to the kernel mentioned in the previous section, and including many bound-state poles in Eq. (5), but neglecting the continuum. The technical details are given in Ref. . Table I lists the results, which are compared with a highly accurate nonrelativistic variational calculations In each symmetry class (s, p, and d), up to 38 virtual states were calculated. The errors reported are absolute deviations from the exact values. The second error under the Be atom excludes the $`2s2p`$ transition, for reasons discussed in the next section.
The effect of neglecting continuum states in these calculations has been investigated by van Gisbergen et al., who performed ALDA calculations from the exact Kohn-Sham potential in a localized basis set. These calculations were done including first only bound states, yielding results identical to those presented here, and then including all positive energy orbitals allowed by their basis set. They found significant improvement in He singlet-singlet excitations, especially for $`1s2s`$ and $`1s3s`$. Other excitations barely changed. Assuming inclusion of the continuum affects results with other approximate kernels similarly, these results do not change the basic reasoning and conclusions presented below, but suggest that calculations including the continuum may prove to be more accurate than those presented here.
## IV Single-pole analysis
The simplest truncation of the eigenvalue equation (5) for the excitation energies is to ignore all coupling between poles, except that between a singlet-triplet pair. This is equivalent to setting $`q|f_{\mathrm{HXC}}|q^{}`$ to zero, for $`qq^{}`$. (We have dropped the spin-index on these contributions, since we deal only with closed shell systems). Then the eigenvalue problem reduces to a simple $`2\times 2`$ problem, with solutions
$`\mathrm{\Omega }_q^+`$ $`=`$ $`\omega _q+2\mathrm{}q|f_{\mathrm{HXC}}q`$ (9)
$`\mathrm{\Omega }_q^{}`$ $`=`$ $`\omega _q+2\mathrm{}q|\mathrm{\Delta }f_{\mathrm{XC}}q,`$ (10)
where
$`f_{\mathrm{HXC}}`$ $`=`$ $`{\displaystyle \frac{1}{4}}{\displaystyle \underset{\sigma \sigma ^{}}{}}f_{\mathrm{HXC}}^{\sigma \sigma ^{}}={\displaystyle \frac{1}{|𝐫𝐫^{}|}}+{\displaystyle \frac{1}{2}}\left(f_{\mathrm{XC}}^{}+f_{\mathrm{XC}}^{}\right),`$ (11)
$`\mathrm{\Delta }f_{\mathrm{XC}}`$ $`=`$ $`{\displaystyle \frac{1}{4}}{\displaystyle \underset{\sigma \sigma ^{}}{}}\sigma \sigma ^{}f_{\mathrm{HXC}}^{\sigma \sigma ^{}}={\displaystyle \frac{1}{2}}\left(f_{\mathrm{XC}}^{}f_{\mathrm{XC}}^{}\right).`$ (12)
Thus $`f_{\mathrm{HXC}}`$ is the spin-summed contribution, which contributes to $`\chi `$, the spin-summed susceptibility, and therefore gives rise to the singlet level, while $`\mathrm{\Delta }f_{\mathrm{XC}}`$ is the spin-flip contribution, also called $`\mu _o^2G_{\mathrm{xc}}`$ in the theory of the frequency-dependent magnetization density . Thus even within the SPA, the KS degeneracy between singlets and triplets is broken, and we identify $`\mathrm{\Omega }_q^{}`$ with the triplet. In Table II, we report results within the single-pole approximation.
At this point, we notice the very strong shift in the Be $`2s2p`$ transition. This is due to the small magnitude of its transition energy, so that the pole of the $`2p`$ energy is very close to the pole of the $`2s`$ energy. Thus the single pole approximation is not expected to work well for this case, and it should be excluded from general statements based on the SPA.
### A Why are Kohn-Sham excitation energies so good?
We see throughout the data that the Kohn-Sham eigenvalues are always inbetween the exact singlet and triplet energy levels. The splitting is much larger for Be than for He, but this obervation is true in both cases. It has already been made by Filippi et al, and explained in terms of quasi-particle amplitudes. Here, we use the single-pole approximation to analyze this result in terms of the known behavior of density functionals. From Eq. (6) we find that the mean energy is given by
$$\overline{\mathrm{\Omega }}_q=\frac{1}{2}(\mathrm{\Omega }_q^++\mathrm{\Omega }_q^{})=\omega _q+\mathrm{}q|\frac{1}{|𝐫𝐫^{}|}+f_{\mathrm{XC}}^{}(𝐫,𝐫^{}|q,$$
(13)
while the energy splitting is given by
$$\mathrm{\Delta }\mathrm{\Omega }_q=(\mathrm{\Omega }_q^+\mathrm{\Omega }_q^{})=2\mathrm{}q|\frac{1}{|𝐫𝐫^{}|}+f_\mathrm{C}^{}q,$$
(14)
since there is no exchange contribution to antiparallel $`f_{\mathrm{XC}}`$. If we further define $`\delta \mathrm{\Omega }_q=\overline{\mathrm{\Omega }}_q\omega _q`$ as the deviation of the mean energy from the Kohn-Sham level, we see that
$$|\delta \mathrm{\Omega }_q|<\mathrm{\Delta }\mathrm{\Omega }_q/2$$
(15)
must be satisfied for the Kohn-Sham level to lie in between the singlet and triplet levels. Within the single-pole approximation, we have a very simple expression for the ratio of these two:
$$\frac{2\delta \mathrm{\Omega }_q}{\mathrm{\Delta }\mathrm{\Omega }_q}=\frac{\mathrm{}q|f_{\mathrm{HXC}}^{}q}{\mathrm{}q|f_{\mathrm{HC}}^{}q}$$
(16)
Consider first the He atom. For two electrons, $`f_\mathrm{X}^{}=1/|𝐫𝐫^{}|`$, exactly cancelling the Hartree term in Eq. (13), leaving only the parallel correlation contribution. (This is reflected in the X column on Table II, where the upshift of the singlet is equal to the downshift of the triplet.) Thus we find
$$\frac{2\delta \mathrm{\Omega }_q}{\mathrm{\Delta }\mathrm{\Omega }_q}=\frac{\mathrm{}q|f_\mathrm{C}^{}q}{\mathrm{}q|\frac{1}{|𝐫𝐫^{}|}+f_\mathrm{C}^{}q}(2\mathrm{e}\mathrm{l})$$
(17)
for two electrons in the single-pole approximation. It is well-known (see, e.g., Ref. ) that for ground-state energies, parallel correlation is much weaker than antiparallel, since antiparallel electrons are not kept apart by the exchange interaction. Thus this ratio is expected to stay well less than 1, as it does for all our He excitations. The effect of the single-pole approximation on this conclusion can be judged by studying the shift in the mean for the X results in Table I.
For Be, and any system with more than two electrons, there is still a good deal of cancellation of the exchange contribution with the direct contribution, but this cancellation is no longer exact. This can be seen in the Be results for X in Table II. By studying the form of $`f_\mathrm{X}`$ given in Eq. (4), we expect this remnant exchange contribution to be of order $`O((N2)/N^2)`$ for unpolarized systems. Thus in exchange-dominated (i.e., high density or weakly correlated) systems, the direct Coulomb term in the denominator will be larger than any remnant exchange term in the numerator. On the other hand, in low-density or strongly correlated systems, if antiparallel correlation continues to dominate over parallel correlation, this ratio will still be less than one. We conclude that the Kohn-Sham levels will usually be close to the true excitations (of single-particle nature).
### B Relation of exact exchange to Görling-Levy perturbation theory
Both time-dependent DFT and Görling-Levy perturbation theory are formally exact methods for extracting electronic excitation energies in density functional theory. In this section, we consider the expansion of the excitation energies in powers of the adiabatic coupling constant $`\lambda `$ to first order. This procedure should give identical results in both theories. Recently, Filippi et al have performed GL first order calculations for the He atom, using the exact Kohn-Sham potential. Their results are numerically identical to ours, using the exact exchange kernel, but only within the single-pole approximation, as given by Table II. Results calculated with the full (i.e., many poles) scheme, i.e., in Table I, differ slightly from theirs. This produces a paradox, in which the easier SPA is more accurate (apparently exact), while the more sophisticated treatment introduces errors.
The resolution of this paradox can be seen most easily in Eq. (1), the RPA-type equation for the susceptibility. Insertion of $`f_\mathrm{X}`$ alone (linear in $`\lambda `$) into these equations will lead to all powers of $`\lambda `$ being present in the solution, since it is a self-consistent integral equation. This is most easily seen by iterating the equations. A simple way to recover the exact first-order GL result is by solving the equations with $`\lambda f_\mathrm{X}`$, and making $`\lambda `$ very small, to find the linear contribution to the change in excitation energies. Insertion of $`\lambda =1`$ into this result will yield the exact GL result.
A far simpler method is to use the single-pole approximation. To see why this works, consider Eq. (2), our matrix whose eigenvalues are at the excitation energies. Since $`f_{\mathrm{HXC}}`$ is at least first-order in $`\lambda `$, all off-diagonal contributions are of O($`\lambda `$). Thus to lowest order, the diagonal dominates, and the off-diagonal corrections contribute $`O(\lambda ^2)`$ corrections. Retaining only diagonal contributions, i.e., the single-pole approximation, yields the exact result to first order in $`\lambda `$. A detailed functional derivation of this result has recently been given by Gonze and Scheffler.
### C A new hybrid functional for the kernel
To illustrate the importance of understanding the origin of errors in density functionals, we use the insight gained within the SPA in the previous sections to construct a hybrid functional for $`f_{\mathrm{XC}}`$. We then apply this both within SPA and in the full calculation. Our original idea, as mentioned in section II, was to use a simple self-interaction correction to produce a better approximation than either exact exchange or ALDA, but the SIC results
of Tables I and II show this does not happen.
Consider Fig. 1, which illustrates the positions of energy levels in the different schemes for the He atom. Our first step is to consider the mean energies. As pointed out in subsection A, these are determined by the parallel contributions to $`f_{\mathrm{XC}}`$ and the Hartree term. We have already seen how these two terms cancel exactly at the exchange level, so that the mean energy is very good in such a calculation. In ALDA, the cancellation (exact for He) of the exchange contributions is lost. This can be seen in the large shifts in the ALDA mean energies in Table II and in Fig. 1. Furthermore, the remaining small parallel-spin correlation contribution can be expected to be grossly overestimated by ALDA, since LDA ground-state correlation energies are usually too large by a factor of 2 or 3. Unfortunately, even SIC-ALDA is not exact for two-electron exchange, and it also suffers from a poor mean energy. Thus we recommend using only exact exchange for the parallel-spin contribution.
On the other hand, the splitting is determined solely by anti-parallel correlation contributions, and the direct term. Thus, an exact exchange treatment misses entirely the significant anti-parallel correlation contribution. This error is highlighted by the far too large splittings in the exchange results in Tables I and II, implying significant cancellation between the direct and antiparallel correlation contributions. So here we advocate use of ALDA. Since the splitting depends on anti-parallel contributions to $`f_{\mathrm{XC}}`$, but the SIC correction only applies to one spin at a time, SIC-ALDA has exactly the same splittings. Our recommended hybrid is therefore
$$f_{\mathrm{XC}}^{}=f_\mathrm{X}^{},f_{\mathrm{XC}}^{}=f_{\mathrm{XC}}^{\mathrm{ALDA}}.$$
(18)
Results shown in Tables I and II indicate that this hybrid decreases almost all errors over either exact exchange or ALDA. On average, the decrease is by about a factor of 3 for the He atom, but much less for Be (about 40%). Notable exceptions are the triplet transitions to $`p`$ states in He and in Be, where the error is increased.
## V Conclusions
We have shown how the results of TDDFT with approximate exchange-correlation kernel functionals may be understood in terms the well-known behavior of the ground-state functionals from which they are derived. We have shown in a simple case how a more accurate functional may be constructed from this insight. We regard this as an initial step toward an accurate approximation for $`f_{\mathrm{XC}}`$. Another obvious analytic tool would be the direct adiabatic decomposition of $`f_{\mathrm{XC}}`$ in terms of $`\lambda `$, which has proved so successful in understanding the hybrids commonly used in ground-state calculations.
###### Acknowledgements.
We gratefully acknowledge support from a NATO collaborative research grant. K.B. also acknowledges support of the Petroleum Research Fund and of NSF grant no. CHE-9875091. M.P. and E.K.U.G. also acknowledge partial support of the DFG. |
warning/0001/cond-mat0001234.html | ar5iv | text | # The effects of nuclear spins on the quantum relaxation of the magnetization for the molecular nanomagnet Fe8
\[
## Abstract
The strong influence of nuclear spins on resonant quantum tunneling in the molecular cluster Fe<sub>8</sub> is demonstrated for the first time by comparing the relaxation rate of the standard Fe<sub>8</sub> sample with two isotopic modified samples: (i) <sup>56</sup>Fe is replaced by <sup>57</sup>Fe, and (ii) a fraction of <sup>1</sup>H is replaced by <sup>2</sup>H. By using a recently developed ”hole digging” method, we measured an intrinsic broadening which is driven by the hyperfine fields. Our measurements are in good agreement with numerical hyperfine calculations. For $`T>`$ 1.5 K, the influence of nuclear spins on the relaxation rate is less important, suggesting that spin–phonon coupling dominates the relaxation rate.
\]
Mesoscopic quantum phenomena are actively investigated both for fundamental science and for future applications, for instance in quantum computing. Magnetic molecular clusters are among the most promising candidates to observe mesoscopic quantum phenomena . One of the most prominent examples is an octanuclear iron(III) cluster, called Fe<sub>8</sub> (Fig. 1), with a spin ground state of $`S=10`$ . Below 360 mK, the magnetization relaxes through a pure tunneling process giving rise to a stepped hysteresis cycle . Furthermore, the tunnel splitting $`\mathrm{\Delta }`$ of Fe<sub>8</sub> shows periodic oscillations when a transverse magnetic field is applied along the hard axis , a long searched phenomenon in magnetism associated with the Berry phase , and predicted several years before . Since $`\mathrm{\Delta }`$ is extremely small for the ground state tunneling, ca. $`10^7`$ K at $`H=0`$, the tunneling process should occur only in an extremely narrow magnetic field range, ca. $`10^8`$ T, and should be practically unobservable. However, a recent theory proposes that the tunneling is mediated by fluctuating hyperfine fields generated by magnetic nuclei , but direct experimental evidence is so far lacking.
In order to study the influence of nuclear spins, we increased the hyperfine coupling by the substitution of <sup>56</sup>Fe with <sup>57</sup>Fe, and decreased it by the substitution of <sup>1</sup>H with <sup>2</sup>H. We found that the relaxation rate of magnetization in the tunneling regime shows a clear isotope effect which we attribute to the changed hyperfine coupling.
The crystals of the standard Fe8 cluster, <sup>st</sup>Fe<sub>8</sub> or Fe<sub>8</sub>, \[Fe<sub>8</sub>(tacn)<sub>6</sub>O<sub>2</sub>(OH)<sub>12</sub>\]Br<sub>8</sub>.9H<sub>2</sub>O where tacn = 1,4,7- triazacyclononane, were prepared as reported by Wieghardt et al. . For the synthesis of the <sup>57</sup>Fe-enriched sample, <sup>57</sup>Fe<sub>8</sub>, a 13 mg foil of 95$`\%`$ inriched <sup>57</sup>Fe was dissolved in a few drops of HCl/HNO<sub>3</sub> (3 : 1) and the resulting solution was used as the iron source in the standard procedure. The <sup>2</sup>H-enriched Fe<sub>8</sub> sample, <sup>D</sup>Fe<sub>8</sub>, was crystallized from pyridine-d<sub>5</sub> and D<sub>2</sub>O (99$`\%`$) under an inert atmosphere at 5C by using a non-deuterated Fe(tacn)Cl<sub>3</sub> precursor. The amount of isotope exchange was not quantitatively evaluated, but it can be reasonably assumed that the H atoms of H<sub>2</sub>O and of the bridging OH groups, as well as a part of those of the NH groups of the tacn ligands are replaced by deuterium while the aliphatic hydrogens are essentially not affected. The crystalline materials were carefully checked by elemental analysis and single-crystal X-ray diffraction.
The magnetic measurements were made on single-crystal samples by using an array of micro-SQUIDs , which measure the magnetic field induced by the magnetization of the crystal. The advantage of this magnetometer lies mainly in its high sensitivity and fast response, allowing short-time measurements down to 1 ms. Furthermore the magnetic field can be changed rapidly and along any direction.
In order to avoid the influence of the crystal shape the relaxation was measured starting at an initial magnetization $`M_{\mathrm{init}}=0`$ where intermolecular dipolar interactions lead to a field distribution with a width of about 50 mT . A small field $`H_z`$ was then applied and the relaxation of magnetization was measured (inset of Fig. 2). For all three samples the relaxation was clearly non-exponential and could be adjusted to a $`\sqrt{\mathrm{\Gamma }_{\mathrm{sqrt}}t}`$law for the short time regime ($`t<100`$ s) for $`T<`$ 0.4 K. Fig. 3 displays the field dependance of $`\mathrm{\Gamma }_{\mathrm{sqrt}}(H_z)`$. The relaxation of the three samples at 40 mK are strikingly different from each other. The <sup>57</sup>Fe<sub>8</sub> sample is the fastest relaxing one whereas the <sup>D</sup>Fe<sub>8</sub> shows the slowest relaxation rate. As a complete theory of the relaxation behavior of crystals of molecular clusters is still missing , we plot in Fig. 2 the time needed to relax one percent of the saturation magnetization $`M_\mathrm{s}`$ as a function of the inverse temperature $`1/T`$. Relaxation and ac susceptibility measurements at $`T>`$ 1.5 K showed no clear difference between the three samples suggesting that above this temperature the relaxation is predominately due to spin-phonon coupling . Although the increased mass of the isotopes changes the spin–phonon coupling, we believe that this effect is small.
In principle, the change of masse does not change the crystalline field of the Fe ions, i.e. the anisotropy constants. Experimentally, this is confirmed with measurements below $`T<`$ 0.35 K, where spin–phonon coupling is negligible, by two observations: (i) relative positions of the resonances as a function of the longitudinal field $`H_z`$ are unchanged , and (ii) all three samples showed the same period of oscillation of $`\mathrm{\Delta }`$ as a function of the transverse field $`H_x`$ , a period which is very sensitive to any change of the anisotropy constants. Finally, we point out that the mass is increased in both isotopically modified samples whereas the effect on the relaxation rate is opposite.
A deeper insight into the relaxation mechanism can be achieved by using our recently developed hole digging method which allows us to estimate the hyperfine level broadening . Starting from the well defined magnetization state $`M_{\mathrm{init}}=0`$, and after applying a small field $`H_{\mathrm{dig}}`$, the sample is let to relax for a time $`t_{\mathrm{dig}}`$, called digging field and digging time, respectively. During the digging time, a small fraction $`\mathrm{\Delta }M_{\mathrm{dig}}`$ of the molecular spins tunnel and reverse the direction of their magnetization. Finally, a field $`H`$ is applied to measure the short time square root relaxation rate $`\mathrm{\Gamma }_{\mathrm{sqrt}}`$ . The entire procedure is then repeated to probe the distribution at other fields yielding the field dependence of the relaxation rate $`\mathrm{\Gamma }_{\mathrm{sqrt}}(H,H_{\mathrm{dig}},t_{\mathrm{dig}})`$ which is more or less proportional to the number of spins which are still free for tunneling. The result of this procedure is that a very sharp ”hole” is dug into the rather broad distribution of $`\mathrm{\Gamma }_{\mathrm{sqrt}}`$ . A typical example is shown in the inset of Fig. 3.
In the limit of very short digging times, the difference between the relaxation rate in the absence and in the presence of digging, $`\mathrm{\Gamma }_{\mathrm{hole}}=\mathrm{\Gamma }_{\mathrm{sqrt}}(H,H_{\mathrm{dig}},t_{\mathrm{dig}}=0)\mathrm{\Gamma }_{\mathrm{sqrt}}(H,H_{\mathrm{dig}},t_{\mathrm{dig}})`$, is approximately proportional to the number of molecules which reversed their magnetization during the time $`t_{\mathrm{dig}}`$. $`\mathrm{\Gamma }_{\mathrm{hole}}`$ is characterized by a width that we call the hole line width $`\sigma `$. In order to find a hole line width that is close to the hyperfine level broadening, all the effects that can broaden the measured hole width must be reduced. The experimental condition giving the smallest line width was found for hole digging in the tails of the dipolar distribution and for small initial magnetization. Under these conditions the spins that tunnel are statistically far from each other allowing us to measure tunneling in the diluted limit. In addition, we applied a transverse field of $`\mu _0H_{\mathrm{trans}}`$ = 200 mT parallel to the hard axis which reduces the tunnel rate allowing us to dig very tiny holes. Fig. 4 displays the hole line width $`\sigma `$ as a function of the reversed fraction $`\mathrm{\Delta }M_{\mathrm{dig}}/2M_\mathrm{s}`$ of molecular spins. A linear extrapolation of $`\sigma `$ to $`\mathrm{\Delta }M_{\mathrm{dig}}/2M_\mathrm{s}=0`$ gives $`\sigma _0`$ which is directly associated to the hyperfine level broadening. Experimentally we found $`\sigma _0`$ to be $`0.6\pm 0.1`$, $`0.8\pm 0.1`$, and $`1.2\pm 0.1`$ mT, for <sup>D</sup>Fe<sub>8</sub>, <sup>st</sup>Fe<sub>8</sub>, and <sup>57</sup>Fe<sub>8</sub>, respectively.
The isotope effect, observed here for the first time in magnetic nanostructures, clearly point out the role of the magnetic nuclei in the relaxation of the magnetization. An evaluation of the hyperfine fields in the three different samples is therefore necessary. The hyperfine interaction between the total spin $`S`$ of the cluster and the magnetic nuclei can be decomposed into the sum of terms related to the interactions between the magnetic moment $`I_i`$ of the $`i`$th nucleus and the individual spin $`S_j`$, assumed to be localized on the $`j`$th iron centre :
$$H_{hf}=\underset{i}{}\mathrm{𝐒𝐀}_i𝐈_i=𝐒\underset{i}{}\left(\underset{j}{}c_j𝐀_{ij}\right)𝐈_i$$
(1)
The projection coefficients $`c_j`$ in Eq. (1) depend on the wave function of the ground state $`S=10`$, which can be calculated by diagonalizing the $`S=10`$ block $`(6328\times 6328)`$ of the exchange spin-Hamiltonian matrix of $`H=_{jk}𝐒_j𝐒_k`$.
With the exchange-coupling parameters that best reproduce the temperature dependence of the magnetic susceptibility , the spin configuration depicted in Fig. 1 provides a large contribution (of ca. 70$`\%`$) of the ground state $`S=10`$. According to this picture, which has been recently confirmed by polarized neutron diffraction data , the projection coefficients are $`c_3=c_4=5/22`$ for the spins pointing down and $`c_1=c_2=c_5=\mathrm{}=c_8=8/33`$ for the remaining iron spins.
The hyperfine interaction described by the $`𝐀_{ij}`$ tensors is both through-space (dipolar) and through bond (contact) in nature. The dipolar components can be easily calculated with the point dipolar approximation. The most important coupling with the <sup>1</sup>H nuclei are those of the bridging OH groups having $`𝐀_{ij}`$ constants as large as 0.045 mT, while the hyperfine coupling with N and Br nuclei does not exceed 0.005 and 0.003 mT respectively. An order of magnitude for the contact terms was estimated using a Density-Functional Theory calculation at the B3LYP level on a model symmetric dimer $`[(`$NH$`{}_{3}{}^{})_4`$Fe$`(`$OH$`)_2`$Fe$`(`$NH$`{}_{4}{}^{}{}_{3}{}^{})]^{4+}`$ .
The hyperfine interaction generates a field which splits the $`m_S=\pm 10`$ states. Since each nuclear spin I splits each state into $`2I+1`$ sublevels, the number of sublevels generated by the coupling of 18 <sup>14</sup>N atoms $`(I=1)`$, 8 <sup>79,81</sup>Br atoms $`(I=3/2)`$, and 120 <sup>1</sup>H atoms (I =1/2) present in <sup>st</sup>Fe<sub>8</sub> is prohibitively large, being $`3^{18}\times 4^8\times 2^{120}`$. We therefore made an approximation taking into account only the most significant terms. In fact the total line width goes as the geometric sum of the individual contributions and therefore the largest contributions dominate over the smaller ones. We evaluated the gaussian broadening determined by the 12 <sup>1</sup>H nuclei of the OH groups, assuming them equivalent with a contact hyperfine coupling constant $`A_{\mathrm{cont}}`$ = 0.05 mT, of the 18 <sup>1</sup>H nuclei of the NH groups assuming $`A_{\mathrm{cont}}`$ = 0.025 mT, and of the 14 N nuclei with $`A_{\mathrm{cont}}=0.2`$ mT. These values introduced in Eq. (1) gave gaussian lines with widths at half-height of 0.2, 0.15, and 0.4 mT, respectively. By combining these we estimate a resulting gaussian distribution with a line width of 0.5 mT, in acceptable agreement with the experimental value of 0.8 mT.
The effect of the <sup>57</sup>Fe nuclear spins in the enriched samples was estimated by assuming that each nucleus only feels its own electron spin. Therefore Eq. (1) is simplified as only the terms with $`i=j`$ are different from zero. Using $`A(^{57}`$Fe) = 1.0 mT, in agreement with reported data , we calculate the stick diagram reported in Fig. 5 which arises from the coupling of six equivalent <sup>57</sup>Fe $`I=1/2`$ with $`A=8/33\times `$1.0 mT and two other nuclear spins $`I=1/2`$ with coupling $`A=5/22\times `$ 1.0 mT. The related histogram can be fitted with a gaussian whose line width is ca 0.8 mT. If we consider the experimental line width of the resonance, ca. 0.8 mT, of <sup>st</sup>Fe<sub>8</sub> and we add the contribution of the <sup>57</sup>Fe nuclei we obtain $`\mathrm{\Delta }H`$ 1.1 mT in close agreement with the experimentally observed value of 1.2 mT.
The partial substitution for the <sup>1</sup>H nuclei of the OH and NH groups with the less magnetic <sup>2</sup>H isotopes leads to a reduction of the line width which in our calculations is estimated to be ca. 0.1 mT which should be compared with the experimental narrowing of ca. 0.25 mT. The difference between the calculated and observed reduction of the line width is similar to the smaller calculated line width compared to the observed one of the <sup>st</sup>Fe<sub>8</sub> sample. This may come from an underestimation of the interactions with the <sup>1</sup>H nuclei.
The present data show the fundamental role of the nuclear spins in the relaxation of the magnetization of Fe<sub>8</sub> in the quantum regime. Indeed, this is in contrast to the familiar role of isotope substitution which are generally associated with phonon coupling and thus proportional to the mass of the nuclei. Here we show that it is the magnetic moment of the nuclei which is important at temperatures well above those at which nuclear spin polarisation is observed.
We are indebted to A. Bencini for his help in DFT calculations, and to P.C.E. Stamp, I. Tupitsyn, N.V. Prokof’ev, and J. Villain for many fruitful and motivating discussions. The financial contributions of Italian MURST, CNR, PFMSTAII, and the French DRET are acknowledged. |
warning/0001/hep-ex0001051.html | ar5iv | text | # Partial-Wave Analysis of the 𝜂𝜋⁺𝜋⁻ System Produced in the Reaction 𝜋⁻𝑝→𝜂𝜋⁺𝜋⁻𝑛 at 18 GeV/c
## I Introduction
In this paper we present results of a partial-wave analysis of the $`\eta \pi ^+\pi ^{}`$ system in the 1210 to 1530 $`\mathrm{MeV}/c^2`$ mass region, as obtained from the study of the reaction
$$\pi ^{}p\eta \pi ^+\pi ^{}n,\eta 2\gamma $$
(1)
at 18.3 GeV/c. The data sample was collected during the summer of 1994 using the Multi-Particle Spectrometer (MPS) at the Alternating Gradient Synchrotron (AGS) facility of Brookhaven National Laboratory (BNL).
The identification of the isoscalar members of the $`J^{PC}=0^+`$ and $`1^{++}`$ nonets has been the subject of considerable interest, particularly with regard to searches for exotic mesons. It is known that such states often have $`a_0(980)\pi `$ decay modes. Since the $`a_0(980)`$ couples to both $`\eta \pi `$ and to $`K\overline{K}`$ final states, comparison of the resonances produced in the $`\eta \pi ^+\pi ^{}`$ and $`K\overline{K}\pi `$ reactions can lead to important information with regard to this identification.
The $`\eta `$$`\pi ^+`$$`\pi ^{}`$ system is complicated, characterized by the large range of accessible quantum numbers ($`J^{PC}=0^+`$, $`0^{}`$, $`1^{}`$, $`1^+`$, $`1^{++}`$, $`2^{}`$ …), a large number of possible $`\eta \pi `$ and $`\pi \pi `$ intermediate isobars ($`\sigma `$, $`\rho (770)`$, $`a_0(980)`$, $`f_2(1270)`$, $`a_2(1320`$))<sup>*</sup><sup>*</sup>*We refer to the $`\pi \pi `$ S-wave as $`\sigma `$. The form used for this is discussed in Section IV A., and the presence of overlapping resonances ($`f_1(1285)`$, $`\eta (1295)`$).
Historically, the low-mass region around the 1300 $`\mathrm{MeV}/c^2`$ enhancement in the $`\eta \pi ^+\pi ^{}`$ and $`K\overline{K}\pi `$ mass spectra was called the D region. Most early analyses made the assumption that a single state existed in this region in the presence of an incoherent (non-interfering) background. The problem was then the determination of the appropriate quantum numbers of this state and its branching ratio to $`\eta `$$`\pi ^+`$$`\pi ^{}`$. Most early experiments showed a preference for $`J^{PC}=1^{++}`$ quantum numbers for this state, now referred to as the $`f_1(1285)`$ .
Later, sufficiently high statistics were collected to carry out a partial wave analysis of the $`\eta \pi ^+\pi ^{}`$ system. Stanton et al. performed an analysis of the reaction $`\pi ^{}p\eta \pi ^+\pi ^{}n`$ at 8.45 GeV/c. The low-mass region was fit with a combination of $`0^+`$, $`1^{++}`$, and $`1^+`$ partial waves. Their analysis suggested the presence of a new state with $`J^{PC}=0^+`$, the $`\eta (1295)`$, as well as the $`f_1(1285)`$. In addition, it was suggested that the fit could be improved considerably if the $`0^+`$ partial waves were not allowed to interfere with the other waves in the fit.
The KEK-E179 collaboration performed two partial wave analyses of the same reaction at 8 GeV/c. They too used a set of $`0^+`$, $`1^{++}`$, and $`1^+`$ partial waves to describe the low-mass region, and observed the $`f_1(1285)`$ and $`\eta (1295)`$. Their analysis also suggested the presence of an additional state, now called the $`\eta (1440)`$, in the high-mass region, which was earlier called the E region.
## II Apparatus
Figure 1 shows the elevation view of the experimental layout. The detector system consists of a charged-particle spectrometer and a downstream 3045-element lead-glass electromagnetic calorimeter (LGD) to provide neutral particle detection.
An 18.3 GeV/c $`\pi ^{}`$ beam is incident on the 30-cm liquid-hydrogen target located at the center of the MPS magnet. Three threshold Ĉerenkov counters in the beam line are used to tag the beam particles as pions. Surrounding the target is a 198-element thallium-doped cesium iodide cylindrical veto array (CsI) used in off-line analysis to reject events with wide-angle, low-energy photons from the decays of baryonic resonances. Between the target and the CsI is a four-plane cylindrical drift chamber (TCYL) for triggering on recoil charged tracks. The downstream half of the magnet is equipped with three proportional wire chambers (TPX1-3) for triggering on forward charged-track multiplicity, and six drift chamber modules (DX1-6) for measuring the momentum of forward charged tracks. Also in the magnet is a window-frame lead-scintillator sandwich photon veto counter (DEA) which covers the solid-angle gap between the CsI and the LGD. Two scintillation counters are mounted on DEA, a window-frame counter (CPVC) to distinguish between charged and neutral particles hitting DEA, and a rectangular counter (CPVB) which covers the hole in the DEA and is used, in conjunction with CPVC, to veto charged tracks in the all-neutral trigger. Beyond the magnet, and just upstream of the LGD, is a final drift chamber (TDX4) consisting of two X-planes, and two scintillation counters (BV and EV) for vetoing non-interacting beam particles and elastic-scattering events. Further details regarding the equipment are given elsewhere .
## III Data Selection and Properties
The trigger for the $`\eta \pi ^+\pi ^{}`$ topology required a Ĉerenkov-tagged $`\pi ^{}`$ incident on the target, two charged tracks emerging forward from the target, no charged recoil track, and an effective mass greater than that of the $`\pi ^0`$ in the LGD as determined by a hardware processor. Forty eight million triggers of this type were taken. From these, a final sample of events consistent with reaction 1 was selected by requiring:
* less than 20 MeV in the CsI to enhance recoil neutron events over $`N^{}`$ events;
* exactly two photons ($`\eta `$) reconstructed in the LGD;
* a reconstructed beam track;
* two forward charged tracks of opposite charge;
* no recoil charged track;
* a 3-constraint SQUAW kinematic fit to reaction 1 with a confidence level greater than $`7\%`$;
* $`|t|<3`$ GeV<sup>2</sup>/c<sup>2</sup> after kinematic fitting, where $`t`$ is defined as the magnitude of the four-momentum transfer squared between the target proton and the neutron in the final state.
The two-gamma mass distributions for about $`10\%`$ of the data are shown before and after the above data selection cuts in Figs. 2(a) and (b) respectively. The $`\eta `$ signal is nearly background free after cuts.
The $`\eta \pi ^+\pi ^{}`$ mass distribution for these events is shown in Figure 3. The $`\eta ^{^{}}(958)`$ is evident. When fit with a Gaussian, a mass of 961 $`\mathrm{MeV}/c^2`$ with $`\sigma `$ =10 $`\mathrm{MeV}/c^2`$ is obtained. This provides a measure of the mass resolution of the apparatus in the 1000 $`\mathrm{MeV}/c^2`$ mass region after kinematic fitting. An enhancement in the 1300 $`\mathrm{MeV}/c^2`$ regionA detailed description of the Dalitz plot in this region is given is elsewhere . is also observed, which, when fit to a Gaussian plus a linear background, yields a mass of 1278 $`\mathrm{MeV}/c^2`$ and $`\sigma `$ = 20 $`\mathrm{MeV}/c^2`$.
In Figs. 4a, 4b and 4c we show the $`\eta \pi ^{}`$, $`\eta \pi ^+`$ and $`\pi ^+\pi ^{}`$ effective mass distributions respectively for a three-body mass between 1200 and 1540 $`\mathrm{MeV}/c^2`$. In Figs. 4a and 4b the $`a_0(980)`$ peak is seen. The $`\rho (770)`$ peak in Fig. 4c is cut off on the high-mass side because of the limited phase space available.
The same distributions are shown in Fig. 5(a)-(c) for the low mass subset of the data between 1200 and 1350 $`\mathrm{MeV}/c^2`$. A very significant asymmetry between the $`\eta \pi ^{}`$ and the $`\eta \pi ^+`$ distributions is evident. This asymmetry is due to the interference between I=0 $`a_0\pi `$ and I=1 $`\rho \eta `$ states, and is well-described in the partial-wave analysis described in the next section.
For the following analysis, 9,082 events were selected from the above data set in the region $`1205M(\eta \pi \pi )1535\mathrm{MeV}/c^2`$.
## IV Partial-wave analysis
### A Fitting Procedure
The formalism used in this analysis is based on the papers of Chung and Chung and Trueman . The analysis techique involved the use of the reflectivity basis to describe the individual partial waves and the maximization of an extended log likelihood function in the fitting procedure. Fits are carried out independently in each $`\eta \pi ^+\pi ^{}`$ mass bin. The procedure and analysis programs are described by Cummings and Weygand .
Due to the large number of possible partial waves accessible to the $`\eta \pi ^+\pi ^{}`$ system, a complete analysis requiring all possible isobars and partial waves is not practical given the limited statistics. In principle one would like to include all possible isobars: $`\sigma ,\rho ,a_0,f_2,a_2`$ and a large set of partial waves ($`J<4`$). Because this analysis is limited to the low-mass region, we can neglect the $`a_2`$ and the $`f_2`$ isobars. The $`f_2(1270)`$ could in principle reach this final state through the $`a_2\pi `$ mode, but this is highly suppressed by phase space. Furthermore, we choose to consider only amplitudes with $`J<2`$ since there are no known states with higher spin in this low-mass region which decay into $`\eta \pi \pi `$.
An incoherent isotropic background was included in some trial fits, but it was not used in the final fit. This type of background is, except for the $`\pi \pi `$ mass dependence of the amplitude, similar to a non-interfering $`J=0`$, $`\sigma \eta `$ partial wave, making them quite difficult to differentiate.
In order to determine whether both spin-flip and non-flip amplitudes at the baryon vertex are required to fit the data, fits were attempted for both rank 1 and 2. (If both types of amplitudes are not necessary, the rank 1 fit will give a good description of the data.) The likelihood function was improved greatly when the fit rank was increased to two. In addition, rank 1 fits to the data were found to become unstable in the absence of a background wave. We conclude that a rank 2 fit is required to fit the data; rank 1 fits were not used.
The $`\rho `$ isobar was modeled by a relativistic Breit-Wigner amplitude with parameters extracted from the Particle Data Book . For the final fit the $`a_0`$ isobar was modeled as a Breit-Wigner form with a mass of 980 $`\mathrm{MeV}/c^2`$ and a width of 72 $`\mathrm{MeV}/c^2`$ . The $`\pi \pi `$ S-wave (the $`\sigma `$) was represented by a parameterization of the $`\pi ^+`$$`\pi ^{}`$ S-wave provided by K. L. Au et al. . Alternate parameterizations of the $`a_0`$ and the $`\sigma `$ were explored . However it was found that the particular choice of parameterization had little effect on the final results.
To determine the appropriate waves for the 1200-1350 $`\mathrm{MeV}/c^2`$ region, a fit was performed using a coarse bin width of 50 $`\mathrm{MeV}/c^2`$ with all waves with $`J<2`$ included. Waves were then discarded from the fit if their removal had little effect on the value of the likelihood function ($`|\mathrm{\Delta }|<5`$). Selected waves were then re-introduced in the fit to insure the stability of the solution. In total, several hundred different sets of partial waves were fit. For each combination of partial waves, the binning, $`t`$ cuts, and starting values of the fit parameters were varied to insure stability of the final solution. The set of waves chosen for the final fit is shown in Table I.
For the final fit, a bin width of 30 $`\mathrm{MeV}/c^2`$ was chosen. This was a compromise between achieving adequate statistics in each mass bin and acquiring the best possible resolution in the entire mass region 1200-1540$`\mathrm{MeV}/c^2`$. The starting values of the fit were randomly chosen and the entire spectrum was re-fit several times to insure stability with the finer bin width. For bin widths smaller than 30 $`\mathrm{MeV}/c^2`$ the fits often became unstable, converging to different solutions depending upon the starting values.
It is interesting to note that the final waves selected for the low-mass region are consistent with those used by Stanton et al. and by Fukui et al. . The only exception is that we do not use a $`1^+a_0\pi `$ wave. While the inclusion of this wave in the fit for the low-mass region is reasonable, providing a natural explanation for the odd-even isospin interference observed in the data, we found that the fit could not distinguish this wave from the $`1^+\rho \eta `$ wave in this mass range. Due to the unambiguous presence of the $`\rho \eta `$ partial wave at higher mass, it was decided not to include the $`1^+a_0\pi `$ partial wave in the final set.
### B Results of Partial Wave Analysis
#### 1 $`J^{PC}=0^+`$ Partial Waves
The fitted intensity distribution for the $`0^+a_0\pi `$ wave as a function of $`\eta \pi ^+\pi ^{}`$ effective mass is shown in Fig. 6a. A sharp peak at $``$ 1300 $`\mathrm{MeV}/c^2`$ is evident, consistent with the observation of $`\eta (1295)a_0\pi `$. Some intensity is seen extending out to 1400 $`\mathrm{MeV}/c^2`$. It should be noted that the majority of the signal for this wave comes from the second rank of the fit. This indicates a different production mechanism than that for the $`1^+`$ and the $`1^{++}`$ waves (which are produced dominantly in the first rank) and means that these latter waves do not interfere with the $`0^+`$ wave (see discussion below).
As shown in Fig. 6b, the $`0^+\sigma \eta `$ wave is double-peaked, with structure suggestive of $`\eta (1295)`$ and $`\eta (1440)`$ production. The dominant nature of the structure seen in the high-mass region in this wave is somewhat inconsistent with previous analyses which observe the presence of a $`\sigma \eta `$ decay of the $`\eta (1440)`$, but do not see it as dominant.
A large fraction of the $`0^+`$ signal occurs in the second rank, especially for the $`\sigma \eta `$ partial waves in the high-mass region. Because these waves do not interfere with the other dominant waves in the fit, reliable relative phase motion could not be obtained.
The $`a_0\pi `$ and $`\sigma \eta 0^+`$ waves were added coherently in each rank and then summed incoherently. The result is shown in Fig. 6c. The $`\eta (1295)`$ and $`\eta (1440)`$ peaks are clearly visible. The spectrum was fit with two spin-0 Breit-Wigner forms plus a quadratic background. Fitted values of the masses and widths are given in Table II. In addition, the $`a_0\pi /\sigma \eta `$ branching ratio were determined from Fig. 6 for both the $`\eta (1295)`$ and the $`\eta (1440)`$. These values are also given in Table II.
#### 2 $`J^{PC}=1^{++}`$ Partial Waves
In Fig. 7a the $`1^{++}a_0\pi `$ partial wave intensity distribution is shown. This wave shows evidence for the $`f_1(1285)`$. The amount of $`f_1`$ signal is comparable to the $`0^+`$ signal in the $`a_0\pi `$ channel. No significant structure is observed at higher mass.
In Fig. 7b the $`1^{++}\sigma \eta `$ partial wave intensity distribution is shown. This wave does not show any structure, but was necessary to the fit for bins above 1450 $`\mathrm{MeV}/c^2`$.
In Fig. 7c the coherent sum of the $`1^{++}`$ partial waves is shown. This sum displays a peak in the vicinity of the $`f_1(1285)`$ and a rise at high mass. A comparison of Fig. 7c and Fig. 6c reveals that the majority of the signal strength in the 1200-1500 $`\mathrm{MeV}/c^2`$ mass region arises not from $`1^{++}`$ partial waves, but from the $`0^+`$ wave. This observation is especially important for the low-mass region because several previous analyses for branching ratio estimates assumed that the low-mass region was dominated by the $`1^{++}`$ wave. We find that the ratio of the $`1^{++}`$ intensity to the sum of the $`1^{++}`$ and $`0^+`$ intensities in the region 1235-1325 $`\mathrm{MeV}/c^2`$ is $`0.19\pm 0.06`$.
The observed $`f_1(1285)`$ signal in Fig. 7c is in the same mass region with a very similar width as the $`\eta (1295)`$. To eliminate the possibility that this $`1^{++}`$ $`(f_1)`$ signal is an artifact, due to “leakage” from the larger $`0^+`$ $`(\eta )`$ signal, a Monte-Carlo study was performed. The measured amplitudes, with the contribution due to the $`1^{++}`$ wave removed, were used to generate Monte-Carlo events taking into account the experimental resolution, acceptance and statistics. These events were then analyzed in exactly the same manner as the data. The resulting $`1^{++}`$ intensity in the low-mass region was found to be less than 2% of the total signal, and consistent with zero. This leads to the conclusion that the observed signals are not artifacts of the analysis or apparatus, and are, in fact, two distinct resonances.
#### 3 $`\rho \eta `$ Partial Waves
The intensity distribution for the $`1^+\rho \eta `$ wave is shown in Fig. 8a. This wave was seen in all previous analyses and is significant in the low-mass region. Previous experiments have claimed this wave to show evidence for production of the $`b_1(1235)`$ with a $`\rho \eta `$ decay mode. However, we do not observe any structure in the $`1^+`$ wave to support this conjecture. As mentioned earlier, this wave is, nevertheless, essential for producing the $`a_0^+/a_0^{}`$ asymmetry observed in the data.
In Fig. 8b the $`1^{}\rho \eta `$ intensity distribution is shown. This is the only negative-reflectivity partial wave in our analysis, and it does not interfere with any other partial waves in the fit. The wave steadily increases throughout the low-mass region, consistent with its being the low-mass tail of the $`\rho (1700)`$.
## V Discussion of the $`f_1(1285)`$ Branching Fractions
Production of the $`\eta (1295)`$ dominates the low-mass peak, accounting for roughly 80% of the signal. This observation has implications on the $`f_1\eta \pi \pi `$ branching fraction. Previous experiments have determined the $`f_1\eta \pi \pi `$ branching fraction without the aid of a partial wave analysis under the assumption that the low-mass peak consists of a single $`f_1`$ state resting on top of an incoherent background. This assumption is clearly incorrect, and values for the earlier determinations of the branching fractions need to be corrected.
Corden et al. studied the reactions $`\pi ^{}p\eta \pi ^+\pi ^{}n`$ and $`\pi ^{}pK\overline{K}\pi n`$ at 15 GeV/c. In their analysis they obtained a $`K\overline{K}\pi /\eta \pi \pi `$ branching ratio in the low-mass region of $`0.5\pm 0.2`$ without the aid of a partial wave analysis. It is reasonable to assume that the relative production of $`f_1(1285)`$ and $`\eta (1295)`$ is the same in the present experiment as in that of Ref. since the experiments are close in energy and study the same final state. It is also reasonable to assume that the $`K\overline{K}\pi `$ decay at low mass is entirely due to $`f_1(1285)`$ decay since this conclusion was reached by a partial wave analysis of the data. Thus the $`K\overline{K}\pi /\eta \pi \pi `$ branching ratio of the $`f_1(1285)`$, as quoted by Corden et al., should be corrected by dividing it by the fraction of the low-mass peak which is due to $`f_1(1285)`$ decay. We thus obtain $`(0.5\pm 0.2)/(0.19\pm 0.06)=2.6\pm 1.4`$ for this branching ratio.
We can perform the same type of estimate using, instead of our own analysis, the results of KEK-E179 for the reaction $`\pi ^{}p\eta \pi ^+\pi ^{}n`$ at 8.95 GeV/c. In that experiment, the fraction of the low-mass peak which is due to $`f_1(1285)`$ decay is claimed to be 50%. Again, using the results of Corden et al. (although in this case the difference in the energies of the experiments is larger), we obtain an alternate estimate of the $`K\overline{K}\pi /\eta \pi \pi `$ branching ratio for $`f_1`$ decay to be $`1.0\pm 0.4`$.
We can estimate the effect which these results can have on the $`f_1(1285)`$ branching fractions by assuming that the low-mass signal observed in the $`K\overline{K}\pi `$, $`\gamma \rho ^0`$, and $`4\pi `$ decay modes is due only to $`f_1(1285)`$ decay. This is the most reasonable for the $`K\overline{K}\pi `$ mode as mentioned above because the other two decay modes ($`\gamma \rho ^0`$, $`4\pi `$) have not been as thoroughly investigated.<sup>§</sup><sup>§</sup>§Of these modes the $`4\pi `$ branching fraction is most suspect due to the large number of interfering partial waves which contribute to a $`4\pi `$ data set. Nevertheless, using the PDG98 branching ratios of $`0.271\pm 0.016`$ for $`K\overline{K}\pi /4\pi `$ and $`0.45\pm 0.18`$ for $`\gamma \rho ^0/2\pi ^+2\pi ^{}`$, the $`f_1(1285)`$ branching fractions can be calculated. (We also assume the branching ratio for $`4\pi /2\pi ^+2\pi ^{}=3`$ as in .) In Table III we list the $`f_1(1285)`$ branching fractions derived by the above procedure.
Assigning systematic errors to these $`f_1(1285)`$ branching fractions is difficult because of the undetermined uncertainties in branching ratios for the $`4\pi `$ and $`\gamma \rho ^0`$ decay modes. However, it is clear from the above exercise that the results from the present experiment and the KEK experiment for the $`f_1(1285)`$ branching fractions are consistent, and those listed in the particle data book need to be corrected. The most significant result is the large reduction in the $`f_1\eta \pi \pi `$ branching fraction.
## VI Summary and Conclusions
A partial wave analysis was performed on 9082 $`\eta \pi ^+\pi ^{}n`$ events in the $`1205M(\eta \pi ^+\pi ^{})1535\mathrm{MeV}/c^2`$ mass interval. The analysis used a rank 2 fit with 30 $`\mathrm{MeV}/c^2`$ bins and a set of 6 partial waves. The partial waves used in the fit were: $`0^+a_0\pi `$, $`0^+\sigma \eta `$, $`1^+\rho \eta `$, $`1^{++}a_0\pi `$, $`1^{++}\sigma \eta `$ and $`1^{}\rho \eta `$.
The low-mass region was found to include a large contribution from the $`0^+`$ wave which indicates the production of $`\eta (1295)`$. Evidence of $`f_1(1285)`$ production was seen in the $`1^{++}`$ wave. The fact that the region is dominated by $`\eta (1295)`$ production leads to significant changes in the $`f_1(1285)`$ branching fractions as discussed in Section V.
The $`\eta (1295)`$ was seen to decay to both $`a_0\pi `$ and $`\sigma \eta `$. The $`a_0\pi /\sigma \eta `$ branching ratio for $`\eta (1295)`$ was estimated to be $`0.48\pm 0.22`$. The mass and width of the $`\eta (1295)`$ were determined to be $`1282\pm 5\mathrm{MeV}/c^2`$ and $`66\pm 13\mathrm{MeV}/c^2`$ respectively. This result is consistent with the PDG98 summary of $`\eta (1295)`$ mass and width of $`1297\pm 2.8\mathrm{MeV}/c^2`$ and $`53\pm 6\mathrm{MeV}/c^2`$, respectively.
The high-mass region is dominated by a large $`0^+\sigma \eta `$ signal present in the second rank of the fit. This signal is consistent with production of a single state, the $`\eta (1440)`$. The mass and width of the $`\eta (1440)`$ are estimated to be $`1404\pm 6\mathrm{MeV}/c^2`$ and $`80\pm 21\mathrm{MeV}/c^2`$. This result is consistent with the PDG98 weighted average value for the mass and width of the $`\eta (1440)`$ determined from the $`\eta \pi \pi `$ mode of $`1405\pm 5\mathrm{MeV}/c^2`$ and $`56\pm 7\mathrm{MeV}/c^2`$, respectively.
The $`\eta (1440)`$ has been previously observed in the reaction $`\pi ^{}pK\overline{K}\pi n`$, in $`p\overline{p}`$ annihilation, and in the radiative decay of $`J/\psi `$, with decays in the $`a_0\pi `$ and $`K\overline{K^{}}`$ modes. Studies of the $`\pi ^{}p\eta \pi ^+\pi ^{}n`$ reaction have yielded both a $`\sigma \eta `$ and an $`a_0\pi `$ component of the $`\eta (1440)`$. In the present analysis, it is found that the $`\sigma \eta `$ decay dominates, while the KEK analyses suggest a larger $`a_0\pi `$ component. The estimate of the $`a_0\pi /\sigma \eta `$ branching ratio for $`\eta (1440)`$ from the present analysis is $`0.15\pm 0.04`$. The systematic errors are unassigned, but assumed to be large due to the difficulty of the fit in distinguishing $`0^+a_0\pi `$ and $`0^+\sigma \eta `$ waves from each other.
In addition to the $`f_1(1285)`$, $`\eta (1295)`$ and the $`\eta (1440)`$ contributions, a large, relatively structureless signal in the $`1^+\rho \eta `$ wave was observed throughout the low mass region. This wave has also been observed in all previous partial wave analyses of the $`\pi ^{}p\eta \pi ^+\pi ^{}n`$ system. There is no obvious resonance interpretation of this structure, but its presence is required to account for the large $`a_0^+/a_0^{}`$ production asymmetry seen in the low mass region. A $`1^{}\rho \eta `$ partial wave, consistent with the low-mass tail of the $`\rho (1700)`$, is also seen.
We would like to express our deep appreciation to the members of the MPS group. Without their outstanding efforts, the results presented here could not have been obtained. We would also like to acknowledge the invaluable assistance of the staffs of the AGS and BNL, and of the various collaborating institutions. This research was supported in part by the National Science Foundation and the US Department of Energy. |
warning/0001/cond-mat0001136.html | ar5iv | text | # Phase transitions in generalized chiral or Stiefel’s models
## I INTRODUCTION
The critical properties of frustrated spin systems are still under discussion . In particular no consensus exists about the nature of the phase transition in canted magnetic systems. One example is the stacked triangular lattice with the nearest neighbor antiferromagnetic interactions (STA) with vector spins $`O(N)`$ where $`N`$ is the number of spin components which is always controversial . The non-collinear ground state due to the frustration leads to a breakdown of symmetry (BS) from $`O(N)`$ in the high temperature to $`O(N2)`$ in the low temperature. This is different from ferromagnets in which the ground state is collinear and the BS is $`O(N)/O(N1)`$. Based on the concept of universality, the class of the transition would be different in the two models. We generalize this chiral model for a BS of the type $`O(N)/O(NP)`$. We obtain the STA model for $`P=2`$ while we obtain new BS for $`NP3`$. For example, the case $`N=P=3`$ should correspond to real experimental systems, this is also applicable in spin glasses where some disorder is present. Several authors have already studied these generalized chiral models applying the Renormalization Group technic . In mean field for $`N>P`$ the model shows a usual second order type, but for $`N=P`$ the transition shows a special behavior. This last result could be interpreted with the BS in this case being $`Z_2SO(N)`$ and the coupling between the two symmetries leading to some special behavior (for example, the case $`N=P=2`$ in two dimensions (d) is always very debated ). The $`d=4ϵ`$ expansion gives more information. The picture is very similar for all $`NP2`$ (for details see and references therein). At the lowest order in $`ϵ`$, there are up to four fixed points, depending on the values of $`N`$ and $`P`$. Amongst them are the trivial Gaussian fixed point and the standard isotropic $`O(NP)`$ Heisenberg fixed point. These two fixed points are unstable. In addition, a pair of new fixed points, one stable and the other unstable, appear if the case is $`NN_c(d)`$ with
$`N_c(d)=5P+2+2\sqrt{6(P+2)(P1)}\left[5P+2+{\displaystyle \frac{25P^2+22P32}{2\sqrt{6(P+2)(P1)}}}\right]ϵ.`$ (1)
For $`P=2`$ we find the standard result $`N_c=21.823.4ϵ`$. On the other hand, for $`P=3`$ we obtain $`N_c=32.533.7ϵ`$ and for $`P=4`$ we obtain $`N_c=42.843.9ϵ`$. A ”tricritical” line exists which divides a second order region for low $`d`$ and large $`N`$ from a first order region for large $`d`$ and small $`N`$. From these results Kawamura, using $`ϵ=1`$ ($`d=3`$), obtained that $`N_c(d=3)<0`$ for all $`P`$. Thus he concluded that the experimental or numerical accessible systems ($`N>0`$) were in the second order region. Unfortunately it has been proved that the results of $`4ϵ`$ are, at best, asymptotic . They have to be resummed to obtain reliable results. Indeed for $`P=2`$ the calculation of the next order in $`ϵ`$, combined with a resummation technic, leads the experimental accessible systems for $`N=2`$ or $`N=3`$ in the first order region . We believe that the same applies for $`P3`$. In order to verify our assumption, we have done some simulations for $`P=3`$ and $`P=4`$ with $`N=P`$ and $`N=P+1`$. The most interesting case is $`P=N=3`$ with some possible experimental realizations and connection with the spin glasses. Moreover it is meaningful to study the generalized model in order to have a better overview. The system we analyze is the Stiefel model . This model is constructed to have the needed BS. It is closely connected to real systems with complicating interactions, which are characterized by the same BS (for the case $`P=2`$ see and reference therein). From the principle of universality, models with the same BS should belong to the same universality class. Moreover we have shown that the use of the Stiefel model allows us to avoid problems which are seen in standard models, such as the presence of a complex fixed point (or minimum in the flow) .
In the following section II the studied models are presented, we describe the details of the simulations and the finite size scaling analysis. Results will be given in section III and the last section is devoted to the conclusion.
## II Stiefel’s models, Monte Carlo simulations and first order transitions
In this section we introduce different models studied in this work.
First the $`V_{3,3}`$ model which is represented in Fig. 8. The energy of the model is
$$H=J\underset{ij}{}\underset{k=1}{\overset{P}{}}\left[𝐞_k(i)𝐞_k(j)\right]$$
(2)
where the $`P`$ mutual orthogonal $`N`$ component unit vectors $`𝐞_k(i)`$ at lattice site $`i`$ interact with the next $`P`$ vectors at the neighboring sites $`j`$. The interaction constant is here negative to favor alignment of the vectors at different sites. Taking a strict orthogonality between the vectors is similar to removing ”irrelevant” modes corresponding to the variation between the spins inside the cell. For example, in the case of a triangular lattice with antiferromagnetic interactions (STA) we force the three spins of each cell to have a rigidity constraint with the sum of all the spins being always zero. The obtained model is equivalent to the STA at the critical temperature and can easily be transformed into the Stiefel’s $`V_{3,2}`$ model (for more detail see ).
We did not use the clusters algorithm because it gives worse results than the standard Metropolis algorithm for first order transitions.
The method for choosing the random vector depends on the number of components $`N`$. For $`N=3`$ we follow the method explained in for the direct-trihedral model. We construct two orthogonal vectors $`𝐞_\mathrm{𝟏}`$ and $`𝐞_\mathrm{𝟐}`$, and the third vector is constructed by the vector product of the first two:
$`𝐞_\mathrm{𝟑}=\sigma 𝐞_\mathrm{𝟏}\times 𝐞_\mathrm{𝟐}`$ (3)
where $`\sigma `$ is a random Ising variable, corresponding to the Ising symmetry present in the $`V_{3,3}`$ model. This is the difference with the direct-trihedral model defined in , where no Ising symmetry is present.
To simulate the $`V_{4,3}`$ model we follow a similar procedure. We construct now three orthogonal unit vectors, $`𝐞_k=(e_k^1,e_k^2,e_k^3,e_k^4)`$ with $`k=1,\mathrm{\hspace{0.17em}2},\mathrm{\hspace{0.17em}3}`$, randomly in four dimensions using six Euler angles. The first $`\theta _0`$ must be chosen with probability $`\mathrm{sin}(\theta _0)^2d\theta _0`$, two other with probability $`\mathrm{sin}(\theta _{1,2})d\theta _{1,2}`$ and the rest three $`\theta _{3,4,5}`$ with probability $`\theta _{3,4,5}`$. We obtain for $`𝐞_\mathrm{𝟏}`$:
$`e_1^1=`$ $`\mathrm{cos}(\theta _3)\mathrm{cos}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _4)`$ (4)
$`\mathrm{cos}(\theta _3)\mathrm{cos}(\theta _1)\mathrm{cos}(\theta _5)\mathrm{sin}(\theta _4)`$ (5)
$`+\mathrm{cos}(\theta _3)\mathrm{sin}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _0)`$ (6)
$`+\mathrm{sin}(\theta _3)\mathrm{sin}(\theta _5)\mathrm{cos}(\theta _2)\mathrm{sin}(\theta _4)`$ (7)
$`\mathrm{sin}(\theta _3)\mathrm{cos}(\theta _5)\mathrm{cos}(\theta _4)`$ (8)
$`e_1^2=`$ $`\mathrm{sin}(\theta _3)\mathrm{cos}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _4)`$ (9)
$`\mathrm{sin}(\theta _3)\mathrm{cos}(\theta _1)\mathrm{cos}(\theta _5)\mathrm{sin}(\theta _4)`$ (10)
$`+\mathrm{sin}(\theta _3)\mathrm{sin}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _0)`$ (11)
$`\mathrm{cos}(\theta _3)\mathrm{sin}(\theta _5)\mathrm{cos}(\theta _2)\mathrm{sin}(\theta _4)`$ (12)
$`+\mathrm{cos}(\theta _3)\mathrm{cos}(\theta _5)\mathrm{cos}(\theta _4)`$ (13)
$`e_1^3=`$ $`\mathrm{sin}(\theta _5)\mathrm{sin}(\theta _2)\mathrm{sin}(\theta _0)`$ (14)
$`e_1^4=`$ $`\mathrm{sin}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _4)`$ (15)
$`+\mathrm{sin}(\theta _1)\mathrm{cos}(\theta _5)\mathrm{sin}(\theta _4)`$ (16)
$`+\mathrm{cos}(\theta _1)\mathrm{sin}(\theta _5)\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _0)`$ (17)
for $`𝐞_\mathrm{𝟐}`$:
$`e_2^1=`$ $`\mathrm{sin}(\theta _0)\mathrm{sin}(\theta _1)\mathrm{cos}(\theta _3)`$ (18)
$`e_2^2=`$ $`\mathrm{sin}(\theta _0)\mathrm{sin}(\theta _1)\mathrm{sin}(\theta _3)`$ (19)
$`e_2^3=`$ $`\mathrm{cos}(\theta _0)`$ (20)
$`e_2^4=`$ $`\mathrm{sin}(\theta _0)\mathrm{cos}(\theta _1)`$ (21)
and for $`𝐞_\mathrm{𝟑}`$:
$`e_3^1=`$ $`\mathrm{cos}(\theta _3)\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _4)\mathrm{cos}(\theta _1)`$ (22)
$`+\mathrm{cos}(\theta _3)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _0)\mathrm{sin}(\theta _1)`$ (23)
$`\mathrm{sin}(\theta _2)\mathrm{sin}(\theta _4)\mathrm{sin}(\theta _3)`$ (24)
$`e_3^2=`$ $`\mathrm{sin}(\theta _3)\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _4)\mathrm{cos}(\theta _1)`$ (25)
$`+\mathrm{sin}(\theta _3)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _0)\mathrm{sin}(\theta _1)`$ (26)
$`+\mathrm{sin}(\theta _2)\mathrm{sin}(\theta _4)\mathrm{cos}(\theta _3)`$ (27)
$`e_3^3=`$ $`\mathrm{cos}(\theta _2)\mathrm{sin}(\theta _0)`$ (28)
$`e_3^4=`$ $`\mathrm{sin}(\theta _2)\mathrm{cos}(\theta _4)\mathrm{sin}(\theta _1)`$ (29)
$`+\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _0)\mathrm{cos}(\theta _1).`$ (30)
From this model we can easily create the direct-quadrihedral model and the $`V_{4,4}`$, these two models being composed of four orthogonal vectors with four components. The differences between the two models are the presence of an Ising variables in the $`V_{4,4}`$ where right handed and left handed are allowed while only one possibility exists in the direct-quadrihedral. The direct-quadrihedral and the $`V_{4,3}`$ models are topologically equivalent, they should have the same low energy physics and therefore belong to the same universality class. The connection between the direct-trihedral and the $`V_{3,2}`$ models is very similar to the above. We form a fourth vector from the vector product of $`𝐞_\mathrm{𝟏}`$, $`𝐞_\mathrm{𝟐}`$ and $`𝐞_\mathrm{𝟑}`$:
$`𝐞_\mathrm{𝟒}=𝐞_\mathrm{𝟏}\times 𝐞_\mathrm{𝟐}\times 𝐞_\mathrm{𝟑}`$ (31)
for the direct-quadrihedral model, and we add a random Ising variable $`\sigma `$ to the $`V_{4,4}`$ model:
$`𝐞_\mathrm{𝟒}=\sigma 𝐞_\mathrm{𝟏}\times 𝐞_\mathrm{𝟐}\times 𝐞_\mathrm{𝟑}.`$ (32)
We follow the standard Metropolis algorithm to update one $`P`$-hedral after the other. In each simulation between 20 000 to 100 000 Monte Carlo steps are made for equilibration and averages. Cubic systems of linear dimensions from $`L=10`$ to $`L=25`$ are simulated.
The order parameter $`M`$ for this model is
$`M={\displaystyle \frac{1}{PL^3}}{\displaystyle \underset{i=1}{\overset{P}{}}}\left|M_i\right|`$ (33)
where $`M_i`$ is the total magnetization given by the sum of the vectors $`𝐞_i`$ over all sites and $`L^3`$ is the total number of sites.
For $`N=P`$ we define a chirality order parameter:
$`\kappa ={\displaystyle \frac{1}{L^3}}𝐞_𝐏.({\displaystyle \underset{i=1}{\overset{P1}{}}}𝐞_𝐢)`$ (34)
where $``$ means the vector product $`\times `$.
We use the histogram MC technique developed by Ferrenberg and Swendsen which is very useful for identifying a first order transition.
The finite size scaling (FSS) for a first order transition has been extensively studied . A first order transition can be identified by some properties and in particular by the following:
* The histogram $`P(E)`$ has a double peak.
* The magnetization, the chirality and the energy have hysteresis.
The double peak in P(E) means that at least two states with different energies coexist in the system at one temperature.
## III Results
We now present our results for the different models. The $`V_{3,3}`$ and the $`V_{4,4}`$ show a strong first order transition. The hysteresis in $`E`$ and $`<M>`$ are shown in Fig. 8 and 8 for the $`V_{3,3}`$ model, and in $`E`$ and $`\kappa `$ for the $`V_{4,4}`$ model in Fig. 8 and 8. This is in accordance with the negative $`\eta `$ exponent for the $`V_{3,3}`$ model found in which describes a first order transition because $`\eta `$ must be positive . This result is understandable because there is a coupling between the Ising symmetry and the $`SO(N)`$ symmetry. We notice that the $`V_{2,2}`$ model is also of first order and that the models have a stronger first order transition if $`N`$ is greater (we obtain the same hysteresis for $`L=20`$ for the $`V_{3,3}`$ as for $`L=10`$ for the $`V_{4,4}`$). Thus we can generalize our result that the transition is always of first order for $`N=P`$.
The $`V_{4,3}`$ model shows no hysteresis. However a double peak structure appears in the energy histogram and becomes more apparent when the size increases (Fig. 8). For greater sizes the two peaks are well separated by a region of zero probability, the transition time from one state to the other grows exponentially with the size of the lattice. We should obtains hysteresis in the thermodynamic quantities when the simulation is not too long. The $`V_{4,3}`$ model has a first order transition but weaker than the direct-quadrihedral model which, for similar sizes, shows hysteresis. As explained above the two models belong to the same universality class, i.e. a first order transition, similar to the dihedral model $`V_{3,2}`$ and the direct-trihedral model . The addition of the fourth leg to the $`V_{4,3}`$ model allows the first order behavior to be more clearly visible. In Fig. 8 we have plotted our hypothesis for the RG diagram flow. Following the initial point, the flow could be under the influence of a ”complex” fixed point (or minimum of the flow ) and the system mimics a second order transition. Well outside the influence of this fixed point the transition is strongly of first order and in the crossover between these two regions the transition is weakly first order. For a more developed discussion see .
## IV Conclusion
We have tried to give a general picture of the transition with a $`O(N)/O(NP)`$ breakdown of symmetry. We have shown by numerical simulations that for $`N=P=3`$ and $`N=P=4`$, the transition is clearly of first order. We have generalized our result for all $`N=P`$. A similar conclusion is obtained for $`N=4`$ and $`P=3`$. Using the fact that for $`N=3`$ and $`P=2`$ the transition is also of first order, we can generalize our result for all $`P=N1`$. This is in contradiction with to the conclusion of Kawamura which is based on two loops of a $`4ϵ`$ expansion. As we have noted the $`ϵ`$ expansion has to be resummed to obtain reliable results. We can try to achieve this by forming simple Padé approximants. For a function $`f=a+bϵ`$ we obtain the approximation $`f=1/(1ϵb/a)`$ which we apply to $`ϵ=1`$ ($`d=3`$) and $`P=`$2, 3 and 4. We obtain $`N_c(P=2)10`$, $`N_c(P=3)16`$ and $`N_c(P=2)21`$. Unfortunately the results can not be perfect and in particular the result for $`P=2`$ is not close enough to the result including the next order expansion $`N_c=3.39`$ which demonstrates that the lower-order $`ϵ`$ expansions are useless in this case. However we remark that the $`N_c`$ ”resummed” increases with $`P`$ which is in agreement with our result, i.e. that the initial point in the renormalization flow is farther away from the mimic of the second order region . Thus the systems will show a stronger first order transition. This result matches with a recent study on the case $`P=N=3`$ which is based on a non perturbative Renormalization Group procedure . We conclude that transitions for $`N=P`$ and $`N=P+1`$ are of first order for all $`N`$.
## V Acknowledgments
This work is supported by the Alexander von Humboldt Foundation. The authors are grateful to Professors B. Delamotte, G. Zumbach, and K.D. Schotte for discussions. |
warning/0001/astro-ph0001460.html | ar5iv | text | # 1 Overview
## 1 Overview
For a period of 3.5 years (at the time of this workshop), the All-Sky Monitor (ASM; Levine et al. 1996) on RXTE has been monitoring the entire sky for new (uncataloged) transient x–ray sources while also recording the intensities of the known (cataloged) sources The current catalog contains about 325 source positions of which about 180 have yielded positive detections. The monitoring of a given source has been reasonably continuous except for times when the sun is relatively close to the source and except for a period of $``$7 weeks shortly after launch when the detectors were turned off due to a temporary breakdown problem. The detected sources include many well known persistent sources as well as a substantial number of ‘transient’ sources. Some of these are recurrent and others are in their first known outburst. Most of the latter have been discovered in the RXTE era, either with other satellites, e.g. CGRO and BSAX, or with RXTE. A few were discovered prior to the launch of RXTE.
Some of the RXTE discoveries have been made during scans with the highly-sensitive, but narrow-field-of-view, PCA instrument while the others have been made with the ASM. In almost all of these cases, whether the transient is new or recurrent or whether it was discovered with RXTE or not, the PCA has carried out pointed observations, and the ASM data have been analyzed to provide a light curve (including upper limits) of the source that extends back to early 1996. The exceptions are faint sources ($`<`$25 mCrab) discovered with the PCA during scans of the crowded galactic-center region. Source confusion renders some of these inaccessible to the ASM.
The number of transient sources thus far observed by RXTE (PCA and ASM) now exceeds 40, excluding Be star systems. We consider a transient to be a source that, for significant periods, has been below the detection level of the proportional-counter experiments (a few millicrabs) and which shows at least an order of magnitude change of intensity that exceeds the one-day ASM threshold ($``$15 mCrab at $`3\sigma `$ for a cataloged source). A more precise definition will be required for quantitative studies of transient occurrence rates. Also, a complete search to these levels requires care because our thresholds vary with position on the sky and with time because of variable source density and variable sky exposure.
Our most recent tabulation (November 1999) shows 23 ‘new’ transients studied with RXTE during the first known outburst, or series of outbursts. Among these, sixteen were discovered with RXTE (9 with the ASM and 7 with the PCA). In addition there are 19 recovered transients, i.e., previously known sources recovered after a period of quiescence. The RXTE/ASM has played a substantial role in the recoveries of these transients. A recovery often leads to the discovery of important new characteristics of the source. An earlier (1997) tabulation of transient sources observed with the RXTE/ASM may be found in Remillard (1999). An updated version will be released later.
The occurrences or reoccurences of transients and also the changes of state of persistent sources provide valuable opportunities for the study of the processes that modulate x-ray emission. The early detection and study of these events require a relatively sensitive all–sky monitor together with a highly sensitive narrow–field x–ray instrument that can be pointed rapidly to any point in the sky. The RXTE combines these capabilities for the first time in the history of x–ray astronomy. Thus new progress on the understanding of the accretion processes in these variable sources is now becoming possible.
## 2 ASM Light Curves
Light curves of transient sources from RXTE/ASM data show examples of repetitive behavior as well as considerable variation within the same source, e.g. from outburst to outburst. There are ‘failed outbursts’ interspersed with ‘normal’ outbursts. There are outbursts that usually turn completely ‘off’ (less than a few mCrab), but sometimes remain luminous at a low level for long periods. Outburst profiles include ‘fast rise with slow exponential-like decay’ sometimes interspersed with flat-topped and more irregular profiles. These light curves provide a quantitative challenge to models of accretion disk instabilities and of the capture of stellar wind from the companion.
In Figures 1–6, we present sample light curves (1.5 – 12 keV) for the 3.5–year period of RXTE’s operations. The intensities are given in ASM ct/s where 75 ct/s corresponds to the x-ray intensity of the Crab nebula. Most of the light curves are displayed in the form of 1–day averages of the 10 to 30 intensity measures typically obtained per day. The dates are given in MJD where:
1996 Jan. 0.0 = MJD 50082.0
1997 Jan. 0.0 = MJD 50448.0
1998 Jan. 0.0 = MJD 50813.0
1999 Jan. 0.0 = MJD 51178.0.
Figure 1 shows the light curves for six neutron-star binary sources. Figures 2, 3, and 4 show the light curves for eight black-hole-candidates, while Figs. 5 and 6 show six additional sources of diverse types. Most of these sources are transients in the classic sense that they are not detectable above some threshold for long periods of time. Figure 5 includes sources that we would not label as transients, but which do exhibit transient behavior in some sense of the word.
The light curves shown are for the entire bandwidth of the ASM detectors, namely 1.5 – 12 keV. In fact, the data are telemetered and stored in three energy channels, 1.5 – 3 keV, 3 – 5 keV, and 5 – 12 keV, which can serve as a further tool in the study of transient sources. The curves are shown here on a rather compressed scale; larger scales reveal substantially more detail. The intensities are available in both graphical and numerical form on the internet sites at MIT and GSFC, e.g., http://heasarc.gsfc.nasa.gov/xte\_weather/ and http://space.mit.edu/XTE/ASM\_lc.html.
Here, we point out features in the light curves of selected transients that should be pertinent to the classification of transient types. The characteristics of the previously known sources are well documented in van Paradijs (1995). The discovery references for a number of the new or recovered sources are listed in Remillard (1999).
We also point the reader to papers that can serve as an introduction to the more recent literature for the sources we have selected for illustration. This serves to give the reader a flavor of the research carried out by RXTE in the past few years. Much of this research has been stimulated by results from the ASM, e.g., because it provides notice of a changing state. This presentation is intended to point out the potential of such data for the understanding of accretion processes.
## 3 Neutron-star binaries; Fig. 1
4U 0115+63. This HMXB (high-mass x–ray binary) is a 3.6–s pulsar. Recently its optical counterpart has been reclassified as an O9e (Unger et al. 1999). In 1999 at MJD $``$51250 (hereafter simply 51250), it underwent a major outburst. Studies with BSAX and RXTE during this event led to the discoveries of high-harmonic cyclotron lines (Heindl et al. 1999; Santangelo et al. 1999). The outburst profile in the ASM light curve is quite symmetric. In addition, there were several weak outbursts in 1996 (50300) whose peaks are spaced by multiples of the 24–d orbital period. In larger scale plots they appear to have profiles similar to the larger outbursts.
X 1608–522. This well known LMXB (low-mass x–ray binary) is an x–ray burster as are many LMXBs. The source was in outburst when RXTE was launched and erupted again at 50850. The profile of the latter outburst shows a remarkably very fast rise, a rapid initial descent, a hesitation at about half maximum followed by a less rapid fall to the lower quiescent level. The source was undetectable ($`<`$10 mCrab) for almost a year (50500 – 50800) before the latter outburst. In contrast, it exhibited a sustained, variable low–level flux at 0.03 – 0.1 Crab for a year or more after each of the outbursts. This could be an indicator that the source is intrinsically unstable between on and off states. This low level (for the ASM) is several orders of magnitude greater than the quiescent luminosity of $`10^{33}`$ erg/s. See Rutledge et al. (1999) for a discussion of the quiescent state.
3A 1942+274. This Ariel source was recovered for the first time in many years by the ASM in late 1998. It was found to be a 16-s pulsar in RXTE PCA observations immediately upon its recovery (Smith & Takeshima 1998). The previously suggested optical counterpart (Israel, Polcaro & Covino 1998) is now considered to be an unlikely candidate (Israel, pvt. comm.). The irregularly spaced multiple peaks in the light curve may represent enhanced accretion near periastron in an $``$80-d elliptical orbit (Campana, Israel & Stella 1999). The source remains active through at least mid November 1999.
XTE J2123–058. This previously unknown source at high galactic latitude ($`36\mathrm{deg}`$) was discovered in ASM data (Levine, Swank & Smith 1998). It is an atoll LMXB with twin x-ray kHz oscillations and x-ray bursts (Homan et al. 1999). An optical counterpart with V = 17 and a 6–h orbit was immediately discovered (Tomsick et al. 1999a) with orbital modulations that continued beyond the x-ray outburst (Soria, Wu & Galloway 1999). The outburst at 51000 is quite weak (85 mCrab); it exhibits the same hesitation during the decay seen in 1608–522. Possible precursor activity $``$100 d before the peak is apparent at $``$15 mCrab, but may be an artifact from x-ray Solar contamination.
Rapid Burster. The ASM data show 6 outbursts of activity since the RXTE launch. They reaffirm the $``$200-d previously-noted quasi-periodicity of these outbursts, but for the first time show the extent and evolution of each outburst. Each outburst lasts about 5 weeks and, in PCA data, exhibits two phases, the first of which is characterized by Type I (thermonuclear) bursts and the second by Type II (accretion) bursts (Guerriero et al. 1999).
Aql X-1. This well-known atoll, bursting, LMXB recurring transient has recently had its optical identification (V1333 Aql) clarified. The counterpart is the western component of a 0.5-arcsec double with V= 21.6 in quiescence and with a late K-star classification (Chevalier et al. 1999). ASM detections of outburst states made possible studies of kHz oscillations in the active state that include demonstrations of (1) a sudden decrease in frequency and flux after an x-ray burst (Yu, Zhang & Zhang 1999) and (2) low-energy lags of $``$1 radian consistent with a Doppler-delay model (Ford 1999).
The ASM light curve of Aql X-1 shows 5 outbursts, including those at the beginning and the end of the plot and another two “failed” bursts at 50270 and 51200. The intervals between these events range from 200 to 300 d. Two of the outburst light curves (50700 and 50900) are roughly symmetric in shape with comparable rise and fall times. The other two have faster rises than decays. The failed outbursts may indicate that the conditions for outburst are marginal, as we suggest for the low-level persistent flux from x1608–522. Aql X-1 lies on the thermal-viscous instability boundary derived by van Paradijs (1996) for neutron stars.
## 4 Black-hole candidates; Figs 2, 3, 4
4U1630–47. This long-known transient and black-hole candidate (from soft-hard spectral components) exhibits outbursts roughly in accord with the reported $``$600–700 d intervals (Kuulkers et al. 1997). Three outbursts are seen in the ASM data with separations $``$700 d and $``$450 d. The longer interval follows the outburst with the largest integrated flux. All three outbursts exhibit sharp maxima at the leading and trailing edges of the profile, and the leading edges rise faster than the trailing edges decay. In spite of this, the latter two outbursts have different shapes; they appear to have failed to attain the full profile of the first outburst. The flux remains markedly above threshold after the second outburst. Again this behavior could be an indicator of marginal instability.
The 1996 outburst (50200) revealed deep narrow (minutes) x-ray absorption dips in RXTE/PCA data (Kuulkers et al. 1998). This outburst is included in the historical review of outburst behavior in 4U 1630–47 (Kuulkers et al. 1997). This review points out the heretofore unrecognized complexity of the outburst behavior of this source The evolution of the x- ray spectral components of the 1998 outburst (50850) from BeppoSAX are presented by Oosterbroek et al. (1998).
GRO J1655–40. This is a well established black-hole system with a dynamical mass for the collapsed object of $``$6–7 solar masses (Orosz & Bailyn 1997, Shabaz et al. 1999). It exhibits superluminal radio jets which establish it as a “microquasar” (Tingay et al. 1995). First discovered in 1994, it was solidly below threshold prior to and after the $``$450–d 1996–1997 outburst. The optical turn-on was found to precede the x-ray emission by $``$5 days (Orosz et al. 1997). Observations with the PCA exhibited several QPOs including 300 Hz when the source spectrum was particularly hard (Remillard et al. 1999b). The evolution of x-ray spectral components is given by Mendez, Belloni & van der Klis (1998), Tomsick et al. (1999b), and Sobczak et al. (1999a). Echo mapping (x-ray to optical) locates the reprocessing region to be in the accretion disk rather than the mass donor star (Hynes et al. 1998).
XTE J1748–288. This RXTE/ASM-discovered transient (Smith, Levine & Wood 1998) exhibited a single outburst at 51000 with a very rapid rise ($``$2 d) and slow exponential-like decay. It was above threshold for 60 d and was detected to at least 100 keV with BATSE. (Harmon et al. 1998). PCA observations showed QPO at 0.5 and 32 Hz (Fox & Lewin 1998). The spectral and QPO evolution has been studied by Revnivtsev, Trudolyubov & Borozdin (1999). Transient radio emission (Hjellming et al., 1998, Fender & Stappers 1998) became extended after 10 days. The jet motion at $`>`$20 mas/d indicated an apparent speed $`>`$ 0.93c for a $``$8 kpc distance. Later the speed decreased and the leading edge of the radio jet brightened due to a shock in the interstellar medium, the first such shock known in a galactic source. There is no reported optical counterpart; its galactic coords. are 0.7, –0.2).
XTE J1755–324. This source, also ASM discovered (Remillard et al. 1997), was noteworthy for its extremely soft color in the ASM (HR2 = 0.3) which suggests black-hole candidacy. (HR2 is the harder color of the two ASM x-ray energy colors.) The intensity profile was similar to XTE J1748–288 in that it rose in $``$2 d and decreased quasi-exponentially. However, unlike 1748–288, but reminiscent of the neutron-star system 1608–522 (Fig. 1), it halted its descent after about 50 days, increased slightly to a second maximum and then descended to non-detectability after a total time of $``$105d. The temporal and spectral evolution is described by Revnivtsev, Gilfanov & Churazov (1998) who find it to be a “canonical” x-ray nova such as Nova Muscae 1991. Again there is no reported optical identification at this writing; galactic coords. 358.0, –3.6.
XTE J2012+381. This transient, discovered with the ASM (Remillard, Levine & Wood 1998), had a hard initial spike and also reached a very low hardness ratio (HR2 = 0.4). In ASCA data it had an ultrasoft spectrum with a hard tail suggesting a black hole (White et al. 1998). A radio counterpart was detected (Hjellming & Rupen 1998a, Pooley 1998). A V = 21.3 star (1.1 arcsec from a foreground 18th magnitude star) lies within 0.4 arcsec of the radio source and is tentatively identified as the optical counterpart (Hynes et al. 1999). The light curve is strongly double peaked with an additional low maximum 150 d after the onset.
GX 339–4. This long-known black-hole candidate with a persistent x-ray flux entered a bright and soft (HR2 $``$0.3) state beginning at $``$50800 and returned to its low state $``$400 d later. It is not generally considered a “transient”; nevertheless the infrequent high soft states could be considered so. The ASM high/soft state is accompanied by a low high-energy (BATSE) flux and a marked reduction of radio emission (Fender et al. 1999a). The transition is reminiscent of the 1996 Cyg X-1 transition to its high/soft state (Wen et al. 1999 and refs. therein). The temporal and spectral characteristics during the rise and at maximum showed a “typical” high/soft state (Belloni et al. 1999b). See also the multiwavelength (radio, optical, x-ray) studies in a 3-paper series (e.g., Smith, Filippenko & Leonard 1999).
XTE J1915+105 (Fig. 3). This remarkable “microquasar” (Mirabel & Rodriguez 1994) has been active continuously since the launch of RXTE. The x–ray light curves show what is probably the most dramatic variability of any known x-ray source. Extensive PCA observations (vertical-line markers at top) have permitted the definition of a number of distinct accretion states (e.g., Greiner, Morgan & Remillard 1997) This source also exhibits superluminal radio jets (e.g. Fender et al. 1999b). A dynamical mass does not exist for this source, but its high luminosity strongly implies a black-hole compact object. The object is replete with unusual x-ray oscillatory modes that repeat on time scales of tens to thousands of seconds (e.g., see Muno et al. 1999). One such mode has been associated with micro-radio outbursts with high confidence (Pooley & Fender 1997, Eikenberry et al. 1998, Mirabel et al. 1998), thus linking specific accretion configurations with jet creation. These events appear to be associated with the loss via accretion of the inner portion of the accretion disk (see Belloni et al. 1997, Pooley & Fender 1997). The radio–x-ray association is also seen on longer time scales in Fig. 3 where increased radio fluxes are sometimes associated with low hard x–ray states, e.g. at 50750, with, in addition, bright flaring at the onset and end of such low states. The source is replete with low frequency x–ray QPOs at frequencies that vary with time and also a 67-Hz QPO that reappears when the spectrum is hard (Morgan, Remillard & Greiner 1997). Recent work has addressed the interplay of the QPOs and the variation of the spectral components, e.g. Muno et al. (1999) and Markwardt, Swank & Taam (1999).
XTEJ1550–564 (Fig. 4). This was the first previously-unknown very–bright transient found in the RXTE era. It was discovered in ASM data with new triggering software soon after its onset (Smith 1998). This permitted PCA observations during the rise, the times of which are marked in the figure. The x-ray light curve with its narrow peak, flat plateau and second maximum is not typical of a prototype x-ray nova, e.g., 0620–00. It reaches 6.8 Crab brightness in the dramatic brief flare at $``$51050. X-ray emission was observed to 200 keV with BATSE (Wilson et al. 1998). The evolution of the x-ray spectral components in the PCA and ASM data were tracked by Sobczak et al. (1999b). The source was found in the very high, high/soft, and intermediate canonical outburst states of black-hole x-ray novae. X- ray QPOs were abundant from 0.05 to 185 Hz (Cui et al. 1999; Remillard et al. 1999a). An optical counterpart brightened about 4 magnitudes over the quiescent B $``$22 state (Jain et al. 1999). The extinction E(B–V) = 0.7 indicated a distance of 2.5 kpc and a luminosity consistent with a K0–K5 star (Sanchez-Fernandez, et al. 1999). A likely radio counterpart was found (Campbell-Wilson et al. 1998).
## 5 Various sources; Fig. 5
Here we illustrate the diversity of x-ray variation in a few other selected sources. Most would not be deemed transients under the conventional definition. However they all exhibit transient behavior in the broader sense.
CI Cam. Here we see the remarkably brief outburst from the symbiotic star CI Cam, discovered with the ASM (Smith et al. 1998). The rise took place over a few hours, and the exponential-like fall (Fig. 6) showed a time scale that began at $``$0.5 d and later became $``$2.3 day (Belloni et al. 1999a). It was above the ASM threshold for only 9 days. The ASM detection enabled studies with ASCA, BSAX and the RXTE PCA as well as in the radio and optical. This object was reported to have corkscrew radio jets reminiscent of SS433 (Hjellming & Mioduszewski 1998). Belloni et al. (1999a) suggest this is a Be-star + neutron-star binary system but do not exclude a black-hole; see Belloni et al. for references to several studies of this outburst.
GS 2138+568 (=Cep X-4?). This source was discovered in 1972 with OSO–7. It was rediscovered with Ginga which found it to be a 66-s pulsar. It is believed to be a Be-star system with a total of four recorded outbursts. A study with BATSE and RXTE (Wilson, Finger & Scott 1998) of outbursts in 1993 and 1997 also serves as a review of the literature. In the ASM data, the source remained above threshold for about 40 days. It is among the fainter sources ($``$30 mCrab) detectable with the ASM in 1–day averages.
4U 1705–44. This modestly bright LMXB burster with no known optical counterpart exhibits large intensity variations that have rise and fall times that quite consistently are on the order of 50 days. This source has historical ‘off’ states but has essentially always been detectable since Feb. 1996 in the ASM data. It also exhibits radio emission. Kilohertz QPO have been discovered with RXTE (Ford, van der Klis & Kaaret 1998). The fast timing behavior at lower frequencies based on the EXOSAT archive are reported by Berger & van der Klis (1998).
SMC X-1. This well-known high-mass system with an eclipsing (3.9 d) x–ray pulsar (0.71s) reveals itself conclusively to have a quasi 60-d period in the ASM data. It is probably a precessing disk system (Wojdowski et al. 1998). The pulsar continues to show a steady spin-up rate (Kahabka & Li 1999).
Cyg X-3. This unusual binary, with a 4.8-h x–ray period (probably orbital) exhibits spectacular radio flares. Its ASM x–ray light curve exhibits one sustained flaring interval and several shorter ones. Bright radio flares occur during the sustained active period, with peak intensities of 3 to 9 J at wavelength 13 cm at 50465, 50485, and 50610 (Ogley et al. 1998). The orbital period evolution has recently been revisited by Matz (1997).
Mkn 501. This blazar exhibits a year-long period of x-ray activity in the ASM data from 50400 to 50800. This active period was accompanied by dramatic flux variations and a high average flux ($``$1.4 Crab) at TeV energies (Quinn et al. 1999; Aharonian et al. 1999). It was on MJD 50607 that an intense rapid (hour scale) TeV flare was detected (Quinn et al.). Aharonian et al. report an intensity correlation with ASM data with zero time lag during this active period. In 1998 the TeV flux had decreased to $``$20% of the Crab. The ASM data serve to guide TeV observers in target selection. The light curve (4-day averages) shows the capability of the ASM for the study of the brighter AGN sources which, even so, are quite weak in the ASM. Mkn 501 reaches 35 mCrab in this plot.
## 6 Conclusion
The light curves shown here are illustrate the richness of the ASM data and improve the prospect for better understanding of the accretion processes underlying transient behavior. The outburst profiles, the marginally-on states, the durations of on and off states, and the hardness parameters should all serve as diagnostics of these processes. A planned reanalysis of the entire data base with improved procedures should provide a uniform search for transients sources down to $``$7 mCrab at positions away from the galactic center. With good fortune the RXTE/ASM will continue to provide comprehensive light curves to the community for at least several more years.
## Acknowledgements
The authors are grateful to the RXTE/ASM team at MIT and GSFC and to NASA for support under Contract NAS5–30612 |
warning/0001/hep-th0001100.html | ar5iv | text | # Gravitating (bi-)sphalerons
## I Introduction
The coupling of non-abelian field theories to Einstein gravity constitutes natural extensions of the Einstein-Maxwell equations. One of the surprising issues of such possibilities has been the discovery of regular, finite energy classical solutions in the Einstein-Yang-Mills equations : the series of Bartnik-Mckinnon solutions.
On the other hand the Weinberg-Salam model theory, which couples the SU(2)$``$U(1) Yang-Mills theory to a doublet of complex Higgs fields emerges more and more as the theory of the unified weak and electromagnetic forces. This particular lagrangian is a member of the family of Yang-Mills-Higgs models among which the Georgi-Glashow model is another distinguished example.
One interesting property of the Weinberg-Salam lagrangian is that it admits an unstable, finite energy classical solution called the sphaleron . Generally, sphaleron solutions are expected to play a role in the understanding of baryon non conserving phenomenon which are allowed to take place in the Weinberg-Salam model . When the parameter determining the mass of the Higgs field increases, additional solutions, the bisphalerons, bifurcate from the sphaleron . This feature seems to be related to the underlying non-linear character of the classical equations and to the spontaneous breakdown of the symmetry .
It is therefore natural to couple the Weinberg-Salam model to the Einstein-Hilbert gravitational lagrangian and to study the response to gravity of the classical solutions available in the flat limit. This problem has been partly investigated in and more recently in a report on the aspects of Einstein-Yang-Mills-Higgs equations but it was not treated in details. In particular the response of the bisphaleron solution to gravity was not investigated. In this paper, we reconsider the classical equations of the Einstein-Weinberg-Salam model for spherically symmetric fields (we assume the Weinberg angle, related to the U(1) part of the gauge group, to be zero). Then the gravitating sphaleron and bisphaleron are constructed and assemble into branches of solutions which evolve with $`\alpha `$, the ratio of the vector-boson mass to the Planck mass. The evolution of the different solutions as functions of $`\alpha `$ and $`M_H`$ are studied in details.
## II The equations
We consider the gauge theory for an SU(2)-Higgs doublet minimally coupled to the Einstein-Hilbert gravitational lagrangian :
$$L=\sqrt{g}[L_G+L_M]$$
(1)
with
$$L_G=\frac{1}{16\pi G}R$$
(2)
$$L_M=\frac{1}{4}F_{\mu \nu }^aF^{a\mu \nu }+(D_\mu \mathrm{\Phi })^{}(D^\mu \mathrm{\Phi })\frac{\lambda }{4}(\mathrm{\Phi }^{}\mathrm{\Phi }\frac{v^2}{2})^2$$
(3)
and with the usual definitions for the fields strenghts and covariant derivatives
$$F_{\mu \nu }^a=_\mu A_\nu ^a_\nu A_\mu ^a+gϵ_{abc}A_\mu ^bA_\nu ^c$$
(4)
$$D_\mu \mathrm{\Phi }=_\mu \mathrm{\Phi }+g(A_\mu ^a\sigma _a)\mathrm{\Phi }$$
(5)
The matter part $`L_M`$ of this Lagrangian approximates the Weinberg-Salam model of electroweak interactions in the limit of vanishing Weinberg angle $`\theta _W`$ (i.e. the gauge group SU(2)$`\times `$U(1) is restricted to SU(2)).
Here we will study the classical, spherically symmetric solutions of the Lagrangian $`L`$. In this purpose, we employ the Schwarzschild-like coordinates for the metric
$$ds^2=A^2Ndt^2+N^1dr^2+r^2(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$
(6)
and we introduce, as usual, the mass function $`m(r)`$ by means of
$$N(r)=1\frac{2m(r)}{r}.$$
(7)
Then we use the standard spherically symmetric ansatz for the spatial components of the gauge field (fields are static and $`A_0=0`$) and for the Higgs fields (the notations of are used)
$$A_i^a=\frac{1f_A(r)}{gr}ϵ_{aij}\widehat{x}_j+\frac{f_B(r)}{gr}(\delta _{ia}\widehat{x}_i\widehat{x}_a)+\frac{f_C(r)}{gr}\widehat{x}_i\widehat{x}_a,$$
(8)
and
$$\mathrm{\Phi }_1=0,\mathrm{\Phi }_2=\frac{v}{\sqrt{2}}[H(r)+iK(r)(\widehat{x}^a\sigma _a)].$$
(9)
It is well known that this ansatz for the matter fields is plagued with a residual gauge symmetry. Along with , we fix this freedom by imposing the axial gauge
$$x^iA_i=0f_C(r)=0.$$
(10)
It is also convenient to introduce the dimensionless coordinate $`x`$ and mass function $`\mu `$ defined by
$$x=g\frac{v}{\sqrt{2}}r,\mu =g\frac{v}{\sqrt{2}}m,$$
(11)
as well as the dimensionless coupling constants $`\alpha `$, $`ϵ`$
$$\alpha ^2=4\pi G\frac{v^2}{2},ϵ=\frac{\lambda }{g^2}=\frac{1}{2}(\frac{M_H}{M_W})^2,$$
(12)
where $`G`$ is Newton’s constant, $`v`$ is the Higgs field expectation value, $`M_H`$ is the Higgs boson mass and $`M_W`$ is the gauge boson mass. If finite, the quantity $`\mu (\mathrm{})`$ defines the mass of the solution.
With these ansatz and definitions, it can be checked after an algebra that the classical equations of the Lagrangian (1) are equivalent to the equations derived from the following two-dimensional action
$$S=𝑑t𝑑x()$$
(13)
$$=A[\frac{1}{2}(N+xN^{}1)+\alpha ^2M]$$
(14)
where the prime means the derivative with respect to $`x`$ and the quantity $`M`$ is defined by
$`M`$ $`=NV_1+V_2`$ (15)
$`V_1`$ $`=(f_A^{})^2+(f_B^{})^2+2x^2((H^{})^2+(K^{})^2)`$ (16)
$`V_2`$ $`={\displaystyle \frac{1}{2x^2}}(f_A^2+f_B^21)^2+ϵx^2(H^2+K^21)^2`$ (18)
$`+(H(f_A1)+Kf_B)^2+(K(f_A+1)Hf_B)^2.`$
The classical equations then reduce to the following system of six non-linear differential equations :
$`\mu ^{}`$ $`=`$ $`\alpha ^2M`$ (19)
$`A^{}`$ $`=`$ $`2A\alpha ^2{\displaystyle \frac{1}{x}}V_1`$ (20)
$`(ANf_A^{})^{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}A({\displaystyle \frac{V_2}{f_A}})`$ (21)
$`(ANf_B^{})^{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}A({\displaystyle \frac{V_2}{f_B}})`$ (22)
$`(x^2ANH^{})^{}`$ $`=`$ $`{\displaystyle \frac{1}{4}}A({\displaystyle \frac{V_2}{H}})`$ (23)
$`(x^2ANK^{})^{}`$ $`=`$ $`{\displaystyle \frac{1}{4}}A({\displaystyle \frac{V_2}{K}})`$ (24)
It is important to remark that the equations are still invariant under the continuous global transformation
$`f_A+if_B`$ $`(\mathrm{exp}(2i\mathrm{\Omega }))(f_A+if_B)`$ (25)
$`H+iK`$ $`(\mathrm{exp}(i\mathrm{\Omega }))(H+iK)`$ (26)
where $`\mathrm{\Omega }`$ is a real constant. Since the regularity of the functional $`V_2`$ at $`x=0`$ clearly requires $`f_A^2(0)+f_B^2(0)=1`$, we can fix the above symmetry by chosing $`f_A(0)=1`$,$`f_B(0)=0`$. This partly fix the boundary conditions which will be discussed more completely in the next section. Then, only a change of sign of the Higgs field can still be chosen arbitrarily.
Let us close this section by discussing the limit $`\alpha 0`$. From the definition (12) it is clear that the limit of vanishing $`\alpha `$ can be considered in different ways. If we keep $`v`$ fixed and let $`G0`$ we obtain the flat limit (gravity decouples). The appropriate parameter which defines the classical energy of the solution is
$$E_c=\mathrm{lim}_x\mathrm{}\frac{1}{\alpha ^2}\mu (x).$$
(27)
It restitutes the physically-meaningful classical energy of the flat (bi-)sphaleron . In order to study the equations in the limit $`v0`$ and $`G`$ fixed, then it is necessary to rescale the radial variable $`x`$ and mass $`\mu `$ according to
$$y\frac{x}{\alpha },\rho \frac{\mu }{\alpha }$$
(28)
and to set $`\alpha =0`$ afterwards in the equations. We then obtain
$`{\displaystyle \frac{d\rho }{dy}}`$ $`=N(({\displaystyle \frac{df_A}{dy}})^2+({\displaystyle \frac{df_B}{dy}})^2)+{\displaystyle \frac{1}{2y^2}}(f_A^2+f_B^21)^2`$ (29)
$`{\displaystyle \frac{dA}{dy}}`$ $`=2A{\displaystyle \frac{1}{y}}(({\displaystyle \frac{df_A}{dy}})^2+({\displaystyle \frac{df_B}{dy}})^2)`$ (30)
$`{\displaystyle \frac{d}{dy}}(AN{\displaystyle \frac{df_J}{dy}})`$ $`=A{\displaystyle \frac{1}{y^2}}f_J(f_A^2+f_B^21),J=A,B.`$ (31)
In particular, the degrees of freedom $`H,K`$ related to the gauge field decouples and, setting $`f_B=0`$ in the equations above, the Bartnik-McKinnon (BM) equations are recovered. In passing note that we have not succeeded in constructing solutions of (31) which are not related to the BM solution by mean of (26).
The corresponding energy is given by
$$E_{BM}=\mathrm{lim}_y\mathrm{}\rho (y)=\mathrm{lim}_x\mathrm{}\frac{\mu (x)}{\alpha }$$
(32)
## III Boundary conditions
The regularity of the solution at the origin, the finiteness of the mass $`\mu (\mathrm{})`$ and the requirement that the metric (6) approaches the Minkowski metric for $`x\mathrm{}`$ lead to definite boundary conditions (BC) for the six radial functions $`\mu ,A,f_A,f_B,H,K`$. As far as the metric functions are concerned we have to impose
$$\mu (0)=0,A(\mathrm{})=1$$
(33)
For the matter field equations two sets of BC lead to regular and finite energy solutions.
### 1 The sphaleron BC
The flat sphaleron (and also the gravitational one) has $`f_B(x)=H(x)=0`$. The remaining functions have to obey
$`f_A(0)=1,`$ $`K(0)=0`$ (34)
$`f_A(\mathrm{})=1,`$ $`K(\mathrm{})=1`$ (35)
### 2 The bisphaleron BC
The flat bisphalerons are characterized by the four non-trivial radial functions and we will see in the next section that they are continuously deformed by gravity. Taking into account the fixing of the global symmetry (26), at the origin the functions have to obey
$`f_A(0)=1,`$ $`f_B(0)=0`$ (36)
$`H^{}(0)=0,`$ $`K(0)=0`$ (37)
In the limit $`x\mathrm{}`$, they have to approach constants in the following way
$`\mathrm{lim}_x\mathrm{}(f_A(x)+if_B(x))`$ $`=\mathrm{exp}(2i\pi q)`$ (39)
$`\mathrm{lim}_x\mathrm{}(H(x)+iK(x))`$ $`=\mathrm{exp}(i\pi q)`$ (40)
So, the solutions of this type are characterized by a real constant $`q[0,1[`$; this parameter has to be determined numerically and depends on $`ϵ`$ and $`\alpha `$.
## IV The solutions
We first describe the solutions for $`ϵ=0.5`$, a generic value of $`ϵ`$ for which the sphaleron is the unique solution of the flat equations. The flat sphaleron is there for $`\alpha =0`$, with an energy $`E_c3.64`$, and gets continuously deformed for $`\alpha >0`$. In particular the function $`N(x)`$ develops a minimum which becomes deeper while $`\alpha `$ increases. This branch of gravitational sphalerons, let us call it $`S_l(\alpha )`$, exists up to a critical value $`\alpha =\alpha _s0.3095`$ and no solution of this type exists for $`\alpha >\alpha _s`$. However there exists a second branch of solutions that we call $`S_u(\alpha )`$ for $`\alpha [0,\alpha _s]`$. For fixed $`\alpha `$ the solution of the branch $`S_u`$ has a higher mass $`\mu _s\mu (\mathrm{})`$ and a deeper minimum of $`N(x)`$ than the corresponding one on the branch $`S_l`$. This is illustrated on Fig.1; the indexes $`l,u`$ in $`S_{l,u}`$ refer to the lower, upper branch when comparing the mass.
In the limit $`\alpha \alpha _s`$, the solutions $`S_u`$ and $`S_l`$ converge to a common limit. It has
$$\mu (\alpha _s)0.290,N_{min}(\alpha _s)0.513$$
(41)
The transition from the branch $`S_l`$ to the branch $`S_u`$ is completely smooth.
In the limit $`\alpha 0`$, the solution on $`S_u`$ tends to the first solution of the BM series. This is recovered by rescaling the radial variable $`x`$ according to $`y=x/\alpha `$ and taking the limit $`\alpha =0`$, as explained in the previous section. The BM solution is well known but, for completeness, we present its profile on Fig. 2. It has
$$\mathrm{lim}_\alpha \mathrm{}\frac{\mu (\alpha )}{\alpha }0.83,N_{min}0.242$$
(42)
We have studied the gravitating sphaleron for a few values of $`ϵ`$ and determined the corresponding critical value $`\alpha _s`$. These are presented in Fig. 3. The special value $`\alpha _s(0)0.376`$ agrees with the result of . We notice the rapid decrease of $`\alpha _s`$ for the low values of $`ϵ`$.
We next discuss the solutions for a value of $`ϵ`$ which allows both sphaleron and bisphaleron solutions to exist. This occurs for $`ϵ>72.0`$ and here we choose generically $`ϵ=100.0`$ for which the energies of the flat bisphaleron and sphaleron are given by
$$E_b4.88,E_s4.93$$
(43)
These two solutions get deformed by gravity when $`\alpha >0`$ and develop two branches of solutions which we will denote $`B_l(\alpha )`$ (with mass $`\mu _b`$) and $`S_l(\alpha )`$ (with mass $`\mu _s`$). They exist respectively up to $`\alpha =\alpha _b0.2247`$ and $`\alpha =\alpha _s0.2218`$. At these critical values, we have respectively
$`\mu _b(\alpha _b)`$ $`0.2080,N_{min}`$ $`0.464`$ (44)
$`\mu _s(\alpha _s)`$ $`0.2047,N_{min}`$ $`0.463`$ (45)
Again, these solutions are continued by upper branches which we denote respectively by $`B_u(\alpha )`$ and $`S_u(\alpha )`$ and which coincide with $`B_l(\alpha )`$ and with $`S_l(\alpha )`$ respectively at $`\alpha =\alpha _b`$ and $`\alpha =\alpha _s`$. The two bisphaleron-solutions corresponding to $`B_l`$ and $`B_u`$ for $`\alpha =0.2`$ are presented on Fig. 4.
All along the two branches the energy of the gravitating bisphaleron stays slightly lower than the one of the sphaleron. Of course both quantities become equal when $`\alpha =\alpha _c`$, this is illustrated by Fig. 5. This figure also clearly indicates that, at the critical value $`\alpha _s`$ (resp. $`\alpha _b`$), the energies of the two gravitating sphaleron $`S_l,S_u`$ (resp. bisphaleron $`B_l,B_u`$) solutions form a cusp. When bisphalerons are present, four solutions are available on some interval of $`\alpha `$ and there are two cusps.
Completely similarly to the case $`ϵ=0.5`$, the branch $`S_u`$ converges to the Bartnik-Mckinnon solution in the limit $`\alpha =0`$. The scenario with the branch $`B_u(\alpha )`$ is different: decreasing $`\alpha `$ from $`\alpha _b`$ we observe that the different radial functions composing this solution uniformly approach their counterparts of sphaleron solution $`S_u`$. This occurs for $`\alpha \alpha _c0.185`$ as illustrated on Fig. 6. On this figure the quantities $`f_B(\mathrm{})`$ and $`H(0)`$ which characterize the bisphaleron are plotted for the different branches. The minimal value $`N_{min}`$ of the function $`N(x)`$ is superposed on the figure.
In view of these results we can say that the branch $`B_u`$ of solutions bifurcates from the branch $`S_u`$ at $`\alpha 0.185`$.
In order to have a qualitative idea of how the gravitating (bi)-sphaleron solutions behave for higher values of $`ϵ`$, we solved the equations for $`ϵ=800.0`$. There the gravitating sphaleron and bisphaleron exist respectively up to $`\alpha _s0.209`$ and $`\alpha _b0.221`$. The branch $`B_u`$ bifurcates from $`S_u`$ at $`\alpha 0.07`$
This suggests that, when $`ϵ`$ increases, the branch $`B_u`$ bifurcates from $`S_u`$ for lower values of $`\alpha `$ and that the gravitating bisphaleron exists on an interval which becomes relatively larger than the interval of existence of the sphaleron.
In the limit of the non-linear sigma model $`ϵ=\mathrm{}`$, it is known that the sphaleron is discontinuous at the origin unlike the bisphaleron which continues to exist as a regular solution. We expect that in this case the bisphaleron branch $`B_u`$ will exist up to $`\alpha =0`$ where it will approach the Bartnik-McKinnon solution.
Let us finally say some words about the stability of the various solutions. For $`ϵ<72.0`$ the flat sphaleron possesses a single direction of instability (a negative mode) and for $`ϵ>72.0`$ the sphaleron has two directions of instability while the bisphaleron has one . In this respect the sphaleron (resp. bisphaleron) is interpreted as the minimal energy barrier bewteen topologically different vacua of the Weinberg-Salam model for $`ϵ<72.0`$ (resp. $`ϵ>72.0`$).
The shape of the plot of the energy, with the two branches terminating into a cusp is typical for catastrophe theory (see e.g. ) and the use of arguments based on Morse theory suggests many useful information about the stability of the different solutions. For instance the number of negative modes for the solutions on the upper branch exceed by one unit the number of negative modes for the solutions on the lower branch. Such a reasonning was demonstrated to be correct e.g. in .
Using the same arguments in the present context indicates that the solutions on the branch $`S_l`$ of Fig. 1 have one direction of instability while the solutions of the branch $`S_u`$ (and therefore also the BM solution) have two.
For the solutions of Fig. 5 we have one (resp. two) directions of instability for the solutions on $`B_l`$ (resp. $`B_u`$) and two directions of instability for the solutions on $`S_l`$. On the branch $`S_u`$ the number of negative modes is equal to two on the interval $`\alpha [0,\alpha _c]`$ (i.e. before the branch $`B_u`$ has bifurcated) and equal to three for $`\alpha [\alpha _c,\alpha _s]`$ (i.e. after the bifurcation). The number of instable modes at the approach of the BM solution is then equal to two, irrespectively of the parameter $`ϵ`$.
Obviously these deductions would need to be confirmed by more elaborated calculations.
## V Conclusion
The coupling of the Yang-Mills-Higgs equations to gravitation often leads to interesting new properties of the available gravitating classical solutions .
The classical equations of the Weinberg-Salam model possess a rich pattern of bifurcations of bispahleron solutions from the sphaleron solution . It is therefore natural to attempt to study the critical phenomenon which occur in the Einstein-Weinberg-Salam equations. In response to gravity, parametrized by the quantity $`\alpha `$ defined in (12), the equations exhibit another type of critical phenomenon : the occurence of two branches of solutions which terminate at a critical value of $`\alpha `$; this was noted in for $`M_H=0`$ but the phenomenon occurs for generic values of $`M_H`$.
Here we were interested only in global solutions (i.e. the metric function $`N`$ has no zero on $`[0,\mathrm{}]`$), but another interesting feature of gravitationally deformed classical solutions (soliton or sphaleron) is the existence of black hole solutions where the function $`N(r)`$ develops an horizon at some finite value $`r=r_h`$, i.e. $`N(r_h)=0`$. We guess that flat (bi-)sphalerons could also be deformed in this way and produce sphaleron black holes.
Figure captions
Fig. 1.
The mass of the gravitating sphaleron solutions (represented by $`\mu _s`$ and by $`\mu _s/\alpha `$) and the minimal value $`N_m`$ of $`N`$ are reported in function of $`\alpha `$ for $`ϵ=0.5`$.
Fig. 2.
The profiles of the functions $`N,A,f_A`$ of the solution $`S_u`$ as functions of $`y=x/\alpha `$ for $`\alpha =0.01`$ and $`ϵ=0.5`$.
Fig. 3.
The critical value $`\alpha _s`$ as a function of $`ϵ`$ for the low values of $`ϵ`$.
Fig. 4.
The profiles of the functions $`N,A,f_A,f_B,H,K`$ of the bisphalerons $`B_l`$ (in dotted) and $`B_u`$ (in solid) for $`\alpha =0.2`$ and $`ϵ=100.0`$.
Fig. 5.
The masses $`\mu _s,\mu _b`$ of the gravitating sphaleron (solid) and bisphaleron (dotted) as functions of $`\alpha `$ in the region of the critical value. The branch $`B_u`$ stops at $`\alpha 0.185`$, as indicated by the star.
Fig. 6.
The values of $`H(0),f_B(\mathrm{}),N_m`$ for the gravitating bisphaleron in function of $`\alpha `$. The corresponding value of $`N_m`$ for the sphaleron is represented by the dotted line. |
warning/0001/astro-ph0001164.html | ar5iv | text | # The Effect of Resistivity on the Nonlinear Stage of the Magnetorotational Instability in Accretion Disks
## 1 Introduction
In recent years a more complete understanding of the origin of angular momentum transport and turbulence in accretion disks has emerged. The discovery that weak magnetic fields render a differentially rotating plasma unstable (Balbus & Hawley 1991; 1992) has directed the focus of research efforts to magnetohydrodynamic (MHD) processes and the onset and evolution of the magnetorotational instability (MRI). The MRI is a linear, local instability whose existence is independent of both field orientation and strength. The action of the MRI directly produces outward angular momentum transport, as required for accretion disks to accrete.
Studies of the nonlinear evolution and saturation of the MRI are required in any comparison between theory and observation; such studies rely on numerical MHD simulations. Many such numerical investigations have been carried out in recent years. Local three-dimensional MHD simulations have shown that turbulence is initiated and sustained by the MRI (Hawley, Gammie, & Balbus 1995, hereafter HGB; Brandenburg et al. 1995; Stone et al. 1996), and that this turbulence supports a significant outward flux of angular momentum. Both the turbulent energy and angular momentum fluxes are dominated by Maxwell rather than Reynolds stress. Simulations that begin with a vanishing mean flux have been used to examine the implications of the MRI for dynamo action in accretion disks (Brandenburg et al. 1995; Hawley, Gammie, & Balbus 1996, hereafter HGB2) These have demonstrated that the MRI is capable of amplifying and sustaining an initially weak magnetic field for many resistive decay times, thus satisfying the minimum definition of a dynamo. Moreover, these studies show that this process cannot be described by kinematic dynamo theory: the effect of the Lorentz force on the flow can never be ignored, even when the field is weak. A comprehensive review of these and many other results pertaining to the MRI and angular momentum transport in accretion disks is given by Balbus & Hawley (1998, hereafter BH).
Most of the numerical studies of the nonlinear stage of the MRI reported to date have adopted the assumption of ideal MHD, i.e. infinite conductivity, so that the field is perfectly frozen-in to the gas. However, in cold, dense plasmas such as might be expected at the centers of protostellar disks (Stone et al. 1998), or disks in dwarf novae systems (Gammie & Menou 1997), the ionization fraction may become so small that this approximation no longer holds. Obviously, the appropriate representation of some region of interest within a partially ionized plasma depends on the specific densities, temperatures, and ionization fractions therein. Different nonideal MHD effects can be important for different physical conditions (see, e.g., Parker 1979). For example, the ambipolar diffusion regime occurs when the neutral-ion collision time is too long to prevent the gas from drifting across magnetic field lines; generally this corresponds to low densities and ionization fractions (more precisely, when the ratio $``$ of the product of the ion and electron gyrofrequencies to the product of the electron-neutral and ion-neutral collision frequencies is $`>1`$). The linear properties of the MRI in the ambipolar diffusion limit have been studied in detail by Blaes & Balbus (1994). Their primary conclusion is that the MRI will grow provided the neutral-ion collision frequency is greater than the orbital frequency. The resistive regime occurs when collisions between the charge carrying species and neutrals damp electric currents (and therefore magnetic fields) in the plasma, a process equivalent to Ohmic resistivity. Resistive effects generally dominate in high density but weakly ionized plasmas (more precisely, when $`<1`$). The local linear stability properties of the MRI with simple resistivity have been examined for both vertical fields (Jin 1996; BH) and toroidal fields (Papaloizou & Terquem 1997). The global stability of resistive disks has been examined by Sano & Miyama (1999). The effect of resistivity on the linear instability is straightforward: if Ohmic diffusion is sufficiently rapid, it can stabilize the MRI. Recently Wardle (1999) has explored the linear properties of the MRI in the a third regime where the conductivity tensor is dominated by the Hall effect. In this limit the MRI exhibits interesting new behavior, including a loss of symmetry with respect to the sign of the background magnetic field.
Just as the linear properties of the MRI in these regimes have been investigated, the nonlinear evolution has also been simulated for a variety of limits. In the so called “strong-coupling limit” (in which the ion inertia is ignored, and the ion density is assumed to be a simple power law of the neutral density), the effect of ambipolar diffusion is to add a nonlinear diffusion term to the induction equation. As a test of a numerical algorithm to solve this term, MacLow et al (1995) reported two-dimensional simulations of the initial growth of the MRI; their results were in agreement with the stability criterion derived by Blaes & Balbus (1994). Brandenburg et al. (1995) performed three-dimensional simulations of the MRI in stratified disks; some of their models included the effects of ambipolar diffusion. They found that sufficiently large diffusivity could damp the MHD turbulence, consistent with the conclusions of Blaes & Balbus (1994).
Hawley & Stone (1998) carried out a full ion-neutral simulation of the MRI in weakly ionized plasmas, where the ions and neutrals are treated as separate fluids coupled only through a collisional drag term. Although they found close agreement with the Blaes and Balbus linear results, the structure and evolution of the saturated state of the MRI in weakly coupled fluids is more complex. Full turbulence is produced only when the collision frequency is
>
100
>
absent100\mathrel{\vbox{\offinterlineskip\hbox{$>$}
\kern 1.29167pt\hbox{$\sim$}}}100 times the orbital frequency. At lower collision frequencies, the nonlinear turbulence is increasingly inhibited by the neutrals, resulting in significantly lowered angular momentum transport rates. These results illustrate that it is possible for nonideal effects to have significant consequences for the nonlinear evolution even when their impact on the linear instability is slight.
The nonlinear evolution of a differentially rotating flow with a nonzero resistivity represents a more straightforward simulation regime. In their study of dynamo action associated with the MRI, HGB2 presented two simulations in which explicit resistive effects were included. These indicated that saturation amplitude and angular momentum transport rates could be significantly decreased by a large resistivity. More recently, Sano, Inutsuka, & Miyama (1998; hereafter SIM) explored saturation of the 2D channel solution in the presence of strong resistivity. They found reconnection of magnetic field lines across the channels could act as a saturation mechanism.
In this paper we will explore in greater depth the nonlinear behavior of the MRI in a single fluid with a finite resistivity. We present an extensive series of resistive MHD simulations in three-dimensions. We study a variety of initial field configurations and strengths over a wide range of resistivities. We find that there are substantial differences between the nonlinear evolution of the MRI in the presence of a net flux compared to the evolution at the resistivity with zero volume-averaged flux. This behavior arises because Ohmic dissipation can never destroy a net field, i.e., one which is supported by currents outside the simulation domain.
As has already been pointed out (BH), a critical dimensionless parameter which may control the behavior of the MRI in dissipative disks is the magnetic Prandtl number, i.e., the ratio of the coefficients of viscosity and resistivity. The present study does not include a physical viscous dissipation; some effective viscous dissipation is, of course, already present due to numerical effects. Thus, the exploration of the role of magnetic Prandtl number in determining the nonlinear outcome of the MRI must await future studies.
This paper is organized as follows. In §2, we discuss our numerical method (including the extension of the ZEUS algorithm to model highly resistive plasmas), and initial and boundary conditions that characterize the simulations. In §3, we discuss the results of simulations with uniform vertical fields, vertical fields with zero net flux, and uniform azimuthal fields. Our conclusions are presented in §4.
## 2 Method
### 2.1 Equations and Algorithms
Our computational model is based on the shearing box approximation developed by HGB. This approximation uses a local expansion of the equations of motion about a fiducial point of radius $`R_{}`$ in cylindrical coordinates $`(R,\varphi ,z)`$. By considering a region whose extent is much less than $`R_{}`$, one can define a local set of Cartesian coordinates $`x=RR_{},y=R_{}(\varphi \mathrm{\Omega }t),z=z`$ that corotate with the disk. Using $`|x|/R_{}1`$, we expand the equations of motion to first order in $`|x|/R_{}`$ to obtain the local equations of compressible MHD (HGB):
$$\frac{\rho }{t}+(\rho 𝐯)=0$$
(1)
$$\frac{𝐯}{t}+𝐯𝐯=\frac{1}{\rho }\left(P+\frac{B^2}{8\pi }\right)+\frac{𝐁𝐁}{4\pi \rho }2𝛀\times 𝐯+3\mathrm{\Omega }^2x\widehat{𝐱}$$
(2)
$$\frac{𝐁}{t}=\times [(𝐯\times 𝐁)\eta 𝐉],$$
(3)
$$\frac{\rho ϵ}{t}+(\rho ϵ𝐯)+P(𝐯)\eta 𝐉^\mathrm{𝟐}=\mathrm{𝟎}$$
(4)
where $`ϵ`$ is the specific internal energy, $`𝐉=\times 𝐁`$ is the current density, and the other symbols have their usual meaning. For our study, the magnetic diffusivity $`\eta `$ is spatially uniform and time independent. The $`z`$ component of gravity is ignored; consequently, there are no vertical buoyancy effects. We adopt an adiabatic equation of state
$$P=\rho ϵ(\gamma 1)$$
(5)
with $`\gamma =5/3`$. A hydrodynamic equilibrium solution to equations (1)–(4) is constant density and pressure and a uniform shear flow, $`𝐯=(3/2)\mathrm{\Omega }x\widehat{𝐲}`$.
The shearing box approximation employs strictly periodic boundary conditions in the angular ($`y`$) and vertical ($`z`$) directions, and shearing-periodic boundary conditions in the radial ($`x`$) direction. These boundary conditions and their implementation are described in more detail in HGB. Briefly, faces along the $`x`$ directions are periodic initially but subsequently shear with respect to each other. Any fluid element that travels off the outer radial boundary reappears at the lower radial boundary at the corresponding sheared position.
The above equations of MHD are solved using the ZEUS code (Stone and Norman 1992a; 1992b). ZEUS is a time explicit MHD code based on finite differences that uses the Method of Characteristics – Constrained Transport (MOCCT) algorithm (Hawley & Stone 1995) to evolve both the induction equation and the Lorentz force. The advantage of MOCCT is that it evolves the magnetic field in such a way as to maintain the constraint $`𝐁=\mathrm{𝟎}`$. Key to this property is the use of Stoke’s Law to write the induction equation in integral form: the rate of change of the magnetic flux through any face of a computational zone is then simply the line integral of the electromotive force (emf) around the edges of the face (Evans & Hawley 1988). To extend the method to include resistivity, we use an operator split solution procedure in which the MOCCT technique is used to update the first term on the RHS of (3) and then the constrained transport formalism is again used to update the magnetic flux using an effective emf defined by the resistive term (i.e. $`\eta 𝐉`$). The current $`𝐉`$ used in this step is computed from the partially updated field resulting from the MOCCT step. Resistive heating \[i.e. the last term on the LHS of eq. (4)\] is computed with this same current, appropriately averaged to the grid center. Since our update of the resistive term is time explicit, we also add a new timestep constraint so that $`t[\mathrm{min}(x,y,z)]^2/\eta `$.
We tested our implementation of the resistivity algorithm by following the diffusion of a magnetic field with an initially gaussian profile in a non-rotating box, a problem whose solution is known analytically. As described in §3, we also have reproduced the linear stability criterion for the MRI in resistive differentially rotating flows.
### 2.2 Initial Conditions
For these simulations we use a computational volume with radial dimension $`L_x`$ = 1, azimuthal dimension $`L_y=2\pi `$, and vertical dimension $`L_z`$ = 1. Most of the runs use a standard grid resolution of $`59\times 123\times 59`$. (Note that this resolution is comparable to the high-resolution runs of HGB and HGB2. The increase in what constitutes a standard grid resolution simply reflects the increase in computational power over the last few years.) Initially the computational domain is filled with a uniform plasma of density $`\rho _0=1`$ and pressure $`P_0=10^6`$. We set $`\mathrm{\Omega }=10^3`$, sound speed $`c_s=(\gamma P/\rho )^{1/2}`$ and vertical scale height $`H=c_s/\mathrm{\Omega }1.3`$. We study the evolution of a variety of initial magnetic field configurations including constant vertical fields $`B_z`$, constant toroidal fields $`B_y`$, and spatially varying $`B_z`$ fields whose volume-average sums to zero (“zero-net field”). The initial magnetic field strength $`B_0`$ is specified by $`\beta =P_0/(B_0^2/8\pi )`$.
Each of these initial field configurations is evolved using a range of resistivities $`\eta `$. The importance of a specific value of resistivity is characterized by the magnetic Reynolds number $`Re_M`$, defined as a characteristic length times velocity divided by $`\eta `$. Here we define $`Re_M`$ in terms of important disk length and velocity scales, namely, the disk sound speed and vertical scale height,
$$Re_M\frac{Hc_s}{\eta }.$$
(6)
With this definition, the magnetic Reynolds number is independent of the initial magnetic field. An alternative definition uses the wavelength of the fastest growing mode of the MRI ($`v_A/\mathrm{\Omega }`$) and $`v_A`$ as the characteristic length and velocity,
$$Re_M^{}=\frac{v_A^2}{\mathrm{\Omega }\eta }.$$
(7)
This definition was used by Sano et al. (1998). The two definitions are related by $`Re_M=Re_M^{}\beta /2`$ using the parameter $`\beta =2c_s^2/v_A^2`$.
The essential properties of resistive MRI can be understood from simple physical scalings. In the nonresistive limit the MRI’s fastest growing wavenumber has $`kv_A\mathrm{\Omega }`$. For a given resistivity $`\eta `$ the magnetic field diffusion rate will be of order $`k^2\eta `$. The MRI will be strongly affected at wavenumbers where the resistive damping rate exceeds the MRI linear growth rate. An important demarcation point is established by setting the resistive damping wavenumber $`k_D`$, defined to be that wavenumber where the diffusion rate is equal to $`\mathrm{\Omega }`$, equal to the wavenumber of the fastest growing MRI mode, $`k_{MRI}=\mathrm{\Omega }/v_A`$. This occurs when $`Re_M^{}=1`$, or
$$Re_M=\beta /2.$$
(8)
In addition to the critical Reynolds number, established by $`k_D=k_{MRI}`$, one can define other important limits. Because the resistive damping rate is proportional to the square of the wavenumber, the largest wavenumbers (shortest wavelengths) of the MRI will be affected first. Small wavenumbers have the potential to remain unstable for larger resistivities. In the small wavenumber (large wavelength) limit, the growth rate of the MRI is proportional to $`kv_A`$. If we equate this growth rate to the resistive damping rate we obtain the condition $`v_A/k\eta =1`$. Thus, for a given scale $`H`$, it is possible to damp all modes with $`k2\pi /H`$, and completely suppress the MRI, if
$$Re_M=2\pi \frac{c_s}{v_A}\beta ^{1/2}.$$
(9)
In a numerical simulation, the largest available scale is set by the dimensions of the computational domain, $`L`$, and this stability limit would also be proportional to the ratio $`H/L`$. In any case, when the diffusion wavenumber is $`k_D=2\pi /L`$ the entire computational box would be dominated by diffusion on a timescale $`\mathrm{\Omega }`$. This Reynolds number is
$$Re_M=\left(\frac{2\pi }{L}\right)^2\left(\frac{Hc_s}{\mathrm{\Omega }}\right)=(2\pi H/L)^240(H/L)^2.$$
(10)
At the other extreme we can also define a Reynolds number for which the diffusion wavenumber is $`k_D=2\pi /\mathrm{\Delta }x`$, where $`\mathrm{\Delta }x`$ is the size of a grid zone. For a computational grid size $`L`$ divided into $`N`$ grid zones the gridscale Reynolds number is
$$Re_M=(2\pi )^2(HN/L)^2.$$
(11)
For Reynolds numbers larger than this, diffusion will not be the dominant effect on dynamical timescales for all computationally resolved lengthscales. The magnetic Reynolds numbers used for our investigations below are all smaller than this limit.
Our numerical simulations also possess an intrinsic numerical resistivity due to truncation error. Because the numerical resistiviy is a nonlinear function of the grid spacing, it must be measured for each individual application by increasing the magnitude of the physical resistivity from zero and noting at what point the solution diverges from the ideal case. As described in section 3.2, we find for our standard resolution our numerical Reynolds number is about 50,000 for vertical fields with zero net flux.
## 3 Results
### 3.1 Uniform Vertical Fields
We begin with simulations of an initial weak uniform poloidal magnetic field in the shearing box. Due to periodic boundary conditions, the net flux associated with this initial mean field remains unchanged for all times. Hawley & Balbus (1992) simulated this problem in 2D in the ideal MHD limit and found that the instability produced an exponentially growing “channel solution” consisting of two oppositely directed radial streams surrounded by radial magnetic field. In 2D, saturation of the instability does not occur; instead, simulations terminate when the plasma channel is squeezed into a thin sheet too small to resolve. In 3D, however, the channels break down into MHD turbulence due to nonaxisymmetric “parasitic” instabilities (Goodman & Xu 1994; HGB; HGB2) so long as the vertical wavelength of the channel mode is less than the radial size of the computational domain.
Even in 2D, however, additional possibilities are created by the addition of a finite resistivity. SIM carried out 2D simulations of the vertical field problem with a resistive plasma for a variety of field strengths, and for resistivities $`0.3Re_M^{}3`$ \[using the definition of Reynolds number given by eq. (7)\]. They found an interesting dichotomy of behavior depending on whether or not $`Re_M^{}`$ was greater or less than one (the critical value). When $`Re_M^{}<1`$, the channel solution saturates via magnetic diffusion and reconnection across the radial streams. Assuming that the resistivity was not so large as to stabilize all possible wavelengths within the computational domain, simulations with $`Re_M^{}<1`$ amplified the magnetic field until $`Re_M^{}1`$ at which point saturation occurred. For stronger initial fields, however, or with weak resistivity such that $`Re_M^{}>1`$ SIM found that the 2D channel solution remained as before, growing without apparent limit. Since saturation by parasitic modes is inherently a nonaxisymmetric process, 3D resistive simulations are required to explore this regime.
In this section we consider a constant vertical magnetic field in an initially uniform 3D shearing box. The linear dispersion relation for a vertical field with resistivity is given by Jin (1996). Ignoring buoyancy terms and assuming a Keplerian background and a displacement $`𝐤=k\widehat{z}`$, a simplified linear dispersion relation can be written
$$\sigma ^4+2\xi \sigma ^3+(2q^2+\xi ^2+1)\sigma ^2+2\xi (q^2+1)\sigma 3q^2+q^4+\xi ^2=0$$
(12)
where $`\sigma `$ is a growth rate in units of orbital frequency, $`q=k_zv_A/\mathrm{\Omega }`$, and $`\xi =k_z^2\eta /\mathrm{\Omega }`$. Growth rates as a function of $`q`$ and $`\xi `$ are plotted in Figure 1.
In a fully conducting plasma, the fastest growing mode of the MRI has $`q=\sqrt{15}/4`$; the corresponding wavelength $`\lambda _c`$ is given by
$$\lambda _c=2\pi \sqrt{16/15}|𝐯_𝐀|/\mathrm{\Omega }.$$
(13)
For these uniform vertical field simulations we select $`\beta =400`$, $`\lambda _c`$ = 0.459. This wavelength extends over 27 grid cells so fast-growing modes are well resolved. The largest vertical wavelength allowed in the box is $`\lambda =L`$ which corresponds to $`q=0.444`$ and has a growth rate of $`\sigma =0.57`$. Wavelengths $`L/2`$ ($`q=0.888`$) and $`L/3`$ ($`q=1.332`$) are also unstable with growth rates $`\sigma =0.75`$ and 0.66 respectively. The zero resistivity run serves as a control model; it uses the same parameters (although a slightly different numerical resolution) as the run labeled Z4 in HGB. We have run four resistive simulations with this initial magnetic field using $`Re_M=1300`$, 520, 260, and 130. For increasing values of magnetic resistivity the most unstable wavelength shifts to larger scales and the growth rate is substantially lower than the ideal MHD rate. The linear growth rates for the three largest wavelengths with $`Re_M=1300`$ are $`\sigma =0.53`$, 0.62, and 0.41. From the dimensional analysis of the critical magnetic Reynolds number, for our given field strength and box size, we expect that the growth rate of the MRI should be significantly affected for $`Re_M\beta =400`$. Indeed, at that level $`\lambda =L/3`$ is no longer unstable, and the growth rates for $`\lambda =L`$ and $`L/2`$ are reduced to 0.45 and 0.36. At $`Re_M=260`$ only the $`\lambda =L`$ wavelength remains unstable with a growth rate of 0.37. This wavelength too becomes marginally stable for $`Re_M=130`$.
Figure 2 is a plot of the time evolution of the volume averaged total magnetic energy in each run. In keeping with the linear analysis, we find no growth when $`Re_M=130`$; all modes have been stabilized by resistivity. Unstable modes are present for all other values of $`Re_M`$, with the most rapid growth for the non-resistive ($`\eta =0`$) model. For each of the models in which unstable modes are present, the evolution of the magnetic energy is similar. An initial phase of exponential growth ends in a sharp peak, after which the magnetic energy declines rapidly before fluctuating about a lower average level. For $`\eta =0`$, this initial peak occurs at 3.5 orbits. Consistent with the decrease in linear growth rates, the peak occurs at increasingly later times for decreasing $`Re_M`$; for $`Re_M=260`$ the magnetic energy peaks at 8 orbits.
In the absence of resistivity, the initial linear growth is associated with development of the channel solution at the wavelength of the fastest growing mode; for $`\eta =0`$ this is $`L_z/2`$. This is also the fastest growing wavelength in the $`Re_M=1300`$ model, which behaves similarly. The growth rate for the $`Re_M=1300`$ run is $`80\%`$ of the maximum growth rate for ideal MHD (0.75$`\mathrm{\Omega }`$), close to what is predicted by the linear analysis. The $`\lambda =L/2`$ channel solution is well established by orbit 2.2 for $`\eta =0`$ and 3.2 for $`Re_M=1300`$. A spectral analysis at this stage indicates that the magnetic energy of the fastest growing wavenumber mode is 3 orders of magnitude greater than all other modes combined.
In the runs with $`Re_M=520`$ and 260, the linear growth rates have been reduced by magnetic diffusion enough to change the nature of the observed channel solution. Ohmic dissipation stabilizes small scale perturbations, and we no longer observe the $`\lambda =L_z/2`$ mode; the dominant mode is now $`\lambda =L`$, i.e., a two stream (one in, one out) mode. With $`Re_M=260`$ the growth rate of this mode has decreased to $`48\%`$ of the ideal MHD value.
As resistivity is increased, the peak in the magnetic energy at the end of the linear growth phase achieves higher amplitudes. This suggests that the fastest growing modes of the parasitic instabilities (the Kelvin-Helmholtz modes) are also affected by resistivity (Goodman & Xu 1994). One reason for this is that the parasitic modes require vertical wavenumbers that are larger than radial wavenumbers. For $`Re_M`$ less than 260, resistivity restricts growth to the smallest vertical wavenumber, inhibiting the onset of the parasitic instability. This allows the channel solution to persist and to reach higher amplitudes before disruption.
Following the peak and subsequent decline, the magnetic energy in all models fluctuates around a mean value. Increasing the resistivity (1) reduces this mean saturated field energy, (2) increases the magnitude of the fluctuations relative to the saturation amplitude, and (3) increases the timescale of the fluctuations. Table 1 lists volume-averaged values for a variety of quantities in the saturated state, time-averaged over the last twenty orbits of each run. There is a systematic decline in both kinetic and magnetic energies with increasing resistivity. For the $`Re_M=260`$ model the mean kinetic energy in the turbulent fluctuations is $`26\%`$ of the mean value in the ideal MHD simulation. In the turbulent flow the magnetic energy resides primarily in the azimuthal component of the field. The background shear flow always favors the growth of the azimuthal field component. The ratio of $`B_y^2`$ to $`B_x^2+B_z^2`$ is 2.6 in the zero resistivity run, increases to 3.1 for the $`Re_M=520`$ run before dropping back to 2.6 with $`Re_M=260`$. The magnetic field energy is greater than the perturbed kinetic energy by a factor of 2.4 with no resistivity; this ratio decreases with increasing resistivity, down to 1.24 in the $`Re_M=260`$ run.
Turbulent shear stress $`W_{xy}=\rho v_x\delta v_yB_xB_y/4\pi `$ transports angular momentum outward, but the total stress decreases with Reynolds number. Using Shakura-Sunyaev scaling, $`W_{xy}=\alpha P_o`$, $`\alpha `$ decreases from 0.307 at $`\eta =0`$ to 0.053 at $`Re_M=260`$. Angular momentum transport is correspondingly reduced. In all cases the Maxwell stress dominates over the Reynolds stress, although the ratio of Maxwell to Reynolds stress decreases from 4.8 for $`\eta =0`$ to 3.38 for $`Re_M=260`$.
Although the time histories of these runs (Figure 2) are superficially similar, the late-time state of the $`Re_M=520`$ and 260 simulations is strikingly different than that of the higher $`Re_M`$ runs. In the ideal MHD case, and in the $`Re_M=1300`$ run, the flows are turbulent. At lower $`Re_M`$ values, however, the fluctuations seen in the saturated state are associated with the periodic re-emergence of the channel solution. Figure 3 shows images of the angular momentum excess $`\delta L\rho (v_y+1.5\mathrm{\Omega }x)`$ overlaid by magnetic field lines in the $`xz`$ plane at $`y=0`$ at orbits 23 and 25 in the ideal ($`\eta =0`$) and $`Re_M=260`$ models. Note from Figure 2 that a strong peak in the magnetic energy begins to develop in the $`Re_M=260`$ model at these times. In the ideal model, the angular momentum excess shows large amplitude, disordered fluctuations; the magnetic field is highly tangled and time-variable, characteristic of MHD turbulence. In contrast, the $`Re_M=260`$ model shows a disordered field and pattern of angular momentum excess at orbit 23, but by orbit 25 (during the growth of the next peak in the magnetic energy) this has changed to a layered profile, combined with a large amplitude sinusoidal variation in the field, both characteristic of the channel solution. A spectral analysis of the magnetic field in the $`Re_M=260`$ model at orbit 25 shows not the power law usually associated with turbulence, but rather the domination of the $`\lambda =L_z`$ mode by two orders of magnitude.
The decrease in turbulent flow and the strength of the resurgent channel solution may be examined by observing the increase in the kinetic energy associated with radial motions. In the $`Re_M=260`$ run, we find that a large increase in the radial kinetic energy accompanies every significant increase in magnetic field energy. For example, at orbit 23 $`<\rho v_x^2>0.01P_o`$ however, by orbit 25 the radial kinetic energy has increased to $`0.1P_o`$. The radial kinetic energy rises by at least one order of magnitude every time magnetic field energy peaks.
For $`Re_M=260`$, the channel solution first re-emerges (after the initial peak) at orbit 17, and thereafter occurs roughly every 4 orbits. These recurrent channels saturate not through parasitic instabilities, but by reconnection across the channels. In their 2D simulations, SIM found that reconnection could be an alternative saturation mechanism for the channel solution; we find that this behavior in 3D as well. The resistive diffusion time for variations of lengthscale $`L_z/2`$ (the vertical extent of one channel) is $`L_z^2/4\eta `$, or about 6 orbits. The frequency of variability is comparable to this resistive diffusion time.
The quasi-periodic emergence of the channel solution could have important consequences for time variability of the accretion rate in disks. Figure 4 plots the shear stress normalized by the initial pressure ($`W_{xy}/P_0=\alpha `$) for the ideal and $`Re_M`$=260 case. Fluctuations in the stress exceed an order of magnitude, although the mean is less than the ideal MHD case. Note that the stress associated with the initial peak results in $`\alpha >1`$. We may conclude that for high values of magnetic diffusion in a mean field, the rate of transport is cyclic and highly variable.
Considerable heating due to Ohmic dissipation occurs, particularly in the high resistivity runs. For example, at $`Re_M=260`$ the internal energy rises by a factor of $`100`$ during the initial decline from the magnetic energy peak at orbit 8. This is a much greater amount of heating than that seen in the other runs. It would appear that most of the magnetic energy extracted from the differential rotation by the growth of the MRI now heats the disk through Ohmic dissipation. This heating is an indication that resistivity is playing a major role in terminating the channel flow. During the later phase of evolution, the internal energy rises in a somewhat stair-step fashion during each re-emergence of the channel solution, although never again at so large a rate. For example, after saturation the internal energy in the $`Re_M=260`$ simulation rises only another 5% by orbit 40. Again this indicates saturation through resistivity. Ohmic heating also drives changes in the plasma $`\beta `$ parameter. Since we have a simple adiabatic equation of state, with no radiative losses, heat simply accumulates. Initially the MRI drives $`\beta `$ toward unity. However, after saturation $`\beta `$ differs considerably for runs with different resistivity, for example $`\eta =0`$ $`\beta 17`$ while for $`Re_m=260`$ $`\beta 3000`$. In these simulations this heating keeps the magnetic pressure from becoming dynamically important, but has limited importance since there is no vertical gravity. In a disk, radiative cooling would limit the growth of $`\beta `$.
### 3.2 Vertical Fields with Zero Net Flux
In the previous section we considered a domain filled with a uniform vertical field. Since the currents generating such a field are outside the computational domain, the background flux cannot change, regardless of the strength of the resistivity. The next set of simulations begins with an initial magnetic field configuration $`B_z=B_o\mathrm{sin}2\pi x`$. With periodic (and shearing-periodic) boundary conditions, the mean field $`B`$ remains zero for all time. Therefore, unlike the uniform vertical field simulations above, physical resistivity can now completely dissipate the field. This particular initial zero-net field configuration was previously studied in the ideal MHD limit by HGB2.
We describe the results of six runs corresponding to six values of magnetic Reynolds number: $`Re_M=\mathrm{}(\eta =0)`$, 65K, 26K, 19.5K, 13K, and 1.3K. These simulations are all performed with a maximum field strength of $`\beta =400`$ and at our standard resolution. Figure 5 is a plot of the time-evolution of the total magnetic energy for these values of the magnetic Reynolds number. Except for the case of $`Re_M=1.3`$K, an initial period of exponential growth is observed, followed by saturation in a sharp peak near 3.2 orbits. As in the uniform vertical field runs, this rise to a strong peak is associated with the $`k_x=k_y=0`$ and $`k_z=4\pi `$ mode, the fastest growing mode from the linear analysis. Note that in all except the ideal MHD case, some diffusion of the initial field can be observed during the first two orbits; for $`Re_M`$=13K and $`Re_M`$=1.3K the field energy is $`89\%`$ and $`24\%`$ of its initial value by orbit 2.2. The diffusion is so rapid for the $`Re_M=1.3`$K model that significant growth in the field energy is never attained. The azimuthal and radial components of the magnetic energy peak at $`10^5P_o`$, only $``$ 4 orders of magnitude greater than their values at orbit 0.5. After orbit 4 the energy in both of these components starts to decline. The vertical component does not exhibit any growth at all. Unlike the radial and azimuthal field components any amplification is insignificant compared to the resistive loss. The growth of the linear mode is, in turn, severely restricted when the background field is rapidly diffusing.
As $`Re_M`$ is decreased, there is a small decline in the peak values of azimuthal and radial field energy; it is the vertical field energy that is most affected. At $`Re_M`$=13K the peak value of vertical field energy is only 1.24 times greater then its initial value. Resistivity affects the evolution not so much by altering the growth rates of the most unstable modes, but by diffusing away the background field upon which those modes are growing. This was not a factor in the uniform initial field models, and the decline in the amplitude of the nonlinear flow at saturation as a function of $`Re_M`$ is a consequence.
After the initial peak, the magnetic energies decline to an approximate mean value about which there are large amplitude, short-timescale fluctuations. Because the spatial variation of the field in the initial state causes the wavelength of the fastest growing mode to vary, parasitic instabilities are not required to cause the channel solution to transition to turbulence. Instead, the structure which results from the growth of the linear modes is already complex enough that turbulence is the inevitable outcome. After the initial peak, the MRI saturates as MHD turbulence for all values of $`Re_M`$ except $`Re_M=1300`$.
Table 2 lists the average value in the turbulent state of selected quantities for several runs. Each quantity has been averaged over the time period 30 – 50 orbits. For $`\eta =0`$, significant growth of the MRI occurs after 2 orbits, with the saturated magnetic energy level $`0.01P_o`$. At 2.2 orbits, $`95\%`$ of the field energy is located in the toroidal field. The magnetic energy is dominated by the azimuthal field with $`85\%`$ of this energy in the azimuthal component. The $`Re_M`$=130K and $`Re_M`$=65K simulations show little apparent difference in energy levels. Angular momentum transport was similar in these three simulations as well, with $`\alpha 0.008`$. Since there is little difference between resistive runs with ReM
>
65
>
𝑅subscript𝑒𝑀65Re_{M}\mathrel{\vbox{\offinterlineskip\hbox{$>$}
\kern 1.29167pt\hbox{$\sim$}}}65K and ideal MHD runs, the effective numerical resistivity would appear to be of the same order, $``$ $`Re_M=65K`$.
Below $`Re_M=65`$K, there is a systematic decline in the saturation amplitude with decreasing $`Re_M`$. The magnetic energy is $`3\times 10^4P_0`$ for $`Re_M=13`$K, compared with $`10^2P_o`$ for the ideal MHD run. Angular momentum transport also declines with decreasing Reynolds number. In the $`Re_M=26`$K run the field saturates at levels above the initial energy, with a small but non-negligible transport rate of α
>
2×103
>
𝛼2superscript103\alpha\mathrel{\vbox{\offinterlineskip\hbox{$>$}
\kern 1.29167pt\hbox{$\sim$}}}2\times 10^{-3}. The $`Re_M=13K`$ run is strongly affected. At orbit 3 it has $`\alpha `$=0.007 and a Maxwell to Reynolds stress ratio = 2.4. The poloidal field is preferentially destroyed as time proceeds; at orbit 3 $`53\%`$ of the field energy is in the $`B_y`$ component but this increases to $`92\%`$ by orbit 10. At orbit 30, $`\alpha `$ = 0.0001 and $`99\%`$ of the remaining energy is in the $`B_y`$ component. Transport has effectively been shut down in the disk, and Ohmic dissipation allows the field energy to continue decreasing. Although the magnetic energy stops declining beyond orbit 30, the mean magnetic energy and effective $`\alpha `$ are so small as to imply the MRI is effectively quenched. At this point the field has large scale organized structure. It is layered: in the upper half of the box field is directed toward the positive azimuthal direction, and in the lower half the field is directed toward the negative azimuthal direction.
We expect the dissipational lengthscale to grow as $`Re_M`$ is lowered. This expectation can be examined quantitatively. Figure 6 is a comparison of the power spectrum of fluctuations in the magnetic energy as a function of wavenumber in the y-dimension $`k_y`$ in the $`Re_M=13K`$ and the $`\eta =0`$ runs. The spectra are time averaged over orbits 15 to 23 for $`Re_M=13K`$ and 16 to 27 for $`\eta =0`$. The $`Re_M=13K`$ spectrum has been normalized to give it the same amplitude as the $`\eta =0`$ case. Both spectra are fit by decreasing power laws, with fluctuations on large scales (small k) fit by a Kolmogorov-like slope (-11/3). On small scales (large k), where dissipation becomes important, the slope of the spectra becomes much steeper. For the $`Re_M=13K`$ run, this change in slope occurs at about $`k_y8(2\pi )/L_y`$, while for the ideal run the change does not occur until $`k_y11(2\pi )/L_y`$. Thus, the large explicit resistivity clearly smooths the turbulence on small scales as expected.
In the ideal MHD simulations of HGB2 using an initial vertical field that varies as $`\mathrm{sin}(x)`$, the energies in the late-time turbulent state were more or less independent of the initial field strength. Here we consider a zero net vertical flux model with $`\beta =1600`$ initially and $`Re_M=19.5K`$, computed at the standard resolution. The evolution of the magnetic energy is qualitatively similar to the $`\beta =400`$ model except that the energy levels of the weaker field simulation always remain a factor four times lower throughout the evolution. The field energy is concentrated in the azimuthal component for both runs. Both runs show similar time behavior with respect to angular momentum transport. An average from orbit 4 through orbit 10 yields $`\alpha =0.012`$ for the $`\beta =400`$ model; this is five times greater then average transport for the $`\beta =1600`$ run. The power spectra for these runs are also similar.
One difference between the $`\beta =400`$ and the weak field $`\beta =1600`$ simulation shows up clearly in the linear stage. The weak field run has a critical wavelength $`\lambda _c=L_z/4`$. This is consistent with linear analysis, which, for $`\beta `$=1600 predicts a critical wavelength $`\lambda _c=0.23`$. Decreasing the field strength while keeping $`Re_M`$ and the box size fixed simply shifts the wavelength of the most unstable mode down closer to the resistive dissipation scale. Since the growth rate of the MRI and the resistive dissipation rate are unchanged, the subsequent time evolution of volume averaged variables is similar to the $`\beta =400`$ run, although the energies remain lower and the detailed structures of the nonlinear flow are different.
### 3.3 Toroidal Fields
In this section we consider the effect of resistivity on the development of the MRI in the presence of a toroidal field. The linear properties of the MRI in the local ideal MHD limit were considered by Balbus & Hawley (1992). The toroidal field instability is nonaxisymmetric, and, for a nonaxisymmetric mode, the radial wavenumber $`k_R`$ evolves with time due to the background shear. For a pure toroidal field, amplification occurs only during that time when $`k/k_z`$ is small. Although peak amplification still occurs for wavenumbers $`kv_A\mathrm{\Omega }`$ ($`k`$ here corresponds to the azimuthal wavenumber for a toroidal field), the toroidal field instability favors large values of $`k_z`$ ($`\mathrm{}`$) and these are the wavenumbers most likely to be affected by resistivity. The vertical field instability, on the other hand, favors finite $`k_z`$ and $`k_R=k_\varphi =0`$ (axisymmetric channel solutions).
Papaloizou and Terquem (1997) examined in some detail the linear stability of a toroidal field configuration with finite resistivity. Consistent with the expectations from the nonresistive linear analysis, they found that mode growth ceased for magnetic Reynolds numbers ($`1000`$) that are larger than those that stabilize the vertical field instability. Interestingly, their condition for transient amplification, their equation (32), corresponds to the zero-frequency limit of the linear dispersion relation for the poloidal field instability. Again many of the qualitative linear properties of the MRI are essentially independent of field strength or orientation.
We have run several simulations of shearing boxes with initial constant toroidal fields. Such simulations were carried out in the nonresistive limit by HGB who examined a variety of initial field strengths. Here we choose an initial field strength of $`\beta =100`$. Models were run at both $`64\times 128\times 64`$ and $`32\times 64\times 32`$ grid resolution using initial random perturbations at about 1% of the sound speed. We will concentrate on the high resolution simulations which were computed for the ideal MHD limit and for $`Re_M=10K`$, 5K and 2K.
The evolution of the toroidal ($`B_y`$) and radial ($`B_x`$) magnetic energies is shown in Figure 7. As the Reynolds number is reduced from the ideal MHD limit, the initial growth rates are reduced and final saturation is delayed. The $`Re_M=2K`$ model shows no growth. Post-saturation time-averaged values for the simulations are given in Table 3. All values show a decline with decreasing Reynolds number, but there is a sharp transition from perturbation growth to perturbation decay in going from $`Re_M=5K`$ to $`2K`$. This transition is consistent with the results for certain specific linear modes carried out by Papaloizou & Terquem (1997).
### 3.4 Resolution
As a step toward investigating the effect of finite resolution on our simulation results, we performed two resolution experiments. We reran the case of vertical field with zero net flux at four different resolutions, once with zero resistivity, and once for $`Re_M=10`$K. The resolutions ran from $`16\times 32\times 16`$ up to $`128\times 256\times 128`$ by powers of two. The initial magnetic field is the same as in §3.2 above, a vertical field that varies as $`\mathrm{sin}(x)`$. A specific set of long-wavelength initial velocity perturbations is applied to each model so that all resolutions begin with the same initial conditions.
The resolution series with zero resistivity behaves in an expected way. The higher the resolution, the larger the rate of growth of the perturbed magnetic field energy, the higher the initial peak, and the earlier that peak occurs. Beyond the initial peak the magnetic energy declines somewhat. The lowest resolution model exhibits a noticeably larger rate of decline, while the other three resolutions look comparable.
Figure 8 plots the time evolution of the magnetic energy in the series of runs with $`Re_M=10`$K. The behavior of the different resolution models is the same as in the zero resistivity run for the first few orbits. After this the magnetic energy declines for all resolutions; the highest resolution run, however, shows the steepest rate of decline in magnetic energy, followed by the $`64\times 128\times 64`$ resolution run. The lowest resolution runs decline less steeply and behave similarly to each other. For this particular Reynolds number, the critical diffusion wavenumber, defined $`\eta k^2=\mathrm{\Omega }`$, corresponds to a wavelength of 0.0714 (where the vertical box size $`L_z=1`$). This wavelength is equal to $`1.14\mathrm{\Delta }z`$ in the lowest resolution simulation (and hence is unresolved), and equal to $`18.3\mathrm{\Delta }z`$ in the highest resolution simulation. Thus we have the interesting observation that by resolving the diffusion lengthscales, turbulence decays more rapidly in the highest resolution grid. Apparently “numerical resistivity” is much less effective at field dissipation compared to a physical resistivity with a diffusive wavelength comparable to the grid zone size.
## 4 Summary
Using numerical MHD simulations, we have studied the shearing box evolution of the MRI in the presence of finite resistivity. We have examined initial field configurations consisting of a uniform vertical field, a uniform toroidal field, and a vertical field that varies sinusoidal in the radial direction. The linear growth rates of the most unstable modes observed in our simulations are in good agreement with a linear analysis. As the resistivity increases, the growth rate for all modes declines. For a fixed box size, all modes are stable at a large enough value of the resistivity (when η
>
vA2/Ω
>
𝜂superscriptsubscript𝑣𝐴2Ω\eta\mathrel{\vbox{\offinterlineskip\hbox{$>$}
\kern 1.29167pt\hbox{$\sim$}}}v_{A}^{2}/\Omega); our numerical results correctly recover this limit. The restrictions imposed upon the toroidal field instability are more severe. We find toroidal field models become linearly stable at larger Reynolds numbers than vertical field models, again, in agreement with the linear analysis (Papaloizou & Terquem 1997). Because the toroidal field instability favors large vertical wavenumbers, the modes that are the most unstable and have the longest period of amplification are precisely those most affected by finite resistivity.
Although the simulations agree with the linear analysis during the linear growth phase, we find that the nonlinear evolution is more complicated, and the linear analyses provide only limited guidance. In particular, finite resistivity can have profound effects on the flow even when the linear modes are still unstable.
The nonlinear outcome of the MRI is profoundly influenced by the presence or absence of a net field. When the shearing box is penetrated by a net vertical field, Ohmic dissipation can never completely destroy the mean field. Thus, provided the resistivity is low enough that at least a few unstable modes are present, the MRI will always exist in such a box. When the resistivity is large, but not so large as to completely stabilize the MRI, the instability leads to a strongly fluctuating magnetic energy and transport rate. These fluctuations are associated with periodic recurrence of the axisymmetric channel solution. High resistivity can then lead to reconnection across channels and rapid decline in magnetic energy, after which the instability grows again. The period of the resurgent channel solution is found to be roughly equal to the resistive diffusion time, of order a few orbits.
It should be stressed that saturation through reconnection is observed only when the initial field is supported by some external currents outside our simulation domain. Also, the effects of stratification are not present in our simulations. These considerations may limit the degree to which our results may be generalized to realistic disk models. However, locally in highly resistive disks threaded by a mean magnetic field transport may be cyclic. This result may be of some importance for our understanding time variable accretion systems.
We find that models with a non-vanishing net magnetic flux display qualitatively different behavior in the nonlinear regime than models with zero net flux. If the initial field configuration contains zero net magnetic flux then the resistive MRI saturates as MHD turbulence, as it does in the ideal MHD case. Finite resistivity, however, significantly modifies the evolution of the turbulence. The amplitude of the magnetic energies and corresponding angular momentum transport rates in these simulations decline with decreasing $`Re_M`$. In fact, for ReM
<
104
<
𝑅subscript𝑒𝑀superscript104Re_{M}\mathrel{\vbox{\offinterlineskip\hbox{$<$}
\kern 1.29167pt\hbox{$\sim$}}}10^{4} the turbulence is completely quenched. This limit is roughly 100 times larger than the Reynolds number required for complete stabilization within the linear theory. Examination of the power spectrum of the turbulence clearly shows a rapid drop off in power at at high wavenumbers (small scales) when resistivity is present compared to the ideal MHD limit.
The finding that finite resistivity can affect the levels of turbulence even when the linear analysis predicts the presence of instability, has potential implications for accretion disk evolution. In particular, Gammie & Menou (1998) point out that finite resistivity leads to magnetic Reynolds numbers $`Re_M10^4`$ in the cool, low states of dwarf novae. Our simulations show that this is indeed an interesting level of resistivity. Many dwarf nova models depend upon different levels of angular momentum transport in the high and low state. Finite resistivity appears to be a viable mechanism by which these different levels could be produced.
This research is supported in part by NASA grant NAG-54278 and NSF grant AST-9528299 (JMS), and NASA grants NAG5-7500 and NAG5-3058, and NSF grant AST-9423187 (JFH). We thank Steve Balbus and Charles Gammie for insightful comments. Supercomputer simulations are supported under an NSF National Resource Allocation grant, and have been carried out on the Origin 2000 system at NCSA, and the T90 and T3E systems at NPACI.
TABLE 1
TIME- AND VOLUME- AVERAGE VALUES FOR UNIFORM VERTICAL FIELD RUNS
| Quantity | $`\eta =0`$ | $`Re_M=1300`$ | $`Re_M=520`$ | $`Re_M=260`$ |
| --- | --- | --- | --- | --- |
| $`B^2/8\pi P_o`$ | 0.450 | 0.277 | 0.140 | 0.062 |
| $`B_x^2/8\pi P_o`$ | 0.093 | 0.055 | 0.025 | 0.012 |
| $`B_y^2/8\pi P_o`$ | 0.324 | 0.204 | 0.105 | 0.045 |
| $`B_z^2/8\pi P_o`$ | 0.033 | 0.018 | 0.009 | 0.005 |
| $`B_xB_y/4\pi P_o`$ | 0.254 | 0.170 | 0.087 | 0.042 |
| $`\rho v_x\delta v_y/P_o`$ | 0.052 | 0.040 | 0.022 | 0.012 |
| $`\rho \delta v^2/2P_o`$ | 0.192 | 0.135 | 0.081 | 0.050 |
| $`\rho v_x^2/2P_o`$ | 0.071 | 0.059 | 0.040 | 0.028 |
| $`\rho \delta v_y^2/2P_o`$ | 0.092 | 0.052 | 0.024 | 0.011 |
| $`\rho v_z^2/2P_o`$ | 0.029 | 0.024 | 0.017 | 0.011 |
| $`\alpha `$ | 0.307 | 0.210 | 0.110 | 0.053 |
| Max/Reyn | 4.8 | 4.23 | 3.87 | 3.38 |
TABLE 2
TIME- AND VOLUME- AVERAGE VALUES FOR ZERO MEAN VERTICAL FIELD RUNS
| Quantity | $`\eta =0`$ | $`Re_M=26K`$ | $`Re_M=19.5K`$ | $`Re_M=13K`$ |
| --- | --- | --- | --- | --- |
| $`B^2/8\pi P_o`$ | 0.010 | 0.003 | 0.003 | 0.00027 |
| $`B_x^2/8\pi P_o`$ | 0.001 | 0.00016 | 0.00014 | $`6.10\times 10^8`$ |
| $`B_y^2/8\pi P_o`$ | 0.0089 | 0.003 | 0.0026 | 0.00027 |
| $`B_z^2/8\pi P_o`$ | 0.0003 | $`5.38\times 10^5`$ | $`4.0\times 10^5`$ | $`8.28\times 10^9`$ |
| $`B_xB_y/4\pi P_o`$ | 0.0045 | 0.001 | 0.0009 | $`3.75\times 10^6`$ |
| $`\rho v_x\delta v_y/P_o`$ | 0.0023 | 0.001 | 0.00069 | $`2.28\times 10^5`$ |
| $`\rho \delta v^2/2P_o`$ | 0.008 | 0.004 | 0.0025 | 0.0001 |
| $`\rho v_x^2/2P_o`$ | 0.0051 | 0.003 | 0.002 | 0.0001 |
| $`\rho \delta v_y^2/2P_o`$ | 0.002 | 0.00075 | 0.00042 | $`2.57\times 10^5`$ |
| $`\rho v_z^2/2P_o`$ | 0.001 | 0.0003 | 0.0002 | $`8.94\times 10^7`$ |
| $`\alpha `$ | 0.0068 | 0.0011 | 0.0016 | $`2.65\times 10^5`$ |
| Max/Reyn | 1.92 | 0.870 | 1.29 | 0.165 |
TABLE 3
TIME- AND VOLUME- AVERAGE VALUES FOR TOROIDAL FIELD RUNS
| Quantity | $`\eta =0`$ | $`Re_M=10K`$ | $`Re_M=5K`$ | $`Re_M=2K`$ |
| --- | --- | --- | --- | --- |
| $`B^2/8\pi P_o`$ | 0.0705 | 0.061 | 0.049 | |
| $`B_x^2/8\pi P_o`$ | 0.0099 | 0.0081 | 0.0059 | |
| $`B_y^2/8\pi P_o`$ | 0.0568 | 0.050 | 0.0403 | 0.01 |
| $`B_z^2/8\pi P_o`$ | 0.0038 | $`0.0031`$ | $`0.0024`$ | |
| $`B_xB_y/4\pi P_o`$ | 0.030 | 0.026 | 0.0102 | |
| $`\rho v_x\delta v_y/P_o`$ | 0.0077 | 0.0036 | 0.00069 | |
| $`\rho \delta v^2/2P_o`$ | 0.0304 | 0.026 | 0.021 | |
| $`\rho v_x^2/2P_o`$ | 0.0138 | 0.012 | 0.0097 | |
| $`\rho \delta v_y^2/2P_o`$ | 0.0113 | 0.0090 | 0.0071 | |
| $`\rho v_z^2/2P_o`$ | 0.0053 | 0.0048 | 0.0045 | |
| $`\alpha `$ | 0.039 | 0.034 | 0.014 | |
| Max/Reyn | 3.33 | 3.38 | 2.83 | | |
warning/0001/astro-ph0001301.html | ar5iv | text | # HEGRA Observations of Galactic Sources
## Summary of observations
The stereoscopic system of imaging atmospheric Cherenkov telescopes (IACTs) of HEGRA has been running since late 1996 with four telescopes. After a fire in the array which also damaged one of these telescopes late in the year 1997, the final configuration of five equal telescopes with identical cameras has become operational in August 1998. Apart from the IACT system, HEGRA successfully operates a stand-alone telescope, called CT1; it is also doing obervations during moon periods. However in this review, I will be concerned with the stereoscopic system alone.
Since 1997 a number of Galactic source candidates has been observed with the IACT system. The Galactic coordinates of the objects discussed in this paper are indicated in Figure 1. The objects analyzed are given in Table 1.
1. The observations of the Crab Nebula were done both at normal (ZA $`30`$) and at high zenith angles (ZA $`60`$). They led to a (combined) energy spectrum up to 20 TeV (Konopelko et al. 1999). It is within the errors compatible with an extension of the power law spectrum inferred from measurements in the TeV energy range (Konopelko et al., these Proceedings). Thus no possible hard hadronic emission component has been identified up to these energies.
2. A search for a periodic signal from the Crab and Geminga Pulsars was also performed (Aharonian et al. 1999a). No evidence for pulsed emission was found. Even though we had expected Geminga to be a major contributor to the distribution of very energetic CR electrons in the neighborhood the Solar System, it showed up as a TeV-quiet object.
3. The observations of the Galactic Microquasar GRS 1915 have been analysed (Kettler 1999). No signals have been found during these observation periods. This is not too surprising since, unfortunately, the source had also been low in other wavelength ranges in those times.
4. The two SNRs Tycho and Cas A have been observed extensively with the stereoscopic system. This is especially true for Cas A, where a deep observation of 128 hr duration has been included in the present analysis (Pühlhofer et al. 1999a). The two objects will be discussed in some detail in section 2.
5. An extensive Galactic Plane scan ($`2`$ hrs of observation time for each point, plus some re-observations) in the TeV band covered the Galactic longitude region from the Galactic Center ($`l,1.5^{}`$) to the Cygnus region ($`l,83.5^{}`$), see Figure 2 .
Sources with a flux above 1/4 Crab units should have been detected, as indicated by Table 2 below. A first analysis reveals no hints for such strong TeV point sources (Pühlhofer et al. 1999b).
6. In a similar vein, a program regarding the search for diffuse VHE $`\gamma `$-ray emission from the Galactic Plane was started. The present analysis is largely of a technical nature (Lampeitl et al. 1999). The observations will be continued during this summer of 1999.
7. Finally, the imaging Cherenkov technique was applied for the first time to the determination of the flux and the TeV energy spectrum of the charged CR protons. For this work background events for the Mkn 501 observations from 1997 were used. Calibration is exclusively by Monte Carlo simulations that include a detailed detector simulation. For physical reasons, the proton detection rate strongly exceeds that for heavier CR nuclei near threshold, around 1.5 TeV. The stereoscopic detection of the air showers permits the effective suppression of air showers induced by heavier particles already at the trigger level, and in addition by software analysis cuts. The results are in good agreement with the recent results of satellite and balloone-borne experiments and reach similar accuracy (Aharonian et al. 1999b). Without any knowledge of the CR composition, the proton spectrum can only be determined with precision near threshold. However, it should be possible to obtain in addition an approximate CR composition, using further specific image cuts (Plyasheshnikov et al. 1998). This will allow an extension of the dynamical range of the spectrum into an energy region that is very costly to cover by direct detection CR experiments. I think it would be important if in addition the large Zenith angle technique could be applied to this problem.
Of course, more than these objects have been observed in the Galaxy. However the data have not been analyzed yet, and are therefore not a subject of this summary.
Let me conclude this section with a general consideration.
As mentioned above, the HEGRA Galactic Plane scan has not yet led to the detection of new sources. In his excellent introductory review, Trevor Weekes (these Proceedings) described this result as ”depressing”.
We were also disappointed. On the other hand, the result is perhaps not too surprising, given the low sensitivity level with which this survey had to be done. The result should also prompt a new discussion about the aims and possibilities of ground-based $`\gamma `$-ray astronomy with imaging telescopes. Space is scarce in these proceedings. So, I will summarize my arguments only briefly in four points, and hope that they open a broader debate: (i) we should of course continue such surveys; any field of astronomy must do this (ii) however it is not too probable that we will find new sources that have not been seen as unusual objects in another wavelength range already, considering the enormous investments in ever more powerful instruments in the radio, infrared, optical, and X-ray domains that have been made over the last two decades (iii) thus, our main activity should perhaps be to look at sources also known in other wavelength-ranges; only then we can hope to obtain a physical understanding of the $`\gamma `$-ray results (iv) given the much higher physical complexity of the acceleration and transport processes for the nonthermal component than for the thermal component, the potential for discovery is one for strong nonthermal activity in known objects, and it is as important as the potential for discovery of previously unknown objects in the more conventional ”thermal” astronomy.
I do not believe that the $`\gamma `$-ray bursts provide a counter argument to this point of view: they are explosive events in previously inconspicuous objects, and could not have been found in a survey with a narrow-FoV instrument like IACTs; an all-sky capability was needed to discover them, and they were difficult to understand for decades before they were detected also in other wavelength ranges. Also Geminga, originally an enigmatic Cos B source, is not really a counterexample, because Geminga could only be physically identified after many years, when ROSAT discovered that it was a long-period Pulsar and determined its period, which was subsequently confirmed by EGRET.
## Supernova Remnants
### Observations
Earlier observations of the SNRs G87.2+2.1 ($`\gamma `$ Cygni) and IC 443 in 1996/97 gave consistent upper limits between Whipple (Buckley et al. 1998) and the HEGRA CT-System (Heß 1998) at effective threshold energies, for the Zenith angles involved, of $`E_\gamma >300`$ GeV and $`E_\gamma >800`$ GeV, respectively. They were slightly above theoretical predictions regarding the $`\pi ^0`$-decay $`\gamma `$-ray emission for a uniform ISM but well within astronomical uncertainties (Völk 1997). Both objects had originally been assumed to interact with interstellar clouds. Under ideal assumptions such an interaction could have increased the $`\pi ^0`$-decay $`\gamma `$-ray luminosity significantly.
These two SNRs are presumably the result of core collapse Supernovae, due to massive ($`M>8\mathrm{M}_{}`$) progenitor stars. If they have masses exceeding roughly $`15M_{}`$, these stars have stellar winds which significantly modify the circumstellar environment. For such ”Wind-SNe” the time history of the $`\pi ^0`$-decay $`\gamma `$-ray emission is much more complex: except within the wind zone, it is much lower than for a uniform ISM of the same density (Berezhko & Völk 1995, Berezhko & Völk 1997).
The recent HEGRA observations concern deep observations of Tycho’s SNR, believed to be a SN Ia in a uniform ISM with strong X-ray lines and no or only a very weak nonthermal X-ray continuum, and of Cas A, assumed to be a SN Ib resulting from the core collapse of a very massive Wolf-Rayet star, with a strong nonthermal X-ray continuum - an archetypical Wind-SN (Pühlhofer et al. 1999a). Both SNRs are very young in an evolutionary sense, presumably still in the sweep-up phase, even though this might only be marginally true for Tycho’s SNR.
The long observation times which we have reserved for these objects imply a change in our ideology: the emphasis is no more on SNR shocks that presumably interact with interstellar clouds, but on very young objects, either in a supposedly uniform ISM or in a strongly modified precursor wind structure.
### Data analysis for Cas A and Tycho
The following data analysis for Tycho and Cas A has been done by G. Pühlhofer. For the angular resolution of the HEGRA CT-system of $`0.05\mathrm{to}0.1^{}`$ Tycho is a $`\gamma `$-ray point source, and Cas A is marginally extended. Due to the available Zenith angles the instrument threshold is at 1 TeV or slightly above (see Table 2). The complete sensitivities of the IACT system are described in the article by M. Panter (these Proceedings). The observation times and significances are given in Table 2.
With $`38`$ hrs of observation no signal has yet been found from Tycho, whereas the full data sample of $`128`$ hrs for Cas A shows evidence for a signal above 1 TeV. For Cas A the event statistics as a function of $`(\mathrm{distance})^2`$ is shown in the left panel of Figure 3 , both for a point source assumption (I), and for a slightly extended source (II). The position of the $`\gamma `$-ray source on the sky is given in the right panel, and is consistent with the radio astronomical position that corresponds to the center of the picture.
Figure 4 shows a model calculation for the energy spectrum of Cas A (Atoyan et al. 1999b). The full and the dashed lines correspond to the expected inverse Compton (IC) emission, as derived phenomenologically from the observed synchrotron spectrum from the radio to the hard X-ray region (Atoyan et al. 1999a), for the two mean magnetic field strengths $`B_1=1\mathrm{m}\mathrm{G}\mathrm{a}\mathrm{n}\mathrm{d}=1.6\mathrm{mG}`$, respectively. The heavily dotted curve corresponds to a $`\pi ^0`$-decay spectrum, produced by an assumed power law spectrum of protons accelerated in the source, with a total energy content of $`W_p=2\times 10^{49}\mathrm{ergs}1/5W_p(t=\mathrm{})`$, where $`W_p(t=\mathrm{})=10^{50}\mathrm{erg}`$ corresponds to an assumed time-asymptotic nonthermal fraction of 10 percent of the total hydrodynamic energy of $`10^{51}\mathrm{erg}`$, generally assumed to be released in Cas A. The proton spectral index assumed is 2.15, with a rather high cutoff at 200 TeV for this Wind-SN already at very early times (Völk & Biermann 1988). The mean thermal gas density seen by the relativistic protons is taken as $`15\mathrm{cm}^3`$. The energetic electrons responsible for the IC emission are assumed to come from three different regions of the SNR interior: the bright compact radio components with magnetic field strength $`B_1=1\mathrm{mG}`$, where electrons are accelerated locally, an extended ”plateau” of shocked circumstellar gas with $`B_2=B_1/4`$ due to global acceleration at the forward SNR shock, and a low-field part of this ”plateau” with $`B_3=0.1\mathrm{mG}`$. The electron spectral index is assumed to be uniformly 2.15, as for the protons. The electron cutoff energy, however, is only 17 TeV, corresponding to the steep drop-off of the observed hard X-ray spectrum with increasing energy.
Clearly, the magnitude of the $`\gamma `$-ray flux at about 1 TeV, if ultimately detected with a significance exceeding $`5\sigma `$, could be equally due to electronic IC or hadronic $`\pi ^0`$-decay emission. However the spectra would be very different for the two cases: an IC spectrum should fall off strongly with energy, in contrast to a $`\pi ^0`$-decay spectrum. Therefore I believe that every effort should be made to obtain a TeV-spectrum of Cas A.
Acknowledgements I am grateful to Gerd Pühlhofer for providing several of the Figures in this paper. |
warning/0001/hep-ph0001125.html | ar5iv | text | # Introduction
## Introduction
The gauge symmetry is known to render the calculation of the elements of the S matrix very intricate. In some future colliders like LHC or NLC, some very complicated scattering processes will be studied. Phenomenologists will have to consider processes with 3, 4 or more particles in the final state. The scattering amplitudes for these processes are in general very complicated because of the very large number of Feynman graphs, and the numerical evaluation of these amplitudes in Monte-Carlo programs suffer from numerical instabilities due for a large part to some huge compensations between the different graphs, which arise from the gauge symmetry. To avoid these numerical instabilities, there are two common methods. The first one consists in using a specific gauge which simplifies the different vertices and propagators . The second one consists in using some algorithms acting on each Feynman graph, based on Ward identities, in order to simplify the expression of the graphs . Both methods lead to the elimination of most of these huge compensations.
In this paper, we consider this problem from another point of view, at the core of Quantum Field Theory. Basically, we raise the question of whether the calculations of the elements of the S matrix can be done directly using some gauge invariant variables. This question can be studied in the context of both methods cited above, using as a fundamental tool the Ward identities. These identities depend on the gauge fixing procedure used in the calculations. We rather look here for a method in which there is no need to break temporarily the gauge symmetry. As a consequence, we must start our formulation from the very beginning of gauge theories, that is to say from the equations of motion. We therefore show in this paper that one can reformulate these equations in terms of local gauge invariant variables for the case where matter fields are scalar.
This new approach may have some interesting applications regarding the quantization of fields. In standard field theory, the quantization procedure is done first on free fields, and therefore matter fields and gauge fields are considered separately, though they are coupled in the equations of motion. A significant consequence is that it is irrelevant to consider the evolution of a free field from a time $`t`$ to an interacting field at time $`t^{}>t`$ through a unitary transformation, because Haag’s theorem says that the field considered at time $`t^{}`$ must be also free (for a good review, see ). Quantum Field Theory is therefore doomed to describe only the transition between asymptotic fields through the LSZ formalism. In experiments where the time variable plays a fundamental role (CP violation experiments in $`K_S^0/K_L^0`$, neutrino oscillations,…) one must use a mixed theory, based in part on classical quantum mechanics (Rabi precession,…) and in part on quantum field theory for the computation of the decay width of the particles. A single theory which would describe completely such experiments is still missing. Since Haag’s theorem does not apply to the case of two constantly interacting fields, the approach presented in this paper opens the prospect of finding an evolution operator between two finite times for an interacting system. That is to say, an asymptotic electron would be described both by its matter field and its surrounding electromagnetic field, in some sense. So we must also find some “realistic” asymptotic solutions to the coupled equations of motion in replacement of the plane waves that are used in standard quantum field theory. The word “realistic” means here that we look for solutions that have a finite conserved momentum. We show that solitons are not possible in this context (for the $`U(1)`$ case), but we conjecture that some periodic-in-time solutions may probably exist.
What is the basic idea of our approach? We know that for a given field-strength tensor, one can compute a corresponding gauge field using the basic cohomological formulas that are reviewed in the appendix. Some authors have already tried to reformulate the Yang-Mills Theory using only the Field-Strength tensor as a basic variable in place of the gauge field . The results of these studies are generally not covariant and non-local, due to the fact that the cohomological formulas are essentially of a non-local nature. In this paper, we rather consider the gauge-currents as fundamental variables, and we keep both locality and covariance of the equations.
The paper is therefore organized as follows:
The first section is devoted to the reformulation of $`U(1)`$ scalar QED in terms of gauge invariant variables.
The second section contains a discussion on asymptotic solutions of the $`U(1)`$ scalar QED. We first show that periodic solutions of Klein Gordon do not have a finite energy, contrary to what is claimed in a recent paper, and therefore we need to consider the coupled equations. We show the impossibility of soliton solutions in this context, and discuss the possibility of periodic (in time) solutions.
The third section presents the non-abelian case, where the gauge group is in a certain class of subgroups of $`U(N)`$. It turns out that the results presented in this paper are in fact a simpler version of some results given by Lunev in 1994 , with in addition the coupling to a scalar multiplet (he only considered a pure Yang-Mills theory).
## 1 A possible reformulation of classical SQED
In this section, we reformulate the classical theory of a scalar field coupled to a $`U(1)`$ gauge field (SQED) in terms of gauge invariant variables. We will then demonstrate the difficulties appearing when one wants to find some “realistic” asymptotic solutions, which would generate a Fock-like space. Using such a space, one could then construct a new formalism for computing cross sections. Let us start with the classical scalar QED lagrangian:
$$=(D_\mu \varphi )^{}D^\mu \varphi m^2\varphi ^{}\varphi \frac{1}{2}_\mu A_\nu (^\mu A^\nu ^\nu A^\mu )$$
(1)
with $`D_\mu =_\mu +ieA_\mu `$. The electrical current is given by $`J_\mu =ie(\varphi ^{}(D_\mu \varphi )(D_\mu \varphi )^{}\varphi )`$ and the probability density $`\rho =\varphi ^{}\varphi `$. Both $`J_\mu `$ and $`\rho `$ are gauge invariant variables and we will show how to rewrite the previous lagrangian as a function of these variables (this treatment will have to be modified in the non-abelian case in which the corresponding expression for these variables are not gauge invariant but gauge “covariant”). First, we shall review the standard equations of motion when $`\varphi `$ and $`A^\mu `$ are taken as field variables:
$`0`$ $`=`$ $`(D_\mu D^\mu +m^2)\varphi `$ (2)
$`0`$ $`=`$ $`(\mathrm{}+m^2)\varphi +2ieA_\mu ^\mu \varphi +ie(A)\varphi e^2(AA)\varphi `$ (3)
$`^\alpha F_{\alpha \beta }`$ $`=`$ $`ie(\varphi ^{}(D_\beta \varphi )(D_\beta \varphi )^{}\varphi )=J_\beta `$ (4)
We shall first note that if one computes $`\varphi ^{}(\text{3})(\text{3})^{}\varphi `$, one obtains $`_\mu J^\mu =0`$, which we would have already obtained by taking the divergence of eq. 4. The redundancy between the last two equations can therefore be removed by making use of $`\varphi ^{}(\text{3})+(\text{3})^{}\varphi `$ instead of Eq. 3. After some algebra, it is not a hard task to make $`J_\mu `$ and $`\rho `$ appear in the equations as we will see later, but for the derivation of the new equations, we rather choose to start from the lagrangian. For this purpose, we will use the following relations:
$`{\displaystyle \frac{J^2}{e^2}}`$ $`=`$ $`(\varphi ^{}(D_\mu \varphi )+(D_\mu \varphi )^{}\varphi )^24(D_\mu \varphi )^{}\varphi \varphi ^{}(D_\mu \varphi )`$ (5)
$`(D_\mu \varphi )^{}D^\mu \varphi `$ $`=`$ $`{\displaystyle \frac{1}{4\rho }}\left((_\mu \rho )^2+{\displaystyle \frac{J^2}{e^2}}\right)`$ (6)
$`=`$ $`(_\mu \sqrt{\rho })^2+{\displaystyle \frac{J^2}{4e^2\rho }}`$ (7)
Throughout the paper, we will conveniently define $`v^\mu `$ such that $`J^\mu =2e^2\rho v^\mu `$ and set $`z(x)=\sqrt{\rho (x)}`$. From the definition of the current, one can also extract the expression of the field strength tensor:
$`2e^2F_{\mu \nu }`$ $`=`$ $`_\mu \left({\displaystyle \frac{J_\nu }{\rho }}\right)_\nu \left({\displaystyle \frac{J_\mu }{\rho }}\right)`$ (8)
$`F_{\mu \nu }`$ $`=`$ $`_\mu (v_\nu )_\nu (v_\mu )`$ (9)
Equations 7 and 9 are the fundamental tools of our formalism. With these, we can write the lagrangian as a function of $`z`$ and $`v^\nu `$ in the following way:
$$=(_\mu z)^2m^2z^2+e^2z^2v^2\frac{1}{4}(_\mu (v_\nu )_\nu (v_\mu ))^2$$
(10)
We have now re-expressed the lagrangian in terms of gauge invariant quantities, and as a by-product the “effective” coupling constant is $`e^2=4\pi \alpha `$ instead of $`e`$. This means that the sign of $`e`$ is not relevant. Although this does not mean that in a perturbative expansion of some solutions, the relevant expansion parameter is necessarily $`e^2`$, it may also be $`\sqrt{4\pi \alpha }`$ or $`|e|`$. From this new lagrangian we can derive the following equations of motion thanks to the Euler-Lagrange equations:
$`(\mathrm{}+m^2)z`$ $`=`$ $`4\pi \alpha zv^2`$ (11)
$`\mathrm{}(v_\nu )_\nu (v)`$ $`=`$ $`8\pi \alpha z^2v_\nu `$ (12)
### 1.1 The Energy-Momentum Tensor
We will further look for asymptotic solutions to the coupled equations with a finite conserved momentum. The symmetrized energy momentum tensor (or Belinfante tensor) can be rewritten this way:
$`T_{\mu \nu }`$ $`\stackrel{def}{=}`$ $`(D_\mu \varphi )^{}D_\nu \varphi +(D_\nu \varphi )^{}D_\mu \varphi F_{\mu \lambda }F_\nu ^\lambda g_{\mu \nu }\left((D_\mu \varphi )^{}D^\mu \varphi m^2\varphi ^{}\varphi {\displaystyle \frac{1}{4}}F_{\alpha \beta }F^{\alpha \beta }\right)`$ (13)
$`T_{\mu \nu }`$ $`=`$ $`2e^2z^2v_\mu v_\nu +2_\mu z_\nu z\left(_\mu (v_\lambda )_\lambda (v_\mu )\right)\left(_\nu (v^\lambda )^\lambda (v_\nu )\right)`$ (14)
$`g_{\mu \nu }\left((_\mu z)^2m^2z^2+e^2z^2v^2{\displaystyle \frac{1}{4}}(_\mu (v_\nu )_\nu (v_\mu ))^2\right)`$
$`P_\mu `$ $`=`$ $`{\displaystyle _\mathrm{\Sigma }}𝑑\sigma ^\nu T_{\nu \mu }`$ (15)
In Eq. 15, $`\mathrm{\Sigma }`$ represents any space-like hyper-surface in the Minkowsky space time, and for the sake of simplicity, we will generally take the $`t=0`$ hypersurface for the computation of $`P_\mu `$.
## 2 Solitons solutions are not normalizable
In general, the spatial extent of the wave function of a free particle (obeying the Klein Gordon equation) increases in time. Here, we will rather look for the possibility to find “soliton-like” solution to the coupled field equations of scalar QED (i.e. Eq. 11 and Eq. 12). First, we will show that for the Klein Gordon equation, we can find some simple “soliton-like” solutions but these solutions are not normalizable (similarly to plane waves). We obtain the same result when the interaction is taken into account, but the arguments used to reject this case are different from the free case. For this reason, the free case is also presented, even if it can be seen as a particular case of the interacting one.
### 2.1 Generalities about solitons
We will say that a function $`f(x)`$ defined on space-time is a soliton if we can find a time-like momentum $`p^\mu `$ such that:
$$p^\mu _\mu f=0$$
(16)
This time-like momentum represents the global momentum of the wave which moves without deformation. To see this trivial fact, Eq. 16 simply means that if we are placed in a frame where $`p^\mu =(m_0,\stackrel{}{0})`$, then the shape of the wave function does not depend on time ($`_0f=0`$). Suppose now that at time $`t=0`$ we look at the shape of the wave function. It is reasonable to say that for an asymptotic solution (supposed to describe a free scalar particle) the probability density is spherically symmetric. We can deduce from that that the function $`f`$ is a function of only one variable. To be more specific, let us consider the two variables $`u=(px)^2p^2x^2`$ and $`\tau =px`$. In the “rest frame”, where $`p^\mu =(m_0,\stackrel{}{0})`$ <sup>(1)</sup><sup>(1)</sup>(1)We shall remark that $`p^2=m_0^2`$, where the mass $`m_0`$ is a priori different from the mass $`m`$ appearing in the Klein Gordon equation., then:
$`u`$ $`=`$ $`(px)^2p^2x^2`$ (17)
$`=`$ $`(m_0t)^2m_0^2(t^2\stackrel{}{x}^2)=m_0^2\stackrel{}{x}^2`$ (18)
$`\tau `$ $`=`$ $`px=m_0t`$ (19)
A “spherically symmetric” scalar function $`f`$ is therefore a function of $`u`$ and $`\tau `$ only. We have seen that $`u0`$ for any $`x`$ and we will often write $`y=\sqrt{u}`$. We have by construction $`u(x^\mu +\lambda p^\mu )=u(x^\mu )`$, which means that a function of the variable $`u`$ is invariant under any translation in the $`p^\mu `$ direction.
For convenience, we will also use the following notations:
$`\lambda ^\alpha `$ $`=`$ $`p^\alpha (px)p^2x^\alpha ={\displaystyle \frac{1}{2}}^\alpha u(x)`$ (20)
$`\lambda ^2`$ $`=`$ $`p^2u(x)=m_0^2u(x)p\lambda =0`$ (21)
$`_\alpha \lambda _\beta `$ $`=`$ $`p_\alpha p_\beta m_0^2g_{\alpha \beta }=\tau _{\alpha \beta }`$ (22)
$`^\alpha \lambda _\alpha `$ $`=`$ $`\tau _\alpha ^\alpha =3m_0^2`$ (23)
$`p^\alpha \tau _{\alpha \beta }`$ $`=`$ $`0\lambda ^\alpha \tau _{\alpha \beta }=m_0^2\lambda _\beta `$ (24)
$`_\mu \sqrt{u}`$ $`=`$ $`{\displaystyle \frac{\lambda _\mu }{\sqrt{u}}}^\mu \left({\displaystyle \frac{\lambda _\mu }{\sqrt{u}}}\right)={\displaystyle \frac{2m_0^2}{\sqrt{u}}}`$ (25)
And in the rest frame, $`\lambda ^\mu =m_0^2(0,\stackrel{}{x})`$. Then, if the scalar function $`f`$ is a “spherically symmetric” soliton, we have:
$`f(x)`$ $`=`$ $`g(\tau ,u=(px)^2p^2x^2)`$ (27)
$`0=p^\mu _\mu f`$ $`=`$ $`p^\mu (p_\mu _0g+2\lambda _\mu _1g)=p^2_0g`$ (28)
Therefore $`g`$ does not depend on $`\tau `$, but only on $`u`$. We therefore obtain a covariant formulation of the notion of a “spherically symmetric” soliton.
### 2.2 Periodic solutions to the Klein Gordon equation
Before we look for some asymptotic solutions to the coupled equations, we must explain why we cannot have some realistic asymptotic states in the free case. Of course finite energy solutions to the Klein-Gordon equation exist, consisting in wave-packets with square integrable momentum densities. But one of the criteria we set in order to define a “realistic” asymptotic field is that the wave-packet must be “bounded” in space-like directions. We consider here that a constantly spreading wave-packet cannot represent the state of a stable free particle. We show in this paragraph that soliton solutions to the Klein-Gordon equation cannot have a finite energy, as a particular case of a stronger result concerning periodic-in-time solutions. The free scalar lagrangian is:
$$=_\mu \mathrm{\Phi }^{}^\mu \mathrm{\Phi }m^2\mathrm{\Phi }^{}\mathrm{\Phi }$$
(29)
The energy-momentum tensor and the corresponding conserved total momentum are:
$`T_{\mu \nu }`$ $`=`$ $`_\mu \mathrm{\Phi }^{}_\nu \mathrm{\Phi }+_\nu \mathrm{\Phi }^{}_\mu \mathrm{\Phi }g_{\mu \nu }`$ (30)
$`P_\nu `$ $`=`$ $`{\displaystyle _\mathrm{\Sigma }}𝑑\sigma ^\mu T_{\mu \nu }`$ (31)
The linearity of the equations of motion allows us to expand the field in a Fourier serie:
$`\varphi (t,\stackrel{}{x})`$ $`=`$ $`\mathrm{\Sigma }_{n=\mathrm{}}^{\mathrm{}}a_n(\stackrel{}{x})e^{in\omega t}`$ (32)
$`=`$ $`\mathrm{\Sigma }_{n=\mathrm{}}^{\mathrm{}}a_n(r)e^{in\omega t}`$ (33)
We first look for solutions of the form $`\varphi =\mathrm{exp}(i\eta px)g(\sqrt{u})`$, where $`\eta `$ is a real parameter. We have:
$`^\mu \varphi `$ $`=`$ $`e^{i\eta px}\left(i\eta g(\sqrt{u})p^\mu +{\displaystyle \frac{\lambda ^\mu }{\sqrt{u}}}g^{}(\sqrt{u})\right)`$ (35)
$`\mathrm{}\varphi `$ $`=`$ $`e^{i\eta px}\left(m_0^2\eta ^2g(\sqrt{u}){\displaystyle \frac{2m_0^2}{\sqrt{u}}}g^{}(\sqrt{u})m_0^2g^{\prime \prime }(\sqrt{u})\right)`$ (36)
$`=`$ $`{\displaystyle \frac{m_0^2}{y}}e^{i\eta px}\left(\eta ^2t(y)+t^{\prime \prime }(y)\right)(g(y)=t(y)/y)`$
$`0`$ $`=`$ $`(\mathrm{}+m^2)\mathrm{\Phi }`$
$`0`$ $`=`$ $`t^{\prime \prime }(y)+t(y)\left(\eta ^2{\displaystyle \frac{m^2}{m_0^2}}\right)`$ (37)
$`t(y)`$ $`=`$ $`Ae^{\sqrt{\frac{m^2}{m_0^2}\eta ^2}y}`$ (38)
We have not considered the other solution that increases as $`y`$ (or $`r`$) increases, because we look for normalized solutions. Thus, the general solution is:
$`\varphi (x)`$ $`=`$ $`{\displaystyle \frac{1}{y}}{\displaystyle \underset{|n|\left[\frac{m}{m_0}\right]}{}}A_ne^{inpx}e^{\sqrt{\frac{m^2}{m_0^2}n^2}y}`$ (39)
$`=`$ $`{\displaystyle \frac{1}{m_0r}}{\displaystyle \underset{|n|\left[\frac{m}{m_0}\right]}{}}A_ne^{inm_0t}e^{\sqrt{m^2n^2m_0^2}r}`$ (40)
The sum has a finite number of terms because we limit ourselves to exponentially decreasing terms. It will be clear in the following that the oscillating solutions for $`|n|>\left[\frac{m}{m_0}\right]`$ will not provide normalizable solutions. Contrary to the claim of Hormuzdiar and Hsu in which considered only the large r behaviour, the solutions are not normalizable. This is due to their small $`r`$ behaviour. This can be shown by computing the conserved momentum:
$`P_0`$ $`=`$ $`{\displaystyle _{t=0}}𝑑\stackrel{}{x}\left[2_0\varphi ^{}_0\varphi g_{00}(_0\varphi ^{}_0\varphi \stackrel{}{}\varphi ^{}\stackrel{}{}\varphi m^2\varphi ^{}\varphi )\right]`$ (41)
$`=`$ $`{\displaystyle _{t=0}}d\stackrel{}{x}[_0\varphi ^{}_0\varphi +\stackrel{}{}\varphi ^{}\stackrel{}{}\varphi +m^2\varphi ^{}\varphi )](0)`$ (43)
$`=`$ $`4\pi {\displaystyle _0^{\mathrm{}}}r^2dr{\displaystyle \frac{1}{r^2}}\left[\right|{\displaystyle \underset{n}{}}inm_0A_ne^{\sqrt{m^2n^2m_0^2}r}|^2+|{\displaystyle \underset{n}{}}A_ne^{\sqrt{m^2n^2m_0^2}r}(\sqrt{m^2n^2m_0^2}+{\displaystyle \frac{1}{r}})|^2`$
$`+m^2\left|{\displaystyle \underset{n}{}}A_ne^{\sqrt{m^2n^2m_0^2}r}|^2\right]`$
and the $`1/r`$ term in the second squared term makes the integral divergent. The integral converges if $`A_n=0`$ but the computation on another space-like hypersurface $`t=t_00`$ would be still divergent, which is an indication that the computation at $`t=0`$ is meaningless, even if it can be accidentally convergent.
### 2.3 Solitons for the coupled SQED equations
The field $`v^\mu `$ may also be written in a simple generic form if we suppose that it obeys the spherically-symmetric soliton condition. The most general form compatible with the symmetries of the solution is given by:
$`v^\mu (x)`$ $`=`$ $`a(u)\lambda ^\mu +b(u)p^\mu `$ (44)
$`v^2`$ $`=`$ $`m_0^2(b^2a^2u)`$ (45)
The first term of $`v^\mu `$ does not contribute to the field strength tensor because if we set $`A=_0^ua(s)𝑑s`$, then $`a(u)\lambda ^\mu =^\mu (A(u)/2)`$ which is a pure gauge term. And we will further demonstrate that this term must vanish. However, we will see in the next sections that for periodic solutions, this term is important.
We will also need to comply with the classical asymptotic conditions at infinity in space-like directions. One must therefore have $`A^\mu `$ decreasing as $`1/r`$ at infinity, and thus $`b(u)C/\sqrt{u}`$ when $`u\mathrm{}`$.
Then we can substitute $`v^\mu `$ and $`z(x)=f(u)`$ in the equations of motion Eq. 11 and Eq. 12 :
$`(\mathrm{}+(m^2e^2v^2))z`$ $`=`$ $`0`$ (46)
$`\mathrm{}(v_\nu )_\nu (v)`$ $`=`$ $`2e^2z^2v_\nu `$ (47)
Using the parameterization of $`v^\mu `$ given in Eq. 44 one gets:
$`_\alpha v_\beta `$ $`=`$ $`2{\displaystyle \frac{d(a)}{du}}\lambda _\alpha \lambda _\beta +2{\displaystyle \frac{d(b)}{du}}\lambda _\alpha p_\beta +a\tau _{\alpha \beta }`$ (48)
$`F_{\mu \nu }=_\mu v_\nu _\nu v_\mu `$ $`=`$ $`2{\displaystyle \frac{d(b)}{du}}(\lambda p)^{\mu \nu }=2m_0^2{\displaystyle \frac{d(b)}{du}}(xp)^{\mu \nu }`$ (49)
$`F_{\mu \lambda }F_\nu ^\lambda `$ $`=`$ $`4m_0^4{\displaystyle \frac{d(b)}{du}}^2(x_\mu x_\nu m_0^2(px)(p_\mu x_\nu +p_\nu x_\mu )+x^2p_\mu p_\nu )`$ (50)
$`F^2`$ $`=`$ $`8m_0^4{\displaystyle \frac{d(b)}{du}}^2u`$ (51)
$`\mathrm{}v_\nu _\nu (v)`$ $`=`$ $`4m_0^2u{\displaystyle \frac{d^2(b)}{du^2}}p_\nu 6m_0^2{\displaystyle \frac{d(b)}{du}}p_\nu `$ (52)
Thus Eq. 47 yields:
$`4m_0^2u{\displaystyle \frac{d^2(b)}{du^2}}p_\nu 6m_0^2{\displaystyle \frac{d(b)}{du}}p_\nu `$ $`=`$ $`2e^2f^2(a\lambda _\nu +bp_\nu )`$ (53)
$`a`$ $`=`$ $`0`$ (54)
$`and4u{\displaystyle \frac{d^2(b)}{du^2}}+6{\displaystyle \frac{d(b)}{du}}`$ $`=`$ $`{\displaystyle \frac{2e^2}{m_0^2}}f^2b`$ (55)
$`\overline{)b(u)={\displaystyle \frac{\stackrel{~}{b}(\sqrt{u})}{\sqrt{u}}}}\stackrel{~}{b}^{\prime \prime }`$ $`=`$ $`{\displaystyle \frac{2e^2}{m_0^2}}f^2\stackrel{~}{b}`$ (56)
Similarly, we will use the change of variable $`f(u)={\displaystyle \frac{m_0t(\sqrt{u})}{\sqrt{u}}}`$ in Eq. 46 and in Eq. 56. We finally obtain this system of coupled differential equations:
$`t^{\prime \prime }(y)\left({\displaystyle \frac{m^2}{m_0^2}}e^2{\displaystyle \frac{\stackrel{~}{b}(y)^2}{y^2}}\right)t`$ $`=`$ $`0`$ (57)
$`\stackrel{~}{b}^{\prime \prime }(y)2e^2{\displaystyle \frac{t(y)^2}{y^2}}\stackrel{~}{b}(y)`$ $`=`$ $`0`$ (58)
#### 2.3.1 Normalization of the solutions
In this paragraph, we compute the conserved momentum of spherically symmetric solitons. We will show that the solutions cannot be normalized. Considering the energy-momentum tensor of Eq. 14, we get for a soliton:
$`\stackrel{}{P}`$ $`=`$ $`\stackrel{}{0}`$ (59)
$`F_{0\lambda }F_0^\lambda `$ $`=`$ $`4m_0^6\left({\displaystyle \frac{db}{du}}\right)^2r^2(restframe,t=0,r=|\stackrel{}{x}|)`$ (60)
$`v^0`$ $`=`$ $`m_0b(u);_0z=0;(_\mu z)^2=4m_0^2u\left({\displaystyle \frac{df}{du}}\right)^2`$ (61)
$`T_{00}`$ $`=`$ $`e^2z^2(2m_0^2b^2m_0^2b^2+m_0^2a^2u)+4m_0^2u\left({\displaystyle \frac{df}{du}}\right)^2+m^2f^2+4m_0^6\left({\displaystyle \frac{db}{du}}\right)^2r^22m_0^4\left({\displaystyle \frac{db}{du}}\right)^2u`$ (62)
$`=`$ $`m_0^2e^2f^2b^2+4m_0^2u\left({\displaystyle \frac{df}{du}}\right)^2+m^2f^2+2m_0^6\left({\displaystyle \frac{db}{du}}\right)^2r^2`$ (63)
$`=`$ $`e^2m_0^4{\displaystyle \frac{t^2\stackrel{~}{b}^2}{y^4}}+m^2{\displaystyle \frac{t^2}{y^2}}+4m_0^6r^2\left({\displaystyle \frac{1}{2y}}{\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{t(y)}{y}}\right)\right)^2+2m_0^6r^2\left({\displaystyle \frac{1}{2y}}{\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{\stackrel{~}{b}(y)}{y}}\right)\right)^2`$ (64)
$`=`$ $`e^2m_0^4{\displaystyle \frac{t^2\stackrel{~}{b}^2}{y^4}}+m^2{\displaystyle \frac{t^2}{y^2}}+m_0^4\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{t(y)}{y}}\right)\right)^2+{\displaystyle \frac{m_0^4}{2}}\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{\stackrel{~}{b}(y)}{y}}\right)\right)^2`$ (65)
$`P_0`$ $`=`$ $`4\pi {\displaystyle _0^{\mathrm{}}}r^2𝑑rT_{00}`$ (66)
$`P_0`$ $`=`$ $`{\displaystyle \frac{4\pi }{m_0^3}}{\displaystyle _0^{\mathrm{}}}y^2𝑑y\left[e^2m_0^4{\displaystyle \frac{t^2\stackrel{~}{b}^2}{y^4}}+m^2{\displaystyle \frac{t^2}{y^2}}+m_0^4\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{t(y)}{y}}\right)\right)^2+{\displaystyle \frac{m_0^4}{2}}\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{\stackrel{~}{b}(y)}{y}}\right)\right)^2\right]`$ (67)
$`=`$ $`4\pi m_0{\displaystyle _0^{\mathrm{}}}𝑑y\left[e^2{\displaystyle \frac{t^2\stackrel{~}{b}^2}{y^2}}+\left({\displaystyle \frac{m}{m_0}}\right)^2t^2+y^2\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{t(y)}{y}}\right)\right)^2+{\displaystyle \frac{y^2}{2}}\left({\displaystyle \frac{d}{dy}}\left({\displaystyle \frac{\stackrel{~}{b}(y)}{y}}\right)\right)^2\right]`$
We have seen that $`\stackrel{~}{b}`$ must tend to a non-vanishing constant at infinity in space-like directions ($`A^\mu 1/r`$), but from Eq. 58 we can conclude that $`\stackrel{~}{b}`$ is a convex function when $`\stackrel{~}{b}>0`$ and the converse for the other sign. From the last term in Eq. 67, we get that $`\stackrel{~}{b}`$ cannot tend to a non-vanishing value in $`y=0`$ (otherwise the integral is divergent). Thus if $`\stackrel{~}{b}`$ vanish in $`y=0`$, it cannot tend to a non-vanishing constant at infinity because it is a convex function if $`\stackrel{~}{b}>0`$ or the converse if $`\stackrel{~}{b}<0`$. The only possibility is $`\stackrel{~}{b}=0`$, and we are then back to the free case, which we have previously rejected.
### 2.4 Is there some periodic solutions to the coupled equations?
Now we introduce a “time” variable $`\tau =px`$ which is dimensionless and $`y=\sqrt{u}`$ like in the soliton case. We have:
$`z(x)`$ $`=`$ $`f(\tau ,y)={\displaystyle \frac{t(\tau ,y)}{y}}`$ (68)
$`v^\mu `$ $`=`$ $`a(\tau ,y)\lambda ^\mu +b(\tau ,y)p^\mu `$ (69)
$`_\mu z`$ $`=`$ $`p_\mu _0f+{\displaystyle \frac{\lambda ^\mu }{y}}_1f`$ (70)
$`\mathrm{}z`$ $`=`$ $`m_0^2\left(_0^2f{\displaystyle \frac{2}{y}}_1f_1^2f\right)={\displaystyle \frac{m_0^2}{y}}\left(_0^2t_1^2t\right)`$ (71)
$`\mathrm{}v_\mu _\mu (v)`$ $`=`$ $`m_0^2p_\mu \left[\rho _0_1a+3_0a_1^2b{\displaystyle \frac{2}{y}}_1b\right]+m_0^2\lambda _\mu \left[_0^2a{\displaystyle \frac{_0_1b}{y}}\right]`$ (72)
From these basic calculations we get for the equations of motion:
$`_0^2t_1^2t+\left({\displaystyle \frac{m^2}{m_0^2}}e^2(b^2y^2a^2)\right)t`$ $`=`$ $`0`$ (73)
$`_0^2a{\displaystyle \frac{_0_1b}{y}}`$ $`=`$ $`2{\displaystyle \frac{e^2t^2}{m_0^2y^2}}a`$ (74)
$`y_0_1a+3_0a_1^2b{\displaystyle \frac{2}{y}}_1b`$ $`=`$ $`2{\displaystyle \frac{e^2t^2}{m_0^2y^2}}b`$ (75)
These equations are much more complicated than in the case of solitons and the fundamental structure of the solutions, even periodic in time is not clear so far. We will restrict ourselves in this paragraph to a description of what is really different in this case and why we conjecture the existence of some normalized periodic solutions.
The conservation of the electromagnetic current leads to the emergence of a kind of pre-potential:
$`_\mu (z^2v^\mu )`$ $`=`$ $`0`$ (76)
$`_1(yt^2a)`$ $`=`$ $`_0(t^2b)`$ (77)
$`yt^2a`$ $`=`$ $`_0\phi (\tau ,y)`$ (78)
$`andt^2b`$ $`=`$ $`_1\phi (\tau ,y)`$ (79)
Introducing this potential in the equations for the electromagnetic field we get:
$`_0\left(_0a{\displaystyle \frac{_1b}{y}}\right)`$ $`=`$ $`2{\displaystyle \frac{e^2}{m_0^2}}_0\left({\displaystyle \frac{\phi }{y^3}}\right)`$ (80)
$`_1\left(y^2_1b\right)_1_0(y^3a)`$ $`=`$ $`2{\displaystyle \frac{e^2}{m_0^2}}_1\phi `$ (81)
These equations can be partially integrated, and we obtain:
$`_0a{\displaystyle \frac{_1b}{y}}`$ $`=`$ $`2{\displaystyle \frac{e^2}{m_0^2}}{\displaystyle \frac{\phi }{y^3}}+A(y)`$ (82)
$`y^2_1b_0(y^3a)`$ $`=`$ $`2{\displaystyle \frac{e^2}{m_0^2}}\phi +B(\tau )`$ (83)
The presence of these two functions A and B enlarges significantly the set of possibilities for the solutions. We therefore hope that some of these might be normalizable, as we shall discuss further.
#### 2.4.1 Normalization of the time-dependent solutions
In order to normalize these periodic solutions, the computation of the conserved momentum gives for Eq. 14:
$`F_{\mu \nu }`$ $`=`$ $`(p\lambda )_{\mu \nu }\left(_0a{\displaystyle \frac{_1b}{y}}\right)`$ (84)
$`F_{\mu \alpha }F_\nu ^\alpha `$ $`=`$ $`m_0^2\left(_0a{\displaystyle \frac{_1b}{y}}\right)^2(y^2p_\mu p_\nu \lambda _\mu \lambda _\nu )F_{0\alpha }F_\mu ^\alpha =m_0^3y^2\left(_0a{\displaystyle \frac{_1b}{y}}\right)^2p_\mu `$ (85)
$`F_{\beta \alpha }F^{\beta \alpha }`$ $`=`$ $`2m_0^4y^2\left(_0a{\displaystyle \frac{_1b}{y}}\right)^2`$ (86)
$`T_{0i}`$ $`=`$ $`2e^2f^2m_0ab\lambda ^i+2m_0_0f{\displaystyle \frac{\lambda ^i}{y}}`$ (87)
$`P_i`$ $`=`$ $`0`$ (88)
$`T_{00}`$ $`=`$ $`e^2m_0^2f^2(b^2+a^2y^2)+m^2f^2+m_0^2\left((_0f)^2+(_1f)^2\right)+{\displaystyle \frac{y^2}{2}}\left(_0a{\displaystyle \frac{_1b}{y}}\right)^2`$ (89)
$`P_0`$ $`=`$ $`{\displaystyle \frac{4\pi }{m_0^3}}{\displaystyle _0^{\mathrm{}}}y^2𝑑yT_{00}`$ (90)
$`=`$ $`{\displaystyle \frac{4\pi }{m_0}}{\displaystyle _0^{\mathrm{}}}y^2𝑑y\left(e^2f^2(b^2+a^2y^2)+{\displaystyle \frac{m^2}{m_0^2}}f^2+\left((_0f)^2+(_1f)^2\right)+{\displaystyle \frac{y^2}{2m_0^2}}\left(_0a{\displaystyle \frac{_1b}{y}}\right)^2\right)`$ (91)
The last term in Eq. 91 also appears in Eq. 82, equation that was absent when we considered soliton solutions. In this equation, the function A is undetermined but if $`\frac{\phi }{y^3}`$ is sufficiently singular at 0, the function A will certainly not compensate the singularity because it is time-independent, and $`\phi `$ is periodic. Thus, if A accidentally compensate $`\frac{\phi }{y^3}`$ at $`y=0`$ for $`t=0`$, it may not be the case at a different time. As a consequence, $`\phi `$ must certainly vanish at $`y=0`$ if one wants the integral to be convergent.
We still have in this case some dramatic constraints on the behaviour of the solutions at $`y=0`$. However, what prevented us from finding normalized periodic solutions in the free case was the finite value of t at $`y=0`$. In the free case, solutions are only composed of exponentially decreasing functions. Here we have another “mass” term in the equation of motion for t. If $`b^2a^2y^2`$ becomes large in the vicinity of the origin, one may obtain solutions that are spatially oscillating (and only near $`y=0`$). Such a possibility allows to have a t function that vanishes at $`y=0`$, while still featuring an exponentially decreasing behaviour at infinity. We expect soon to be able to confirm this conjecture by numerical simulations, before we can get more rigorous answers to this problem.
## 3 The non-abelian case
### 3.1 The standard equations of motion
In this case we consider a scalar field $`\mathrm{\Phi }`$ lying in an N-dimensional vector space of representation of the Lie group $`𝒢`$ (a subgroup of $`U(N)`$). The results presented in this section will not work for all the possible gauge groups, yet our method is valid for $`U(N)`$ or $`SU(N)`$. There are very few constraints that may be imposed on a generic gauge group. Probably the most important one is that there must exist a scalar product on the Lie algebra which is invariant under an inner automorphism. One can then demonstrate that the solvable part (in the Levi decomposition of the group) must be abelian. Thus if we also constrain the group to be compact, it is relevant to consider gauge groups as being a sum of $`U(1)`$ terms, plus any semi-simple part like $`SU(N)`$. Since the $`U(1)`$ case has been previously solved, we focus here on $`SU(N)`$ groups. Actually, we will see that our formalism works if the orbit of any vector $`\mathrm{\Phi }`$ under the gauge group is $`C^N`$, which is the case for $`SU(N)`$.
We will denote by $`i𝒜`$ the real Lie algebra, such that the matrices lying in $`𝒜`$ are hermitian. If $`\rho `$ stands for the representation of the Lie algebra, we can endow the algebra with the following scalar product for the computation of the Yang-Mills part of the lagrangian: $`(A,B)_\rho =\mathrm{Tr}[\rho (A)^{}\rho (B)]`$. The scalar product generally used with a semi-simple group is the Killing form applied to the field strength tensor. Since this scalar product is proportional to any scalar product of the form $`(A,B)_\rho `$ (the coefficient being the Dynkin index of $`\rho `$), we will simply use this scalar product $`(A,B)=\mathrm{Tr}[A^{}B]=\mathrm{Tr}[AB]`$. In the following, we give the lagrangian and the corresponding equations of motion, using $`\mathrm{\Phi }`$ and $`W_\mu `$ as variables.
$``$ $`=`$ $`_0+_{YM}`$
$`_0`$ $`=`$ $`(D_\mu \mathrm{\Phi })^{}D^\mu \mathrm{\Phi }m^2\mathrm{\Phi }^{}\mathrm{\Phi }`$ (92)
$`=`$ $`_\mu \mathrm{\Phi }^{}^\mu \mathrm{\Phi }ig\mathrm{\Phi }^{}W_\mu ^\mu \mathrm{\Phi }+ig_\mu \mathrm{\Phi }^{}W^\mu \mathrm{\Phi }+g^2\mathrm{\Phi }^{}W_\mu W^\mu \mathrm{\Phi }m^2\mathrm{\Phi }^{}\mathrm{\Phi }`$
$`_{YM}`$ $`=`$ $`{\displaystyle \frac{1}{4}}(F_{\mu \nu },F^{\mu \nu })={\displaystyle \frac{1}{4}}\{\mathrm{Tr}\left[(_\mu W_\nu _\nu W_\mu )(^\mu W^\nu ^\nu W^\mu )\right]`$ (93)
$`g^2\mathrm{Tr}\left[[W_\mu ,W_\nu ][W^\mu ,W^\nu ]\right]+2ig\mathrm{Tr}\left[(_\mu W_\nu _\nu W_\mu )[W^\mu ,W^\nu ]\right]\}`$
$`{\displaystyle \frac{}{(_\mu W_\nu )}}(\mathrm{\Omega }_\nu )`$ $`=`$ $`\mathrm{Tr}[\mathrm{\Omega }_\nu G^{\mu \nu }]`$ (94)
$`_\mu {\displaystyle \frac{}{(_\mu W_\nu )}}(\mathrm{\Omega }_\nu ){\displaystyle \frac{}{(W_\nu )}}(\mathrm{\Omega }_\nu )`$ $`=`$ $`\mathrm{Tr}[\mathrm{\Omega }_\nu _\mu G^{\mu \nu }]\left(ig\mathrm{\Phi }^{}\mathrm{\Omega }_\mu ^\mu \mathrm{\Phi }+ig_\mu \mathrm{\Phi }^{}\mathrm{\Omega }^\mu \mathrm{\Phi }+g^2\mathrm{\Phi }^{}\{\mathrm{\Omega }_\mu ,W^\mu \}\mathrm{\Phi }\right)`$ (96)
$`+ig\mathrm{Tr}\left[[W_\mu ,\mathrm{\Omega }_\nu ]G^{\mu \nu }\right]`$
$`=`$ $`\mathrm{Tr}[\mathrm{\Omega }_\nu _\mu G^{\mu \nu }]+ig\mathrm{Tr}\left[\mathrm{\Omega }_\nu \left(D^\nu \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }(D^\nu \mathrm{\Phi })^{}\right)\right]`$
$`+ig\mathrm{Tr}\left[\mathrm{\Omega }_\nu [G^{\mu \nu },W_\mu ]\right]`$
$`0`$ $`=`$ $`\mathrm{Tr}\left[\mathrm{\Omega }_\nu \left(𝒟_\mu G^{\mu \nu }+ig(D^\nu \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }(D^\nu \mathrm{\Phi })^{})\right)\right](\mathrm{\Omega }_\nu 𝒜)`$ (97)
$`𝒟^\mu (G_{\mu \nu })`$ $`=`$ $`\mathrm{\Pi }_𝒜\left[ig\left(D_\nu \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }(D_\nu \mathrm{\Phi })^{}\right)\right]=\mathrm{\Pi }_𝒜(J_\nu )`$ (98)
$`J_\mu `$ $`=`$ $`ig\left(_\mu \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }_\mu \mathrm{\Phi }^{}+ig\{W_\mu ,\mathrm{\Phi }\mathrm{\Phi }^{}\}\right)`$ (99)
$`_\mu {\displaystyle \frac{}{(_\mu \mathrm{\Phi }^{})}}{\displaystyle \frac{}{(\mathrm{\Phi }^{})}}=00`$ $`=`$ $`(D_\mu D^\mu +m^2)\mathrm{\Phi }`$ (100)
$`=`$ $`(\mathrm{}+m^2)\mathrm{\Phi }+2igW_\alpha ^\alpha \mathrm{\Phi }+ig(_\alpha W^\alpha )\mathrm{\Phi }g^2W_\mu W^\mu \mathrm{\Phi }`$
We shall note that in Eq. (97), the fact that the equation is only valid for $`\mathrm{\Omega }_\nu 𝒜`$ is very important. It comes from the fact that in the variationnal principle leading to the Euler-Lagrange equations, the variation of the gauge field ($`\mathrm{\Omega }_\nu `$) must lie in the Lie algebra also. If this equation were valid for any matrix $`\mathrm{\Omega }_\nu `$, then we would have $`𝒟^\alpha (G_{\alpha \beta })`$ (which is in $`𝒜`$) equal to $`ig\left(D_\beta \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }(D_\beta \mathrm{\Phi })^{}\right)`$, which is not necessarily in $`𝒜`$, and this is why there is this projection operator on the Lie algebra $`\mathrm{\Pi }_𝒜`$ in Eq. 98. For $`su(N)`$ algebras, this projection is simply $`MM\mathrm{Tr}(M)\frac{I}{N}`$, where $`I`$ stands for the identity matrix. Eq. 98 and Eq. 100 are the equations of motion respectively for the gauge fields and for the scalar fields, which can be related to the abelian equations of Eq. 4 and Eq. 3. The non-abelian equivalent of the current is now extracted from Eq. 98 and is given by the matrix: $`J_\nu =ig\left(D_\nu \mathrm{\Phi }\mathrm{\Phi }^{}\mathrm{\Phi }(D_\nu \mathrm{\Phi })^{}\right)`$
Contrary to the abelian case, this current is not gauge invariant anymaore but rather gauge covariant, that is $`J_\nu =UJ_{}^{}{}_{\nu }{}^{}U^1`$ under a gauge transformation.
We now operate as in the previous section, and observe that if we compute $`(\text{100})\mathrm{\Phi }^{}\mathrm{\Phi }(\text{100})^{}`$, we get:
$`0`$ $`=`$ $`(D_\mu D^\mu \mathrm{\Phi })\mathrm{\Phi }^{}\mathrm{\Phi }(D_\mu D^\mu \mathrm{\Phi })^{}\mathrm{\Phi }`$ (101)
$`=`$ $`𝒟_\mu \left((D^\mu \mathrm{\Phi })\mathrm{\Phi }^{}\mathrm{\Phi }(D^\mu \mathrm{\Phi })^{}\right)`$
$`0`$ $`=`$ $`𝒟_\mu \left(J^\mu \right)`$ (102)
If one projects this equation on the Lie algebra, the resulting equation is redundant with Eq. 98 on which we apply the operator $`𝒟^\nu `$. Like in the abelian case, we find a redundancy, but it is important to note at this stage that eq. 102 is a stronger condition than if we just applied $`𝒟^\nu `$ on Eq. 98. It seems that we missed some degrees of freedom in Eq. 98. The fundamental structure of the gauge group is responsible for this fact. For instance, in the case of a $`u(N)`$ algebra, $`\mathrm{\Pi }_𝒜(M)=M`$ if M is hermitian, and all the “degrees of freedom” of $`J_\mu `$ are concerned with this redundancy between the equation for the matter and the equation for the gauge field.
We therefore have too much information in the set of equations of the matter field and one should replace eq. (100) by $`(\text{100})\mathrm{\Phi }^{}+\mathrm{\Phi }(\text{100})^{}`$, i.e.:
$`0`$ $`=`$ $`(D_\mu D^\mu \mathrm{\Phi })\mathrm{\Phi }^{}+\mathrm{\Phi }(D_\mu D^\mu \mathrm{\Phi })^{}+2m^2\mathrm{\Phi }\mathrm{\Phi }^{}`$ (103)
$`=`$ $`𝒟_\mu \left((D^\mu \mathrm{\Phi })\mathrm{\Phi }^{}+\mathrm{\Phi }(D^\mu \mathrm{\Phi })^{}\right)2D_\mu \mathrm{\Phi }(D^\mu \mathrm{\Phi })^{}+2m^2\mathrm{\Phi }\mathrm{\Phi }^{}`$ (104)
$`=`$ $`\left(𝒟_\mu 𝒟^\mu (\mathrm{\Phi }\mathrm{\Phi }^{})2(D_\mu \mathrm{\Phi })(D^\mu \mathrm{\Phi })^{}+2m^2\mathrm{\Phi }\mathrm{\Phi }^{}\right)`$ (105)
### 3.2 The gauge invariant variables
The procedure used to obtain Eq. 8 consists in eliminating the two first terms of $`\frac{J_\mu }{ie}=\phi ^{}_\mu \phi _\mu \phi ^{}\phi +2ie\rho A_\mu `$ in order to extract the gauge field. We have $`\frac{J_\mu }{ie\rho }=2ieA_\mu +_\mu \mathrm{\Lambda }`$ and the pure gauge term disappears in $`F_{\mu \nu }`$. But in our case we have a matrix and this procedure does not work. However, the extraction of $`A_\mu `$ can be seen in another way. In the abelian case, we could also have taken a unitary gauge, that is to say a gauge in which $`\phi `$ is real. This automatically eliminates the desired terms. We may proceed here in a similar way. The essential hypothesis is that any two scalar fields $`\mathrm{\Phi }`$ and $`\mathrm{\Psi }_0`$ can be related by an element of the gauge group. It is the case for $`U(N)`$ or $`SU(N)`$. Thus, the central point of the method is to choose a constant unitary vector $`\mathrm{\Psi }_0`$, and therefore one can find $`U`$ in the gauge group such that:
$`\mathrm{\Phi }`$ $`=`$ $`zU\mathrm{\Psi }_0`$ (106)
$`z`$ $`=`$ $`\sqrt{\mathrm{\Phi }^{}\mathrm{\Phi }}=\sqrt{\rho }`$ (107)
A consequence is that if $`W_{}^{}{}_{\mu }{}^{}`$ is the gauge field in the “unitary” gauge obtained by the matrix $`U`$ we have from Eq. 99:
$`J_{}^{}{}_{\mu }{}^{}=U^1J_\mu U`$ $`=`$ $`ig\left(z_\mu (z)\mathrm{\Psi }_0\mathrm{\Psi }_0^{}\mathrm{\Psi }_0\mathrm{\Psi }_0^{}z_\mu (z)+ig\rho \{W_{}^{}{}_{\mu }{}^{},\mathrm{\Psi }_0\mathrm{\Psi }_0^{}\}\right)`$ (108)
$`=`$ $`(ig)^2\rho \{W_{}^{}{}_{\mu }{}^{},\mathrm{\Psi }_0\mathrm{\Psi }_0^{}\}`$ (109)
However, it is in general impossible to reconstruct the entire gauge field $`W_{}^{}{}_{\mu }{}^{}`$ from this equation, except for the $`SU(2)`$ case because of the relation $`\{\sigma ^i,\sigma ^j\}=2\delta ^{ij}`$ (and this anticommutator has no residue lying in the Lie algebra). Using this property, the traceless part of $`J_{}^{}{}_{\mu }{}^{}`$ gives $`(ig)^2\rho W_{}^{}{}_{\mu }{}^{}`$. Since it works only for $`SU(2)`$, we need to find a way to get the missing degrees of freedom of the gauge field. The method consists in constructing an orthonormal basis of $`C^N`$, starting from $`\mathrm{\Psi }_0`$: $`(\mathrm{\Psi }_0,\mathrm{\Psi }_1,\mathrm{},\mathrm{\Psi }_{N1})`$, which does not depend on space-time coordinates. If we set $`\mathrm{\Phi }_k=zU\mathrm{\Psi }_k`$ ($`k1`$), then $`(\rho ^1\mathrm{\Phi },\rho ^1\mathrm{\Phi }_1,\mathrm{},\rho ^1\mathrm{\Phi }_{N1})`$ forms also an orthonormal basis. A gauge transformation will naturally apply also to these new scalar fields, and we consider the gauge invariant variables:
$`J_{mn\mu }`$ $`=`$ $`ig\left(\mathrm{\Phi }_m^{}D_\mu \mathrm{\Phi }_n(D_\mu \mathrm{\Phi }_m)^{}\mathrm{\Phi }_n\right)`$ (110)
The simple reason why we do not consider some other gauge invariant variables, by taking the sum of the two terms above instead of their difference is that $`\mathrm{\Phi }_m^{}D_\mu \mathrm{\Phi }_n+(D_\mu \mathrm{\Phi }_m)^{}\mathrm{\Phi }_n=_\mu (\mathrm{\Phi }_m^{}\mathrm{\Phi }_n)=_\mu (\rho \delta _{m,n})`$ and thus they can be expressed using the gauge invariant variable $`z=\sqrt{\rho }`$. In the unitary gauge, these gauge invariant variables allow to reconstruct the gauge field completely:
$`J_{mn\mu }`$ $`=`$ $`2(ig)^2\rho \mathrm{\Psi }_m^{}W_{}^{}{}_{\mu }{}^{}\mathrm{\Psi }_n(J_{nm\mu }=J_{mn\mu }^{})`$ (111)
$`v_{mn\mu }`$ $`=`$ $`{\displaystyle \frac{1}{2g^2\rho }}J_{mn\mu }`$ (112)
$`W_{}^{}{}_{\mu }{}^{}`$ $`=`$ $`{\displaystyle \underset{m,n}{}}v_{mn\mu }\mathrm{\Psi }_n\mathrm{\Psi }_m^{}`$ (113)
The equations of motion for the gauge field in Eq. 98 can then be rewritten in the unitary gauge (note that $`\mathrm{\Pi }_𝒜(U^1JU)=U^1\mathrm{\Pi }_𝒜(J)U`$)
$`G_{}^{}{}_{\mu \nu }{}^{}`$ $`=`$ $`{\displaystyle \underset{m,n}{}}\left(_\mu \left(v_{mn\nu }\right)_\nu \left(v_{mn\mu }\right)+ig{\displaystyle \underset{k}{}}(v_{mk\mu }v_{kn\nu }v_{mk\nu }v_{kn\mu })\right)\mathrm{\Psi }_n\mathrm{\Psi }_m^{}`$ (114)
$`𝒟_{}^{}{}_{}{}^{\mu }(G_{}^{}{}_{\mu \nu }{}^{})`$ $`=`$ $`^\mu G_{}^{}{}_{\mu \nu }{}^{}+ig[W_{}^{}{}_{\mu }{}^{},G_{}^{}{}_{\mu \nu }{}^{}]`$ (116)
$`=`$ $`{\displaystyle \underset{m,n}{}}\left(\mathrm{}\left(v_{mn\nu }\right)_\nu \left(v_{mn}\right)+ig{\displaystyle \underset{k}{}}^\mu (v_{mk\mu }v_{kn\nu }v_{mk\nu }v_{kn\mu })\right)\mathrm{\Psi }_n\mathrm{\Psi }_m^{}`$
$`+ig{\displaystyle \underset{m,n}{}}\mathrm{\Psi }_n\mathrm{\Psi }_m^{}[{\displaystyle \underset{l}{}}v_{ml}^\mu (_\mu \left(v_{ln\nu }\right)_\nu \left(v_{ln\mu }\right)+ig{\displaystyle \underset{k}{}}(v_{lk\mu }v_{kn\nu }v_{lk\nu }v_{kn\mu }))`$
$`(_\mu \left(v_{ml\nu }\right)_\nu \left(v_{ml\mu }\right)+ig{\displaystyle \underset{k}{}}(v_{mk\mu }v_{kl\nu }v_{mk\nu }v_{kl\mu }))v_{ln}^\mu ]`$
$`=`$ $`g^2z^2{\displaystyle \underset{m}{}}\mathrm{\Pi }_𝒜\left(v_{0m\mu }\mathrm{\Psi }_m\mathrm{\Psi }_0^{}+v_{m0\mu }\mathrm{\Psi }_0\mathrm{\Psi }_m^{}\right)`$ (117)
$`=`$ $`g^2z^2{\displaystyle \underset{m}{}}\left(v_{0m\mu }\mathrm{\Psi }_m\mathrm{\Psi }_0^{}+v_{m0\mu }\mathrm{\Psi }_0\mathrm{\Psi }_m^{}2{\displaystyle \frac{v_{00\mu }}{N}}\mathrm{\Psi }_m\mathrm{\Psi }_m^{}\right)`$ (118)
The last equality is only valid for $`SU(N)`$. One must adapt this formula for another gauge group. Projecting these equations on the basis of matrices $`\mathrm{\Psi }_m\mathrm{\Psi }_n^{}`$ leads to a large set of $`N^2`$ equations in which only gauge invariant variables are present. In the $`SU(N)`$ case we can also separate these equations into four different classes depending on the indices m and n, because of the specific form of the current matrix projected on the Lie algebra. The four cases correspond to the diagonal case with indices in the form $`(m,m)`$ ($`m>0`$), the case with indices in the form $`(0,m)`$ or $`(m,0)`$ ($`m>0`$), and finally the case where $`m=n=0`$. The projection on these different cases can be easily done and we will not present them here. It is clear that the gauge fields $`W_{}^{}{}_{\mu }{}^{}`$ expressed in the basis of the $`\mathrm{\Psi }_k`$’s is nothing but the matrix composed of the gauge invariant coefficients $`(v_{m,n\mu })`$. Of course, these coefficients depend on the constant basis we choose, but physical solutions must be independent of this choice. It remains to demonstrate that these equations of motion can be re-expressed using only variables that are also independent from the constant basis chosen: we can consider some objects of the form $`\mathrm{Tr}[(W_{}^{}{}_{\mu }{}^{})^n]`$, or equivalently the characteristic polynomial of $`W_{}^{}{}_{\mu }{}^{}`$. We expect to have new results in the near future.
We may conclude this last section with the equation of motion for the matter fields. The simplest way is to look at the lagrangian and to use the following equality:
$$\frac{1}{g^2}\underset{m,n}{}J_{mn\mu }J_{nm}^\mu =N_\mu \rho ^\mu \rho 4\rho (D_\mu \mathrm{\Phi })^{}(D^\mu \mathrm{\Phi })$$
(119)
The matter part of the lagrangian can then be written:
$$_0=N_\mu z^\mu zm^2z^2+g^2z^2\underset{m,n}{}v_{mn\mu }v_{nm}^\mu $$
(120)
And the equation of motion for the scalar field is finally:
$$\left(\mathrm{}+\frac{m^2}{N}\right)z=\frac{g^2}{N}z\underset{m,n}{}v_{mn\mu }v_{nm}^\mu $$
(121)
## 4 Conclusion
In this paper, we give a certain number of results which are really encouraging for the purpose of reformulating gauge theories using only gauge-invariant variables. Within the prospects of this work, a short-term project would naturally be to find an equivalent formulation when fermions are involved. Then, the quantization of the theory has to be constructed. Within this subtopic, it would be interesting to revisit the general formalism of quantization in QFT. An equation like $`\frac{dA}{dt}=i[H,A]`$ is a very old non-relativistic formula which is surprisingly still used in textbooks about relativistic quantum field theory. Instead of the Hamiltonian, one would naturally consider an operator of the form $`_\mathrm{\Sigma }𝑑\sigma ^\mu T_{\mu \nu }`$ in order to quantize a theory. This has not been done yet and one of the possible reasons is that there is no unique expression for the energy-momentum tensor $`T_{\mu \nu }`$. There are some current research activities on this topic , in order to find the “best” criteria to define uniquely $`T_{\mu \nu }`$. So far, it seems that the Belinfante tensor is a good candidate, since it is gauge-invariant. Therefore it can be naturally inserted in the formalism presented in this paper. Finally, in the long-term we hope to be able to compute some scattering cross sections using directly gauge invariant variables, and also to provide a revised version of Quantum Field Theory which would apply to unstable particles and more generally, to physical systems that evolve on a “long-time” scale, (CP violation, neutrino oscillations,…) as mentionned in the introduction.
## 5 Appendices
### 5.1 Review of basic cohomological formulas
As noted in this paper, one of the main problems regarding gauge independence is to have a method to find the set of gauge fields with a given Field-Strength tensor $`F^{\mu \nu }`$. We will separate the abelian case from the non-abelian one, because the curvature tensor depends linearly on the gauge field in the abelian case, quadratically in the latter case. Linearity is lost in the non-abelian case, which renders the problem much more complicated.
The problem can be summarized as follows: if one has a specific tensor F of rank $`n`$, we look for another tensor of rank $`n1`$ such that $`F=dA`$ where $`d`$ represents the exterior derivative. The tensor $`F`$ must obey $`dF=0`$ because of the property $`d^2=0`$. So we want to find $`A`$ from a given $`F`$, assumed that $`F`$ is a closed form (i.e. $`dF=0`$). Given a solution $`A`$, one can find another solution $`A^{}`$ by adding to $`A`$ any term of the form $`d\mathrm{\Lambda }`$, again because $`d^2=0`$. Therefore, we will say that two tensors of rank $`n1`$ are co-homologous if there exists $`\mathrm{\Lambda }`$ such that $`AA^{}=d\mathrm{\Lambda }`$. It is an equivalence relation and the equivalence classes are called cohomology classes (for the de-Rahm cohomology, and we will further explain why it is important to make this distinction when the non-abelian case is involved).
### 5.2 Abelian gauge fields
Let $`M=R^4`$ be the Minkowsky space-time, and consider $`X^\mu (u,x)`$ an application from $`[0,1]\times M`$ into $`M`$ such that:
$`xM,X^\mu (0,x)`$ $`=`$ $`x_0^\mu `$ (122)
$`xM,X^\mu (1,x)`$ $`=`$ $`x^\mu `$ (123)
We also assume that $`X^\mu `$ is infinitely smooth. It is then called a “contraction”. The reader will recover the standard Poincaré formula by taking $`X^\mu (u,x)=ux^\mu `$. Suppose $`A^\mu (x)`$ is a vector field with vanishing curvature, then if we define $`V(x)`$ as follows:
$`V(x)`$ $`=`$ $`{\displaystyle _0^1}𝑑u{\displaystyle \frac{X^\mu }{u}}A_\mu (X(u,x))`$ (124)
$`Then_\mu V`$ $`=`$ $`A_\mu (x){\displaystyle _0^1}𝑑u{\displaystyle \frac{X^\alpha }{u}}{\displaystyle \frac{X^\beta }{x^\mu }}F_{\alpha \beta }(X)`$ (125)
Therefore, if the curvature of $`A`$ vanishes, $`V(x)`$ is a possible solution for the potential. Also, if one replaces explicitly $`A^\mu `$ by $`^\mu V^{}`$ in eq. 124, one gets $`V^{}(x)V^{}(x_0)`$, and not $`V^{}(x)`$. $`V(x)`$ is therefore not a “fixed point solution” of an integral equation, but can be defined as the solution for which $`V(x_0)=0`$. The rest in the expression of $`_\mu V`$ vanishes explicitly for a vanishing curvature, but when the curvature is not $`0`$, this formula provides us with an explicit expression for $`A^\mu `$ as a function of $`F^{\mu \nu }`$ up to a gauge transformation by $`^\mu V`$. Thus we have already the next step, and if we consider a given field-strength tensor $`F^{\mu \nu }`$, we can define the following vector field:
$$A^\mu (x)=_0^1𝑑u\frac{X^\alpha }{u}\frac{X^\beta }{x^\mu }F_{\alpha \beta }(X(u,x))$$
(126)
Then, with this definition we have:
$$_\mu A_\nu _\nu A_\mu =F_{\mu \nu }(x)_0^1𝑑u\frac{X^\alpha }{u}\frac{X^\beta }{x^\mu }\frac{X^\gamma }{x^\nu }(_\alpha F_{\beta \gamma }+_\beta F_{\gamma \alpha }+_\gamma F_{\alpha \beta })$$
(127)
The last term vanishes if $`dF=0`$, and we recognize here the homogeneous Maxwell equations. In this case, the expression we have chosen for $`A^\mu `$ is a possible gauge field, and this formula is of course very important because it allows us to “parameterize” the orbits of gauge fields. It is possible to go on with this scheme, and for a given 3-form $`\omega _{\alpha \beta \gamma }`$ we can define $`F^{\mu \nu }`$ using:
$$F_{\mu \nu }=_0^1\frac{X^\alpha }{u}\frac{X^\beta }{x^\mu }\frac{X^\gamma }{x^\nu }\omega _{\alpha \beta \gamma }(X(u,x))$$
(128)
and when $`d\omega =0`$, we have $`_\alpha F_{\beta \gamma }+_\beta F_{\gamma \alpha }+_\gamma F_{\alpha \beta }=\omega _{\alpha \beta \gamma }(x)`$, and so on (but there is actually only one next step because we have assumed here that we are in four space-time dimensions and any four form is proportional to the Levi-Civita pseudo-tensor).
To summarize, given an $`n`$ form $`F`$ such that $`dF=0`$, we have been able to exhibit a $`n1`$ form $`A`$ such that $`F=dA`$. This element $`A`$ can be interpreted as an element of an equivalent class of cohomology with a given curvature. In other words, we have “computed” the cohomology. Expressed this way, it looks simple but hides the real difficulties, which are of a topological nature. In all these calculations, we have assumed the existence of $`X^\mu `$, which imposes some constraints on the topology of the four dimensional space-time. If the whole Minkowsky space is taken under consideration, no topological problem occurs, and more generally, this is true if we consider a simply connected space. Then, one can find $`X^\mu `$ and proceed to the previous calculations.
### 5.3 Conventions for the Non-abelian case
$`iA`$ and $`iB`$ are supposed to lie in the real Lie algebra corresponding to the Lie Group $`𝒢`$, which is a subgroup of $`U(N)`$ here. Therefore $`A`$ and $`B`$ are hermitian. We set $`A=UA^{}U^1`$. $`X`$ and $`Y`$ are vectors lying in the same representation as the matter field $`\mathrm{\Phi }`$.
$`\mathrm{\Phi }`$ $`=`$ $`U\mathrm{\Phi }^{}=e^{iT}\mathrm{\Phi }^{}(Tsmall)`$ (129)
$`W_\mu `$ $`=`$ $`UW_{}^{}{}_{\mu }{}^{}U^1+{\displaystyle \frac{i}{g}}_\mu (U)U^1W_{}^{}{}_{\mu }{}^{}=U^1W_\mu U{\displaystyle \frac{i}{g}}U^1_\mu (U)`$ (130)
$`\delta W_\mu =W_{}^{}{}_{\mu }{}^{}W_\mu `$ $`=`$ $`{\displaystyle \frac{i}{g}}𝒟_{}^{}{}_{\mu }{}^{}(U)U^1={\displaystyle \frac{i}{g}}U^1𝒟_\mu (U)`$ (131)
$`UW_{}^{}{}_{\nu }{}^{}U^1W_{}^{}{}_{\nu }{}^{}+{\displaystyle \frac{i}{g}}(_\nu U)U^1`$ $`=`$ $`e^{iT}W_{}^{}{}_{\nu }{}^{}e^{iT}W_{}^{}{}_{\nu }{}^{}+{\displaystyle \frac{i}{g}}(_\nu e^{iT})e^{iT}`$ (132)
$``$ $`[iT,W_{}^{}{}_{\nu }{}^{}]{\displaystyle \frac{1}{g}}(_\nu T)={\displaystyle \frac{1}{g}}𝒟_\nu (A)`$ (133)
$`D_\mu \mathrm{\Phi }`$ $`=`$ $`(_\mu +igW_\mu )\mathrm{\Phi }D_\mu \mathrm{\Phi }=D_\mu (U\mathrm{\Phi }^{})=UD_{}^{}{}_{\mu }{}^{}\mathrm{\Phi }^{}`$ (134)
$`𝒟_\mu (A)`$ $`=`$ $`_\mu A+ig[W_\mu ,A]𝒟_\mu (UA^{}U^1)=U(𝒟_{}^{}{}_{\mu }{}^{}A^{})U^1`$ (135)
$`𝒟_\mu (AB)`$ $`=`$ $`𝒟_\mu (A)B+A𝒟_\mu (B)`$ (136)
$`𝒟_\mu (XY^{})`$ $`=`$ $`D_\mu (X)Y^{}+X(D_\mu (Y))^{}`$ (137)
$`D_\mu (AX)`$ $`=`$ $`𝒟_\mu (A)X+AD_\mu (X)`$ (138)
$`[D_\mu ,D_\nu ]\mathrm{\Phi }`$ $`=`$ $`ig(_\mu W_\nu _\nu W_\mu +ig[W_\mu ,W_\nu ])\mathrm{\Phi }=igG_{\mu \nu }\mathrm{\Phi }`$ (139)
$`G_{\mu \nu }`$ $`=`$ $`UG_{}^{}{}_{\mu \nu }{}^{}U^1`$ (140)
$`[𝒟_\alpha ,𝒟_\beta ](A)`$ $`=`$ $`ig[G_{\alpha \beta },A]`$ (141)
$`0`$ $`=`$ $`[D_\nu ,[D_\rho ,D_\sigma ]]\mathrm{\Phi }+[D_\rho ,[D_\sigma ,D_\nu ]]\mathrm{\Phi }+[D_\sigma ,[D_\nu ,D_\rho ]]\mathrm{\Phi }`$ (142)
$`0`$ $`=`$ $`\epsilon ^{\mu \nu \rho \sigma }[D_\nu ,G_{\rho \sigma }](\mathrm{\Phi })(\mathrm{\Phi })`$ (143)
$`0`$ $`=`$ $`𝒟_\nu (\stackrel{~}{G}^{\mu \nu })(Bianchi)`$ (144)
For $`SU(N)`$ gauge groups, it may be useful to use the relation:
$$\mathrm{\Phi }\mathrm{\Phi }^{}=\mathrm{\Phi }^{}\mathrm{\Phi }\frac{1}{N}I+A_\varphi $$
(145)
where $`I`$ stands for the identity matrix in $`N`$ dimensions, $`A_\mathrm{\Phi }`$ lies therefore in the Lie algebra $`su(N)`$, and we will conveniently denote by $`\rho _\mathrm{\Phi }=\mathrm{\Phi }^{}\mathrm{\Phi }`$ the probability density of $`\mathrm{\Phi }`$.
### 5.4 Non abelian case and the Path Ordered Exponential
If $`A`$ is an operator valued function of the real variable $`\lambda `$, a solution to the differential equation $`f^{}(\lambda )=A(\lambda )f(\lambda )`$ is given by (see ):
$`f(x)`$ $`=`$ $`[1+{\displaystyle _0^x}d\lambda A(\lambda )+{\displaystyle _0^x}d\lambda _1A(\lambda _1){\displaystyle _0^{\lambda _1}}d\lambda _2A(\lambda _2)+\mathrm{}`$ (146)
$`+{\displaystyle _0^x}d\lambda _1A(\lambda _1)\mathrm{}{\displaystyle _0^{\lambda _{n1}}}d\lambda _nA(\lambda _n)]f(0)`$
$`=`$ $`\underset{}{𝑒}^{_0^x𝑑\lambda A(\lambda )}f(0)=\underset{ds0}{lim}{\displaystyle \underset{k=n}{\overset{1}{}}}e^{A(s_k)ds}f(0)withs_k={\displaystyle \frac{k\times x}{n}}`$ (147)
$`\underset{}{𝑒}^{_0^x𝑑\lambda A(\lambda )}`$ $`=`$ $`\underset{}{𝑒}^{_y^x𝑑\lambda A(\lambda )}\underset{}{𝑒}^{_0^y𝑑\lambda A(\lambda )}`$ (148)
$`{\displaystyle \frac{d}{ds}}\underset{}{𝑒}^{_0^sF(v)𝑑v}`$ $`=`$ $`F(s)\underset{}{𝑒}^{_0^sF(v)𝑑v}{\displaystyle \frac{d}{ds}}\underset{}{𝑒}^{_s^1F(v)𝑑v}=\underset{}{𝑒}^{_s^1F(v)𝑑v}F(s)`$ (149)
Note that the product in Eq. 147 is done “from right to left”. In the following, we list a few properties of the path order exponential:
$`\left(\underset{}{𝑒}^{\scriptscriptstyle A}\right)^1=\underset{}{𝑒}^{\scriptscriptstyle A}`$ $`=`$ $`1{\displaystyle _0^1}A(u)𝑑u+{\displaystyle _0^1}𝑑u_1{\displaystyle _0^{u_1}}𝑑u_2A(u_2)A(u_1)+\mathrm{}`$ (150)
$`\underset{}{𝑒}^{_a^xA^{}(s)A^1(s)𝑑s}`$ $`=`$ $`A(x)A^1(a)`$ (151)
$`P(x)=\underset{}{𝑒}^{_a^xA(s)𝑑s}\underset{}{𝑒}^{_a^xA(s)+B(s)ds}`$ $`=`$ $`P(x)\underset{}{𝑒}^{_a^xP^1(s)B(s)P(s)𝑑s}`$ (152)
$`A(x)\underset{}{𝑒}^{_a^xB(s)𝑑s}A^1(a)`$ $`=`$ $`\underset{}{𝑒}^{_a^x\left(A^{}(s)A^1(s)+A(s)B(s)A^1(s)\right)𝑑s}`$ (153)
$`{\displaystyle \frac{}{\lambda }}\underset{}{𝑒}^{_a^bA(u,\lambda )𝑑u}`$ $`=`$ $`{\displaystyle _a^b}𝑑s\underset{}{𝑒}^{_s^bA(u,\lambda )𝑑u}{\displaystyle \frac{A(s,\lambda )}{\lambda }}\underset{}{𝑒}^{_a^sA(u,\lambda )𝑑u}`$ (154)
The last formula can be demonstrated easily if one uses the product form of the ordered exponential (Eq. 147)
#### 5.4.1 Introduction of a space-time contraction
If we now consider a contraction $`X_\mu (u,x)`$ where $`X_\mu (0,x)=x_0`$ and $`X_\mu (1,x)=x`$ (see Eq. 123), we obtain the following definition:
$`F(u,x)`$ $`=`$ $`{\displaystyle \frac{X_\mu }{u}}(u,x)A^\mu (X_\mu (u,x))`$ (155)
$`_\mu X_\alpha A^\alpha (X_\mu (u,x))|_{u=1}`$ $`=`$ $`A_\mu (x)`$ (156)
$`f(x)=\underset{}{𝑒}^{ig_\gamma 𝑑l^\mu A_\mu }(x)`$ $`=`$ $`1+(ig){\displaystyle _0^1}𝑑uF(u,x)+(ig)^2{\displaystyle _0^1}𝑑u_1F(u_1,x){\displaystyle _0^{u_1}}𝑑u_2F(u_2,x)+\mathrm{}`$ (157)
$`+(ig)^n{\displaystyle _0^1}𝑑u_1F(u_1,x)\mathrm{}{\displaystyle _0^{u_{n1}}}𝑑u_nF(u_n,x)+\mathrm{}`$
$`=`$ $`{\displaystyle \underset{k}{}}(ig)^k{\displaystyle _{[0;1]^k}}𝑑u_1\mathrm{}𝑑u_k\theta (u_1,\mathrm{},u_k)F(u_1,x)\mathrm{}F(u_k,x)`$ (158)
$`=`$ $`\mathrm{exp}\left(ig{\displaystyle _0^1}𝑑uF(u,x)\right)(if[A(x),A(x^{})]=0)`$ (159)
$`\theta (u_1,\mathrm{},u_k)`$ $`=`$ $`1iffu_1u_2\mathrm{}u_k,0ifnot`$ (160)
$`=`$ $`H(u_1u_2)H(u_2u_3)\mathrm{}H(u_{k1}u_k)`$ (161)
Each term in the sum can be obtain by the following recursion:
$`J_0(a,b,x)`$ $`=`$ $`1`$ (162)
$`J_n(a,b,x)`$ $`=`$ $`{\displaystyle _a^b}𝑑sF(s,x)J_{n1}(a,s)`$ (163)
$`J_n(a,a,x)`$ $`=`$ $`0n,x`$ (164)
$`J_n(a,b,x)`$ $`=`$ $`{\displaystyle _a^b}𝑑s_sX_\mu A^\mu (X_\mu (s,x))J_{n1}(a,s)`$ (165)
Let $`\mathrm{\Phi }`$ be a solution (if it exists) to the system of PDE $`_\mu \mathrm{\Phi }=igW_\mu (x)\mathrm{\Phi }`$, then:
$`{\displaystyle \frac{X^\mu }{u}}_\mu \mathrm{\Phi }`$ $`=`$ $`ig{\displaystyle \frac{X^\mu }{u}}W_\mu (x)\mathrm{\Phi }`$ (167)
$`{\displaystyle \frac{d}{du}}\mathrm{\Phi }(X(u,x))`$ $`=`$ $`F(u,x)\mathrm{\Phi }(X(u,x))`$ (168)
$`\mathrm{\Phi }(X(u,x))`$ $`=`$ $`\underset{}{𝑒}^{_0^u𝑑vF(v,x)}\mathrm{\Phi }_0`$ (169)
$`\mathrm{\Phi }(x)=\mathrm{\Phi }(X(1,x))`$ $`=`$ $`\underset{}{𝑒}^{_0^1𝑑vF(v,x)}\mathrm{\Phi }_0`$ (170)
$`=`$ $`\underset{}{𝑒}^{ig_0^1𝑑u\frac{X^\mu }{u}W_\mu (X(u,x))}\mathrm{\Phi }_0`$ (171)
If $`\mathrm{\Phi }`$ is a square matrix and $`\mathrm{\Phi }_0=I`$, then $`\mathrm{\Phi }`$ is invertible because $`det(\mathrm{\Phi })=e^{ig{\scriptscriptstyle \mathrm{Tr}F}}0`$, thus $`W_\mu =\frac{i}{g}_\mu \mathrm{\Phi }\mathrm{\Phi }^1`$ which is a right invariant form, the curvature of which vanishes. It is not surprising to get such a constraint. Already in the abelian case, if $`\varphi =e^{ig{\scriptscriptstyle A}}`$ then $`_\mu \varphi =ig\left(A_\mu +_uX^\alpha _\mu X^\beta F_{\alpha \beta }\right)\varphi `$ (see Eq. 125) and we explicitly show the presence of a curvature term as an obstacle to solve the system of differential equations. To obtain a similar formula in the non-abelian case, let us take the partial derivatives of Eq. 171. We get:
$`{\displaystyle \frac{i}{g}}_\mu \mathrm{\Phi }`$ $`=`$ $`{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_\mu \left({\displaystyle \frac{X^\nu }{s}}W_\nu (X(s,x))\right)\underset{}{𝑒}^{ig_0^sW_x}`$ (172)
$`=`$ $`{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}\left(_\mu _sX^\nu W_\nu (X(s,x))+_sX^\nu _\mu X^\rho _\rho W_\nu (X(s,x))\right)\underset{}{𝑒}^{ig_0^sW_x}`$ (178)
$`=`$ $`{\displaystyle _0^1}ds\underset{}{𝑒}^{ig_s^1W_x}\{_\mu _sX^\nu W_\nu (X)+_uX^\nu _\mu X^\rho G_{\rho \nu }(X)`$
$`+_sX^\nu _\mu X^\rho (_\nu W_\rho (X)ig[W_\rho ,W_\nu ])\}\underset{}{𝑒}^{ig_0^sW_x}`$
$`=`$ $`{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_s\left(_\mu X^\nu W_\nu (X)\right)\underset{}{𝑒}^{ig_0^sW_x}`$
$`+{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}\{ig_sX^\nu _\mu X^\rho [W_\rho ,W_\nu ]+_sX^\nu _\mu X^\rho G_{\rho \nu }\}\underset{}{𝑒}^{ig_0^sW_x}`$
$`=`$ $`{\displaystyle _0^1}ds_s\left(\underset{}{𝑒}^{ig_s^1W_x}_\mu X^\nu W_\nu (X)\underset{}{𝑒}^{ig_0^sW_x}\right)`$
$`{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}(ig_sX^\nu W_\nu (X))_\mu X^\rho W_\rho (X)\underset{}{𝑒}^{ig_0^sW_x}`$
$`+{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_\mu X^\rho W_\rho (X)(ig_sX^\nu W_\nu (X))\underset{}{𝑒}^{ig_0^sW_x}`$
$`+{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}\{ig_sX^\nu _\mu X^\rho [W_\rho ,W_\nu ]+_sX^\nu _\mu X^\rho G_{\rho \nu }\}\underset{}{𝑒}^{ig_0^sW_x}`$
$`=`$ $`W_\mu (x)\underset{}{𝑒}^{ig_0^1W_x}0+{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_sX^\nu _\mu X^\rho G_{\rho \nu }\underset{}{𝑒}^{ig_0^sW_x}`$ (179)
$`=`$ $`W_\mu (x)\underset{}{𝑒}^{ig_0^1W_x}{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_sX^\nu _\mu X^\rho G_{\nu \rho }\underset{}{𝑒}^{ig_0^sW_x}`$ (180)
where Eq. 178 and Eq. 178 make use of Eq. 149. The result of Eq. 180 is nothing but the non-abelian equivalent of Eq. 125, and it can be interesting to rewrite it as follows:
$`W_\mu (x)`$ $`=`$ $`{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_sX^\nu _\mu X^\rho G_{\nu \rho }\underset{}{𝑒}^{ig_0^sW_x}\left(\underset{}{𝑒}^{ig_0^1W_x}\right)^1+{\displaystyle \frac{i}{g}}_\mu \mathrm{\Phi }\mathrm{\Phi }^1`$ (181)
This expression gives $`W_\mu (x)`$ as a gauge equivalent of (see Eq. 130):
$`W_{}^{}{}_{\mu }{}^{}`$ $`=`$ $`\left(\underset{}{𝑒}^{ig_0^1W_x}\right)^1{\displaystyle _0^1}𝑑s\underset{}{𝑒}^{ig_s^1W_x}_sX^\nu _\mu X^\rho G_{\nu \rho }\underset{}{𝑒}^{ig_0^sW_x}`$ (182)
$`=`$ $`{\displaystyle _0^1}𝑑s\left(\underset{}{𝑒}^{ig_0^sW_x}\right)^1_sX^\nu _\mu X^\rho G_{\nu \rho }\left(\underset{}{𝑒}^{ig_0^sW_x}\right)`$ |
warning/0001/hep-ph0001169.html | ar5iv | text | # The Reconstruction of Trilinear Higgs Couplings**footnote *Proceedings, Physics with a High-Luminosity 𝑒⁺𝑒⁻ Linear Collider, DESY/ECFA LC Workshop, DESY 99-123F.
## 1. Introduction
1. The Higgs mechanism is a cornerstone in the electroweak sector of the Standard Model (SM) . The electroweak gauge bosons and the fundamental matter particles acquire masses through the interaction with a scalar field. The self-interaction of the scalar field leads to a non-zero field strength $`v=(\sqrt{2}G_F)^{1/2}246`$ GeV in the ground state, inducing the spontaneous breaking of the electroweak $`\mathrm{SU}(2)_\mathrm{L}\times \mathrm{U}(1)_\mathrm{Y}`$ symmetry down to the electromagnetic $`\mathrm{U}(1)_{\mathrm{EM}}`$ symmetry.
To establish the Higgs mechanism sui generis experimentally, the characteristic self-energy potential of the Standard Model,
$`V=\lambda \left[|\phi |^2\frac{1}{2}v^2\right]^2`$ (1)
with a minimum at $`\phi _0=v/\sqrt{2}`$, must be reconstructed once the Higgs particle will be discovered. This experimental task requires the measurement of the trilinear and quadrilinear self-couplings of the Higgs boson. The self-couplings are uniquely determined in the Standard Model by the mass of the Higgs boson which is related to the quadrilinear coupling $`\lambda `$ by $`M_H=\sqrt{2\lambda }v`$. Introducing the physical Higgs field $`H`$ in the neutral component of the doublet, $`\phi _0=(v+H)/\sqrt{2}`$, the multiple Higgs couplings can be derived from the potential $`V`$:
$`V={\displaystyle \frac{M_H^2}{2}}H^2+{\displaystyle \frac{M_H^2}{2v}}H^3+{\displaystyle \frac{M_H^2}{8v^2}}H^4`$ (2)
The trilinear and quadrilinear vertices of the Higgs field $`H`$ are given by the coefficients:
$`\lambda _{HHH}`$ $`=`$ $`3M_H^2/M_Z^2[\mathrm{unit}:\lambda _0=M_Z^2/v33.8GeV]`$ (3)
$`\lambda _{HHHH}`$ $`=`$ $`3M_H^2/M_Z^4[\mathrm{unit}:\lambda _0^2]`$ (4)
For a Higgs mass $`M_H=110`$ GeV, the trilinear coupling amounts to $`\lambda _{HHH}\lambda _0/M_Z=1.6`$ for a typical energy scale $`M_Z`$, whereas the quadrilinear coupling $`\lambda _{HHHH}\lambda _0^2=0.6`$ is suppressed with respect to the trilinear coupling by a factor close to the size of the weak gauge coupling.
The trilinear Higgs self-coupling can be measured directly in pair-production of Higgs particles at hadron and high-energy $`e^+e^{}`$ linear colliders. Several mechanisms that are sensitive to $`\lambda _{HHH}`$ can be exploited for this task. Higgs pairs can be produced through double Higgs-strahlung off $`W`$ or $`Z`$ bosons , $`WW`$ or $`ZZ`$ fusion \[??\]; moreover through gluon-gluon fusion in $`pp`$ collisions \[??\] and high-energy $`\gamma \gamma `$ fusion at photon colliders. In a precursor to this report it was recently shown that for collider energies up to about 1 TeV double Higgs-strahlung is the most promising process for measuring the trilinear coupling:
$`\begin{array}{ccccc}\text{double Higgs-strahlung}\hfill & :\hfill & e^+e^{}\hfill & & ZHH\hfill \\ & & & & \\ & & & Z& \end{array}`$ (8)
As evident from Fig. 1, the trilinear coupling is involved only in one diagram of this process. However, the two other diagrams are generated by the electroweak gauge interactions, and can thus be assumed known since the Higgs-gauge coupling is measured directly in the basic process of single Higgs-strahlung.
After the decay of the Higgs bosons into $`b`$ and $`\tau `$ pairs many reducible electroweak and QCD background processes contribute to the final state. It has been demonstrated in careful experimental simulations and phenomenological analyses that the signal can nevertheless be isolated in a clean form.
With values typically of the order of a few fb and below, the cross sections are small at $`e^+e^{}`$ linear colliders for masses of the Higgs boson in the intermediate mass range. High luminosities are therefore needed to produce a sufficiently large sample of Higgs-pair events and to isolate the signal from the background.
2. If a light Higgs boson with a mass below about 130 GeV will be discovered, the Standard Model is likely embedded in a supersymmetric theory. The minimal supersymmetric extension of the Standard Model (MSSM) includes two iso-doublets of Higgs fields $`\phi _1`$, $`\phi _2`$ which, after three components are absorbed to provide longitudinal degrees of freedom to the electroweak gauge bosons, gives rise to a quintet of physical Higgs boson states: $`h`$, $`H`$, $`A`$, $`H^\pm `$ . While an upper bound of about 130 GeV can be derived on the mass of the light CP-even neutral Higgs boson $`h`$ , the heavy CP-even and CP-odd neutral Higgs bosons $`H`$, $`A`$, and the charged Higgs bosons $`H^\pm `$ may have masses of the order of the electroweak symmetry scale $`v`$ up to about 1 TeV. This extended Higgs system can be described by two parameters at the tree level: one mass parameter which is generally identified with the pseudoscalar $`A`$ mass $`M_A`$, and tan$`\beta `$, the ratio of the vacuum expectation values of the two neutral fields in the two iso-doublets.
The mass parameters and the couplings in the self-interaction potential of the two Higgs doublets are affected by top and stop-loop radiative corrections. Radiative corrections in the one-loop leading $`M_t^4`$ approximation are parameterized by
$`ϵ{\displaystyle \frac{3G_FM_t^4}{\sqrt{2}\pi ^2\mathrm{sin}^2\beta }}\mathrm{log}{\displaystyle \frac{\stackrel{~}{M}^2}{M_t^2}}`$ (9)
where the scale of supersymmetry breaking is characterized by a common squark-mass value $`\stackrel{~}{M}`$ which will be set to $`1`$ TeV in the numerical analyses; if stop mixing effects are modest at the SUSY scale, they can be accounted for by shifting $`\stackrel{~}{M}^2`$ in $`ϵ`$ by the amount
$`\begin{array}{ccccc}\stackrel{~}{M}^2\stackrel{~}{M}^2+\mathrm{\Delta }\stackrel{~}{M}^2& :\hfill & \hfill \mathrm{\Delta }\stackrel{~}{M}^2& =& \widehat{A}^2[1\widehat{A}^2/(12\stackrel{~}{M}^2)]\hfill \\ & & \hfill \widehat{A}& =& A\mu \mathrm{cot}\beta \hfill \end{array}`$ (12)
where $`A`$ and $`\mu `$ correspond to the trilinear coupling in the top sector and the higgsino mass parameter in the superpotential, respectively. The neutral CP-even Higgs boson masses, and the mixing angle $`\alpha `$ in the neutral sector are given in this approximation by
$`M_{h,H}^2`$ $`=`$ $`\frac{1}{2}\left[M_A^2+M_Z^2+ϵ\sqrt{\left(M_A^2+M_Z^2+ϵ\right)^24M_A^2M_Z^2\mathrm{cos}^22\beta 4ϵ\left(M_A^2\mathrm{sin}^2\beta +M_Z^2\mathrm{cos}^2\beta \right)}\right]`$
$`\mathrm{tan}2\alpha `$ $`=`$ $`\mathrm{tan}2\beta {\displaystyle \frac{M_A^2+M_Z^2}{M_A^2M_Z^2+ϵ/\mathrm{cos}2\beta }}\text{with}{\displaystyle \frac{\pi }{2}}\alpha 0`$ (13)
when expressed in terms of the mass $`M_A`$ and tan $`\beta `$.
The set of trilinear couplings between the neutral physical Higgs bosons can be written in units of $`\lambda _0`$ as
$`\lambda _{hhh}`$ $`=`$ $`3\mathrm{cos}2\alpha \mathrm{sin}\left(\beta +\alpha \right)+3{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{cos}\alpha }{\mathrm{sin}\beta }}\mathrm{cos}^2\alpha `$
$`\lambda _{Hhh}`$ $`=`$ $`2\mathrm{sin}2\alpha \mathrm{sin}\left(\beta +\alpha \right)\mathrm{cos}2\alpha \mathrm{cos}\left(\beta +\alpha \right)+3{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{sin}\alpha }{\mathrm{sin}\beta }}\mathrm{cos}^2\alpha `$
$`\lambda _{HHh}`$ $`=`$ $`2\mathrm{sin}2\alpha \mathrm{cos}\left(\beta +\alpha \right)\mathrm{cos}2\alpha \mathrm{sin}\left(\beta +\alpha \right)+3{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{cos}\alpha }{\mathrm{sin}\beta }}\mathrm{sin}^2\alpha `$
$`\lambda _{HHH}`$ $`=`$ $`3\mathrm{cos}2\alpha \mathrm{cos}\left(\beta +\alpha \right)+3{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{sin}\alpha }{\mathrm{sin}\beta }}\mathrm{sin}^2\alpha `$
$`\lambda _{hAA}`$ $`=`$ $`\mathrm{cos}2\beta \mathrm{sin}\left(\beta +\alpha \right)+{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{cos}\alpha }{\mathrm{sin}\beta }}\mathrm{cos}^2\beta `$
$`\lambda _{HAA}`$ $`=`$ $`\mathrm{cos}2\beta \mathrm{cos}\left(\beta +\alpha \right)+{\displaystyle \frac{ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{sin}\alpha }{\mathrm{sin}\beta }}\mathrm{cos}^2\beta `$ (14)
In the decoupling limit $`M_A^2M_H^2M_{H^\pm }^2v^2/2`$, the self-coupling of the light CP-even neutral Higgs boson $`h`$ approaches the SM value.
In the subsequent numerical analysis the complete one-loop and the leading two-loop corrections to the MSSM Higgs masses and to the trilinear couplings are included, as presented in Ref. . Mixing effects due to non-vanishing $`A`$, $`\mu `$ parameters primarily affect the light Higgs mass; the upper limit on $`M_h`$ depends strongly on the size of the mixing parameters, raising this value for tan $`\beta 2.5`$ beyond the reach of LEP2, cf. Ref. . The couplings however are affected less by higher-order corrections when evaluated for the physical Higgs masses. The variation of the trilinear couplings with $`M_A`$ is shown for two values $`\mathrm{tan}\beta =3`$ and $`50`$ in Figs. 2a and 2b. The region in which the couplings vary rapidly, corresponds to the $`h/H`$ cross-over region of the two mass branches in the neutral CP-even sector, cf. eq. (13). The trilinear couplings between $`h`$, $`H`$ and the pseudoscalar pair $`AA`$ are in general significantly smaller than the trilinear couplings among the CP-even Higgs bosons.
In contrast to the Standard Model, resonance production of the heavy neutral Higgs boson $`H`$ followed by subsequent decays $`Hhh`$, plays a dominant role in part of the parameter space for moderate values of $`\mathrm{tan}\beta `$ and $`H`$ masses between 200 and 350 GeV, Ref. . In this range, the branching ratio, derived from the partial width
$`\mathrm{\Gamma }[Hhh]={\displaystyle \frac{\sqrt{2}G_FM_Z^4}{32\pi M_H}}\lambda _{Hhh}^2\beta _h`$ (15)
is neither too small nor too close to unity to be measured directly. \[The decay of either $`h`$ or $`H`$ into a pair of pseudoscalar states, $`h/HAA`$, is kinematically not possible in the parameter range which the present analysis is based upon; if realized, the couplings $`\lambda _{hAA}`$ and $`\lambda _{HAA}`$ can be determined in the same way.\] If double Higgs production is mediated by the resonant production of $`H`$, the total production cross section of light Higgs pairs increases by about an order of magnitude .
The trilinear Higgs-boson couplings are involved in a large number of processes at $`e^+e^{}`$ linear colliders among which double Higgs-strahlung and triple Higgs production are the preferred channels :
$`\begin{array}{cccccc}\text{double Higgs-strahlung}:\hfill & e^+e^{}\hfill & ZH_iH_j\hfill & \mathrm{and}\hfill & ZAA\hfill & [H_{i,j}=h,H]\hfill \\ \text{triple Higgs production}:\hfill & e^+e^{}\hfill & AH_iH_j\hfill & \mathrm{and}\hfill & AAA\hfill & \end{array}`$ (18)
The trilinear couplings which enter for various final states, cf. Fig. 3, are marked by a cross in the matrix Table 1. In the ideal case the system could be solved for all $`\lambda ^{}`$s, up to discrete ambiguities, based on double Higgs-strahlung, $`Ahh`$ and triple $`A`$ production \[”bottom-up approach”\]. This can easily be inferred from the correlation matrix Table 1. From $`\sigma (ZAA)`$ and $`\sigma (AAA)`$ the couplings $`\lambda (hAA)`$ and $`\lambda (HAA)`$ can be extracted. In a second step, $`\sigma (Zhh)`$ and $`\sigma (Ahh)`$ can be used to solve for $`\lambda (hhh)`$ and $`\lambda (Hhh)`$; subsequently, $`\sigma (ZHh)`$ for $`\lambda (HHh)`$; and, finally, $`\sigma (ZHH)`$ for $`\lambda (HHH)`$. The remaining triple Higgs cross sections $`\sigma (AHh)`$ and $`\sigma (AHH)`$ could provide additional redundant information on the trilinear couplings.
In practice, not all the cross sections will be large enough to be accessible experimentally, preventing the straightforward solution for the complete set of couplings. In this situation however the reverse direction can be followed \[”top-down approach”\]. The trilinear Higgs couplings can stringently be tested by comparing the theoretical predictions of the cross sections with the experimental results for the accessible channels of double and triple Higgs production.
If, as expected in the MSSM, the couplings $`hAA`$ and $`HAA`$ are very small, the $`ZAA`$ and $`AAA`$ final states can be left out of the analysis. This conclusion can be checked experimentally in a model-independent way, assuming nothing but the knowledge of pre-determined gauge boson-Higgs couplings. The system is then reduced to the trilinear couplings among the CP-even Higgs bosons $`h`$, $`H`$ \[in the double-line box of Table 1\] which can be measured in the analysis chain outlined in the previous paragraph, based solely on double Higgs-strahlung $`ZH_iH_j`$ and triple Higgs-production $`Ahh`$. The $`hH_iH_j`$ couplings involving the light Higgs boson $`h`$ with any combination of CP-even Higgs bosons can thus be determined in total.
The processes $`e^+e^{}`$$`ZH_iA`$ \[$`H_i=h,`$ $`H`$\] of mixed CP-even/CP-odd Higgs final states are generated through gauge interactions alone, mediated by virtual $`Z`$ bosons decaying to the CP even–odd Higgs pair, $`Z^{}H_iA`$. These parity-mixed processes do not involve trilinear Higgs-boson couplings.
3. In this report we summarize the results of Ref. for the production of Higgs boson pairs in the Standard Model and in the minimal supersymmetric extension. Other aspects have been discussed in Refs. . The comparison with LHC channels has been presented in Refs. . The analyses have been carried out for $`e^+e^{}`$ linear colliders , which are currently designed for an initial phase in the range $`\sqrt{s}=500`$ GeV to 1 TeV. The small cross sections require high luminosities as foreseen in the TESLA design with targets of $`=300`$ and 500 fb<sup>-1</sup> per annum for $`\sqrt{s}=500`$ and $`800`$ GeV, respectively .
The report is divided into two parts. In Section 2 we discuss the measurement of the trilinear Higgs coupling in the Standard Model for double Higgs-strahlung at $`e^+e^{}`$ linear colliders. In Section 3 this program, including the triple Higgs production, is extended to the Minimal Supersymmetric Standard Model MSSM.
## 2. Higgs Pair–Production in the Standard Model
### 2.1 Double Higgs-strahlung
The (unpolarized) differential cross section for the process of double Higgs-strahlung $`e^+e^{}`$$`ZHH`$, cf. Fig. 1, can be cast into the form
$`{\displaystyle \frac{d\sigma [e^+e^{}ZHH]}{dx_1dx_2}}={\displaystyle \frac{\sqrt{2}G_F^3M_Z^6}{384\pi ^3s}}{\displaystyle \frac{v_e^2+a_e^2}{(1\mu _Z)^2}}𝒵(x_1,x_2)`$ (19)
after the angular dependence is integrated out. The vector and axial-vector $`Z`$ charges of the electron are defined as usual, by $`v_e=1+4\mathrm{sin}^2\theta _W`$ and $`a_e=1`$. $`x_{1,2}=2E_{1,2}/\sqrt{s}`$ are the scaled energies of the two Higgs particles, $`x_3=2x_1x_2`$ is the scaled energy of the $`Z`$ boson, and $`y_i=1x_i`$; the square of the reduced masses is denoted by $`\mu _i=M_i^2/s`$, and $`\mu _{ij}=\mu _i\mu _j`$. In terms of these variables, the coefficient $`𝒵`$ may be written as:
$`𝒵=𝔷^2f_0+{\displaystyle \frac{1}{4\mu _Z(y_1+\mu _{HZ})}}[{\displaystyle \frac{f_1}{y_1+\mu _{HZ}}}+{\displaystyle \frac{f_2}{y_2+\mu _{HZ}}}+2\mu _Z𝔷f_3]+\{y_1y_2\}`$ (20)
with
$`𝔷={\displaystyle \frac{\lambda _{HHH}}{y_3\mu _{HZ}}}+{\displaystyle \frac{2}{y_1+\mu _{HZ}}}+{\displaystyle \frac{2}{y_2+\mu _{HZ}}}+{\displaystyle \frac{1}{\mu _Z}}`$ (21)
The coefficients $`f_i`$ are listed in Ref. . The first term in the coefficient $`𝔷`$ includes the trilinear coupling $`\lambda _{HHH}`$. The other terms are related to sequential Higgs-strahlung amplitudes and the 4-gauge-Higgs boson coupling; the individual terms can easily be identified by examining the characteristic propagators.
Since double Higgs-strahlung is mediated by s-channel $`Z`$-boson exchange, the cross section doubles if oppositely polarized electron and positron beams are used.
The cross sections for double Higgs-strahlung in the intermediate mass range are presented in Fig. 4a for total $`e^+e^{}`$ energies of $`\sqrt{s}=500`$ GeV and 1 TeV. The cross sections are shown for polarized electrons and positrons, $`\lambda _e^{}\lambda _{e^+}=1`$. As a result of the scaling behavior, the cross section for double Higgs-strahlung decreases with rising energy beyond the threshold region. The cross section increases with rising trilinear self-coupling in the vicinity of the SM value. The sensitivity to the $`HHH`$ self-coupling is demonstrated in Fig. 4b by varying the trilinear coupling $`\kappa \lambda _{HHH}`$ within the range $`\kappa =0.8`$ and $`1.2`$; the sensitivity is also illustrated by the vertical arrows in Fig. 4a for a variation of $`\kappa `$ in the same range. Evidently the cross section $`\sigma (`$$`e^+e^{}`$$`ZHH)`$ is sensitive to the value of the trilinear coupling; the sensitivity is not swamped by the irreducible background diagrams involving only the Higgs-gauge couplings. While the irreducible background diagrams become more important for rising energies, the sensitivity to the trilinear Higgs coupling is very large just above the kinematical threshold for the $`ZHH`$ final state. Near the threshold the value of the propagator of the intermediate virtual Higgs boson connecting to the two real Higgs bosons through $`\lambda _{HHH}`$ in the final state, is maximal. The maximum cross section for double Higgs-strahlung is reached at energies $`\sqrt{s}2M_H+M_Z+200`$ GeV, i.e. for Higgs masses in the lower part of the intermediate range at $`\sqrt{s}500`$ GeV.
Below 1 TeV one can always find collider energies at which the cross section for Higgs-strahlung $`e^+e^{}ZHH`$ is larger than the cross section for $`WW`$ fusion of two Higgs bosons, $`e^+e^{}\overline{\nu }_e\nu _eHH`$. However, for collider energies above 1 TeV the logarithmic increase of the $`t`$-channel fusion process dominates over the Higgs-strahlung mechanism which scales in the energy.
In recent experimental simulations of the Higgs-strahlung process it has been shown that the signal for two-Higgs boson production can be extracted despite the multi-channel reducible background . A sensitivity better than $`20\%`$ can be expected for the measurement of the trilinear Higgs self-coupling in the lower part of the intermediate Higgs mass range of the Standard Model.
The complete reconstruction of the Higgs potential in the Standard Model requires the measurement of the quadrilinear coupling $`\lambda _{HHHH}`$, too. This coupling is suppressed relative to the trilinear coupling by a factor which is effectively of the order of the weak gauge coupling for masses in the lower part of the intermediate Higgs mass range. Access to the quadrilinear coupling can be obtained directly only through the production of three Higgs bosons: $`e^+e^{}`$$`ZHHH`$. However, this cross section is strongly reduced by three orders of magnitude compared to the corresponding double-Higgs channel. As argued before, the signal amplitude involving the four-Higgs coupling \[as well as the irreducible Higgs-strahlung amplitudes\] is suppressed, leading to a reduction by a factor $`[\lambda _{HHHH}^2\lambda _0^4/16\pi ^2]/[\lambda _{HHH}^2\lambda _0^2/M_Z^2]10^3`$. Irreducible background diagrams are suppressed by a ratio of similar size. Moreover, the phase space is reduced by the additional heavy particle in the final state. A few illustrative examples of cross sections for triple Higgs-strahlung are listed in Table 2.
## 3. The Supersymmetric Higgs Sector
A large ensemble of Higgs couplings are present in supersymmetric theories. Even in the minimal realization MSSM, six different trilinear couplings $`hhh`$, $`Hhh`$, $`HHh`$, $`HHH`$, $`hAA`$, $`HAA`$ are generated among the neutral particles, and many more quadrilinear couplings. Since in major parts of the MSSM parameter space the Higgs bosons $`H`$, $`A`$, $`H^\pm `$ are quite heavy, we will focus primarily on the production of light neutral pairs $`hh`$, yet the production of heavy Higgs bosons will also be discussed where appropriate. The channels in which trilinear Higgs couplings can be probed in $`e^+e^{}`$ collisions, have been catalogued in Table 1.
Barring the exceptional case of very light pseudoscalar $`A`$ states, $`\lambda _{Hhh}`$ is the only trilinear coupling that may be measured in resonance decays, $`Hhh`$, while all the other couplings must be accessed in continuum pair production. The relevant mechanisms have been categorized in Fig. 3 for double Higgs-strahlung and associated triple Higgs production.
### 3.1 Double Higgs-strahlung
The (unpolarized) cross section for double Higgs-strahlung, $`e^+e^{}Zhh`$, is modified with regard to the Standard Model by $`H`$,$`A`$ exchange diagrams, cf. Fig. 3:
$`{\displaystyle \frac{d\sigma [e^+e^{}Zhh]}{dx_1dx_2}}`$ $`=`$ $`{\displaystyle \frac{\sqrt{2}G_F^3M_Z^6}{384\pi ^3s}}{\displaystyle \frac{v_e^2+a_e^2}{(1\mu _Z)^2}}𝒵_{11}(x_1,x_2)`$ (22)
with
$`𝒵_{11}`$ $`=`$ $`𝔷_{11}^2f_0+{\displaystyle \frac{𝔷}{2}}\left[{\displaystyle \frac{\mathrm{sin}^2\left(\beta \alpha \right)f_3}{y_1+\mu _{1Z}}}+{\displaystyle \frac{\mathrm{cos}^2\left(\beta \alpha \right)f_3}{y_1+\mu _{1A}}}\right]+{\displaystyle \frac{\mathrm{sin}^4\left(\beta \alpha \right)}{4\mu _Z\left(y_1+\mu _{1Z}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{1Z}}}+{\displaystyle \frac{f_2}{y_2+\mu _{1Z}}}\right]`$ (23)
$`+`$ $`{\displaystyle \frac{\mathrm{cos}^4\left(\beta \alpha \right)}{4\mu _Z\left(y_1+\mu _{1A}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{1A}}}+{\displaystyle \frac{f_2}{y_2+\mu _{1A}}}\right]+{\displaystyle \frac{\mathrm{sin}^22\left(\beta \alpha \right)}{8\mu _Z\left(y_1+\mu _{1A}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{1Z}}}+{\displaystyle \frac{f_2}{y_2+\mu _{1Z}}}\right]`$
$`+`$ $`\left\{y_1y_2\right\}`$
and
$`𝔷_{11}=\left[{\displaystyle \frac{\lambda _{hhh}\mathrm{sin}\left(\beta \alpha \right)}{y_3\mu _{1Z}}}+{\displaystyle \frac{\lambda _{Hhh}\mathrm{cos}\left(\beta \alpha \right)}{y_3\mu _{2Z}}}\right]+{\displaystyle \frac{2\mathrm{sin}^2\left(\beta \alpha \right)}{y_1+\mu _{1Z}}}+{\displaystyle \frac{2\mathrm{sin}^2\left(\beta \alpha \right)}{y_2+\mu _{1Z}}}+{\displaystyle \frac{1}{\mu _Z}}`$ (24)
The notation follows the Standard Model, with $`\mu _1=M_h^2/s`$ and $`\mu _2=M_H^2/s`$. In parameter ranges in which the heavy neutral Higgs boson $`H`$ or the pseudoscalar Higgs boson $`A`$ becomes resonant, the decay widths are implicitly included by shifting the masses to complex Breit-Wigner values.
The total cross sections are shown in Fig. 5 for the $`e^+e^{}`$ collider energy $`\sqrt{s}=500`$ GeV. The parameter $`\mathrm{tan}\beta `$ is chosen to be 3 and 50, and the mixing parameters $`A=1`$ TeV and $`\mu =1`$ TeV and $`1`$ TeV, respectively. If $`\mathrm{tan}\beta `$ and the mass $`M_h`$ are fixed, the masses of the other heavy Higgs bosons are predicted in the MSSM . Since the vertices are suppressed by $`\mathrm{sin}/\mathrm{cos}`$ functions of the mixing angles $`\beta `$ and $`\alpha `$, the continuum $`hh`$ cross sections are suppressed compared to the Standard Model. The size of the cross sections increases for moderate $`\mathrm{tan}\beta `$ by nearly an order of magnitude if the $`hh`$ final state can be generated in the chain $`e^+e^{}ZHZhh`$ via resonant $`H`$ Higgs-strahlung. If the light Higgs mass approaches the upper limit for a given value of $`\mathrm{tan}\beta `$, the decoupling theorem drives the cross section of the supersymmetric Higgs boson back to its Standard Model value since the Higgs particles $`A`$, $`H`$, $`H^\pm `$ become asymptotically heavy in this limit. As a result of the decoupling theorem, resonance production is not effective for large tan$`\beta `$. If the $`H`$ mass is large enough to allow for decays to $`hh`$ pairs, the $`ZZH`$ coupling is already too small to generate a sizable cross section.
While the basic structure for the cross sections of the other $`ZH_iH_j`$ \[$`H_{i,j}=h`$, $`H`$\] final states remains the same, the complexity increases due to unequal masses of the final-state particles. The double differential cross section of the process $`e^+e^{}ZH_iH_j`$ is given for unpolarized beams by the expression
$`{\displaystyle \frac{d\sigma [e^+e^{}ZH_iH_j]}{dx_1dx_2}}`$ $`=`$ $`{\displaystyle \frac{\sqrt{2}G_F^3M_Z^6}{384\pi ^3s}}{\displaystyle \frac{v_e^2+a_e^2}{(1\mu _Z)^2}}𝒵_{ij}(x_1,x_2)`$ (25)
The coefficients $`𝒵_{ij}`$ in the cross sections can be written as
$`𝒵_{ij}`$ $`=`$ $`𝔷_{ij}^2f_0+{\displaystyle \frac{𝔷_{ij}}{2}}\left[{\displaystyle \frac{d_id_jf_3}{y_1+\mu _{iZ}}}+{\displaystyle \frac{c_ic_jf_3}{y_1+\mu _{iA}}}\right]+{\displaystyle \frac{\left(d_id_j\right)^2}{4\mu _Z\left(y_1+\mu _{iZ}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{iZ}}}+{\displaystyle \frac{f_2}{y_2+\mu _{jZ}}}\right]`$ (26)
$`+`$ $`{\displaystyle \frac{\left(c_ic_j\right)^2}{4\mu _Z\left(y_1+\mu _{iA}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{iA}}}+{\displaystyle \frac{f_2}{y_2+\mu _{jA}}}\right]+{\displaystyle \frac{d_id_jc_ic_j}{2\mu _Z\left(y_1+\mu _{iA}\right)}}\left[{\displaystyle \frac{f_1}{y_1+\mu _{iZ}}}+{\displaystyle \frac{f_2}{y_2+\mu _{jZ}}}\right]`$
$`+`$ $`\left\{(y_1,\mu _i)(y_2,\mu _j)\right\}`$
with
$`𝔷_{ij}=\left[{\displaystyle \frac{d_1\lambda _{hH_iH_j}}{y_3\mu _{1Z}}}+{\displaystyle \frac{d_2\lambda _{HH_iH_j}}{y_3\mu _{2Z}}}\right]+{\displaystyle \frac{2d_id_j}{y_1+\mu _{iZ}}}+{\displaystyle \frac{2d_id_j}{y_2+\mu _{jZ}}}+{\displaystyle \frac{\delta _{ij}}{\mu _Z}}`$ (27)
The expressions for $`f_0`$ to $`f_3`$ have been denoted in Ref. . The modifications of the SM Higgs-gauge coupling in the MSSM are accounted for by the mixing parameters:
$`\begin{array}{cccccc}VVh:\hfill & d_1=\hfill & \mathrm{sin}(\beta \alpha )\hfill & VVH:\hfill & d_2=\hfill & \mathrm{cos}(\beta \alpha )\hfill \\ VAh:\hfill & c_1=\hfill & \mathrm{cos}(\beta \alpha )\hfill & VAH:\hfill & c_2=\hfill & \mathrm{sin}(\beta \alpha )\hfill \end{array}`$ (30)
for $`V=Z`$ and $`W`$. The reduction of the $`Zhh`$ cross section is partly compensated by the $`ZHh`$ and $`ZHH`$ cross sections so that their sum adds up approximately to the SM value, if kinematically possible, as demonstrated in Fig. 6a for tan $`\beta =3`$ at $`\sqrt{s}=500`$ GeV and $`hh`$, $`Hh`$ and $`HH`$ final states.
### 3.2 Triple-Higgs Production
The 2-particle processes $`e^+e^{}`$$`ZH_i`$ and $`e^+e^{}`$$`AH_i`$ are among themselves and mutually complementary to each other in the MSSM , coming with the coefficients $`\mathrm{sin}^2(\beta \alpha )/\mathrm{cos}^2(\beta \alpha )`$ and $`\mathrm{cos}^2(\beta \alpha )/\mathrm{sin}^2(\beta \alpha )`$ for $`H_i=h,`$ $`H`$, respectively. Since multi-Higgs final states are mediated by virtual $`h,`$ $`H`$ bosons, the two types of self-complementarity and mutual complementarity are also operative in double-Higgs production: $`e^+e^{}`$$`ZH_iH_j,`$ $`ZAA`$ and $`AH_iH_j,`$ $`AAA`$. As the different mechanisms are intertwined, the complementarity between these 3-particle final states is of more complex matrix form, as evident from Fig. 3.
We will analyze in this section in detail the processes involving only the light neutral Higgs boson $`h`$, $`e^+e^{}`$$`Ahh`$. The more cumbersome analyses for heavy neutral Higgs bosons $`H`$ and other channels have been presented in Ref. .
In the first case one finds for the unpolarized cross section
$`{\displaystyle \frac{d\sigma [e^+e^{}Ahh]}{dx_1dx_2}}={\displaystyle \frac{G_F^3M_Z^6}{768\sqrt{2}\pi ^3s}}{\displaystyle \frac{v_e^2+a_e^2}{(1\mu _Z)^2}}𝔄_{11}(x_1,x_2)`$ (31)
where the function $`𝔄_{11}`$ reads
$`𝔄_{11}`$ $`=`$ $`\left[{\displaystyle \frac{c_1\lambda _{hhh}}{y_3\mu _{1A}}}+{\displaystyle \frac{c_2\lambda _{Hhh}}{y_3\mu _{2A}}}\right]^2{\displaystyle \frac{g_0}{2}}+{\displaystyle \frac{c_1^2\lambda _{hAA}^2}{\left(y_1+\mu _{1A}\right)^2}}g_1+{\displaystyle \frac{c_1^2d_1^2}{\left(y_1+\mu _{1Z}\right)^2}}g_2`$ (32)
$`+`$ $`\left[{\displaystyle \frac{c_1\lambda _{hhh}}{y_3\mu _{1A}}}+{\displaystyle \frac{c_2\lambda _{Hhh}}{y_3\mu _{2A}}}\right]\left[{\displaystyle \frac{c_1\lambda _{hAA}}{y_1+\mu _{1A}}}g_3+{\displaystyle \frac{c_1d_1}{y_1+\mu _{1Z}}}g_4\right]+{\displaystyle \frac{c_1^2\lambda _{hAA}^2}{2\left(y_1+\mu _{1A}\right)\left(y_2+\mu _{1A}\right)}}g_5`$
$`+`$ $`{\displaystyle \frac{c_1^2d_1\lambda _{hAA}}{\left(y_1+\mu _{1A}\right)\left(y_1+\mu _{1Z}\right)}}g_6+{\displaystyle \frac{c_1^2d_1\lambda _{hAA}}{\left(y_1+\mu _{1A}\right)\left(y_2+\mu _{1Z}\right)}}g_7+{\displaystyle \frac{c_1^2d_1^2}{2\left(y_1+\mu _{1Z}\right)\left(y_2+\mu _{1Z}\right)}}g_8`$
$`+`$ $`\left\{y_1y_2\right\}`$
with $`\mu _{1,2}=M_{h,H}^2/s`$. The coefficients $`g_k`$ are listed in Ref. .
The general form of the double differential cross section of the process $`e^+e^{}AH_iH_j`$ for unpolarized beams reads in the same notation as above:
$`{\displaystyle \frac{d\sigma [e^+e^{}AH_iH_j]}{dx_1dx_2}}={\displaystyle \frac{G_F^3M_Z^6}{768\sqrt{2}\pi ^3s}}{\displaystyle \frac{v_e^2+a_e^2}{(1\mu _Z)^2}}𝔄_{ij}(x_i,x_2)`$ (33)
with the function $`𝔄_{ij}(x_1,x_2)`$ defined by
$`𝔄_{ij}`$ $`=`$ $`\left[{\displaystyle \frac{\lambda _{hH_iH_j}c_1}{y_3\mu _{1A}}}+{\displaystyle \frac{\lambda _{HH_iH_j}c_2}{y_3\mu _{2A}}}\right]^2g_0+{\displaystyle \frac{\lambda _{H_jAA}^2c_i^2}{\left(y_1+\mu _{iA}\right)^2}}g_1+{\displaystyle \frac{\lambda _{H_iAA}^2c_j^2}{\left(y_2+\mu _{jA}\right)^2}}g_1^{}`$ (34)
$`+{\displaystyle \frac{c_j^2d_i^2}{\left(y_1+\mu _{iZ}\right)^2}}g_2+{\displaystyle \frac{c_i^2d_j^2}{\left(y_2+\mu _{jZ}\right)^2}}g_2^{}+\left[{\displaystyle \frac{\lambda _{hH_iH_j}c_1}{y_3\mu _{1A}}}+{\displaystyle \frac{\lambda _{HH_iH_j}c_2}{y_3\mu _{2A}}}\right]`$
$`\times \left[{\displaystyle \frac{\lambda _{H_jAA}c_i}{y_1+\mu _{iA}}}g_3+{\displaystyle \frac{\lambda _{H_iAA}c_j}{y_2+\mu _{jA}}}g_3^{}+{\displaystyle \frac{c_jd_i}{y_1+\mu _{iZ}}}g_4+{\displaystyle \frac{c_id_j}{y_2+\mu _{jZ}}}g_4^{}\right]`$
$`+{\displaystyle \frac{\lambda _{H_iAA}\lambda _{H_jAA}c_ic_j}{\left(y_1+\mu _{iA}\right)\left(y_2+\mu _{jA}\right)}}g_5+{\displaystyle \frac{c_ic_jd_id_j}{\left(y_1+\mu _{iZ}\right)\left(y_2+\mu _{jZ}\right)}}g_8+{\displaystyle \frac{\lambda _{H_jAA}c_ic_jd_i}{\left(y_1+\mu _{iA}\right)\left(y_1+\mu _{iZ}\right)}}g_6`$
$`+{\displaystyle \frac{\lambda _{H_iAA}c_ic_jd_j}{\left(y_2+\mu _{jA}\right)\left(y_2+\mu _{jZ}\right)}}g_6^{}+{\displaystyle \frac{\lambda _{H_jAA}c_i^2d_j}{\left(y_1+\mu _{iA}\right)\left(y_2+\mu _{jZ}\right)}}g_7+{\displaystyle \frac{\lambda _{H_iAA}c_j^2d_i}{\left(y_2+\mu _{jA}\right)\left(y_1+\mu _{iZ}\right)}}g_7^{}`$
The coefficients $`g_k`$ were given in Ref. .
The size of the total cross section $`\sigma [e^+e^{}Ahh]`$ is compared with double Higgs-strahlung $`\sigma [e^+e^{}Zhh]`$ in Fig. 6b for tan $`\beta =3`$ at $`\sqrt{s}=1`$ TeV. The cross section involving the pseudoscalar Higgs boson is small in the continuum. The effective coupling in the chain $`Ah_{virt}Ahh`$ is $`\mathrm{cos}(\beta \alpha )\lambda _{hhh}`$ while in the chain $`AH_{virt}Ahh`$ it is $`\mathrm{sin}(\beta \alpha )\lambda _{Hhh}`$; both products are small either in the first or in the second coefficient. Only for resonance $`H`$ decays $`AHAhh`$ the cross section becomes very large as expected from the decoupling theorem.
### 3.4 Sensitivity Areas
The results obtained in the preceding sections can be summarized in compact form by constructing sensitivity areas for the trilinear SUSY Higgs couplings based on the cross sections for double Higgs-strahlung and triple Higgs production. $`WW`$ double-Higgs fusion can provide additional information on the Higgs self-couplings, in particluar for large collider energies.
The sensitivity areas will be defined in the $`[M_A,`$ tan$`\beta ]`$ plane . The criteria for accepting a point in the plane as accessible for the measurement of a specific trilinear coupling are set as follows:
$`\begin{array}{cc}(i)\hfill & \sigma [\lambda ]>0.01\mathrm{fb}\hfill \\ (ii)\hfill & \mathrm{eff}\{\lambda 0\}>2\mathrm{st}.\mathrm{dev}.\mathrm{for}=2\mathrm{ab}^1\hfill \end{array}`$ (37)
The first criterion demands at least 20 events in a sample collected for an integrated luminosity of 2 ab<sup>-1</sup>, corresponding to the lifetime of a high-luminosity machine such as TESLA. The second criterion demands at least a 2 standard-deviation effect of the non-zero trilinear coupling away from zero. Even though the two criteria may look quite loose, the tightening of $`(i)`$ and/or $`(ii)`$ does not have a large impact on the size of the sensitivity areas in the $`[M_A,`$ tan$`\beta ]`$ plane, see Ref. . The second criterion was defined in Ref. slightly different by introducing the relative variation of the trilinear couplings with respect to the MSSM. \[Due to an algorithmic error, the sensitivity areas in Ref. had been overestimated.\] For the sake of simplicity, mixing effects are neglected in the analysis.
Sensitivity areas of the trilinear couplings among the scalar Higgs bosons $`h`$, $`H`$ in the correlation matrix Table 1, are depicted in Figs. 7a and 7b. If at most one heavy Higgs boson is present in the final state, the lower energy $`\sqrt{s}=500`$ GeV is more preferable in the case of double Higgs-strahlung. $`HH`$ final states in double Higgs-strahlung and triple Higgs production including $`A`$ give rise to larger sensitivity areas at the high energy $`\sqrt{s}=1`$ TeV. Apart from small regions in which interference effects play a major role, the magnitude of the sensitivity regions in the parameter tan$`\beta `$ is readily explained by the magnitude of the parameters $`\lambda \mathrm{sin}(\beta \alpha )`$ and $`\lambda \mathrm{cos}(\beta \alpha )`$, shown individually in Figs. 2a and 2b. For large $`M_A`$ the sensitivity criteria cannot be met any more either as a result of phase space effects or due to the suppression of the $`H`$, $`A`$ propagators for large masses. While the trilinear coupling of the light neutral CP-even Higgs boson is accessible in nearly the entire MSSM parameter space, the regions for the $`\lambda `$’s involving heavy Higgs bosons are rather restricted.
Since neither experimental efficiencies nor background related cuts are considered in this paper, the areas shown in Figs. 7a and 7b must be interpreted as maximal envelopes. They are expected to shrink when experimental efficiencies are properly taken into account; more elaborate cuts on signal and backgrounds, however, may help reduce their impact.
## 4. Conclusions
In this report we have analyzed the production of Higgs boson pairs and triple Higgs final states at $`e^+e^{}`$ linear colliders up to energies of 1 TeV. They will allow the measurement of the fundamental trilinear Higgs self-couplings. The first theoretical steps into this area have been taken by calculating the production cross sections in the Standard Model for Higgs bosons in the intermediate mass range and for Higgs bosons in the minimal supersymmetric extension.
The cross sections in the Standard Model for double Higgs-strahlung and triple Higgs production are small so that high luminosities are needed to perform these experiments. Even though the $`e^+e^{}`$ cross sections are smaller than the corresponding $`pp`$ cross sections, the strongly reduced number of background events renders the search for the Higgs-pair signal events, through $`bbbb`$ final states for instance, easier in the $`e^+e^{}`$ environment than in jetty LHC final states. For sufficiently high luminosities even the first phase of these colliders with an energy of 500 GeV will allow the experimental analysis of self-couplings for Higgs bosons in the intermediate mass range.
The extended Higgs spectrum in supersymmetric theories gives rise to a plethora of trilinear and quadrilinear couplings. The $`hhh`$ coupling is generally quite different from the Standard Model. It can be measured in $`hh`$ continuum production at $`e^+e^{}`$ linear colliders. Other couplings between heavy and light scalar Higgs bosons can be measured as well, though only in restricted areas of the $`[M_A,`$ tan$`\beta ]`$ parameter space as illustrated in the set of Figs. 7a and 7b. The trilinear couplings including the pseudoscalar Higgs boson $`A`$ are predicted to be small in the MSSM; future experimental analyses are therefore expected to give rise to upper bounds on these couplings if the MSSM is realized in Nature. For more general supersymmetric theories a model-independent analysis must be performed in triple Higgs channels.
### Acknowledgements
We gratefully acknowledge discussions with P. Gay, P. Lutz, L. Maiani, P. Osland and F. Richard. |
warning/0001/hep-ph0001004.html | ar5iv | text | # References
BOSE SYMMETRY INTERFERENCE EFFECTS OF $`4\pi `$ FINAL STATES
Jie Chen<sup>a,b,c</sup>, Xue-Qian Li<sup>a,b,c</sup>, Bing-Song Zou<sup>a,c</sup>
a. CCAST(World Laboratory), P.O.Box 8730, Beijing 100080, China
Abstract
We carefully analyze the relative branching ratios of $`4\pi `$ final states $`\pi ^+\pi ^{}\pi ^+\pi ^{}`$, $`\pi ^+\pi ^{}\pi ^0\pi ^0`$ and $`\pi ^0\pi ^0\pi ^0\pi ^0`$, from various resonances of $`J^{PC}=0^{++}`$, $`0^+`$, $`2^{++}`$. We find that the Bose symmetry interference effects would make their ratios to obviously differ from the naive counting values without considering these effects. The results should be applied to estimate correctly various $`4\pi `$ decay branching ratios of relevant resonances.
PACS number(s): 14.40Cs, 13.25Jx, 13.75Cs, 13.25Gv
I. Introduction
In the energy range of $`12.5`$ GeV, there exist very rich hadronic resonance spectra, including possible $`0^{++}`$, $`0^+`$, and $`2^{++}`$ glueballs. An important source of information about the nature of these resonances is their various decay branching ratios. Among the observed decay modes for the $`0^{++}`$, $`0^+`$ and $`2^{++}`$ resonances, the $`4\pi `$ final state is a very important one.
There are three kinds of $`4\pi `$ final states: $`\pi ^+\pi ^{}\pi ^+\pi ^{}`$, $`\pi ^+\pi ^{}\pi ^0\pi ^0`$ and $`\pi ^0\pi ^0\pi ^0\pi ^0`$. Usually, due to specialty and limitation of each detector and other reasons, one experiment is only good at studying one kind of the $`4\pi `$ final states. For example, the Crystal Barrel (CBAR) detector is particularly good at studying neutral final states and therefore studied resonances decaying into $`\pi ^0\pi ^0\pi ^0\pi ^0`$ final state; the BES (Beijing Spectrometer) detector is good at detecting charged particles and only has studied $`J/\mathrm{\Psi }`$ radiative decaying into $`\pi ^+\pi ^{}\pi ^+\pi ^{}`$ final state. The problem is that from the measured rate for one kind of the $`4\pi `$ final states how to deduce the rates for other two kinds of the $`4\pi `$ final states.
The $`4\pi `$ final states are usually produced via 2-meson intermediate states $`M_1`$ and $`M_2`$, namely the parent-resonance decays into two mesons which then result in the $`4\pi `$ final states:
$$MM_1+M_24\pi .$$
The parent-meson which we are interested in are $`0^{++},0^+`$ and $`2^{++}`$ etc., because they have the same quantum numbers as glueballs which are under intensive discussions at present. $`M_1`$ and $`M_2`$ can be various mesons among which $`\sigma \sigma ,\rho \rho `$ and $`f_2\sigma `$ are the most possible candidates and therefore discussed in this work.
Naive counting for the $`4\pi `$ final states from simple isospin decomposition would result in
$`\mathrm{\Gamma }(M\pi ^+\pi ^{}\pi ^0\pi ^0):\mathrm{\Gamma }(M\pi ^+\pi ^{}\pi ^+\pi ^{}):\mathrm{\Gamma }(M\pi ^0\pi ^0\pi ^0\pi ^0)`$
$`=4:4:1,\mathrm{for}\mathrm{f}_2\sigma ,\sigma \sigma \mathrm{intermediate}\mathrm{states};`$ (1)
$`\mathrm{\Gamma }(M\pi ^+\pi ^{}\pi ^0\pi ^0):\mathrm{\Gamma }(M\pi ^+\pi ^{}\pi ^+\pi ^{}):\mathrm{\Gamma }(M\pi ^0\pi ^0\pi ^0\pi ^0)`$
$`=2:1:0,\mathrm{for}\rho \rho \mathrm{intermediate}\mathrm{states},`$ (2)
where all interference effects are neglected.
Since all pions can be treated as identical particles and each mode has the same production amplitude up to an SU(2) factor, so unless all the interference terms cancel each other after integration over the invariant phase space of final states, their effects can be important. In earlier literatures, the naive counting was employed to evaluate production rate of one mode from others. Even though this counting way is simple and valid in certain cases, it may bring up remarkable errors in some cases.
In this work, we carefully analyze the relative ratios of $`B(M\pi ^+\pi ^{}\pi ^0\pi ^0):B(M\pi ^+\pi ^{}\pi ^+\pi ^{}):B(M\pi ^0\pi ^0\pi ^0\pi ^0)`$ by including interference terms precisely for various masses of the parent meson of $`0^{++},0^+,2^{++}`$, and intermediate 2-meson states.
In Sec.II, we present the formulation for analysis; the numerical results and discussion are given in Sec.III.
II. Formulation
For $`M\pi _1(p_1)\pi _2(p_2)\pi _3(p_3)\pi _4(p_4)`$ where $`p_i^{}s(i=1\mathrm{}4)`$ are the four-momenta of the four produced pions and we use notation
$$p_{ab}p_a+p_b,a,b=1,\mathrm{},4,\mathrm{and}ab.$$
The propagators take the Breit-Wigner form
$`F_{ab}`$ $`=`$ $`{\displaystyle \frac{i}{p_{ab}^2m^2+i\mathrm{\Gamma }m}}\mathrm{for}\mathrm{scalar}\mathrm{mesons};`$
$`D_{ab}^{\alpha \beta }`$ $`=`$ $`{\displaystyle \frac{i}{p_{ab}^2m^2+i\mathrm{\Gamma }m}}\stackrel{~}{g}^{\alpha \beta }\mathrm{for}\mathrm{massive}\mathrm{vector};`$
$`D_{ab}^{\alpha \beta \gamma \delta }`$ $`=`$ $`{\displaystyle \frac{i}{p_{ab}^2m^2+i\mathrm{\Gamma }m}}[{\displaystyle \frac{1}{2}}(\stackrel{~}{g}^{\alpha \gamma }\stackrel{~}{g}^{\beta \delta }+\stackrel{~}{g}^{\alpha \delta }\stackrel{~}{g}^{\beta \gamma }){\displaystyle \frac{1}{3}}\stackrel{~}{g}^{\alpha \beta }\stackrel{~}{g}^{\gamma \delta }]`$ (3)
$`\mathrm{for}\mathrm{spin}2\mathrm{tensor}\mathrm{meson},`$
where
$$\stackrel{~}{g}^{\alpha \beta }g^{\alpha \beta }+\frac{p_{ab}^\alpha p_{ab}^\beta }{m^2},$$
(4)
and $`m`$ is the mass of the concerned intermediate meson of spin 0 or 1 or 2. In the following we explicitly present the expressions with $`MM_1+M_24\pi `$ for various $`M,M_1,M_2`$ identities.
1. Decay of $`M(0^{++})4\pi `$.
To investigate the interference effects of $`M4\pi `$, we distinguish the processes caused by different intermediate states. Here we first ignore possible interferences between $`M\sigma \sigma 4\pi `$ and $`M\rho \rho 4\pi `$ and later we will show that except for special cases, it is legitimate. Then we argue that the conclusion can be generalized to most situations.
(a) The squares of amplitudes corresponding to $`\sigma \sigma `$ intermediate state are
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^2}{2}}|F_{12}^\sigma F_{34}^\sigma |^2\mathrm{for}f_0\pi ^+\pi ^{}\pi ^0\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^2}{4}}|F_{12}^\sigma F_{34}^\sigma +F_{14}^\sigma F_{32}^\sigma |^2\mathrm{for}f_0\pi ^+\pi ^{}\pi ^+\pi ^{};`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^2}{24}}|F_{12}^\sigma F_{34}^\sigma +F_{13}^\sigma F_{24}^\sigma +F_{14}^\sigma F_{32}^\sigma |^2\mathrm{for}f_0\pi ^0\pi ^0\pi ^0\pi ^0.`$ (5)
(b) via $`\rho \rho `$ intermediate states.
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{}_{}{}^{}2}{2}}|(p_1p_2)(p_3p_4)F_{12}^\rho F_{34}^\rho +(p_1p_4)(p_3p_2)F_{14}^\rho F_{32}^\rho |^2`$
$`\mathrm{for}f_0\pi ^+\pi ^0\pi ^{}\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{}_{}{}^{}2}{4}}|(p_1p_2)(p_3p_4)F_{12}^\rho F_{34}^\rho +(p_1p_4)(p_3p_2)F_{14}^\rho F_{32}^\rho |^2`$ (6)
$`\mathrm{for}f_0\pi ^+\pi ^{}\pi ^+\pi ^{},`$
where $`p_i^{}s`$ are the momenta of the outgoing pions and $`F_{ij}^\sigma ,F_{ij}^\rho `$ are the propagators of $`\sigma `$meson and $`\rho `$meson respectively. It is noted that there is no process $`f_0\rho \rho 4\pi ^0`$, because $`\rho ^0\pi ^0\pi ^0`$ is forbidden by isospin symmetry.
2. Decay of $`0^+`$ mesons.
It is obvious that $`0^+\sigma \sigma `$ is forbidden by parity and angular-momentum conservations, it decays into $`4\pi `$ only via $`\rho \rho `$ intermediate states.
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{}_{}{}^{}2}{2}}|ϵ_{\mu \nu \lambda \rho }(p_1^\mu p_2^\nu p_3^\lambda p_4^\rho F_{12}^\rho F_{34}^\rho p_1^\mu p_4^\nu p_3^\lambda p_2^\rho F_{14}^\rho F_{32}^\rho )|^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}0}^+\pi ^+\pi ^0\pi ^{}\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{}_{}{}^{}2}{4}}|ϵ_{\mu \nu \lambda \rho }(p_1^\mu p_2^\nu p_3^\lambda p_4^\rho F_{12}^\rho F_{34}^\rho p_1^\mu p_4^\nu p_3^\lambda p_2^\rho F_{14}^\rho F_{32}^\rho )|^2`$ (7)
$`\mathrm{for}\mathrm{\hspace{0.33em}0}^+\pi ^+\pi ^{}\pi ^+\pi ^{}.`$
3. Decay of $`2^{++}`$ mesons.
(a) Via $`\sigma \sigma `$ intermediate states,
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{{}_{}{}^{\prime \prime }2}}{2}}|\sqrt{{\displaystyle \frac{1}{6}}}(r^2+3r_Z^2)F_{12}^\sigma F_{34}^\sigma |^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}\pi ^+\pi ^{}\pi ^0\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{{}_{}{}^{\prime \prime }2}}{4}}|\sqrt{{\displaystyle \frac{1}{6}}}(r^2+3r_Z^2)F_{12}^\sigma F_{34}^\sigma +\sqrt{{\displaystyle \frac{1}{6}}}(r^{{}_{}{}^{\prime \prime }2}+3r_Z^{{}_{}{}^{\prime \prime }2})F_{14}^\sigma F_{32}^\sigma |^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}\pi ^+\pi ^{}\pi ^+\pi ^{};`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{24}}g^{{}_{}{}^{\prime \prime }2}|\sqrt{{\displaystyle \frac{1}{6}}}(r^2+3r_Z^2)F_{12}^\sigma F_{34}^\sigma +\sqrt{{\displaystyle \frac{1}{6}}}(r^{}_{}{}^{}2+3r_Z^{}_{}{}^{}2)F_{13}^\sigma F_{24}^\sigma +`$ (8)
$`\sqrt{{\displaystyle \frac{1}{6}}}(r^{{}_{}{}^{\prime \prime }2}+3r_Z^{{}_{}{}^{\prime \prime }2})F_{14}^\sigma F_{32}^\sigma |^2\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}\pi ^0\pi ^0\pi ^0\pi ^0,`$
where
$$rp_{12}p_{34},r^{^{}}p_{13}p_{24},r^{^{\prime \prime }}p_{14}p_{32}.$$
(b) Via $`\rho \rho `$ intermediate states.
For $`2^{++}`$ decays, $`\sigma \sigma `$ production can only occur at d-wave, which is more suppressed, therefore, the $`\rho \rho `$ mode may dominate in the $`2^{++}4\pi `$ decays. The amplitudes are
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{{}_{}{}^{\prime \prime }2}}{2}}|(p_{12}^xp_{34}^xp_{12}^yp_{34}^y+2p_{12}^zp_{34}^z)F_{12}^\rho F_{34}^\rho +(p_{14}^xp_{32}^xp_{14}^yp_{32}^y+2p_{14}^zp_{32}^z)F_{14}^\rho F_{32}^\rho |^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}\pi ^+\pi ^0\pi ^{}\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{g^{{}_{}{}^{\prime \prime }2}}{4}}|(p_{12}^xp_{34}^xp_{12}^yp_{34}^y+2p_{12}^zp_{34}^z)F_{12}^\rho F_{34}^\rho +(p_{14}^xp_{32}^xp_{14}^yp_{32}^y+2p_{14}^zp_{32}^z)F_{14}^\rho F_{32}^\rho |^2`$ (9)
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}\pi ^+\pi ^{}\pi ^+\pi ^{}.`$
4. Decay of $`2^{++}4\pi `$ via $`f_2\sigma `$ intermediate states.
Here we have a simplified expression for the decay modes instead of using the propagator given in eqs.(S0.Ex5) as
$$|M|^2=|(T^{11}T^{22}+2T^{33})_{(ab)}F_{f_2}(ab)F_\sigma (cd)|^2,$$
(10)
where
$$T^{ii}=q^iq^i+\frac{1}{3}(1+p^ip^i/M_{f_2}^2)|\stackrel{}{q}|^2,$$
(11)
and
$$p_{ab}=p_a+p_b,q_{ab}=p_ap_b,F_{f_2}=\frac{1}{M_{f_2}^2s_{ab}iM_{f_2}\mathrm{\Gamma }_{f_2}},s_{ab}=p_{ab}^2.$$
The subscript $`(ab)`$ denotes the argument indices in the tensor $`T^{ii}`$, Thus
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{2}}|(T^{11}T^{22}+2T^{33})_{12}F_{f_2}(12)F_\sigma (34)|^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}f_2\sigma \pi ^+\pi ^{}\pi ^0\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{4}}|(T^{11}T^{22}+2T^{33})_{12}F_{f_2}(12)F_\sigma (34)+(T^{11}T^{22}+2T^{33})_{14}F_{f_2}(14)F_\sigma (32)|^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}f_2\sigma \pi ^+\pi ^{}\pi ^+\pi ^{};`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{24}}|(T^{11}T^{22}+2T^{33})_{12}F_{f_2}(12)F_\sigma (34)+(T^{11}T^{22}+2T^{33})_{13}F_{f_2}(13)F_\sigma (24)+`$ (12)
$`(T^{11}T^{22}+2T^{33})_{14}F_{f_2}(14)F_\sigma (32)|^2`$
$`\mathrm{for}\mathrm{\hspace{0.33em}2}^{++}f_2\sigma \pi ^0\pi ^0\pi ^0\pi ^0.`$
5. The interference between channels with $`\sigma \sigma `$ and $`\rho \rho `$ intermediate states in $`M4\pi `$.
(a) Above, we have ignored possible interference between channels with $`\sigma \sigma `$ and $`\rho \rho `$ intermediate states for the $`4\pi `$ final states, just because we assume that one of the two modes would overwhelm over the other. This allegation might deviate from reality. So in this subsection, we study this interference effects. As an example, we only concentrate on the $`0^{++}`$ decays. Then we have the squares of amplitudes as
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{2}}|g^{^{}}(p_1p_2)(p_3p_4)F_{12}^\rho F_{34}^\rho +g^{^{}}(p_1p_4)(p_3p_2)F_{14}^\rho F_{32}^\rho +gF_{13}^\sigma F_{24}^\sigma |^2`$
$`\mathrm{for}f_0\pi ^+\pi ^0\pi ^{}\pi ^0;`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{4}}|g^{^{}}(p_1p_2)(p_3p_4)F_{12}^\rho F_{34}^\rho +g^{^{}}(p_1p_4)(p_3p_2)F_{14}^\rho F_{32}^\rho +g(F_{12}^\sigma F_{34}^\sigma +F_{14}^\sigma F_{32}^\sigma )|^2`$
$`\mathrm{for}f_0\pi ^+\pi ^{}\pi ^+\pi ^{};`$
$`|M|^2`$ $`=`$ $`{\displaystyle \frac{1}{24}}|gF_{12}^\sigma F_{34}^\sigma +gF_{13}^\sigma F_{24}^\sigma +gF_{14}^\sigma F_{32}^\sigma |^2`$ (13)
$`\mathrm{for}f_0\pi ^0\pi ^0\pi ^0\pi ^0.`$
(b) As an illustration let us study the decay of $`f_0(1750)`$, because there are data available for $`f_0(1750)\rho \rho `$ and $`\sigma \sigma `$ . The partial decay widths are
$`\mathrm{\Gamma }(f_0\sigma \sigma )`$ $`=`$ $`{\displaystyle \frac{g_f^2}{16\pi M_f}}(1{\displaystyle \frac{4m_\sigma ^2}{M_f^2}})^{1/2},`$
$`\mathrm{\Gamma }(f_0\rho \rho )`$ $`=`$ $`{\displaystyle \frac{g_f^{}_{}{}^{}2}{16\pi M_f}}(1{\displaystyle \frac{4m_\rho ^2}{M_f^2}})^{1/2}[3+{\displaystyle \frac{1}{4m_\rho ^4}}(M_f^44M_f^2m_\rho ^2)],`$
$`\mathrm{\Gamma }(\rho 2\pi )`$ $`=`$ $`{\displaystyle \frac{g_\rho ^2m_\rho }{48\pi }}(1{\displaystyle \frac{4m_\pi ^2}{m_\rho ^2}})^{3/2},`$
$`\mathrm{\Gamma }(\sigma \pi \pi )`$ $`=`$ $`{\displaystyle \frac{g_\sigma ^2}{16\pi m_\sigma }}(1{\displaystyle \frac{4m_\pi ^2}{m_\sigma ^2}})^{1/2}.`$ (14)
So we have
$$gg_fg_\sigma ^2,g^{^{}}g_f^{^{}}g_\rho ^2.$$
(15)
By setting $`\mathrm{\Gamma }^{th}=\mathrm{\Gamma }^{exp}`$, we have obtained all the coupling constants straightforwardly.
III. Numerical results and discussion
We have employed the Monte-Carlo program to carry out the calculations of the widths. And in the practical calculation, we need to multiply the propagators of vector and tensor $`F_\rho `$ and $`F_{f_2}`$ in all the equations by the Blatt-Weisskopt barrier factor $`B_l(p)`$ ($`B_0(p)=1`$), which is widely used in partial-wave analyses. Namely, in our program the $`F_\rho `$ is replaced by $`F_\rho B_l(p)`$ and $`F_{f_2}`$ by $`F_{f_2}B_2(p)`$ respectively. For the $`\sigma `$propagator, there are various forms. Here we only use two typical ones. The first is ,
$`F_\sigma `$ $`=`$ $`{\displaystyle \frac{1}{M_\sigma ^2siM_\sigma (\mathrm{\Gamma }_1(s)+\mathrm{\Gamma }_2(s))}},`$ (16)
$`\mathrm{where}`$ (17)
$`\mathrm{\Gamma }_1(s)=G_1{\displaystyle \frac{\sqrt{14m_\pi ^2/s}}{\sqrt{14m_\pi ^2/M_\sigma ^2}}}{\displaystyle \frac{sm_\pi ^2/2}{M_\sigma ^2m_\pi ^2/2}}e^{(sM_\sigma ^2)/4\beta ^2},`$
$`\mathrm{\Gamma }_2(s)=G_2{\displaystyle \frac{\sqrt{116m_\pi ^2/s}}{1+exp(\mathrm{\Lambda }(s_0s))}}{\displaystyle \frac{1+exp(\mathrm{\Lambda }(s_0M_\sigma ^2))}{\sqrt{116m_\pi ^2/M_\sigma ^2}}}`$
with $`M_\sigma =1.067`$ GeV, $`G_1=1.378`$ GeV, $`\beta =0.7`$ GeV, $`G_2=0.0036`$ GeV, $`\mathrm{\Lambda }=3.5`$ GeV<sup>-2</sup> and $`s_0=2.8`$ GeV<sup>2</sup>.
The second is ,
$`F_\sigma `$ $`=`$ $`{\displaystyle \frac{e^{2i\varphi }1}{2i}}+{\displaystyle \frac{g_1\rho _1e^{2i\varphi }}{M_R^2si(\rho _1g_1+\rho _2g_2)}},`$
$`\mathrm{with}`$ (19)
$`e^{2i\varphi }={\displaystyle \frac{1+a_1s+a_2s^2+i\rho _1[b_1(sM_\pi ^2/2)+b_2s^2]}{1+a_1s+a_2s^2i\rho _1[b_1(sM_\pi ^2/2)+b_2s^2]}},`$
where $`a_1=0.3853GeV^2`$, $`a_2=0.4237GeV^4`$, $`b_1=3.696GeV^2`$, $`b_2=1.462GeV^4`$, $`g_1=0.1108`$, $`g_2=0.4229`$, $`M_R=0.9535GeV`$, $`\rho _1=\sqrt{14m_\pi ^2/s}`$, $`\rho _2=\sqrt{14m_\kappa ^2/s}`$, and s is the invariant mass squared of the system. This one is in fact the full $`\pi \pi `$ S-wave scattering amplitude corresponding to CERN-Münich $`\pi \pi `$ S-wave phase shifts and is very close to the AMP amplitude, and hence includes contributions from several $`0^{++}`$ resonances.
It is noted that when we only concern the relative values of the branching ratios, as in most parts of this work we do not need the concrete coupling constants in eqs.(S0.Ex14) through (S0.Ex37) and our numerical results of branching ratios may differ from the real values by a constant. Obviously this does not affect our conclusion at all.
Below we will present our results in graphs and make also discussions. In next section, we will summarize what we have learned from this investigation.
1. In Fig.1 we present the relative branching ratios of $`B(0^{++}\sigma \sigma 4\pi )`$ which are normalized by $`B(0^{++}\sigma \sigma 4\pi ^0)`$. It is obvious that the ratios of $`B(\pi ^+\pi ^{}\pi ^0\pi ^0):B(\pi ^+\pi ^{}\pi ^+\pi ^{}):B(\pi ^0\pi ^0\pi ^0\pi ^0)`$ decline from the naive counting 4:4:1. The curves $`\pi ^+\pi ^{}\pi ^0\pi ^0a`$ and $`\pi ^0\pi ^0\pi ^0\pi ^0b`$ are evaluated in terms of eq.(16) for the $`\sigma `$propagator while the others correspond to eq.(S0.Ex42). The same conventions apply to Fig.4 and Fig.6.
2. To study the significance of the interference effects in $`M4\pi `$, we define a quantity $`R`$ as
$$R=\frac{[LIPS]_i|A_i|^2}{[LIPS]|_iA_i|^2},$$
(20)
where the sum runs over all channels which contribute to the same $`4\pi `$ products, so the channels may interfere among each other and the integration is carried out over the Lorentz Invariant Phase Space (LIPS).
Fig.2 gives the R-values for $`0^{++}\rho \rho \pi ^+\pi ^{}\pi ^0\pi ^0`$ and $`0^{++}\rho \rho \pi ^+\pi ^{}\pi ^+\pi ^{}`$, because $`0^{++}\rho \rho \pi ^0\pi ^0\pi ^0\pi ^0`$ is forbidden.
3. Fig.3 is for $`0^+\rho \rho 4\pi `$ in analog to Fig.2.
4. Fig.4 is for $`2^{++}\sigma \sigma 4\pi `$ and Fig.5 is for $`2^{++}\rho \rho 4\pi `$.
5. Recently, the BES collaboration discovered a new possible channel $`f_2\sigma `$ in the $`2^{++}`$resonance decay, thus as an intermediate state, it can also contribute to the $`4\pi `$ final states. Fig.6 shows the relative branching ratios of $`2^{++}f_2\sigma \pi ^+\pi ^{}\pi ^0\pi ^0,\pi ^+\pi ^{}\pi ^+\pi ^{},\pi ^0\pi ^0\pi ^0\pi ^0`$ respectively where as usual they are also normalized by $`B(2^{++}\pi ^0\pi ^0\pi ^0\pi ^0)`$.
6. As aforementioned, we deliberately ignore the interference among different intermediate channels. It is true if one of the channels prevails over the others. Here we study the interference between $`\sigma \sigma `$ and $`\rho \rho `$ intermediate channels in $`0^{++}4\pi `$ decays. This theoretical estimation depends on the effective couplings $`g^{^{}}`$ and $`g`$ which are formulated in eq.(15). In Fig.7 and Fig.8 we deal with the $`\pi ^+\pi ^{}\pi ^0\pi ^0`$ and $`\pi ^0\pi ^0\pi ^0\pi ^0`$ final states respectively. And for the convenience, we compute $`|M|_{\sigma \sigma }^2\mathrm{according}\mathrm{to}\mathrm{eq}.(\text{S0.Ex8})`$ and $`|M|_{\rho \rho }^2\mathrm{according}\mathrm{to}\mathrm{eq}.(\text{S0.Ex11})`$, then obtain the interference term as $`B_I=`$$`(|M|^2|M|_{\rho \rho }^2|M|_{\sigma \sigma }^2)`$ for various masses of the parent meson where the formula for $`|M|^2`$ is given in eq.$`(\text{S0.Ex33})`$. In Fig.7, g’/g takes 4.5 whereas 8.5 in Fig.8, we choose the form(16) for a $`\sigma `$propagator.
For the $`f_0(1500)`$, $`f_0(1750)`$ and $`f_0(2100)`$, the $`\sigma \sigma `$ intermediate state dominates over the $`\rho \rho `$ intermediate states. Hence the interference between $`\sigma \sigma `$ and $`\rho \rho `$ intermediate states is negligible for these states.
Now we apply our results to estimate the branching ratios of the channels which are not measured yet.
Below we tabulate the numerical results for some branching ratios of $`J/\mathrm{\Psi }`$ radiative decays to $`4\pi `$ states where only one of the three $`4\pi `$-modes ($`\pi ^+\pi ^{}\pi ^+\pi ^{}`$) is experimentally measured. In table 1, the first column contains the values of $`B(\pi ^+\pi ^{}\pi ^+\pi ^{})`$ measured by the BES collaboration , while the other two columns are for the ones evaluated in terms of our scheme where interference effects are carefully considered. The table 2 is similar but based on the MARKIII data .
Table 1. Branching ratios of $`J/\mathrm{\Psi }`$ radiative decays to $`4\pi `$ based on the BES data
| parent meson | $`B(\pi ^+\pi ^{}\pi ^+\pi ^{})`$ (measured) | $`B(\pi ^+\pi ^{}\pi ^0\pi ^0)`$ | $`B(\pi ^0\pi ^0\pi ^0\pi ^0)`$ | $`B(4\pi )`$ |
| --- | --- | --- | --- | --- |
| $`f_0(1500)`$ | $`(3.1\pm 0.2\pm 1.1)\times 10^4`$ | $`1.75\times 10^4`$ | $`1.05\times 10^4`$ | $`5.9\times 10^4`$ |
| $`f_0(1740)`$ | $`(3.1\pm 0.2\pm 1.1)\times 10^4`$ | $`1.73\times 10^4`$ | $`1.03\times 10^4`$ | $`5.9\times 10^4`$ |
| $`f_0(2100)`$ | $`(5.1\pm 0.3\pm 1.8)\times 10^4`$ | $`3.08\times 10^4`$ | $`1.71\times 10^4`$ | $`9.9\times 10^4`$ |
| $`f_2(1950)`$ | $`(5.5\pm 0.3\pm 1.9)\times 10^4`$ | $`5.58\times 10^4`$ | $`1.63\times 10^4`$ | $`12.7\times 10^4`$ |
Table 2. Branching ratios of $`J/\mathrm{\Psi }`$ radiative decays to $`4\pi `$ based on the MARK III data
| parent meson | $`B(\pi ^+\pi ^{}\pi ^+\pi ^{})`$ (measured) | $`B(\pi ^+\pi ^{}\pi ^0\pi ^0)`$ | $`B(\pi ^0\pi ^0\pi ^0\pi ^0)`$ | $`B(4\pi )`$ |
| --- | --- | --- | --- | --- |
| $`f_0(1505)`$ | $`(2.5\pm 0.4)\times 10^4`$ | $`1.41\times 10^4`$ | $`0.84\times 10^4`$ | $`4.8\times 10^4`$ |
| $`f_0(1750)`$ | $`(4.3\pm 0.6)\times 10^4`$ | $`2.41\times 10^4`$ | $`1.43\times 10^4`$ | $`8.1\times 10^4`$ |
| $`f_0(2104)`$ | $`(3.0\pm 0.8)\times 10^4`$ | $`1.82\times 10^4`$ | $`1.00\times 10^4`$ | $`5.8\times 10^4`$ |
Using Crystal Barrel results and our results here, we get the relative branching ratio of $`Br(f_0(1500)4\pi )/Br(f_0(1500)2\pi )`$ to be ($`2.1\pm 0.6`$), which is compatible with the result $`1.5\pm 0.4`$ from $`\pi \pi `$ scattering phase shifts, instead of $`3.3\pm 0.8`$ by assuming Eq.(1).
In conclusion, our numerical results indicate that interference effects would make the ratios
$$B(M\pi ^+\pi ^{}\pi ^0\pi ^0):B(M\pi ^+\pi ^{}\pi ^+\pi ^{}):B(M\pi ^0\pi ^0\pi ^0\pi ^0)$$
much deviating from the naive counting 4:4:1 for isoscalar $`0^{++}`$ and $`2^{++}`$ mesons. The graphs provided in this work can serve as a standard reference that once one of the $`4\pi `$ modes $`\pi ^+\pi ^{}\pi ^0\pi ^0,\pi ^+\pi ^{}\pi ^+\pi ^{},\pi ^0\pi ^0\pi ^0\pi ^0`$ is measured, we can determine the branching ratios of the other modes.
Acknowledgment: This work is partly supported by the National Natural Science Foundation of China (NNSFC). We would like to thank the BES collaboration for supporting their work. One of us (Zou) thanks D.V.Bugg for useful discussions. |
warning/0001/hep-ph0001311.html | ar5iv | text | # Contents
## 1 Introduction
During many years neutrino physics was a very important branch of elementary particle physics. In the last few years the interest to neutrinos particularly increased. This is connected first of all with the success of the Super-Kamiokande experiment in which very convincing evidence in favour of oscillations of atmospheric neutrinos were obtained.
It is plausible that tiny neutrino masses and neutrino mixing are connected with the new large scale in physics. This scale determines the smallness of neutrino masses with respect to the masses of charged leptons and quarks. In such a scenario neutrinos with definite masses are truly neutral Majorana particles (quarks and leptons have charges and are Dirac particles) It is evident, however, that many new experiments are necessary to reveal the real origin of neutrino masses and mixing.
Experimental neutrino physics is a very difficult and exciting field of research. Now it is a time when many new ideas and methods are being proposed. In CERN and other laboratories projects of new neutrino experiments are developing. Possibilities of new neutrino facility, neutrino factory, are investigated in different laboratories. Thus it is a very appropriate time to discuss neutrino physics at the CERN-JINR school.
In these lectures I will consider different possibilities of neutrino mixing. Then, I will discuss in some details neutrino oscillations in vacuum and in matter. In the last part of the lectures I will consider the present experimental situation.
I tried to give in these lectures some important results and details of derivation of some results. I hope that lectures will be useful for those who want to study physics of massive neutrinos. More results and details can be found in the books and reviews .
Most references to original papers can be found in
## 2 Neutrino mixing
According to the Standard Model of electroweak interaction the Lagrangian of the interaction of neutrinos with other particles is given by the Charged Current (CC) and the Neutral Current (NC) Lagrangians:
$`_I^{\mathrm{CC}}={\displaystyle \frac{g}{2\sqrt{2}}}j_\alpha ^{\mathrm{CC}}W^\alpha +\mathrm{h}.\mathrm{c}.,`$ (2.1)
$`_I^{\mathrm{NC}}={\displaystyle \frac{g}{2\mathrm{cos}\theta _W}}j_\alpha ^{\mathrm{NC}}Z^\alpha .`$ (2.2)
Here $`g`$ is the electroweak interaction constant, $`\theta _W`$ is the weak (Weinberg) angle and $`W^\alpha `$ and $`Z^\alpha `$ are the fields of the $`W^+`$ and $`Z^0`$ vector bosons. If neutrino masses are equal to zero in this case CC and NC interactions conserve electron $`L_e`$, muon $`L_\mu `$ and tauon $`L_\tau `$ lepton numbers
$`{\displaystyle L_e}=const,{\displaystyle L_\mu }=const,{\displaystyle L_\tau }=const`$ (2.3)
The values of the lepton numbers of charged leptons, neutrinos and other particles are given in the Table 2.
According to the *neutrino mixing hypothesis* masses of neutrinos are different from zero and *neutrino mass term* does not conserve lepton numbers. For the fields of $`\nu _{lL}`$ that enter into CC and NC Lagrangians (2.1) and (2.2) we have, in this case,
$$\nu _{lL}=\underset{i}{}U_{li}\nu _{iL}$$
(2.4)
where $`\nu _i`$ is the field of neutrino with mass $`m_i`$ and $`U`$ is the unitary mixing matrix.
The relation (2.4) leads to violation of lepton numbers due to small neutrino mass differences and neutrino mixing. To reveal such effects special experiments (neutrino oscillation experiments, neutrinoless double $`\beta `$-decay experiments and others) are necessary. We will discuss such experiments later. Now we will consider different possibilities of neutrino mixing.
Let us notice first of all that the relation 2.4 is similar to the analogous relation in the quark case. The standard CC current of quarks have the form
$$j_\alpha ^{CC}=2(\overline{u_L}\gamma _\alpha d_L^{}+\overline{c}_L\gamma _\alpha s_L^{}+\overline{t}_L\gamma _\alpha b_L^{})$$
(2.5)
Here
$$d_L^{}=\underset{q=d,s,b}{}V_{uq}q_L,s_L^{}=\underset{q=d,s,b}{}V_{cq}q_L,b_L^{}=\underset{q=d,s,b}{}V_{tq}q_L$$
(2.6)
where $`V`$ is Cabibbo–Kobayashi–Maskawa quark mixing matrix. There can be, however, a fundamental difference between mixing of quarks and neutrino mixing. Quarks are charged four–component Dirac particles: quarks and antiquarks have different charges.
For neutrinos with definite masses there are two possibilities:
1. In case the total lepton number $`L=L_e+L_\mu +L_\tau `$ is conserved, neutrino with definite masses $`\nu _i`$ are four–component Dirac particles (neutrinos and antineutrinos differ by the sign of $`L`$);
2. If there are no conserved lepton numbers, neutrinos with definite masses $`\nu _i`$ are two-component Majorana particles (there are no quantum numbers in this case that can allow to distinguish neutrino from antineutrino).
The nature of neutrino masses and the character of neutrino mixing is determined by the neutrino mass term.
### 2.1 Dirac Neutrinos
If the neutrino mass term is generated by the same standard Higgs mechanism, that is responsible for the mass generation of quarks and charged leptons, then for the neutrino mass term we have
$$^\mathrm{D}=\underset{l,l^{}}{}\overline{\nu _{l^{}R}}M_{l^{}l}^\mathrm{D}\nu _{lL}+\mathrm{h}.\mathrm{c}.$$
(2.7)
where $`M^D`$ is the complex $`3\times 3`$ matrix and $`\nu _{lR}`$ is the right–handed singlet. In the case of mass term (2.7) the total Lagrangian is invariant under global gauge invariance
$$\nu _{lL}e^{i\alpha }\nu _{lL},\nu _{lR}e^{i\alpha }\nu _{lR},le^{i\alpha }l,$$
(2.8)
where $`\alpha `$ is a constant that does not depend on the flavor index $`l`$. The invariance under the transformation (2.8) means that the total lepton number $`L=L_e+L_\mu +L_\tau `$ is conserved
$$L=const$$
(2.9)
Now let us diagonalize the mass term (2.7). The complex matrix $`M^\mathrm{D}`$ can be diagonalized by biunitary transformation
$$M^\mathrm{D}=VmU^{},$$
(2.10)
where $`V^{}V=1`$, $`U^{}U=1`$ and $`m_{ik}=m_i\delta _{ik},m_i>0`$.
With the help of (2.10), from (2.7) for the neutrino mass term we obtain the standard expression
$$^\mathrm{D}=\underset{l^{},l,i}{}\overline{\nu _{l^{}R}}V_{l^{}i}m_i(U^{})_{il}\nu _{lL}+\mathrm{h}.\mathrm{c}.=\underset{i=1}{\overset{3}{}}m_i\overline{\nu _i}\nu _i$$
(2.11)
Here
$$\nu _i=\nu _{iL}+\nu _{iR}(i=1,2,3)$$
and
$$\nu _{iL}=\mathrm{\Sigma }_l(U^{})_{il}\nu _{lL}$$
$$\nu _{iR}=\mathrm{\Sigma }_l(V^{})_{il}\nu _{lR}$$
For the neutrino mixing we have
$$\nu _{lL}=\underset{i}{}U_{li}\nu _{iL}$$
(2.12)
Processes in which the total lepton number is conserved, like $`\mu e+\gamma `$ and others, are, in principle, allowed in the case of mixing of Dirac massive neutrinos. It can be shown, however, that the probabilities of such processes are much smaller than the experimental upper bounds.
Neutrinoless double $`\beta `$–decay,
$$(A,Z)(A,Z+2)+e^{}+e^{},$$
due to the conservation of the total lepton number is forbidden in the case of Dirac massive neutrinos.
### 2.2 Majorana neutrinos
Neutrino mass terms that are generated in the framework of the models beyond the Standard Model, like the Grand Unified SO(10) Model, do not conserve lepton numbers $`L_e`$, $`L_\mu `$ and $`L_\tau `$. Let us build the most general neutrino mass term that does not conserve $`L_e`$, $`L_\mu `$ and $`L_\tau `$.
Neutrino mass term is a linear combination of the products of left–handed and right-handed components of neutrino fields. Notice that $`(\nu _L)^C=C(\overline{\nu }_L)^T`$ is the right–handed component and $`(\nu _R)^C=C(\overline{\nu }_R)^T`$ is the left–handed component. Here C is the charge conjugation matrix, that satisfies the relations $`C\gamma _\alpha ^TC^1=\gamma _\alpha `$, $`C^T=C`$, $`C^{}C=1`$ <sup>2</sup><sup>2</sup>2In fact, $`L`$ and $`R`$ components satisfy the relations
$$\frac{1+\gamma _5}{2}\nu _L=0\frac{1\gamma _5}{2}\nu _R=0$$
From the first of these relations we have $`\overline{\nu }_L(1\gamma _5)/2=0`$. Further, from this last relation we obtain $`[(1\gamma _5)/2]^T\overline{\nu }_L^T=0.`$ Multiplying this relation by the matrix $`C`$ from the left and taking into account that $`C\gamma _5^TC^1=\gamma ^5`$ we have $`[(1\gamma _5)/2](\nu _L)^C=0`$. Thus, $`(\nu _L)^C`$ is right–handed component. Analogously we can show that $`(\nu _R)^C`$ is left-handed component.
The most general Lorentz–invariant neutrino mass term in which flavor neutrino fields $`\nu _{lL}`$ and right–handed singlet fields $`\nu _{lR}`$ enter has the following form
$$^{\mathrm{D}\mathrm{M}}=\frac{1}{2}\overline{(n_L)^C}Mn_L+\mathrm{h}.\mathrm{c}.$$
(2.13)
Here
$$n_L=\left(\begin{array}{c}\nu _L^{}\\ (\nu _R^{})^C\end{array}\right)\text{with}\nu _L^{}=\left(\begin{array}{c}\nu _{eL}\\ \nu _{\mu L}\\ \nu _{\tau L}\end{array}\right)\text{and}\nu _R^{}=\left(\begin{array}{c}\nu _{eR}\\ \nu _{\mu R}\\ \nu _{\tau R}\end{array}\right),$$
(2.14)
$`M`$ is complex $`6\times 6`$ matrix. Taking into account that $`\overline{(\nu _L)^C}=\nu _L^TC^1`$ we have
$$^{\mathrm{D}\mathrm{M}}=\frac{1}{2}n_L^T𝒞^1Mn_L+\mathrm{h}.\mathrm{c}..$$
(2.15)
From this expression it is obvious that there is no global gauge invariance in the case of the mass term (2.13), i.e. that the mass term (2.13) does not conserve lepton numbers.
The matrix $`M`$ is symmetric. In fact, taking into account the commutation properties of fermion fields we have
$$n_L^T𝒞^1Mn_L=n_L^T(𝒞^T)^1M^Tn_L=n_L^T𝒞^1M^Tn_L.$$
(2.16)
From this relation it follows that
$$M^T=M$$
The symmetric $`6\times 6`$ matrix can be presented in the form
$$M=\left(\begin{array}{cc}M_L& (M_\mathrm{D})^T\\ M_\mathrm{D}& M_R\end{array}\right).$$
(2.17)
where $`M_L=M_L^T`$, $`M_R=M_R^T`$ and $`M^D`$ are $`3\times 3`$ matrices. With the help of (2.17) for the mass term (2.15) we have
$$^{\mathrm{D}\mathrm{M}}=_L^\mathrm{M}+^\mathrm{D}+_R^\mathrm{M}.$$
(2.18)
Here $`^D`$ is the Dirac mass term, that we have considered before, and the new terms
$$_L^\mathrm{M}=\frac{1}{2}\underset{l^{},l}{}\overline{(\nu _{l^{}L})^c}M_{l^{}l}^L\nu _{lL}+\mathrm{h}.\mathrm{c}.,$$
(2.19)
$$_R^\mathrm{M}=\frac{1}{2}\underset{l^{},l}{}\overline{(\nu _{l^{}R})^c}M_{l^{}l}^R\nu _{lR}+\mathrm{h}.\mathrm{c}.,$$
(2.20)
which do not conserve lepton numbers are called left–handed and right–handed Majorana mass terms, respectively. The mass term (2.13) is called Dirac–Majorana mass term.
A symmetrical matrix can be diagonalized with the help of unitary transformation
$$M=(U^{})^TmU^{}.$$
Here $`U`$ is unitary matrix and $`m_{ik}=m_i\delta _{ik},m_i>0`$. Using the relation (11) we can write the mass term (2.15) in the standard form
$$^{\mathrm{D}\mathrm{M}}=\frac{1}{2}(\overline{U^{}n_L})^CmU^+n_L+\mathrm{h}.\mathrm{c}.=\frac{1}{2}\overline{\nu }m\nu =\frac{1}{2}\underset{i=1}{\overset{6}{}}m_i\overline{\nu }_i\nu _i,$$
(2.21)
where
$$\nu =U^+n_L+(U^+n_L)^C=\left(\begin{array}{c}\nu _1\\ \nu _2\\ \mathrm{}\\ \nu _6\end{array}\right),$$
(2.22)
Thus the fields $`\nu _i`$ (i=1,2…6) are the fields of neutrinos with mass $`m_i`$. From (2.22) it follows that the fields $`\nu _i`$ satisfy the Majorana condition
$$\nu _i^C=\nu _i,$$
(2.23)
Let us obtain now the relation that connects the left–handed flavor fields $`\nu _{lL}`$ with the massive fields $`\nu _{iL}`$. From (2.22) for the left–handed components we have
$$n_L=U\nu _L.$$
(2.24)
From this relation for the flavor field $`\nu _{lL}`$ it follows
$$\nu _{lL}=\underset{i=1}{\overset{6}{}}U_{li}\nu _{iL}(l=e,\mu ,\tau ),$$
(2.25)
Thus, in the case of Dirac–Majorana mass term, the flavor fields are linear combinations of left–handed components of six massive Majorana fields. From (2.25) it follows that the fields $`\nu _{lR}^C`$ are orthogonal linear combinations of the same massive Majorana fields
$$(\nu _{lR})^C=\underset{i=1}{\overset{6}{}}U_{\overline{l}i}\nu _{iL}.$$
(2.26)
In the case of Majorana field particles and antiparticles, quanta of the field, are identical. In fact, for fermion fields $`\nu (x)`$ we have in general case
$$\nu (x)=\frac{1}{(2\pi )^{3/2}}\frac{1}{\sqrt{2p^0}}\left[c_r(p)u^r(p)e^{ipx}+d_r^{}(p)C\left(\overline{u}^r(p)\right)^Te^{ipx}\right]d^3p$$
(2.27)
where $`c_r(p)(d_r^{}(p))`$ is the operator of absorption of particle (creation of antiparticle) with momentum $`p`$ and helicity $`r`$. If the field $`\nu (x)`$ satisfies the Majorana condition (2.23), then we have
$$c_r(p)=d_r(p)$$
(2.28)
Let us stress that it is natural that the neutrinos with definite masses in the case of Dirac–Majorana mass term are Majorana neutrinos: in fact there are no conserved quantum numbers that could allow us to distinguish particles and antiparticles.
### 2.3 The simplest case of one generation (Majorana neutrinos)
It is instructive to consider in detail the Dirac–Majorana mass term in the simplest case of one generation. We have
$`^{\mathrm{D}\mathrm{M}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}m_L(\overline{\nu }_L)^c\nu _Lm_\mathrm{D}\overline{\nu }_R\nu _L{\displaystyle \frac{1}{2}}m_R\overline{\nu }_R(\nu _R)^c+\mathrm{h}.\mathrm{c}.`$ (2.29)
$`=`$ $`{\displaystyle \frac{1}{2}}(\overline{n}_L)^cMn_L+\mathrm{h}.\mathrm{c}.,`$
where
$$n_L\left(\begin{array}{c}\nu _L\\ (\nu _R)^c\end{array}\right),M\left(\begin{array}{cc}m_L& m_\mathrm{D}\\ m_\mathrm{D}& m_R\end{array}\right).$$
(2.30)
Let us assume that the parameters $`m_L,m_R`$ and $`m_D`$ are real (the case of CP invariance). In order to diagonalize the mass term (2.29) let us write the matrix $`M`$ in the form
$$M=\frac{1}{2}\mathrm{Tr}M+\underset{¯}{M},$$
(2.31)
where Tr $`M=m_L+m_D`$ and
$$\underset{¯}{M}=\left(\begin{array}{cc}\frac{1}{2}(m_Rm_L)& m_\mathrm{D}\\ m_\mathrm{D}& \frac{1}{2}(m_Rm_L)\end{array}\right).$$
(2.32)
For the symmetrical real matrix we have
$$\underset{¯}{M}=𝒪\underset{¯}{m}𝒪^T.$$
(2.33)
Here
$$𝒪=\left(\begin{array}{cc}\hfill \mathrm{cos}\vartheta & \hfill \mathrm{sin}\vartheta \\ \hfill \mathrm{sin}\vartheta & \hfill \mathrm{cos}\vartheta \end{array}\right)$$
(2.34)
is an orthogonal matrix, and $`\underset{¯}{m}_{ik}=\underset{¯}{m}_i\delta _{ik}`$, where
$$\underset{¯}{m}_{1,2}=\frac{1}{2}\sqrt{(m_Rm_L)^2+4m_D^2}$$
(2.35)
are eigenvalues of the matrix $`\underset{¯}{M}`$.
From (2.33), (2.34) and (2.35) for the parameters $`\mathrm{cos}\vartheta `$ and $`\mathrm{sin}\vartheta `$ we easily find the following expressions
$$\mathrm{cos}2\vartheta =\frac{m_Rm_L}{\sqrt{(m_Rm_L)^2+4m_D^2}},\mathrm{tan}2\vartheta =\frac{2m_D}{(m_Rm_L)}.$$
(2.36)
For the matrix $`M`$ from (2.33) and (2.35) we have
$$M=Om^{}O^T$$
where
$$m_{1,2}^{}=\frac{1}{2}(m_R+m_L)\sqrt{(m_Rm_L)^2+4m_D^2}.$$
(2.37)
The eigenvalues $`m_i^{}`$ can be positive or negative. Let us write
$$m_i^{}=m_i\eta _i,$$
(2.38)
where $`m_i=|m_i|`$ and $`\eta _i`$ is the sign of the i-eigenvalue. With the help of (2.33) and (2.38) we have
$$M=(U^{})^TmU^{}$$
Here
$$U^{}=\sqrt{\eta }O^T$$
where $`\sqrt{\eta }`$ takes the values 1 and $`i`$.
Now using the general formulas (2.21) and (2.22) for the mass term we have
$$^{DM}=\frac{1}{2}\underset{i=1,2}{}m_i\overline{\nu }_i\nu _i$$
(2.39)
Here $`\nu _i=\nu _i^C`$ is the field of the Majorana particles with mass $`m_i`$. The fields $`\nu _L`$ and $`(\nu _R)^C`$ are connected with massive fields by the relation
$$\left(\begin{array}{c}\nu _L\\ (\nu _R)^C\end{array}\right)=U\left(\begin{array}{c}\nu _{1L}\\ \nu _{2L}\end{array}\right),$$
(2.40)
where $`U=O(\sqrt{\eta })^{}`$ is a $`2\times 2`$ mixing matrix.
Let us consider now three special cases.
1. No mixing
Assume $`m_D`$ = 0. In this case $`\theta =0`$, $`m_1=m_L`$, $`m_2=m_R`$ and $`\eta =1`$ (assuming that $`m_L`$ and $`m_R`$ are positive). From (2.40) we have
$$\nu _L=\nu _{1L}(\nu _R)^C=\nu _{2L}.$$
(2.41)
Thus, if $`m_\mathrm{D}=0`$ there is no mixing. For the Majorana fields $`\nu _1`$ and $`\nu _2`$ we have
$$\nu _1=\nu _L+(\nu _L)^C$$
(2.42)
$$\nu _2=\nu _R+(\nu _R)^C.$$
(2.43)
2. Maximal mixing
Assume $`m_R=m_L`$, $`m_D0`$. From Eq. (2.36), (2.37) and (2.40) we have
$$\theta =\frac{\pi }{4},m_{1,2}=m_Lm_D.$$
(2.44)
(assuming $`|m_D|<m_L`$) and
$$\nu _L=\frac{1}{\sqrt{2}}\nu _{1L}+\frac{1}{\sqrt{2}}\nu _{2L};(\nu _R)^C=\frac{1}{\sqrt{2}}\nu _{1L}+\frac{1}{\sqrt{2}}\nu _{2L}.$$
(2.45)
Thus if the diagonal elements of the mass matrix $`M`$ are equal, then we have maximal mixing.
3. See–saw mechanism of neutrino mass generation
Assume $`m_L=0`$ and
$$m_Dm_R$$
(2.46)
From (2.35) and (2.37) we have in this case
$$m_1\frac{m_D^2}{m_R},m_2m_R,\theta \frac{m_D}{m_R}(\eta _1=1,\eta _2=1).$$
(2.47)
Neglecting terms linear in $`m_D/m_R1`$, from (2.40) we have
$$\nu _Li\nu _{1L},(\nu _R)^C\nu _{2L}.$$
(2.48)
For the Majorana fields we have
$$\nu _1i\nu _Li(\nu _L)^C,\nu _2=\nu _R+(\nu _R)^C.$$
(2.49)
Thus if the condition (2.46) is satisfied, in the spectrum of masses of Majorana particles there are one light particle with the mass $`m_1<<m_D`$ and one heavy particle with the mass $`m_1>>m_D`$. The condition $`m_L=0`$ means that the lepton number is violated only by the right–handed term $`\frac{1}{2}m_R\overline{\nu }_R(\nu _R)^C`$ that is characterized by the large mass $`m_R`$. It is natural to assume that the parameter $`m_D`$ which characterizes the Dirac term $`m_D\overline{\nu }_R\nu _L`$ is of the order of lepton or quark masses. The mass of the light Majorana neutrino $`m_1`$ will be in this case much smaller than the mass of lepton or quark. This is famous see-saw mechanism. This mechanism connects the smallness of the neutrino masses with respect to the masses of other fundamental fermions with violation of the lepton numbers at very large scale (usually $`m_DM_{GUT}10^{16}`$ GeV.
In the case of the see–saw for three families in the spectrum of masses of Majorana particles there are three light masses $`m_1,m_2,m_3`$ (masses of neutrinos) and three very heavy masses $`M_1,M_2,M_3`$. Masses of neutrinos are connected with the masses of heavy Majorana particles by the see–saw relation
$$m_i\frac{(m_\mathrm{f}^i)^2}{M_i}m_\mathrm{f}^i(i=1,2,3).$$
(2.50)
where $`m_\mathrm{f}^i`$ is the mass of lepton or quark in $`i`$-family. The see–saw mechanism is a plausible explanation of the experimentally observed smallness of neutrino masses. Let us stress that if neutrino masses are of the see-saw origin then
a. neutrinos with definite masses are Majorana particles;
b. there are three massive neutrinos;
c. there must be a hierarchy of neutrino masses $`m_1m_2m_3`$.
## 3 Neutrino oscillations
The most important consequences of the neutrino mixing are so called neutrino oscillations. Neutrino oscillations were first considered by B. Pontecorvo many years ago in 1957-58. Only one type of neutrino was known at that time and there was general belief that neutrino is a massless two–component particle. B. Pontecorvo draw attention that there is no known principle which requires neutrino to be massless (like gauge invariance for the photon) and that the investigation of neutrino oscillations is a very sensitive method to search for effects of small neutrino masses. We will consider here in detail the phenomenon of neutrino oscillations.
Assume that there is neutrino mixing
$$\nu _{\alpha L}=\underset{i}{}U_{\alpha i}\nu _{iL}.$$
(3.51)
where $`U^{}U=1`$ and $`\nu _i`$ is the field of neutrino (Dirac or Majorana) with the mass $`m_i`$. The field $`\nu _{\alpha L}`$ in (3.51) are flavor fields ($`\alpha =e,\mu ,\tau `$) and in general also sterile ones ($`\alpha =s_1,\mathrm{}`$). Let us assume that neutrino mass differences are small and different neutrino masses cannot be resolved in neutrino production and detection processes.
For the state of neutrino with momentum $`\stackrel{}{p}`$ we have
$$|\nu _\alpha =\underset{i}{}U_{\alpha i}^{}|\nu _i.$$
(3.52)
where $`|\nu _i`$ is the vector of state of neutrino with momentum $`\stackrel{}{p}`$, energy
$$E_i=\sqrt{p^2+m_i^2}p+\frac{m_i^2}{2p}(pm_i).$$
(3.53)
and (up to the terms $`m_i^2/p^2`$) helicity is equal to -1. If at the initial time $`t=0`$ the state of neutrino is $`|\nu _\alpha `$ at the time $`t`$ for the neutrino state we have
$$|\nu _\alpha _t=\underset{i}{}U_{\alpha i}^{}e^{iE_it}|\nu _i.$$
(3.54)
The vector $`|\nu _\alpha `$ is the superposition of the states of all types of neutrino. In fact, from (3.52), using unitarity of the mixing matrix, we have
$$|\nu _i=\underset{\alpha ^{}}{}|\nu _\alpha ^{}U_{\alpha ^{}i}.$$
(3.55)
From (3.54) and (3.55) we have
$$|\nu _\alpha _t=\underset{\alpha ^{}}{}|\nu _\alpha ^{}𝒜_{\nu _\alpha ^{};\nu _\alpha }(t).$$
(3.56)
where
$$𝒜_{\nu _\alpha ^{};\nu _\alpha }(t)=\underset{i}{}U_{\alpha ^{}i}e^{iE_it}U_{\alpha i}^{}.$$
(3.57)
is the amplitude of the transition $`\nu _\alpha \nu _\alpha ^{}`$ at the time $`t`$. The transition amplitude $`𝒜_{\alpha ^{};\alpha }(t)`$ has a simple meaning: the term $`U_{\alpha i}^{}`$ is the amplitude of the transition from the state $`|\nu _\alpha `$ to the state $`|\nu _i`$; the term $`e^{iE_it}`$ describes the evolution in the state with energy $`E_i`$; the term $`U_{\alpha ^{}i}`$ is the transition amplitude from the state $`|\nu _i`$ to the state $`|\nu _\alpha ^{}`$.
The different $`|\nu _i`$ gives coherent contribution to the amplitude $`𝒜_{\nu _\alpha ^{};\nu _\alpha }(t)`$. From (3.57) it follows that the transitions between different states can take place only if: i) at least two neutrino masses are different; ii) the mixing matrix is non–diagonal. In fact, if all neutrino masses are equal we have $`a(t)=e^{iEt}U_{\alpha ^{}i}U_{\alpha i}^{}=e^{iEt}\delta _{\alpha ^{}\alpha }`$. If the mixing matrix is diagonal (no mixing), we have $`𝒜_{\nu _\alpha ^{};\nu _\alpha }(t)=e^{iE_\alpha t}\delta _{\alpha ^{}\alpha }`$.
Let us numerate neutrino masses in such a way that $`m_1<m_2<\mathrm{}<m_n`$. For the transition probability, from (3.57), we have the following expression:
$`P_{\nu _\alpha \nu _\alpha ^{}}`$ $`=\left|{\displaystyle \underset{i}{}}U_{\alpha ^{}i}\left[\left(e^{i(E_iE_1)t}1\right)+1\right]U_{\alpha i}^{}\right|^2`$
$`=\left|\delta _{\alpha \alpha ^{}}+{\displaystyle \underset{i}{}}U_{\alpha ^{}i}U_{\alpha i}^{}\left(e^{i\mathrm{\Delta }m_{i1}^2\frac{L}{2p}}1\right)\right|^2,`$
where $`\mathrm{\Delta }m_{i1}^2=m_i^2m_1^2`$ and $`Lt`$ is the distance between neutrino source and neutrino detector. Thus the neutrino transition probability depends on the ratio $`\frac{L}{E}`$, the range of values of which is determined by the conditions of an experiment.
It follows from Eq. (3) that the transition probability depends in the general case on $`(n1)`$ neutrino mass squared differences and parameters that characterize the mixing matrix $`U`$. The $`n\times n`$ matrix $`U`$ is characterized by $`n_\theta =n(n1)/2`$ angles. The number of phases for Dirac and Majorana cases is different. If neutrino with definite masses $`\nu _i`$ are Dirac particles the number of phases is equal to $`n_\varphi ^D=(n1)(n2)/2`$. If $`\nu _i`$ are Majorana particles the number of phases is equal to $`n_\varphi ^{M_j}=n(n1)/2`$.
Notice that from (3) it follows that transition probability is invariant under the transformation
$$U_{\alpha i}e^{i\beta _\alpha }U_{\alpha i}e^{i\alpha _i}$$
(3.59)
where $`\beta _\alpha `$ and $`\alpha _i`$ are arbitrary real phases. From (3.59) it follows that the number of phases that enter into the transition probability is equal to $`n_\varphi =(n1)(n2)/2`$ in both Dirac and Majorana cases. We come to the conclusion that additional Majorana phases do not enter into the transition probability . Thus, by investigation of neutrino oscillations it is impossible to distinguish the case of Dirac neutrinos from the case of Majorana neutrinos.
Let us consider now oscillations of antineutrinos. For the vector of state of antineutrino with momentum $`\stackrel{}{p}`$ from (3.51) we have
$$|\overline{\nu }_\alpha =\underset{i}{}U_{\alpha i}|\overline{\nu }_i(\mathrm{Dirac}\mathrm{case})$$
(3.60)
$$|\overline{\nu }_\alpha =\underset{i}{}U_{\alpha i}|\nu _i(\mathrm{Majorana}\mathrm{case})$$
(3.61)
where $`|\overline{\nu }_i`$ ($`|\nu _i`$) is the state of antineutrino (neutrino) with momentum $`\stackrel{}{p}`$, energy $`E_i=\sqrt{p^2+m_i^2}p+m_i^2/2p`$ and helicity equal to +1 ( up to $`m_i^2/p^2`$ terms).
In analogy with (3.57) for the amplitude of the transition $`\overline{\nu }_\alpha \overline{\nu }_\alpha ^{}`$ in both Dirac and Majorana cases we have
$$𝒜_{\overline{\nu }_\alpha ^{};\overline{\nu }_\alpha }(t)=\underset{i}{}U_{\alpha ^{}i}^{}e^{iE_it}U_{\alpha i}.$$
(3.62)
If we compare (3.57) and (3.62) we come to the conclusion that
$$𝒜_{\overline{\nu }_\alpha ^{};\overline{\nu }_\alpha }(t)=𝒜_{\nu _\alpha ;\nu _\alpha ^{}}(t).$$
(3.63)
Thus for the transition probabilities we have the following relation
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\mathrm{P}(\overline{\nu }_\alpha ^{}\overline{\nu }_\alpha ).$$
(3.64)
This relation is the consequence of CPT invariance. If CP invariance in the lepton sector takes place then for Dirac neutrinos we have
$$U_{\alpha i}^{}=U_{\alpha i}$$
(3.65)
while for Majorana neutrinos, from CP invariance, we have
$$U_{\alpha i}\eta _i=U_{\alpha i}^{};$$
(3.66)
where $`\eta _i=\pm i`$ is the CP parity of the Majorana neutrino with mass $`m_i`$. From (3.57), (3.63), (3.65) and (3.66) it follows that in case of CP invariance we have
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\mathrm{P}(\overline{\nu }_\alpha \overline{\nu }_\alpha ^{}).$$
(3.67)
Let us go back to the Eq. (3). It is obvious from (3) that if the conditions of an experiment are such that $`\mathrm{\Delta }m_{i1}^2\frac{L}{p}1`$ for all $`i`$ then neutrino oscillations cannot be observed. To observe neutrino oscillations it is necessary that for at least one neutrino mass squared difference the condition $`\mathrm{\Delta }m^2\frac{L}{p}>1`$ is satisfied. We will discuss this condition later.
### 3.1 Two neutrino oscillations
Let us consider in details the simplest case of the oscillations between two neutrinos $`\nu _\alpha \nu _\alpha ^{}`$ ($`\alpha ^{}\alpha ;\alpha ,\alpha ^{}`$ are equal to $`\mu `$, e or $`\tau `$, $`\mu `$,…). The index $`i`$ in Eq. (3) takes values 1 and 2 and for the transition probability we have
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=|\delta _{\alpha ^{}\alpha }+U_{\alpha _2^{}}U_{\alpha _2}^{}(e^{i\mathrm{\Delta }m_{21}^2\frac{L}{2p}}1)|^2$$
(3.68)
For $`\alpha ^{}\alpha `$ we have from (3.68)
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\mathrm{P}(\nu _\alpha ^{}\nu _\alpha )=\frac{1}{2}\mathrm{A}_{\alpha ^{}\alpha }(1\mathrm{cos}\mathrm{\Delta }m^2\frac{L}{2p})$$
(3.69)
Here the amplitude of oscillations is equal to
$$\mathrm{A}_{\alpha ^{};\alpha }=4|U_{\alpha ^{}2}|^2|U_{\alpha 2}|^2$$
(3.70)
and $`\mathrm{\Delta }m^2=m_2^2m_1^2`$. Due to unitarity of the mixing matrix
$$|U_{\alpha 2}|^2+|U_{\alpha ^{}2}|^2=1(\alpha ^{}\alpha )$$
(3.71)
Let us introduce the mixing angle $`\theta `$
$$|U_{\alpha 2}|^2=\mathrm{sin}^2\theta |U_{\alpha ^{}2}|^2=\mathrm{cos}^2\theta $$
(3.72)
Thus the oscillation amplitude $`\mathrm{A}_{\alpha ^{};\alpha }`$ is equal to
$$\mathrm{A}_{\alpha ^{};\alpha }=\mathrm{sin}^22\theta $$
(3.73)
The survival probabilities $`\mathrm{P}(\nu _\alpha \nu _\alpha )`$ and $`\mathrm{P}(\nu _\alpha ^{}\nu _\alpha ^{})`$ can be obtained from (3.68) or from the condition of the conservation of the total probability $`\mathrm{P}(\nu _\alpha \nu _\alpha )+\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=1`$. We have
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=\mathrm{P}(\nu _\alpha ^{}\nu _\alpha ^{})=1\frac{1}{2}\mathrm{sin}^22\theta (1\mathrm{cos}\frac{\mathrm{\Delta }m^2L}{2p})$$
(3.74)
Thus in the case of two neutrinos the transition probabilities are characterized by two parameters $`\mathrm{sin}^22\theta `$ and $`\mathrm{\Delta }m^2`$.
Let us notice that in the case of transitions between two neutrinos only moduli of the elements of the mixing matrix enter into expressions for the transition probabilities. This means that in this case the CP relation (3.64) is satisfied automatically. Thus, in order to observe effects of CP violation in the lepton sector the transitions between three neutrinos must take place (this is similar to the quark case: for two families of quarks CP is conserved due to unitarity of the mixing matrix).
We also notice that the expression (3.69) for the transition probability can be written in the form
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\frac{1}{2}\mathrm{sin}^22\theta \left(1\mathrm{cos}2\pi \frac{L}{L_0}\right)$$
(3.75)
where
$$L_0=4\pi \frac{E}{\mathrm{\Delta }m}$$
(3.76)
is the oscillation length. The expression (3.69) is written in the units $`\mathrm{}=c=1`$. We can write it in the form
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\frac{1}{2}\mathrm{sin}^22\theta \left(1\mathrm{cos}\mathrm{\hspace{0.17em}2.54}\mathrm{\Delta }m^2\frac{E}{L}\right)$$
(3.77)
where $`\mathrm{\Delta }m^2`$ is neutrino mass squared difference in eV<sup>2</sup>, $`L`$ is the distance in m (km) and $`E`$ is the neutrino energy in MeV (GeV). For the oscillation length we have
$$L_0=2.47\frac{E(\mathrm{MeV})}{\mathrm{\Delta }m^2(\mathrm{eV}^2)}\mathrm{m}$$
(3.78)
The Eq. (3.69) and (3.74) describe periodical transitions (oscillations) between different types of neutrinos due to difference of neutrino masses and to neutrino mixing. The transition probability depends periodically on $`L/E`$. At the values of $`L/E`$ at which the condition $`2.54\mathrm{\Delta }m^2(L/E)=\pi (2n+1)(n=0,1,\mathrm{})`$ is satisfied, the transition probability is equal to the maximal value $`\mathrm{sin}^22\theta `$. If the condition $`2.54\mathrm{\Delta }m^2(L/E)=2\pi n`$ is satisfied, the transition probability is equal to zero.
In order to see neutrino oscillations it is necessary that the parameter $`\mathrm{\Delta }m^2`$ is large enough so that the condition $`\mathrm{\Delta }m^2(L/E)1`$ is satisfied. This condition allows us to estimate the minimal value of the parameter $`\mathrm{\Delta }m^2`$ that can be revealed in an experiment on the search for neutrino oscillations. For short and long baseline experiments with accelerator (reactor) neutrinos for $`\mathrm{\Delta }m_{min}^2`$ we have, respectively 10 – 1 eV<sup>2</sup>, $`10^2`$ – 10<sup>-3</sup> eV<sup>2</sup> (10<sup>-1</sup> – 10<sup>-2</sup> eV<sup>2</sup>, 10<sup>-2</sup> – 10<sup>-3</sup> eV$`{}_{}{}^{2})`$. For atmospheric and solar neutrinos for $`\mathrm{\Delta }m_{min}^2`$ we have 10<sup>-2</sup> – 10<sup>-3</sup> eV<sup>2</sup> and 10<sup>-10</sup> – 10<sup>-11</sup> eV<sup>2</sup> , respectively. Let us notice that in the case of $`\mathrm{\Delta }m^2(L/E)1`$, due to averaging over neutrino spectrum and over distances between neutrino production and detection points, the term $`\mathrm{cos}\mathrm{\Delta }m^2(L/2p)`$ in the transition probability disappears and the averaged transition probabilities are given by $`\overline{\mathrm{P}}(\nu _\alpha \nu _\alpha ^{})=\frac{1}{2}\mathrm{sin}^22\theta `$ and $`\overline{\mathrm{P}}(\nu _\alpha \nu _\alpha )=1\frac{1}{2}\mathrm{sin}^22\theta `$.
### 3.2 Three neutrino oscillations in the case of neutrino <br>mass hierarchy
The two neutrino transition probabilities (3.69) and (3.74) are usually used for the analysis of experimental data. Let us consider now the case of the transitions between three flavor neutrinos.
General expressions for transition probabilities between three neutrino types are characterized by 6 parameters and have a rather complicated form. We will consider the case of hierarchy of neutrino masses
$$m_1m_2m_3$$
which corresponds to the oscillations of solar and atmospheric neutrinos (we have in mind that $`\mathrm{\Delta }m_{21}^2`$ can be relevant for oscillations of solar neutrinos and $`\mathrm{\Delta }m_{31}^2`$ can be relevant for oscillations of atmospheric neutrinos; from the analysis of the experimental data it follows that $`\mathrm{\Delta }m_{sol}^210^5`$ eV<sup>2</sup> (or $`10^{10}`$ eV<sup>2</sup>) and $`\mathrm{\Delta }m_{atm}^210^3`$ eV<sup>2</sup>; see later). We will see that transition probabilities have in this case the rather simple two–neutrino form.
Let us consider neutrino oscillations in experiments for which the largest neutrino mass squared difference $`\mathrm{\Delta }m_{31}^2`$ is relevant. For such experiments
$$\mathrm{\Delta }m_{12}^2\frac{L}{2p}1$$
(3.79)
and for the probability of the transition $`\nu _\alpha \nu _\alpha ^{}`$, from (3) we obtain the following expression
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\left|\delta _{\alpha ^{}\alpha }+U_{\alpha ^{}3}U_{\alpha 3}^{}\left(e^{i\mathrm{\Delta }m_{31}^2\frac{L}{2p}}1\right)\right|^2$$
(3.80)
For the transition probability $`\nu _\alpha \nu _\alpha ^{}`$ $`(\alpha ^{}\alpha )`$ from (3.80) we have
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\frac{1}{2}\mathrm{A}_{\alpha ^{};\alpha }\left(1\mathrm{cos}\mathrm{\Delta }m_{31}^2\frac{L}{2p}\right)$$
(3.81)
where the amplititude of oscillations is given by
$$\mathrm{A}_{\alpha ^{};\alpha }=4|U_{\alpha ^{}3}|^2|U_{\alpha 3}|^2$$
(3.82)
Using unitarity of the mixing matrix, for the survival probability we obtain, from (3.81) and (3.82),
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=1\underset{\alpha ^{}\alpha }{}\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=1\frac{1}{2}\mathrm{B}_{\alpha ;\alpha }\left(1\mathrm{cos}\mathrm{\Delta }m_{31}^2\frac{L}{2p}\right)$$
(3.83)
where
$$\mathrm{B}_{\alpha ;\alpha }=4|U_{\alpha 3}|^2(1|U_{\alpha 3}|^2)$$
(3.84)
It is natural that Eq. (3.81) and (3.82) have the same dependence on the parameter $`L/E`$ as the standard two–neutrino formulas (3.68) and (3.74): only the largest $`\mathrm{\Delta }m^2`$ is relevant for the oscillations. The oscillation amplitudes $`\mathrm{A}_{\alpha ;\alpha }`$ and $`\mathrm{B}_{\alpha ;\alpha }`$ depend on the moduli squared of the mixing matrix elements that connect neutrino flavors with the heaviest neutrino $`\nu _3`$. Further, from the unitarity of the mixing matrix it follows that
$$|U_{e3}|^2+|U_{\mu 3}|^2+|U_{\tau 3}|^2=1$$
(3.85)
Thus, in three–neutrino case with hierarchy of neutrino masses, the transition probabilities in experiments for which $`\mathrm{\Delta }m_{31}^2`$ is relevant are described by three parameters: $`\mathrm{\Delta }m_{31}^2`$, $`|U_{e3}|^2`$ and $`|U_{\mu 3}|^2`$ (remember that in the two neutrino case there are two parameters, $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta `$ ).
Since only moduli of the elements of the mixing matrix enter into transition probabilities, the relation
$$\mathrm{P}(\nu _\alpha \nu _\alpha ^{})=\mathrm{P}(\overline{\nu }_\alpha \overline{\nu }_\alpha ^{})$$
(3.86)
holds (as in the two–neutrino case). Thus the violation of the CP–invariance in the lepton sector cannot be revealed in the case of three neutrinos with mass hierarchy. Notice that the relation
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=\mathrm{P}(\nu _\alpha ^{}\nu _\alpha ^{}),$$
(3.87)
which takes place in the case of two neutrino oscillations, is not valid in the three–neutrino case.
Let us consider now neutrino oscillations in the case of experiments for which $`\mathrm{\Delta }m_{21}^2`$ is relevant ($`\mathrm{\Delta }m_{21}^2\frac{L}{2p}>1`$). From (3.57) for the survival probability we obtain in this case the following expression
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=\left|\underset{i=1,2}{}|U_{\alpha i}|^2e^{i\mathrm{\Delta }m_{i1}^2\frac{L}{2p}}+|U_{\alpha 3}|^2e^{i\mathrm{\Delta }m_{31}^2\frac{L}{2p}}\right|^2$$
(3.88)
Due to averaging over neutrino spectra and source–detector distances, the interference term $`\mathrm{cos}\mathrm{\Delta }m_{31}^2(L/2p)`$ in Eq. (3.88) disappears and for the probability we have
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=|\underset{i=1,2}{}|U_{\alpha i}|^2e^{i\mathrm{\Delta }m_{i1}^2\frac{L}{2p}}|^2+|U_{\alpha 3}|^4$$
(3.89)
Further, from the unitarity relation $`_{i=1}^3|U_{\alpha i}|^2=1`$ we have
$$\underset{i=1,2}{}|U_{\alpha i}|^4=(1|U_{\alpha 3}|^2)^22|U_{\alpha 1}|^2|U_{\alpha 2}|^2$$
(3.90)
Using (3.90) we can present the survival probability in the form
$$\mathrm{P}(\nu _\alpha \nu _\alpha )=(1|U_{\alpha 3}|^2)^2\mathrm{P}^{(1,2)}(\nu _\alpha \nu _\alpha )+|U_{\alpha 3}|^4$$
(3.91)
Here
$$\mathrm{P}^{(1,2)}(\nu _\alpha \nu _\alpha )=1\frac{1}{2}\mathrm{sin}^22\overline{\theta }_{12}(1\mathrm{cos}^2\mathrm{\Delta }m_{21}^2\frac{L}{2p})$$
(3.92)
and the angle $`\overline{\theta }_{12}`$ is determined by the relations
$$\mathrm{cos}^2\overline{\theta }_{12}=\frac{|U_{\alpha 1}|^2}{_{i=1,2}|U_{\alpha i}|^2},\mathrm{sin}^2\overline{\theta }_{12}=\frac{|U_{\alpha 2}|^2}{_{i=1,2}|U_{\alpha i}|^2},$$
(3.93)
The probability $`\mathrm{P}^{(1,2)}(\nu _e\nu _e)`$ has the two–neutrino form and it is characterized by two parameters: $`\mathrm{\Delta }m_{31}^2`$ and $`\mathrm{sin}^22\overline{\theta }_{12}`$. We have derived the expression (3.92) for the case of the oscillations in vacuum. Let us notice that similar expression is valid for the case of the neutrino transitions in matter.
The expressions (3.81), (3.83) and (3.92) can be used to describe neutrino oscillations in atmospheric and long baseline neutrino experiments (LBL) as well as in solar neutrino experiments. In the framework of neutrino mass hierarchy, in the probabilities of transition of atmospheric (LBL) and solar neutrinos enter different $`\mathrm{\Delta }m^2`$ ($`\mathrm{\Delta }m_{31}^2`$ and $`\mathrm{\Delta }m_{2,1}^2`$, respectively) and the only element that connects oscillations of atmospheric (LBL) and solar neutrinos is $`|U_{e3}|^2`$. From LBL reactor experiment CHOOZ and Super–Kamiokande experiment it follows that this element is small (see later). This means that oscillations of atmospheric (LBL) and solar neutrinos are described by different elements of the neutrino mixing matrix.
## 4 Neutrino in matter
Up to now we have considered oscillations of neutrinos in vacuum. If there is neutrino mixing the effects of the matter can significantly enhance the probability of the transitions between different types of neutrinos (MSW effect). We will consider here this effect in some details.
Let consider neutrinos with momentum $`\stackrel{}{p}`$. The equation of the motion for a free neutrino has the form
$$i\frac{|\psi (t)}{t}=H_0|\psi (t)$$
(4.94)
Let us develop the state $`|\psi (t)`$ over states of neutrinos with definite flavor $`|\nu _\alpha `$ ($`\alpha =e,\mu ,\tau `$). We have
$$|\psi (t)=\underset{\alpha }{}|\nu _\alpha a_\alpha (t)$$
(4.95)
where $`a_\alpha (t)`$ is the wave function of neutrino in the flavor representation. From (4.94) for $`a_\alpha (t)`$ we obtain the equation
$$i\frac{a_\alpha (t)}{t}=\underset{\alpha ^{}}{}\nu _\alpha |H_0|\nu _\alpha ^{}a_\alpha ^{}(t)$$
(4.96)
Now we will develop the state $`|\nu _\alpha `$ over the eigenstates $`|\nu _i`$ of the free Hamiltonian H<sub>0</sub>:
$`H_0|\nu _i=E_i|\nu _i,`$ (4.97)
$`E_i=\sqrt{p^2+m_i^2}p+{\displaystyle \frac{m_i^2}{2p}}.`$ (4.98)
We have:
$$|\nu _\alpha =\underset{i}{}|\nu _i\nu _i|\nu _\alpha $$
(4.99)
If we compare (4.99) and (3.52) we find
$$\nu _i|\nu _\alpha =U_{\alpha i}^{}\nu _\alpha |\nu _i=U_{\alpha i}$$
(4.100)
Further we have
$$\nu _\alpha |H_0|\nu _\alpha ^{}=\underset{i}{}\nu _\alpha |\nu _i\nu _i|H_0|\nu _i\nu _i|\nu _\alpha ^{}=\underset{i}{}U_{\alpha i}\frac{m_i^2}{2p}U_{i\alpha ^{}}^{}+p\delta _{\alpha \alpha ^{}}$$
(4.101)
The last term of 4.101, which is proportional to unit matrix, cannot change the flavor state of neutrino. This term can be excluded from the equation of motion by redefining the phase of the function $`a(t)`$. We have:
$$i\frac{a(t)}{t}=U\frac{m^2}{2p}U^{}a(t)$$
(4.102)
This equation can be easily solved. Let us multiply (4.102) by the matrix $`U^{}`$ from the left. Taking into account unitarity of the mixing matrix we have:
$$i\frac{a^{}(t)}{t}=\frac{m^2}{2p}a^{}(t)$$
(4.103)
where $`a^{}(t)=U^{}a(t)`$. The solution of equation (4.103) has the form
$$a^{}(t)=e^{i\frac{\mathrm{\Delta }m^2}{2p}t}a^{}(0).$$
(4.104)
For the function $`a(t)`$ in flavor representation, from (4.103) and (4.104), we find
$$a(t)=Ue^{i\frac{\mathrm{\Delta }m^2}{2p}t}U^{}a(0)$$
(4.105)
and for the amplitude of the $`\nu _\alpha \nu _\alpha ^{}`$ transition in vacuum from (4.105) we obtain the expression
$$𝒜_{\nu _\alpha ^{};\nu _\alpha }(t)=\underset{i}{}U_{\alpha ^{}i}e^{i\frac{\mathrm{\Delta }m_i^2}{2p}}U_{\alpha i}^{}$$
(4.106)
which (up to the irrelevant factor $`e^{ipt}`$) coincides with (3.57).
Let us now introduce the effective Hamiltonian of interaction of flavor neutrino with matter. Due to coherent scattering of neutrino in matter, the refraction index of neutrino is given by the following classical expression:
$$n(x)=1+\frac{2\pi }{p^2}\mathrm{f}(0)\rho (x)$$
(4.107)
Here $`\mathrm{f}(0)`$ is the amplitude of elastic neutrino scattering in forward direction, and $`\rho (x)`$ is the number density of matter (the axis $`x`$ is the direction of $`\stackrel{}{p}`$). The effective interaction of neutrinos with matter is determined by the second term of Eq. (4.107) :
$$H_I(x)=p[n(x)1]=\frac{2\pi }{p}\mathrm{f}(0)\rho (x)$$
(4.108)
NC scattering of neutrinos on electrons and nucleons (due to the Z-exchange) cannot change the flavor state of neutrinos. This is connected with $`\nu _e\nu _\mu \nu _\tau `$ universality of NC: the corresponding effective Hamiltonian is proportional to the unit matrix<sup>3</sup><sup>3</sup>3 Let us notice that if there are flavour and sterile neutrinos NC interactions with matter must be taken into account..
CC interaction (due to the W-exchange) gives contribution only to the amplitude of the elastic $`\nu _e`$ -$`e`$ scattering
$$\nu _e+e\nu _e+e$$
(4.109)
For the corresponding effective Hamiltonian we have
$$_I(x)=\frac{G_F}{\sqrt{2}}2\overline{\nu }_{eL}\gamma ^\alpha \nu _{eL}\overline{e}\gamma _\alpha (1\gamma _5)e+h.c.$$
(4.110)
The amplitude of process (4.109) is given by
$$\mathrm{f}_{\nu _ee}=\frac{1}{\sqrt{2}\pi }G_Fp$$
(4.111)
and, from (4.108) and (4.111), for the effective Hamiltonian in flavor representation we have
$$H_I(x)=\sqrt{2}G_F\rho _e(x)\beta $$
(4.112)
where $`(\beta )_{\nu _e;\nu _e}=1`$, while all other elements of the matrix $`\beta `$ are equal to zero and $`\rho _e(x)`$ is the electron number density at the point $`x`$.
The effective Hamiltonian of the neutrino interaction with matter can be also obtained by calculating of the average value of the Hamiltonian (4.110) in the state which describes matter and neutrino with momentum $`\stackrel{}{p}`$ and negative helicity . Taking into account that for non–polarized media
$`\mathrm{mat}|\overline{e}(\stackrel{}{x})\gamma ^\alpha e(\stackrel{}{x})|\mathrm{mat}=\rho _e(\stackrel{}{x})\delta _{\alpha 0},`$ (4.113)
$`\mathrm{mat}|\overline{e}(\stackrel{}{x})\gamma ^\alpha \gamma _5e(\stackrel{}{x})|\mathrm{mat}=0,`$ (4.114)
from (4.110) we obtain (4.112).
The evolution equation of neutrino in matter can be written, from (4.102) and (4.112), in the following form ($`t=x`$):
$$i\frac{a(x)}{x}=(U\frac{m^2}{2p}U^{}+\sqrt{2}G_F\rho _e(x)\beta )a(x)$$
(4.115)
Let consider in detail the simplest case of two flavor neutrinos (say, $`\nu _e`$ and $`\nu _\mu `$). In this case we have
$$U=\left(\begin{array}{cc}\hfill \mathrm{cos}\vartheta & \hfill \mathrm{sin}\vartheta \\ \hfill \mathrm{sin}\vartheta & \hfill \mathrm{cos}\vartheta \end{array}\right)$$
(4.116)
where $`\theta `$ is the mixing angle. Further it is convenient to write the Hamiltonian in the form
$$H=\frac{1}{2}\mathrm{Tr}H+H^m$$
(4.117)
where Tr $`H=\frac{1}{2p}(m_1^2+m_2^2)+\sqrt{2}G_F\rho _e`$. The first term of (4.117), which is proportional to the unit matrix, can be omitted. For the Hamiltonian we have then
$$H^m(x)=\frac{1}{4p}\left(\begin{array}{cc}\mathrm{\Delta }m^2\mathrm{cos}2\vartheta +A(x)& \mathrm{\Delta }m^2\mathrm{sin}2\vartheta \\ \mathrm{\Delta }m^2\mathrm{sin}2\vartheta & \mathrm{\Delta }m^2\mathrm{cos}2\vartheta A(x)\end{array}\right)$$
(4.118)
where $`\mathrm{\Delta }m^2=m_2^2m_1^2`$ and $`A(x)=2\sqrt{2}G_F\rho _e(x)p`$. The effect of matter is described by the quantity $`A(x)`$. Notice that this quantity enters only into the diagonal elements of the Hamiltonian and has the dimensions of $`M^2`$.
Let us first consider the case of constant density. In order to solve equation of motion we will diagonalize the Hamiltonian. We have:
$$H^m=U^mE^mU_{}^{m}{}_{}{}^{}$$
(4.119)
where $`E_i^m`$ is the eigenvalue of the matrix $`H^m`$ and
$$U^m=\left(\begin{array}{cc}\hfill \mathrm{cos}\vartheta ^m& \hfill \mathrm{sin}\vartheta ^m\\ \hfill \mathrm{sin}\vartheta ^m& \hfill \mathrm{cos}\vartheta ^m\end{array}\right)$$
(4.120)
It is easy to see that
$$E_{1,2}^m=\frac{1}{4p}\sqrt{(\mathrm{\Delta }m^2\mathrm{cos}2\theta A)^2+(\mathrm{\Delta }m^2\mathrm{sin}2\theta )^2}.$$
(4.121)
Now, with the help of Eq. (4.119) – (4.121), for the angle $`\theta ^m`$ we have
$$\mathrm{tan}2\theta ^m=\frac{\mathrm{\Delta }m^2\mathrm{sin}2\theta }{\mathrm{\Delta }m^2\mathrm{cos}2\theta A};\mathrm{cos}2\theta ^m=\frac{\mathrm{\Delta }m^2\mathrm{cos}2\theta A}{\sqrt{(\mathrm{\Delta }m^2\mathrm{cos}2\theta A)^2+(\mathrm{\Delta }m^2\mathrm{sin}2\theta )^2}}$$
(4.122)
The states of flavor neutrinos are given by
$$|\nu _e=\mathrm{cos}\theta ^m|\nu _{1m}+\mathrm{sin}\theta ^m|\nu _{2m};|\nu _\mu =\mathrm{sin}\theta ^m|\nu _{1m}+\mathrm{cos}\theta ^m|\nu _{2m}$$
(4.123)
where $`|\nu _{im}`$ ($`i=1,2`$) are eigenvectors of the Hamiltonian of neutrino in matter and $`\theta ^m`$ is the mixing angle of neutrino in matter.
The solution of the evolution equation
$$i\frac{a(x)}{x}=H_ma(x)$$
(4.124)
can be now easily found. With the help of (4.119) we have
$$i\frac{a^{}(x)}{x}=E^ma^{}(x)$$
(4.125)
where
$$a^{}(x)=(U^m)^{}a(x).$$
(4.126)
The equation (4.125) has the following solution:
$$a^{}(x)=e^{iE^m(xx_0)}a^{}(x_0)$$
(4.127)
where $`x_0`$ is the point where the neutrino was produced. Finally, from (4.126) and (4.127), we have
$$a(x)=U^me^{iE^m(xx_0)}(U^m)^{}a(x_0)$$
(4.128)
The amplitude of the $`\nu _\alpha \nu _\alpha ^{}`$ transition in matter turns out to be
$$𝒜_{\nu _\alpha ^{};\nu _\alpha }=\underset{i=1,2}{}U_{\alpha ^{}i}^me^{iE_i^m(xx_0)}U_{\alpha i}^{}$$
(4.129)
and, from (4.129) and (4.120), we obtain the following transition probabilities, in full analogy with the two–neutrino vacuum case:
$`P^m(\nu _e\nu _\mu )=P^m(\nu _\mu \nu _e)={\displaystyle \frac{1}{2}}\mathrm{sin}^22\theta ^m(1\mathrm{cos}\mathrm{\Delta }E^mL),`$ (4.130)
$`P^m(\nu _e\nu _e)=P^m(\nu _\mu \nu _\mu )=(1P^m(\nu _e\nu _\mu ).`$ (4.131)
Here $`\mathrm{\Delta }E^m=E_2^mE_1^m=\frac{1}{2p}\sqrt{(\mathrm{\Delta }m^2\mathrm{cos}2\theta A)^2+(\mathrm{\Delta }m^2\mathrm{sin}2\theta )^2}`$ and $`L=xx_0`$ is the distance that neutrino passes in matter.
For the oscillation length of neutrino in matter with constant density we have
$$L_0^m=4\pi \frac{p}{\sqrt{(\mathrm{\Delta }m^2\mathrm{cos}2\theta A)^2+(\mathrm{\Delta }m^2\mathrm{sin}2\theta )^2}}$$
(4.132)
The mixing angle and oscillation length in matter can differ significantly from the vacuum values. It follows from (4.122) that if the condition<sup>4</sup><sup>4</sup>4Eq. (4.131) is the condition at which the diagonal elements of the Hamiltonian of neutrino in matter vanish. It is evident that in such a case the mixing is maximal.
$$\mathrm{\Delta }m^2\mathrm{cos}2\theta =A=2\sqrt{2}G_F\rho _ep$$
(4.133)
is satisfied, the mixing in matter is maximal ($`\theta ^m=\pi /4`$) independently on the value of the vacuum mixing angle $`\theta `$. Notice also that if the condition (4.133) is satisfied, the distance between the energy levels of neutrinos in matter is minimal and the oscillation length in matter is maximal. We have
$$L_0^m=\frac{L_0}{\mathrm{sin}2\theta }$$
(4.134)
where $`L_0=4\pi p/(\mathrm{\Delta }m)`$ is the oscillation length in vacuum. If the distance $`L`$ in the transition probabilities (4.131) is large (as in the Sun case) the effect of $`\nu _e\nu _\mu `$ transitions is large even in case of a small vacuum mixing angle $`\theta `$. The relation (4.133) is called resonance condition.
The density of electrons in the Sun is not constant. It is maximal in the center of the Sun and decreases practically exponentially to its periphery. The consideration of the dependence of $`\rho _e`$ on $`x`$ allowed to discover possibilities for the large effects of the transitions of solar $`\nu _e`$’s into other states in matter (MSW effect).
Let us consider the evolution equation when the Hamiltonian depends on the distance $`x`$ that neutrino passes in matter
$$i\frac{a(x)}{x}=H^m(x)a(x)$$
(4.135)
The Hermitian Hamiltonian $`H^m(x)`$ can be diagonalized by a unitary transformation
$$H^m(x)=U^m(x)E^m(x)U_{}^{m}{}_{}{}^{}(x)$$
(4.136)
where $`U^m(x)U_{}^{m}{}_{}{}^{}(x)=1`$ and $`E_i^m`$(x) are eigenvalues of $`H^m(x)`$. From (4.135) and (4.136) we have
$$U_{}^{m}{}_{}{}^{}(x)i\frac{a(x)}{t}=E^m(x)a^{}(x)$$
(4.137)
where
$$a^{}(x)=U_{}^{m}{}_{}{}^{}(x)a(x)$$
(4.138)
Further, by taking into account that
$$U_{}^{m}{}_{}{}^{}(x)i\frac{a(x)}{x}=i\frac{a^{}(x)}{x}+iU_{}^{m}{}_{}{}^{}(x)\frac{U^m(x)}{x}a^{}(x),$$
(4.139)
we have the following equation for $`a^{}(x)`$:
$$i\frac{a^{}(x)}{x}=\left(E^m(x)iU_{}^{m}{}_{}{}^{}(x)\frac{U^m(x)}{x}\right)a^{}(x).$$
(4.140)
In the case $`\rho _e`$ = const the equation (4.140) coincides with (4.125).
Let us now assume that the function $`\rho _e(x)`$ depends weakly on $`x`$ and the second term in Eq. (4.138) can be dropped (adiabatic approximation). It is evident that the solution of the equation
$$i\frac{a_i^{}(x)}{x}=E_i^m(x)a_i^{}(x)$$
(4.141)
has the form
$$a_i^{}(x)=e^{i{\displaystyle _{x_0}^x}E_i^m\left(x\right)𝑑x}a_i^{}(x_0)$$
(4.142)
($`x_0`$ being the initial point).
It follows from (4.141) and (4.142) that, in the adiabatic approximation, a neutrino on the way from the point $`x_0`$ to the point $`x`$ remains in the same energy level. From (4.138) and (4.142) we obtain the following solution of the evolution equation in flavor representation:
$$a(x)=U^m(x)e^{i_{x_0}^xE^m(x)𝑑x}U_{}^{m}{}_{}{}^{}(x_0)A(X_0).$$
(4.143)
Moreover the amplitude of $`\nu _\alpha \nu _\alpha ^{}`$ transition in adiabatic approximation is given by
$$𝒜_{\nu _\alpha ^{};\nu _\alpha }=U_{\alpha ^{}i}^m(x)e^{i_{x_0}^xE_i^m(x)𝑑x}U_{\alpha i}^m(x_0).$$
(4.144)
The latter is similar to the expressions (4.106) and (4.129) for the amplitudes of transition in vacuum and in matter with $`\rho _e`$ = const.
For the case of the two flavor neutrinos
$$U^m(x)=\left(\begin{array}{cc}\hfill \mathrm{cos}\vartheta ^m(x)& \hfill \mathrm{sin}\vartheta ^m(x)\\ \hfill \mathrm{sin}\vartheta ^m(x)& \hfill \mathrm{cos}\vartheta ^m(x)\end{array}\right)$$
(4.145)
and $`\mathrm{tan}2\theta ^m(x)`$ and $`cos2\theta ^m(x)`$ are given by Eq. (4.122) in which
$$\mathrm{A}(x)=2\sqrt{2}G_F\rho _e(x)p$$
(4.146)
The eigenvalues of the Hamiltonian $`H^m(x)`$ are given by Eq. (4.121). From (4.145) we have
$$U_{}^{m}{}_{}{}^{}(x)\frac{U^m(x)}{x}=\left(\begin{array}{cc}0& \frac{\theta ^m(x)}{x}\\ \frac{\theta ^m(x)}{x}& 0\end{array}\right)$$
(4.147)
and the exact equation (4.140) takes the form
$$i\frac{}{x}\left(\begin{array}{c}a_1^{}\\ a_2^{}\end{array}\right)=\left(\begin{array}{cc}E_1^m& i\frac{\theta ^m}{x}\\ i\frac{\theta ^m}{x}& E_2^m\end{array}\right)\left(\begin{array}{c}a_1^{}\\ a_2^{}\end{array}\right)$$
(4.148)
The Hamiltonian $`H^m`$ in the right–hand side of this equation can be written in the form
$$H_m=\frac{1}{2}(E_1^m+E_2^m)+\left(\begin{array}{cc}\frac{1}{2}\mathrm{\Delta }E^m& i\frac{\theta ^m}{x}\\ i\frac{\theta ^m}{x}& \frac{1}{2}\mathrm{\Delta }E^m\end{array}\right)$$
(4.149)
where $`\mathrm{\Delta }E^m=E_2^mE_1^m`$. As we stressed several times, the term of the Hamiltonian which is proportional to the unit matrix is not important for flavor evolution.
From Eq. (4.149) it follows that adiabatic approximation is valid if the condition
$$\left|\frac{\theta ^m}{x}\right|\frac{1}{2}\mathrm{\Delta }E^m$$
(4.150)
is satisfied. With the help of (4.122) it is easy to show that (4.150) can be written in the form
$$4\sqrt{2}G_Fp^2\mathrm{\Delta }m^2\mathrm{sin}2\theta \left|\frac{\rho _e}{x}\right|\left[(\mathrm{\Delta }m^2\mathrm{cos}2\theta A)^2+(\mathrm{\Delta }m^2\mathrm{sin}2\theta )^2\right]^{3/2}.$$
(4.151)
If the resonance condition
$$\mathrm{\Delta }m^2\mathrm{cos}2\theta =A(x_R)$$
(4.152)
is satisfied at the point $`x=x_R`$, the condition of validity of the adiabatic approximation can be written in the form
$$\frac{2p\mathrm{cos}2\theta \left|\frac{}{x}\mathrm{ln}\rho _e(x_R)\right|}{\mathrm{\Delta }m^2\mathrm{sin}^22\theta }1.$$
(4.153)
From Eq. (4.144) we obtain the following probability for the $`\nu _\alpha \nu _\alpha ^{}`$ transition in the adiabatic approximation:
$`P(\nu _\alpha \nu _\alpha ^{})`$ $`={\displaystyle \underset{i}{}}|U_{\alpha ^{}i}^m(x)|^2|U_{\alpha i}^m(x_0)|^2+`$
$`+2Re{\displaystyle \underset{i<k}{}}U_{\alpha ^{}i}^m(x)U_{\alpha ^{}k}^{m}{}_{}{}^{}e^{i_{x_0}^x(E_i^mE_k^m)𝑑x}U_{\alpha i}^{m}{}_{}{}^{}(x_0)U_{\alpha k}^m(x_0).`$
For solar neutrinos the second term in the r.h.s. of this expression disappears due to averaging over the energy and the region in which neutrinos are produced. Hence for the averaged transition probability we have
$$\overline{P}(\nu _\alpha \nu _\alpha ^{})=\underset{i}{}|U_{\alpha ^{}i}^m(x)|^2|U_{\alpha i}^m(x_0)|^2$$
(4.155)
Thus, in the adiabatic approximation, the averaged transition probability is determined by the elements of the mixing matrix in matter at the initial and final points. For the case of two neutrino flavors we have the following simple expression for the $`\nu _e`$ survival probability
$`\overline{P}(\nu _e\nu _e)`$ $`=\mathrm{cos}^2\theta ^m(x)\mathrm{cos}^2\theta ^m(x_0)+\mathrm{sin}^2\theta ^m(x)\mathrm{sin}^2\theta ^m(x_0)`$ (4.156)
$`={\displaystyle \frac{1}{2}}\left(1+\mathrm{cos}2\theta ^m(x)\mathrm{cos}2\theta ^m(x_0)\right)`$
From Eq. (4.156) it is easy to see that if the neutrino passes the point $`x=x_R`$ where the resonance condition is satisfied, a large effect of disappearance of $`\nu _e`$ will be observed. In fact, the condition (4.152) is fulfilled if $`\mathrm{cos}2\theta >0`$ (neutrino masses are labelled in such a way that $`\mathrm{\Delta }m^2>0`$). At the production point $`x_0`$ the density is larger than at point $`x_R`$ and $`A(x_0)>\mathrm{\Delta }m^2\mathrm{cos}2\theta `$. From (4.122) it follows than $`\mathrm{cos}2\theta (x_0)<0`$. Thus, if the resonance condition is fulfilled, we see from Eq. (4.156) that $`P(\nu _e\nu _e)<\frac{1}{2}`$. If the condition
$$A(x_0)\mathrm{\Delta }m^2$$
(4.157)
is satisfied for neutrinos produced in the center of the Sun,then $`\mathrm{cos}2\theta ^m(x_0)1`$ and, for neutrinos passing through the Sun, the survival probability is equal to:
$$\overline{P}(\nu _e\nu _e)\frac{1}{2}(1\mathrm{cos}2\theta )$$
(4.158)
It is obvious from this expression that the $`\nu _e`$ survival probability at small $`\theta `$ is close to zero: all $`\nu _e`$’s are transformed into $`\nu _\mu `$’s.
Let us consider evolution of neutrino states in such a case. From Eq. (4.122) it follows that, at the production point, $`\theta ^m(x_0)\pi /2`$. From (4.123) we have then
$$|\nu _e|\nu _{2m};|\nu _\mu =|\nu _{1m}(x=x_0)$$
(4.159)
Thus at the production point flavor states are states with definite energy. In the adiabatic approximation there are no transitions between energy levels. In the final point $`\rho _e`$ = 0 and at small $`\theta `$ we have
$$|\nu _2|\nu _\mu ,|\nu _1|\nu _e(x=x_0)$$
(4.160)
Thus, all $`\nu _e`$’s transfer to $`\nu _\mu `$’s. The resonance condition (4.152) was written in units $`\mathrm{}=c=1`$. We can rewrite it in the following form
$$\mathrm{\Delta }m^2\mathrm{cos}2\theta 0.710^7E\rho \mathrm{eV}^2$$
where $`\rho `$ is the density of matter in g$``$ cm<sup>-3</sup> and $`E`$ is the neutrino energy in MeV. In the central region of the Sun $`\rho 10^2\mathrm{g}\mathrm{cm}^3`$ and the energy of the solar neutrinos is $`1MeV`$. Thus the resonance condition is satisfied at $`\mathrm{\Delta }m^210^5`$ eV<sup>2</sup>.
The expression (4.155) gives the averaged survival probability in the adiabatic approximation. In the general case we have
$$\overline{P}(\nu _\alpha \nu _\alpha ^{})=|U_{\alpha ^{}i}^m(x)|^2P_{ik}|U_{\alpha k}^m(x_0)|^2$$
(4.161)
where $`P_{ik}`$ is the probabilty of transition from the state with energy $`E_k^m`$ to the state with energy $`E_i^m`$. Let us consider the simplest case of the transition between two types of neutrinos. From the conservation of the total probability we have
$$P_{11}=1P_{21},P_{22}=1P_{12},P_{12}=P_{21}$$
(4.162)
Thus in the case of two neutrinos all transition probabilities $`P_{ik}`$ are expressed through $`P_{12}`$. With the help of (4.145), (4.161) and (4.162), for the $`\nu _e`$ survival probability we have:
$$\overline{P}(\nu _e\nu _e)=\frac{1}{2}+\left(\frac{1}{2}P_{12}\right)\mathrm{cos}2\theta ^m(x)\mathrm{cos}2\theta ^m(x_0)$$
(4.163)
In the literature there exist different approximate expressions for the transition probability $`P_{12}`$. In the Landau–Zenner approximation, based on the assumption that the transition occurs mainly in the resonance region,
$$P_{12}=e^{\frac{\pi }{2}\gamma _RF}$$
(4.164)
where
$$\gamma _R=\frac{\frac{1}{2}\mathrm{\Delta }E^m}{|\theta ^m/x|}=\frac{\mathrm{\Delta }m^2\mathrm{sin}^22\theta }{2p\mathrm{cos}2\theta |\frac{}{x}\mathrm{ln}\rho _e(x_R)|}.$$
(4.165)
In the above equation $`F=1`$ for linear density and $`F=1\mathrm{tan}^2\theta `$ for exponential density. The adiabatic approximation is valid if $`\gamma _R1`$ \[see (4.150)\]. In this case $`P_{12}0`$.
This concludes the considerations on the phenomenological theory of neutrino mixing and on the theory of neutrino oscillations in vacuum and in matter. We will start now the discussion of experimental data. There are three methods to search for the effects of neutrino masses and mixing:
I. The precise measurement of the high energy part of $`\beta `$–spectrum;
II. The search for neutrinoless double $`\beta `$–decay;
III. The investigation of neutrino oscillations .
We shall discuss now the results which have been obtained in some of the most recent experiments.
## 5 Search for effects of neutrino mass <br>in experiments on the measurement <br>of the $`\beta `$-spectrum of <sup>3</sup>H
We will discuss here briefly the results of searching for effects of neutrino masses in experiments on the measurement of the high-energy part of the $`\beta `$–spectrum in the decay
$${}_{}{}^{3}\mathrm{H}{}_{}{}^{3}\mathrm{He}+e^{}+\overline{\nu _e}$$
(5.166)
The process (5.166) is a superallowed $`\beta `$–decay: the nuclear matrix element is constant and the $`\beta `$–spectrum is determined by the phase–space factor and the Coulomb interaction of the final $`e^{}`$ and <sup>3</sup>He. For the $`\beta `$–spectrum we have
$$\frac{dN}{dT}=CpE(QT)\sqrt{(QT)^2m_\nu ^2}F(E)$$
(5.167)
Here $`p`$ is electron momentum, $`E=m_e+T`$ is the total electron energy, $`Q=m_{{}_{}{}^{3}\mathrm{H}}m_{{}_{}{}^{3}\mathrm{He}}m_e18.6`$ keV is the energy release, $`C=`$ const and $`F(E)`$ is the Fermi function, which describes the Coulomb interaction of the final particles. In the Eq. (5.167) the term $`(QT)`$ is the neutrino energy (the recoil energy of <sup>3</sup>He can be neglected) and the neutrino mass enters through the neutrino momentum $`p_\nu =\sqrt{(QT)^2m_\nu ^2}`$. Notice that in the derivation of Eq. (5.167) the simplest assumption was done that $`\nu _e`$ is the particle with mass $`m_\nu `$.
The Kurie function is then determined as follows
$$K(T)=\sqrt{\frac{dN}{dt}\frac{1}{pEF(E)}}=\sqrt{C}\sqrt{(QT)\sqrt{(QT)^2m_\nu ^2}}$$
(5.168)
If $`m_\nu `$ = 0, the Kurie function is the stright line $`K(T)=\sqrt{C}(QT)`$ and $`T_{max}=0`$. If $`m_\nu 0`$ then $`T_{max}=Qm_\nu `$ and at small $`m_\nu `$ the Kurie function deviates from the stright line in the region close to the maximum allowed energy. Thus, if $`m_\nu 0`$ in the end point part of the spectrum a deficit of observed events must be measured (with respect to the number of events expected at $`m_\nu =0`$).
In experiments on the search for effects of neutrino mass by <sup>3</sup>H–method no positive indications in favour of $`m_\nu 0`$ were found. In these experiments some anomalies were observed. First, practically in all experiments the best–fit values of $`m_\nu ^2`$ are negative. This means that instead of a deficit of events, some excess is observed. Second, in the Troitsk experiment a peak in electron spectrum is observed at the distance of a few eV from the end. The position of the peak is changed periodically with time. There are no doubts that new, more precise experiments are necessary. The results of two running experiments are presented in Table 5.
## 6 Neutrinoless double $`\beta `$-decay
The decay
$$(A,Z)(A,Z+2)+e^{}+e^{}$$
(6.169)
is possible only if the total lepton number $`L`$ is not conserved, i.e. if neutrinos with definite masses are Majorana particles. There are many experiments in which neutrinoless double $`\beta `$-decay ($`(\beta \beta )_{0\nu }`$–decay) of <sup>76</sup>Ge, <sup>136</sup>Xe, <sup>130</sup>Te, <sup>82</sup>Se, <sup>100</sup>Mo and other even–even nuclei is searched for.
Let consider the process (6.169) in the framework of neutrino mixing. The standard CC Hamiltonian of the weak interaction has the form
$$H_I=\frac{G_F}{\sqrt{2}}2\overline{e}_L\gamma ^\alpha \nu _{eL}j_\alpha +h.c.$$
(6.170)
Here $`j_\alpha `$ is the weak hadronic current and
$$\nu _{eL}=U_{ei}\nu _{iL}$$
(6.171)
where $`\nu _i`$ is the Majorana neutrino field with mass $`m_i`$.
The ($`\beta \beta )_{0\nu }`$ decay is a process of second order in $`G_F`$ with an intermediate virtual neutrino. Neutrino masses and mixing enter into the neutrino propagator<sup>5</sup><sup>5</sup>5 We have used the relation $`\nu _i^T=\nu _iC`$ that follows from the Majorana condition $`\nu _i^C=C\overline{\nu }_i^T=\nu _i`$. It is obvious that in the case of Dirac neutrino the propagator is equal to zero.
$`\nu _{eL}^{}{}_{}{}^{}(x_1)\nu _{eL}^{T}{}_{}{}^{}(x_2)`$ $`={\displaystyle \underset{i}{}}U_{ei}^2\nu _{iL}^{}{}_{}{}^{}(x_1)\nu _{iL}^{T}{}_{}{}^{}(x_2)={\displaystyle U_{ei}^2\frac{(1\gamma _5)}{2}\nu _{i}^{}{}_{}{}^{}(x_1)\overline{\nu }_{i}^{}{}_{}{}^{}(x_2)\frac{(1\gamma _5)}{2}C}`$ (6.172)
$`={\displaystyle U_{ei}^2\frac{(1\gamma _5)}{2}\frac{i}{(2\pi )^4}\frac{e^{ip(x_1x_2)}(\text{/}p+m_i)}{p^2m_i^2}d^4p\frac{(1\gamma _5)}{2}C}`$
Taking into account that
$$\frac{(1\gamma _5)}{2}(\text{/}p+m_i)\frac{(1\gamma _5)}{2}=m_i\frac{(1\gamma _5)}{2},$$
(6.173)
we come to the conclusion that the matrix element of $`(\beta \beta )_{0\nu }`$–decay is proportional to<sup>6</sup><sup>6</sup>6 The term $`m_i^2`$ in denominator is small with respect to characteristic $`p`$ in nuclei ($``$ 10 MeV) and can be neglected.
$$<m>=U_{ei}^2m_i$$
(6.174)
From (6.173) it is evident that the proportionality of the matrix element of $`(\beta \beta )_{0\nu }`$–decay to $`<m>`$ is due to the fact that the standard CC interaction is the left–handed one. If neutrino masses are equal to zero $`(\beta \beta )_{0\nu }`$–decay is forbidden (conservation of helicity). Notice that, if there is some small admixture of right–handed currents in the interaction Hamiltonian, the $`LR`$ interference gives a contribution proportional to the $`\text{/}p`$ term in the neutrino propagator. Other mechanisms of $`(\beta \beta )_{0\nu }`$–decay are also possible (SUSY with violation of R-parity ect.).
In the experiments on the search for $`(\beta \beta )_{0\nu }`$-decay very strong bounds on the life–time of this process were obtained. The results of some of the latest experiments are presented in Table 6. From these data upper bounds for $`|<m>|`$ can be obtained. The upper bounds depend on the values of the nuclear matrix elements, the calculation of which is a complicated problem. From <sup>76</sup>Ge data it follows
$$|<m>|<(0.51)\mathrm{eV}$$
(6.175)
In the future experiments on the search for $`(\beta \beta )_{0\nu }`$–decay (Heidelberg-Moscow, NEMO, CUORE and others) the sensitivity $`|<m>|<0.1`$ eV is planned to be achieved.
## 7 Neutrino oscillation experiments
We will discuss now the existing experimental data on the search for neutrino oscillations. There exist at present convincing evidences in favour of neutrino oscillations, which were obtained in atmospheric neutrino experiments and first of all in the Super–Kamiokande experiment. Strong indications in favour of neutrino masses and mixing were obtained in all solar neutrino experiments. Finally, some indications in favour of $`\nu _\mu \nu _e`$ transitions were obtained in the LSND accelerator experiment. In many reactor and accelerator short baseline experiments and in the reactor long baseline experiments CHOOZ no indication in favour of neutrino oscillations was found. We will start with the discussion of the results of solar neutrino experiments.
### 7.1 Solar neutrinos
The energy of the Sun is generated in the reactions of the thermonuclear pp and CNO cycles. The main pp–cycle is illustrated in Fig.7.1.
The energy of the sun is produced in the transition
$$4p+2e^{}{}_{}{}^{4}\mathrm{He}+2\nu _e,$$
(7.176)
If we assume that solar $`\nu _e`$’s do not transfer into other neutrino types ($`P(\nu _e\nu _e)=1`$) we can obtain a relation between the luminosity of the Sun, $`L_{}`$ and the flux of solar neutrinos. Let us consider neutrino with energy $`E`$. From (7.176) it follows that
$$\frac{1}{2}(Q2E)$$
(7.177)
is the luminous energy corresponding to the emission of one neutrino. Here
$$Q=4m_p+2m_em_{{}_{}{}^{4}\mathrm{He}}26.7MeV$$
(7.178)
is the energy release in the transition (7.176). If we multiply (7.177 ) by the total flux of solar $`\nu _e`$’s from different reactions and integrate over the neutrino energy $`E`$ we will obtain the flux of luminous energy from the Sun
$$\frac{1}{2}(Q2E)\underset{i}{}I_i(E)dE=\frac{L_{}}{4\pi R^2}.$$
(7.179)
Here $`L_{}3.8610^{33}`$ erg/s is the luminosity of the Sun, $`R`$ is the Sun–Earth distance and $`I_i^0(E)`$ is the flux of neutrinos from the source $`i`$ ($`i`$ = pp, …). Notice that in the derivation of the relation (7.179) we have assumed that the Sun is in a stationary state.
The luminosity relation (7.179) is solar model independent constraint on the solar neutrino fluxes. The flux $`I_i(E)`$ can be written in the form
$$I_i(E)=X_i(E)\mathrm{\Phi }_i$$
(7.180)
where $`\mathrm{\Phi }_i`$ is the total flux and the function $`X_i(E)`$ describes the form of the spectrum ($`X_i(E)𝑑E=1`$). The functions $`X_i(E)`$ are known functions, determined by the weak interaction. The luminosity relation (7.179) can be written in the form
$$Q\underset{i}{}\left(12\frac{\overline{E}_i}{Q}\right)\mathrm{\Phi }_i=\frac{L_{}}{2\pi R^2}$$
(7.181)
where $`\overline{E}_i=EX_i(E)𝑑E`$ is the average energy of neutrinos from the source $`i`$. The main sources of solar neutrinos are listed in Table 7.1
As it is seen from the Table, the main source of solar neutrinos is the reaction $`p+pd+e^++\nu _e`$. This reaction is the source of low energy neutrinos. The source of monochromatic medium energy neutrinos is the process
$$e{}_{}{}^{}+{}_{}{}^{7}\mathrm{Be}\nu _e+{}_{}{}^{7}\mathrm{Li}.$$
(7.182)
The reaction $`{}_{}{}^{8}\mathrm{B}{}_{}{}^{8}\mathrm{Be}+e^++\nu _e`$ is the source of the rare high energy neutrinos. The results of solar neutrino experiments are presented in Table 7.1.
Homestake, GALLEX and SAGE are radiochemical experiments. In the Kamiokande and the Super–Kamiokande experiments recoil electrons (angle and energy) in the elastic neutrino–electron scattering are detected. In these experiments the direction of neutrinos is determined and it is confirmed that the detected events are from solar neutrinos.
In the Homestake experiment, because of high threshold ($`E_{th}=0.81`$ MeV) mainly <sup>8</sup>B neutrinos are detected: $`77`$% of events are due to <sup>8</sup>B neutrinos and $`15`$% of events are due to <sup>7</sup>Be neutrinos. In GALLEX and SAGE experiments ($`E_{th}=0.23`$ MeV) neutrinos from all reactions are detected: $`54`$% of events are due to $`pp`$ neutrinos, $`27`$% of events are due to <sup>7</sup>Be neutrinos and $`10`$% of events are due to <sup>8</sup>B neutrinos. In the Kamiokande and Super–Kamiokande experiments due to the high threshold ($`E_{th}=7`$ MeV for Kamiokande and $`E_{th}=5.5`$ MeV for the Super–Kamiokande) only high energy <sup>8</sup>B neutrinos are detected.
The results of the solar neutrino experiments are presented in Table7.1. As it is seen from the Table, the detected event rates in all solar neutrino experiments are significantly smaller than the predicted one.<sup>7</sup><sup>7</sup>7Notice that in the framework of neutrino oscillations the possibility of deficit of solar $`\nu _e`$’s was discussed by B. Pontecorvo in 1968 before the results of the Homestake experiment were obtained. The most natural explanation of the data of solar neutrino experiments can be obtained in the framework of neutrino mixing. In fact, if neutrinos are massive and mixed, solar $`\nu _e`$’s on the way to the earth can transfer into neutrinos of the other types that are not detected in the radiochemical Homestake, GALLEX and SAGE experiments. In Kamiokande and Super–Kamiokande experiments all flavor neutrinos $`\nu _e`$, $`\nu _\mu `$ and $`\nu _\tau `$ are detected. However, the cross section of $`\nu _\mu `$ ($`\nu _\tau `$) $`e`$ scattering is about six times smaller than the cross section of $`\nu _ee`$ scattering.
All existing solar neutrino data can be explained if we assume that solar neutrino fluxes are given by the Standard Solar Model (SSM) and that there are transitions between two neutrino types determined by the two parameters: mass squared difference $`\mathrm{\Delta }m^2`$ and mixing parameter $`\mathrm{sin}^22\theta `$. We will present the results of such analysis of the data later on.
Now we will make some remarks about a model independent analysis of the data. First of all from the luminosity relation (7.179) for the total flux of solar neutrinos we have the following lower bound
$$\mathrm{\Phi }=\underset{i}{}\mathrm{\Phi }_i\frac{L_{}}{2\pi R^2Q}$$
(7.183)
Furthermore, for the counting rate in the gallium experiments we have
$$Q_{Ga}=_{E_{th}}\sigma (E)I_i(E)dE=\underset{i}{}\overline{\sigma }_i\mathrm{\Phi }_i\overline{\sigma }_{pp}\mathrm{\Phi }=(76\pm 2)\mathrm{SNU}$$
(7.184)
By comparing this lower bound with the results of the GALLEX and SAGE experiments (see Table 7.1) we come to the conclusion that there is no contradiction between experimental data and luminosity constraint if we assume that there are no transitions of solar neutrinos into other states ($`P(\nu _e\nu _e)=1`$).
It is possible, however, to show in a model independent way that the results of different solar neutrino experiments are not compatible if we assume $`P(\nu _e\nu _e)=1`$. In fact, let us compare the results of the Homestake and the Super–Kamiokande experiments. We will consider the total neutrino fluxes $`\mathrm{\Phi }_i`$ as free parameters. From the results of Super-Kamiokande experiment we can determine the flux of <sup>8</sup>B neutrinos, $`\mathrm{\Phi }_{{}_{}{}^{8}B}`$ (see Table7.1). If we calculate now the contribution of <sup>8</sup>B neutrinos into the counting rate of the Homestake experiment we get
$$Q_{Cl}^{{}_{}{}^{8}B}=(2.78\pm 0.27)\mathrm{SNU}$$
(7.185)
The difference between measured event rate and $`Q_{Cl}^{{}_{}{}^{8}B}`$ gives the contribution to the Chlorine event rate of <sup>7</sup>Be and other neutrinos. We have
$$Q_{Cl}^{{}_{}{}^{7}Be+\mathrm{}}=Q_{Cl}^{ex}Q_{Cl}^{{}_{}{}^{8}B}=(0.22\pm 0.35)\mathrm{SNU}$$
(7.186)
All existing solar models predict much larger contribution of <sup>7</sup>Be neutrinos to the Chlorine event rate:
$$Q_{Cl}^{{}_{}{}^{7}Be}(SSM)=(1.15\pm 0.1)SNU$$
(7.187)
The large suppression of the flux of <sup>7</sup>Be neutrinos (together with the observation of <sup>8</sup>B neutrinos) is the problem for any solar model. The <sup>8</sup>B nuclei are produced in the reaction $`p+{}_{}{}^{7}\mathrm{Be}{}_{}{}^{8}\mathrm{B}+\gamma `$ and in order to observe neutrinos from <sup>8</sup>B decay enough <sup>7</sup>Be nuclei must exist in the Sun interior. We can come to the same conclusion about the suppression of the flux of <sup>7</sup>Be neutrinos if we compare the results of Gallium and Super–Kamiokande experiments.
All existing solar neutrino data can be described if there are oscillation between two neutrino flavors, the neutrino fluxes being given by the SSM values. If we assume that the oscillation parameters $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta `$ are in the region in which matter MSW effect can be important, then from the fit of the data two allowed regions of the oscillation parameters can be obtained. For the best fit values it was found
$`\mathrm{\Delta }m^2`$ $`=510^6\mathrm{eV}^2\mathrm{sin}^22\theta =510^3`$ $`(\mathrm{SMA})`$ (7.188)
$`\mathrm{\Delta }m^2`$ $`=210^5\mathrm{eV}^2\mathrm{sin}^22\theta =0.76`$ $`(\mathrm{LMA})`$ (7.189)
The data can be also described if we assume that the oscillation parameters are in the region in which matter effects can be neglected (the case of vacuum oscillations). For the best fit values it was found in this case
$$\mathrm{\Delta }m^2=4.310^{10}\mathrm{eV}^2\mathrm{sin}^22\theta =0.79(\mathrm{VO}).$$
(7.190)
In the Super–Kamiokande experiment during 825 days 11240 solar neutrino events were observed. Such a large statistics allows the Super–Kamiokande collaboration to measure the energy spectrum of the recoil electrons and day/night asymmetry. No significant deviation from the expected spectrum was observed (may be with the exception of the high energy part of the spectrum). For the day/night asymmetry the following value was obtained
$$\frac{1}{2}\left(\frac{ND}{N+D}\right)=0.065\pm 0.031\pm 0.013$$
(7.191)
From the analysis of the latest Super–Kamiokande data the following best–fit values of the oscillation parameters were found:
$`\mathrm{\Delta }m^2`$ $`=510^6\mathrm{eV}^2\mathrm{sin}^22\theta =510^3`$ $`(\mathrm{SMA})`$ (7.192)
$`\mathrm{\Delta }m^2`$ $`=3.210^5\mathrm{eV}^2\mathrm{sin}^22\theta =0.8`$ $`(\mathrm{LMA})`$ (7.193)
$`\mathrm{\Delta }m^2`$ $`=4.310^{10}\mathrm{eV}^2\mathrm{sin}^22\theta =0.79`$ $`(\mathrm{VO})`$ (7.194)
These values are compatible with the ones in Eq. (7.188), (7.189) and (7.190), which were found from the analysis of the event rates measured in all solar neutrino experiments.
The new solar neutrino experiment SNO started recently in Canada. The target in this experiment is heavy water (1 kton of $`D_2O`$) and Cerenkov light is detected by $`10^4`$ photomultipliers. Neutrinos will be detected through the observation of the CC reaction
$$\nu _e+de^{}+p+p$$
(7.195)
as well as of the NC reaction
$$\nu +d\nu +n+p$$
(7.196)
and $`\nu e`$ elastic scattering
$$\nu +e\nu +e$$
(7.197)
The detection of neutrinos via the CC process (7.195) will allow to measure the spectrum of $`\nu _e`$ on the Earth. The detection of neutrinos via the NC process (7.196) (neutrons will be detected) will allow to determine the total flux of flavor neutrinos $`\nu _e,\nu _\mu ,\nu _\tau `$. From the comparison of NC and CC event rates model independent conclusions on the transition of solar $`\nu _e`$’s into other flavor states can be made.
Next solar neutrino experiment will be BOREXINO. In this experiment 300 tons of liquid scintillator of very high purity will be used. Solar neutrinos will be detected through the observation of the recoil electrons in the process
$$\nu +e\nu +e.$$
(7.198)
The energy threshold in the BOREXINO experiment will be very low, about 250 keV. That will allow to detect the monoenergetic <sup>7</sup>Be neutrinos. If vacuum oscillations are the origin of the solar neutrino problem, a seasonal variation of the <sup>7</sup>Be neutrino signal (due to excentricity of the Earth orbit) will be observed.
### 7.2 Atmospheric neutrinos
Atmospheric neutrinos are produced mainly in the decays of pions and muons
$$\pi \mu +\nu _\mu ,\mu e+\nu _e+\nu _\mu $$
(7.199)
pions being produced in the interaction of cosmic rays in the Earth atmosphere. Notice that in the existing detectors neutrino and antineutrino events cannot be distinguished. At small energies, $`1`$ GeV, the ratio of fluxes of $`\nu _\mu `$’s and $`\nu _e`$’s from the chain (7.199) is equal to two. At the higher energies this ratio is larger than two (not all muons decay in the atmosphere) but it can be predicted with accuracy better than 5% (the absolute fluxes of muon and electron neutrinos are predicted presently with accuracy not better than 20 – 25%). This is the reason why the results of the measurements of total fluxes of atmospheric neutrinos are presented in the form of a double ratio
$$R=\frac{\left(N_\mu /N_e\right)_{\mathrm{data}}}{\left(N_\mu /N_e\right)_{\mathrm{MC}}}$$
(7.200)
where $`(N_\mu /N_e)_{\mathrm{data}}`$ is the ratio of the total number of observed muon and electron events and $`(N_\mu /N_e)_{\mathrm{MC}}`$ is the ratio predicted from Monte Carlo simulations.
We will discuss the results of the Super–Kamiokande experiment. A large water Cerenkov detector is used in this experiment. The detector consists of two parts: the inner one of 50 kton (22.5 kton fiducial volume) is covered with 11146 photomultipliers and the outer part, 2.75 m thick, is covered with 1885 photomultipliers. The electrons and muons are detected through the observation of the Cerenkov radiation. The efficiency of particle identification is larger than 98%. The observed events are divided in fully contained events (FC) for which Cerenkov light is deposited in the inner detector and partially contained events (PC) in which the muon track deposits part of its Cerenkov radiation in the outer detector. FC events are further divided into sub-GeV events ($`E_{vis}`$ 1.33 GeV) and multi-GeV events $`E_{vis}`$ 1.33 GeV). In the Super-Kamiokande experiment for sub-GeV events and multi-GeV events (FC and PC) the following values of the double ratio $`R`$ were obtained, respectively (848.3 days):
$$\begin{array}{cc}\hfill R& =0.680_{0.022}^{+0.023}\pm 0.053\hfill \\ & \\ \hfill R& =0.678_{0.039}^{+0.042}\pm 0.080\hfill \end{array}$$
(7.201)
These values are in agreement with the values of $`R`$ obtained in other water Cerenkov experiments (Kamiokande and IMB) and in the Soudan2 experiment in which the detector is iron calorimeter.
$`R`$ $`=0.65\pm 0.05\pm 0.08(\mathrm{Kamiokande})`$ (7.202)
$`R`$ $`=0.54\pm 0.05\pm 0.11(\mathrm{IMB})`$ (7.203)
$`R`$ $`=0.61\pm 0.15\pm 0.05(\mathrm{Soudan2})`$ (7.204)
The fact that the double ratio $`R`$ is significantly less than one is an indication in favor of neutrino oscillations.
The important evidence in favour of neutrino oscillations was obtained by the Super–Kamiokande collaboration. These data were first reported at NEUTRINO98 conference in Japan, in June 1998. A significant up–down asymmetry of multi–GeV muon events was discovered in the Super–Kamiokande experiment.
For atmospheric neutrinos the distance between production region and detector changes from about 20 km for down–going neutrinos ($`\theta =0`$, $`\theta `$ being the zenith angle) up to about 13,000 km for up–going neutrinos ($`\theta =\pi `$). In the Super–Kamiokande experiment for the multi–GeV events the zenith angle $`\theta `$ can be determined. In fact, charged leptons follow the direction of neutrinos (the averaged angle between the charged lepton and the neutrino is $`15^o20^o`$). The possible source of the zenith angle dependence of neutrino events is the magnetic field of the Earth. However, for neutrinos with energies larger than 2 – 3 GeV, within a few % no $`\theta `$-dependence of neutrino events is expected.
The Super-Kamiokande collaboration found a significant zenith angle dependence of the multi–GeV muon neutrinos. For the integral up–down asymmetry of multi–GeV muon neutrinos (FC and PC) the following value was obtained
$$A_\mu =0.311\pm 0.043\pm 0.010$$
(7.205)
Here
$$A=\frac{UD}{U+D}$$
(7.206)
where $`U`$ is the number of up–going neutrinos ($`\mathrm{cos}\theta 0.2`$) and $`D`$ is the number of down–going neutrinos (($`\mathrm{cos}\theta 0.2`$). No asymmetry of the electron neutrinos was found:
$$A_e=0.036\pm 0.067\pm 0.02$$
(7.207)
The Super–Kamiokande data can be described if we assume that there are $`\nu _\mu \nu _\tau `$ oscillations. The following best–fit values of the oscillation parameters were found from the analysis of FC events
$$\mathrm{\Delta }m^2=3.0510^3eV^2,\mathrm{sin}^22\theta =0.995$$
(7.208)
($`\chi _{min}^2=55.4`$ at 67 d.o.f.). Let us notice that if we assume that there are no oscillations, then in this case $`\chi ^2=177`$ at 69 d.o.f. From the combined analysis of all data it was found
$$\mathrm{\Delta }m^2(26)10^3eV^2,\mathrm{sin}^22\theta >0.84$$
(7.209)
If $`\nu _\mu \nu _s`$ oscillations are assumed, at large energies matter effects must be important. From the investigation of the high energy events (PC and upward–going muon events, muons being produced by neutrinos in the rock under the detector) the Super–Kamiokande collaboration came to the conclusion that $`\nu _\mu \nu _s`$ oscillations are disfavoured at 95% C.L.
The range of oscillation parameters which was obtained from the analysis of the atmospheric neutrino data will be investigated in details in long–baseline experiments. The results of the first LBL reactor experiment, CHOOZ, were recently published (in this experiment the distance between reactors and detector is $`1`$ km). No indication in favour ofthe transitions of $`\overline{\nu _e}`$ into other states was found in this experiment. For the ratio $`R`$ of the number of measured and expected events it was found
$$R=1.01\pm 2.8\%(\mathrm{stat})\pm 2.7\%(\mathrm{syst})$$
(7.210)
These data allow to exclude $`\mathrm{\Delta }m^2>710^4\mathrm{eV}^2`$ at $`\mathrm{sin}^22\theta =1`$ (90% C.L.).
In LBL Kam-Land experiment $`\overline{\nu _e}`$’s from reactors at the distance of $`150200`$ km from the detector will be detected. Neutrino oscillations $`\overline{\nu }_e\overline{\nu }_x`$ with $`\mathrm{\Delta }m^2>10^5\mathrm{eV}^2`$ and large values of $`\mathrm{sin}^22\theta `$ will be explored. The BOREXINO collaboration plans to detect $`\overline{\nu }_e`$ from reactors at the distance of about 800 km from the detector.
The first LBL accelerator experiment K2K is running now. In this experiment $`\nu _\mu `$’s with average energy of $`1.4`$ GeV, produced at KEK accelerator, will be detected in the Super–Kamiokande detector (at a the distance of about $`250`$ km). The disappearance channel $`\nu _\mu \nu _\mu `$ and the appearance channel $`\nu _\mu \nu _e`$ will be investigated in detail. This experiment will be sensitive to $`\mathrm{\Delta }m^2210^3\mathrm{eV}^2`$ at large $`\mathrm{sin}^22\theta `$.
The LBL MINOS experiment between Fermilab and Soudan (the distance is of about $`730`$ km) is under the construction. In this experiment all the possible channels of $`\nu _\mu `$ transitions will be investigated in the atmospheric neutrino range of $`\mathrm{\Delta }m^2`$.
The LBL CERN-Gran Sasso experiments (the distance is of about $`730`$ km) ICARUS, NOE and others, are under constraction at CERN and Gran Sasso. The direct detection of $`\tau `$’s from $`\nu _\mu \nu _\tau `$ transition will be one of the major goal of these experiments.
### 7.3 LSND experiment
Some indications in favour of $`\nu _\mu \nu _e`$ oscillations were found in short–baseline LSND accelerator experiment. This experiment was done at the Los Alamos linear accelerator (with protons of $`800`$ MeV energy). This is a beam–stop experiment: most of $`\pi ^+`$’s in the beam, produced by protons, come to a rest in the target and decay (mainly by $`\pi ^+\mu ^+\nu _\mu `$); $`\mu ^+`$’s also come to a rest in the target and decay by $`\mu _+e^+\nu _e\overline{\nu _\mu }`$. Thus, the beam–stop target is the source of $`\nu _\mu ,\nu _e`$ and $`\overline{\nu }_\mu `$ (no $`\overline{\nu }_e`$ are produced in the decays).
The large scintillator neutrino detector LSND was located at a distance of about 30 m from the neutrino source. In the detector $`\nu _e`$’s were searched for through the observation of the process
$$\overline{\nu }_e+pe^++n$$
(7.211)
Both $`e^+`$ and delayed 2.2 MeV $`\gamma `$’s from the capture $`npd\gamma `$ were detected.
In the LSND experiment $`33.9\pm 8.0`$ events were observed in the interval of $`e^+`$ energies $`30<E<60`$ MeV. Assuming that these events are due to $`\overline{\nu }_\mu \overline{\nu }_e`$ transitions, for the transition probability it was found
$$P(\overline{\nu }_\mu \overline{\nu }_e)=(0.31\pm 0.09\pm 0.06)10^3$$
(7.212)
From the analysis of LSND data the allowed region in $`\mathrm{sin}^22\theta \mathrm{\Delta }m^2`$ plot was obtained. If the results of SBL reactor experiments and SBL accelerator experiments on the search for $`\nu _\mu \nu _e`$ transitions are taken into account for the allowed values of the oscillation parameters it was found
$$0.2<\mathrm{\Delta }m^2<2\mathrm{e}\mathrm{V}^2210^3<\mathrm{sin}^22\theta <410^2$$
(7.213)
The indications in favour of $`\nu _\mu \nu _e`$ oscillations obtained in the LSND experiment will be checked by BOONE experiment at Fermilab, scheduled for 2001-2002.
## 8 Conclusions
The problem of neutrino masses and mixing is the central problem of today’s neutrino physics. More than 40 different experiments all over the world are dedicated to the investigation of this problem and many new experiments are in preparation. The investigation of the properties of neutrinos is one of the most important direction in the search for a new scale in physics. These investigations will be very important for the understanding of the origin of tiny neutrino masses and neutrino mixing which, according to the existing data, is very different from CKM quark mixing.
If all existing data will be confirmed by the future experiments it would mean that at least four massive neutrinos exist in nature (in order to to provide three independent neutrino mass squared differences: $`\mathrm{\Delta }m_{\mathrm{solar}}^210^5\mathrm{eV}^2`$ (or $`10^{10}`$ eV<sup>2</sup>), $`\mathrm{\Delta }m_{\mathrm{atm}}^210^3\mathrm{eV}^2`$ and $`\mathrm{\Delta }m_{\mathrm{LSND}}^21\mathrm{e}\mathrm{V}^2`$ ). From the phenomenological analysis of all existing data it follows that in the spectrum of masses of four massive neutrinos there are two close masses separated by the ”large” one, by the about 1 eV LSND gap. Taking into account big–bang nucleosynthesis constraint on the number of neutrinos it can be shown that the dominant transition of the solar neutrinos is $`\nu _e\nu _{\mathrm{sterile}}`$ one and the dominant transition of the atmospheric neutrinos is $`\nu _\mu \nu _\tau `$.
If the LSND indication in favour of $`\nu _\mu \nu _e`$ oscillations will be not confirmed by the future experiments, the mixing of three massive neutrinos with mass hierarchy is plausible scenario.
The nature of massive neutrinos (Dirac or Majorana?) can be determined from the experiments on the search for neutrinoless double $`\beta `$\- decay. It can be shown that from the existing neutrino oscillation data it follows that effective Majorana mass $`<m>`$ in the case of three massive Majorana neutrinos with mass hierarchy is not larger than $`10^2`$ eV (the present bound is $`|<m>|0.5`$ eV and the sensitivity of the next generation of experiments will be $`|<m>|0.1`$ eV).
The sensitivity $`|<m>|10^2`$ eV is very important problem of experiments on the search for neutrinoless double $`\beta `$-decay.
I would like to express my deep gratitude to R. Bernabei, W.M. Alberico and S. Bilenkaia for their great help in preparing these lecture notes. |
warning/0001/hep-th0001043.html | ar5iv | text | # A new massive vector field theory
## 1 Introduction
After the introduction of Dirac’s equation, the search began for similar equations for higher spins. In the past, various approaches have been tried$``$ equations describing many masses and spins particle, non-Lagrangian theories and theories with indefinite metric et al. At first, it was observed that, apart from spin-1/2, none of the other spins obeys a single-particle relativistic wave equation. For example, it was generally believed that for spins 0 and 1, the Klein-Gordon- and Proca equations were unique, respectively. However, more than 60 years ago, it was found that the Kemmer-Duffin-Petian- equations(KDF equations) can describe both spin-0 and spin-1 objects. Since then, many more systems of equations for arbitrary spins, which originate from different assumptions after considering their invariance under Lorentz group, have been found.
Unfortunately, it has been known for a long time that there still exist many difficulties in the construction of higher spins field theories, which has turned out to be the most intriguing and challenging in theoretical physics. Especially, such a theory has touched upon some of the most basic ingredients of present-day physical theory. For example, in those theories, either the usual connection between spin and statics is violated, or the law of causation do not hold, or the negative energy difficulty is still encountered after second quantization having been accomplished, or, in the presence of interactions, the complex energy eigenvalues, superluminal propagation of waves and many other undesirable features- were found too. In particular, these behaviors were exhibited by both the KDP and the Proca equations.
On the other hand, both the Maxwell equations for electromagnetic field and Yang-Mills equations for non-Abelian gauge field can be restated in terms of a spinor notation- resembling the one for Dirac equation, which motivates people to extend the idea to a massive system with arbitrary spin-. In this paper, we set up a new massive vector field equation(called as Dirac-like equation), which takes a form similar to the Dirac equation but involves some six-by-six matrices. The equation is no longer equivalent to any other existed ones such as the KDP equation(in the spin-1 case), the Proca equation and the Weinberg equation, , etc. Moreover, it is observed, on one hand, the general solution of a relativistic wave equation can be, not only the sum of positive-frequency and negative-frequency parts(denoted by $`\phi _1`$), but also the difference of them(denoted by $`\phi _2`$). On the other hand, the positive-frequency and negative-frequency solutions of an equation are linearly independent such that $`\phi _1`$ and $`\phi _2`$ transform in the same way under the Lorentz and the gauge transformations, Therefore, these two types of general solutions($`\phi _1`$ and $`\phi _2`$) are simultaneously used to construct the Lagrangian of the vector field. As a result, all those difficulties mentioned above are swept away.
The paper is organized as follows. In Sec. II, the Dirac-like equation is put forward and the corresponding plane wave solution is discussed. In Sec. III, by choosing a suitable Lagrangian, from which the Dirac-like equation can be derived, we quantilize the Dirac-like field naturally according to Bose-Einstein statistics, where the energy is positive-definite too. In Sec. IV, an interesting feature of the field is shown. From the Lagrangian, we construct the Feynman propagator for the field in Sec. V, with the causality being preserved. In Sec. VI, the Lorentz invariance of the theory is discussed, where the action of the transversal field is proved to be Lorentz invariant. In Sec. VII, we develop the Feynman rules for the vector QED, where, as an example, the polarization cross section for the process $`e^+e^{}f^+f^{}`$ is calculated. At last, in Sec. VIII, we show that vector QED is a renormalizable theory. The system of natural units and Bjorken conventions are used throughout in the paper.
## 2 A new massive vector field equation
In analogy with the construction of Dirac equation, we can set up the following free relativistic ”Dirac-like” equation as follow:
$`(i\beta ^\mu _\mu m)\phi (x)=0,`$ (1)
where $`m`$ is the mass and $`\beta ^\mu =(\beta ^0,\stackrel{}{\beta })`$ satisfies
$`\{\begin{array}{cc}& (\stackrel{}{\beta }\stackrel{}{p})^3=\stackrel{}{p}^2(\stackrel{}{\beta }\stackrel{}{p})\hfill \\ & \beta ^0\beta ^i+\beta ^i\beta ^0=0\hfill \\ & (\beta ^0)^2=1,\hfill \end{array}`$ (5)
where $`\stackrel{}{p}`$ is an arbitrary three dimensional vector such as the spatial component of a 4-momentum. To express the $`\beta `$ matrix explicitly, we choose
$`\begin{array}{cc}\beta ^0=\left(\begin{array}{ccc}I_{3\times 3}\hfill & 0& \\ 0\hfill & I_{3\times 3}& \end{array}\right),& \stackrel{}{\beta }=\left(\begin{array}{ccc}0\hfill & \stackrel{}{\tau }& \\ \stackrel{}{\tau }\hfill & 0& \end{array}\right),\end{array}`$ (11)
where $`I_{3\times 3}`$ is the $`3\times 3`$ unit matrix and $`\stackrel{}{\tau }=(\tau _1,\tau _2,\tau _3)`$, in which
$`\begin{array}{ccc}\tau _1=\left(\begin{array}{ccc}0\hfill & 0& \hfill 0\\ 0\hfill & 0& \hfill i\\ 0\hfill & i& \hfill 0\end{array}\right),& \tau _2=\left(\begin{array}{ccc}0\hfill & 0& \hfill i\\ 0\hfill & 0& \hfill 0\\ i\hfill & 0& \hfill 0\end{array}\right),& \tau _3=\left(\begin{array}{ccc}0\hfill & i& \hfill 0\\ i\hfill & 0& \hfill 0\\ 0\hfill & 0& \hfill 0\end{array}\right).\end{array}`$ (22)
It is easy to check that $`\beta ^\mu `$ given above satisfy the relation (5). Let $`\stackrel{}{\alpha }=\beta ^0\stackrel{}{\beta }`$, Eq. (1) can be rewritten as
$`i_t\phi (x)=(\stackrel{}{\alpha }\widehat{p}+\beta ^0m)\phi (x),`$ (23)
where $`\widehat{p}=i`$ is the momentum operator. Accordingly, the Hamiltonian is $`\widehat{H}=\stackrel{}{\alpha }\widehat{p}+\beta ^0m`$. Let $`\widehat{L}=\stackrel{}{x}\times \widehat{p}`$ represent the operator of orbit angular momentum, we have
$`\begin{array}{cc}[\widehat{H},\widehat{L}]=i\stackrel{}{\alpha }\times \widehat{p},& [\widehat{H},\widehat{L}+\stackrel{}{S}]=0,\end{array}`$ (25)
where
$`\stackrel{}{S}=\left(\begin{array}{ccc}\stackrel{}{\tau }\hfill & 0& \\ 0\hfill & \stackrel{}{\tau }& \end{array}\right)`$ (28)
is the spin matrix. Since $`\stackrel{}{S}^2=2`$, the corresponding particle(called Dirac-like particle) has spin 1, which will be demonstrated further in Sec. VI. Namely, Eq. (1) represents the equation of a massive vector field. As we will see later, the Dirac-like field is different from any other vector fields, and hence is a completely new one.
Substituting the plane wave solution $`\phi (p)e^{ipx}`$ into Eq. (23), we obtain the representation of Eq. (23) in momentum space as follow:
$`(E\stackrel{}{\alpha }\widehat{p}\beta ^0m)\phi (p)=0,`$ (29)
where $`E=p^0`$ is the energy. When $`\{\widehat{H},\widehat{p},\frac{\stackrel{}{s}\widehat{p}}{|\stackrel{}{p}|}\}`$ are chosen as dynamical completeness operators, the fundamental solutions of Eq. (29) are derived
$`\begin{array}{cc}|\stackrel{}{p},s_+=\sqrt{\frac{E_s+m}{2m}}\left(\begin{array}{cc}\eta _s& \\ \frac{\stackrel{}{\tau }\stackrel{}{p}}{E_s+m}\eta _s& \end{array}\right),& |\stackrel{}{p},s_{}=\sqrt{\frac{E_s+m}{2m}}\left(\begin{array}{cc}\frac{\stackrel{}{\tau }\stackrel{}{p}}{E_s+m}\eta _s& \\ \eta _s& \end{array}\right),\end{array}`$ (35)
where $`|\stackrel{}{p},s_+`$ and $`|\stackrel{}{p},s_{}`$ correspond to the positive and the negative energy solutions, respectively. $`s=1,0,1`$ and
$`E_1=E_1=E_{}=\sqrt{\stackrel{}{p}^2+m^2},`$ $`E_0=E_L=m,`$ (36)
$`\eta _1={\displaystyle \frac{1}{\sqrt{2}|\stackrel{}{p}|}}\left(\begin{array}{ccc}\frac{p_1p_3ip_2|\stackrel{}{p}|}{p_1ip_2}\hfill & & \\ \frac{p_2p_3+ip_1|\stackrel{}{p}|}{p_1ip_2}\hfill & & \\ (p_1+ip_2)\hfill & & \end{array}\right),`$ $`\eta _1=\eta _1^{},`$ $`\eta _0={\displaystyle \frac{1}{|\stackrel{}{p}|}}\left(\begin{array}{ccc}p_1\hfill & & \\ p_2\hfill & & \\ p_3\hfill & & \end{array}\right).`$ (43)
where $`\eta _1^{}`$ is the complex conjugate of $`\eta _1`$. Let $`\eta _s^{}`$ stands for the Hermitian conjugate of $`\eta _s`$, we have
$`\{\begin{array}{cc}& \eta _s^{}\eta _s^{}=\delta _{ss^{}}\hfill \\ & \underset{s}{}\eta _s\eta _s^{}=I_{3\times 3},\hfill \end{array}`$ (46)
which means that $`\{\eta _s\}`$ form a complete orthonormal basis. It can be verified that
$`{\displaystyle \frac{1}{|\stackrel{}{p}|}}\stackrel{}{\tau }\stackrel{}{p}\eta _s=\lambda _s\eta _s,`$ (47)
where $`\lambda _1=1`$, $`\lambda _1=1`$ and $`\lambda _0=0`$. Obviously, the solutions with $`s=\pm 1`$ correspond to the transversal polarization solutions of Eq. (29), while another one corresponds to the longitudinal polarization solution. Besides, an interesting result can be read from Eq. (36): the transversal polarization particles have energy $`\sqrt{\stackrel{}{p}^2+m^2}`$ while the longitudinal one has energy m, whose physical reason will be explained in Sec. IV.
As for the wave functions, the ones in momentum space are chosen as
$`\chi (\stackrel{}{p},s)=|\stackrel{}{p},s_+,`$ $`y(\stackrel{}{p},s)=|\stackrel{}{p},s_{}=\sqrt{{\displaystyle \frac{E_s+m}{2m}}}\left(\begin{array}{cc}\frac{\stackrel{}{\tau }\stackrel{}{p}}{E_s+m}\eta _s& \\ \eta _s& \end{array}\right).`$ (50)
and the corresponding ones in position space are
$`\phi _{p,s}(x)=\sqrt{{\displaystyle \frac{m}{vE_s}}}\chi (\stackrel{}{p},s)e^{ipx},\phi _{p,s}(x)=\sqrt{{\displaystyle \frac{m}{vE_s}}}y(\stackrel{}{p},s)e^{ipx}.`$ (51)
Taking into account Eq. (46), we have(here $`\overline{\chi }=\chi ^{}\beta ^0`$, and so on)
$`\{\begin{array}{cc}& \overline{\chi }(\stackrel{}{p},s)\chi (\stackrel{}{p},s^{})=\overline{y}(\stackrel{}{p},s)y(\stackrel{}{p},s^{})=\delta _{ss^{}}\hfill \\ & \overline{\chi }(\stackrel{}{p},s)y(\stackrel{}{p},s^{})=\overline{y}(\stackrel{}{p},s)\chi (\stackrel{}{p},s^{})=0.\hfill \end{array}`$ (54)
$`\{\begin{array}{cc}& \underset{s}{}\chi (\stackrel{}{p},s)\overline{\chi }(\stackrel{}{p},s)=\frac{\beta p_{}+m}{2m}(I_{2\times 2}\underset{s=\pm 1}{}\eta _s\eta _s^{})+\frac{\beta p_L+m}{2m}(I_{2\times 2}\eta _0\eta _0^{})\hfill \\ & \underset{s}{}y(\stackrel{}{p},s)\overline{y}(\stackrel{}{p},s)=\frac{\beta p_{}m}{2m}(I_{2\times 2}\underset{s=\pm 1}{}\eta _s\eta _s^{})+\frac{\beta p_Lm}{2m}(I_{2\times 2}\eta _0\eta _0^{}).\hfill \end{array}`$ (57)
where $``$ is the direct product symbol. Obviously, Eq. (54) and Eq. (57) complete the orthonormality relations and the spin summation relations for the massive vector field, from which the completeness relation can be derived too.
## 3 Quantization of the Dirac-like field
In this section, we will quantize the Dirac-like field in a manner similar to that used for Dirac field but obeying the Bose-Einstein statistics. Two types of general solutions of Eq. (1) can be constructed simultaneously as follows:
$`\phi _1(x)\phi _+(x)+\phi _{}(x)={\displaystyle \underset{\stackrel{}{p},s}{}}[a(\stackrel{}{p},s)\phi _{p,s}(x)+b^{}(\stackrel{}{p},s)\phi _{p,s}(x)],`$ (58)
$`\phi _2(x)\phi _+(x)\phi _{}(x)={\displaystyle \underset{\stackrel{}{p},s}{}}[a(\stackrel{}{p},s)\phi _{p,s}(x)b^{}(\stackrel{}{p},s)\phi _{p,s}(x)],`$ (59)
where
$`\phi _+(x)={\displaystyle \underset{\stackrel{}{p},s}{}}a(\stackrel{}{p},s)\phi _{p,s}(x),`$ $`\phi _{}(x)={\displaystyle \underset{\stackrel{}{p},s}{}}b^{}(\stackrel{}{p},s)\phi _{p,s}(x),`$ (60)
are the positive and the negative frequency parts of the general solutions, respectively, and $`a(\stackrel{}{p},s)`$, $`b^{}(\stackrel{}{p},s)`$ are coefficients.
To deduce the free Dirac-like equation, the free Lagrangian is chosen as
$`L=\overline{\phi }_2(x)(i\beta ^\mu _\mu m)\phi _1(x).`$ (61)
Then the canonical momentum conjugates to $`\phi _1(x)`$ is
$`\pi (x)=L/\dot{\phi _1}=i\phi _2^{}(x),`$ (62)
and thus, the Hamiltonian $`H`$ and the momentum $`\stackrel{}{p}`$ are, respectively,
$`\begin{array}{c}H=[\pi (x)\dot{\phi }_1(x)L]d^3x=\phi _2^{}(i\stackrel{}{\alpha }+\beta ^0m)\phi _1(x)d^3x,\\ \stackrel{}{p}=\pi (x)\phi _1d^3x=i\phi _2^{}(x)\phi _1d^3x.\end{array}`$ (65)
Now, we promote $`\phi _1(x)`$ and $`\pi (x)`$ to operators and the canonical equal time commutation relations become
$`\begin{array}{cc}[\phi _{1\alpha }(\stackrel{}{x},t),\pi _\beta (\stackrel{}{x}^{},t)]=i\delta _{\alpha \beta }\delta ^3(\stackrel{}{x}\stackrel{}{x}^{}),\hfill & \end{array}`$ (67)
with the others vanishing. In term of $`a(\stackrel{}{p},s)`$ and $`b(\stackrel{}{p},s)`$, we get the following commutation relations
$`[a(\stackrel{}{p},s),a^{}(\stackrel{}{p}^{},s^{})]=[b(\stackrel{}{p},s),b^{}(\stackrel{}{p}^{},s^{})]=\delta _{\stackrel{}{p}\stackrel{}{p}^{}}\delta _{ss^{}},`$ (68)
and all other commutators vanish. Making use of Eq. (54), (58), (59) and (68), Eq. (65) transforms into
$`\begin{array}{c}H=\underset{\stackrel{}{p},s}{}E_s[a^{}(\stackrel{}{p},s)a(\stackrel{}{p},s)+b^{}(\stackrel{}{p},s)b(\stackrel{}{p},s)+1],\hfill \\ \stackrel{}{p}=\underset{\stackrel{}{p},s}{}[a^{}(\stackrel{}{p},s)a(\stackrel{}{p},s)+b^{}(\stackrel{}{p},s)b(\stackrel{}{p},s)],\hfill \end{array}`$ (71)
where $`a^{}(\stackrel{}{p},s)`$, $`a(\stackrel{}{p},s)`$ are the creation and annihilation operators of particles, respectively, while $`b^{}(\stackrel{}{p},s)`$, $`b(\stackrel{}{p},s)`$ are the corresponding ones of antiparticles. Obviously, the Dirac-like field obeys the Bose-Einstein statics and the corresponding energy is positive-definite, which is just what we desire.
## 4 Character of the new vector field
Let us turn our attention to the expectation value $`\stackrel{}{v}`$ of the velocity operator $`\dot{\stackrel{}{x}}=i[\widehat{H},\stackrel{}{x}]=\stackrel{}{\alpha }`$
$`\stackrel{}{v}={\displaystyle \phi _2^{}(x)\dot{\stackrel{}{x}}\phi _1(x)d^3x}={\displaystyle \phi _2^{}\stackrel{}{\alpha }\phi _1d^3x}.`$ (72)
In fact, $`\phi _2^{}\stackrel{}{\alpha }\phi _1`$ can be regarded as the probability current density and hence $`\stackrel{}{v}`$ corresponds to the probability current. Substituting Eq. (LABEL:var1), (LABEL:var2) into Eq. (72), we obtain
$`\stackrel{}{v}=\stackrel{}{v}_{}+\stackrel{}{v}_L,`$ (73)
where $`\stackrel{}{v}_L`$ represents the current related to transversal particle only, while $`\stackrel{}{v}_{}`$ corresponds to the current with the contribution of longitudinal particles involved in too. The explicit expression of them are
$`\begin{array}{cc}\stackrel{}{v}_L\hfill & =\underset{\stackrel{}{p}}{}\underset{s=\pm 1}{}\frac{\stackrel{}{p}}{E}[a^{}(\stackrel{}{p},s)a(\stackrel{}{p},s)b^{}(\stackrel{}{p},s)b(\stackrel{}{p},s)]\hfill \\ & +\underset{\stackrel{}{p}}{}\underset{s=\pm 1}{}(1)^{\frac{(s1)}{2}}\frac{m}{E}\stackrel{}{\eta }_0[a^{}(\stackrel{}{p},s)b^{}(\stackrel{}{p},s)e^{i2Et}a(\stackrel{}{p},s)b(\stackrel{}{p},s)e^{i2Et}],\hfill \\ \stackrel{}{v}_{}\hfill & =\underset{\stackrel{}{p}}{}\underset{s=\pm 1}{}\sqrt{\frac{1}{2E(E+m)}}|\stackrel{}{p}|\{[a^{}(\stackrel{}{p},0)a(\stackrel{}{p},s)e^{i(mE)t}+b^{}(\stackrel{}{p},s)b(\stackrel{}{p},0)e^{i(mE)t}]\stackrel{}{\eta }_s+h.c.\}\hfill \\ & +\underset{\stackrel{}{p}}{}\underset{s=\pm 1}{}\sqrt{\frac{E+m}{2E}}\{[a^{}(\stackrel{}{p},0)b^{}(\stackrel{}{p},s)\stackrel{}{\eta }_s+a^{}(\stackrel{}{p},s)b^{}(\stackrel{}{p},0)\stackrel{}{\eta }_s^{}]e^{i(m+E)t}h.c.\},\hfill \end{array}`$ (78)
respectively, in which $`E=E_{}=\sqrt{\stackrel{}{p}^2+m^2}`$ and $`\stackrel{}{\eta }_s`$ is the vector representation of Eq. (43).
Before going on, we will give some discussions about Eq. (78):
(1) As far as $`\stackrel{}{v}_L`$ is concerned, the first term corresponds to the classic current with group velocity $`\frac{\stackrel{}{p}}{E}`$ of the wave packet, while the second term corresponds to the zitterbewegung current of the transversal particle. The zitterbewegung current does not vanish as $`\stackrel{}{p}0`$, which infers that it is intrinsic and independent of macroscopic classic motion.
(2) As for $`\stackrel{}{v}_{}`$, the first term corresponds to the current resulting from the interference between the transversal and the longitudinal particle, and the second term corresponds to the zitterbewegung current containing the contribution of longitudinal particle.
(3) $`\stackrel{}{\eta _0}`$ is parallel to $`\stackrel{}{p}`$ while $`\stackrel{}{\eta }_{\pm 1}`$ is vertical to $`\stackrel{}{p}`$(denoted by $`\stackrel{}{\eta _0}\stackrel{}{p}`$ and $`\stackrel{}{\eta _0}\stackrel{}{p}`$, respectively), thus $`\stackrel{}{v}_L\stackrel{}{p}`$ while $`\stackrel{}{v}_{}\stackrel{}{p}`$(or, $`\stackrel{}{v}_{}\stackrel{}{p}=0`$), On the other hand, the longitudinal particle does not contribute to $`\stackrel{}{v}_L`$(which contributes only to $`\stackrel{}{v}_{}`$). Therefore, the longitudinal particle makes no contribution to the current in the direction of momentum $`\stackrel{}{p}`$, which is consistent with the statement that $`E_L=m`$ obtained earlier(see Eq. (36)). As a consequence, we can regard the longitudinal particle as the one corresponding to standing wave. In fact, the behaviors of the new vector field are similar to those of the electromagnetic wave propagating in hollow metallic waveguide, where the longitudinal component of the electromagnetic wave makes no contribution to the flow of energy(or the Poynting vector) along the waveguide.
## 5 The free propagator of the Dirac-like field
The free propagator of Dirac-like field is defined as
$`iR_f(x_1x_2)0|T\phi _1(x_1)\overline{\phi }_2(x_2)|0,`$ (79)
where T is the time order symbol. Considering Eq. (57), we obtain
$`iR_f(x_1x_2)=iR_f(x_1x_2)+iR_{fL}(x_1x_2),`$ (80)
where $`iR_f(x_1x_2)`$ is the free propagator of transversal field and $`iR_{fL}(x_1x_2)`$ is the longitudinal one, which read
$`iR_f(x_1x_2)={\displaystyle \frac{d^4p}{(2\pi )^4}\frac{i\mathrm{\Omega }_{}}{p_0^2E_{}^2+i\epsilon }e^{ip(x_1x_2)}}`$ (81)
$`iR_{fL}(x_1x_2)={\displaystyle \frac{d^4p}{(2\pi )^4}\frac{i\mathrm{\Omega }_L}{p_0^2E_L^2+i\epsilon }e^{ip(x_1x_2)}},`$
respectively, where $`\epsilon `$ is a infinitesimal real quantity, $`p_\mu =(p_0,\stackrel{}{p})`$ and
$`\mathrm{\Omega }_{}=(i\beta +m)A_{}`$ (82)
$`\mathrm{\Omega }_L=(i\beta +m)A_L,`$
in which $`A_{}=I_{2\times 2}\underset{s=\pm 1}{}\eta _s\eta _s^{}`$ and $`A_L=I_{2\times 2}\eta _0\eta _0^{}`$. By applying $`p_0^2E_{}^2=p^2m^2,(\beta p)^2A_{}=p^2A_{}`$ and $`p_0^2E_L^2=p_0^2m^2,(\beta p)^2A_L=p^2A_L`$, we have
$`iR_f(x_1x_2)={\displaystyle \frac{d^4p}{(2\pi )^4}\frac{iA_{}}{\beta pm+i\epsilon }e^{ip(x_1x_2)}}`$ (83)
$`iR_{fL}(x_1x_2)={\displaystyle \frac{d^4p}{(2\pi )^4}\frac{iA_L}{\beta pm+i\epsilon }e^{ip(x_1x_2)}}.`$
Due to $`A_{}+A_L=1`$,
$`iR_f(x_1x_2)={\displaystyle \frac{d^4p}{(2\pi )^4}\frac{i}{\beta pm+i\epsilon }e^{ip(x_1x_2)}},`$ (84)
which takes a form similar to the free propagator of Dirac field but involves the matrices $`\beta ^\mu `$ instead of Dirac matrices $`\gamma ^\mu `$
The representation of $`iR_f(x_1x_2)`$ in momentum space is
$`iR_f(p)=iR_f(p)+iR_{fL}(p)={\displaystyle \frac{i}{\beta pm+i\epsilon }},`$ (85)
where
$`iR_f(p)={\displaystyle \frac{i}{\beta pm+i\epsilon }}A_{},iR_{fL}(p)={\displaystyle \frac{i}{\beta pm+i\epsilon }}A_L.`$ (86)
It is easy to find that
$`(i\beta m)R_f(x_1x_2)=\delta ^4(x_1x_2).`$ (87)
Namely, $`R_f(x_1x_2)`$ is the Green’s function of free Dirac-like equation. To make Eq. (86) more explicit, we choose a frame in which $`\stackrel{}{p}=(0,0,p_3)`$, then
$`\begin{array}{cc}A_{}=I_{2\times 2}\left(\begin{array}{ccc}1\hfill & 0& \hfill 0\\ 0\hfill & 1& \hfill 0\\ 0\hfill & 0& \hfill 0\end{array}\right),A_L=I_{2\times 2}\left(\begin{array}{ccc}0\hfill & 0& \hfill 0\\ 0\hfill & 0& \hfill 0\\ 0\hfill & 0& \hfill 1\end{array}\right),& \end{array}`$ (95)
## 6 Lorentz invariance of the theory
Considering the fact that the positive and the negative frequency parts of Dirac-like field are linearly independent, one can readily verify that $`\phi _1(x)`$ and $`\phi _2(x)`$(given by Eq. (58) and (59), respectively) transform in the same way under a Lorentz transformation.
In following special cases, Lorentz boost along the direction of the Dirac-like field’s motion or Lorentz boost $`L(\stackrel{}{p})`$ which makes a Dirac-like particle from rest to momentum $`\stackrel{}{p}`$, the Lagrangian $`L_\phi `$ given by Eq. (61) is easily proved to be Lorentz invariant.
Now, we consider the variation of Lagrangian under an arbitrary Lorentz transformation. Before embarking on the process, the following fact should be noted: the longitudinal field makes no contribution to the current in the direction of the momentum of Dirac-like field and the energy of it is $`E_L=m`$, so we regard it as the one that exists in a standing wave form or in a virtual form. Meanwhile, for the independence of the longitudinal field with the interaction related to the current, the longitudinal field is taken as an unobservable one. Certainly, the unobservable longitudinal field in one frame can be turned into the observable transversal field in another frame and vice versa. However, the action of the transversal field is Lorentz invariant(just as we will show later). As far as the noncovariance of the transversal condition for the transversal field is concerned, it won’t do any hurt to the theory. The Lorentz invariant transition amplitude still can be obtained from a noncovariant Hamiltonian formulation of field theory, just as what we have done in electromagnetic field theory with the noncovariant Coulomb gauge. For the reasons mentioned above, we will demonstrate only the invariance of the action of transversal field in the following.
For generality, let us consider the case that the Lagrangian contains an interaction term, in which the Dirac-like field is coupled to the electromagnetic field in the minimal form, namely
$`L_\phi =\overline{\phi }_2(x)[i\beta D_\mu m]\phi _1(x),`$ (96)
where $`\phi _1`$ and $`\phi _2`$ contain only the transversal parts, $`D_\mu =_\mu ieA_\mu `$, e is a dimensionless coupling constant and $`A_\mu `$ is the electromagnetic field. The Lorentz invariant free term for $`A_\mu `$ is omitted. Under Lorentz transformation
$`x^\mu x^\mu =a^{\mu \nu }x_\nu ,`$ $`D_\mu D_\mu ^{}=a_{\mu \nu }D^\nu ,`$ $`d^4xd^4x^{}=d^4x,`$ (97)
$`\phi _1`$ and $`\phi _2`$ transform linearly in the same way,
$`\phi _1(x)\phi _1^{}(x^{})=\mathrm{\Lambda }\phi _1(x),\overline{\phi }_2(x)\overline{\phi }_2^{}(x^{})=\overline{\phi }_2\beta ^0\mathrm{\Lambda }^{}\beta ^0,`$ (98)
and thus the Lagrangian transforms as
$`L_\phi L_\phi ^{}=\overline{\phi }_2(x)\beta ^0\mathrm{\Lambda }^{}\beta ^0[i\beta ^\mu a_{\mu \nu }D^\nu m]\mathrm{\Lambda }\phi _1(x).`$ (99)
Owing to the invariance of mass term $`m\overline{\phi }_2(x)\phi _1(x)`$, $`\beta ^0\mathrm{\Lambda }^{}\phi ^0=\mathrm{\Lambda }^1`$. Therefore,
$`T(x)L_\phi ^{}L_\phi =\overline{\phi }_2(x)i[\mathrm{\Lambda }^1\beta ^\mu a_{\mu \nu }D^\nu \beta ^\mu D_\mu ]\phi _1(x).`$ (100)
Obviously, the conclusion of Lorentz invariance of the action $`L_\phi d^x`$ can be drawn once
$`{\displaystyle T(x)d^4x}={\displaystyle L_\phi ^{}d^4x^{}}{\displaystyle L_\phi d^4x}=0.`$ (101)
Under the infinitesimal Lorentz transformation, we have
$`a_{\mu \nu }=g_{\mu \nu }+\epsilon _{\mu \nu },`$ $`\mathrm{\Lambda }=1{\displaystyle \frac{i}{2}}\epsilon ^{\mu \nu }s_{\mu \nu },`$ (102)
where $`g_{\mu \nu }`$ is the metric tensor, $`\epsilon _{\mu \nu }`$ is the infinitesimal antisymmetric tensor and $`s_{\mu \nu }`$ is the unknown coefficient. With the help of $`\epsilon _{\mu \nu }\beta ^\mu D^\nu =\frac{1}{2}\epsilon _{\rho \tau }(\beta _\rho D_\tau \beta _\tau D_\rho )`$, $`T(x)`$ becomes
$`T(x)={\displaystyle \frac{i}{2}}\epsilon ^{\rho \tau }\overline{\phi }_2(x)[(\beta _\rho D_\tau \beta _\tau D_\rho )i[\beta ^\sigma ,s_{\rho \tau }]D_\sigma ]\phi _1(x).`$ (103)
To prove Eq. (101), all possible cases of Eq. (106) are discussed as follows:
(1) As $`\rho =\tau `$, $`\epsilon ^{\rho \tau }=0`$ and hence $`T(x)=0`$;
(2) As $`\rho =l`$ and $`\tau =m`$, namely, in the spatial rotation case, we have $`s_{lm}=s_{ml}=ϵ_{lmn}\left(\begin{array}{cc}\tau _n0& \\ 0\tau _n& \end{array}\right)`$, where $`\tau _n`$ has been given by Eq. (95) and $`ϵ_{lmn}`$ is the full antisymmetric tensor($`ϵ_{123}=1`$), so $`T(x)=0`$. As we can see, $`S_{23}^2+S_{31}^2+S_{12}^2=2`$, the Dirac-like field thus has spin 1.
(3) As $`\rho =l`$ and $`\tau =0`$, that is, in the general Lorentz boost case, the situation is a little more complicated compared with that in previous cases. In the present case, $`s_{l0}=s_{0l}=\frac{i}{2}\beta ^0\beta ^l\frac{i}{2}\alpha ^l`$, and $`T(x)`$ becomes
$`\begin{array}{cc}T(x)\hfill & =\frac{i}{2}\epsilon ^{l0}\phi _2^{}(x)[D_l+\frac{1}{2}(\alpha ^m\alpha ^l+\alpha ^l\alpha ^m)D_m]\phi _1(x)\hfill \\ & =\frac{i}{4}\epsilon ^{l0}[(\stackrel{}{F^{}}^{}F_l)+(\stackrel{}{G^{}}^{}G_l)],\hfill \end{array}`$ (106)
where $`\stackrel{}{F}^{}=(F_1^{},F_2^{},F_3^{})`$(it is similar for $`\stackrel{}{G}^{}`$), $`\stackrel{}{F}`$ and $`\stackrel{}{G}`$, the definition of them are
$`\begin{array}{cc}\phi _1(x)=\left(\begin{array}{ccc}F(x)\hfill & & \\ iG(x)\hfill & & \end{array}\right),& \phi _2(x)=\left(\begin{array}{ccc}F^{}(x)\hfill & & \\ iG^{}(x)\hfill & & \end{array}\right),\end{array}`$ (112)
in which $`F=\left(\begin{array}{c}F_1\hfill \\ F_2\hfill \\ F_3\hfill \end{array}\right)`$, $`G`$, $`F^{}`$ and $`G^{}`$ are in a similar form. Since $`\phi _1(x)`$ and $`\phi _2(x)`$ contain only the transversal parts, they satisfy transversal conditions
$`\stackrel{}{D}\stackrel{}{F}=\stackrel{}{D}\stackrel{}{G}=\stackrel{}{D}\stackrel{}{F^{}}=\stackrel{}{D}\stackrel{}{G^{}}=0`$ (113)
or
$`\stackrel{}{D}^{}\stackrel{}{F}^{}=\stackrel{}{D}^{}\stackrel{}{G}^{}=\stackrel{}{D}^{}\stackrel{}{F^{}}^{}=\stackrel{}{D}^{}\stackrel{}{G^{}}^{}=0.`$ (114)
In fact, by using Eq. (35) and Eq. (43), we can obtain Eq. (113) and Eq. (114) with $`A_\mu =0`$.
It can be verified also that
$`(\alpha ^m\alpha ^l+\alpha ^l\alpha ^m)D_m=(2D_lM_lM_l^T),`$ (115)
where $`M_L^T`$ is the transpose of $`M_L`$ and
$`\begin{array}{ccc}M_1=I_{2\times 2}\left(\begin{array}{ccc}D_1\hfill & 0& \hfill 0\\ D_2\hfill & 0& \hfill 0\\ D_3\hfill & 0& \hfill 0\end{array}\right),& M_2=I_{2\times 2}\left(\begin{array}{ccc}0\hfill & D_1& \hfill 0\\ 0\hfill & D_2& \hfill 0\\ 0\hfill & D_3& \hfill 0\end{array}\right),& M_3=I_{2\times 2}\left(\begin{array}{ccc}0\hfill & 0& \hfill D_1\\ 0\hfill & 0& \hfill D_2\\ 0\hfill & 0& \hfill D_3\end{array}\right).\end{array}`$ (126)
Taking use of Eq. (113), (114) and (115), we obtain Eq. (106). As we know, the physical field vanishes as $`|\stackrel{}{x}|\mathrm{}`$, so
$`{\displaystyle T(x)d^4x}={\displaystyle \frac{i}{4}}\epsilon ^{l0}{\displaystyle [(\stackrel{}{F^{}}^{}F_L)+(\stackrel{}{G^{}}^{}G_l)]d^4x}=0.`$ (127)
After all the possible cases (1), (2) and (3) have been considered , the conclusion $`T(x)d^4x0`$ becomes obvious. That is to say, the action of transversal Dirac-like field is Lorentz invariant.
Since the action $`L_\phi d^4x`$ is Lorentz invariant while the corresponding Lagrangian $`L_\phi `$ is not, the Lorentz invariance has a special implication in our theory. We will discuss it further in the next section.
## 7 Feynman rules and Polarization cross section for $`e^+e^{}f^+f^{}`$
Under the gauge transformation $`\phi _1(x)e^{i\theta (x)}\phi _1(x)`$(where $`\theta (x)`$ is a real parameter), the field quantity $`\phi _2(x)`$ transforms in the same way for the reason that the positive and the negative frequency parts of $`\phi _1(x)`$ or $`\phi _2(x)`$ are linearly independent. Obviously, the Lagrangian given by Eq. (61) is invariant under the global gauge transformation and the corresponding conserved charge is
$`Q={\displaystyle \phi _2^{}(x)\phi _1(x)d^3x}={\displaystyle \underset{\stackrel{}{p},s}{}}[a^{}(\stackrel{}{p},s)a(\stackrel{}{p},s)b^{}(\stackrel{}{p},s)b(\stackrel{}{p},s)],`$ (128)
just as what we expect. Minimal electromagnetic coupling is easily introduced into Eq. (61) by making the replacement $`_\mu D_\mu =_\mu ieA_\mu `$ and Eq. (61) turns into Eq. (96). One can verify that Eq. (96) is invariant also under local gauge transformation
$`\{\begin{array}{cc}& \phi _1(x)e^{i\theta (x)}\phi _1(x)(hence\phi _2(x)e^{i\theta (x)}\phi _2(x))\hfill \\ & A_\mu (x)A_\mu (x)\frac{1}{e}_\mu \theta (x).\hfill \end{array}`$ (131)
Now, we develop the relevant perturbative theory(called ”vector QED” or ”VQED”) in adiabatic approximation, where the S-matrix can be calculated from Dyson’s formula
$`S={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}S^{(n)}={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(i)^n}{n!}}{\displaystyle d^4x_1\mathrm{}d^4x_nT\{H_I(x_1)\mathrm{}H_I(x_n)\}},`$ (132)
where $`H_I(x)`$ is the interaction Hamiltonian written in interaction picture. After a tedious calculation, we give the Feynman rules for VQED in momentum space as follows(for simplicity, the Dirac-like particle or antiparticle is denoted by DL or anti-DL, respectively):
1. Propagators:
Photon’s$`=TA_\mu (x_1)A_\nu (x_2)=\frac{ig_{\mu \nu }}{q^2+i\epsilon }`$,
DL’s$`=T\phi _1(x_1)\overline{\phi }_2(x_2)=\frac{iA_{}}{\beta pm+i\epsilon }`$.
2. External lines:
Photon annihilation$`=ϵ_\mu (p)`$, Photon creation$`=ϵ_\mu ^{}(p)`$.
DL annihilation$`=\sqrt{\frac{m}{VE}}\chi (p,s)`$, DL creation$`=\sqrt{\frac{m}{VE}}\overline{\chi }(p,s)`$.
anti-DL annihilation$`=\sqrt{\frac{m}{VE}}\overline{y}(p,s)`$, anti-DL creation$`=\sqrt{\frac{m}{VE}}y(p,s)`$.
3. Vertex(DL with photon): $`ie\beta _\mu `$.
4. Impose momentum conservation at each vertex.
5. Integrate over each free loop momentum: $`\frac{d^4p}{(2\pi )^4}`$.
6. Divide by symmetry factor.
As an application, we will calculate the polarization cross section for process $`e^{}(p,s)+e^+(q,t)f^{}(p^{},s^{})+f^+(q^{},t^{})`$: the annihilation of an electron with a positron to create a pair of Dirac-like particles $`f^+`$ and $`f^{}`$. For simplicity, we work in the center-of-mass(CM) frame and compute the relevant transition amplitude in the lowest order.
In our case, the initial state is $`|e=c^{}(p,s)d^{}(q,t)|0`$ and the final state is $`|f=a^{}(p^{},s^{})b^{}(q^{},t^{})|0`$, where $`c^{}`$, $`d^{}`$ are the creation operators of electron and positron and $`s`$, $`t(=1,2)`$ are their spin indices. $`s^{}`$, $`t^{}(=\pm 1)`$ are the transversal polarization indices while $`p`$, $`q`$, $`p^{}`$ and $`q^{}`$ are the 4-momentum. The interaction Hamiltonian related to the process is
$`H_I(x)=e\overline{\psi }(x)\gamma ^\mu \psi (x)A_\mu (x)+e\overline{\phi }_2(x)\beta ^\mu \phi _1(x)A_\mu (x),`$ (133)
where $`\gamma ^\mu `$ is Dirac matrix and $`\overline{\psi }(x)=\psi ^{}(x)\gamma ^0`$, in which
$`\psi (x)={\displaystyle \underset{\stackrel{}{p},s}{}}\sqrt{{\displaystyle \frac{m_0}{VE_0}}}[c(p,s)u(p,s)e^{ipx}+d^{}(p,s)\nu (p,s)e^{ipx}]`$ (134)
is the electronic field, $`m_0,E_0`$ are the mass and the energy of electron, respectively. Then the corresponding transition amplitude is
$`\begin{array}{ccc}S_{fe}^{(2)}\hfill & =\hfill & f|S^{(2)}|e\hfill \\ & =\hfill & (2\pi )^4\delta ^4(p+qp^{}q^{})\sqrt{\frac{m}{vE}}\overline{\chi }(p^{},s^{})(ie\beta _\mu )\sqrt{\frac{m}{vE}}y(q^{},t^{})\hfill \\ & & \frac{ig^{\mu \nu }}{(p+q)^2}\sqrt{\frac{m_0}{vE_0}}\overline{\nu }(q,t)(ie\gamma _\nu )\sqrt{\frac{m_0}{vE_0}}u(p,s),\hfill \end{array}`$ (138)
where $`S^{(2)}`$ is defined by Eq. (132).
However, since $`e^+`$ and $`e^{}`$ have spin 1/2 while $`f^+`$ and $`f^{}`$ have spin 1, conservation of angular momentum requires that both the total spin of $`e^+e^{}`$ and that of $`f^+f^{}`$ are zero, which implies that both the initial state current(denoted by $`J_0^\mu `$) and final state current(denoted by $`J^\mu `$) must be spin singlets, so the relevant cross section is a polarization one(denoted by $`\sigma _{polar}`$). In the CM frame,
$`J_0^\mu ={\displaystyle \frac{1}{\sqrt{2}}}[\overline{\nu }(q,1)\gamma ^\mu u(p,1)\overline{\nu }(q,2)\gamma ^\mu u(p,2)],`$ (139)
$`J^\mu ={\displaystyle \frac{1}{\sqrt{2}}}[\overline{\chi }(p^{},1)\beta ^\mu y(q^{},1)\overline{\nu }(p^{},1)\beta ^\mu y(p,1)].`$ (140)
It is not difficult to infer that
$`\sigma _{polar}=`$ $`{\displaystyle (2\pi )^4\delta ^4(p+qp^{}q^{})\frac{m_0^2m^2}{2|\stackrel{}{p}|E_0E^2}\frac{e^4}{(p+q)^4}\frac{d^3p^{}}{(2\pi )^3}\frac{d^3q^{}}{(2\pi )^3}M^2},`$ (141)
where
$`M^2=|J_\mu J_0^\mu |^2.`$ (142)
Obviously, $`\frac{M}{(p+q)^2}`$ corresponds to the invariant transition amplitude. As a result, $`M^2`$ must be Lorentz invariant. After a cumbersome calculation, we obtain
$`M^2=4\mathrm{cos}^2\theta ,`$ (143)
where $`\theta `$ stands for the angle between the incoming electrons and the outgoing Dirac-like particles. Lorentz invariance of $`M^2`$ requires that $`\theta =0`$ or $`\pi `$(as we know, only in these cases, the $`\mathrm{cos}^2\theta `$ is Lorentz invariant). Namely, the final state momentum is purely longitudinal. Therefore, the Lorentz invariance has an incidental meaning for our theory(i.e. VQED): it provides an additional constraint for a possible physical process, just as the conservation of quantum number provides a selection rule for a possible particle reaction. In other words, VQED contains not only the Lorentz invariant process(and hence this process is allowed), but also the process violating the Lorentz invariance and hence being forbidden. In a sense, this paper provides a new approach to construct certain quantum field theories still unknown to us.
In a word, $`M^2=4`$. Seeing that $`E_0=E`$ in the CM frame and $`|\stackrel{}{p}|E_0`$ in the high energy approximation, we obtain
$`\sigma _{polar}={\displaystyle \frac{\pi \alpha ^2m_0^2m^2}{E_0^6}}v^3,`$ (144)
where $`\alpha =\frac{e^2}{4\pi }`$ and $`v^{}`$ is the velocity of $`f^+`$(or $`f^{}`$). It is found that $`v^{}=\frac{\sqrt{3}}{3}`$ maximizes $`\sigma _{polar}`$, that is
$`(\sigma _{polar})_{max}={\displaystyle \frac{2\sqrt{3}\pi \alpha ^2}{243}}{\displaystyle \frac{m_0^2}{m^4}}.`$ (145)
While in the high energy limit, $`v^{}1`$, and
$`\sigma _{polar}={\displaystyle \frac{\pi \alpha ^2m_0^2m^2}{E_0^6}}.`$ (146)
To sum up, the following conclusions follows:
(1), In the high energy approximation, $`m,m_eE_0=E`$, $`\sigma _{polar}0`$.
(2), The momentum of $`f^+`$ and $`f^{}`$ is purely longitudinal.
In the light of (1) and (2), we can explain the fact that none of the Dirac-like particles have been found so far.
## 8 Renormalizability of VQED
After the introduction of VQED in Sec. VII, the question whether VQED is renormalizable then arises. Before proceeding discussions, let us pause a moment to mention the following fact: Let the Dirac-like field quantity $`\phi =\left(\begin{array}{c}F\hfill \\ iG\hfill \end{array}\right)`$ and set $`m=0`$, where $`F=\left(\begin{array}{c}F_1\hfill \\ F_2\hfill \\ F_3\hfill \end{array}\right)`$ and $`G=\left(\begin{array}{c}G_1\hfill \\ G_2\hfill \\ G_3\hfill \end{array}\right)`$. In term of F and G, Eq. (1) or Eq. (23) can be reexpressed as Maxwell equations written in free vacuum, where F and G correspond to the electric field intensity $`E^{}`$ and the magnetic field intensity $`H^{}`$, respectively. However, we can’t regard F and G directly as E’ and H’ correspondingly, for two reasons:
(1), Under an infinitesimal Lorentz boost parametrized by the infinitesimal velocity v, the transformation properties of F(or G) are different from those of E’(or H’). The transformation of F(or G) contains the factor of $`\frac{v}{2}`$ while the transformation of E’(or H’) contains the factor v.
(2), The dimension of F(or G) is $`\frac{3}{2}`$ while that of E’(or H’) is 2. In fact, the canonical dimension of the Dirac-like vector is different from that of electromagnetic field.
In a word, the Dirac-like field is a new kind of vector field.
Now, let us pay attention to the renormalizability mentioned above. Fortunately, VQED can be developed in term of the same formalism as those in the electron-photon interaction theory, namely, spinor QED(denoted by SQED). In particular, the free propagator and the vertex in VQED take the same forms as those in SQED except for involving matrices $`\beta ^\mu `$(or $`\beta ^\mu A_{}`$) instead of the Dirac matrices $`\gamma ^\mu `$. As a result, starting from the Feynman integral(i. e. the loop momentum integral) of SQED, we can obtain the corresponding Feynman integral of VQED by replacing the $`\gamma ^\mu `$ with $`\beta ^\mu `$(or $`\beta ^\mu A_{}`$). Owing to these facts, the discussions of the renormalization of VQED is just a step by step business.
Let us consider a n-vertex one-particle-irreducible(1PI) diagram in VQED, in which there are N external Dirac-like particle lines and N’ external photon lines. Taking use of the similarity between VQED and SQED, we obtain the superficial degree of divergence of the 1PI diagram, say D, as follow:
$`D=4N^{}{\displaystyle \frac{3}{2}}N,`$ (147)
from which we draw the following conclusions:
1, D is independent of the number n of the vertices(i.e. the order n of perturbation).
2, D depends only on the number of external lines of the 1PI diagram, and there are only a finite number of external lines with $`D0`$.
From (1) and (2), we come to the conclusion that VQED is renormalizable. |
warning/0001/hep-th0001075.html | ar5iv | text | # The ℝ𝑃² Valued Sigma and Baby Skyrme Models
## 1 Introduction
The Skyrme model has long been of interest as an effective field theory of nucleons . The 2 dimensional baby Skyrme model allows us to study a more tractable analogue of the Skyrme model . The related O(3) sigma model is applicable to some condensed matter systems. Nematic liquid crystals are a condensed matter system where the target space is not $`S^2`$ but $`P^2`$. In this paper we describe our work on the variants of the baby Skyrme model and O(3) sigma model using $`P^2`$ as a target space.
### 1.1 Models
The O(3) sigma model describes maps from $`^2`$ (the “physical” space) to $`S^2`$ (the “target” space). $`^2`$ is the normal Euclidean plane, whereas $`S^2`$ is the 2 dimensional surface of a unit sphere embedded in 3 Euclidean dimensions. The lagragian of the O(3) sigma model is given by
$$=\frac{1}{4}^\mu \stackrel{}{\varphi }_\mu \stackrel{}{\varphi },$$
(1)
where $`\stackrel{}{\varphi }^3`$ such that $`\stackrel{}{\varphi }\stackrel{}{\varphi }=1`$. The condition that $`\stackrel{}{\varphi }\stackrel{}{\varphi }_{vac}`$ as $`r\mathrm{}`$, where $`r`$ is spatial radius, is usually imposed to keep the action finite. This effectively compactifies the physical space from $`^2`$ to $`S^2`$. This in turn implies that the configurations $`\stackrel{}{\varphi }`$ must fall into disjoint classes characterised by the elements of the second homotopy group $`\pi _2(S^2)`$. Field configurations which fall into any of the classes corresponding to non-trivial elements of $`\pi _2(S^2)`$ describe ‘lumps’ of energy. These lumps are not stable, and as a result are not true solitons. This instability is due to the conformal invariance of the model – the lumps have no intrinsic scale, and so may change scale without any energetic penalty. This scale invariance may be removed by the addition of new terms to the Lagrangian, as seen below.
The first model we examined was the baby Skyrme model with the so called ‘new’ potential term. The Lagrangian is
$$=\frac{1}{4}^\mu \stackrel{}{\varphi }_\mu \stackrel{}{\varphi }+\theta _1\frac{1}{4}((^\mu \stackrel{}{\varphi }^\nu \stackrel{}{\varphi })(_\mu \stackrel{}{\varphi }_\nu \stackrel{}{\varphi })(^\mu \stackrel{}{\varphi }_\mu \stackrel{}{\varphi })^2)+\theta _2_V,$$
(2)
where $`\theta _1`$ and $`\theta _2`$ are real positive constants. The term containing $`\theta _1`$ is the so called (2 dimensional) Skyrme term. This term tends to cause solitons to broaden, and is the only possible Lorentz covariant term of fourth order with only first order time derivatives. The term containing $`\theta _2`$ is normally refered to as the potential term. This term is chosen so that it acts to shrink the size of the soliton. Consequently a stable equilibrium may be reached where the effects of the Skyrme term and potential term balance. A number of different potential terms have been studied , but our $`P^2`$ target space forces us to use a potential term which is invariant under the transformation $`\stackrel{}{\varphi }\stackrel{}{\varphi }`$. A good choice would be the so called ‘new’ potential
$$_V=\frac{1}{4}(1(\stackrel{}{\varphi }\stackrel{}{\varphi }_{vac})^2).$$
(3)
### 1.2 The Geometry and Topology of $`P^2`$
#### 1.2.1 Geometry
$`P^2`$ may be thought of as a hemisphere, and so may be parameterised by a unit 3-vector $`\stackrel{}{\varphi }`$ where $`\stackrel{}{\varphi }(x)\stackrel{}{\varphi }(x)`$. $`P^2`$ may also be described locally by a 2-vector $`\stackrel{}{\chi }`$ where
$$\chi _j^i=\frac{\varphi _j}{\varphi _i}i=\mathrm{1..3}j=\mathrm{1..3}.$$
Here $`i`$ labels which chart is being used, and $`j`$ labels the components of $`\stackrel{}{\chi }`$. $`\stackrel{}{\chi }`$ appears to have 3 components in this definition but, as the component $`\chi _i^i`$ is always 1, $`\stackrel{}{\chi }`$ has only 2 degrees of freedom. The chart labeled $`i`$ is ill defined on the equator $`\varphi _i=0`$, and so all three charts are required to cover $`P^2`$ – the first covers all of the manifold except for an equator, the second all of this equator except for 2 antipodal points and the third those two points.
A neat way to combine all three $`\stackrel{}{\chi }`$ charts is to use the matrix $`P_{ij}(x)=\varphi _i(x)\varphi _j(x)`$. Each chart is easily extracted as $`\chi _j^i=\frac{P_{ij}}{P_{ii}}`$ with no sum implied by repeated indices.
#### 1.2.2 Topology
The $`n`$th homotopy group of $`P^2`$ can be found quickly for $`n2`$ as $`S^2`$ is a universal covering space of $`P^2`$. A theorem found in states that this implies that $`\pi _n(S^2)\pi _n(P^2)n2`$. $`\pi _1(P^2)`$ may be found to be the second symmetric group, $`_2`$. To summarise:
$`\pi _1(S^2)e\pi _1(P^2)_2,`$ (4)
$`\pi _2(S^2)\pi _2(P^2),`$ (5)
$`\pi _3(S^2)\pi _3(P^2),`$ (6)
#### 1.2.3 Winding Number
The class of field configuration $`\stackrel{}{\varphi }`$ may be characterised by the degree of map, or winding number. To calculate the winding number we must calculate the pullback of the two form on the target space; the map $`\stackrel{}{\varphi }`$ from the physical space to the target space induces a natural map (the “pullback”) from the two form on the target space to a two form on the physical space. If the two form on the physical space is then integrated over all space the result must be a multiple of an integer. By renormalising we get a formula for the integer valued winding number:
$$T=\frac{1}{8\pi }d^2xϵ_{ij}ϵ_{abc}\varphi _a(_i\varphi _b)(_j\varphi _c).$$
(7)
It should be noted that the transformation $`\stackrel{}{\varphi }\stackrel{}{\varphi }`$ leads to $`TT`$. This is because $`P^2`$ is a non-orientable manifold – a two form is not well defined on $`P^2`$ as its sense is changed by a translation along a non-trivial loop. In fact it may be shown that field configurations of $`\stackrel{}{\varphi }`$ are characterised not by $`\pi _2()`$ but by $`\pi _2()/\pi _1()`$. In the case of an $`P^2`$ target space this means that field configurations are characterised by $`/_2`$.
### 1.3 Numerical Techniques
The models outlined above are highly nonlinear and more or less completely intractable analytically. The models are, however, open to study using numerical techniques. One may write a simulation on a discrete grid and analyse the evolution of an arbitrary initial state. With a clever ansatz for the initial conditions, with or without the use of a relaxation routine, the behaviour of minimal enery field configurations may be studied.
We simulated the sigma and Skyrme models numerically, using $`P^2`$ as the target space. Our simulations were based on a 200 $`\times `$ 200 point grid using the nine point laplacian and derivatives outlined in , together with a fourth order Runge-Kutta algorithm for simulating time evolution. The timestep length was set to half the gridpoint spacing. To give the $`P^2`$ topology the state vector was stored using the $`P`$ matrix described in section (1.2.1) for position on the target space, and $`Q_{ij}=\dot{P}_{ij}=\varphi _i\dot{\varphi }_j+\dot{\varphi }_i\varphi _j`$ for the rate of change of position on the target space with time, where a dot is used to denote derivatives with respect to time. The simulation then evaluated derivatives by determining the field as the equivalent $`\stackrel{}{\varphi }`$ for any given set of nine points, with the central point mapped to the north pole of $`S^2`$. Reflections from the boundaries of the grid were reduced by damping out kinetic energy in a region near the boundary. The fields were kept on manifold by applying the following transformations each timestep:
$`P_{ij}^{}`$ $`={\displaystyle \frac{P_{ij}}{P_{kk}}},`$ (8)
$`Q_{ij}^{}`$ $`={\displaystyle \frac{Q_{ij}P_{ij}^{}Q_{kk}}{\sqrt{P_{kk}}}}.`$ (9)
These transformations ensure that $`Tr(P)=1`$ (i.e. the field lies on $`P^2`$) and that $`Tr(Q)=0`$ (i.e. the rate of change of the field is tangential to the surface of the field manifold).
## 2 The $`P^2`$ Valued New Baby Skyrme Model
Simulations of this model were in close qualitative and quantitative agreement with simulations using an $`S^2`$ target space.
### 2.1 The Hedgehog Anzatz
As in the $`S^2`$ model the simplest ansatz for a skyrmion is the radially symmetric ansatz, usually refered to as the ‘hedgehog’ ansatz . For a skyrmion of topological charge one this ansatz gives:
$$\stackrel{}{\varphi }=\left(\begin{array}{c}\mathrm{sin}[f(r)]\mathrm{cos}(\theta \gamma )\\ \mathrm{sin}[f(r)]\mathrm{sin}(\theta \gamma )\\ \mathrm{cos}[f(r)]\end{array}\right).$$
(10)
Here $`f(r)`$ is known as the profile function and $`\gamma `$ is an arbitrary phase parameter. This profile function is arbitrary up to the boundary condition that $`f(0)=(2n+1)\pi `$ and $`f(\mathrm{})=2m\pi `$ where $`n,m`$. To minimise the energy of the ansatz one may determine the energy of the field as a functional of the profile function and its first derivative, as shown below:
$$E=2\pi _0^{\mathrm{}}r𝑑r\left(f^{}_{}{}^{}2+\frac{\mathrm{sin}^2f}{r^2}(1+2\theta _1f^{}_{}{}^{}2)+\theta _2V(f)\right).$$
(11)
One may then use the calculus of variations to extremise this energy with the appropriate limits at $`r=0`$ and $`r=\mathrm{}`$. The resulting Euler-Lagrange equation is a non linear second order ODE for the profile function. We solved this numerically, using the shooting method. This ansatz leads to a stable single skyrmion. Note that our ansatz is identical to that of the $`S^2`$ skyrmion, and so our result agrees with the result of Weidig . Simulations showed this ansatz to be stable, again in agreement with Weidig. Note that the energy has been defined such that the energy of a hedgehog ansatz soliton with $`\theta _1=\theta _2=0`$ is given by $`E=1`$.
### 2.2 Two Skyrmions
As for the $`S^2`$ model, a two skyrmion ansatz may be arrived at by taking the stereographic projection from the north pole of a one skyrmion:
$$W=\frac{\varphi _1+i\varphi _2}{1\varphi _3}.$$
(12)
This may then be combined with another skyrmion using the ansatz
$$\frac{1}{W_T}=\frac{1}{W_1}+\frac{1}{W_2},$$
(13)
so that the final field $`W_T`$ is approximately equal to the field for any skyrmion near that skyrmion (where $`W=0`$) and the final field takes the vacuum value of $`W=\mathrm{}`$ at spatial infinity. Using this ansatz we were able to reproduce the attractive and repulsive channels found in , as well as the ninety degree scattering found in the attractive channel.
### 2.3 Equivalence of $`S^2`$ and $`P^2`$ models for Smooth Maps
In fact the two models are exactly equivalent as may be seen from the following argument. Consider maps $`f`$, $`\varphi `$:
$$f:I^nS^m,$$
(14)
where $`n<m`$, and
$$\varphi :S^mM.$$
(15)
As $`\pi _n(S^m)en<mf`$ may always be smoothly deformed to map all of $`I^n`$ to a point. This implies that, for all smooth maps $`\varphi `$, all $`n`$-loops in $`S^m`$ shall be the pre-image of an $`n`$-loop in $`M`$ corresponding to the identity element of $`\pi _n(M)`$.
One consequence of this is that for $`MP^2`$, $`n=1`$ and $`m=2`$, all maps $`\varphi `$ may be described as maps to $`MS^2`$, as $`P^2`$ is universally covered by $`S^2`$, so the $`P^2`$ new baby Skyrme model is identical to the $`S^2`$ baby Skyrme model if $`\stackrel{}{\varphi }`$ is well defined at all points.
It is also worthy of note that this argument suggests that simulations of Fadeev-Hopf solitons using $`P^2`$ as a target space should reproduce the simulation of such solitons with an $`S^2`$ target space unless defects are present.
## 3 The Sigma Model with Defects
The implication of the argument put forward above is that to discover new behaviour in an $`P^2`$ model we must introduce a discontinuity into the field. For such a point to be stable a circle around the discontinuity in physical space must map to a non-trivial loop in the target space. If this is not the case then the discontinuity can be removed by a continuous variation in the field. These (point) discontinuities are analogous to disclination lines in a liquid crystal. We shall refer to them as “defects”. One consequence of having an odd number of defects in the system is that a contour around the edge of the physical space must map to a non-trivial loop on $`P^2`$. This means that it is no longer possible to have the field tend to a single value at spatial infinity. As a result, it is no longer easy to use a conventional potential term such as (3) in the Lagrangian. We may use the $`P^2`$ sigma model, so our model has neither Skyrme nor a potential terms, or we may use an unconventional potential term. One possible potential term would be of the form
$$_V=\frac{1}{4}(\stackrel{}{\varphi }\stackrel{}{\varphi }_{mass})^2),$$
(16)
which we shall call the “easy plane” potential. This effectively gives a mass to one of the fields ($`\varphi _{mass}`$) where the potential (3) gives mass to both fields orthogonal to $`\varphi _{vac}`$. Easy plane baby skyrmions have not previously been studied and so are worthy of attention in their own right. We have confined our work to the sigma model in all that follows.
It should be noted that if there is no single value for the field at infinity we are no longer compactifying the physical space to $`S^2`$. As a result winding number is no longer necessarily conserved. In fact, the physical space is topologically $`S^1\times ^1`$ as not only is the space not compactified at infinity, it also has no well defined map at the point of the defect. Whilst the mapping from the target space is frozen at infinity, it is not frozen around the defect. The circle around the defect necessarily maps to a non-trivial curve in $`P^2`$. If we represent $`P^2`$ as a sphere, where antipodal identification is allowed, a non-trivial curve is a line between a pair of (arbitrary) poles. If we have a defect-soliton system, one of these curves represents both the image of a circle at infinity in the physical space, and the image of an infinitesimal circle around the defect. As a circle is contracted in from infinity to the defect, the image curve rotates around the sphere, hinged on the polar identification, wrapping the sphere once. As the image of the infinitesimal circle around the defect is able to move, it may rotate around the identified poles, unwinding the soliton like object.
We decided to study the behaviour of a soliton like lump in the presence of a point defect. To do this we need an ansatz for a soliton-defect system to use as an initial conditon for our simulation. Such an ansatz needs to be sufficiently close to equilibrium to avoid large quantities of radiation perturbing the system – if the system is far from equilibrium much of the excess energy will be shed from the solitonic objects as radiation.
### 3.1 Defect Ansatz
The most obvious ansatz for a defect is a radially symmetric, planar field. A defect at $`r=0`$ may be written in polar coordinates as
$$\stackrel{}{\varphi }=\left(\begin{array}{c}\mathrm{sin}\frac{\theta }{2}\\ 0\\ \mathrm{cos}\frac{\theta }{2}\end{array}\right),$$
(17)
which, using the stereographic projection from the south pole,
$$W=\frac{\varphi _1+i\varphi _2}{1+\varphi _3},$$
(18)
is given by:
$$W=\mathrm{tan}\frac{\theta }{4}.$$
(19)
This ansatz is defined for $`0\theta <2\pi `$. On the line $`\theta =0`$ there appears, at a first glance, to be a discontinuity in the field, but as $`\stackrel{}{\varphi }`$ is identified with $`\stackrel{}{\varphi }`$ there is no such discontinuity. In fact any field configuration which contains a defect must have a line from the defect to infinity along which the identification between $`\stackrel{}{\varphi }`$ and $`\stackrel{}{\varphi }`$ is made. It is simple to show analytically that the above field configuration is in equilibrium. Simulations also show that this field configuration is stable. It should be noted that this field configuration has an infinite action arising from a logarithmic divergence as $`r`$ tends to infinity or as $`r`$ tends to zero. This is not a problem if the physical system we are modeling is a liquid crystal. This is because such a system is, firstly, finite in extent, and, secondly, a discrete system. The field approximation is clearly not appropriate at $`r=0`$. Care must be taken when simulating such an object numerically to place such an object away from a grid point.
### 3.2 ‘Glued’ Ansatze
#### 3.2.1 The Three Region Ansatz
One ansatz for a defect-soliton system which we tried early on involved ‘glueing’ together three regions in an attempt to graft together a known ansatz for the soliton to a defect like object – see figure (2).
In region I the field followed the standard lump ansatz for the O(3) sigma model. This is given by the hedgehog ansatz (10) with the profile function given by:
$$f(r)=\mathrm{tan}^1\frac{2\lambda r}{1\lambda ^2r^2},$$
(20)
where $`\lambda `$ is an arbitrary (real) parameter describing the (inverse) ‘width’ of the lump. In region II the field is constant and equal to the vacuum value of region I – there should be little discrepancy at the border between regions I and II as a result, and any variation may be interpolated across several grid points in the simulation. In the region III the field is given by:
$$\stackrel{}{\varphi }=\left(\begin{array}{c}\mathrm{cos}\theta _d\\ 0\\ \mathrm{sin}\theta _d\end{array}\right),$$
(21)
for $`\pi \theta _d<\pi `$. The region II and region III fields match along the line $`\theta =\frac{\pi }{2}`$ and, due to the identification $`\stackrel{}{\varphi }\stackrel{}{\varphi }`$, along the line $`\theta =\frac{\pi }{2}`$. This places a defect at the origin. Unfortunately this defect is far from stable – the energy of this defect is double that of the defect described by (17). Thus in the simulations of this total field configuration we have seen the emission of waves of radiation by the defect. These waves affected the skyrmion. To reduce these problems we have modified our ansatz as detailed below.
#### 3.2.2 The Two Region Ansatz
We modified the above ansatz by replacing the field in regions II and III by the field below:
$$W=\mathrm{tan}\frac{\theta _d}{4}e^{k(r_dr_0)},$$
(22)
where $`0\theta _d<2\pi `$ and W is the stereographic projection from the south pole. At a first glance this field looks as though it might tend to the vacuum value in all directions whilst looking like a defect at small radii. In fact, this is not the case – the field is smooth away from the origin, and therefore a loop around this field at infinity must be non-trivial. The field must vary dramatically as $`\theta `$ approaches $`2\pi `$ and $`r`$ approaches infinity. We had hoped that this would not be a problem in the (finite) region of the simulation, but discretisation brought its own problems, as shown in figure (3). The line where $`W=1`$ must cross the last line of grid points on the large $`\theta _d`$ side of the $`\theta _d=0`$ line.
This leads to a point where the variation in the field from one grid point to the next is as large as it can be, resulting in a region of high energy density. In fact this region also has defect type winding number – the ansatz has created a second defect in a far from equilibrium configuration. In fact if one constructs an ansatz which tends to constant field at spatial infinity (or the edge of the simulation region) then the space must contain an even number of defects. A loop at large r which maps to one point in $`P^2`$ must correspond to the trivial element of $`\pi _2(P^2)=_2`$, so the number of defects inside this loop must be even.
### 3.3 The Stereographic Defect - Soliton Ansatz
To create an ansatz for a field configuration including both a soliton and a defect we must re-express our field configurations using the stereographic map, $`W`$. A projection from the south pole causes the north pole, $`\varphi _3=1`$, to map to $`W=0`$ and the south pole to map to $`W=\mathrm{}`$. Note that the antipodal identification under this map is given by
$$W\frac{1}{W^{}}.$$
(23)
Our ansatz must a) look like a soliton near the soliton, b) look like a defect near the defect, c) look like a defect at infinity and d) be smooth at all points away from the defect.
A single soliton at the origin may be described by the field
$$\stackrel{}{\varphi }=\left(\begin{array}{c}\alpha \mathrm{cos}(\theta \gamma )\\ \alpha \mathrm{sin}(\theta \gamma )\\ \beta \end{array}\right),$$
(24)
where $`\alpha =\frac{2\lambda r}{1+\lambda ^2r^2}`$, $`\beta =\frac{1\lambda ^2r^2}{1+\lambda ^2r^2}`$, $`\gamma `$ is an arbitrary (real) phase parameter, $`\lambda `$ is an arbitrary (real) parameter characterising the width of the soliton, and $`(r,\theta )`$ are polar coordinates. Note that this is the hedgehog ansatz with the profile function given by (20). To combine this with a defect we choose the point of projection of both fields such that the soliton at infinity maps to a number of modulus $`1`$, and the line along which the defect field identifies approaches $`0`$ from one side and $`\mathrm{}`$ from the other. These fields are given below.
$$W_s=\frac{\alpha \mathrm{sin}(\theta \gamma )+i\beta }{1+\alpha \mathrm{cos}(\theta \gamma )},$$
(25)
$`W_d`$ $`={\displaystyle \frac{\mathrm{sin}(\frac{\theta }{2})}{1+\mathrm{cos}(\frac{\theta }{2})}}`$
$`=\mathrm{tan}({\displaystyle \frac{\theta }{4}}).`$ (26)
If we multiply these fields we find
$$W_T=\frac{\alpha \mathrm{sin}(\theta _s\gamma )+i\beta }{1+\alpha \mathrm{cos}(\theta _s\gamma )}\mathrm{tan}(\frac{\theta _d}{4}),$$
(27)
where $`\theta _d`$ is the polar angle coordinate with respect to the position of the defect and $`(r,\theta _s)`$ are the polar coordinates with respect to the position of the soliton. This ansatz obeys conditions c) and d) above. c) is satisfied as at infinity $`W_T=i\times W_d`$. This is acceptable as multiplying a field configuration by a number of unit modulus is merely a rotation about the axis of projection. d) is also satisfied as the line of identification in $`W_d`$ is preserved by the multiplication – $`0\times W_s=0`$ and (crudely speaking) $`\mathrm{}\times W_s=\mathrm{}`$. Condition a) is fulfilled if the soliton and defect are sufficiently separated for the variation in $`W_d`$ to be small across the scale of the soliton. Condition b), however, is only fulfulled if $`(\theta _s\gamma )=(2n+1)\frac{\pi }{2}`$ where $`n`$.
### 3.4 The Inhomogeneous Defect - Soliton Ansatz
An alternative ansatz involves using the inhomogeneous coordinates outlined in section (1.2). If we define a complex number
$$𝒲=\frac{\varphi _1+i\varphi _2}{\varphi _3}$$
(28)
then we can express the soliton field as a complex number which tends to $`0`$ at $`r`$ tends to $`0`$ and $`\mathrm{}`$. If we add this to the defect field then the result must satisfy conditions c) and d) outlined in section (3.3). c) is satisfied because the map used is a map of $`P^2`$, not $`S^2`$ and so needs no line of identification for the defect ansatz, whilst d) is obeyed as the soliton field vanishes at infinity, leaving only the defect field. Explicitly, the fields look like this:
$`𝒲_s={\displaystyle \frac{\alpha }{\beta }}e^{i(\theta _s\gamma )},`$ (29)
$`𝒲_d=\mathrm{tan}({\displaystyle \frac{\theta _d}{2}}),`$ (30)
$`𝒲_T={\displaystyle \frac{\alpha }{\beta }}e^{i(\theta _s\gamma )}+\mathrm{tan}({\displaystyle \frac{\theta _d}{2}}).`$ (31)
This expression obeys a) if $`𝒲_d`$ is small around the position of the soliton. Condition b) is only met if $`𝒲_s`$ is small in the region of the defect, and in the example above condition b) is less well satisfied if $`𝒲_s`$ has an imaginary component at the defect.
### 3.5 Simulation methods
For simulations of a defect-soliton system we introduced a conformal grid in a similar manner to Leese et al in . This involved changing the physical coordinates $`(x,y)`$ to $`(x^{},y^{})`$ such that
$$X^{}=\frac{X}{1+|X|}.$$
(32)
This allows a much larger area to be simulated. The grid was $`199\times 199`$ points with $`dX^{}=0.01`$. This conformal grid was introduced to reduce boundary effects. Again, the timestep length was set to half the gridpoint spacing.
Another alteration to the simulation for defect-soliton systems was the introduction of a period of relaxation before the simulation was allowed to run freely. $`Q_{ij}`$ was set to zero throughout the grid every $`10`$ timesteps for the first $`100`$ timesteps to take the simulation as close as possible to its minimal energy before the simulation proper began.
### 3.6 Results
The soliton-defect system displayed certain generic characteristics – a “spreading” channel and a “spiking” channel according to the phase of the soliton. In the spreading channel the soliton would become broader and broader until it significantly overlapped the defect, at which point it would unwind, and the energy would be released to infinity as radiation. In the spiking channel the soliton would become more and more localised until the variation in the field became numerically untenable. The defect would remain stationary in all simulations, whilst the soliton would move only a very small distance, to the extent that variations in the width of the soliton would be far more significant than any movement of the maxima of the soliton.
#### 3.6.1 Stereographic Ansatz
We used the stereographic ansatz (equation(27)) as our initial condition with $`\lambda =2.5`$, and the initial soliton position relative to the defect at $`\theta _d=\pi `$, $`r_d=2.5`$. We shall discuss the results of simulations with $`\gamma `$ set to $`\frac{\pi }{2}`$, $`\frac{\pi }{2}`$, $`0`$ and $`\frac{\pi }{4}`$. After the initial period of relaxation energy was conserved to better than $`0.05\%`$ in all of these simulations. The period of relaxation ended at $`t=0.5`$, after which the simulation ran freely until $`t=9.995`$, unless the simulation became untenable due to soliton spiking.
When the simulation was run with the phase initially set to $`\gamma =\frac{\pi }{2}`$ the soliton became narrower with time as described above (see figure (5)). This resulted in the simulation ending at $`t=3.745`$.
When $`\gamma =\frac{\pi }{2}`$ the soliton spread out and eventually unwound as described above (see figure (4)). When $`\gamma =0`$ was simulated the soliton was still subject to spreading and unwinding – the spreading channel is wider than the spiking channel.
For $`\gamma =\frac{\pi }{4}`$ the soliton began to spike, but after a period of time the the rate of spiking slowed and the soliton began to spread, with the soliton eventually unwinding (see figure(6)). This also suggests that the spiking channel is unstable.
Table (1) shows the energies of various simulations with initial conditions given by this ansatz. Note that these energies depend heavily on the lattice spacing at the defect, and are therefore only of value when comparing the different channels.
#### 3.6.2 Inhomogeneous Ansatz
We also simulated the soliton defect system using the inhomogeneous ansatz (equation (31)) as the initial condition, with the initial $`\lambda =2.5`$, and the initial soliton position relative to the defect at $`\theta _d=0`$, $`r_d=2.5`$, so that the contribution to the ansatz from the soliton is significant compared to that from the defect in the region of the soliton. With this ansatz the soliton spikes for $`\gamma =\frac{\pi }{2}`$ (see figure (8)) and spreads at $`\gamma =\frac{\pi }{2}`$ (see figure (7)). If $`\gamma =\frac{\pi }{4}`$ in the initial condition, the soliton starts to spike, but then later spreads (see figure(9)). Table (2) shows the energies for simulations with these initial conditions. Again, these energies are only included for their value in comparing channels and to show that energy is conserved during the free run of the simulation.
### 3.7 Collective Coordinate Approach
To further the understanding of our numerical results we carried out some approximate analytic work based on the so called collective coordinate approach. We substituted one of our ansatze into the Lagrangian density, integrated over the spatial variables to find a true Lagrangian and then considered one or more of the ansatz parameters as dynamic variables, so for example Lagrangian (1) together with
$$\stackrel{}{\varphi }=\stackrel{}{\varphi }(\lambda ,\dot{\lambda },x,y),$$
(33)
implies that
$$=(\lambda ,\dot{\lambda },x,y),$$
(34)
which may then be integrated over to give
$$L=(\lambda ,\dot{\lambda },x,y)𝑑x𝑑y=L(\lambda ,\dot{\lambda }),$$
(35)
allowing us to construct an equation of motion for $`\lambda `$.
This effectively constrains the solutions of the field theory to the submanifold defined by the ansatz – in the example above our solutions are constrained to a 2 dimensional submanifold of the infinite dimensional phase space of $`\stackrel{}{\varphi }(x,y)`$. This does not allow for soliton unwinding or the release of radiation with our ansatze, as the ansatze never have radiation and always have a soliton.
If we take the ansatz (27) together with the Lagrangian (1) expressed in terms of stereographic coordinates
$$=\frac{_\mu W^\mu W^{}}{4(1+|W|^2)^2},$$
(36)
we may construct the Lagrangian density in terms of position and a few parameters. If we take the initial soliton and defect position outlined in section (3.6.1) then $`W`$ becomes
$$W=\frac{2\lambda (r\mathrm{sin}(\theta \gamma )2.5\mathrm{sin}\gamma )+i(1\lambda ^2(r^2+5r\mathrm{cos}\theta +6.25))}{1+\lambda ^2(r^2+5r\mathrm{cos}\theta +6.25)+2\lambda (r\mathrm{cos}(\theta \gamma )+2.5\mathrm{cos}\gamma )}\mathrm{tan}\frac{\theta }{4},$$
(37)
where ($`r,\theta `$) are polar coordinates, $`\lambda `$ parameterises the (inverse) width of the soliton and $`\gamma `$ is the phase of the soliton. Note that the soliton has position ($`2.5`$, $`\pi `$) and the defect is at the origin in this coordinate system. Taking spatial derivatives of $`W`$ is then a straight forward if somewhat tedious process. Time derivatives may be found by treating one or more of the parameters as dynamic - if we consider $`\lambda `$ to be dynamic and the other parameters to be static the $`_t=\dot{\lambda }_\lambda `$. This substitution of an ansatz and use of parameters as dynamic variables is equivalent to assuming that the field configuration moves quasi-statically from one configuration to another with different values of the dynamic variables. This assumption is valid only if our ansatz is close to equilibrium and our dynamic parameter only varies slowly with time (i.e. in this case $`\dot{\lambda }`$ is small).
Using this approximation our Lagrangian density becomes
$$=\frac{r^2_rW_rW^{}+_\theta W_\theta W^{}r^2\dot{\lambda }^2_\lambda W_\lambda W^{}}{4r^2(1+|W|^2)^2}.$$
(38)
So our approximate Lagrangian with a time dependant $`\lambda `$ becomes
$$L=A(\lambda )\dot{\lambda }^2B(\lambda ),$$
(39)
where
$$A(\lambda )=\frac{r^2_rW_rW^{}+_\theta W_\theta W^{}}{4r^2(1+|W|^2)^2}r𝑑r𝑑\theta $$
(40)
and
$$B(\lambda )=\frac{_\lambda W_\lambda W^{}}{4(1+|W|^2)^2}r𝑑r𝑑\theta .$$
(41)
The Euler-Lagrange equation then gives us
$$\ddot{\lambda }=\frac{(A^{}(\lambda )+\dot{\lambda }^2B^{}(\lambda ))}{2B(\lambda )}.$$
(42)
We considered the example above, where all variables are static except for $`\lambda `$, but found an analytic integration to be intractable. We carried out the integration and time evolution numerically, using a fourth order Runge-Kutta algorithm for the time evolution. We also carried out a similar analysis using the inhomogeneous ansatz (31).
In figures (10, 11) we show the results of this treatment against those of the full simulation. The two treatments produce broadly similar results, although the rapidity of the broadening and spiking is faster in the full simulation. This is not entirely unreasonable considering that we have moved from around 160,000 degrees of freedom to 2! The curves from the full simulation were found by finding the maximum winding number density on the grid and then finding the value of lambda that would give a single soliton of the form of equation (24) this maximum winding number density. One consequense of this technique is that for broad solitons the maximum winding number for the soliton may be smaller than the maximum on the grid – this leads to the curves for the broadening channel becoming unreliable at around $`t=5`$. Naturally, the collective coordinate simulation began at the point when the relaxation ended in the full simulation.
## 4 Conclusions
We have examined the sigma model and baby Skyrme model with $`P^2`$ as a target space and found these models to be identical to their $`S^2`$ counterparts in the absence of defects. We examined the interaction between defects and soliton-like lumps in the sigma model and found the interaction to depend on the relative phase of the lump. We found a channel which causes the soliton to broaden and another which causes the soliton to spike. When the soliton overlapped the defect significantly, the soliton would unwind. We broadly reproduced this behaviour with the collective coordinate approach, using only two collective coordinates.
A range of possibilities for future work present themselves – studying the interaction between two defects, the interactions of two lumps in the presence of a defect and the behaviour of the $`P^2`$ sigma model on a torus all have the potential to exhibit new and interesting behaviour.
### Acknowledgements
I would like to thank Wojtek Zakrzewski for his advice and guidance. I would also like to thank PPARC for funding this research. |
warning/0001/cond-mat0001036.html | ar5iv | text | # Roton Instability of the Spin Wave Excitation in the Fully Polarized Quatum Hall State and the Phase Diagram at 𝜈=2
## I Introduction
The interplay between the electron’s spin degree of freedom and the inter-electron interaction has been of interest in the condensed matter physics, in particular in the jellium model where the neutralizing background is taken to be rigid and uniform. The relative strength of the interaction is conventionally measured through the parameter $`r_S`$ which is the interparticle distance measured in units of the Bohr radius $`a_Bϵ\mathrm{}^2/me^2`$, $`ϵ`$ being the dielectric constant and $`m`$ the band mass of electron. For electrons confined to two dimensions, which will be our focus in this paper, we have
$$r_S\frac{(\pi \rho )^{1/2}}{a_B},$$
(1)
where $`\rho `$ is the two-dimensional density of electrons. The interaction strength is enhanced relative to the kinetic energy as the system becomes more dilute, i.e., as $`r_S`$ increases. It has been predicted that as $`r_S`$ is increased, the electron liquid eventually becomes spontaneously spin polarized to gain in the exchange energy, ultimately going into a Wigner Crystal at $`r_S37`$ .
The two-dimensional electron systems, obtained experimentally at the interface of two semiconductors, constitute an almost ideal realization of the jellium model for several reasons. Samples with mobility in access of 10 million cm<sup>2</sup>/Vs are available , minizing the effect of disorder. Furthermore, the density of electrons, which controls the strength of the interaction relative to the kinetic energy, can be varied by a factor of 20 . We will consider electrons in the presence of a magnetic field, specifically, at filling factor $`\nu =2`$, which is a particularly clean test case for the kind of physics in which we are interested. There are three relevant energy scales here: $`\mathrm{}\omega _C=\mathrm{}eB/mc`$ is the cyclotron energy, $`V_C=e^2/ϵl_0`$ is the typical Coulomb energy, $`l_0=\sqrt{\mathrm{}c/eB}`$ being the magnetic length, and $`E_Z`$ is the Zeeman splitting energy which is the Zeeman energy cost necessary for a single spin flip. (To an extent, the Zeeman and the cyclotron energies can be varied independently by application of the magnetic field at an angle; while the former depends on the total magnetic field, the latter is determined by the normal component only.) The ground state is known in two limits. When $`\mathrm{}\omega _C`$ dominates, the ground state is a spin singlet, with 0$``$ and 0$``$ Landau levels fully occupied. On the other hand, when $`E_Z`$ is the largest energy, the ground state is fully polarized; when the Coulomb interaction is not strong enough to cause substantial Landau level mixing (i.e., in the limit of $`E_Z>>\mathrm{}\omega _C>>V_C`$), the ground state has 0$``$ and 1$``$ Landau levels occupied. When the ground state is described in terms of filled Landau levels, we will denote it by $`(N`$$`:N`$$`)`$, where $`N`$$``$ and $`N`$$``$ are the numbers of occupied Landau levels for up and down spin electrons. The possible filled Landau level states at $`\nu =2`$ are then $`(1:1)`$ and $`(2:0)`$, the unpolarized and the fully polarized states, respectively. Our interest will be in situations when $`V_C`$ becomes comparable to or greater than the cyclotron energy. This again corresponds to $`r_S1`$, where at $`\nu =2`$, $`r_S`$ can be seen to be given by
$$r_S=\frac{V_C}{\mathrm{}\omega _C},$$
(2)
and is clearly a measure of the strength of the interaction relative to the cyclotron energy. Here, Landau level mixing becomes crucial and may destabilize the above states. Our principal goal will be to determine the phase diagram of the fully polarized and the unpolarized states $`(2:0)`$ and $`(1:1)`$.
Our work has been motivated by the recent experiments of Eriksson et al. where they investigate by inelastic light scattering both the spin and the charge density collective modes at $`\nu =2`$ for samples with $`r_S`$ as large as 6, corresponding to densities as low as $`\rho =0.9\times 10^{10}`$ cm<sup>-2</sup>. They find a qualitative change in the number and the character of collective modes at approximately $`r_S2`$. This was interpreted in a Landau Fermi liquid approach, where the magnetic field was treated as a perturbation on the zero field Fermi liquid. However, the Fermi liquid approach, which is suitable at small magnetic field, is not an obviously valid starting point for the problem at hand, and other approaches are desirable. A comparison between the ground state energies of the unpolarized and the fully polarized Hartree Fock state shows that a transition between them takes place at $`r_S2.1`$ for $`E_Z=0`$, as we will see in Sec.V. This raises the question: Is the ground state at large $`r_S`$ fully polarized? If true, this would indeed be an interesting example of an interaction driven ferromagnetism. If it is indeed fully polarized, is the $`(2:0)`$ state a reasonable starting point for its study? Besides being fully polarized, $`(2:0)`$ incorporates the effect of Landau level mixing, albeit in a very special manner, through promoting all electrons in 0$``$ to 1$``$.
A reliable treatment of Landau level mixing lies at the crux of the problem. We shall incorporate Landau level mixing in a perturbative time-dependent Hartree-Fock scheme, i.e., by incorporating vertex corrections through ladder diagrams in the random phase approximation (RPA). The most crucial approximation in our calculations will be a restriction to the subspace of a single-particle hole pair; within this subspace, however, the Landau level mixing is treated accurately. Clearly, this approach is not quantitatively valid except at small $`r_S`$, but we believe that it gives an insight into the physics even when $`r_S`$ is not small. For a better quantitative description at large $`r_S`$, it would be important to deal with screening by more than one particle-hole pair, but that is outside the scope of the present paper.
As mentioned above, there certainly are parameters for which the fully polarized $`(2:0)`$ state describes the actual ground state; in particular, it may occur even when $`r_S`$ is large provided that the Zeeman energy is sufficiently strong. Our approach here will be to take it as the starting point and investigate the regime of its stability by calculating the dispersion of the charge and the spin density collective modes. The collective modes have a simple interpretation when the Landau level mixing is negligible ($`V_C/\mathrm{}\omega _C0`$). The lowest charge density excitation mode corresponds to the excitation of one electron from 1$``$ to 2$``$ Landau level. The kinetic energy change of any collective mode is well defined when the Landau level mixing is weak, and will be used to label the various modes (this notation will be used even when the Landau level mixing is significant, by looking at the evolution of the modes from small to large $`r_S`$). The 1$``$$``$2$``$ mode will be referred to as the $`m=1`$ mode. For spin-density excitation, there are three modes of primary interest, corresponding to excitations 0$``$$``$0$``$, 1$``$$``$1$``$, and 1$``$$``$0$``$ which are depicted schematically in (a), (b), and (c) in Fig.1 respectively. The first two are $`m=0`$ modes and the last one is $`m=1`$ mode. These are of course coupled, and a diagonalization of the problem produces the usual spin-wave excitation mode, with the energy approaching the Zeeman splitting in the long wave length limit, in accordance with the Goldstone theorem, as well as two massive spin-density modes. Our principal result is that while the charge density collective mode shows no instability, the spin-density collective mode develops a deep roton minimum in the presence of substantial Landau level mixing and becomes soft in certain parameter regimes. (Indeed, as $`r_S`$ is increased, there will eventually also be an instability in the charge density channel, indicating the formation of a Wigner crystal, but we have not explored this question since our perturbation theory is not valid at very large $`r_S`$.) Both the existence of the roton minimum and its softening as the Zeeman energy is reduced are experimentally testable predictions of our theory. The phase diagram thus obtained later is shown in Fig.2. It is instructive to compare it with the phase diagram in the absence of interactions, which consists of the fully polarized state $`(2:0)`$ for $`E_Z>\mathrm{}\omega _C`$ and the unpolarized state $`(1:1)`$ for $`E_Z<\mathrm{}\omega _C`$ with a transition taking place at precisely $`E_Z=\mathrm{}\omega _C`$. At small $`r_S`$, interactions make the $`(2:0)`$ state more stable, indicated by the fact that it survives even for $`E_Z<\mathrm{}\omega _C`$. However, at large $`r_S`$, $`E_Z>\mathrm{}\omega _C`$ is required to stabilize the fully polarized state. It is noteworthy that at small $`E_Z`$, the fully polarized state is found to be unstable at arbitrary $`r_S`$. We also consider the collective modes of $`(1:1)`$, expected to be valid at small $`E_Z`$ and $`r_S`$. It becomes unstable as $`r_S`$ is increased, consistent with an earlier conclusion of MacDonald .
What about the state at large $`r_S`$ but small Zeeman energy? As discovered in our study, here the $`\nu =2`$ state is not described by either of the two aforementioned Hartree Fock states. The finite wave vector spin-wave instability of the fully polarized state suggests that it is some kind of spin density wave state. It is not possible to be more definitive about it based on our present study. Of course, at extremely large $`r_S`$, when $`V_C`$ is much larger than the cyclotron energy, our calculation is unreliable, but indicates that even if a fully polarized state occurs, as expected based on the zero field result, it will most likely not be described by the $`(2:0)`$ Hartree Fock state.
There have been many theoretical studies of the collective excitations of integer quantum Hall effect (IQHE) states in the past, but, to our knowledge, the spin-density wave excitations of the fully polarized $`(2:0)`$ state have not been considered previously. For other collective modes, our results reduce to the earlier results in appropriate limits. If possible transitions of the electron and hole between different Landau levels are ignored and if self-energy corrections are omitted, the problem of collective excitation is reduced to determining the binding energy of two oppositely charged particles strictly confined in their respective Landau levels. In this case the wave function for the bound state is independent of interaction potential, and is uniquely determined by the wave vector $`\stackrel{}{k}`$. The transition of the electron and hole or the recombination of the particle-hole pair has been considered in the random phase approximation(RPA). Later the RPA was incorporated with the self-energy correction and the binding energy term by Kallin and Halperin where a number of interesting collective excitations from the unpolarized and the partially polarized ground state were considered in the absence of Landau level mixing (valid when $`\mathrm{}\omega _C>>V_C`$). The Landau level mixing was considered by MacDonald in Ref , treating the mixing matrix elements between various modes as small parameters and applying a second order perturbation theory. Our calculation will be formulated in terms of diagrams, following Kallin and Halperin in Ref., and will be performed with a full treatment of the mixing matrix elements within the subspace of a single particle-hole pair at any given instant. The diagrammatic formulation of the problem is presented in detail in Sec.II below.
In Sec.III we describe the diagrammatic formalism used to compute dispersion curves of the collective excitation from the fully polarized ground state at $`\nu =2`$. We will concentrate on the spin density excitation which will be responsible for an instability of the IQHE state in the parameter regime under consideration. Similarly, Sec.IV is devoted to the collective excitations of the unpolarized ground state at $`\nu =2`$. Dispersion curves of the spin density excitation is computed for various values of $`r_S`$ and Zeeman splitting energy, $`E_Z`$. The phase diagram as a function of $`r_S`$ and $`E_Z`$ is obtained in Sec.V by determining the critical $`E_Z`$ at which the energy of the spin density excitation vanishes. From the phase diagram we will learn that for large $`r_S`$ and small $`E_Z`$ neither the fully polarized nor the unpolarized state is stable against a spin-density wave state. The paper is concluded in Sec.VI, where we also discuss the implications of our study for the fractional quantum Hall effect (FQHE).
## II Diagrammatic formalism of collective excitation
### A Algebra in the symmetric gauge
Even though the choice of gauge does not affect physical quantities, we find it convenient to use the symmetric gauge. We will start by establishing the basic algebra in the symmetric gauge closely following Ref..
The Hamiltonian for an electron moving in a two dimensional space under a perpendicular magnetic field is given by
$$H=\frac{\stackrel{}{\pi }^2}{2m}$$
(3)
where the kinetic momentum is written down as
$$\stackrel{}{\pi }=i\mathrm{}\stackrel{}{}+\frac{e\stackrel{}{A}}{c}.$$
(4)
When the magnetic field is uniform, we can use a convenient algebraic method analogous to the solution by the ladder operator of the one-dimensional harmonic oscillator. In order to construct the algebraic formalism we first note that the $`x`$ and $`y`$ components of the kinetic momentum are canonically conjugate coordinates:
$$[\pi _x,\pi _y]=\frac{i\mathrm{}e}{c}\widehat{z}(\stackrel{}{}\times \stackrel{}{A})=\frac{i\mathrm{}^2}{l_{0}^{}{}_{}{}^{2}}.$$
(5)
From the commutation relationship between $`\pi _x`$ and $`\pi _y`$ we can define a ladder operator so that the ladder operator and its Hermitian conjugate satisfy the same commutation relation as those of the one-dimensional harmonic oscillator. That is,
$$[a,a^{}]=1,$$
(6)
where
$$a^{}\frac{l_0/\mathrm{}}{\sqrt{2}}(\pi _x+i\pi _y).$$
(7)
The Hamiltonian can now be written in the form of a one-dimensional harmonic oscillator.
$$H=\frac{\mathrm{}\omega _C}{2}(a^{}a+aa^{})$$
(8)
Therefore the eigenvalues are $`\mathrm{}\omega _C(n+1/2)`$ where $`n`$ is a non-negative integer which is known as the Landau level index. The eigenstates, however, cannot be fully determined by the Landau level index alone because the energy does not depend on the coordinates of the cyclotron orbit center, indicating a degeneracy of the Landau level. Let us define
$$\stackrel{}{C}=\stackrel{}{r}+\frac{\widehat{z}\times \stackrel{}{\pi }}{m\omega _C}$$
(9)
which is convetionally known as the guiding-center operator. The $`x`$ and $`y`$ components of the guiding-center operator are canonically conjugate coordinates similar to those of the kinetic momentum.
$$[C_x,C_y]=il_{0}^{}{}_{}{}^{2}$$
(10)
Therefore we can define another ladder operator by
$$b\frac{1}{\sqrt{2}l_0}(C_x+iC_y)$$
(11)
which satisfies
$$[b,b^{}]=1$$
(12)
and
$$[a,b]=[a^{},b]=[H,b]=0.$$
(13)
The fact that the ladder operator $`b`$ commutes with the Hamiltonian shows that the degeneracy of the lowest Landau level is actually related to the positioning of the guiding-center coordinate. Since we identify two independent sets of ladder operators in a two-dimensional space, the full set of eigenstates can be generated by repeatedly applying raising operators on the ground state:
$`|n,m={\displaystyle \frac{(a^{})^n(b^{})^m}{\sqrt{n!m!}}}|0,0`$ (14)
For a magnetic field pointing in the positive $`z`$ direction, the vector potential in the symmetric gauge is given by
$$\stackrel{}{A}=\frac{B}{2}(y,x,0)$$
(15)
and the ladder operators may be written in terms of $`z(x+iy)`$ as follows.
$$a^{}=\frac{i}{\sqrt{2}}\left(\frac{z}{2}2\frac{}{\overline{z}}\right)$$
(16)
$$a=\frac{i}{\sqrt{2}}\left(\frac{\overline{z}}{2}+2\frac{}{z}\right)$$
(17)
$$b^{}=\frac{1}{\sqrt{2}}\left(\frac{\overline{z}}{2}2\frac{}{z}\right)$$
(18)
$$b=\frac{1}{\sqrt{2}}\left(\frac{z}{2}+2\frac{}{\overline{z}}\right)$$
(19)
with the Hamiltonian given by
$$H=\frac{\mathrm{}\omega _C}{2}\left(4\frac{}{z}\frac{}{\overline{z}}+z\frac{}{z}\overline{z}\frac{}{\overline{z}}+\frac{z\overline{z}}{4}\right),$$
(20)
Here and in the rest of the paper, we use the convention that $`\mathrm{}=c=e=l_0=1`$, as well as the Area$`=1`$. In particular, this implies that the total number of flux quanta piercing the system, $`N_\varphi =\frac{1}{2\pi }`$.
Up to this point we have concentrated on the single particle Hamiltonian. In order to compute concrete physical quantities with the interaction treated perturbatively, we will need to use certain matrix elements and their various properties, which we now list .
$``$ *Plane Wave Matrix Elements*
The matrix element of a plane wave $`\mathrm{exp}(i\stackrel{}{k}\stackrel{}{r})`$ is given by
$$n_\beta ,m_\beta |e^{i\stackrel{}{k}\stackrel{}{r}}|n_\alpha ,m_\alpha =(i)^{n_\beta n_\alpha }g_{n_\beta n_\alpha }(\overline{\kappa })g_{m_\beta m_\alpha }(\kappa )e^{k^2/2},$$
(21)
where $`k=\sqrt{k_{x}^{}{}_{}{}^{2}+k_{y}^{}{}_{}{}^{2}}`$, $`\kappa =k_x+ik_y`$ and
$`g_{n_\beta n_\alpha }(\kappa )`$ $``$ $`n_\beta |\mathrm{exp}({\displaystyle \frac{i}{\sqrt{2}}}\kappa b^{})\mathrm{exp}({\displaystyle \frac{i}{\sqrt{2}}}\overline{\kappa }b)|n_\alpha `$ (22)
$`=`$ $`\left({\displaystyle \frac{2^{n_\alpha }n_\alpha !}{2^{n_\beta }n_\beta !}}\right)^{1/2}(i\kappa )^{n_\beta n_\alpha }L_{n_\alpha }^{n_\beta n_\alpha }(k^2/2)`$ (23)
where, for $`n_\beta >n_\alpha `$, $`L_{n_\alpha }^{n_\beta n_\alpha }`$ is the associated Laguerre polynomial, defined as:
$$L_n^m(x)=\underset{s=0}{\overset{n}{}}\frac{(x)^s}{s!}\left(\genfrac{}{}{0pt}{}{n+m}{ns}\right)$$
(24)
For $`n_\beta <n_\alpha `$ we define:
$$L_{n_\alpha }^{n_\beta n_\alpha }(x)=\frac{n_\beta !}{n_\alpha !}(x)^{n_\alpha n_\beta }L_{n_\beta }^{n_\alpha n_\beta }$$
(25)
Eq.(21) can be evaluated first by rewriting $`\stackrel{}{k}\stackrel{}{r}=(\kappa \overline{z}+\overline{\kappa }z)/2`$, expressing $`z`$ and $`\overline{z}`$ in terms of the ladder operators, and moving all annihilation operators to the right using $`e^Ae^B=e^Be^Ae^{[A,B]}`$.
$``$ *Matrix Products*
One of the most important properties of the matrix $`g_{mm^{}}(\kappa )`$ is its product
$$\underset{l}{}g_{n_\alpha l}(\kappa _1)g_{ln_\beta }(\kappa _2)=e^{\overline{\kappa _1}\kappa _2/2}g_{n_\alpha n_\beta }(\kappa _1+\kappa _2)$$
(26)
This can be derived by using the definition in Eq. (23) and the completeness of the ladder operator eigenstates.
$``$ *Eigenfunctions in the Symmetric Gauge*
Similar to duality of the position-space and momentum-space in the one-dimensional harmonic oscillator, we have the orbital wave function closely related to the plane-wave matrix element.
$$\stackrel{}{r}|n,m\varphi _{n,m}(\stackrel{}{r})=(i)^ng_{mn}(i\overline{z})\frac{e^{r^2/4}}{\sqrt{2\pi }}$$
(27)
To derive this, first note that where $`|0,0`$ is given by
$$\stackrel{}{r}|0,0=\frac{1}{\sqrt{2\pi }}\mathrm{exp}[\frac{1}{4}z\overline{z}]$$
(28)
which is annihilated by both $`a`$ and $`b`$. The eigenfunction for a general $`n`$ and $`m`$ is therefore given by
$$\stackrel{}{r}|n,m=\frac{i^n}{\sqrt{2\pi 2^{m+n}n!m!}}(\frac{z}{2}2\frac{}{\overline{z}})^n(\frac{\overline{z}}{2}2\frac{}{z})^m\mathrm{exp}[\frac{1}{4}z\overline{z}]$$
(29)
Now write $`e^{z\overline{z}/4}=e^{z\overline{z}/4}e^{z\overline{z}/2}`$ and use
$$\mathrm{exp}[\frac{1}{4}z\overline{z}](\frac{\overline{z}}{2}2\frac{}{z})^m(\frac{z}{2}2\frac{}{\overline{z}})^n\mathrm{exp}[\frac{1}{4}z\overline{z}]=(2)^{m+n}(\frac{}{\overline{z}})^n(\frac{}{z})^m$$
(30)
Defining $`t=z\overline{z}/2=r^2/2`$ one gets
$$\stackrel{}{r}|n,m=\frac{(i)^n}{\sqrt{2\pi 2^{n+m}n!m!}}e^{t/2}2^mz^{nm}(\frac{}{t})^nt^me^t$$
(31)
which reduces to Eq. (27) with the standard definition of the associated Laguerre polynomial.
$``$ *Trace*
The trace of $`g_{mm^{}}(\kappa )`$ is a constant because the charge density of a completely filled Landau level is a constant.
$$\frac{1}{A}\underset{m=0}{\overset{N_\varphi }{}}g_{mm}(\kappa )=\frac{N_\varphi }{A}\delta _{\stackrel{}{k},0}=\frac{1}{2\pi }\delta _{\stackrel{}{k},0}$$
(32)
where $`A`$ stands for the area of system and therefore $`N_\varphi /A`$ is $`1/2\pi `$.
Using the definition of $`g_{mm^{}}(\kappa )`$, Eq.(23), and properties of the Laguerre polynomial we find the transpose and complex conjugate as follows.
$``$ *Transpose*
$$g_{n_\alpha n_\beta }(\kappa )=g_{n_\beta n_\alpha }(\overline{\kappa }).$$
(33)
$``$ *Complex Conjugate*
$$\overline{g}_{n_\alpha n_\beta }(\kappa )=g_{n_\alpha n_\beta }(\overline{\kappa })$$
(34)
$``$ *Orthogonality*
The final property we will use is the orthogonality between the plane-wave matrix elements which will be useful in computing the self-energy:
$$d^2\stackrel{}{k}e^{k^2/2}g_{n_\alpha n_\beta }(\kappa )g_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(\overline{\kappa })=2\pi \delta _{n_\alpha ,n_{\alpha }^{}{}_{}{}^{}}\delta _{n_\beta ,n_{\beta }^{}{}_{}{}^{}}.$$
(35)
### B Response function
The response function is a very important quantity from which we can deduce many physical observables assuming linear response. In particular, the collective excitations correspond to the poles of the response function. In order to calculate the response function we use the zero temperature limit of the Matsubara formalism, using the standard analytic continuation ($`i\omega \omega +i\delta `$) in order to get the retarded response function. We will describe below in detail only the charge-density response function, since the spin-density response function can be obtained with a straightforward modification. The charge-density response function is related to the density-density correlation function as follows:
$$\chi (k,i\omega )_0^{\mathrm{}}𝑑\tau e^{i\omega \tau }T_\tau \widehat{\rho }(\stackrel{}{k},\tau )\widehat{\rho }(\stackrel{}{k},0).$$
(36)
where the density operator, $`\widehat{\rho }(\stackrel{}{k},\tau )`$, is given in the symmetric gauge by
$$\widehat{\rho }(\stackrel{}{k},\tau )\underset{n_\alpha m_\alpha }{}\underset{n_\beta m_\beta }{}n_\alpha ,m_\alpha |e^{i\stackrel{}{k}\stackrel{}{r}}|n_\beta ,m_\beta c_{n_\alpha m_\alpha }^{}(\tau )c_{n_\beta m_\beta }(\tau ).$$
(37)
The actual computation of the density response function is performed in the perturbation theory. Assuming at any given time there is a single particle-hole pair in the system, we write the response function which is depicted by the Feynman diagram in Fig.3.
$$\chi (k,i\omega )=\underset{n_\alpha n_\beta }{}\underset{m_\alpha m_\beta }{}n_\alpha ,m_\alpha |e^{i\stackrel{}{k}\stackrel{}{r}}|n_\beta ,m_\beta D_{n_\alpha n_\beta }(i\omega )\mathrm{\Gamma }_{n_\alpha n_\beta }^{m_\alpha m_\beta }(k,i\omega )$$
(38)
where the vertex function $`\mathrm{\Gamma }_{n_\alpha n_\beta }^{m_\alpha m_\beta }(k,i\omega )`$ satisfies the following equation:
$`\mathrm{\Gamma }_{n_\alpha n_\beta }^{m_\alpha m_\beta }`$ $`(`$ $`k,i\omega )=n_\beta ,m_\beta |e^{i\stackrel{}{k}\stackrel{}{r}}|n_\alpha ,m_\alpha `$ (39)
$`+`$ $`n_\beta ,m_\beta |e^{i\stackrel{}{k}\stackrel{}{r}}|n_\alpha ,m_\alpha \stackrel{~}{v}(k){\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}{\displaystyle \underset{m_{\alpha }^{}{}_{}{}^{}m_{\beta }^{}{}_{}{}^{}}{}}n_{\alpha }^{}{}_{}{}^{},m_{\alpha }^{}{}_{}{}^{}|e^{i\stackrel{}{k}\stackrel{}{r}}|n_{\beta }^{}{}_{}{}^{},m_{\beta }^{}{}_{}{}^{}`$ (40)
$`\times `$ $`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^{m_\alpha ^{}m_{\beta }^{}{}_{}{}^{}}(k,i\omega )`$ (41)
$``$ $`{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}{\displaystyle \underset{m_{\alpha }^{}{}_{}{}^{}m_{\beta }^{}{}_{}{}^{}}{}}\left[{\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\varphi _{n_\alpha ,m_\alpha }(\stackrel{}{r_1})\overline{\varphi }_{n_\beta ,m_\beta }(\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\overline{\varphi }_{n_{\alpha }^{}{}_{}{}^{},m_{\alpha }^{}{}_{}{}^{}}(\stackrel{}{r_1})\varphi _{n_{\beta }^{}{}_{}{}^{},m_{\beta }^{}{}_{}{}^{}}(\stackrel{}{r_2})}\right]`$ (42)
$`\times `$ $`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^{m_\alpha ^{}m_{\beta }^{}{}_{}{}^{}}(k,i\omega ).`$ (43)
where $`v(r)`$ is a potential and $`\widehat{v}(k)`$ is its Fourier transform. The Feynman diagrams corresponding to each term on the right-hand side of the Eq. (43) are shown in Fig.3: The first term describes just the bare vertex and the second and third terms depict the RPA correction (Bubble diagram) and the binding energy term (Ladder diagram), respectively. In the Eq. (38) and Eq. (43) $`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )`$ is defined as the frequency integral of the product of two Green functions:
$`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega ){\displaystyle \frac{d\omega ^{}}{2\pi }G_{n_{\alpha }^{}{}_{}{}^{}}(i\omega ^{}i\omega )G_{n_{\beta }^{}{}_{}{}^{}}(i\omega ^{})}`$ (44)
and the Green function is given by
$$G_n(i\omega )=\frac{1}{i\omega (n\mu _0)\omega _C\mathrm{\Sigma }_n^{n_0}},$$
(45)
where $`n`$ is the Landau level index of the electron or hole and $`\mu _0`$ is the chemical potential which is halfway between the highest occupied Landau level and the lowest unoccupied Landau level. We note that the Green function is fully dressed, containing the self energy ($`\mathrm{\Sigma }`$) correction, and is also independent of the $`m`$ quantum number. In general the self energy $`\mathrm{\Sigma }`$ is a very complicated function of $`\stackrel{}{k}`$ and $`\omega `$, but it turns out that in this system the self energy is real and depends on just the Landau level index but not $`\stackrel{}{k}`$ and $`\omega `$. We will use this result below, postponing the explicit derivation to the next section.
Plugging the explicit form of Green function into Eq.(44) gives
$`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )`$ $``$ $`{\displaystyle \frac{d\omega ^{}}{2\pi }G_{n_{\alpha }^{}{}_{}{}^{}}(i\omega ^{}i\omega )G_{n_{\beta }^{}{}_{}{}^{}}(i\omega ^{})}`$ (46)
$`=`$ $`{\displaystyle \frac{\theta \left((\mu _0n_{\beta }^{}{}_{}{}^{})\omega _C\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0}\right)\theta \left((\mu _0n_{\alpha }^{}{}_{}{}^{})\omega _C\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\right)}{i\omega (n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{})\omega _C(\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0})}}`$ (47)
$``$ $`{\displaystyle \frac{\theta (\mu _0n_{\beta }^{}{}_{}{}^{})\theta (\mu _0n_{\alpha }^{}{}_{}{}^{})}{i\omega (n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{})\omega _C(\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0})}}`$ (48)
where
$$\theta (x)=\{\begin{array}{cc}1& \text{if }x>0\\ 0& \text{if }x<0\\ 1/2& \text{if }x=0\end{array}$$
(49)
and it is assumed that the self energy does not modify the Fermi level significantly but moves the position of the pole in the denominator of Eq.(48).
In order to solve the vertex equation, Eq.(43), we first eliminate the angular momentum index by defining a new vertex function $`\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )`$:
$`\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )`$ $``$ $`e^{k^2/2}{\displaystyle \underset{m_\alpha m_\beta }{}}g_{m_\alpha m_\beta }(\kappa )\mathrm{\Gamma }_{n_\alpha n_\beta }^{m_\alpha m_\beta }(k,i\omega )`$ (50)
$``$ (51)
$`\mathrm{\Gamma }_{n_\alpha n_\beta }^{m_\alpha m_\beta }(k,i\omega )`$ $`=`$ $`2\pi \overline{g}_{m_\alpha m_\beta }(\kappa )\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega ).`$ (52)
Eq.(52) is obtained from Eq.(50) through the property of the plane-wave matrix product Eq.(26). Noting that the bare vertex is just a plane-wave matrix element defined in the Eq.(21) and utilizing the properties of the plane-wave matrix elements, we succeed in summing over all the angular momentum indices to get a new vertex equation.
$`\mathrm{\Gamma }_{n_\alpha n_\beta }`$ $`(`$ $`k,i\omega )={\displaystyle \frac{(i)^{n_\beta n_\alpha }}{2\pi }}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa })`$ (53)
$`+`$ $`(i)^{n_\beta n_\alpha }e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa }){\displaystyle \frac{\stackrel{~}{v}(k)}{2\pi }}{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}(i)^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}g_{n_{\alpha }^{}{}_{}{}^{}n_\beta ^{}}(\overline{\kappa })D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )`$ (54)
$``$ $`{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}\left[2\pi {\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\overline{\mathrm{\Phi }}_{n_{\alpha }^{}{}_{}{}^{},n_{\beta }^{}{}_{}{}^{}}^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})}\right]D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )`$ (55)
where
$$\mathrm{\Phi }_{n_\alpha n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})e^{k^2/4}\underset{m_\alpha m_\beta }{}g_{m_\alpha m_\beta }(\kappa )\varphi _{n_\alpha m_\alpha }(\stackrel{}{r_1})\overline{\varphi }_{n_\beta m_\beta }(\stackrel{}{r_2}).$$
(56)
Similarly, the density response function can be written in terms of the new vertex function, $`\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )`$.
$$\chi (k,i\omega )=\underset{n_\alpha n_\beta }{}(i)^{n_\alpha n_\beta }g_{n_\alpha n_\beta }(\overline{\kappa })D_{n_\alpha n_\beta }(i\omega )\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )$$
(57)
The third term on the right hand side of Eq.(55) contains a formal expression for the Coulomb-potential matrix elements between the wave function $`\mathrm{\Phi }_{n_\alpha n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ and $`\mathrm{\Phi }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ which prove to be the eigenfunctions of the two-body Hamiltonian for oppositely charged particles confined in the Landau levels $`(n_\alpha ,n_\beta )`$ and $`(n_{\alpha }^{}{}_{}{}^{},n_{\beta }^{}{}_{}{}^{})`$, respectively . Therefore, following Kallin and Halperin, we will call it the binding energy. The second term in Eq.(55) has the physical interpretation of the exchange energy of particle-hole pair which is the only term in the usual RPA. For convenience let us define $`n_{\alpha \beta }=n_\alpha n_\beta `$, and denote the binding energy and the exchange energy as $`V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ and $`U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ respectively:
$$U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)i^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa })\frac{\stackrel{~}{v}(k)}{2\pi }g_{n_{\alpha }^{}{}_{}{}^{}n_\beta ^{}}(\overline{\kappa })$$
(58)
$$V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)2\pi d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\overline{\mathrm{\Phi }}_{n_{\alpha }^{}{}_{}{}^{},n_{\beta }^{}{}_{}{}^{}}^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})$$
(59)
Then Eq. (55) becomes
$`\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )`$ $`=`$ $`{\displaystyle \frac{i^{n_{\alpha \beta }}}{2\pi }}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa })`$ (60)
$``$ $`{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}\left[V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)\right]D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega ).`$ (61)
The Green function $`D_{n_\alpha ,n_\beta }`$ contains a factor $`\left(\theta (n_0+1/2n_\beta )\theta (n_0+1/2n_\alpha )\right)`$ which vanishes for certain choices of $`(n_\alpha ,n_\beta )`$. However, these choices co not contribute to $`\chi `$ in Eq. (57). Therefore, restricting to only those values of $`(n_\alpha ,n_\beta )`$ for which $`D_{n_\alpha ,n_\beta }`$ is non-zero, we transform the equation into an associated set of linear equations:
$`{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}[\delta _{n_\alpha ,n_{\alpha }^{}{}_{}{}^{}}\delta _{n_\beta ,n_{\beta }^{}{}_{}{}^{}}D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(i\omega )+V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $``$ $`U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)]D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )`$ (62)
$`=`$ $`{\displaystyle \frac{i^{n_{\alpha \beta }}}{2\pi }}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa }),`$ (63)
where
$$D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(i\omega )=\left(i\omega (n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{})\omega _C(\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0})\right)\times \left(\theta (n_0+1/2n_{\beta }^{}{}_{}{}^{})\theta (n_0+1/2n_{\alpha }^{}{}_{}{}^{})\right).$$
(64)
(Note that here $`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(i\omega )`$ is the inverse of the matrix element, not the matrix element of the inverse.) Eq.(63) can be written in the form of a matrix equation by defining a matrix $`M`$ as follows:
$$M_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )\delta _{n_\alpha ,n_{\alpha }^{}{}_{}{}^{}}\delta _{n_\beta ,n_{\beta }^{}{}_{}{}^{}}D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(i\omega )+V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k).$$
(65)
$`M_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )`$ can be viewed as a matrix element of $`M`$ if we consider the set of indices $`(n_\alpha n_\beta )`$ to be a collective index. Therefore we write Eq.(63) in the following form:
$`{\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}M_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\omega )\mathrm{\Gamma }_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )={\displaystyle \frac{i^{n_{\alpha \beta }}}{2\pi }}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa }).`$ (66)
Inverting the matrix $`M`$ amounts to solving the vertex equation. That is,
$`D_{n_\alpha n_\beta }(i\omega )\mathrm{\Gamma }_{n_\alpha n_\beta }(k,i\omega )={\displaystyle \underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}}{\displaystyle \frac{i^{n_{\alpha \beta }^{}{}_{}{}^{}}}{2\pi }}e^{k^2/2}(M^1)_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )\overline{g}_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(\overline{\kappa }).`$ (67)
where $`M^1`$ is the inverse matrix of $`M`$. Finally, we obtain the density response function in terms of the matrix $`M`$.
$$\chi (k,i\omega )=\underset{n_\alpha n_\beta }{}\underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}\frac{(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}}{2\pi }e^{k^2/2}g_{n_\alpha n_\beta }(\overline{\kappa })(M^1)_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,i\omega )\overline{g}_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(\overline{\kappa })$$
(68)
Since the summation of the Landau level indices should be performed over all the possible states, the number of terms is actually infinite. However, it is possible to obtain an accurate estimate keeping a reasonably small number of terms.
### C Collective excitation
The collective excitations are the poles of the response function, which, from Eq.(68), correspond to energies for which the inverse of $`M`$ becomes singular, that is to say, when
$$Det\left[M(i\omega \omega +i\delta )\right]=0.$$
(69)
We carry out detailed computations of the binding energy, RPA energy and self energy in order to explicitly evaluate the dispersion curve of the collective excitation. We start with the binding energy.
$``$ Binding energy (Ladder diagram contribution)
The explicit form of $`\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ is obtained from Eq.(56) by utilizing the plane-wave matrix product formula, Eq.(26):
$`\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ $`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }}}{2\pi }}e^{k^2/4}{\displaystyle \underset{m_\alpha m_\beta }{}}g_{m_\alpha m_\beta }(\kappa )e^{z_1\overline{z_1}/4}g_{m_\alpha n_\alpha }(i\overline{z_1})e^{z_2\overline{z_2}/4}\overline{g}_{m_\beta n_\beta }(i\overline{z_2})`$ (70)
$`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }}}{2\pi }}e^{k^2/4}e^{(r_{1}^{}{}_{}{}^{2}+r_{2}^{}{}_{}{}^{2})/4}{\displaystyle \underset{m_\alpha m_\beta }{}}g_{n_\alpha m_\alpha }(iz_1)g_{m_\alpha m_\beta }(\kappa )g_{m_\beta n_\beta }(iz_2)`$ (71)
$`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }}}{2\pi }}e^{k^2/4}e^{(r_{1}^{}{}_{}{}^{2}+r_{2}^{}{}_{}{}^{2})/4}e^{i\overline{\kappa }z_2/2}{\displaystyle \underset{m_\alpha }{}}g_{n_\alpha m_\alpha }(iz_1)g_{m_\alpha n_\beta }(\kappa +iz_2)`$ (72)
$`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }}}{2\pi }}e^{k^2/4}e^{(r_{1}^{}{}_{}{}^{2}+r_{2}^{}{}_{}{}^{2})/4}e^{i(\overline{\kappa }z_2+\kappa \overline{z_1})/2}e^{\overline{z_1}z_2/2}g_{n_\alpha n_\beta }(\kappa iz_1+iz_2)`$ (73)
$`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }}}{2\pi }}e^{i\stackrel{}{R}(\stackrel{}{k}+\stackrel{}{r}\times \widehat{z}/2)}e^{\frac{|\stackrel{}{r}+\stackrel{}{k}\times \widehat{z}|^2}{4}}\overline{g}_{n_\alpha n_\beta }(i(\overline{z}+i\overline{k})).`$ (74)
Following Kallin and Halperin, the wave function $`\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ can be shown to be an eigenstate of the following Hamiltonian where the particles 1 and 2 are projected onto the Landau levels with indices $`n_\alpha `$ and $`n_\beta `$ respectively.
$`\widehat{H}`$ $`=`$ $`\widehat{P}_{n_\alpha ,n_\beta }H\widehat{P}_{n_\alpha ,n_\beta }`$ (75)
$`=`$ $`\widehat{P}_{n_\alpha ,n_\beta }\left[{\displaystyle \frac{\stackrel{}{\pi _1}^2}{2m}}+{\displaystyle \frac{\stackrel{}{\pi _2}^2}{2m}}+v_{arb}(|\stackrel{}{r_1}\stackrel{}{r_2}|)\right]\widehat{P}_{n_\alpha ,n_\beta }`$ (76)
$`=`$ $`\mathrm{}\omega _C(n_\alpha +n_\beta +1)+\widehat{P}_{n_\alpha ,n_\beta }v_{arb}(|\stackrel{}{r_1}\stackrel{}{r_2}|)\widehat{P}_{n_\alpha ,n_\beta }`$ (77)
where $`\stackrel{}{\pi _1}`$ $`(\stackrel{}{\pi _2})`$ is the kinetic momentum for the particle 1 (2), $`v_{arb}(r)`$ is an arbitrary (attractive) potential and $`\widehat{P}_{n_\alpha ,n_\beta }`$ is the operator projecting the particles onto the Landau levels with indices $`n_\alpha `$ and $`n_\beta `$. One can prove that $`\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ is the eigenfunction of $`\widehat{H}`$ by computing the overlap between $`v_{arb}(|\stackrel{}{r_1}\stackrel{}{r_2}|)\times \mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ and any arbitrary basis state in the projected Hilbert space, for example $`\overline{\varphi }_{n_\alpha m_\gamma }(\stackrel{}{r_1})\varphi _{n_\beta m_\delta }(\stackrel{}{r_2})`$. Then we note that it is proportional to the overlap between $`\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})`$ and $`\overline{\varphi }_{n_\alpha m_\gamma }(\stackrel{}{r_1})\varphi _{n_\beta m_\delta }(\stackrel{}{r_2})`$. Of course, the proportionality constant is the eigenvalue which is equal to the binding energy previously defined in Eq.(59) in the case of $`n_{\alpha }^{}{}_{}{}^{}=n_\alpha `$ and $`n_{\beta }^{}{}_{}{}^{}=n_\beta `$.
Now let us get the explicit formula for the binding energy.
$`V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $`=`$ $`2\pi {\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\mathrm{\Phi }_{n_\alpha ,n_\beta }^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\overline{\mathrm{\Phi }}_{n_{\alpha }^{}{}_{}{}^{},n_{\beta }^{}{}_{}{}^{}}^\kappa (\stackrel{}{r_1},\stackrel{}{r_2})}`$ (78)
$`=`$ $`{\displaystyle \frac{(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}}{2\pi }}{\displaystyle d^2\stackrel{}{R}d^2\stackrel{}{r}e^{\frac{|\stackrel{}{r}+\stackrel{}{k}\times \widehat{z}|^2}{2}}v(\stackrel{}{r})\overline{g}_{n_\alpha n_\beta }(i(\overline{z}+i\overline{k}))g_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i(\overline{z}+i\overline{k}))}`$ (79)
$`=`$ $`(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r}\stackrel{}{k}\times \widehat{z})e^{r^2/2}\overline{g}_{n_\alpha n_\beta }(i\overline{z})g_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\overline{z})}`$ (80)
$`=`$ $`(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}{\displaystyle \frac{d^2\stackrel{}{q}}{(2\pi )^2}\stackrel{~}{v}(q)e^{i\stackrel{}{q}\stackrel{}{k}\times \widehat{z}}\frac{d^2\stackrel{}{r}}{2\pi }e^{i\stackrel{}{q}\stackrel{}{r}}e^{r^2/2}\overline{g}_{n_\alpha n_\beta }(i\overline{z})g_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(i\overline{z})}`$ (81)
$`=`$ $`(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}\left({\displaystyle \frac{2^{n_\beta }2^{n_{\beta }^{}{}_{}{}^{}}n_\beta !n_{\beta }^{}{}_{}{}^{}!}{2^{n_\alpha }2^{n_{\alpha }^{}{}_{}{}^{}}n_\alpha !n_{\alpha }^{}{}_{}{}^{}!}}\right)^{1/2}{\displaystyle \frac{d^2\stackrel{}{q}}{(2\pi )^2}\stackrel{~}{v}(q)e^{i\stackrel{}{q}\stackrel{}{k}\times \widehat{z}}}`$ (82)
$`\times `$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }e^{i\stackrel{}{q}\stackrel{}{r}}e^{r^2/2}z^{n_{\alpha \beta }}\overline{z}^{n_{\alpha \beta }^{}}L_{n_\beta }^{n_{\alpha \beta }}(\frac{r^2}{2})L_{n_{\beta }^{}{}_{}{}^{}}^{n_{\alpha \beta }^{}{}_{}{}^{}}(\frac{r^2}{2})},`$ (83)
where $`L_n^m`$ is an associated Laguerre polynomial. By using the fact $`_0^{2\pi }𝑑\theta e^{i(x\mathrm{cos}\theta +n\theta )}=2\pi i^nJ_n(x)`$ we perform the angle integrations to get the final formula:
$`V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $`=`$ $`e^{i(n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{})\theta _\stackrel{}{k}}\left({\displaystyle \frac{2^{n_\beta }2^{n_{\beta }^{}{}_{}{}^{}}n_\beta !n_{\beta }^{}{}_{}{}^{}!}{2^{n_\alpha }2^{n_{\alpha }^{}{}_{}{}^{}}n_\alpha !n_{\alpha }^{}{}_{}{}^{}!}}\right)^{1/2}{\displaystyle _0^{\mathrm{}}}𝑑q{\displaystyle \frac{q}{2\pi }}\stackrel{~}{v}(q)J_{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}(qk)`$ (84)
$`\times `$ $`{\displaystyle _0^{\mathrm{}}}𝑑re^{r^2/2}r^{n_{\alpha \beta }+n_{\alpha \beta }^{}{}_{}{}^{}+1}L_{n_\beta }^{n_{\alpha \beta }}({\displaystyle \frac{r^2}{2}})L_{n_{\beta }^{}{}_{}{}^{}}^{n_{\alpha \beta }^{}{}_{}{}^{}}({\displaystyle \frac{r^2}{2}})J_{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}(qr),`$ (85)
where $`n_{\alpha \beta }=n_\alpha n_\beta `$, $`n_{\alpha \beta }^{}{}_{}{}^{}=n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}`$ and $`J_n`$ is a Bessel function and $`\theta _\stackrel{}{k}`$ is the angle of $`\stackrel{}{k}`$ measured from the $`x`$-axis.
$``$ Random phase approximation energy (Bubble diagram contribution)
The exchange energy from the RPA is rather straightforward to calculate since there is no integration involved. The explicit form is given by
$`U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $`=`$ $`i^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}e^{k^2/2}\overline{g}_{n_\alpha n_\beta }(\overline{\kappa }){\displaystyle \frac{\stackrel{~}{v}(k)}{2\pi }}g_{n_{\alpha }^{}{}_{}{}^{}n_\beta ^{}}(\overline{\kappa })`$ (86)
$`=`$ $`e^{i(n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{})\theta _\stackrel{}{k}}\left({\displaystyle \frac{2^{n_\beta }2^{n_{\beta }^{}{}_{}{}^{}}n_\beta !n_{\beta }^{}{}_{}{}^{}!}{2^{n_\alpha }2^{n_{\alpha }^{}{}_{}{}^{}}n_\alpha !n_{\alpha }^{}{}_{}{}^{}!}}\right)^{1/2}{\displaystyle \frac{\stackrel{~}{v}(k)}{2\pi }}e^{k^2/2}k^{n_{\alpha \beta }+n_{\alpha \beta }^{}{}_{}{}^{}}L_{n_\beta }^{n_{\alpha \beta }}({\displaystyle \frac{k^2}{2}})L_{n_{\beta }^{}{}_{}{}^{}}^{n_{\alpha \beta }^{}{}_{}{}^{}}({\displaystyle \frac{k^2}{2}}).`$ (87)
Incidentally, a comparison between Eq.(85) and Eq.(87) reveals that $`U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ and $`V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ have the same phase factor, $`e^{i(n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{})\theta _\stackrel{}{k}}`$. Therefore the phase factor can be eliminated in a consistent way, which is expected because the system is uniform and isotropic.
$``$ Self energy
The last diagram in the Fig.3 is the self-energy contribution to the “full” Green function. As before, a single particle-hole pair is assumed, which is reflected in the Feynman diagram through the unscreened Coulomb line. The corresponding Dyson equation is solved in the conventional way:
$`G_{n_\alpha }(i\omega )`$ $`=`$ $`G_{n_\alpha }^{(0)}(i\omega )+G_{n_\alpha }^{(0)}(i\omega )\mathrm{\Sigma }_{n_\alpha }^{n_0}G_{n_\alpha }(i\omega )`$ (88)
$`=`$ $`{\displaystyle \frac{1}{i\omega (n_\alpha \mu _0)\omega _C\mathrm{\Sigma }_{n_\alpha }^{n_0}}}`$ (89)
where
$$G_{n_\alpha }^{(0)}(i\omega )=\frac{1}{i\omega (n_\alpha \mu _0)\omega _C}$$
(90)
and
$`\mathrm{\Sigma }_{n_\alpha }^{n_0}`$ $``$ $`{\displaystyle \underset{n_\beta m_\beta }{}}\left[{\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\overline{\varphi }_{n_\alpha m_\alpha }(\stackrel{}{r_1})\varphi _{n_\alpha m_\alpha }(\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\varphi _{n_\beta m_\beta }(\stackrel{}{r_1})\overline{\varphi }_{n_\beta m_\beta }(\stackrel{}{r_2})}\right]`$ (91)
$`\times `$ $`{\displaystyle \frac{d\omega }{2\pi }G_{n_\beta }^{(0)}(i\omega )}`$ (92)
$`=`$ $`{\displaystyle \underset{n_\beta m_\beta }{}}\left[{\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\overline{\varphi }_{n_\alpha m_\alpha }(\stackrel{}{r_1})\varphi _{n_\alpha m_\alpha }(\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\varphi _{n_\beta m_\beta }(\stackrel{}{r_1})\overline{\varphi }_{n_\beta m_\beta }(\stackrel{}{r_2})}\right]`$ (93)
$`\times `$ $`\theta (n_0+1/2n_\beta )`$ (94)
$`=`$ $`{\displaystyle \underset{n_\beta =0}{\overset{n_0}{}}}\left[{\displaystyle d^2\stackrel{}{r_1}d^2\stackrel{}{r_2}\overline{\varphi }_{n_\alpha m_\alpha }(\stackrel{}{r_1})\varphi _{n_\alpha m_\alpha }(\stackrel{}{r_2})v(\stackrel{}{r_1}\stackrel{}{r_2})\underset{m_\beta }{}\varphi _{n_\beta m_\beta }(\stackrel{}{r_1})\overline{\varphi }_{n_\beta m_\beta }(\stackrel{}{r_2})}\right]`$ (95)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r_1}}{2\pi }\frac{d^2\stackrel{}{r_2}}{2\pi }e^{(r_{1}^{}{}_{}{}^{2}+r_{2}^{}{}_{}{}^{2}z_2\overline{z_1})/2}v(\stackrel{}{r_1}\stackrel{}{r_2})\overline{g}_{m_\alpha n_\alpha }(i\overline{z_1})g_{m_\alpha n_\alpha }(i\overline{z_2})\underset{n_\beta =0}{\overset{n_0}{}}g_{n_\beta n_\beta }(i(\overline{z_1}\overline{z_2}))}`$ (96)
In the above equation the explicit form of the single particle eigenstate, Eq.(27), and the plane-wave matrix product formula, Eq.(26), have been used. It is convenient for the computation of integrals to change the variables from $`\stackrel{}{r_1}`$ and $`\stackrel{}{r_2}`$ to the center of mass coordinate $`\stackrel{}{R}`$ and the relative coordinate $`\stackrel{}{r}`$. In the form of a complex number $`\stackrel{}{R}Z(z_1+z_2)/2`$ and $`\stackrel{}{r}zz_1z_2`$.
$`\mathrm{\Sigma }_{n_\alpha }^{n_0}`$ $`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{3r^2/8}\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})\frac{d^2\stackrel{}{R}}{2\pi }e^{R^2/2}e^{(z\overline{Z}\overline{z}Z)/4}\overline{g}_{m_\alpha n_\alpha }(i\overline{Z}+\frac{i}{2}\overline{z})g_{m_\alpha n_\alpha }(i\overline{Z}\frac{i}{2}\overline{z})}`$ (97)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})\frac{d^2\stackrel{}{R}}{2\pi }\mathrm{exp}\left(\frac{1}{2}|Z+z/2|^2+\frac{1}{2}(Z+z/2)\overline{z}\right)}`$ (98)
$`\times `$ $`\overline{g}_{m_\alpha n_\alpha }(i\overline{Z}+{\displaystyle \frac{i}{2}}\overline{z})g_{m_\alpha n_\alpha }(i\overline{Z}{\displaystyle \frac{i}{2}}\overline{z})`$ (99)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})\frac{d^2\stackrel{}{R}^{}}{2\pi }\mathrm{exp}\left(\frac{1}{2}|Z^{}|^2\right)}`$ (100)
$`\times `$ $`\overline{g}_{m_\alpha n_\alpha }(i\overline{Z}^{})\mathrm{exp}\left({\displaystyle \frac{1}{2}}Z^{}\overline{z}\right)g_{m_\alpha n_\alpha }(i\overline{Z}^{}i\overline{z})`$ (101)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})\frac{d^2\stackrel{}{R}^{}}{2\pi }\mathrm{exp}\left(\frac{1}{2}|Z^{}|^2\right)}`$ (102)
$`\times `$ $`g_{n_\alpha m_\alpha }(i\overline{Z}^{}){\displaystyle \underset{l}{}}g_{m_\alpha l}(i\overline{Z}^{})g_{ln_\alpha }(i\overline{z})`$ (103)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})\underset{l}{}g_{ln_\alpha }(i\overline{z})\underset{=\delta _{l,n_\alpha }\text{ : orthogonality}}{\underset{}{\frac{d^2\stackrel{}{R}^{}}{2\pi }\mathrm{exp}\left(\frac{1}{2}|Z^{}|^2\right)g_{n_\alpha m_\alpha }(i\overline{Z}^{})g_{lm_\alpha }(iZ^{})}}}`$ (104)
$`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}g_{n_\alpha n_\alpha }(i\overline{z})\underset{n_\beta =0}{\overset{n_0}{}}L_{n_\beta }^0(\frac{r^2}{2})}`$ (105)
The angular momentum index $`m_\alpha `$ is naturally eliminated after the integration over the center-of-mass coordinate by using Eq.(35). The plane-wave matrix product formula has also been used when we proceed from the third step to the fourth step. Using $`_{n=0}^{n_0}L_n^0(x)=L_{n_0}^1(x)`$ one can finally write down the self energy as follows:
$`\mathrm{\Sigma }_{n_\alpha }^{n_0}`$ $`=`$ $`{\displaystyle \frac{d^2\stackrel{}{r}}{2\pi }v(\stackrel{}{r})e^{r^2/2}L_{n_\alpha }^0(\frac{r^2}{2})L_{n_0}^1(\frac{r^2}{2})}`$ (106)
$`=`$ $`{\displaystyle _0^{\mathrm{}}}𝑑q{\displaystyle \frac{q}{2\pi }}\stackrel{~}{v}(q){\displaystyle _0^{\mathrm{}}}𝑑rrL_{n_\alpha }^0({\displaystyle \frac{r^2}{2}})L_{n_0}^1({\displaystyle \frac{r^2}{2}})J_0(qr)e^{r^2/2}.`$ (107)
### D Spin degree of freedom
In order to see what modifications need to be made in the above analysis to include the spin degree of freedom, it is instructive to recall the physical meaning of the three parts of the collective exciton energy. The binding energy due to the ladder diagram is the *direct* interaction between the excited electron and the hole. Therefore it will not be affected by the presence of the spin degree of freedom. On the other hand, the RPA energy is the *exchange* energy between the excited electron and the hole. So it will vanish if the excited electron has a different spin than the hole, as in the case of the spin density excitation. In other words the electron-hole pair with the same spin cannot recombine through the Coulomb potential. In the case of the charge density excitation, however, the RPA energy depends on the polarization of the ground state. We have computed the RPA energy in the previous section assuming that all the electrons have the same spin, which corresponds to the fully polarized state. If the ground state is unpolarized, the RPA energy will be twice as large as that of the fully polarized state simply because the particle-hole pair can be created and annihilated as either a spin-up or spin-down pair. Formally speaking, the vertex equation, Eq.(43), will have two identical RPA terms for a given set of indices. The self energy term is due to the exchange energy between a given electron and the rest of electrons in the system while there is no direct term because we assume a neutralizing positive charge background. Since it is an exchange term, we have to include only the interaction between the electrons with the same spin. Therefore Eq.(107) is generalized to include the spin degree of freedom as follows:
$$\mathrm{\Sigma }_n^{n(\sigma )}=_0^{\mathrm{}}𝑑q\frac{q}{2\pi }\stackrel{~}{v}(q)_0^{\mathrm{}}𝑑rrL_n^0(\frac{r^2}{2})L_{n(\sigma )}^1(\frac{r^2}{2})J_0(qr)e^{r^2/2}.$$
(108)
where
$$\sigma =\{\begin{array}{cc}1/2& \text{for the spin antiparallel to B field}\\ 1/2& \text{for the spin parallel to B field}\end{array}$$
(109)
and $`n(\sigma )`$ is the index of the highest Landau level occupied by the electron with spin $`\sigma `$. Using this new self energy one can rewrite Eq.(48) as follows:
$$D_{n_\alpha n_\beta }(i\omega )=\frac{\theta (\mu (\sigma _\beta )n_\beta )\theta (\mu (\sigma _\alpha )n_\alpha )}{i\omega (n_\alpha n_\beta )\omega _C(\sigma _\alpha \sigma _\beta )E_Z(\mathrm{\Sigma }_{n_\alpha }^{n(\sigma _\alpha )}\mathrm{\Sigma }_{n_\beta }^{n(\sigma _\beta )})},$$
(110)
where the chemical potential $`\mu (\sigma )`$ was defined earlier, and the Zeeman coupling is included through the term $`\sigma E_z`$.
### E Solutions for the pole of the response function
Computing the dispersion curve of collective excitations for a general $`r_S`$ requires solving the equation for the pole of the response function, Eq.(69). In the limit $`r_S0`$, when there is no Landau level mixing, the matrix has a block diagonal form since collective modes with different kinetic energies do not couple, and each block, which has a finite dimension, can be diagonalized separately to obtain the collective mode energies . At non-zero values of $`r_S`$, however, the full matrix must be diagonalized. Strictly speaking, the matrix $`M`$ in Eq.(69) is of infinite dimension, but in practice, we work with a finite size matrix, keeping a sufficient number of Landau levels to ensure a convergence of the collective mode energy. For the lowest energy collective modes, which are our primary concern, and for $`r_S<6`$, we find that it is adequate to work with $`M`$ of dimension of up to 20.
We also find it useful to convert the Eq. (69) into an eigenvalue equation, the solution of which can be obtained using standard linear algebraic methods. In this formulation it is natural to define an effective Hamiltonian to be the matrix in the eigenvalue problem. The details of the procedure are discussed in the remainder of this section.
For convenience we write down Eq.(68) here after the analytic continuation.
$$\chi (k,\omega )=\underset{n_\alpha n_\beta }{}\underset{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}{}\frac{(i)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}}{2\pi }e^{k^2/2}g_{n_\alpha n_\beta }(\overline{\kappa })(M^1)_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,\omega )\overline{g}_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(\overline{\kappa }).$$
(111)
For an arbitrary set of indices $`(n_\alpha ,n_\beta )`$ where $`n_\alpha `$ is greater than $`n_\beta `$, there is a reversed set $`(n_\beta ,n_\alpha )`$ whose kinetic energy cost is negative. The mode described by the reversed set of Landau level indices was called a negative-energy mode by MacDonald because of its negative kinetic energy cost. . MacDonald considered mixing between positive and negative energy modes as well as between the positive energy modes in second-order perturbation theory in order to compute the collective excitations at general $`r_S`$ in the spin unpolarized ground state. In the present work we directly solve the pole equation, Eq.(69), instead of approximating it to the second-order. In computing the lowest lying mode it is especially important to consider mixing with the negative-energy modes since the lowest positive-energy mode is energetically closest to the negative-energy modes.
In order to make our discussion concrete and transparent, let us explicitly write down the matrix elements in the case of the full spin polarization. If $`n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}m>0`$, $`n_{\alpha }^{}{}_{}{}^{}>\mu _0`$ and $`n_{\beta }^{}{}_{}{}^{}<\mu _0`$, Eq.(64) is written, after the analytic continuation, as follows:
$$D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(\omega )=\left(\omega m\omega _C(\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0})\right).$$
(112)
If the order of the indices is reversed, we get
$`D_{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}^1(\omega )`$ $`=`$ $`\left(\omega +m\omega _C+(\mathrm{\Sigma }_{n_{\alpha }^{}{}_{}{}^{}}^{n_0}\mathrm{\Sigma }_{n_{\beta }^{}{}_{}{}^{}}^{n_0})\right)\times (1)`$ (113)
$`=`$ $`D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(\omega )`$ (114)
Therefore the matrix $`M`$ defined in Eq.(65) and its counterpart with reversed indices are
$$M_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k,\omega )=\delta _{n_\alpha ,n_{\alpha }^{}{}_{}{}^{}}\delta _{n_\beta ,n_{\beta }^{}{}_{}{}^{}}D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(\omega )+V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)$$
(115)
and
$$M_{n_\beta n_\alpha }^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}(k,\omega )=\delta _{n_\alpha ,n_{\alpha }^{}{}_{}{}^{}}\delta _{n_\beta ,n_{\beta }^{}{}_{}{}^{}}D_{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}^1(\omega )+(1)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}\left(V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)\right)$$
(116)
where we set $`\theta _\stackrel{}{k}=0`$ without loss of generality and therefore $`V_{n_\beta n_\alpha }^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}(U_{n_\beta n_\alpha }^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}})=(1)^{n_{\alpha \beta }n_{\alpha \beta }^{}{}_{}{}^{}}V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}})`$.
The sign in front of the second term in Eq.(116) can be interpreted so that the negative-energy mode has the angle of wave vector equal to $`\pi `$: for the negative-energy mode $`\theta _\stackrel{}{k}=\pi `$ in the phase factor of $`V_{n_\beta n_\alpha }^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}`$ and $`U_{n_\beta n_\alpha }^{n_{\beta }^{}{}_{}{}^{}n_{\alpha }^{}{}_{}{}^{}}`$. According to Eq.(116) the negative-energy mode has a negative frequency and the opposite direction of wave vector relative to the positive-energy mode. Therefore when the positive-energy mode is viewed as a plane wave $`e^{i(\omega t\stackrel{}{k}\stackrel{}{r})}`$, the negative-energy mode is the complex conjugate plane wave $`e^{i(\omega t\stackrel{}{k}\stackrel{}{r})}`$. In this interpretation the requirement of negative-energy mode is natural since an arbitrary plane wave with $`\stackrel{}{k}`$ is written as a linear combination of $`e^{i(\omega t\stackrel{}{k}\stackrel{}{r})}`$ and $`e^{i(\omega t\stackrel{}{k}\stackrel{}{r})}`$. Therefore in general a collective excitation with $`\stackrel{}{k}`$ should be described by both the positive-energy and negative-energy modes. Incidentally, we mention that the mode describing an excitation within the same Landau level, for example the spin-wave Goldstone mode, does not have the negative-energy counterpart because there should not be any double counting in Eq.(111).
In any case we realize that the pole equation, Eq.(69), is not an eigenvalue equation as it stands because of the sign change in $`\omega `$ for the negative-energy mode. But it can be transformed to an eigenvalue equation as follows. Let us denote $`M`$ in terms of sub-matrices.
$$M(k,\omega )=\left[\begin{array}{cc}M_{00}(k,\omega )& M_{01}(k)\\ M_{10}(k)& M_{11}(k,\omega )\end{array}\right]$$
(117)
where $`M_{00}(M_{11})`$ is associated with mixing between the positive-energy (negative-energy) modes and $`M_{01}(=M_{10})`$ is between the positive and negative energy modes. Thanks to a property of determinant we can obtain the solution of $`Det[M(k,\omega )]=0`$ by solving the following equation.
$$Det[\stackrel{~}{M}(k,\omega )]=0,$$
(118)
where
$$\stackrel{~}{M}(k,\omega )=\left[\begin{array}{cc}M_{00}(k,\omega )& M_{01}(k)\\ M_{10}(k)& M_{11}(k,\omega )\end{array}\right]$$
(119)
Now one can define *an effective Hamiltonian matrix* $`H(k)`$ using $`\stackrel{~}{M}(k,\omega )`$.
$$\stackrel{~}{M}(k,\omega )=H(k)\omega I$$
(120)
Therefore Eq.(118) amounts to the eigenvalue equation of the effective Hamiltonian matrix $`H`$, which, however, is not a Hermitian matrix because of the sign change.
Solving the eigenvalue equation of a non-Hermitian matrix is complicated by the fact that the eigenvalues of a non-Hermitian matrix can be highly sensitive to small changes in the matrix elements . The sensitivity of eigenvalues to rounding errors during the execution of some algorithms can be reduced by the procedure of *balancing*. The idea of balancing is to use similarity transformations to make corresponding rows and columns of a matrix have comparable norms, thus reducing the overall norm of the matrix while leaving the eigenvalues unchanged. Then the general strategy for finding the eigenvalues of a matrix is to reduce the matrix to a simpler form, and perform an iterative procedure on the simplified matrix. The simpler structure we use is called the Hessenberg form. An upper Hessenberg matrix has zeros everywhere below the diagonal except for the first subdiagonal row. Then one can find the eigenvalues by applying the *QR algorithm* repeatedly to the Hessenberg form until convergence is reached.
Finally, we mention that even though the above formalism has been developed for a general situation, we will concentrate on the physics at $`\nu =2`$ in the following sections taking the Coulomb potential as the interaction. When the finite thickness effect of the 2D system is of interest, one can replace the Coulomb potential by an effective potential such as the Stern-Howard potential .
## III Collective excitations of the fully polarized IQHE state at $`\nu =2`$
Equipped with the explicit formulas for the binding energy, the RPA energy and the self energy, we compute the dispersion curves of the collective excitation from the fully polarized ground state. First we study the large B field limit, i.e. the small $`r_S`$ limit, where the (time-dependent) Hatree-Fock approximation is valid. In the small $`r_S`$ limit the effective Hamiltonian $`H`$ defined in Eq. (120) is already block-diagonalized so that only the matrix elements within the Hilbert subspace of the same kinetic energy survive. The off-diagonal terms due to the interaction energy become negligible compared to the kinetic energy. Therefore the pole equation is simple to solve in this case. For an arbitrary value of $`r_S`$ the Hamiltonian is generally complicated and needs to be diagonalized over whole Hilbert space.
When we consider the small $`r_S`$ limit, the integrals we encounter in Eq.(85), Eq.(87) and Eq.(107) can be expressed in terms of $`f_n(\alpha ,\beta )`$ and $`g_n(\alpha )`$, defined as follows:
$$f_n(\alpha ,\beta )_0^{\mathrm{}}𝑑xx^nJ_0(\beta x)e^{\alpha x^2}$$
(121)
and
$$g_n(\alpha )_0^{\mathrm{}}𝑑xx^ne^{\alpha x^2}.$$
(122)
where $`n`$ is an integer. The explicit functional forms of $`f_n(\alpha ,\beta )`$ and $`g_n(\alpha )`$ are categorized into ones with even $`n`$ or with odd $`n`$:
$`f_{2m}(\alpha ,\beta )`$ $`=`$ $`\left({\displaystyle \frac{}{\alpha }}\right)^mf_0(\alpha ,\beta )`$ (123)
$`f_{2m+1}(\alpha ,\beta )`$ $`=`$ $`\left({\displaystyle \frac{}{\alpha }}\right)^mf_1(\alpha ,\beta )`$ (124)
where
$`f_0(\alpha ,\beta )`$ $`=`$ $`\sqrt{{\displaystyle \frac{\pi }{4\alpha }}}e^{\beta ^2/8\alpha }I_0({\displaystyle \frac{\beta ^2}{8\alpha }}),`$ (125)
$`f_1(\alpha ,\beta )`$ $`=`$ $`{\displaystyle \frac{1}{2\alpha }}e^{\beta ^2/4\alpha }`$ (126)
and $`I_0`$ is a modified Bessel function. Similarly,
$`g_{2m}(\alpha )`$ $`=`$ $`\left({\displaystyle \frac{}{\alpha }}\right)^mg_0(\alpha )`$ (127)
$`g_{2m+1}(\alpha )`$ $`=`$ $`\left({\displaystyle \frac{}{\alpha }}\right)^mg_1(\alpha )`$ (128)
where
$`g_0(\alpha )`$ $`=`$ $`\sqrt{{\displaystyle \frac{\pi }{4\alpha }}},`$ (129)
$`g_1(\alpha )`$ $`=`$ $`{\displaystyle \frac{1}{2\alpha }}`$ (130)
### A Charge density excitations
Since the lowest mode in the energy spectrum is most relevant, we consider the mode where an electron is taken from the $`(n=1)`$ Landau level and promoted to the $`(n=2)`$ Landau level without flipping the spin in order to obtain a charge density excitation. Since its kinetic energy is $`\mathrm{}\omega `$, we will call the mode the $`m=1`$ mode. As discussed before, this mode is not mixed with other modes in the limit of small $`r_S`$, i.e., the matrix $`M`$ is already diagonal. For the lowest charge density excitation we have the following equation:
$`M_{2,1}^{2,1}(k,\omega )=\omega \omega _C{\displaystyle \frac{e^2}{ϵl_0}}(\stackrel{~}{\mathrm{\Sigma }^1}_{n=2}\stackrel{~}{\mathrm{\Sigma }^1}_{n=1}){\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{U}_{2,1}^{2,1}(k)+{\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{V}_{2,1}^{2,1}(k)=0.`$ (131)
In the above equation we defined dimensionless matrix elements so that
$`\stackrel{~}{V}_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $`=`$ $`V_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)/(e^2/ϵl_0)`$ (132)
$`\stackrel{~}{U}_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)`$ $`=`$ $`U_{n_\alpha n_\beta }^{n_{\alpha }^{}{}_{}{}^{}n_{\beta }^{}{}_{}{}^{}}(k)/(e^2/ϵl_0)`$ (133)
$`\stackrel{~}{\mathrm{\Sigma }}_{n_\alpha }^{n_0}`$ $`=`$ $`\mathrm{\Sigma }_{n_\alpha }^{n_0}/(e^2/ϵl_0)`$ (134)
Let us denote the solution of the pole equation as $`\mathrm{\Delta }(k)`$ from now on. Then the dispersion curve of the lowest charge density excitation is given by
$`\mathrm{\Delta }(k)`$ $`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}(\stackrel{~}{\mathrm{\Sigma }}_{n=2}^1\stackrel{~}{\mathrm{\Sigma }}_{n=1}){\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{V}_{2,1}^{2,1}(k)+{\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{U}_{2,1}^{2,1}(k)`$ (135)
$`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}[g_2(\alpha ){\displaystyle \frac{1}{2}}g_4(\alpha )+{\displaystyle \frac{1}{16}}g_6(\alpha )`$ (136)
$``$ $`f_0(\alpha ,\beta )+{\displaystyle \frac{3}{2}}f_2(\alpha ,\beta ){\displaystyle \frac{5}{8}}f_4(\alpha ,\beta )+{\displaystyle \frac{1}{16}}f_6(\alpha ,\beta )]_{\alpha =1/2,\beta =k}`$ (137)
$`+`$ $`{\displaystyle \frac{e^2}{ϵl_0}}e^{k^2/2}k\left(1{\displaystyle \frac{k^2}{4}}\right)^2`$ (138)
As explained earlier, Landau level mixing in the non-zero $`r_S`$ regime is included by diagonalizing the effective Hamiltonian defined in Eq.(120). Using the term $`m=1`$ mode for the sake of convenience to indicate the lowest charge density excitation in the fully polarized state at a general $`r_S`$, we plot their dispersion curves in Fig.4 which shows that the charge density excitation modes do not exhibit any sign of an instability in the parameter range of $`r_S`$ considered here. A non-trivial check of our calculations is to make sure that the collective excitations computed within our approximation satisfy the exact Kohn’s theorem which states that the $`m=1`$ mode energy must approach $`\omega _C`$ as $`k0`$ . Fig.4 shows that Kohn’s theorem is satisfied not only for the pure mode but also the mode with Landau level mixing in general $`r_S`$. Incidentally, the Zeeman coupling will not affect the dispersion curves because the spin is not flipped.
### B Spin density excitations
Following the convention used in the previous sections, we indicate the excitation modes in terms of $`m`$, the kinetic energy of the mode in units of $`\mathrm{}\omega _C`$ in the limit of small $`r_S`$. We shall see that, unlike the charge density excitation, the lowest lying spin excitation can be either $`m=1`$ mode or the lower one of the two $`m=0`$ modes, depending on the value of $`r_S`$. The $`m=1`$ mode describes the process whereby an electron in the $`n=1`$ Landau level is demoted to the $`n=0`$ Landau level with its spin reversed, whereas the $`m=0`$ mode has an electron with its spin flipped in the same Landau level. In the latter case, there are two possible modes: spin density excitation within $`n=1`$ Landau level or $`n=0`$ Landau level. At small $`r_S`$ the $`m=1`$ mode the lowest excitation while for large $`r_S`$ the lowest spin density excitation is a $`m=0`$ mode. Also, without the Zeeman coupling, the spin reversed mode always causes an instability of the fully polarized (2:0) state. A determination of the Zeeman splitting energy $`E_Z`$ required to make the (2:0) state stable for general $`r_S`$ will be one of the main goals when we try to obtain the phase diagram of the spin polarization as a function of $`r_S`$ and $`E_Z`$. The dispersion curves of the pure spin density excitation, however, can be computed in the limit of small $`r_S`$ without recourse to the actual value of $`E_Z`$. We assume that it is large enough to stablize the excitation because the Zeeman energy is just a constant shift in this limit.
$``$ $`m=1`$ mode
As in the case of the charge density excitation, we first solve the pole equation in the small $`r_S`$ limit to get the pure mode without any Landau level mixing. With matrix elements
$$M_{0,1}^{0,1}(k,\omega )=\omega +\omega _C\frac{e^2}{ϵl_0}\left[\stackrel{~}{\mathrm{\Sigma }}_{n=1}^1\stackrel{~}{V}_{0,1}^{0,1}(k)\right]=0$$
(139)
the solution is
$$\mathrm{\Delta }(k)=\omega _C+\frac{e^2}{ϵl_0}\left[\stackrel{~}{\mathrm{\Sigma }}_{n=1}^1\stackrel{~}{V}_{0,1}^{0,1}(k)\right]$$
(140)
where
$`\stackrel{~}{\mathrm{\Sigma }}_{n=1}^1=\left[2g_0(\alpha ){\displaystyle \frac{3}{2}}g_2(\alpha )+{\displaystyle \frac{1}{4}}g_4(\alpha )\right]_{\alpha =1/2}={\displaystyle \frac{5}{4}}\sqrt{{\displaystyle \frac{\pi }{2}}}`$ (141)
and
$`\stackrel{~}{V}_{0,1}^{0,1}(k)`$ $`=`$ $`\left[f_0(\alpha ,\beta )+{\displaystyle \frac{1}{2}}f_2(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}`$ (142)
$`=`$ $`{\displaystyle \frac{1}{2}}\sqrt{{\displaystyle \frac{\pi }{2}}}e^{k^2/4}\left[(1+{\displaystyle \frac{k^2}{2}})I_0({\displaystyle \frac{k^2}{4}}){\displaystyle \frac{k^2}{2}}I_1({\displaystyle \frac{k^2}{4}})\right]`$ (143)
The dispersion curve of this $`m=1`$ mode is plotted in Fig.5.
$``$ $`m=0`$ modes and the spin wave mode
Since there are two possible $`m=0`$ modes, the pole equation becomes a matrix equation even in the small $`r_S`$ limit:
$$\left|\begin{array}{cc}M_{0,0}^{0,0}(k,\omega )& M_{1,1}^{0,0}(k)\\ M_{0,0}^{1,1}(k)& M_{1,1}^{1,1}(k,\omega )\end{array}\right|=0$$
(144)
The matrix elements are given by
$`M_{0,0}^{0,0}(k,\omega )`$ $`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}\left[\stackrel{~}{\mathrm{\Sigma }}_{n=0}^1\stackrel{~}{V}_{0,0}^{0,0}(k)\right]`$ (145)
$`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}\left[2g_0(\alpha ){\displaystyle \frac{1}{2}}g_2(\alpha )f_0(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}`$ (146)
$`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}\sqrt{{\displaystyle \frac{\pi }{2}}}\left[{\displaystyle \frac{3}{2}}e^{k^2/4}I_0({\displaystyle \frac{k^2}{4}})\right],`$ (147)
$`M_{1,1}^{0,0}(k)`$ $`=`$ $`{\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{V}_{1,1}^{0,0}(k)`$ (148)
$`=`$ $`{\displaystyle \frac{e^2}{ϵl_0}}\left[{\displaystyle \frac{1}{2}}f_2(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}`$ (149)
$`=`$ $`{\displaystyle \frac{e^2}{ϵl_0}}{\displaystyle \frac{1}{2}}\sqrt{{\displaystyle \frac{\pi }{2}}}e^{k^2/4}\left[(1{\displaystyle \frac{k^2}{2}})I_0({\displaystyle \frac{k^2}{4}})+{\displaystyle \frac{k^2}{2}}I_1({\displaystyle \frac{k^2}{4}})\right]`$ (150)
$`=`$ $`M_{0,0}^{1,1}(k),`$ (151)
and (152)
$`M_{1,1}^{1,1}(k,\omega )`$ $`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}\left[\stackrel{~}{\mathrm{\Sigma }}_{n=1}^1\stackrel{~}{V}_{0,0}^{1,1}(k)\right]`$ (153)
$`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}\left[2g_0(\alpha ){\displaystyle \frac{3}{2}}g_2(\alpha )+{\displaystyle \frac{1}{4}}g_4(\alpha )f_0(\alpha ,\beta )+f_2(\alpha ,\beta ){\displaystyle \frac{1}{4}}f_4(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}`$ (154)
$`=`$ $`\omega {\displaystyle \frac{e^2}{ϵl_0}}{\displaystyle \frac{5}{4}}\sqrt{{\displaystyle \frac{\pi }{2}}}`$ (155)
$`+`$ $`{\displaystyle \frac{e^2}{ϵl_0}}{\displaystyle \frac{1}{4}}\sqrt{{\displaystyle \frac{\pi }{2}}}e^{k^2/4}\left[(3k^2+{\displaystyle \frac{3}{8}}k^4)I_0({\displaystyle \frac{k^2}{4}})+k^2(1{\displaystyle \frac{k^2}{2}})I_1({\displaystyle \frac{k^2}{4}})+{\displaystyle \frac{k^4}{8}}I_2({\displaystyle \frac{k^2}{4}})\right]`$ (156)
Since spontaneous symmetry breaking occurs in the fully polarized state, there must be a spin-wave Goldstone mode whose energy approaches the unshifted Zeeman splitting energy in the long wavelength limit. Similar to the case of the charge density excitation it is important to check if the spin density excitations computed within our approximation satisfy this exact theorem. When we take the $`k0`$ limit of Eq.(144), we have the following equation:
$$\left|\begin{array}{cc}\omega \frac{e^2}{ϵl_0}\frac{1}{2}\sqrt{\frac{\pi }{2}}& \frac{e^2}{ϵl_0}\frac{1}{2}\sqrt{\frac{\pi }{2}}\\ \frac{e^2}{ϵl_0}\frac{1}{2}\sqrt{\frac{\pi }{2}}& \omega \frac{e^2}{ϵl_0}\frac{1}{2}\sqrt{\frac{\pi }{2}}\end{array}\right|=0.$$
(157)
The solutions are
$`\mathrm{\Delta }_1(k=0)`$ $`=`$ $`0`$ (158)
$`\mathrm{\Delta }_2(k=0)`$ $`=`$ $`{\displaystyle \frac{e^2}{ϵl_0}}\sqrt{{\displaystyle \frac{\pi }{2}}}.`$ (159)
This confirms the existence of a Goldstone (spin-wave) mode. Furthermore it can be shown that the massless spin-wave mode exists with an arbitrary potential and Landau level mixing within our approximation. The dispersion curves of the two $`m=0`$ modes are also plotted in Fig.5 along with that of $`m=1`$ mode.
As before, the dispersion curves for general $`r_S`$ and $`E_Z`$ are obtained from diagonalization of the effective Hamiltonian. We take the dispersion curves at $`r_S=`$ 1.0 and 3.0 as examples and plot them in Fig.6 to illustrate the qualitative difference between the small and large $`r_S`$ regimes. We will sometimes use the term Goldstone mode in order to indicate spin-wave excitation mode for general $`r_S`$ since its energy approaches $`E_Z`$ as $`k`$ goes to zero. We will also use the term $`m=1`$ mode for the lowest excitation for small $`r_S`$ since it has the lowest energy in the limit of vanishing $`r_S`$. The physical implications and the corresponding phase diagram will be discussed in more details in the later section.
## IV Collective excitations from the unpolarized IQHE state at $`\nu =2`$
The collective excitations in the spin unpolarized state are computed in this section in complete analogy with the previous section. First, let us consider the small $`r_S`$ limit with zero $`E_Z`$. The energy of the lowest charge density excitation is
$`\mathrm{\Delta }(k)`$ $`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}[\stackrel{~}{\mathrm{\Sigma }}_{n=1}^0\stackrel{~}{\mathrm{\Sigma }}_{n=0}^0]{\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{V}_{1,0}^{1,0}(k)+{\displaystyle \frac{e^2}{ϵl_0}}2\stackrel{~}{U}_{1,0}^{1,0}(k)`$ (160)
$`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}ke^{k^2/4}+{\displaystyle \frac{e^2}{ϵl_0}}\left[g_0(\alpha ){\displaystyle \frac{1}{2}}g_2(\alpha )f_0(\alpha ,\beta )+{\displaystyle \frac{1}{2}}f_2(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}.`$ (161)
And the energy of the lowest spin density exciation is
$`\mathrm{\Delta }(k)`$ $`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}[\stackrel{~}{\mathrm{\Sigma }}_{n=1}^0\stackrel{~}{\mathrm{\Sigma }}_{n=0}^0]{\displaystyle \frac{e^2}{ϵl_0}}\stackrel{~}{V}_{1,0}^{1,0}(k)`$ (162)
$`=`$ $`\omega _C+{\displaystyle \frac{e^2}{ϵl_0}}\left[g_0(\alpha ){\displaystyle \frac{1}{2}}g_2(\alpha )f_0(\alpha ,\beta )+{\displaystyle \frac{1}{2}}f_2(\alpha ,\beta )\right]_{\alpha =1/2,\beta =k}.`$ (163)
The dispersion curves of the above pure modes are plotted in Fig.7. We can use Kohn’s theorem to check if our approximation is reasonable. Fig.7 shows that the energy of charge density excitation approach $`\omega _C`$ as $`k0`$. Since the charge density excitation does not show any instability, we will concentrate only on the spin density excitation to obtain the phase boundary which is the critical Zeeman splitting energy needed for the stable excitation. Fig.8 shows the dispersion curves of the lowest spin density excitation for $`r_S=1.0`$ and various values of $`E_Z`$.
## V The phase diagram
The phase diagram of the state at $`\nu =2`$ can be obtained at two levels of sophistication.
### A Phase diagram in Hartree Fock approximation
The simplest approximation is that of non-interacting electrons. In this case, the phase boundary is given simply by $`E_Z=\mathrm{}\omega _C`$, as shown in Fig.9. As we shall see, this is sensible only in the limit of $`r_S0`$; interactions modify the phase diagram substantially elsewhere.
In the simplest approximation, interaction can be incorporated by comparing the energies of the fully polarized and unpolarized ground states in the Hatree-Fock approximation, that is, by assuming that the ground state contains either 0$``$ and 0$``$ Landau levels fully occupied or 0$``$ and 1$``$. The contributions of the kinetic energy and the Zeeman coupling to the ground state energy are straightforward. The exchange interaction energy in the Hartree-Fock approximation can be evaluated in terms of the self-energy defined in the previous section. The self-energy $`\mathrm{\Sigma }_n^{n(\sigma )}`$ is the exchange interaction between an electron in the Landau level with index $`n`$ and all the other electrons with the same spin. The exchange interaction energy per particle is then the sum of the self-energies for all electrons divided by two times the number of electrons where the factor of two prevents a double counting. That is to say,
$$V_{ex}=\frac{1}{2\nu }\underset{\sigma =1/2}{\overset{1/2}{}}\left(\underset{n^{}=0}{\overset{n(\sigma )}{}}\mathrm{\Sigma }_n^{}^{n(\sigma )}\right).$$
(164)
Therefore the ground state energy per particle is
$$E_g=\frac{1}{\nu }\underset{\sigma =1/2}{\overset{1/2}{}}\underset{n^{}=0}{\overset{n(\sigma )}{}}\left[(n^{}+\frac{1}{2})\mathrm{}\omega _C+\sigma E_Z+\frac{1}{2}\mathrm{\Sigma }_n^{}^{n(\sigma )}\right].$$
(165)
The average energy of the fully polarized ground state is
$`{\displaystyle \frac{E_g(2:0)}{\mathrm{}\omega _C}}{\displaystyle \frac{1}{2}}={\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{2}}{\displaystyle \frac{E_Z}{\mathrm{}\omega _C}}+{\displaystyle \frac{1}{4}}(\stackrel{~}{\mathrm{\Sigma }}_{n=0}^1+\stackrel{~}{\mathrm{\Sigma }}_{n=1}^1)r_S={\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{2}}{\displaystyle \frac{E_Z}{\mathrm{}\omega _C}}{\displaystyle \frac{11}{16}}\sqrt{{\displaystyle \frac{\pi }{2}}}r_S.`$ (166)
Similarly the average energy of the unpolarized ground state is
$`{\displaystyle \frac{E_g(1:1)}{\mathrm{}\omega _C}}{\displaystyle \frac{1}{2}}={\displaystyle \frac{1}{2}}\stackrel{~}{\mathrm{\Sigma }}_{n=0}^0r_S={\displaystyle \frac{1}{2}}\sqrt{{\displaystyle \frac{\pi }{2}}}r_S.`$ (167)
The phase boundary is given by the solution of the following equation:
$`{\displaystyle \frac{E_g(2:0)E_g(1:1)}{\mathrm{}\omega _C}}=0.`$ (168)
Therefore the critical $`E_Z/\mathrm{}\omega _C`$ as a function of $`r_S`$ is
$`{\displaystyle \frac{E_Z}{\mathrm{}\omega _C}}=1{\displaystyle \frac{3}{8}}\sqrt{{\displaystyle \frac{\pi }{2}}}r_S`$ (169)
shown in Fig.9.
### B Phase diagram from collective mode instability
The phase diagram obtained by a comparison of the energies of the Hartree Fock states is not fully reliable for two reasons. First, it neglects Landau level mixing, which is crucial for the issue of interest here. Secondly, it does not allow for the possibility of other states in the phase diagram. Therefore, it is more appropriate to look for instabilities of the two states by asking when one of the collective modes becomes soft. Indeed, a first order phase transition may (and likely will) occur even before the collective mode energy approaches zero, but we believe that the phase diagram obtained by considering instabilities ought to be reliable qualitatively and even semi-quantitatively.
As noted previously, there is no instability in the charge density wave collective mode in the parameter range considered here. We have determined the onset of the spin density collective mode instability both in the fully polarized and the unpolarized states by varying $`r_S`$ and $`E_Z`$, as shown in Fig.6 and Fig.8. Fig.9 shows the phase diagram thus obtained. The following features are noteworthy.
(i) The nature of instability is different depending on whether $`r_S`$ is small or large. At small $`r_S`$, where the interactions are negligible and the physics is dictated by the Zeeman energy, the lowest energy spin-density excitation is clearly the $`m=1`$ mode. This continues to be the case for $`r_S2`$; here the $`m=1`$ spin-density mode is responsible for the instability of the (2:0) state, as shown in Fig.6. However, for $`r_S2`$, interactions are sufficiently strong that the spin-wave mode becomes the lowest energy mode and causes the instability. One way to understand why the $`m=0`$ spin wave mode has lower energy at large $`r_S`$ than the $`m=1`$ mode is because whereas the former has an energy equal to $`E_Z`$ in the long wave length limit no matter what $`r_S`$, as guaranteed by the Goldstone theorem, the energy of the latter is determined by the interactions.
(ii) At small $`r_S`$, the interactions make the fully polarized (2:0) state more stable as compared to the non-interacting problem, as evidenced by the fact that the transition out of it takes place at a Zeeman energy smaller than $`\mathrm{}\omega _C`$. This is precisely as expected from the exchange physics, as also captured by the Hartree-Fock phase diagram.
(iii) The instability in the spin-wave mode occurs through the development of a roton minimum, the energy of which vanishes at certain Zeeman energy. The roton minimum in turn is caused by the Landau level mixing, underscoring the important role of Landau level mixing at large $`r_S`$. Without Landau level mixing the spin-wave mode does not show any instability as shown in Fig.5.
(iv) For the unpolarized state, the lowest spin density excitation is the $`m=1`$ mode at all $`r_S`$, the long wavelength limit of the excitation energy of which is fixed to be $`\mathrm{}\omega _CE_Z`$ independent of $`r_S`$. Here also, the instability occurs through a roton, which becomes deeper as the $`r_S`$ is increased (and interactions become stronger). Since a spin flip is favored by exchange, the energy of the $`m=1`$ mode decreases with increasing $`r_S`$, consistent with the feature that the critical $`E_Z`$ in this case is monotonically decreasing as a function of $`r_S`$, as is shown in Fig.2.
(v) At small $`r_S`$ there is a small region in Fig.2 where the two phases coexist. This clearly is an artifact of various approximations involved in our calculation, and implies that the actual locations of the phase boundaries cannot be taken too seriously. As mentioned earlier, a first order transition is likely to occur before the roton energy vanishes.
(vi) At large $`r_S`$ and small $`E_Z`$, the ground state is derived neither from (2:0) nor from (1:1). Since we find a finite wave vector instability in the spin density wave excitation, it is natural to expect that the state here has a spin-density wave state. Further work will be required to establish the nature of this state in more detail.
## VI Conclusion
The principal outcome of our calculations is the phase diagram in Fig.9 which shows the regions where the fully polarized and the unpolarized Hatree Fock states (2:0) and (1:1) are valid. We believe that it should be possible to investigate the roton minimum in the spin-wave excitation of the fully polarized state as well as its instability in inelastic light scattering experiments.
Another situation where similar physics may apply is in the case of composite fermions at effective filling factor $`\nu ^{}=2`$, which corresponds to the electron filling factor $`\nu =2/5`$. It is easier in this case to see a transition between the fully polarized and the unpolarized states because the effective cyclotron energy for composite fermions is substantially small compared to the cyclotron energy of electrons, which makes it possible to obtain $`E_Z`$ comparable to or larger than the effective cyclotron energy in tilted field experiments. The critical Zeeman energy at the transition was calculated by Park and Jain by comparing the ground state energies , in reasonable agreement with the experiments of Kukushkin, von Klitzing, and Eberl . Park and Jain also estimated a mass for composite fermions, the “polarization mass” by equating the critical Zeeman energy to the effective cyclotron energy. There is one subtlety though. In the case of composite fermions both the effective cyclotron energy and the effective interactions derive from the same underlying energy, namely the Coulomb interaction between the electrons, and therefore neither the interactions between composite fermions nor a mixing between composite fermion Landau levels can, in principle, be neglected in any realistic limit. These would provide a correction to the mass obtained in Ref.. However, we note that the mass was reliable to no more than 20-30%, and the corrections may be negligible compared to that.
Interestingly, there is experimental evidence that the transition between the fully polarized and the unpolarized composite fermion states does not occur directly but through an intermediate state with a partial spin polarization. Murthy has proposed that this state is a Hofstadter lattice of composite fermions, and has half the maximum possible polarization. It would be interesting to see if similar physics obtains for $`\nu =2`$ as well. In particular, the phase diagram of Fig.9 predicts that for $`r_S1`$, the transition from the fully polarized state to the unpolarized state as a function of the Zeeman energy is not direct but through another, not yet fully identified state. (We suspect that this may be true at any arbitrary $`r_S`$, although not captured by our calculated phase diagram.)
This work was supported in part by the National Science Foundation under grant no. DMR-9986806. We thank G. Murthy for discussions. |
warning/0001/nucl-th0001048.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The properties of hadrons in hot and/or dense nuclear matter are of central interest for the nuclear physics community as one expects to learn about precurser effects for chiral symmetry restoration or to explore the vicinity of a quark-gluon plasma (QGP) phase transition. Whereas the early ’big bang’ most likely evolved through equilibrium configurations from the QGP to a hadronic phase, this is not the case for the hot/dense systems produced in collisions of heavy ions at relativistic energies.
Nowadays, the dynamical description of strongly interacting systems out of equilibrium is dominantly based on transport theories and efficient numerical recipies have been set up for the solution of the coupled channel transport equations (and Refs. therein). These transport approaches have been derived either from the Kadanoff-Baym equations or from the hierarchy of connected equal-time Green functions by applying a Wigner transformation and restricting to first order in the derivatives of the phase-space variables ($`X,P`$). Whereas theoretical formulations of off-shell quantum transport have been limited to the formal level for a couple of years only recently a tractable semiclassical form has been derived for testparticles in the eight dimensional phase-space of a particle .
In this contribution a brief review is given in Section 2 on the present understanding of ’low mass’ dilepton data from $`pp`$ to $`AA`$ collisions and the necessity for a quantum transport description is pointed out. A short reminder of the steps for a derivation of such transport theories is presented in Section 3 as well as generalized testparticle equations of motion that are extracted in the semiclassical limit. Section 4 is devoted to a presentation of the most important off-shell effects in nucleus-nucleus collisions from GANIL to AGS energies.
## 2 Low mass dileptons
Since the first dilepton studies from nucleus-nucleus collisions at the BEVALAC by the DLS Collaboration the field of dilepton measurements has rapidly expanded at CERN/SPS (see Refs. ); the new detector HADES will complement the experimental programm at the SIS . The data taken by the DLS Collaboration on the elementary $`pp`$ reaction can reasonably be well described by the production of intermediate baryonic resonances as well as $`\pi ^0,\eta ,\omega ,\rho ^0`$ and $`\varphi `$ mesons and their Dalitz or direct decays to $`e^+e^{}`$ pairs. This has been demonstrated in detail in Refs. . A similar statement holds true for the $`p+Be`$ and $`p+W`$ reactions at 450 GeV and 200 GeV , respectively, which are fully described by the meson Dalitz or direct decays to dileptons since baryon resonance decays play no longer a substantial role at SPS energies. Whereas the $`e^+e^{}`$ and $`\mu ^+\mu ^{}`$ differential spectra of the CERES and HELIOS-3 Collaborations – that show an excess of dileptons in the invariant mass regime 0.3 GeV $`M`$ 0.6 GeV – can be described within the ’dropping mass’ scenario or the ’melting’ $`\rho `$-meson scenario – that involves a strong broadening of the $`\rho `$ spectral function in the medium due to the coupling to dressed pions and resonance-hole loops – the latter concepts seem to work no longer for the DLS data at 1 A GeV as worked out by Bratkovskaya et al. . Here, the $`e^+e^{}`$ invariant mass spectra for $`Ca+Ca`$ are underestimated in the regime 0.2 GeV $`M`$ 0.6 GeV by a factor of 6–7 when involving ’vacuum’ spectral functions for the $`\rho `$ and $`\omega `$ mesons and by a factor of 2–3 within the ’dropping mass’ and ’melting’ $`\rho `$ scenarios (DLS-puzzle). In part this discrepancy might be attributed to an improper ’use’ of the vector-dominance-model (VDM) in elementary reactions or unknown isospin dependencies in reactions involving neutrons as pointed out by Mosel at this workshop and should be examined experimentally first in $`\gamma p`$ or $`\pi ^{}p`$ and $`\pi ^{}d`$ reactions . These elementary reactions also are expected to provide valuable insight into $`\rho /\omega `$ mixing and their individual production processes .
A general survey on low mass dilepton production in $`Au+Au`$ collisions from SIS to RHIC energies has been given in Refs. ; here the energy regime from 2 – 10 A GeV is found to be most promising for studies of the vector meson spectral functions at high baryon density since the systems show high density regimes above $`23`$ $`\rho _0`$ for more than 10 fm/c which are large compared to the $`\rho `$ and $`\omega `$ life times in the medium. Furthermore, at a couple of A GeV the scalar quark condensate $`<\overline{q}q>`$ is expected to approach zero, i.e. a chirally restored phase, for substantial space-time volumes . At SPS energies and above most vector mesons are produced dominantly by meson-meson channels in the longitudinally expanding fireball at rather low baryon density (but high temperature) . This lowers the perspectives of low mass dilepton measurements at RHIC energies; however, intermediate mass dileptons from 1.2 – 2.5 GeV of invariant mass might show a signal from the QGP phase provided that open charm is strongly suppressed when propagating through a colored QGP medium .
In view of these more general considerations an accurate understanding of the vector meson properties in the medium – or the imaginary part of the current-current correlation functions – can only be achieved if the full dynamical evolution of the hadron spectral functions – also for nonequilibrium phase-space configurations – can be followed throughout all stages of a nucleus-nucleus collision. The next Section is devoted to a formulation of such type of off-shell transport theory following Refs. .
## 3 Extended semiclassical transport equations
The general starting point for the derivation of a transport equation for particles with a finite and dynamical width are the Dyson-Schwinger equations for the retarded and advanced Green functions $`S^{ret}`$, $`S^{adv}`$ and for the non-ordered Green functions $`S^<`$ and $`S^>`$ . In the case of scalar bosons – which is considered in the following for simplicity – these Green functions are defined by
$$iS_{xy}^<:=<\mathrm{\Phi }^{}(y)\mathrm{\Phi }(x)>,iS_{xy}^>:=<\mathrm{\Phi }(x)\mathrm{\Phi }^{}(y)>,$$
(1)
$$iS_{xy}^{ret}:=\mathrm{\Theta }(x_0y_0)<[\mathrm{\Phi }(x),\mathrm{\Phi }^{}(y)]>,iS_{xy}^{adv}:=\mathrm{\Theta }(y_0x_0)<[\mathrm{\Phi }(x),\mathrm{\Phi }^{}(y)]>.$$
They depend on the space-time coordinates $`x,y`$ as indicated by the indices $`_{xy}`$. The Green functions are determined via Dyson-Schwinger equations by the retarded and advanced self energies $`\mathrm{\Sigma }^{ret},\mathrm{\Sigma }^{adv}`$ and the collisional self energy $`\mathrm{\Sigma }^<`$:
$`\widehat{S}_{0x}^1S_{xy}^{ret}=\delta _{xy}+\mathrm{\Sigma }_{xz}^{ret}S_{zy}^{ret},\widehat{S}_{0x}^1S_{xy}^{adv}=\delta _{xy}+\mathrm{\Sigma }_{xz}^{adv}S_{zy}^{adv},`$ (2)
$`\widehat{S}_{0x}^1S_{xy}^<=\mathrm{\Sigma }_{xz}^{ret}S_{zy}^<+\mathrm{\Sigma }_{xz}^<S_{zy}^{adv},`$ (3)
where Eq. (3) is the well-known Kadanoff-Baym equation. Here $`\widehat{S}_{0x}^1`$ denotes the (negative) Klein-Gordon differential operator which is given for bosonic field quanta of (bare) mass $`M_0`$ by $`\widehat{S}_{0x}^1=(_x^\mu _\mu ^x+M_0^2)`$; $`\delta _{xy}`$ represents the four-dimensional $`\delta `$-distribution $`\delta _{xy}\delta ^{(4)}(xy)`$ and the symbol $``$ indicates an integration (from $`\mathrm{}`$ to $`\mathrm{}`$) over all common intermediate variables (cf. ).
For the derivation of a semiclassical transport equation one now changes from a pure space-time formulation into the Wigner-representation with variable $`X=(x+y)/2`$ and the four-momentum $`P`$, which is introduced by Fourier-transformation with respect to the relative space-time coordinate $`(xy)`$. In any semiclassical transport theory one, furthermore, keeps only contributions up to the first order in the space-time gradients. To simplify notation the operator $`\mathrm{}`$ is introduced as
$`\mathrm{}\{F_1\}\{F_2\}:={\displaystyle \frac{1}{2}}\left({\displaystyle \frac{F_1}{X^\mu }}{\displaystyle \frac{F_2}{P_\mu }}{\displaystyle \frac{F_1}{P_\mu }}{\displaystyle \frac{F_2}{X^\mu }}\right),`$ (4)
which is a four-dimensional generalization of the well-known Poisson-bracket. In first order gradient expansion one then obtains algebraic relations between the real and the imaginary part of the retarded Green functions. On the other hand Eq. (3) leads to a ’transport equation’ for the Green function $`S^<`$. In order to obtain real quantities one separates all retarded and advanced functions – Green functions and self energies – into real and imaginary parts,
$`S_{XP}^{ret,adv}=ReS_{XP}^{ret}{\displaystyle \frac{i}{2}}A_{XP},\mathrm{\Sigma }_{XP}^{ret,adv}=Re\mathrm{\Sigma }_{XP}^{ret}{\displaystyle \frac{i}{2}}\mathrm{\Gamma }_{XP}.`$ (5)
The imaginary part of the retarded propagator is given (up to a factor 2) by the normalized spectral function
$`A_{XP}=i\left[S_{XP}^{ret}S_{XP}^{adv}\right]=2ImS_{XP}^{ret},{\displaystyle \frac{dP_0^2}{4\pi }A_{XP}}=\mathrm{\hspace{0.33em}1},`$ (6)
while the imaginary part of the self energy corresponds to half the width $`\mathrm{\Gamma }_{XP}`$. By separating the complex equations into their real and imaginary contributions one obtains an algebraic equation between the real and the imaginary part of $`S^{ret}`$ (Eq. (12) of ), an algebraic solution for the spectral function (in first order gradient expansion) as
$`A_{XP}={\displaystyle \frac{\mathrm{\Gamma }_{XP}}{(P^2M_0^2Re\mathrm{\Sigma }_{XP}^{ret})^2+\mathrm{\Gamma }_{XP}^2/4}},`$ (7)
as well as for the real part of the retarded propagator .
The (Wigner-transformed) Kadanoff-Baym equation (3) allows for the construction of a transport equation for the Green function $`S^<`$. When separating the real and the imaginary contribution of this equation one obtains i) a generalized transport equation (Eq. (15) in ) and ii) a generalized mass-shell constraint (Eq. (16) in ). Furthermore, according to Botermans and Malfliet , in the transport equation the collisional self energy $`\mathrm{\Sigma }^<`$ has to be replaced by $`S^<\mathrm{\Gamma }/A`$ to gain a consistent first order gradient expansion scheme. Finally, the general transport equation (in first order gradient expansion) reads
$$A_{XP}\mathrm{\Gamma }_{XP}[\mathrm{}\{P^2M_0^2Re\mathrm{\Sigma }_{XP}^{ret}\}\{S_{XP}^<\}$$
$$\frac{1}{\mathrm{\Gamma }_{XP}}\mathrm{}\{\mathrm{\Gamma }_{XP}\}\{(P^2M_0^2Re\mathrm{\Sigma }_{XP}^{ret})S_{XP}^<\}]=i[\mathrm{\Sigma }_{XP}^>S_{XP}^<\mathrm{\Sigma }_{XP}^<S_{XP}^>].$$
(8)
It has also been independently derived by Ivanov et al. and Leupold . Its formal structure is fixed by the approximations applied. Note, however, that the dynamics is fully determined by the different self energies, i.e. $`Re\mathrm{\Sigma }_{XP}^{ret},\mathrm{\Gamma }_{XP},\mathrm{\Sigma }_{XP}^<`$ and $`\mathrm{\Sigma }_{XP}^>`$ that have to be specified for the physical systems of interest.
Besides the drift term (i.e. $`\mathrm{}\{P^2M_0^2\}\{S^<\}=P^\mu _\mu ^XS^<)`$ and the Vlasov term (i.e. $`\mathrm{}\{Re\mathrm{\Sigma }^{ret}\}\{S^<\}`$) a third contribution appears on the l.h.s. of (8) (i.e. $`\mathrm{}\{\mathrm{\Gamma }_{XP}\}\{\mathrm{}S_{XP}^<\}`$), which vanishes in the quasiparticle limit and incorporates – as shown in – the off-shell behaviour in the particle propagation that has been neglected so far in transport studies<sup>3</sup><sup>3</sup>3This also holds true for the recent numerical studies in Ref. . The r.h.s. of (8) consists of a collision term with its characteristic gain ($`\mathrm{\Sigma }^<S^>`$) and loss ($`\mathrm{\Sigma }^>S^<`$) structure, where scattering processes of particles into and out of a given phase-space cell are described.
### 3.1 Testparticle approximation
In order to obtain an approximate solution to the transport equation (8) a testparticle ansatz is used for the Green function $`S^<`$, more specifically for the real and positive semidefinite quantity $`F_{XP}=A_{XP}N_{XP}=iS_{XP}^<`$,
$`F_{XP}{\displaystyle \underset{i=1}{\overset{N}{}}}\delta ^{(3)}(\stackrel{}{X}\stackrel{}{X}_i(t))\delta ^{(3)}(\stackrel{}{P}\stackrel{}{P}_i(t))\delta (P_0ϵ_i(t)).`$ (9)
In the most general case (where the self energies depend on four-momentum $`P`$, time $`t`$ and the spatial coordinates $`\stackrel{}{X}`$) the equations of motion for the testparticles read
$$\frac{d\stackrel{}{X}_i}{dt}=\frac{1}{1C_{(i)}}\frac{1}{2ϵ_i}\left[\mathrm{\hspace{0.17em}2}\stackrel{}{P}_i+\stackrel{}{}_{P_i}Re\mathrm{\Sigma }_{(i)}^{ret}+\frac{ϵ_i^2\stackrel{}{P}_i^2M_0^2Re\mathrm{\Sigma }_{(i)}^{ret}}{\mathrm{\Gamma }_{(i)}}\stackrel{}{}_{P_i}\mathrm{\Gamma }_{(i)}\right],$$
(10)
$$\frac{d\stackrel{}{P}_i}{dt}=\frac{1}{1C_{(i)}}\frac{1}{2ϵ_i}\left[\stackrel{}{}_{X_i}Re\mathrm{\Sigma }_i^{ret}+\frac{ϵ_i^2\stackrel{}{P}_i^2M_0^2Re\mathrm{\Sigma }_{(i)}^{ret}}{\mathrm{\Gamma }_{(i)}}\stackrel{}{}_{X_i}\mathrm{\Gamma }_{(i)}\right],$$
(11)
$$\frac{dϵ_i}{dt}=\frac{1}{1C_{(i)}}\frac{1}{2ϵ_i}\left[\frac{Re\mathrm{\Sigma }_{(i)}^{ret}}{t}+\frac{ϵ_i^2\stackrel{}{P}_i^2M_0^2Re\mathrm{\Sigma }_{(i)}^{ret}}{\mathrm{\Gamma }_{(i)}}\frac{\mathrm{\Gamma }_{(i)}}{t}\right],$$
(12)
where the notation $`F_{(i)}`$ implies that the function is taken at the coordinates of the testparticle, i.e. $`F_{(i)}F(t,\stackrel{}{X}_i(t),\stackrel{}{P}_i(t),ϵ_i(t))`$. These equations also have been derived by Leupold in the nonrelativistic limit recently .
In (10-12) a common multiplication factor $`(1C_{(i)})^1`$ appears, which contains the energy derivatives of the retarded self energy
$`C_{(i)}={\displaystyle \frac{1}{2ϵ_i}}\left[{\displaystyle \frac{}{ϵ_i}}Re\mathrm{\Sigma }_{(i)}^{ret}+{\displaystyle \frac{ϵ_i^2\stackrel{}{P}_i^2M_0^2Re\mathrm{\Sigma }_{(i)}^{ret}}{\mathrm{\Gamma }_{(i)}}}{\displaystyle \frac{}{ϵ_i}}\mathrm{\Gamma }_{(i)}\right]`$ (13)
and yields a shift of the system time $`t`$ to the ’eigentime’ of particle $`i`$ defined by $`\stackrel{~}{t}_i=t/(1C_{(i)})`$ . As the reader immediately verifies, the derivatives with respect to the ’eigentime’, i.e. $`d\stackrel{}{X}_i/d\stackrel{~}{t}_i`$, $`d\stackrel{}{P}_i/d\stackrel{~}{t}_i`$ and $`dϵ_i/d\stackrel{~}{t}_i`$ then emerge without this renormalization factor for each testparticle $`i`$ when neglecting higher order time derivatives in line with the semiclassical approximation scheme.
For momentum-independent self energies one regains the transport equations as derived in . Furthermore, in the limiting case of particles with vanishing gradients of the width $`\mathrm{\Gamma }_{XP}`$ these equations of motion reduce to the well-known transport equations of the quasiparticle picture.
Furthermore, the variable $`M^2=P^2Re\mathrm{\Sigma }^{ret}`$ is taken as an independent variable instead of $`P_0`$. Eq. (12) then turns to
$`{\displaystyle \frac{dM_i^2}{dt}}={\displaystyle \frac{M_i^2M_0^2}{\mathrm{\Gamma }_{(i)}}}{\displaystyle \frac{d\mathrm{\Gamma }_{(i)}}{dt}}`$ (14)
for the time evolution of the testparticle $`i`$ in the invariant mass squared .
### 3.2 Collision terms
The collision term of the Kadanoff-Baym equation can only be worked out by giving explicit approximations for $`\mathrm{\Sigma }^<`$ and $`\mathrm{\Sigma }^>`$. As in the conventional transport theory the formulation of collision terms is based on Dirac-Brueckner theory for the transition amplitudes. The formal structure resembles somewhat the on-shell formulations , however, includes additionally the spectral functions of all hadrons in the initial and final channels and four-momentum integrations instead of three-momentum integrals . Furthermore, now all off-shell transition amplitudes enter that so far are scarcely known for strong interaction processes. The collisional width $`\mathrm{\Gamma }_{coll}(X,\stackrel{}{P},M^2)`$ for each hadron then is defined via the loss-term of the corresponding collision integral .
For the numerical studies mentioned below the off-shell transition amplitudes squared have been taken as the on-shell values employed in the conventional HSD approach , however, corrected by the final phase-space for the particles with off-shell masses. For a detailed description of the numerical recipies and approximations employed the reader is refered to Refs. .
## 4 Off-shell effects in nucleus-nucleus collisions
The off-shell propagation of hadrons in nucleus-nucleus collisions has been examined from GANIL to AGS energies in Refs. and for infinite nuclear matter problems in Ref. . The main results are briefly summarized:
* The numerical implementations of off-shell dynamics can be shown to lead to the proper equilibrium mass distributions (for $`t\mathrm{}`$) in line with the quantum statistical limit when employing the same equilibrium spectral functions for all hadrons .
* Numerical studies of systems in a finite box with periodic boundary conditions show that the off-shell dynamics lead to practically the same equilibration times as the on-shell dynamics .
* At GANIL energies ($``$ 95 A GeV) the off-shell effects (OSE) are most pronounced since the excitation of $`\mathrm{\Delta }`$-resonances is strongly suppressed. OSE show up especially in the high momentum tails of the collisional $`\sqrt{s}`$ distribution, in high momentum proton spectra and most pronounced in high energy photon production as demonstrated in comparison to the $`\gamma `$ spectra from the TAPS collaboration for $`Ar+Au`$ at 95 A GeV .
* At SIS energies (1–2 A GeV) the OSE lead only to a slight enhancement of the high transverse momentum spectra since the high mass tail of the nucleon spectral function remains small compared to the $`\mathrm{\Delta }`$ mass distribution . The latter again is only marginally enhanced at high masses relative to the on-shell dynamics which leads to almost the same pion spectra in this energy domain. It’s worth noting that pions in the high density phase become somewhat harder, which leads to an enhanced production of kaons and antikaons. The enhancement of $`K^{}`$ for $`Ni+Ni`$ reactions at 1.8 A GeV amounts to a factor of about 2 such that in-medium antikaon potentials should be substantially less attractive than suggested in .
* The ’DLS puzzle’ mentioned in Section 2 might be solved within the off-shell dynamics once there is a strong coupling of low-mass $`\rho ^0`$ mesons to baryons at high baryon density, which approximately ’thermalizes’ the $`\rho `$ mass distribution. Since this is presently a speculation, no definite conclusions can be drawn so far.
* At AGS energies of 11 A GeV the particle production is dominated by ’string’ continuum excitations, which themselves are similar to very broad hadron spectral functions. Since all productions thresholds for $`K,\overline{K},\rho ,\omega ,\varphi `$ are by far exceeded in the initial collisions, no substantial effects from the off-shell dynamics could be established within the numerical accuracy on rapidity distributions and transverse mass spectra.
In closing it is necessary to point out that although the general equations for off-shell transport are available by now and efficient numerical solution schemes exist, the final understanding of this approach requires the knowledge of all off-shell (in-medium) transition amplitudes! The present recipe of using final state phase-space corrections is only a first step on this way. |
warning/0001/hep-ph0001253.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Among the central targets of the $`B`$-factories is a measurement of the time-dependent CP asymmetry of the decay $`B_d\pi ^+\pi ^{}`$ , which can be expressed as follows:
$`a_{\mathrm{CP}}(B_d(t)\pi ^+\pi ^{}){\displaystyle \frac{\text{BR}(B_d^0(t)\pi ^+\pi ^{})\text{BR}(\overline{B_d^0}(t)\pi ^+\pi ^{})}{\text{BR}(B_d^0(t)\pi ^+\pi ^{})+\text{BR}(\overline{B_d^0}(t)\pi ^+\pi ^{})}}`$ (1)
$`=𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})\mathrm{cos}(\mathrm{\Delta }M_dt)+𝒜_{\mathrm{CP}}^{\mathrm{mix}}(B_d\pi ^+\pi ^{})\mathrm{sin}(\mathrm{\Delta }M_dt).`$
Here $`𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})`$ and $`𝒜_{\mathrm{CP}}^{\mathrm{mix}}(B_d\pi ^+\pi ^{})`$ are due to “direct” and “mixing-induced” CP violation, respectively. In the summer of 1999, the CLEO collaboration reported the first observation of the long-awaited $`B_d\pi ^+\pi ^{}`$ transition, with the following CP-averaged branching ratio :
$$\text{BR}(B_d\pi ^+\pi ^{})\frac{1}{2}\left[\text{BR}(B_d^0\pi ^+\pi ^{})+\text{BR}(\overline{B_d^0}\pi ^+\pi ^{})\right]=\left(4.3_{1.4}^{+1.6}\pm 0.5\right)\times 10^6.$$
(2)
This channel usually appears in the literature as a tool to determine the angle $`\alpha =180^{}\beta \gamma `$ of the unitarity triangle of the Cabibbo–Kobayashi–Maskawa matrix (CKM matrix) . However, penguin topologies are expected to affect this determination severely. Although there are several strategies on the market to control these penguin uncertainties , they are usually very challenging from an experimental point of view. Constraints on $`\alpha `$ from $`B_d\pi ^+\pi ^{}`$ were considered in .
In a recent paper , a strategy was proposed, where $`B_d\pi ^+\pi ^{}`$ is combined with its $`U`$-spin counterpart $`B_sK^+K^{}`$ to extract $`\varphi _d=2\beta `$ and $`\gamma `$. If the phase-convention independent quantity $`\varphi _d`$, which is related to the $`B_d^0`$$`\overline{B_d^0}`$ mixing phase and can be determined straightforwardly with the help of the “gold-plated” mode $`B_dJ/\psi K_\mathrm{S}`$ , is used as an input, the $`U`$-spin arguments in the extraction of $`\gamma `$ can be minimized. This approach, which relies only on the $`U`$-spin flavour symmetry and is not affected by any final-state-interaction effects , is very promising for “second-generation” $`B`$-physics experiments at hadron machines, such as LHCb or BTeV . There is a variant of this strategy for the asymmetric $`e^+e^{}`$ $`B`$-factories operating at the $`\mathrm{{\rm Y}}(4S)`$ resonance (BaBar and BELLE), where $`B_s`$ decays cannot be explored, if $`B_sK^+K^{}`$ is replaced by $`B_d\pi ^{}K^\pm `$, and a certain dynamical assumption concerning “exchange” and “penguin annihilation” topologies is made. Although $`B_sK^+K^{}`$ should be accessible at HERA-B and Run II of the Tevatron, a measurement of $`B_d\pi ^{}K^\pm `$ may be easier for these “first-generation” hadronic $`B`$ experiments. At HERA-B, for instance, one expects to collect 260 and 35 decay events per year of $`B_d\pi ^{}K^\pm `$ and $`B_sK^+K^{}`$, respectively . The present result for the CP-averaged $`B_d\pi ^{}K^\pm `$ branching ratio from the CLEO collaboration is as follows :
$$\text{BR}(B_d\pi ^{}K^\pm )\frac{1}{2}\left[\text{BR}(B_d^0\pi ^{}K^+)+\text{BR}(\overline{B_d^0}\pi ^+K^{})\right]=\left(17.2_{2.4}^{+2.5}\pm 1.2\right)\times 10^6;$$
(3)
a first result for the corresponding direct CP asymmetry is also available :
$$𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^{}K^\pm )\frac{\text{BR}(B_d^0\pi ^{}K^+)\text{BR}(\overline{B_d^0}\pi ^+K^{})}{\text{BR}(B_d^0\pi ^{}K^+)+\text{BR}(\overline{B_d^0}\pi ^+K^{})}=0.04\pm 0.16.$$
(4)
In this paper, we point out that the CLEO results (2) and (3) imply – among other things – a rather restricted range for the ratio of the “penguin” to “tree” contributions of the decay $`B_d\pi ^+\pi ^{}`$, and upper bounds on the direct CP asymmetries $`𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})`$ and $`𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^{}K^\pm )`$. If in addition mixing-induced CP violation in $`B_d\pi ^+\pi ^{}`$ is measured and $`\varphi _d`$ is fixed through $`B_dJ/\psi K_\mathrm{S}`$, we may obtain moreover interesting constraints on $`\gamma `$. An extraction of this angle becomes possible, if direct CP violation in $`B_d\pi ^+\pi ^{}`$ or $`B_d\pi ^{}K^\pm `$ is observed.
The outline of this paper is as follows: in Section 2, we have a brief look at the general structure of the relevant decay amplitudes and observables. The constraints on the penguin parameters and the direct CP asymmetries are discussed in Section 3, whereas the bounds on $`\gamma `$ are the subject of Section 4. Finally, the conclusions and an outlook are given in Section 5.
## 2 Decay Amplitudes and Observables
The transition amplitude of the $`\overline{b}\overline{d}`$ decay $`B_d^0\pi ^+\pi ^{}`$ can be written as follows :
$$A(B_d^0\pi ^+\pi ^{})=\lambda _u^{(d)}\left(A_{\mathrm{cc}}^u+A_{\mathrm{pen}}^u\right)+\lambda _c^{(d)}A_{\mathrm{pen}}^c+\lambda _t^{(d)}A_{\mathrm{pen}}^t,$$
(5)
where $`A_{\mathrm{cc}}^u`$ is due to “current–current” contributions, the amplitudes $`A_{\mathrm{pen}}^j`$ describe “penguin” topologies with internal $`j`$ quarks ($`j\{u,c,t\})`$, and the
$$\lambda _j^{(d)}V_{jd}V_{jb}^{}$$
(6)
are the usual CKM factors. Making use of the unitarity of the CKM matrix and applying the Wolfenstein parametrization , generalized to include non-leading terms in $`\lambda `$ , yields
$$A(B_d^0\pi ^+\pi ^{})=e^{i\gamma }\left(1\frac{\lambda ^2}{2}\right)𝒞\left[1de^{i\theta }e^{i\gamma }\right],$$
(7)
where
$$𝒞\lambda ^3AR_b\left(A_{\mathrm{cc}}^u+A_{\mathrm{pen}}^{ut}\right),$$
(8)
with $`A_{\mathrm{pen}}^{ut}A_{\mathrm{pen}}^uA_{\mathrm{pen}}^t`$, and
$$de^{i\theta }\frac{1}{(1\lambda ^2/2)R_b}\left(\frac{A_{\mathrm{pen}}^{ct}}{A_{\mathrm{cc}}^u+A_{\mathrm{pen}}^{ut}}\right).$$
(9)
The quantity $`A_{\mathrm{pen}}^{ct}`$ is defined in analogy to $`A_{\mathrm{pen}}^{ut}`$, and the CKM factors are given as usual by $`\lambda |V_{us}|=0.22`$, $`A|V_{cb}|/\lambda ^2=0.81\pm 0.06`$ and $`R_b|V_{ub}/(\lambda V_{cb})|=0.41\pm 0.07`$. The “penguin parameter” $`de^{i\theta }`$, which measures – sloppily speaking – the ratio of the $`B_d\pi ^+\pi ^{}`$ “penguin” to “tree” contributions, will play a central role in this paper.
Using the Standard-Model parametrization (7), we obtain
$`𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})`$ $`=`$ $`\left[{\displaystyle \frac{2d\mathrm{sin}\theta \mathrm{sin}\gamma }{12d\mathrm{cos}\theta \mathrm{cos}\gamma +d^2}}\right]`$ (10)
$`𝒜_{\mathrm{CP}}^{\mathrm{mix}}(B_d\pi ^+\pi ^{})`$ $`=`$ $`+\left[{\displaystyle \frac{\mathrm{sin}(\varphi _d+2\gamma )2d\mathrm{cos}\theta \mathrm{sin}(\varphi _d+\gamma )+d^2\mathrm{sin}\varphi _d}{12d\mathrm{cos}\theta \mathrm{cos}\gamma +d^2}}\right],`$ (11)
where $`\varphi _d=2\beta `$ can be determined with the help of the “gold-plated” mode $`B_dJ/\psi K_\mathrm{S}`$ through
$$𝒜_{\mathrm{CP}}^{\mathrm{mix}}(B_dJ/\psi K_\mathrm{S})=\mathrm{sin}\varphi _d.$$
(12)
Strictly speaking, mixing-induced CP violation in $`B_dJ/\psi K_\mathrm{S}`$ probes $`\varphi _d+\varphi _K`$, where $`\varphi _K`$ is related to the weak $`K^0`$$`\overline{K^0}`$ mixing phase and is negligibly small in the Standard Model. Due to the small value of the CP-violating parameter $`\epsilon _K`$ of the neutral kaon system, $`\varphi _K`$ can only be affected by very contrived models of new physics .
In the case of $`B_sK^+K^{}`$, we have
$$A(B_s^0K^+K^{})=e^{i\gamma }\lambda 𝒞^{}\left[1+\left(\frac{1\lambda ^2}{\lambda ^2}\right)d^{}e^{i\theta ^{}}e^{i\gamma }\right],$$
(13)
where
$$𝒞^{}\lambda ^3AR_b\left(A_{\mathrm{cc}}^u^{}+A_{\mathrm{pen}}^{ut^{}}\right)$$
(14)
and
$$d^{}e^{i\theta ^{}}\frac{1}{(1\lambda ^2/2)R_b}\left(\frac{A_{\mathrm{pen}}^{ct^{}}}{A_{\mathrm{cc}}^u^{}+A_{\mathrm{pen}}^{ut^{}}}\right)$$
(15)
correspond to (8) and (9), respectively. The primes remind us that we are dealing with a $`\overline{b}\overline{s}`$ transition. It should be emphasized that (7) and (13) are completely general parametrizations of the $`B_d^0\pi ^+\pi ^{}`$ and $`B_s^0K^+K^{}`$ decay amplitudes within the Standard Model, relying only on the unitarity of the CKM matrix. In particular, these expressions take into account also final-state-interaction effects, which received a lot of attention in the recent literature .
Since the decays $`B_d\pi ^+\pi ^{}`$ and $`B_sK^+K^{}`$ are related to each other by interchanging all down and strange quarks, the $`U`$-spin flavour symmetry of strong interactions implies
$$de^{i\theta }=d^{}e^{i\theta ^{}}.$$
(16)
Interestingly, this relation is not affected by $`U`$-spin-breaking corrections within a certain model-dependent approach (a modernized version of the “Bander–Silverman–Soni mechanism” ), making use – among other things – of the “factorization” hypothesis to estimate the relevant hadronic matrix elements . It would be interesting to investigate the $`U`$-spin-breaking corrections to (16) also within the “QCD factorization” approach, which was recently proposed in Ref. . In this paper, it was argued that there is a heavy-quark expansion for non-leptonic $`B`$-decays into two light mesons, and that non-factorizable corrections, as well as final-state-interaction processes, are suppressed by $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_b`$. We shall come back to this approach in Section 3, where a comparison of its prediction for the penguin parameter $`de^{i\theta }`$ is made with the constraints that are implied by the CLEO results (2) and (3).
For the following considerations, it is useful to introduce the observable
$$H\frac{1}{ϵ}\left|\frac{𝒞^{}}{𝒞}\right|^2\left[\frac{M_{B_d}}{M_{B_s}}\frac{\mathrm{\Phi }(M_K/M_{B_s},M_K/M_{B_s})}{\mathrm{\Phi }(M_\pi /M_{B_d},M_\pi /M_{B_d})}\frac{\tau _{B_s}}{\tau _{B_d}}\right]\left[\frac{\text{BR}(B_d\pi ^+\pi ^{})}{\text{BR}(B_sK^+K^{})}\right],$$
(17)
where
$$ϵ\frac{\lambda ^2}{1\lambda ^2},$$
(18)
and
$$\mathrm{\Phi }(x,y)\sqrt{\left[1(x+y)^2\right]\left[1(xy)^2\right]}$$
(19)
denotes the usual two-body phase-space function. The CP-averaged branching ratio BR$`(B_sK^+K^{})`$ can be extracted from the corresponding “untagged” rate , where no rapid oscillatory $`\mathrm{\Delta }M_st`$ terms are present . In the strict $`U`$-spin limit, we have $`|𝒞^{}|=|𝒞|`$. Corrections to this relation can be calculated within the “factorization” approximation, yielding
$$\left|\frac{𝒞^{}}{𝒞}\right|_{\mathrm{fact}}=\frac{f_K}{f_\pi }\frac{F_{B_sK}(M_K^2;0^+)}{F_{B_d\pi }(M_\pi ^2;0^+)}\left(\frac{M_{B_s}^2M_K^2}{M_{B_d}^2M_\pi ^2}\right),$$
(20)
where $`f_K`$ and $`f_\pi `$ denote the kaon and pion decay constants, and the form factors $`F_{B_sK}(M_K^2;0^+)`$ and $`F_{B_d\pi }(M_\pi ^2;0^+)`$ parametrize the hadronic quark-current matrix elements $`K^{}|(\overline{b}u)_{\mathrm{V}\mathrm{A}}|B_s^0`$ and $`\pi ^{}|(\overline{b}u)_{\mathrm{V}\mathrm{A}}|B_d^0`$, respectively . If we employ (7) and (13), we obtain the expression
$$H=\frac{12d\mathrm{cos}\theta \mathrm{cos}\gamma +d^2}{ϵ^2+2ϵd^{}\mathrm{cos}\theta ^{}\mathrm{cos}\gamma +d^2},$$
(21)
which will play a key role in the following sections. Let us also note that there is an interesting relation between $`H`$ and the corresponding direct CP asymmetries :
$$𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_sK^+K^{})=ϵH\left(\frac{d^{}\mathrm{sin}\theta ^{}}{d\mathrm{sin}\theta }\right)𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{}).$$
(22)
Since the decays $`B_sK^+K^{}`$ and $`B_d\pi ^{}K^\pm `$ differ only in their spectator quarks, we have
$$𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_sK^+K^{})𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^{}K^\pm )$$
(23)
$$\text{BR}(B_sK^+K^{})\text{BR}(B_d\pi ^{}K^\pm )\frac{\tau _{B_s}}{\tau _{B_d}},$$
(24)
and obtain
$$H\frac{1}{ϵ}\left(\frac{f_K}{f_\pi }\right)^2\left[\frac{\text{BR}(B_d\pi ^+\pi ^{})}{\text{BR}(B_d\pi ^{}K^\pm )}\right]=7.4\pm 3.0.$$
(25)
Here we have also taken into account the CLEO results (2) and (3), and have added the experimental errors in quadrature. The advantage of (25) is that it allows the determination of $`H`$ without a measurement of the decay $`B_sK^+K^{}`$. However, it should be kept in mind that this relation relies not only on $`SU(3)`$ flavour-symmetry arguments, but also on a certain dynamical assumption. The point is that $`B_sK^+K^{}`$ receives also contributions from “exchange” and “penguin annihilation” topologies, which are absent in $`B_d\pi ^{}K^\pm `$. It is usually assumed that these contributions play a minor role . However, they may be enhanced through certain rescattering effects . Although these topologies do not lead to any problems in the strategies discussed below if $`H`$ is fixed through a measurement of $`B_sK^+K^{}`$ – even if they should turn out to be sizeable – they may affect (23)–(25). The importance of the “exchange” and “penguin annihilation” topologies contributing to $`B_sK^+K^{}`$ can be probed – in addition to (23) and (24) – with the help of the decay $`B_s\pi ^+\pi ^{}`$. The naïve expectation for the corresponding branching ratio is $`𝒪(10^8)`$; a significant enhancement would signal that the “exchange” and “penguin annihilation” topologies cannot be neglected. Another interesting decay in this respect is $`B_dK^+K^{}`$, for which already stronger experimental constraints exist .
## 3 Constraining the Penguin Parameters and the <br>Direct CP Asymmetries
If we make use of (21) and apply the $`U`$-spin relation (16), the observable $`H`$ allows us to determine the quantity
$$C\mathrm{cos}\theta \mathrm{cos}\gamma $$
(26)
as a function of $`d`$:
$$C=\frac{ad^2}{2bd},$$
(27)
where
$$a=\frac{1ϵ^2H}{H1}\text{and}b=\frac{1+ϵH}{H1}.$$
(28)
In Ref. , a similar function of strong and weak phases was considered for the $`B_d\pi ^{}K^\pm `$, $`B^\pm \pi ^\pm K`$ system, and it was pointed out that this quantity plays an important role to derive interesting constraints. Since $`C`$ is the product of two cosines, it has to lie between $`1`$ and $`+1`$, thereby implying an allowed range for $`d`$. If we take into account (27) and (28), we obtain (for $`H<1/ϵ^2`$)
$$\frac{1ϵ\sqrt{H}}{1+\sqrt{H}}d\frac{1+ϵ\sqrt{H}}{|1\sqrt{H}|}.$$
(29)
An alternative derivation of this range can be found in Ref. . In the special case of $`H=1`$, there is only a lower bound on $`d`$, which is given by $`d_{\mathrm{min}}=(1ϵ)/2`$; for $`H<1`$, $`C`$ takes a minimal value that implies an allowed range for $`\gamma `$:
$$|\mathrm{cos}\gamma |C_{\mathrm{min}}=\frac{\sqrt{(1ϵ^2H)(1H)}}{1+ϵH}\sqrt{1H}.$$
(30)
From a conceptual point of view, this bound on $`\gamma `$ is completely analogous to the one derived in . Unfortunately, it is only of academic interest in the present case, as (25) indicates $`H>1`$, which we shall assume in the following discussion. So far, we have treated $`\theta `$ and $`\gamma `$ as “unknown”, free parameters. However, for a given value of $`\gamma `$, we have
$$|\mathrm{cos}\gamma |C+|\mathrm{cos}\gamma |,$$
(31)
and obtain constraints on $`d`$ that are stronger than (29):
$$d_{\mathrm{min}}^{\mathrm{max}}=\pm b|\mathrm{cos}\gamma |+\sqrt{a+b^2\mathrm{cos}^2\gamma }.$$
(32)
In Fig. 1, we show the dependence of $`C`$ on $`d`$ for the values of the observable $`H`$ given in (25). Interestingly, the large values of $`H`$ imply a rather restricted range for $`d`$. In particular, we get the lower bound $`d0.2`$. The “diamonds” in Fig. 1 represent the results obtained within the “QCD factorization” approach , representing the state-of-the-art technology in the calculation of the penguin parameter $`de^{i\theta }`$:
$$de^{i\theta }|_{\mathrm{QCD}\mathrm{fact}}=0.09[0.18]e^{i\mathrm{\hspace{0.17em}193}[187]^{}}.$$
(33)
Here a certain formally power-suppressed contribution, which is “chirally enhanced” through the factor
$$r_\chi =\frac{2M_\pi ^2}{(m_u+m_d)m_b},$$
(34)
has been neglected \[included at leading order\]. The “error bars” in Fig. 1 correspond to the presently allowed range for $`\gamma `$ that is implied by the usual “indirect” fits of the unitarity triangle :
$$36^{}\gamma 97^{},$$
(35)
and the “diamonds” are evaluated with (33) for the preferred (central) value of $`\gamma =62^{}`$. The horizontal dotted lines in Fig. 1 represent $`C=\pm \mathrm{cos}36^{}`$. It is an interesting feature of the contours in the $`d`$$`C`$ plane that they allow in principle the determination of $`\mathrm{cos}\gamma `$ with the help (33), i.e. if $`d`$ and $`\theta `$ are known. However, as can be seen in Fig. 1, the most recent CLEO data on $`B_d\pi ^+\pi ^{}`$ and $`B_d\pi ^{}K^\pm `$ are not in favour of an interpretation of the “QCD factorization” result (33) within the Standard Model; a solution could be obtained for $`d0.2`$ and $`C1`$. However, since (33) gives $`\mathrm{cos}\theta 1`$, we would then conclude that $`\mathrm{cos}\gamma 1`$, which would be in conflict with the Standard-Model range (35). Arguments for $`\mathrm{cos}\gamma <0`$ using $`BPP`$, $`PV`$ and $`VV`$ decays were also given in Ref. .
Before we discuss the origin of a possible discrepancy of the “QCD factorization” results with the contours in the $`d`$$`C`$ plane, let us have a closer look at the impact of corrections to (16). To this end, we generalize this relation as follows:
$$d^{}=\xi d,\theta ^{}=\theta +\mathrm{\Delta }\theta ,$$
(36)
yielding
$$C\mathrm{cos}\theta \mathrm{cos}\gamma =\left(\frac{1}{1+u^2}\right)\left[\frac{ad^2}{2bd}\pm u\sqrt{(1+u^2)\mathrm{cos}^2\gamma \left(\frac{ad^2}{2bd}\right)^2}\right],$$
(37)
where $`a`$ and $`b`$ correspond to the following generalization of (28):
$$a=\frac{1ϵ^2H}{\xi ^2H1},b=\frac{1+ϵ\xi H\mathrm{cos}\mathrm{\Delta }\theta }{\xi ^2H1},$$
(38)
and
$$u=\frac{ϵ\xi H\mathrm{sin}\mathrm{\Delta }\theta }{1+ϵ\xi H\mathrm{cos}\mathrm{\Delta }\theta }.$$
(39)
Since the parameter $`u`$ is doubly suppressed by $`ϵ`$ and $`\mathrm{\Delta }\theta `$, it is a small quantity. In the case of $`\mathrm{\Delta }\theta =20^{}`$, $`\xi =1`$ and $`H=7.4`$, we have, for example, $`u=0.10`$. In Fig. 2, we illustrate the impact of $`\xi 1`$ and $`\mathrm{\Delta }\theta 0`$ on the contour in the $`d`$$`C`$ plane corresponding to $`H=7.4`$. In contrast to (27), the general expression (37) depends also on the CKM angle $`\gamma `$ for $`\mathrm{\Delta }\theta 0`$. However, since the major effect in Fig. 2 is due to possible corrections to $`d^{}=d`$, we shall assume $`\theta ^{}=\theta `$ in the remainder of this paper. In this case, (37) takes the same form as (27).
Although it is too early to draw any definite conclusions, let us note that there would be basically two different explanations for a discrepancy of the “QCD factorization” results with the contours shown in Figs. 1 and 2: hadronic effects or physics beyond the Standard Model. Concerning the former case, the $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_b`$ terms and the “chirally enhanced” contributions may actually play an important role. Interestingly, the inclusion of the latter ones at leading order shifts the value of $`d`$ in the right direction. In order to get the full picture, it would be an important task to analyse (20) and (36) in the “QCD factorization” approach. Using present data, it seems that the “QCD factorization” results (33) can only be accommodated – if at all possible – for values of $`\gamma `$ sizeably larger than $`90^{}`$, which would be in conflict with (35), and a possible sign for new physics. Since the parameter $`de^{i\theta }`$ is governed by penguin topologies, i.e. by flavour-changing neutral-current (FCNC) processes, it may well be affected by physics beyond the Standard Model . Moreover, it should be kept in mind that the unitarity of the CKM matrix has been used in the calculation of the contours shown in Figs. 1 and 2. Further studies and better data are needed to explore these exciting issues in more detail.
Let us now turn to the constraints on the direct CP asymmetries (see also ). Before turning to the general case, it is instructive to consider $`\gamma =90^{}`$. In this case, we obtain
$$𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})|_{\gamma =90^{}}=\left[\frac{2d\mathrm{sin}\theta }{1+d^2}\right],𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_sK^+K^{})|_{\gamma =90^{}}=+\left[\frac{2ϵd^{}\mathrm{sin}\theta ^{}}{ϵ^2+d^2}\right],$$
(40)
and
$$H|_{\gamma =90^{}}=\frac{1+d^2}{ϵ^2+d^2}.$$
(41)
The CP asymmetries given in (40) take their extremal values for $`\theta =\theta ^{}=\pm 90^{}`$, and (41) allows us to determine $`d`$:
$$d|_{\gamma =90^{}}=\sqrt{\frac{1ϵ^2H}{\xi ^2H1}},$$
(42)
where we have also used $`d^{}=\xi d`$. Consequently, we obtain
$$\left|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})\right|_{\gamma =90^{}}^{\mathrm{max}}=2\sqrt{\frac{(1ϵ^2H)(\xi ^2H1)}{\left(\xi ^2ϵ^2\right)^2H^2}}\frac{2}{\xi \sqrt{H}}$$
(43)
and
$$\left|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_sK^+K^{})\right|_{\gamma =90^{}}^{\mathrm{max}}=2ϵ\xi \sqrt{\frac{(1ϵ^2H)(\xi ^2H1)}{\left(\xi ^2ϵ^2\right)^2}}2ϵ\sqrt{H}.$$
(44)
Let us emphasize that (44) is essentially unaffected by any corrections to the $`U`$-spin relation (16) for $`H=𝒪(10)`$; its theoretical accuracy is practically only limited by (20), which enters in the determination of $`H`$ through (17).
In the general case $`\gamma 90^{}`$, we employ (27) to eliminate the CP-conserving strong phase $`\theta `$ in (10). Following these lines, we obtain $`𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})`$ as a function of $`d`$ for a given value of $`\gamma `$. If we keep $`\gamma `$ fixed, and vary $`d`$ within the allowed range corresponding to (32), we find that $`|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})|`$ takes the following maximal value:
$$|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^+\pi ^{})|_{\mathrm{max}}=2|\mathrm{sin}\gamma |\sqrt{\frac{a+b^2\mathrm{cos}^2\gamma }{(1+a)^24(ab)(1+b)\mathrm{cos}^2\gamma }},$$
(45)
where $`a`$ and $`b`$ are given in (38) for $`\mathrm{\Delta }\theta =0`$ (see the comment after (39)). In the case of $`B_sK^+K^{}`$, we obtain
$`|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^{}K^\pm )|_{\mathrm{max}}|𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_sK^+K^{})|_{\mathrm{max}}`$ (46)
$`=2ϵ\xi H|\mathrm{sin}\gamma |\sqrt{{\displaystyle \frac{a+b^2\mathrm{cos}^2\gamma }{(1+a)^24(ab)(1+b)\mathrm{cos}^2\gamma }}}.`$
For $`\gamma =90^{}`$, these expressions reduce to (43) and (44), respectively. In Fig. 3, we show the dependence of (45) and (46) on $`\gamma `$ for the values of $`H`$ given in (25). The shaded regions correspond to a variation of the parameter $`\xi d^{}/d`$ within the interval $`[0.8,1.2]`$ for $`H=7.4`$. In contrast to (45), (46) is essentially unaffected by a variation of $`\xi `$, as we have already noted above. The range for $`H`$ given in (25) disfavours large direct CP violation in $`B_sK^+K^{}`$ and $`B_d\pi ^{}K^\pm `$ (see also ), which is also consistent with the 90% C.L. interval of $`0.22𝒜_{\mathrm{CP}}^{\mathrm{dir}}(B_d\pi ^{}K^\pm )+0.30`$ reported recently by the CLEO collaboration . On the other hand, there is a lot of space for large direct CP violation in $`B_d\pi ^+\pi ^{}`$. As can be seen in Fig. 3, a measurement of non-vanishing CP asymmetries $`|𝒜_{\mathrm{CP}}^{\mathrm{dir}}|_{\mathrm{exp}}`$ would allow us to exclude immediately a certain range of $`\gamma `$ around $`0^{}`$ and $`180^{}`$, as values of $`\gamma `$ corresponding to $`|𝒜_{\mathrm{CP}}^{\mathrm{dir}}|_{\mathrm{exp}}>|𝒜_{\mathrm{CP}}^{\mathrm{dir}}|_{\mathrm{max}}`$ are excluded. However, in order to constrain this CKM angle, the mixing-induced CP asymmetry $`𝒜_{\mathrm{CP}}^{\mathrm{mix}}(B_d\pi ^+\pi ^{})`$ appears to be more powerful.
Before we turn to these bounds in the following section, let us note that the observables of the decay $`B_d\pi ^+\pi ^{}`$ were combined with the CP-averaged $`B_d\pi ^{}K^\pm `$ and $`B_sK^+K^{}`$ branching ratios in Refs. and , respectively, to derive constraints on the penguin effects in the extraction of the CKM angle $`\alpha `$. In the present paper, we combine the experimental information provided by these modes in a different way, which appears more favourable to us. In particular, we use the mixing-induced CP asymmetry of the “gold-plated” mode $`B_dJ/\psi K_\mathrm{S}`$ as an additional input , and derive bounds on the CKM angle $`\gamma `$. The utility of $`B_d\pi ^{}K^\pm `$ decays to control the penguin effects on CP violation in $`B_d\pi ^+\pi ^{}`$ was also emphasized in Ref. .
## 4 Constraining the CKM Angle $`𝜸`$
In the following discussion, we assume that $`\mathit{\varphi }_𝒅\mathbf{=}\mathrm{𝟐}𝜷`$ has been measured at the $`𝑩`$-factories through (12), which is one of the major goals of these experiments. The presently allowed range for $`𝜷`$ that is implied by the usual “indirect” fits of the unitarity triangle is given as follows :
$$\mathrm{𝟏𝟔}^{\mathbf{}}\mathbf{}𝜷\mathbf{}\mathrm{𝟑𝟓}^{\mathbf{}}\mathbf{,}$$
(47)
with a preferred (central) value of $`𝜷\mathbf{=}\mathrm{𝟐𝟓}^{\mathbf{}}`$, which is also consistent with the present experimental result $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝑱\mathbf{/}𝝍𝑲_𝐒\mathbf{)}\mathbf{=}\mathbf{}\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathrm{𝟐}𝜷\mathbf{)}\mathbf{=}\mathbf{}\mathbf{0.79}_{\mathbf{}\mathbf{0.41}}^{\mathbf{+}\mathbf{0.44}}`$ of the CDF collaboration . A measurement of this mixing-induced CP asymmetry allows us to determine only $`\mathrm{𝐬𝐢𝐧}\mathit{\varphi }_𝒅`$, i.e. to fix $`\mathit{\varphi }_𝒅`$ up to a twofold ambiguity. Several strategies were proposed in the literature to resolve this ambiguity . In the $`𝑩`$-factory era, an experimental uncertainty of $`𝚫\mathrm{𝐬𝐢𝐧}\mathit{\varphi }_𝒅\mathbf{|}_{\mathrm{𝐞𝐱𝐩}}\mathbf{=}\mathbf{0.05}`$ seems to be achievable after a few years of taking data, which corresponds to an uncertainty of $`𝚫\mathit{\varphi }_𝒅\mathbf{=}\mathbf{\pm }\mathrm{𝟓}^{\mathbf{}}`$ for the central value of $`\mathit{\varphi }_𝒅\mathbf{=}\mathrm{𝟓𝟎}^{\mathbf{}}`$.
If we assume, for a moment, that there are no penguin effects present in $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$, i.e. $`𝒅\mathbf{=}\mathrm{𝟎}`$, we would simply have
$$𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{|}_{𝒅\mathbf{=}\mathrm{𝟎}}\mathbf{=}\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}\mathrm{𝟐}𝜸\mathbf{)}\mathbf{,}$$
(48)
as can be seen in (11). Since the unitarity of the CKM matrix implies $`\mathit{\varphi }_𝒅\mathbf{+}\mathrm{𝟐}𝜸\mathbf{=}\mathbf{}\mathrm{𝟐}𝜶`$, this CP asymmetry is usually written as $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{|}_{𝒅\mathbf{=}\mathrm{𝟎}}\mathbf{=}\mathbf{}\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathrm{𝟐}𝜶\mathbf{)}`$, and would allow a direct measuerment of $`𝜶`$. However, (48) is the “generic” interpretation of this CP asymmetry, allowing us to determine $`𝜸`$, if $`\mathit{\varphi }_𝒅`$ is fixed through $`𝑩_𝒅\mathbf{}𝑱\mathbf{/}𝝍𝑲_𝐒`$. In the case of large penguin contributions, this interpretation of $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}`$ actually appears to be more favourable than the usual one in terms of $`𝜶`$, which was employed, for example, in Refs. . Since we definitely have to worry about penguin effects in $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$, as we have pointed out in the previous section, we shall use the corresponding mixing-induced CP asymmetry to contrain the CKM angle $`𝜸`$ in this section.
Concerning the search for new physics, $`𝜸`$ is actually the interesting aspect of the mixing-induced $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ CP asymmetry. If $`\mathit{\varphi }_𝒅`$ is affected by new physics, these effects could be seen, for example, by comparing the $`𝑩_𝒅\mathbf{}𝑱\mathbf{/}𝝍𝑲_𝐒`$ results with the “indirect” range (47). Since this channel is governed by $`\overline{𝒃}\mathbf{}\overline{𝒄}𝒄\overline{𝒔}`$ “tree” processes, its decay amplitude is not expected to be affected significantly by new-physics effects, and allows the determination of $`\mathit{\varphi }_𝒅`$ even in the presence of physics beyond the Standard Model. In order to search for indications of new physics, the values of $`𝜸`$ implied by the CP-violating effects in $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ could be compared with the “indirect” range arising from the usual fits of the unitarity triangle, or with theoretically clean extractions from pure “tree” decays, such as $`𝑩_𝒅\mathbf{}𝑫^\mathbf{}\mathbf{\pm }𝝅^{\mathbf{}}`$ or $`𝑩\mathbf{}𝑫𝑲`$ (see also the brief discussion of new-physics effects in Section 3).
If we look at the expressions (10) and (11) for the direct and mixing-induced CP asymmetries of the decay $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$, we observe that the CP-conserving strong phase $`𝜽`$ enters only in the form of $`\mathrm{𝐜𝐨𝐬}𝜽`$ in the latter case. Consequently, using $`\mathrm{𝐜𝐨𝐬}𝜽\mathbf{=}𝑪\mathbf{/}\mathrm{𝐜𝐨𝐬}𝜸`$ and (27), we obtain
$$𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{=}\frac{\mathbf{\left[}𝒃\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}\mathrm{𝟐}𝜸\mathbf{)}\mathrm{𝐜𝐨𝐬}𝜸\mathbf{}𝒂\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}𝜸\mathbf{)}\mathbf{\right]}\mathbf{+}\mathbf{\left[}\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}𝜸\mathbf{)}\mathbf{+}𝒃\mathrm{𝐬𝐢𝐧}\mathit{\varphi }_𝒅\mathrm{𝐜𝐨𝐬}𝜸\mathbf{\right]}𝒅^\mathrm{𝟐}}{\mathbf{\left[}\mathbf{(}𝒃\mathbf{}𝒂\mathbf{)}\mathbf{+}\mathbf{(}\mathrm{𝟏}\mathbf{+}𝒃\mathbf{)}𝒅^\mathrm{𝟐}\mathbf{\right]}\mathrm{𝐜𝐨𝐬}𝜸}\mathbf{,}$$
(49)
where $`𝒂`$ and $`𝒃`$ are given in (38) for $`𝚫𝜽\mathbf{=}\mathrm{𝟎}`$, i.e. the small corrections due to $`𝚫𝜽\mathbf{}\mathrm{𝟎}`$ have been neglected for simplicity (see the comment after (39)). Since (49) is a monotonic function of the variable $`𝒅^\mathrm{𝟐}`$, it takes its extremal values for the minimal and maximal values of $`𝒅`$ given in (32); inserting them into (49) yields
$$𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{|}_{\mathrm{𝐞𝐱𝐭𝐫}\mathbf{.}}\mathbf{=}\frac{\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}\mathrm{𝟐}𝜸\mathbf{)}\mathbf{+}𝒂\mathrm{𝐬𝐢𝐧}\mathit{\varphi }_𝒅\mathbf{+}\mathrm{𝟐}𝒘_\mathbf{\pm }\mathbf{\left[}\mathrm{𝐬𝐢𝐧}\mathbf{(}\mathit{\varphi }_𝒅\mathbf{+}𝜸\mathbf{)}\mathbf{+}𝒃\mathrm{𝐜𝐨𝐬}𝜸\mathrm{𝐬𝐢𝐧}\mathit{\varphi }_𝒅\mathbf{\right]}}{\mathrm{𝟏}\mathbf{+}𝒂\mathbf{+}\mathrm{𝟐}𝒘_\mathbf{\pm }\mathbf{(}\mathrm{𝟏}\mathbf{+}𝒃\mathbf{)}\mathrm{𝐜𝐨𝐬}𝜸}\mathbf{,}$$
(50)
where
$$𝒘_\mathbf{\pm }\mathbf{=}𝒃\mathrm{𝐜𝐨𝐬}𝜸\mathbf{\pm }\sqrt{𝒂\mathbf{+}𝒃^\mathrm{𝟐}\mathrm{𝐜𝐨𝐬}^\mathrm{𝟐}𝜸}\mathbf{.}$$
(51)
In Fig. 4, we illustrate the resulting allowed range for $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}`$ in the case of $`𝑯\mathbf{=}\mathbf{7.4}`$ and $`\mathit{\varphi }_𝒅\mathbf{=}\mathrm{𝟓𝟎}^{\mathbf{}}`$ (shaded region). The impact of a deviation of the parameter $`𝝃`$ from 1 is illustrated by the dotted and dot-dashed lines, which correspond to $`𝝃\mathbf{=}\mathbf{0.8}`$ and $`\mathbf{1.2}`$, respectively. For a given value of $`𝜸`$, the allowed range for the mixing-induced $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ CP asymmetry is usually very large. However, a measured value of $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}`$ would, on the other hand, imply a rather restricted range for $`𝜸`$. If we assume, for example, that $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{=}\mathbf{0.4}`$ has been measured, and take into account that the experimental value of $`𝜺_𝑲`$ implies $`𝜸\mathbf{}\mathbf{[}\mathrm{𝟎}^{\mathbf{}}\mathbf{,}\mathrm{𝟏𝟖𝟎}^{\mathbf{}}\mathbf{]}`$, we would conclude that $`\mathrm{𝟒𝟏}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟕𝟒}^{\mathbf{}}`$ or $`\mathrm{𝟏𝟓𝟖}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟏𝟕𝟎}^{\mathbf{}}`$. Allowing $`𝝃\mathbf{}\mathbf{[}\mathbf{0.8}\mathbf{,}\mathbf{1.2}\mathbf{]}`$, i.e. symmetry-breaking corrections of $`\mathrm{𝟐𝟎}\mathbf{\%}`$, we would obtain the slightly modified ranges $`\mathrm{𝟑𝟗}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟖𝟎}^{\mathbf{}}`$ $`\mathbf{}`$ $`\mathrm{𝟏𝟓𝟓}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟏𝟕𝟎}^{\mathbf{}}`$ and $`\mathrm{𝟒𝟑}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟕𝟏}^{\mathbf{}}`$ $`\mathbf{}`$ $`\mathrm{𝟏𝟔𝟎}^{\mathbf{}}\mathbf{}𝜸\mathbf{}\mathrm{𝟏𝟕𝟎}^{\mathbf{}}`$ for $`𝝃\mathbf{=}\mathbf{0.8}`$ and 1.2, respectively. Since the allowed region for $`𝒅`$ is enlarged (reduced) for smaller (larger) values of $`𝑯`$, the bounds on $`𝜸`$ become weaker (stronger) in this case.
Let us finally note that if in addition to a measurement of $`𝑯`$ and $`𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐦𝐢𝐱}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}`$ direct CP violation in $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ or $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ is observed, we have three independent observables at our disposal, which depend on $`𝜸`$, $`𝒅`$ and $`𝜽`$. Consequently, we are then not only in a position to constrain these “unknown” parameters, but also to determine them . Moreover, the normalization $`\mathbf{|}𝓒\mathbf{|}`$ of the $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ decay amplitude (see (7)) can be extracted from the corresponding CP-averaged branching ratio, and can be compared with theoretical predictions. The $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ decays offer also alternative strategies to determine $`𝜸`$ and certain hadronic quantities, if these transitions are combined with other $`𝑩\mathbf{}𝝅𝑲`$ modes .
## 5 Conclusions and Outlook
The decays $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ and $`𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}`$ provide interesting strategies to extract the CKM angle $`𝜸`$ and hadronic penguin parameters at “second-generation” $`𝑩`$-physics experiments of the LHC era. In this paper, we have considered a variant of this approach for the “first-generation” $`𝑩`$-factories, where the $`𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}`$ decays are replaced by $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ modes.
We have pointed out that the CP-averaged $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ and $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ branching ratios allow us to fix contours in the $`𝒅`$$`\mathbf{[}\mathrm{𝐜𝐨𝐬}𝜽\mathrm{𝐜𝐨𝐬}𝜸\mathbf{]}`$ plane, which can be compared with theoretical results for the $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ “penguin parameter” $`𝒅𝒆^{𝒊𝜽}`$, for example with those of the “QCD factorization” approach. Although it is too early to draw any definite conclusions, it is interesting to note that the most recent CLEO data are not in favour of an interpretation of the “QCD factorization” results within the Standard Model. This feature may be due to hadronic effects or new physics. Further theoretical studies and better experimental data are required to investigate these exciting issues in more detail.
Another interesting aspect of the recent CLEO results for the CP-averaged $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ and $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ branching ratios is that they imply upper bounds on the corresponding direct CP asymmetries, which are given by $`\mathbf{|}𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐝𝐢𝐫}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}\mathbf{)}\mathbf{|}_{\mathrm{𝐦𝐚𝐱}}\mathbf{}\mathbf{<}\mathbf{0.8}`$ and $`\mathbf{|}𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐝𝐢𝐫}}\mathbf{(}𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }\mathbf{)}\mathbf{|}_{\mathrm{𝐦𝐚𝐱}}\mathbf{}\mathbf{|}𝓐_{\mathrm{𝐂𝐏}}^{\mathrm{𝐝𝐢𝐫}}\mathbf{(}𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}\mathbf{)}\mathbf{|}_{\mathrm{𝐦𝐚𝐱}}\mathbf{}\mathbf{<}\mathbf{0.3}`$. The latter bound is remarkably stable under $`𝑼`$-spin-breaking corrections – in contrast to the former one – and may also play an important role to search for new physics.
If in addition to the CP-averaged $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ and $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$ branching ratios mixing-induced CP violation in the former decay is measured, and the $`𝑩_𝒅^\mathrm{𝟎}`$$`\overline{𝑩_𝒅^\mathrm{𝟎}}`$ mixing phase is fixed through $`𝑩_𝒅\mathbf{}𝑱\mathbf{/}𝝍𝑲_𝐒`$, interesting constraints on $`𝜸`$ can be obtained. A further step in this programme would be the observation of direct CP violation in $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ or $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$, which would allow a determination of $`𝜸`$, $`𝒅𝒆^{𝒊𝜽}`$ and $`\mathbf{|}𝓒\mathbf{|}`$. In this way, two of the major goals of the $`𝑩`$-factories – time-dependent analyses of the benchmark modes $`𝑩_𝒅\mathbf{}𝑱\mathbf{/}𝝍𝑲_𝐒`$ and $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ – can be combined with each other to probe the CKM angle $`𝜸`$ and to obtain valuable insights into the world of penguins.
Another important step would be a measurement of the CP-averaged $`𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}`$ branching ratio, which may be possible at HERA-B and Run II of the Tevatron. Using this observable, a certain dynamical assumption concerning “exchange” and “penguin annihilation” topologies can be avoided, which has to be made in the case of $`𝑩_𝒅\mathbf{}𝝅^{\mathbf{}}𝑲^\mathbf{\pm }`$. The theoretical accuracy would then only be limited by $`𝑼`$-spin-breaking effects and would not be affected by any final-state-interaction processes. The final goal is a measuerment of the CP-violating observables of $`𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}`$, which should be possible at LHCb and BTeV. At these experiments, the physics potential of $`𝑩_𝒅\mathbf{}𝝅^\mathbf{+}𝝅^{\mathbf{}}`$ and $`𝑩_𝒔\mathbf{}𝑲^\mathbf{+}𝑲^{\mathbf{}}`$ can be fully exploited, and in addition to an extraction of $`𝜸`$ at the level of a few degrees, also interesting consistency checks of the basic $`𝑼`$-spin relations can be performed. |
warning/0001/cond-mat0001040.html | ar5iv | text | # Critical exponents at the superconductor-insulator transition in dirty-boson systems
## Abstract
I obtain the inverse of the correlation length exponent at the superfluid-Bose glass quantum critical point as a series in small parameter $`\sqrt{d1}`$, with $`d`$ being the dimensionality of the system, and compute the first two terms. For $`d=2`$ I find $`\nu _s=0.81`$ and $`\nu _c=1.03`$, for short-range and Coulomb interactions between bosons, respectively. When combined with the exact values of the dynamical critical exponents, these results are in quantitative agreement with the experiments on onset of superfluidity in $`{}_{}{}^{4}He`$ in porous glasses, and on superconductor-insulator transition in homogeneous metallic films.
The phenomenon of superconductor-insulator (SI) transition occurs in plethora of low-dimensional electronic systems, examples ranging from Josephson junction arrays and homogeneous thin films , to high-temperature superconductors , and is believed to represent a prototypical quantum (zero-temperature) phase transition. At low temperatures, as some controlling parameter like thickness of a film is varied, the resistivity changes from a sharply decreasing to a continuously increasing function of temperature . The good collapse of the resistivity data under scaling and near-universality of the critical value of the conductivity indicate a quantum ($`T=0`$) critical point that separates two many-body ground states with different symmetries. A natural question arises: what is the mechanism of destruction of the superfluid ground state in a disordered system? One possibly universal answer is provided by the so-called dirty-boson theory, which postulates that it is the loss of the superconducting phase coherence due to localization of Cooper pairs which is ultimately responsible for the SI transition , , . Since on the scale of the diverging phase-coherence correlation length Cooper pairs will appear as point particles, the SI transition would, under this hypothesis, in general fall into the same universality class as the onset of superfluidity in $`{}_{}{}^{4}He`$ in disordered media , corrected for the long-rangeness of the Coulomb interaction. In principle, a way to assess the validity of this idea is to compare the measured critical exponents with the calculations for the dirty-boson Hamiltonian. A strongly-coupled nature of the dirty-boson critical point, however, poses a fundamental obstacle to this procedure, and makes any but most qualitative understanding of the superfluid-Bose glass transition very difficult. The absence of a useful non-interacting starting point for disordered bosons forces one to rely on uncontrollable approximation schemes or turn to numerical calculations . This seems to be a common problem for the theories of interacting disordered low-dimensional quantum systems, apparent also in the fermionic systems of this type . In fact, the dirty-boson Hamiltonian may be the simplest quantum problem that irreducibly contains the physics of interactions and disorder , and as such has received a lot of attention through the years .
Recently, a new approach to the dirty-boson criticality has been suggested , according to which the strongly coupled superfluid-Bose glass critical point in two dimensions ($`d=2`$) could be understood as smoothly evolving from the zero-disorder critical point in $`d=1`$. The idea is to note that, by preventing the clean superfluid ground state in $`d=1`$ to exhibit a true long-range order, the Mermin-Wagner theorem forces the SI fixed point in $`d=1`$ to lie precisely at zero disorder . In $`d=1+ϵ`$ a true long-range order becomes possible and the superfluid thus becomes more resilient to disorder, which causes the SI fixed point to shift to a finite, but small, value of disorder, controlled by the parameter $`ϵ`$ . Although the dirty-boson transition probably lacks the upper critical dimension , it has the lower critical dimension $`d_l=1`$, and this in principle allows one to compute the universal quantities at the transition perturbatively in small parameter $`ϵ=d1`$. Within this formalism, a particular symmetry of the low-energy action present in $`d=1`$ guarantees that the dynamical critical exponent is $`z=d`$ ($`z=1`$ for Coulomb interaction) exactly , in agreement with the expectation based on the compressible nature of the Bose-glass . The second, correlation length exponent $`\nu `$, however, turns out to be a perturbation series in $`\sqrt{ϵ}`$. On the experimental side, a directly measurable quantity is typically the product of the two exponents $`z\nu `$ , , and a meaningful comparison with experiment requires knowledge of the exponent $`\nu `$ to some accuracy. In this Letter I present a field-theoretic method for higher-order calculations within the $`ϵ`$-expansion for the superfluid-Bose glass transition, and use it to compute the correlation length exponent to two lowest orders in $`\sqrt{ϵ}`$. The result for both short-range and Coulomb interaction between bosons (see Table I) leads to values of $`\nu `$ in $`d=2`$ in a very good agreement with the experiments on the onset of superfluidity in $`{}_{}{}^{4}He`$ in aerogel and on the SI transition in thin metallic films , as well as with the Monte Carlo calculations on the dirty-boson Hamiltonian . My calculation supports the idea that the SI transition in disordered electronic systems falls into the dirty-boson universality class, and establishes a way for a quantitative understanding of the SI criticality, as presently exists for the thermal critical phenomena . The effort involved in higher-order calculations of the correlation length exponent and of the universal critical conductivity is discussed.
To be specific, consider the effective low-energy $`T=0`$ action for the disordered superfluid in $`d=1`$ , :
$`S={\displaystyle \frac{K}{\pi }}{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑x𝑑\tau \{c^2[_x\theta _i(x,\tau )]^2+[_\tau \theta _i(x,\tau )]^2\}`$ (1)
$`D{\displaystyle \underset{i,j=1}{\overset{N}{}}}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑x𝑑\tau 𝑑\tau ^{}\mathrm{cos2}[\theta _i(x,\tau )\theta _j(x,\tau ^{})].`$ (2)
The coupling $`K`$ is inversely proportional to the superfluid density at some microscopic cutoff length $`\mathrm{\Lambda }^1`$, $`c`$ is the velocity of low-energy phonons, and $`D`$ is proportional to the width of the Gaussian random potential. The interaction between particles in (1) is taken to be short-ranged, and the standard limit on the number of replicas $`N0`$ is assumed. The field $`\theta (x,\tau )`$ is the dual phase , , and the above theory describes the destruction of the second sound mode in the superfluid due to unbinding of topological defects (phase slips) at the point of transition in $`d=1`$. Its role is similar to that of the sine-Gordon theory for the Kosterlitz-Thouless transition in the 2D XY model . To determine the macroscopic state of the system, one is in principle interested in fate of the couplings $`K`$, $`c`$, and $`D`$ as the cutoff in the theory is lowered. Under a change $`\mathrm{\Lambda }\mathrm{\Lambda }/s`$ the combinations of the coupling constants $`u=3\eta ^1`$, (where $`\eta =Kc`$), $`\kappa =1/Kc^2`$, and $`WD`$ (to be precisely defined shortly), in $`d=1+ϵ`$ dimensions are expected to renormalize according to the $`\beta `$-functions :
$$\dot{u}=ϵ(u3)+W+auW+O(W^2),$$
(3)
$$\dot{W}=uW+bW^2+O(uW^2),$$
(4)
$$\dot{\kappa }=(dz)\kappa ,$$
(5)
where $`\dot{x}=dx/d\mathrm{ln}(s)`$, and $`a`$ and $`b`$ are numerical coefficients. The $`d`$-dependent terms in the recursion relations (2)-(4) can be understood as deriving from the scaling of the superfluid density, $`\rho _{sf}K^1\xi ^{2zd}`$, and the compressibility, $`\kappa \xi ^{zd}`$ , , where $`\xi `$ is the diverging correlation length near the critical point, and $`z`$ the dynamical exponent. Adopting the logic of the minimal subtraction scheme , the disorder-dependent terms in the recursion relations should be computed precisely in $`d=1`$, where one has the dual representation (1) of the low-energy theory on disposal. The symmetry of the interaction term in (1) under a transformation $`\theta _i(x,\tau )\theta _i(x,\tau )+f(x)`$ for arbitrary $`f(x)`$ implies then that there could be no disorder-dependent terms in Eq. (4) , so $`z=d`$ at the fixed point. The correlation length exponent $`\nu `$ follows from the linearization of the first two equations near the critical point at $`W^{}u^{}ϵ`$. It is then straightforward to check that to the second order in $`ϵ^{1/2}`$
$$\nu _s^1=3^{\frac{1}{2}}ϵ^{\frac{1}{2}}+\frac{1+3(a+b)}{2}ϵ+O(ϵ^{\frac{3}{2}}),$$
(6)
for short-range interactions. The $`O(ϵ^{3/2})`$ term follows from the higher-order terms in Eqs. (2)-(3).
While the outlined procedure is conceptually simple, its implementation is made difficult by the fact that the action (1) has a compact form only in real space, and the interaction term contains all powers of the dual field. A similar obstacle appears in the calculation of the Kosterlitz recursion relations beyond the lowest order in fugacity . Here I will introduce a field-theoretic approach to the problem, which also enables one to avoid the usual pitfalls of the momentum-shell renormalization group when applied beyond the lowest order. The gist of the method is to recognize that in $`d=1`$, right at $`u=0`$ the coupling constant $`W`$ becomes dimensionless, and the theory (1) appears to be just renormalizable. One then expects that the logarithmic divergences at the renormalizable point $`d=1`$ and $`u=0`$ will at small finite $`u`$ show as poles when $`u0`$. Since the coupling $`W`$ acquires a finite scaling dimension for $`u0`$, the coefficients in the $`\beta `$-functions are expected to stay finite as $`u0`$. This is analogous to the standard procedure of dimensional regularization, commonly used to study thermal critical phenomena near the upper critical dimension , except that here the coupling constant $`u`$ plays the role of dimension, while the real physical dimension is at first fixed at $`d=1`$. When finally $`d1+ϵ`$, the $`\beta `$-functions are deformed into Eqs. (2)-(4).
With this strategy in mind, consider the self-energy defined by the propagator of the dual phase in $`d=1`$ as $`G^1(k,\omega )=(2K/\pi )(\omega ^2+c^2k^2)+\mathrm{\Sigma }(\omega )`$. It will prove useful to separate the first and the second order contributions to $`\mathrm{\Sigma }`$ by writing it as $`\mathrm{\Sigma }(\omega )=\mathrm{\Sigma }_1(\omega )+\mathrm{\Sigma }_2(\omega )+O(D^3)`$, where $`\mathrm{\Sigma }_nD^n`$. Simple calculation then gives
$$\mathrm{\Sigma }_1(\omega )=8D_{\mathrm{}}^{\mathrm{}}𝑑\tau (1\mathrm{e}^{i\omega \tau })e^{f(\tau )},$$
(7)
where the two-point correlation function $`f(\tau )=4\theta (0,0)(\theta (0,0)\theta (0,\tau ))`$ is given by the integral
$$f(\tau )=(3u)_0^{c\mathrm{\Lambda }|\tau |}\frac{dx}{x}(1e^x).$$
(8)
When $`u0`$, it readily follows that at small frequencies
$$\mathrm{\Sigma }_1(\omega )=\omega ^2\frac{2W}{\pi c}(\frac{1}{u}+O(1)),$$
(9)
where I introduced the frequency-dependent, dimensionless coupling $`W=(4\pi D/(c^2\mathrm{\Lambda }^3))(c\mathrm{\Lambda }/\omega )^u`$. After a tedious but straightforward algebra one similarly finds
$$\mathrm{\Sigma }_2(\omega )=\frac{\pi }{2K}(\frac{\mathrm{\Sigma }_1(\omega )}{\omega })^2+I(\omega ),$$
(10)
where
$`I(\omega )=8D^2{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑y𝑑\tau 𝑑\tau ^{}𝑑v(1e^{i\omega \tau })\mathrm{e}^{f(\tau )f(\tau ^{})}`$ (11)
$`[F(y,v,\tau ,\tau ^{})(1+\mathrm{e}^{i\omega v}{\displaystyle \frac{1}{2}}(1\mathrm{e}^{i\omega \tau ^{}})1],`$ (12)
and $`F`$ denotes a four-point correlation function:
$$F(y,v,\tau ,\tau ^{})=\mathrm{e}^{4(\theta (0,0)\theta (0,\tau ))(\theta (y,v)\theta (y,v+\tau ^{}))}.$$
(13)
When $`\omega 0`$, after rescaling the imaginary times and the length in the integral as $`\omega \tau \tau `$ and $`\omega y/cy`$, the leading divergence in $`I(\omega )`$ as $`u0`$ comes from the integration over small values of $`\tau `$ and $`\tau ^{}`$. To obtain the leading divergence in $`I(\omega )`$ it therefore suffices to expand $`F`$ to the lowest order in $`\tau `$ and $`\tau ^{}`$, to find that at small frequencies
$$I(\omega )=\omega ^2\frac{6}{\pi c}W^2(\frac{1}{u^2}+O(\frac{1}{u})).$$
(14)
The last equation is the central result of this work. Collecting all the terms back into the self-energy one recognizes the renormalized coupling $`\eta _r`$ as the coefficient of $`\omega ^2`$-term in the propagator. In general,
$$\eta _r=\eta +\frac{W}{u}+x\frac{W^2}{u^2}+O(\frac{W^2}{u}),$$
(15)
where the terms finite when $`u0`$ have been discarded, and $`x`$ is a number determined from Eq. (12). After judiciously defining a renormalized disorder in (13) as $`W_r=W+2xW^2/u`$, and rescaling it to bring the coefficient of $`O(W)`$-term in the Eq. (2) to unity as $`9W_rW_r`$, a differentiation with respect to $`\mathrm{ln}(c\mathrm{\Lambda }/\omega )`$ leads to the Eqs. (2)-(3) for the renormalized couplings $`u_r`$ and $`W_r`$, with the coefficients $`a=2/3`$ and $`b=2x/9`$, with $`b=0`$. The subleading $`W^2/u`$ term in the Eq. (13) determines the next, $`O(W^2)`$, term in (2), and the next-order correction to $`\nu _s^1`$.
A remarkable feature of the perturbation series for $`\nu _s`$ is its independence of the renormalization procedure, i. e. of the non-universal finite parts of the self-energy which have been dropped in the last equation. To see this to the order of my calculation consider the most general redefinition of the coupling constants to the order $`W^2`$ :
$$u_r^{}=u_r+\alpha W_r+\beta u_rW_r+\gamma W_r^2,$$
(16)
$$W_r^{}=W_r+\delta u_rW_r+\sigma W_r^2,$$
(17)
where the coefficients $`\{\alpha ,\mathrm{}\sigma \}`$ are finite, and dependable on the finite parts of the self-energy. It is easy to check that the recursion relations for the new couplings have the same form as the Eqs. (2)-(3), but with the coefficients $`a^{}=a+\alpha \delta `$ and $`b^{}=b\alpha +\delta `$. Interestingly, while the coefficients $`a`$ and $`b`$ by themselves are non-universal, the critical exponent depends only on their sum, which is perfectly universal, i. e. scheme independent. An invariant similar to $`a+b`$ appears also in the $`\beta `$-functions for the sine-Gordon model , where it determines the first correction to Kosterlitz-Thouless scaling.
I expect the presented $`ϵ`$-expansion to lead to divergent series; the point $`D=0`$ in the action (1) is non-analytic, since for $`D<0`$ the Gaussian probability distribution for the random potential becomes unbounded. Nevertheless, the hope is that the series in Eq. (5) for example, will be asymptotic, and that the few lowest terms may already lead to useful results. Indeed, estimating $`\nu _s`$ for $`d=2`$ from the simple sum of the first two terms gives $`\nu _s=0.81`$, within bounds found in the Monte Carlo calculations . The experimental data of Crowell et al. on the onset of superfluidity and on the specific heat of $`{}_{}{}^{4}He`$ in aerogel at low temperatures on its face value are consistent with the effective dimensionality of $`d=2`$. Under this assumption the product of the two exponents in their experiment is $`z\nu =1.60\pm 0.08`$. Assuming further that $`z_s=2`$ at the transition in $`d=2`$ gives $`\nu =0.80\pm 0.04`$. Although the uncertainty cited here should be taken with some reservation, and the accuracy of the measurement falls short of the standards in thermal critical phenomena, the result appears to be in excellent agreement with my calculation (see Table I). It is worth noting that the inequality $`\nu 2/d`$ seems to be violated both by the experiment and by my estimate. It has been argued recently that the above inequality is an artifact of the particular averaging procedure, and that the true exponent is in fact not bound from below. It would be interesting to see if the higher-order corrections eventually push the value of $`\nu _s`$ above unity in $`d=2`$, or indeed $`\nu _s<1`$ as the experiment and the present calculation suggest.
To make a comparison with the experiments on SI transition in homogeneous thin films with thickness as the tuning parameter it is necessary to take into account the long-range Coulomb interaction between the Cooper pairs. As explained in detail elsewhere , within the present scheme this may be simply accomplished by defining the Coulomb interaction as $`V_c(\stackrel{}{r})=e^2d^d\stackrel{}{q}\mathrm{exp}(i\stackrel{}{q}\stackrel{}{r})/q^{d1}`$, so to coincide with the $`\delta `$-function repulsion in $`d=1`$. The only change in the calculation then is that $`z_c=1`$, and that $`ϵϵ/2`$ in the Eq. (2), while the disorder-dependent terms in the recursion relations, which follow from $`d=1`$, remain the same. The first two terms in the series then give $`\nu _c=1.03`$, in accord with the Monte Carlo results (see Table I). Experiment finds $`z_c=1.0\pm 0.1`$, by suppressing the transition temperature with the magnetic field or by scaling of resistance with the electric field . Collapsing the resistivity data then gives the experimental value of $`\nu _c=1.2\pm 0.2`$, again in a very good agreement with my result.
As mentioned earlier, the next term in the series (5) requires only the computation of the subleading, $`O(W^2/u)`$ term in Eq. (13). Here, however, it appears that it is no more enough to know the correlator $`F`$ (after rescaling the lengths with $`\omega `$) only at small rescaled $`\tau `$ and $`\tau ^{}`$, as it was for the leading divergence in Eq. (12). In light of the likely asymptotic nature of the expansion, this is left for future work.
Another universal quantity of interest is the critical conductivity in $`d=2`$ , which, aside from the universal unit $`e_{}^2/h`$, for bosons of charge $`e_{}`$ can be obtained as a Laurent series in $`ϵ`$ . The lowest order term was obtained in , and for $`d=2`$ the result $`\sigma _c=0.69e_{}^2/h`$ agrees with the low-temperature experiment on Bismuth films quite well. It would nevertheless be useful to compute the next-order correction, and see if it pushes the result towards the self-dual value of $`e_{}^2/h`$, to which a large number of experimental results seems to converge. Calculating the next-order term in the critical conductivity would in principle proceed along the same lines as here, except that one needs to perform the analytic continuation to real frequencies to obtain the real-time dc response at low temperature. This problem was solved in to the lowest order, but applying the same trick in the next order term seems not straightforward. A preliminary analysis also suggests that the second-order contribution to the universal conductivity requires a computation of both leading and subleading divergences in the second-order contribution to the self-energy.
In conclusion, it is shown that the expansion around the lower critical dimension $`d_l=1`$ for the superfluid-Bose glass critical point allows a field-theoretic formulation that facilitates a systematic higher-order calculation of the critical exponents. The computation to two lowest orders yields results for the correlation length critical exponents in $`d=2`$ in respectable agreement with the experiment and Monte Carlo calculations, both for the short-range and the Coulomb interaction between particles. The results suggest that the superconductor-insulator transition in homogeneous thin films is in the universality class of disordered bosons.
This work has been supported by NSERC of Canada. |
warning/0001/math0001022.html | ar5iv | text | # Random vicious walks and random matrices
## 1 Introduction
In , two types of random vicious walkers models, *random turn walker model* and *lock step walker model*, are considered. In these models, walkers are on one-dimensional integer lattice, and time is discrete. For their applications and earlier results, see, for example, and references therein. In this paper, we present results on lock step model showing a relation to random matrix theory. For random turn walker model, see and discussions following Theorem 1.1 below.
At time $`t=0`$, infinitely many walkers are located at the sites $`\{0,2,4,6,\mathrm{}\}`$. We label the walkers by $`P_1,P_2,P_3,\mathrm{}`$ from the left to the right. In the lock step model, at each time $`t=n`$, all the particles move either to their right or to their left with equal probability. The only constraint is that no two particles can occupy the same site at the same time. This is why the walkers are called “vicious”. One typical path configuration is shown in Figure 1.
This model can also be thought of as a certain totally asymmetric exclusion process in discrete time. Initially there are infinitely many particles at $`\{1,2,3,\mathrm{}\}`$. A particle is called *left-movable* if its left site is vacant. Particles $`P_{j+1},P_{j+2},\mathrm{},P_k`$ are called *successors* of a particle $`P_j`$ at a certain time if they are next to each other in the order of the indices. At each (discrete) time step, a left-movable particle either moves to its left site *together* with arbitrarily taken number of its successors, or stays at the same site with equal probability. It is easy to see that this process is equivalent to the above lock step model ; right move of lock step corresponds staying at the same site in the exclusion process. Figure 2 represent an exclusion process equivalent to the lock step model in Figure 1.
Suppose that during $`N`$ time steps, total $`k`$ left moves are made by all the particles. In the example of Figure 1, $`N=6`$ and $`k=14`$. We denote by $`\mathrm{P}(N,k)`$ the set of all path configurations during $`N`$ time steps with total $`k`$ left moves. Then each configuration has equal probability, $`1`$ over the cardinal of $`\mathrm{P}(N,k)`$. Hence our probability space is $`\mathrm{P}(N,k)`$ with uniform probability given by $`1/|\mathrm{P}(N,k)|`$. We denote by $`L_j(N,k)(\pi )`$ the number of left moves made by the particle $`P_j`$ in a path $`\pi \mathrm{P}(N,k)`$. We are interested in the limiting statistics of the random variables $`L_j(N,k)`$ as $`N,k\mathrm{}`$.
To state the main result, we need a definition. Let $`u(x)`$ be the solution of the differential equation
$$u_{xx}=2u^3+xu,u(x)Ai(x)\text{as}x+\mathrm{},$$
(1.1)
where $`Ai`$ is the Airy function. The above equation is called Painlevé II equation. It is known that there is unique global solution to (1.1) (see, e.g. and references in it). Define the function, called the *GOE Tracy-Widom distribution function*, by
$$F_1(x)=\mathrm{exp}\left\{\frac{1}{2}_x^{\mathrm{}}(sx)(u(s))^2𝑑s+\frac{1}{2}_x^{\mathrm{}}u(s)𝑑s\right\}.$$
(1.2)
This is indeed a distribution function, and the decay rate is given by
$`F_1(x)`$ $`=`$ $`1+O(e^{cx^{3/2}}),\text{as }x+\mathrm{}\text{,}`$ (1.3)
$`F_1(x)`$ $`=`$ $`O(e^{c|x|^3}),\text{as }x\mathrm{}\text{,}`$ (1.4)
for some $`c>0`$ (see, e.g. (2.11)-(2.14) of ). In , Tracy and Widom proved that $`F_1`$ is the limiting distribution of the properly centered and scaled largest eigenvalue of a random matrix taken from a Gaussian orthogonal ensemble. The subscription $`1`$ in $`F_1`$ is a general convention : there are also GUE and GSE Tracy-Widom distribution functions $`F_2`$ and $`F_4`$ . One can find general discussion for random matrices in .
Now the main theorem is
###### Theorem 1.1.
For fixed $`0<t<1`$, let
$$\eta (t)=\frac{2t}{1+t},\rho (t)=\frac{\left(t(1t)\right)^{1/3}}{1+t}.$$
(1.5)
Let $`F_1(x)`$ be the GOE Tracy-Widom distribution function. Under the condition that in $`N`$ time steps total $`k`$ left moves are made, the (conditional) probability distribution of the number $`L_1(N,k)`$ of left moves made by the leftmost particle satisfies
$$\underset{N\mathrm{}}{lim}\left(\frac{L_1(N,k)\eta (t)N}{\rho (t)N^{1/3}}x\right)=F_1(x),\text{when }k=\frac{t^2}{1t^2}N^2+o(N^{4/3}).$$
(1.6)
Also all the moments of the scaled random variable converge to the corresponding moments of the limiting random variable.
In other words, in the large $`N`$ limit, the (conditional) fluctuation of the displacement of the leftmost particle in lock step model is identical to the fluctuation of the largest eigenvalue of a random GOE matrix. Naturally we expect that the $`k^{\text{th}}`$ particle corresponds to the $`k^{\text{th}}`$ eigenvalue of random GOE matrix.
It is interesting to compare the above result with the results for random turn walker model. Initially there are infinitely many walkers $`Q_1,Q_2,Q_3,\mathrm{}`$ at the position $`\{1,2,3,\mathrm{}\}`$. We again call a walker *left-movable* if its left site is vacant. At each time, we select *one* walker at random among left-movable walkers, and move it to its left site. Hence there is one and only one movement at each time and all the movements are to the left. An example of random turn walker path configuration is in Figure 3.
Let $`X_j(N)`$ be the displacement of the $`j^{\text{th}}`$ walker after $`N`$ time step. It is shown in that for $`j=1,2`$, we have
$$\underset{N\mathrm{}}{lim}\left(\frac{X_j(N)2\sqrt{N}}{N^{1/6}}x\right)=F_1^{(j)}(x),j=1,2,$$
(1.7)
where $`F_1^{(j)}`$ is the limiting distribution of the (scaled) $`j^{\text{th}}`$ largest eigenvalue of a random GOE matrix. Especially we have $`F_1(x)=F_1^{(1)}(x)`$. On the contrary, if we assume that the walkers move to their left in the first $`N`$ time steps, and then move to their right in the next $`N`$ time steps so that at the end walkers come back to their original positions, then we obtain the GUE Tracy-Widom distribution in the limit . Indeed in this case, a lot more are known. The general $`j^{\text{th}}`$ row statistics and also the correlation functions converge to the corresponding quantities of random GUE matrix in the limit .
The first step to prove the above theorem is to map the path statistics into tableaux statistics following . By definition, a semistandard Young tableau (SSYT) is an array of positive integers top and left adjusted as in Figure 4 so that the numbers in each row increase weakly and the numbers in each column increase strictly. A reference for tableaux is , and we freely use the notations in it. In , Guttmann, Owczarek and Viennot established a simple bijection between path configurations of lock step model and the set of SSYT : for a path configuration, we write down the time steps at which the $`j^{\text{th}}`$ particle made movement to its left on the $`j^{\text{th}}`$ column. Hence the top row is the array of time steps the particles made first movement to their left, the second row is the array of time steps the particles made their second movement to their left, and so on. If we draw boxes around each number, the result is a SSYT. See figure 4 for the tableau corresponding to the path configuration of Figure 1.
This map is a bijection between $`\mathrm{P}(N,k)`$ and the set of SSYT of size $`k`$ with fillings taken from $`\{1,2,\mathrm{},N\}`$. Moreover, under this bijection, $`L_j(N,k)`$ is equal to the number of boxes in the $`j^{\text{th}}`$ column of the corresponding SSYT. Therefore the statistics of $`L_j`$ is identical to the $`j^{\text{th}}`$ column statistics of random tableaux.
After the work of Guttmann, Owczarek and Viennot, Forrester in observed that similar bijection can be established between path configuration of random turn model and the set of standard Young tableaux (SYT). The above result (1.7) is obtained based on this bijection and the recent work on statistics of the first row of random SYT. Also as mentioned above, if we assume that the walkers move to their left in the first $`N`$ time steps, and then move to their right in the next $`N`$ time steps so that at the end walkers come back to their original positions, the limiting fluctuation is not GOE type, but GUE type. This difference comes from the fact that in this case, the path configuration is in bijection with the *pairs* of SYT. The statistics of pairs of SYT is more well studied than that of single SYT , and we have stronger results mentioned earlier.
This paper is organized as follows. In Section 2, using the result of , we express the generating function for the probability of the first column of random tableaux in terms of a Hankel determinant. It is a general fact that Hankel determinant is related to orthogonal polynomials on the unit circle. The asymptotics of orthogonal polynomials is obtained via Riemann-Hilbert problem and summarized separately in Section 3. We can obtain the limiting statistics of the first column from the knowledge of Hankel determinant asymptotics. This work occupies the second half of Section 2. The proof of Theorem 1.1 is given at the end of Section 2.
Acknowledgments. The author would like to thank Percy Deift for his interest and encouragement.
## 2 Proof
Let $`d_\lambda (N)`$ be the number of semistandard Young tableaux of shape $`\lambda `$ with fillings taken from $`\{1.2.\mathrm{},N\}`$, and let $`\mathrm{}(\lambda )`$ be the number of rows of $`\lambda `$ (parts of $`\lambda `$, or the length of the first column). From the bijection of path configurations and tableaux, the number of path $`\pi \mathrm{P}(N,k)`$ satisfying $`L_1(N,k)(\pi )l`$ is equal to
$$\underset{\begin{array}{c}\lambda k\\ \mathrm{}(\lambda )l\end{array}}{}d_\lambda (N).$$
(2.1)
In our analysis (see also ), it turns out that in addition to the number of rows, the number of odd columns plays an important role in describing a tableau. For a partition $`\lambda `$, we define $`\lambda ^{}`$ to be the transpose of $`\lambda `$, $`f(\lambda )`$ to be the number of odd row in $`\lambda `$, and $`|\lambda |`$ to be the size of $`\lambda `$. Let $`b(N,j,m,l)`$ be the number of semistandard Young tableaux of size $`2j+m`$ with $`m`$ odd columns with at most $`l`$ columns with fillings taken from $`\{1,2,\mathrm{},N\}`$ :
$$b(N,j,m,l)=\underset{\begin{array}{c}\lambda 2j+m\\ f(\lambda ^{})=m\\ \mathrm{}(\lambda )l\end{array}}{}d_\lambda (N).$$
(2.2)
We use the notation $`b(N,j,m,\mathrm{})`$ for the sum above without restriction on $`\mathrm{}(\lambda )`$. Now we define a generating function
$$\begin{array}{cc}\hfill \varphi (N,l,t,\beta )& :=(1t^2)^{N(N1)/2}(1\beta t)^N\underset{\mathrm{}(\lambda )l}{}t^{|\lambda |}\beta ^{f(\lambda ^{})}d_\lambda (N)\hfill \\ & =(1t^2)^{N(N1)/2}(1\beta t)^N\underset{j,m0}{}t^{2j}(\beta t)^mb(N,j,m,l),\hfill \end{array}$$
(2.3)
where the sum in the first expression is taken over all the partitions $`\lambda `$ satisfying $`\mathrm{}(\lambda )l`$.
The starting point of our analysis is the following result of .
###### Lemma 2.1.
Let $`\varphi (N,l,t,\beta )`$ be defined as in (2.3). We have
$$\varphi (N,2l+1,t,\beta )=(1t^2)^{N(N1)/2}det(H_l),$$
(2.4)
where $`H_l=(h_{jk})_{0j,k<l}`$ is the $`l\times l`$ Hankel determinant with
$$h_{jk}=h_{jk}(N,t)=\frac{2^{j+k+1}}{\pi }_1^1x^{j+k}(1+t^22tx)^N(1x^2)^{1/2}𝑑x.$$
(2.5)
###### Remark.
Note that the right hand side of (2.4) does not depend on $`\beta `$.
###### Proof.
This proof is in in a slightly different form. Let $`s_\lambda (x)`$, $`x=(x_1,x_2,\mathrm{})`$, be the Schur function, and define $`H(u;y)`$ with $`y=(y_1,y_2,\mathrm{})`$ by
$$H(u;y)=\underset{j}{}(1uy_j)^1.$$
(2.6)
In (5.65) of , it is proved that
$$\underset{\mathrm{}(\lambda )2l+1}{}\beta ^{f(\lambda ^{})}s_\lambda (x)=H(\beta ;x)𝔼_{USp(2l)}det(H(U;x)),$$
(2.7)
which is an identity as a formal power series in $`x`$. But the combinatorial definition of the Schur function is (see, e.g. Chapter 7.10 of )
$$s_\lambda (x)=\underset{T}{}x_1^{\alpha _1(T)}x_2^{\alpha _2(T)}\mathrm{},$$
(2.8)
where the sum is over all semistandard Young tableaux $`T`$ of shape $`\lambda `$, and $`\alpha _j(T)`$ is the number of parts of $`T`$ equal to $`j`$ (type of $`T`$). Since $`_j\alpha _j(T)=|\lambda |`$, if we take the special case $`x=(t,t,\mathrm{},t,0,0,\mathrm{})`$ where the first $`N`$ elements are $`t`$ and the rest are $`0`$, then $`s_\lambda (x)`$ becomes
$$t^{|\lambda |}d_\lambda (N).$$
(2.9)
Hence for this special choice of $`x`$ (2.7) is now
$$\underset{\mathrm{}(\lambda )2l+1}{}\beta ^{f(\lambda ^{})}t^{|\lambda |}d_\lambda (N)=(1\beta t)^N𝔼_{USp(2l)}det((1tU)^N).$$
(2.10)
Now using Weyl’s integration formula for $`Sp(2l)`$ (see, e.g. ), the expectation in (2.10) becomes
$$𝔼_{USp(2l)}det((1tU)^N)=\frac{2^{l^2}}{l!(2\pi )^l}_{[0,2\pi ]^l}\underset{1j<kl}{}(\mathrm{cos}\theta _j\mathrm{cos}\theta _k)^2\underset{j=1}{\overset{l}{}}\mathrm{sin}^2\theta _j(1+t^22t\mathrm{cos}\theta _j)^Nd\theta _j.$$
(2.11)
Standard Vandermonde determinant manipulations yield
$$\frac{2^{l^2}}{(2\pi )^l}det\left(_0^{2\pi }\mathrm{cos}^{j+k}\theta \mathrm{sin}^2\theta (1+t^22t\mathrm{cos}\theta )^N𝑑\theta \right)_{0j,k<l},$$
(2.12)
which again after change of variables $`x=\mathrm{cos}\theta `$, is equal to
$$\frac{2^{l^2}}{\pi ^l}det\left(_1^1U_j(x)U_k(x)(1+t^22tx)^N(1x^2)^{1/2}𝑑x\right)_{0j,k<l}$$
(2.13)
where $`U_j(x)=\frac{\mathrm{sin}((j+1)\theta )}{\mathrm{sin}\theta }`$, $`x=\mathrm{cos}\theta `$, is the Chebyshev polynomial of the second kind. Note that $`U_j(x)=2^jx^j+\mathrm{}`$. Hence elementary row and column operations yield Lemma 2.1. ∎
Using this expression, we first obtain the asymptotic result for the generating function. The limit is insensitive to $`\beta `$ since so is $`\varphi `$.
###### Proposition 2.2.
Let $`0<t<1`$ and $`\beta >0`$ be fixed satisfying $`0<\beta t<1`$. For each $`l`$ and $`N`$, define $`x`$ by
$$x=\frac{l\eta (t)N}{\rho (t)N^{1/3}},$$
(2.14)
where $`\eta (t)`$ and $`\rho (t)`$ are defined in (1.5) Then there exits a positive constant $`M_0`$ such that for $`M>M_0`$, there are constants $`C,c>0`$, independent of $`M`$, and $`C(M)`$ which may depend on $`M`$ so that
$$|\varphi (N,l,t,\beta )F_1(x)|\frac{C(M)}{l^{1/3}}+Ce^{cM^{3/2}},M<x<M.$$
(2.15)
Also we have
$`01\varphi (N,l,t,\beta )Ce^{cx^{3/2}},`$ $`xM_0,`$ (2.16)
$`0\varphi (N,l,t,\beta )Ce^{c|x|^3},`$ $`xM_0.`$ (2.17)
###### Proof.
It is enough to consider the limit for $`\varphi (N,2l+1,t,\beta )`$ since from the definition (2.3), $`\varphi `$ is monotone increasing in $`l`$. First we related the determinant in (2.4) with certain quantities of orthogonal polynomials on the circle.
Let $`p_j(x)=x^j+\mathrm{}`$ be the $`j^{\text{th}}`$ monic orthogonal polynomial with respect to the weight $`w(x)dx=(1+t^22tx)^N(1x^2)^{1/2}dx`$ on the interval $`(1,1)`$, and let $`C_j`$ be the norm of $`p_j`$ :
$$_1^1p_j(x)p_k(x)w(x)𝑑x=C_j\delta _{jk}.$$
(2.18)
It is a well known result of orthogonal polynomial theory (see, e.g. ) that $`C_j=det(\stackrel{~}{H}_{j+1})/det(\stackrel{~}{H}_j)`$, where $`\stackrel{~}{H}_l=(\stackrel{~}{h}_{jk})_{0j,k<l}`$ with
$$\stackrel{~}{h}_{jk}=_1^1x^{j+k}(1+t^22tx)^N(1x^2)^{1/2}𝑑x.$$
(2.19)
which is equal to $`\frac{\pi }{2^{j+k+1}}h_{jk}`$ (recall (2.5)). Hence $`det(\stackrel{~}{H}_j)=\frac{\pi ^j}{2^{j^2}}det(H_j)`$. Since the Szegö strong limit theorem for Hankel determinants (see, e.g. ) implies that $`lim_l\mathrm{}det(H_l)=(1t^2)^{N(N1)/2}`$ for fixed $`N`$, we have
$$\varphi (N,2l+1,t,\beta )=\underset{k\mathrm{}}{lim}\underset{j=l}{\overset{k}{}}\frac{det(H_j)}{det(H_{j+1})}=\underset{j=l}{\overset{\mathrm{}}{}}\frac{\pi }{2^{2j+1}}C_j^1.$$
(2.20)
Now we use the relation between orthogonal polynomials on the unit circle and those on the interval $`(1,1)`$. Let $`\pi _j(z)=z^j+\mathrm{}`$ be the $`j^{\text{th}}`$ monic orthogonal polynomials on the unit circle $`\{|z|=1\}`$ with respect to the weight
$$\phi (z)\frac{dz}{2\pi i}=(1tz)^N(1tz^1)^N\frac{dz}{2\pi iz},$$
(2.21)
and let $`N_j`$ be the norm of $`\pi _j(z)`$ :
$$_{|z|=1}\pi _j(z)\overline{\pi _k(z)}\phi (z)\frac{dz}{2\pi iz}=N_j\delta _{jk}.$$
(2.22)
There is a simple relation between orthogonal polynomials $`p_j`$ on the unit circle and orthogonal polynomials $`\pi _j`$ on the interval (see the forth equation of (11.5.2) in ) :
$$C_j^{1/2}p_j(x)=\sqrt{\frac{2}{\pi }}\left(1+\pi _{2j+2}(0)\right)^{1/2}N_{2j+1}^{1/2}\frac{z^j\pi _{2j+1}(z)z^j\pi _{2j+1}(z^1)}{zz^1},x=\frac{1}{2}(z+z^1).$$
(2.23)
Especially comparing the coefficient of the leading term $`x^j`$, we have the relation
$$C_j=\frac{\pi }{2^{2j+1}}\left(1+\pi _{2j+2}(0)\right)N_{2j+1}.$$
(2.24)
But we also have $`(1\pi _{n+1}(0)^2)N_n=N_{n+1}`$ (see (11.3.6) in ). Hence (2.24) is equal to
$$C_j=\frac{\pi }{2^{2j+1}}\left(1\pi _{2j+2}(0)\right)^1N_{2j+2}.$$
(2.25)
Therefore (2.20) becomes
$$\varphi (N,2l+1,t,\beta )=\underset{j=l}{\overset{\mathrm{}}{}}\left(1\pi _{2j+2}(0)\right)N_{2j+2}^1.$$
(2.26)
This argument appeared in Corollary 2.7 of . Now using Proposition 3.2, computations similar to Lemma 7.1 of (also Corollary 7.2 and Corollary 7.6 of ) yield the result. ∎
Interpreting the notation $`b(N,j,m,\mathrm{})`$ as the sum without constraints on $`\mathrm{}(\lambda )`$ in (2.2), the number of path configuration in $`\mathrm{P}(N,k)`$ is equal to
$$|\mathrm{P}(N,k)|=\underset{2j+m=k}{}b(N,j,m,\mathrm{}),$$
(2.27)
and the probability of interest that $`L_1(N,k)l`$ for $`\pi \mathrm{P}(N,k)`$ is equal to
$$\frac{1}{|\mathrm{P}(N,k)|}\underset{2j+m=k}{}b(N,j,m,l)=\frac{1}{|\mathrm{P}(N,k)|}\underset{2j+m=k}{}p(N,j,m,l)b(N,j,m,\mathrm{}),$$
(2.28)
where
$$p(N,j,m,l)=\frac{b(N,j,m,l)}{b(N,j,m,\mathrm{})}.$$
(2.29)
For fixed $`N`$, as $`l\mathrm{}`$, the Szegö strong limit theorem for Hankel determinants (see, e.g. ) implies that (2.4) becomes $`1`$. Thus we have the identity
$$\underset{j,m0}{}t^{2j}(\beta t)^mb(N,j,m,\mathrm{})=(1t^2)^{N(N1)/2}(1\beta t)^N.$$
(2.30)
By taking Taylor expansion of the right hand side in $`t`$ and $`\beta `$, we obtain
$$b(N,j,m,\mathrm{})=\left(\genfrac{}{}{0pt}{}{\frac{N(N1)}{2}+j1}{j}\right)\left(\genfrac{}{}{0pt}{}{N+m1}{m}\right).$$
(2.31)
There is a more direct way to see this. See the remark after Lemma 2.6 below. Now from (2.27), the total number of paths in $`\mathrm{P}(N,k)`$ is equal to
$$|\mathrm{P}(N,k)|=\underset{2j+m=k}{}\left(\genfrac{}{}{0pt}{}{\frac{N(N1)}{2}+j1}{j}\right)\left(\genfrac{}{}{0pt}{}{N+m1}{m}\right).$$
(2.32)
Now it is straightforward to obtain the following result on the number of all paths.
###### Lemma 2.3.
Let $`0<t<1`$ and let
$$k=\left[\frac{t^2}{1t^2}N^2+o(N^{4/3})\right].$$
(2.33)
As $`N\mathrm{}`$, we have
$$|\mathrm{P}(N,k)|=\frac{\mathrm{exp}\left\{\frac{N^2t^2}{1t^2}\mathrm{log}t\mu N\mathrm{log}t\frac{1}{4}\left(\frac{\mu (1t^2)}{t}1\right)^2\right\}}{\sqrt{\pi }tN(1t)^N(1t^2)^{N(N1)/21}}\left(1+o(1)\right)$$
(2.34)
where the term $`o(1)`$ vanishes as $`N\mathrm{}`$, and $`\mu `$ is defined by
$$\mu :=\frac{k}{N}\frac{t^2}{1t^2}N$$
(2.35)
which is of order $`o(N^{1/3})`$ from (2.33). Moreover, the main contribution to the sum comes from $`|m\frac{t}{1t}N|N^{1/2+ϵ/2}`$ for some $`0<ϵ<\frac{1}{3}`$ ; precisely, there is a constant $`c>0`$ such that for any $`0<ϵ<\frac{1}{3}`$, we have
$$|\mathrm{P}(N,k)|=\left[\underset{\begin{array}{c}|m\frac{t}{1t}N|N^{1/2+ϵ/2}\\ km\text{ is even}\end{array}}{}b(N,\frac{km}{2},m,\mathrm{})\right]\left(1+O(e^{cN^ϵ})\right).$$
(2.36)
###### Proof.
From (2.32), we have
$$|\mathrm{P}(N,k)|=\underset{m=0}{\overset{[\frac{k}{2}]}{}}a(m),a(m)=\left(\genfrac{}{}{0pt}{}{\frac{N(N1)}{2}+\frac{km}{2}1}{\frac{km}{2}}\right)\left(\genfrac{}{}{0pt}{}{N+m1}{m}\right).$$
(2.37)
The ratio of $`a(m)`$ is
$$\frac{a(m+2)}{a(m)}=\frac{(N+m+1)(N+m)(km)}{(m+2)(m+1)(N(N1)+km2)}.$$
(2.38)
One can directly check that under the condition (2.33), the above ratio is decreasing in $`m`$, and becomes closest to $`1`$ at
$$m_c=\left[\frac{t}{1t}N+o(N^{1/3})\right].$$
(2.39)
Hence $`a(m)`$ is unimodal: it is increasing for $`m<m_c`$ and is decreasing for $`m>m_c`$. Now consider the neighborhood $`𝒩`$ of $`m_c`$ of size $`N^{1/2+ϵ/2}`$ for some fixed $`0<ϵ<\frac{1}{3}`$. For $`m`$ in $`𝒩`$, set
$$m=\frac{t}{1t}N+x,|x|N^{1/2+ϵ/2}.$$
(2.40)
For any $`M,x>0`$, Stirling’s formula yields
$$(M+x)!=\sqrt{2\pi M}M^{M+x}e^{M+\frac{x^2}{2M}}(1+O\left(\frac{x}{M}\right)+O(\frac{x^3}{M^2}\left)\right).$$
(2.41)
Using (2.33), (2.40) and (2.41), we have for $`m`$ in $`𝒩`$,
$$\left(\genfrac{}{}{0pt}{}{N+m1}{m}\right)=\frac{(\frac{1}{1t}N+x1)!}{(N1)!(\frac{t}{1t}N+x)!}=\frac{\mathrm{exp}\left\{(\frac{tN}{1t}+x)\mathrm{log}t\frac{(1t)^2x^2}{2tN}\right\}}{\sqrt{2\pi tN}(1t)^{N1}}\left(1+O(N^{\frac{1}{2}+\frac{3}{2}ϵ})\right)$$
(2.42)
and
$$\begin{array}{cc}& \left(\genfrac{}{}{0pt}{}{\frac{N(N1)}{2}+\frac{km}{2}1}{\frac{km}{2}}\right)=\frac{\left(\frac{N(N1)}{2(1t^2)}+(\mu \frac{t}{1t^2})\frac{N}{2}\frac{x}{2}1\right)!}{\left(\frac{N(N1)}{2}1\right)!\left(\frac{t^2N(N1)}{2(1t^2)}+(\mu \frac{t}{1t^2})\frac{N}{2}\frac{x}{2}\right)!}\hfill \\ & =\frac{\mathrm{exp}\left\{(\frac{t^2}{1t^2}N^2+(\mu \frac{t}{1t})Nx)\mathrm{log}t\frac{1}{4}(\frac{(1t^2)\mu }{t}1)^2\right\}}{\sqrt{\pi }tN(1t^2)^{N(N1)/21}}\left(1+o(1)\right).\hfill \end{array}$$
(2.43)
Thus we have
$$a(m)=\frac{\mathrm{exp}\left\{\frac{N^2t^2}{1t^2}\mathrm{log}tN\mu \mathrm{log}t\frac{1}{4}\left(\frac{\mu (1t^2)}{t}1\right)^2\frac{(1t)^2x^2}{2tN}\right\}}{\sqrt{2}\pi t^{3/2}N^{3/2}(1t)^N(1t^2)^{N(N1)/21}}\left(1+o(1)\right).$$
(2.44)
Let
$$|\mathrm{P}(N,k)|=\underset{}{}a(m)+\underset{}{}a(m),$$
(2.45)
where $``$ denotes the set $`𝒩`$ of $`m`$ satisfying (2.40) and $``$ denotes the rest of the range of $`m`$. From (2.44), the first sum over $``$ is equal to the right hand side of (2.34). Also from the unimodality, $`a(m)`$ in $``$ is less than or equal to the largest of $`a(m_+)`$ and $`a(m_{})`$ where $`m_\pm =[\frac{t}{1t}N\pm N^{1/2+ϵ}]`$. The number of summand in $``$ is of order $`N^2`$. Hence using (2.44) again for $`m_\pm `$, if we take $`c=\frac{(1t)^2}{4t}`$, for large $`N`$, we obtain
$$\underset{}{}a(m)=\left(\underset{}{}a(m)\right)e^{cN^ϵ},$$
(2.46)
which establishes the proof. ∎
Now we rewrite (2.3) as
$$\varphi (N,l,t,\beta )=(1t^2)^{N(N1)/2}(1\beta t)^N\underset{j,m0}{}t^{2j}(\beta t)^mb(N,j,m,\mathrm{})p(N,j,m,l).$$
(2.47)
The asymptotics of $`\varphi (N,l,t,\beta )`$ and $`p(N,j,m,l)`$ are related as follows.
###### Lemma 2.4.
For any $`d>0`$, there are constants $`C_0,c_0>0`$ such that for all $`l0`$,
$$p(N,\mu _+(N),\nu _+(N),l)\frac{C_0}{N^d}\varphi (N,l,t,\beta )p(N,\mu _{}(N),\nu _{}(N),l)+\frac{C_0}{N^d}$$
(2.48)
where
$`\mu _\pm (N)`$ $`=`$ $`\left[{\displaystyle \frac{t^2}{2(1t^2)}}N^2\pm c_0N\sqrt{\mathrm{log}N}\right],`$ (2.49)
$`\nu _\pm (N)`$ $`=`$ $`\left[{\displaystyle \frac{\beta t}{1\beta t}}N\pm c_0\sqrt{N\mathrm{log}N}\right].`$ (2.50)
The proof follows by using the following Lemma twice for $`j`$ and $`m`$ indices together with the Lemma 2.6. (Recall the (2.31)).
###### Lemma 2.5.
For a sequence $`\{q_j\}_{j0}`$, we define the following generating function
$$G(N)=(1a)^N\underset{j=0}{\overset{\mathrm{}}{}}a^j\left(\genfrac{}{}{0pt}{}{N+j1}{j}\right)q_j,0<a<1,N=1,2,\mathrm{}.$$
(2.51)
For each $`d>0`$, there are constants $`C_1,c_10`$ such that for any sequence $`\{q_j\}_{j0}`$ satisfying (i) $`q_jq_{j+1}`$ and (ii) $`0q_j1`$,
$$q_N^{}\frac{C_1}{N^d}G(N)q_N^{}+\frac{C_1}{N^d},N1,$$
(2.52)
where
$`N^{}`$ $`=`$ $`{\displaystyle \frac{a}{1a}}Nc_1\sqrt{N\mathrm{log}N},`$ (2.53)
$`N^{}`$ $`=`$ $`{\displaystyle \frac{a}{1a}}N+c_1\sqrt{N\mathrm{log}N}.`$ (2.54)
###### Proof.
This proof is parallel to that of the de-Poissonization lemma in . We have
$$G(N)=\underset{j=0}{\overset{\mathrm{}}{}}f_jq_j,f_j=(1a)^Na^j\left(\genfrac{}{}{0pt}{}{N+j1}{j}\right).$$
(2.55)
Stirling’s formula yields for $`n,m1`$,
$$\left(\genfrac{}{}{0pt}{}{m+n1}{m}\right)C\mathrm{exp}\{m\left[(1+\frac{n}{m}\mathrm{log}(1+\frac{n}{m})\frac{n}{m}\mathrm{log}\frac{m}{n}]\right\},$$
(2.56)
with some constant $`C`$. Thus we have
$$f_jC\mathrm{exp}\{Nh(j/N)\},h(x)=(1+x)\mathrm{log}(1+x)x\mathrm{log}x+x\mathrm{log}a+\mathrm{log}(1a).$$
(2.57)
One can directly check the following estimates of $`h`$ :
$`h(x)`$ $``$ $`{\displaystyle \frac{(1a)^2}{2a}}\left(x{\displaystyle \frac{a}{1a}}\right)^2,0x{\displaystyle \frac{2a}{1a}},`$ (2.58)
$`h(x)`$ $``$ $`\left[\mathrm{log}2{\displaystyle \frac{1+a}{2a}}\mathrm{log}(1+a)\right]x,x{\displaystyle \frac{2a}{1a}}.`$ (2.59)
We take a constant $`c_1>0`$ satisfying $`\frac{(1a)^2}{2a}c_1^21d`$. From (2.58) and the condition (ii), we have
$$\underset{jN^{}}{}f_jq_j\frac{a}{1a}CNe^{\frac{(1a)^2}{2a}c^2\mathrm{log}N}\frac{C}{N^d},$$
(2.60)
with a new constant $`C`$. Similarly,
$$\underset{N^{}j\frac{2a}{1a}N}{}f_jq_j\frac{C}{N^d}.$$
(2.61)
Also using (2.59), we have
$$\underset{j\frac{2a}{1a}N}{}f_jq_jC^{}e^{c^{}N}$$
(2.62)
for some constants $`c^{},C^{}`$. Thus we have
$$\left|G(N)\underset{N^{}jN^{}}{}f_jq_j\right|\frac{C}{N^d},$$
(2.63)
with a possibly different constant $`C`$. Now from the monotonicity condition (i), we have
$$\underset{N^{}jN^{}}{}f_jq_j\left(\underset{N^{}jN^{}}{}f_j\right)q_N^{}q_N^{},$$
(2.64)
and
$$\underset{N^{}jN^{}}{}f_jq_j\left(\underset{N^{}jN^{}}{}f_j\right)q_N^{}\left(1\frac{C}{N^d}\right)q_N^{}q_N^{}\frac{C}{N^d},$$
(2.65)
using the equality (2.63) for the second equality. Thus we obtained the desired result. ∎
To use the above Lemma to $`\varphi `$, we need monotonicity in $`l`$. It is more convenient now to view semistandard Young tableaux (SSYT) as generalized permutations. A two-rowed array
$$\pi =\left(\begin{array}{ccc}i_1& \mathrm{}& i_k\\ j_1& \mathrm{}& j_k\end{array}\right)$$
(2.66)
is called a generalized permutation if either $`i_r<i_{r+1}`$ or $`i_r=i_{r+1}`$, $`j_rj_{r+1}`$. Suppose the elements in the upper row of $`\pi `$ come from $`\{1,2,\mathrm{},M\}`$ and the elements in the bottom row come from $`\{1,2,\mathrm{},N\}`$. One can represent a generalized permutation as a $`M\times N`$ matrix $`(a_{ik})`$ where $`a_{ik}`$ is the number of times when $`\left(\genfrac{}{}{0pt}{}{i}{k}\right)`$ occurs in $`\pi `$. For example, the generalized permutation
$$\left(\begin{array}{ccccccccc}1& 1& 1& 2& 2& 2& 2& 3& 3\\ 1& 3& 3& 2& 2& 2& 4& 3& 4\end{array}\right)$$
(2.67)
corresponds to
$$\left(\begin{array}{cccc}1& 0& 2& 0\\ 0& 3& 0& 1\\ 0& 0& 1& 1\end{array}\right).$$
(2.68)
In the proof of the Lemma below, we regard a generalized permutation as a $`M\times N`$ square board with stacks of $`a_{ik}`$ balls in each position $`(i,k)`$. We denote by $`L(\pi )`$ the length of the longest strictly decreasing subsequence of $`\pi `$. In the example (2.67), $`L(\pi )=2`$.
Let $`M_{j,m}`$ be the set of $`N\times N`$ matrices $`\pi =(a_{ik})`$ which is symmetric $`a_{ik}=a_{ki}`$, and satisfies $`_{i=1}^Na_{ii}=m`$ and $`_{1i<kN}a_{ik}=j`$. This is a certain subset of the set of generalized permutations. The celebrated Robinson-Schensted-Knuth correspondence establishes a bijection between $`M_{j,m}`$ and the set of SSYT of size $`2j+m`$ with $`m`$ odd columns with fillings taken from $`\{1,2,\mathrm{},N\}`$. Moreover, under this bijection, $`L(\pi )`$ for $`\pi M_{j,m}`$ (viewed as a generalized permutation) is equal to the number of rows of the corresponding SSYT.
With this preliminary, we can prove the following.
###### Lemma 2.6 (monotonicity).
For any $`j,m0`$, we have
$$p(N,j+1,m,l)p(N,j,m,l),p(N,j,m+1,l)p(N,j,m,l).$$
(2.69)
###### Proof.
We first consider the second inequality. From (2.29), we need to show that
$$(m+1)b(N,j,m+1,l)(N+m)b(N,j,m,l).$$
(2.70)
By the definition (2.2) and the Robinson-Schensted-Knuth correspondence, $`b(N,j,m,l)`$ is equal to the number of $`\pi M_{j,m}`$ satisfying $`L(\pi )l`$.
Consider all possible distinct (strict) upper triangular parts of elements in $`M_{j,m}`$. It is equal to putting $`j`$ identical balls into $`N(N1)/2`$ boxes ; $`K=\left(\genfrac{}{}{0pt}{}{N(N1)/2+j1}{j}\right)`$ distinct ways. Hence we have a disjoint union $`M_{j,m}=_{i=1}^KS_{i,m}`$ where each $`S_{i,m}`$ consists of $`\pi M_{j,m}`$ with same upper triangular part, and elements in $`S_{i,m}`$ and $`S_{i^{},m}`$ have different upper triangular parts when $`ii^{}`$. Similarly, $`M_{j,m+1}=_{i=1}^KS_{i,m+1}^{}`$ where $`\sigma S_{i,m+1}^{}`$ has the upper triangular part same as that of $`\pi S_{i,m}`$.
Now for each $`\pi =(a_{rs})S_{i,m}`$, we generate $`N+m`$ elements in $`S_{i,m+1}^{}`$ as follows. For $`1rN`$, assign $`a_{rr}+1`$ identical $`\pi ^{}=(a_{kl}^{})`$ such that $`a_{rr}^{}=a_{rr}+1`$ and $`a_{kl}^{}=a_{kl}`$ for $`(k,l)(r,r)`$. (One can think this as adding a new ball in an array of $`a_{rr}`$ balls ; there are $`a_{rr}+1`$ ways.) Since $`_{1rN}a_{rr}+1=m+N`$, there result $`(m+N)|S_{i,m}|`$ (many identical) elements of $`S_{i,m+1}`$. Note that under this assignment,
$$L(\pi ^{})L(\pi ).$$
(2.71)
Now fix $`\sigma =(b_{kl})S_{i,m+1}`$. Since there are $`m+1`$ diagonal entries, there are exactly $`m+1`$ (many identical) elements of $`S_{i,m}`$ from which $`\sigma `$ is generated under the above assignment. (Considering each entry as a ball, each $`m+1`$ balls on the diagonal can be a newly added one.) Thus we have the identity $`(m+N)|S_{i,m}|=(m+1)|S_{i,m+1}|`$. Furthermore, from the remark regarding (2.71), we have $`(m+1)|R(i,m+1,l)|(m+N)|R_{i,m,l}|`$, where $`R_{i,m,l}`$ is the subset of $`\pi S_{i,m}`$ satisfying $`L(\pi )l`$. Therefore the second inequality in the Lemma is obtained.
The first inequality follows from a similar argument. ∎
###### Remark.
As mentioned before, using the generalized permutation interpretation of SSYT, we can see (2.31) directly. From non-negative integer matrix representation of generalized permutations, $`b(N,j,m,\mathrm{})`$ is the number of $`N\times N`$ matrices $`(a_{rs})`$ with non-negative integer entries such that $`_{r=1}^Na_{rr}=m`$ and $`_{1r<sN}a_{rs}=j`$. It is equivalent to placing $`m`$ identical balls of color 1 into $`N`$ boxes and $`j`$ identical balls of color 2 into $`N(N1)/2`$ boxes. Therefore we obtain (2.31).
Now we give the proof of the theorem.
###### proof of Theorem 1.1.
In (2.28), we have
$$(L_1(N,k)l)=\frac{1}{|P(N,k)|}\underset{2j+m=k}{}p(N,j,m,l)b(N,j,m,\mathrm{}).$$
(2.72)
where $`P(N,k)`$, $`p(N,j,m,l)`$ and $`b(N,j,m,\mathrm{})`$ are given in (2.32), (2.31) and (2.29), and $`b(N,j,m,l)`$ is given in (2.2). We split the above sum into two pieces. One part is the sum over (1) $`|m\frac{t}{1t}N|N^{1/2+ϵ/2}`$ and (2) the rest, where $`0<ϵ<\frac{1}{3}`$ is fixed. Then since $`0p(N,j,m,l)1`$, we have from (2.36)
$$\frac{1}{|P(N,k)|}\underset{(2)}{}p(N,j,m,l)b(N,j,m,\mathrm{})=O(e^{cN^ϵ})$$
(2.73)
for some $`c>0`$.
We use Lemma 2.4 to estimate $`p(N,j,m,l)`$ for $`(j,m)`$ in (1). Set
$`\stackrel{~}{t}^2={\displaystyle \frac{km2c_0N\sqrt{\mathrm{log}N}}{N^2+km2c_0N\sqrt{\mathrm{log}N}}},\beta ={\displaystyle \frac{mc_0\sqrt{N\mathrm{log}N}}{N+mc_0\sqrt{N\mathrm{log}N}}}\stackrel{~}{t}^1,`$ (2.74)
where $`k`$ satisfies the condition in (1.6), and we take $`\stackrel{~}{t}>0`$. For $`(j,m)`$ in (1), they satisfy
$`\stackrel{~}{t}=t+o(N^{2/3}),\beta =1+O(N^{1/2+ϵ/2}).`$ (2.75)
The first inequality of (2.48) yields $`p(N,j,m,l)\varphi (N,l,\stackrel{~}{t},\beta )+C_0N^d`$. Set $`\stackrel{~}{x}`$ by (2.14) where $`t`$ is replaced by $`\stackrel{~}{t}`$ and $`l`$ is given by $`l=[\eta (t)N+x\rho (t)N^{1/3}]`$. Let $`M\mathrm{?}M_0`$ satisfies $`M<2x<M`$ where $`M_0`$ is given in Proposition 2.2. From (2.75), we have
$$\stackrel{~}{x}=x+o(1),$$
(2.76)
Proposition 2.2 implies that
$$|\varphi (N,l,\stackrel{~}{t},\beta )F_1(\stackrel{~}{x})|\frac{C(M)}{l^{1/3}}+Ce^{cM^{3/2}},l=[\eta (t)N+x\rho (t)N^{1/3}]$$
(2.77)
for large $`N`$. Since $`F_1^{}(x)=\frac{1}{2}(u(x)+v(x))F_1(x)`$ is bounded for $`x`$, we have from (2.76) that $`F_1(\stackrel{~}{x})=F_1(x)+o(1)`$. Thus we have for large $`N`$,
$$p(N,j,m,l)F_1(x)+o(1),$$
(2.78)
where $`o(1)`$ term is independent of $`(j,m)`$ in (1) and vanishes as $`N\mathrm{}`$. Thus using Lemma 2.3, we have for large $`N`$,
$$\begin{array}{cc}\hfill (L_1(N,k)l)& \frac{1}{|P(N,k)|}\underset{(1)}{}(F_1(x)+o(1))b(N,j,m,\mathrm{})+O(e^{cN^{2ϵ}})\hfill \\ & =F_1(x)+o(1),l=[\eta (t)N+x\rho (t)N^{1/3}]\hfill \end{array}$$
(2.79)
Similarly, we obtain the lower bound using the second inequality of (2.48). Thus we proved (1.6).
The convergence of moments is also similar using (2.16) and (2.17) (cf. Section 8 of ). ∎
## 3 Asymptotics of orthogonal polynomials
This section is devoted to asymptotics of orthogonal polynomials used in the proof of Proposition 2.2. The key ingredient is the equilibrium measure (see ).
On the unit circle, the equilibrium measure $`d\mu _V(z)=\psi (\theta )\frac{d\theta }{2\pi }`$ for $`V(z)`$ and its support are uniquely determined by the following Euler-Lagrange variational conditions :
$$\begin{array}{cc}& \text{there exits a real constant }l\text{ such that},\hfill \\ & 2_\mathrm{\Sigma }\mathrm{log}|zs|d\mu _V(s)V(z)+l=0\text{for }z\overline{J},\hfill \\ & 2_\mathrm{\Sigma }\mathrm{log}|zs|d\mu _V(s)V(z)+l0\text{for }z\mathrm{\Sigma }\overline{J}.\hfill \end{array}$$
(3.1)
###### Lemma 3.1.
Let $`\gamma 0`$ and $`0<t<1`$, and let
$`V(z)`$ $`=`$ $`\gamma \mathrm{log}(1tz)(1tz^1).`$ (3.2)
Then their equilibrium measure $`\psi (\theta )d\theta /2\pi `$ is given as follows.
* When $`0\gamma \frac{1+t}{2t}`$, we have $`J=\mathrm{\Sigma }`$, $`l=0`$, and
$`\psi (\theta )`$ $`=`$ $`1\gamma +{\displaystyle \frac{\gamma (1t^2)}{1+t^22t\mathrm{cos}\theta }}`$ (3.3)
$`=`$ $`1+\gamma t\left({\displaystyle \frac{z}{1tz}}+{\displaystyle \frac{z^1}{1tz^1}}\right),z=e^{i\theta }.`$ (3.4)
* When $`\gamma >\frac{1+t}{2t}`$, $`J=\{e^{i\theta }:|\theta |\theta _c\}`$, where $`\mathrm{sin}^2\frac{\theta _c}{2}=\frac{(1t)^2(2\gamma 1)}{4t(\gamma 1)^2}`$, $`0<\theta _c<\pi `$, or
$$|\xi t|=\frac{\gamma (1t)}{\gamma 1},\xi =e^{i\theta _c},$$
(3.5)
and
$$l=2\gamma \mathrm{log}\frac{(2\gamma 1)(1t)}{2(\gamma 1)}\mathrm{log}\frac{(2\gamma 1)(1t)^2}{4t(\gamma 1)^2},$$
(3.6)
and finally
$`\psi (\theta )`$ $`=`$ $`{\displaystyle \frac{4(\gamma 1)\mathrm{cos}\frac{\theta }{2}}{\frac{(1t)^2}{t}+4\mathrm{sin}^2\frac{\theta }{2}}}\sqrt{\mathrm{sin}^2{\displaystyle \frac{\theta _c}{2}}\mathrm{sin}^2{\displaystyle \frac{\theta }{2}}}`$ (3.7)
$`=`$ $`{\displaystyle \frac{(\gamma 1)(1+z)}{(zt)(zt^1)}}\sqrt{(z\xi )(z\xi ^1)_+},z=e^{i\theta },`$ (3.8)
where $`\sqrt{(z\xi )(z\xi ^1)_+}`$ denotes the limit of $`z`$ from inside the unit circle. And in this case, the inequality of the second variational condition in (3.1) is strict for $`z\mathrm{\Sigma }\overline{J}`$.
###### Proof.
The proof given here is similar to the proof of Lemma 4.3 in , whose main ingredient is the following results of Lemma 4.2 in . Let $`d\mu (s)=u(\theta )d\theta `$ be an absolutely continuous probability measure on the unit circle $`\mathrm{\Sigma }`$ and $`u(\theta )=u(\theta )`$. Define
$$g(z)=_\mathrm{\Sigma }\mathrm{log}(zs)𝑑\mu (s),$$
(3.9)
where for fixed $`s=e^{i\theta _0}\mathrm{\Sigma }`$, $`\mathrm{log}(zs)`$ is defined to be analytic in $`\left((\mathrm{},1]\{e^{i\theta }:\pi \theta \theta _0\}\right)`$, and $`\mathrm{log}(zs)\mathrm{log}z`$ for $`z+\mathrm{}`$ with $`z`$. Then for $`z=e^{i\varphi }\mathrm{\Sigma }`$, we have
$`g_+(z)+g_{}(z)`$ $`=`$ $`2{\displaystyle _\mathrm{\Sigma }}\mathrm{log}|zs|d\mu (s)+i(\varphi +\pi ),`$ (3.10)
$`g_+(z)g_{}(z)`$ $`=`$ $`2\pi i{\displaystyle _\varphi ^\pi }u(\theta )𝑑\theta .`$ (3.11)
Also evenness of $`u`$ yields $`g(0)=\pi i`$.
* When $`0\gamma \frac{1+t}{2t}`$ : By residue calculation, it is easy to check $`_\pi ^\pi \psi (\theta )\frac{d\theta }{2\pi }=1`$. Define $`g(z)=_\pi ^\pi \mathrm{log}(ze^{i\theta })\psi (\theta )\frac{d\theta }{2\pi }`$ as in (3.9). Then by direct residue calculation, we obtain
$$g^{}(z)=\{\begin{array}{cc}\frac{\gamma t}{1tz},|z|<1,\hfill & \\ \frac{\gamma tz^2}{1tz^1}+\frac{1}{z},|z|>1.\hfill & \end{array}$$
(3.12)
Since $`g(0)=\pi i`$ and $`g(z)\mathrm{log}(z)`$ as $`z\mathrm{}`$, we have
$$g(z)=\{\begin{array}{cc}\gamma \mathrm{log}(1tz)+\pi i,|z|<1,\hfill & \\ \gamma \mathrm{log}(1tz^1)+\mathrm{log}z,|z|>1.\hfill & \end{array}$$
(3.13)
Thus, from (3.10), the variational condition (3.1) is satisfied with $`J=\mathrm{\Sigma }`$ and $`l=0`$.
* When $`\gamma >\frac{1+t}{2t}`$ : Set $`\beta (z)=\sqrt{(z\xi )(z\xi ^1)}`$ which is analytic in $`\overline{J}`$ and $`\beta (z)z`$ as $`z+\mathrm{}`$ with $`z`$. Then we have
$$\beta (0)=1,\beta (t)=|\xi t|,\beta (t^1)=\frac{1}{t}|\xi t|.$$
(3.14)
First, direct residue calculation using (3.14) shows that $`_{\theta _c}^{\theta _c}\psi (\theta )\frac{d\theta }{2\pi }=1`$, hence $`\psi (\theta )d\theta /(2\pi )`$ is a probability measure. Now we define $`g(z)=_{\theta _c}^{\theta _c}\mathrm{log}(ze^{i\theta })\psi (\theta )\frac{d\theta }{2\pi }`$ as before. Using (3.14) again, residue calculations yield
$$g^{}(z)=\frac{1}{2}\left(\frac{1}{z}+\frac{\gamma tz^2}{1tz^1}\frac{\gamma t}{1tz}\right)\frac{(\gamma 1)(z+1)}{2z(zt)(zt^1)}\beta (z).$$
(3.15)
Thus we have for $`|z|>1`$, $`z(\mathrm{},1)[t^1,\mathrm{})`$,
$$g(z)=\frac{\gamma }{2}\mathrm{log}(1tz)(1tz^1)+\frac{1}{2}\mathrm{log}z_{1_{+0}}^z\frac{(\gamma 1)(s+1)}{2s(st)(st^1)}\beta (s)𝑑s+g_{}(1)\gamma \mathrm{log}(1t),$$
(3.16)
and for $`|z|<1`$, $`z(1,t]`$,
$$g(z)=\frac{\gamma }{2}\mathrm{log}(1tz)(1tz^1)+\frac{1}{2}\mathrm{log}z_{1_0}^z\frac{(\gamma 1)(s+1)}{2s(st)(st^1)}\beta (s)𝑑s+g_+(1)\gamma \mathrm{log}(1t).$$
(3.17)
Now we compute $`g_+(1)+g_{}(1)`$. From (3.10) with $`z=1`$ and the evenness of $`\psi `$, we have
$$g_+(1)+g_{}(1)\pi i=\frac{8(\gamma 1)}{\pi }_0^{\theta _c}\mathrm{log}\left(2\mathrm{sin}\frac{\theta }{2}\right)\frac{\mathrm{cos}\frac{\theta }{2}}{\frac{(1t)^2}{t}+4\mathrm{sin}^2\frac{\theta }{2}}\sqrt{\mathrm{sin}^2\frac{\theta _c}{2}\mathrm{sin}^2\frac{\theta }{2}}𝑑\theta .$$
(3.18)
Substituting $`x=(\mathrm{sin}\frac{\theta _c}{2})^1\mathrm{sin}\frac{\theta }{2}`$, the above becomes
$$\frac{4(\gamma 1)}{\pi }_0^1\mathrm{log}\left(2\mathrm{sin}\frac{\theta _c}{2}x\right)\frac{\sqrt{1x^2}}{x^2+p^2}𝑑x,p:=\frac{1t}{2\sqrt{t}\mathrm{sin}\frac{\theta _c}{2}}=\frac{\gamma 1}{\sqrt{2\gamma 1}}.$$
(3.19)
Using $`\frac{2(\gamma 1)}{\pi }_0^1\frac{\sqrt{1x^2}}{x^2+p^2}𝑑x=1`$ which is a consequence of the fact that $`\psi (\theta )d\theta /(2\pi )`$ is a probability measure, we have
$$g_+(1)+g_{}(1)\pi i=2\mathrm{log}\left(2\mathrm{sin}\frac{\theta _c}{2}\right)+\frac{4(\gamma 1)}{\pi }_0^1\mathrm{log}(x)\frac{\sqrt{1x^2}}{x^2+p^2}𝑑x.$$
(3.20)
Now we use $`_0^1\mathrm{log}x\frac{dx}{\sqrt{1x^2}}=_0^{\pi /2}\mathrm{log}(\mathrm{sin}\theta )𝑑\theta =\frac{\pi }{2}\mathrm{log}2`$ to rewrite the last term in the above as
$$2(\gamma 1)\mathrm{log}2+\frac{4(\gamma 1)(1+p^2)}{\pi }_0^1\frac{\mathrm{log}x}{(x^2+p^2)\sqrt{1x^2}}𝑑x.$$
(3.21)
Also one can show by residue calculations that
$$_0^1\frac{\mathrm{log}x}{(x^2+p^2)\sqrt{1x^2}}𝑑x=\frac{\pi }{2}_1^{\mathrm{}}\frac{dx}{(x^2+p^2)\sqrt{x^21}}+\frac{\pi \mathrm{log}p}{2p\sqrt{1+p^2}}.$$
(3.22)
But one can directly verify that
$$\frac{dx}{(x^2+p^2)\sqrt{x^21}}=\frac{1}{2p\sqrt{1+p^2}}\mathrm{log}\left|\frac{p\sqrt{x^21}+\sqrt{1+p^2}x}{p\sqrt{x^21}\sqrt{1+p^2}x}\right|+C.$$
(3.23)
Thus from (3.20), (3.21) and the definition of $`p`$ in (3.19), we have
$$g_+(1)+g_{}(1)\pi i=2\gamma \mathrm{log}\frac{2(\gamma 1)}{2\gamma 1}+\mathrm{log}\frac{(2\gamma 1)(1t)^2}{4t(\gamma 1)^2},$$
(3.24)
which is equal to $`2\gamma \mathrm{log}(1t)l`$. Now for $`z\overline{J}`$, from (3.16) and (3.17), we have
$$g_+(z)+g_{}(z)=V(z)+\mathrm{log}zl+\pi i.$$
(3.25)
Thus (3.10) yields that the first variational condition (3.1) is satisfied. On the other hand, for $`z\mathrm{\Sigma }\overline{J}`$, $`g_+(z)+g_{}(z)`$ is equal to the right hand side of (3.25) plus
$$_{C_1}\frac{(\gamma 1)(s+1)}{2s(st)(st^1)}\beta (s)𝑑s,$$
(3.26)
where $`C_1=\{e^{i\theta }:\theta _c\theta \mathrm{arg}z\}`$ oriented from $`\xi `$ to $`z`$ if $`\mathrm{arg}z>0`$, and $`C_1=\{e^{i\theta }:\mathrm{arg}z\theta \theta _c\}`$ oriented from $`\xi ^1`$ to $`z`$ if $`\mathrm{arg}z<0`$. For $`\theta _c<\mathrm{arg}z<\pi `$, (3.26) is equal to
$$_{\theta _c}^{\mathrm{arg}z}\frac{\sqrt{2}(\gamma 1)\mathrm{cos}\frac{\theta }{2}}{t+t^12\mathrm{cos}\theta }\sqrt{\mathrm{cos}\theta _c\mathrm{cos}\theta }𝑑\theta <0,$$
(3.27)
and for $`\pi <\mathrm{arg}z<\theta _c`$, it is equal to
$$_{\mathrm{arg}z}^{\theta _c}\frac{\sqrt{2}(\gamma 1)\mathrm{cos}\frac{\theta }{2}}{t+t^12\mathrm{cos}\theta }\sqrt{\mathrm{cos}\theta _c\mathrm{cos}\theta }𝑑\theta <0.$$
(3.28)
Thus the second variational condition of (3.1) is satisfied and the inequality is strict.
For fixed $`0<t<1`$, we define a weight function on the unit circle $`\mathrm{\Sigma }`$ by
$$\phi (z;N):=(1tz)^N(1tz^1)^N.$$
(3.29)
Let $`\pi _n(z;N)=z^n+\mathrm{}`$ be the $`n^{\text{th}}`$ monic orthogonal polynomial with respect to the measure $`\phi (z;N)dz/(2\pi iz)`$ on the unit circle, and let $`N_n(N)`$ be the norm of $`\pi _n(z;N)`$ :
$$_\mathrm{\Sigma }\pi _n(z;N)\overline{\pi _m(z;N)}(1tz)^N(1tz^1)^N\frac{dz}{2\pi iz}=N_n(N)\delta _{nm}.$$
(3.30)
We also define
$$\pi _n^{}(z;N)=z^n\pi _n(z^1;N).$$
(3.31)
Define a $`2\times 2`$ matrix-valued function $`Y`$ of $`z\mathrm{\Sigma }`$ by
$$Y(z;n;N):=\left(\begin{array}{cc}\pi _n(z;N)& _\mathrm{\Sigma }\frac{\pi _n(s;N)}{sz}\frac{\phi (s;N)ds}{2\pi is^n}\\ N_{n1}(N)^1\pi _{n1}^{}(z;N)& N_{n1}(N)^1_\mathrm{\Sigma }\frac{\pi _{n1}^{}(s;N)}{sz}\frac{\psi (s;N)ds}{2\pi is^n}\end{array}\right),n1.$$
(3.32)
Then $`Y(;n;N)`$ solves the following Riemann-Hilbert problem (RHP) (see Lemma 4.1 in ) :
$$\{\begin{array}{cc}Y(z;n;N)\text{is analytic in}z\mathrm{\Sigma },\hfill & \\ Y_+(z;n;N)=Y_{}(z;n;N)\left(\begin{array}{cc}1& \frac{1}{z^n}\phi (z;N)\\ 0& 1\end{array}\right),\text{on}z\mathrm{\Sigma },\hfill & \\ Y(z;n;N)\left(\begin{array}{cc}z^n& 0\\ 0& z^n\end{array}\right)=I+O(\frac{1}{z})\text{as}z\mathrm{}.\hfill & \end{array}$$
(3.33)
Here the notation $`Y_+(z;n:N)`$ (resp., $`Y_{}`$) denotes the limit of $`Y(z^{};n;N)`$ as $`z^{}z`$ satisfying $`|z^{}|<1`$ (resp., $`|z^{}|>1`$). Note that $`n`$ and $`N`$ play the role of external parameters in the above RHP. One can easily show that the solution of the above RHP is unique, hence (3.32) is the unique solution of the above RHP. This RHP formulation of orthogonal polynomials on the unit circle is an adaptation of a result of Fokas, Its and Kitaev in where orthogonal polynomials on the real line are considered.
From (3.32), it is easy to check that
$`N_{n1}(N)^1`$ $`=`$ $`Y_{21}(0;n;N),`$ (3.34)
$`\pi _n(z;N)`$ $`=`$ $`Y_{11}(z;n;N),`$ (3.35)
$`\pi _n^{}(z;N)`$ $`=`$ $`z^nY_{11}(z^1;n;N)=Y_{21}(z;n+1;N)(Y_{21}(0;n+1;N))^1.`$ (3.36)
The asymptotics of the above quantities can be obtained by applying Deift-Zhou method for Riemann-Hilbert problem (3.33). A reference for Deift-Zhou method is . In and , similar asymptotics are obtained for different weight function $`e^{t(z+z^1)}`$ as $`t,n\mathrm{}`$. It is interesting to compare the following results with Proposition 5.1 in .
###### Proposition 3.2.
For each $`n`$ and $`N`$, define $`x`$ by
$$\frac{2t}{1+t}\frac{N}{n}=1\left[\frac{1t}{2(1+t)^2}\right]^{1/3}\frac{x}{n^{2/3}}.$$
(3.37)
Also let
$$l:=\frac{2N}{n}\mathrm{log}\frac{(2Nn)(1t)}{2(Nn)}\mathrm{log}\frac{n(2Nn)(1t)^2}{4t(Nn)^2}$$
(3.38)
and for $`\frac{N}{n}>\frac{1+t}{2t}`$, let
$$\mathrm{sin}\frac{\theta _c}{2}:=\frac{(1t)}{2(Nn)}\sqrt{\frac{n(2Nn)}{t}}.$$
(3.39)
There exits $`M_0>0`$ such that as $`n,N\mathrm{}`$, the following asymptotic results hold.
1. If $`0\frac{2t}{1+t}\frac{N}{n}a`$ for $`0<a<1`$, then
$$|N_{n1}(N)^11|,|\pi _n(0;N)|Ce^{cn},$$
(3.40)
for some constants $`C`$, $`c`$ which may depend on $`a`$.
2. If $`a\frac{2t}{1+t}\frac{N}{n}1Mn^{2/3}`$ for some $`M>M_0`$ and $`0<a<1`$, then
$$|N_{n1}(N)^11|,|\pi _n(0;N)|\frac{C}{n^{1/3}}e^{cx^{3/2}}$$
(3.41)
for some constant $`C`$ and $`c`$ depending on $`M`$.
3. If $`MxM`$ for some $`M>0`$, then
$$\left|N_{n1}(N)^11\left[\frac{2(1+t)^2}{1t}\right]^{1/3}\frac{v(x)}{n^{1/3}}\right|,\left|\pi _n(0;N)+(1)^n\left[\frac{2(1+t)^2}{1t}\right]^{1/3}\frac{u(x)}{n^{1/3}}\right|\frac{C}{n^{2/3}}$$
(3.42)
for some constant $`C`$ depending on $`M`$.
4. If $`1+Mn^{2/3}\frac{2t}{1+t}a`$ for some $`M>M_0`$ and $`a>1`$, then
$$\left|\frac{e^{nl}}{\mathrm{sin}\frac{\theta _c}{2}}N_{n1}(N)^11\right|,\left|\frac{(1)^n}{\mathrm{cos}\frac{\theta _c}{2}}\pi _n(0;n;N)1\right|\frac{C}{\frac{2t}{1+t}Nn}$$
(3.43)
for some constant $`C`$ depending on $`M`$.
5. If $`a\frac{2t}{1+t}`$ for some $`a>1`$,
$$\left|\frac{e^{nl}}{\mathrm{sin}\frac{\theta _c}{2}}N_{n1}(N)^11\right|,\left|\frac{(1)^n}{\mathrm{cos}\frac{\theta _c}{2}}\pi _n(0;n;N)1\right|\frac{C}{n}$$
(3.44)
for some constant $`C`$ depending on $`a`$.
###### Remark.
From the calculations analogous to Section 10 of , in addition to the above asymptotics results, we can obtain more results similar to those in Section 5 of . For example, suppose $`x`$ defined in (3.37) above satisfies $`c_1xc_2`$ for some constants $`c_1,c_2`$ (hence we are in the case (iii) of above proposition). For $`\alpha >0`$, define $`w`$ by
$$\alpha =1\left[\frac{2(1+t)^2}{1t}\right]^{1/3}\frac{2w}{n^{1/3}}.$$
(3.45)
Then we have for $`w>0`$ fixed,
$`\underset{n\mathrm{}}{lim}(1)^n(1+t\alpha )^N\pi _n(\alpha ;N)`$ $`=`$ $`m_{12}(iw;x),`$ (3.46)
$`\underset{n\mathrm{}}{lim}(1+t\alpha )^N\pi _n^{}(\alpha ;N)`$ $`=`$ $`m_{22}(iw;x),`$ (3.47)
and for $`w<0`$ fixed,
$`\underset{n\mathrm{}}{lim}(1)^n(1+t\alpha )^N\pi _n(\alpha ;N)`$ $`=`$ $`m_{11}(iw;x)e^{\frac{8}{3}w^32xw},`$ (3.48)
$`\underset{n\mathrm{}}{lim}(1+t\alpha )^N\pi _n^{}(\alpha ;N)`$ $`=`$ $`m_{21}(iw;x)e^{\frac{8}{3}w^32xw},`$ (3.49)
where $`m(z;x)`$ is the solution to the Riemann-Hilbert problem for Painlevé II equation with special monodromy data $`p=q=1`$, $`r=0`$ (see, e.g. (2.15) of ). These results are parallel to (5.21), (5.22), (5.25), (5.26) of . But in this paper, we only need Proposition 3.2 above.
We are not going to present the detail of the proof because the computation is parallel to that of Lemma 5.1 and Lemma 6.3 of (see also Proposition 5.1 of ) where the authors consider the same asymptotic problem with different weight function $`e^{\sqrt{\lambda }\mathrm{cos}\theta }`$. Instead we give some indication why we have Painlevé II function in the case (iii). Let us consider only when $`x>0`$. Define a $`2\times 2`$ matrix valued function $`m(z)`$ by
$`m(z):=\{\begin{array}{cc}Y(z;n;N)\left(\begin{array}{cc}0& (1tz)^N\\ (1tz)^N& 0\end{array}\right),\hfill & |z|<1\hfill \\ Y(z;n;N)\left(\begin{array}{cc}z^n(1tz^1)^N& 0\\ 0& z^n(1tz^1)^N\end{array}\right),\hfill & |z|>1,\hfill \end{array}`$ (3.50)
where $`Y`$ is defined in (3.32). From the RHP for $`Y`$, $`m`$ solves a new RHP : $`m`$ is analytic in $`z\mathrm{\Sigma }`$, $`m(z)=I+O(\frac{1}{z})`$ as $`z\mathrm{}`$, and satisfies the jump condition $`m_+(z)=m_{}(z)v(z)`$ on $`z\mathrm{\Sigma }`$ where
$$v(z)=\left(\begin{array}{cc}1& z^n(1tz)^N(1tz^1)^N\\ z^n(1tz)^N(1tz^1)^N& 0\end{array}\right).$$
(3.51)
The choice of $`m`$ above is related to the equilibrium measure in Lemma 3.1. The role of equilibrium measure in RHP for orthogonal polynomials is discussed in (see also ). The two RHP’s for $`Y`$ and $`m`$ are algebraically related and are equivalent in the sense that a solution to one RHP implies a solution to the other RHP. Note that the jump matrix $`v(z)`$ has the factorization $`v(z)=v_{}(z)v_+(z)`$ where
$$v_{}(z)=\left(\begin{array}{cc}1& 0\\ z^n(1tz)^N(1tz^1)^N& 1\end{array}\right),v_+(z)=\left(\begin{array}{cc}1& z^n(1tz)^N(1tz^1)^N\\ 0& 1\end{array}\right).$$
(3.52)
Hence by usual deformation technique of RHP, we can bring the matrix $`v_+`$ to a contour in $`|z|<1`$, and the matrix $`v_{}`$ to a contour in $`|z|>1`$. By the assumption of $`N`$ and $`n`$ in case (iii), except in a neighborhood of $`z=1`$, we can find a new contour where the off diagonal entries of $`v_\pm `$ decay exponentially as $`n\mathrm{}`$. Hence the main contribution to the RHP as $`n\mathrm{}`$ comes only from a neighborhood of $`z=1`$. This is exactly related to the fact that the support of the equilibrium measure is Lemma 3.1 has special point $`z=1`$ at which a new gap opens up when $`\gamma (=\frac{N}{n})=\frac{1+t}{2t}`$. Now let us focus on the neighborhood of $`z=1`$. The (12) entry of $`v`$ is $`e^{h(z)}`$ where
$$h(z)=n\mathrm{log}zN\mathrm{log}(1tz)+N\mathrm{log}(1tz^1).$$
(3.53)
Set $`z=1+s`$. Using the definition of $`x`$ in (3.37), expansion of (3.53) becomes
$$\begin{array}{cc}\hfill h(z)& =n\mathrm{log}(1)x\left[\frac{1t}{2(1+t)^2}\right]^{1/3}(n^{1/3}s)\frac{x}{2n^{1/3}}\left[\frac{1t}{2(1+t)^2}\right]^{1/3}(n^{1/3}s)^2+\frac{1t}{6(1+t)^2}(n^{1/3}s)^3+\mathrm{}\hfill \\ & n\mathrm{log}(1)2x\stackrel{~}{s}+\frac{8}{3}\stackrel{~}{s}^3\text{as }n\mathrm{}\text{,}\hfill \end{array}$$
(3.54)
where $`\stackrel{~}{s}:=\frac{1}{2}\left[\frac{1t}{2(1+t)^2}\right]^{1/3}(n^{1/3}s)`$. Thus we are lead to a RHP with the jump matrix
$$\stackrel{~}{v}(z)=\left(\begin{array}{cc}1& (1)^ne^{2(x\stackrel{~}{s}+\frac{4}{3}\stackrel{~}{s}^3)}\\ (1)^ne^{2(x\stackrel{~}{s}+\frac{4}{3}\stackrel{~}{s}^3)}& 0\end{array}\right)$$
(3.55)
on the line $`i`$ oriented from $`+i\mathrm{}`$ to $`i\mathrm{}`$. After rotation by $`\pi /2`$, this is precisely the jump matrix for the Painlevé II equation with parameter $`p=q=1`$, $`r=0`$ (see, e.g. (2.15) of : the term $`(1)^n`$ in off diagonal entries can be simply conjugated out). Thus the $`m`$, therefore $`Y`$, can be expressed in terms of Painlevé II solution in the limit $`n\mathrm{}`$ in the case (iii) with $`x>0`$. |
warning/0001/gr-qc0001061.html | ar5iv | text | # Boson Stars: Alternatives to primordial black holes?
## I Introduction
### A Dark matter — Issue of missing mass
The rotation velocities of spiral galaxies can be accurately measured from the Doppler effect. At large radii where the stellar surface brightness is falling exponentially, velocities are obtained for clouds of neutral hydrogen using the 21 cm hyperfine line. The resulting ‘rotation curves’ are found to be roughly flat out to the maximum observed radii of about 50 kpc, which implies an enclosed mass increasing linearly with radius. This mass profile is much more extended than the distribution of starlight, which typically converges within $`10`$ kpc; thus, the galaxies are presumed to be surrounded by extended halos of dark matter.
Perhaps the most compelling evidence for dark matter comes from clusters of galaxies. These are structures of about 1 Mpc size containing more than 100 galaxies, representing an overdensity of about a factor 1000 relative to the mean galaxy density. It is assumed that they are gravitationally bound since the time for galaxies to cross the cluster lasts only about 10% of the age of the Universe. The cluster masses are determined in several independent ways: First, the virial theorem uses the radial velocities of individual galaxies as ‘test particles’. Second, observations of hot gas at about $`10^7`$ K contained in the clusters, which is observed in X-rays via thermal bremsstrahlung. The gas temperature is derived from the X-ray spectrum, and the density profile from the map of the X-ray surface brightness. Assuming the gas is pressure-supported against the gravitational potential leads to a mass profile for the cluster. The third method is gravitational lensing of background objects by the cluster potential. There are two regimes: the ‘strong lensing’ regime at small radii, which leads to arcs and multiple images, and the weak lensing regime at large radii, which causes background galaxies to be preferentially stretched in the tangential direction. All three methods yield estimates for cluster masses which show that visible stars contribute only a few percent of the observed mass, and the hot X-ray gas only about 10–20%, hence, clusters are dominated by dark matter.
On the largest scales, there is further evidence for dark matter: ‘streaming motions’ of galaxies (e.g., towards nearby superclusters such as the “Great Attractor”) can be compared to maps of the galaxy density from redshift surveys to yield estimates of $`\mathrm{\Omega }`$. Here the theory is more straightforward since the density perturbations are still in the linear regime, but the observations are less secure. A similar estimate may be derived by comparing our Galaxy’s 600 km/s motion towards the Virgo cluster relative to the cosmic rest frame, confirmed by the observed temperature dipole in the cosmic microwave background (CMB).
Furthermore, it is possible to connect the observed large-scale structure in the galaxy distribution with the results of the CMB anisotropies if the universe is dominated by non-baryonic dark matter. Commonly, the present matter/energy density $`\mathrm{\Omega }_0=\mathrm{\Omega }_\mathrm{M}+\mathrm{\Omega }_\mathrm{\Lambda }`$ of the Universe is decomposed into two components. There is accumulating evidence for $`\mathrm{\Omega }_0=1\pm 0.2`$ and (total) matter density $`\mathrm{\Omega }_\mathrm{M}=0.4\pm 0.1`$ which implies a vacuum energy or dimensionless cosmological constant of $`\mathrm{\Omega }_\mathrm{\Lambda }=0.85\pm 0.2`$ , cf. . Theories of inflation prefer a flat Universe with $`\mathrm{\Omega }_0=1`$ as its most ‘natural’ value; this also requires non-baryonic dark matter.
### B Dark matter — Candidates
What are realistic candidates for dark matter? Hot gas appears to be excluded by limits on the Compton distortion of the blackbody CMB spectrum; atomic hydrogen due to 21 cm observations; and ordinary stars by faint star counts. Asteroids are very unlikely since stars do not process hydrogen into heavy elements very efficiently and hydrogen ‘snowballs’ should evaporate or lead to excessive cratering on the Moon. Black holes more massive than $`10^5`$ $`M_{}`$ would destroy small globular clusters by tidal effects.
Today’s most viable dark matter candidates fall into two broad classes: astrophysical size objects called MAssive Compact Halo Objects (MACHOs), and so-called Weakly Interacting Massive Particles (WIMPs). Actually these classes could possibly be interrelated, as we are going to show.
Several different objects belong to the first class: Jupiter-size brown dwarfs consisting of hydrogen and helium less massive than 0.08 $`M_{}`$ are the most prominent possibility. Below this limit, the central temperature is not sufficient in order to ignite hydrogen fusion, so these objects just radiate very weakly in the infrared due to gravitational contraction. Other MACHO candidates include stellar remnants such as cool white dwarfs, neutron stars, and primordial black holes or as a result of a supernova.
The WIMP candidates are the invisible axion (hypothesized to solve the strong CP problem or reemerging ‘universally’ in effective string Lagrangians), one of the neutrinos (provided it has a mass of about 10 eV), and the lightest supersymmetric particle, the neutralino, which is expected to be stable. All these have to have a very weak interaction, so that they could not be detected so far.
### C Gravitational microlensing
The conclusion of gravitational microlensing of stars within the Large Magellanic Cloud (LMC) is that MACHOs in the planetary mass range $`10^6`$ to 0.05 $`M_{}`$ do not contribute a substantial fraction of the Galactic dark halo. In the two-year data of the LMC events of the MACHO group 8 events could be detected which are well in excess of the predicted background of approximately 1.1 events arising from known stellar populations. Hence, MACHOs in a dark halo appear to be a natural explanation.
A statistical analysis of the galactic halo via microlensing suggests that MACHOs account for a significant part ($`>`$ 20%) of the total halo mass of our galaxy. Their most likely mass range seems to be in the range 0.3 – 0.8 $`M_{}`$, with an average mass of 0.5 $`M_{}`$, cf. . If the bulge is more massive than the standard halo model assumes, the average MACHO mass will be somewhat lower at $`0.1`$ $`M_{}`$.
This can be viewed as an indication that MACHOs form an distinct large class of old objects that cannot be easily extrapolated from any familiar stellar population, such as brown or white dwarfs.
However, there are some astrophysical difficulties with this interpretation, mainly arising from the estimated mass $`0.5`$ $`M_{}`$ for the lenses. These cannot be hydrogen-burning stars in the halo since such objects are limited to less than 3% of the halo mass by deep star counts . Modifying the halo model to slow down the lens velocities can reduce the implied lens mass somewhat, but probably not below the substellar limit 0.08 $`M_{}`$. Old white dwarfs have about the right mass and can evade the direct-detection constraints, but it is difficult to form them with high efficiency, and there may be problems with overproduction of metals and overproduction of light at high redshifts from the luminous stars which were the progenitors of the white dwarfs . Primordial black holes are a viable possibility, though one has to appeal to a coincidence to have them in a stellar mass range.
Due to these difficulties of getting MACHOs in the inferred mass range without violating other constraints, there have been a number of suggestions for explaining the LMC events without recourse to a dark population: most of these suggestions construct some non-standard distribution of ‘ordinary’ stars along the LMC line of sight. However, these proposals appear somewhat contrived , but can be tested observationally in the near future.
### D Boson stars or axion stars as alternatives?
It has been recently suggested that MACHOs could rather be primordial black holes formed during the early QCD epoch in the inflationary scenario. For cosmological dark matter, bound states of gravitational waves, so-called ‘gravitational geons’ built from spin–2 bosons, are also considered recently .
Since the standard model of elementary particles as well as their superstring extensions involve also Higgs type scalar fields, we will analyze here the alternative possibility that primordial boson stars account for this non-baryonic part of dark matter . Boson stars are descendants of the so-called geons of Wheeler , except that they are built from scalar particles (spin–0) instead of electromagnetic fields, i.e. spin–1 bosons. If scalar fields exist in nature, such localized soliton-type configurations kept together by their self-generated gravitational field can form within Einstein’s general relativity.
In the case of complex scalar fields, an absolutely stable branch of such non-topological solitons with conserved particle number exists. In the spherically symmetric case, we have shown via catastrophe theory that these boson stars have a stable branch with a wide range of masses and radii.
Kaup’s first investigation of the spherically symmetric boson star (BS) was based on massive scalar particles. Lateron, a nonlinear $`U(|\mathrm{\Phi }|^2)`$ potential was introduced by Mielke and Scherzer , where also solutions with nodes, i.e. “principal quantum number” $`n>1`$, were found. In building macroscopic boson stars, a nonlinear Higgs type self-interaction potential $`U(|\mathrm{\Phi }|^2)`$ was later considered as an additional repulsive interaction. Thereby the Kaup limit for boson stars can even exceed the limiting mass of 3.23 $`M_{}`$ for neutron stars .
Three surveys summarize the present status of the non-rotating case, a more recent survey including the rotating BS can be found in .
Recently, we construct for the first time the corresponding localized differentially rotating configurations via numerical integration of the coupled Einstein–Klein–Gordon equations. Due to gravito–magnetic effect, the ratio of conserved angular momentum and particle number turns out to be an integer $`a`$, the azimuthal quantum number of our soliton–type stars. The resulting axisymmetric metric, the energy density and the Tolman mass are completely regular.
### E Are fundamental scalar fields part of nature?
The physical nature of the spin–0–particles out of which the boson star (BS) consists is still an open issue. Until now, no fundamental elementary scalar particle has been found in accelerator experiments which could serve as the main constituent of the boson star. In the electroweak theory of Glashow, Weinberg, and Salam, a Higgs boson dublett $`(\mathrm{\Phi }^+,\mathrm{\Phi }^0)`$ and its anti-dublett $`(\mathrm{\Phi }^{},\overline{\mathrm{\Phi }}^0)`$ are necessary ingredients in order to generate masses for the $`W^\pm `$ and $`Z^0`$ gauge vector bosons . After symmetry breaking, only one scalar particle, the Higgs particle $`h:=(\mathrm{\Phi }^0+\overline{\mathrm{\Phi }}^0)/\sqrt{2}`$, remains free and occurs as a state in a constant scalar field background. Nowadays, calculations of the two–loop electroweak effects enhanced by powers of the mass of the rather heavy top quark has lead to an indirect determination of the Higgs mass, cf. . For $`M_\mathrm{t}=173.8\pm 5`$ GeV/$`c^2`$, one finds $`m_\mathrm{h}=104_{49}^{+93}`$ GeV/$`c^2`$. So far, the experimental constraints are weak; even for the unrealistic case of a Higgs mass up to 1000 GeV/$`c^2`$, the discrepancies for, e.g., the mass of the $`W`$ boson would be small. Above 1.2 TeV/$`c^2`$, however, the self–interaction $`U(\mathrm{\Phi })`$ of the Higgs field is so large that the perturbative approach of the standard model becomes unreliable. Therefore a conformal extension of the standard model with gravity included could be necessary, see . Fermilab’s upgraded tevatron has a mass reach of $`135<m_\mathrm{h}<186`$ GeV/c<sup>2</sup>, while the high–energy experiments at the LHC at Cern will ultimately reveal if these Higgs particles really exist in nature.
As free particles, the Higgs boson is unstable with respect to the decays $`hW^++W^{}`$ and $`hZ^0+Z^0`$. In an hypothetical compact object like the BS, these decay channels are expected to be in equilibrium with the inverse process $`Z^0+Z^0h`$, for instance. Presumably, this is in full analogy with the neutron star or quark star , where one finds an equilibrium of $`\beta `$– and inverse $`\beta `$–decay of the neutrons or quarks and thus stability of the macroscopic star with respect to radioactive decay. Such Higgs sector nontopological solitons may also be candidates for cold dark matter. Nishimura and Yamaguchi constructed a neutron star using an equation of state of an isotropic fluid built from Higgs bosons.
Nowadays there are many attempts of unifying the standard model with gravity on the quantum level, like string theory . Commonly, the four–dimensional effective models make the prediction that the tensor field $`g_{\mu \nu }`$ of gravity is accompanied by one or several scalar fields.
In string effective supergravities , the mass of the dilaton $`\phi `$ can be related to the supersymmetry breaking scale $`m_{\mathrm{SUSY}}`$ by $`m_\phi 10^3(m_{\mathrm{SUSY}}/`$ TeV/c$`{}_{}{}^{2})^2`$ eV/c<sup>2</sup> with interesting astrophysical implications , but this is not the only possibility.
The other scalar field of the effective string Lagrangian is the ‘universal’ invisible axion $`\sigma `$, a pseudo–scalar potential for the Kalb–Ramond three form $`H:=e^\phi {}_{}{}^{}d\sigma `$. (There are some speculations to identify it with the the axial part of a possible torsion of spacetime). From the Hubble scale $`H_{\mathrm{eq}}10^{27}`$ eV/c<sup>2</sup> of matter–radiation equilibrium and the temperature $`T_\mathrm{m}100`$ MeV of mass generation at the epoch of chiral symmetry breaking, one can derive the condition $`m_\sigma >(T_\mathrm{m}/\mathrm{eV})^2H_{\mathrm{eq}}`$. This allows a very light axion mass $`m_\sigma =7.4\times (10^7\mathrm{GeV}/f_\sigma )`$ eV/c<sup>2</sup> $`>10^{11}`$ eV/c<sup>2</sup> with decay constant $`f_\sigma `$ close to the inverse Planck time, thus a prime candidate for dark matter. (This should not be confused with the Goldstone boson $`a`$ of the Peccei–Quinn symmetry of standard QCD, for which a recent experiment has excluded the range of $`m_\mathrm{a}10^6`$ eV/c<sup>2</sup>. From cooling neutron stars, there can be inferred an upper limit of $`m_\mathrm{a}<0.06`$ – 0.3 eV/c<sup>2</sup>, depending on the equation of state of the nucleon fluid.
## II Boson stars
In a 1968 perspective paper, Kaup has studied for the first time the full generally relativistic coupling of linear Klein–Gordon fields to gravity in a localized configuration. This ‘Klein–Gordon geon’ is nowadays christened mini–boson star and can be regarded as a macroscopic quantum state. It was already realized that no Schwarzschild type event horizon occurs in such numerical solutions. Moreover, the problem of the stability of the resulting scalar ‘geons’ with respect to radial perturbations is treated. It is shown that such objects are, below a well-defined critical mass, resistant to gravitational collapse. These considerations are on a semiclassical level, since the Klein–Gordon field is treated as a classical field.
The Lagrangian density of gravitationally coupled complex scalar field $`\mathrm{\Phi }`$ reads
$$_{\mathrm{BS}}=\frac{\sqrt{g}}{2\kappa }\{R+\kappa [g^{\mu \nu }(_\mu \mathrm{\Phi }^{})(_\nu \mathrm{\Phi })U(\mathrm{\Phi }^2)]\},$$
(1)
where $`\kappa =8\pi G`$ is the gravitational constant. Using the principle of variation, one finds the coupled Einstein–Klein–Gordon equations
$`G_{\mu \nu }:=R_{\mu \nu }{\displaystyle \frac{1}{2}}g_{\mu \nu }R`$ $`=`$ $`\kappa T_{\mu \nu }(\mathrm{\Phi }),`$ (2)
$`\left(\mathrm{}+{\displaystyle \frac{dU}{d\mathrm{\Phi }^2}}\right)\mathrm{\Phi }`$ $`=`$ $`0,`$ (3)
where
$$T_{\mu \nu }(\mathrm{\Phi })=\frac{1}{2}[(_\mu \mathrm{\Phi }^{})(_\nu \mathrm{\Phi })+(_\mu \mathrm{\Phi })(_\nu \mathrm{\Phi }^{})]g_{\mu \nu }(\mathrm{\Phi })/\sqrt{g}$$
(4)
is the stress–energy tensor and $`\mathrm{}:=\left(1/\sqrt{g}\right)_\mu \left(\sqrt{g}g^{\mu \nu }_\nu \right)`$ the generally covariant d’Alembertian.
### A Spherically symmetric solutions
The stationarity ansatz
$$\mathrm{\Phi }(r,t)=P(r)e^{i\omega t}$$
(5)
describes a spherically symmetric bound state of the scalar field with frequency $`\omega `$.
In the case of spherical symmetry and isotropic cordinates, the line-element reads
$$ds^2=e^{\nu (r)}dt^2e^{\lambda (r)}\left[dr^2+r^2\left(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2\right)\right],$$
(6)
in which the functions $`\nu =\nu (r)`$ and $`\lambda =\lambda (r)`$ depend on the radial coordinate $`r:=\sqrt{x^2+y^2+z^2}`$.
In a nut–shell, a boson star is a stationary solution of a (non-linear) Klein–Gordon equation in its own gravitational field; cf. . As in the case of a prescribed Schwarzschild background , the spacetime curvature affects the resulting radial Schrödinger equation
$$\left[_{r^2}V_{\mathrm{eff}}(r)+\omega ^2m^2\right]P=0$$
(7)
for the radial function $`P(r)`$ essentially via an effective gravitational potential $`V_{\mathrm{eff}}(r)`$, when written in terms of the tortoise coordinate $`dr^{}:=e^{(\lambda \nu )/2}dr`$ . Then it can be easily realized that localized solutions fall off asymptotically as $`P(r)(1/r)\mathrm{exp}\left(\sqrt{m^2\omega ^2}r\right)`$ in a Schwarzschild-type asymptotic background.
The energy–momentum tensor becomes diagonal, i.e. $`T_\mu {}_{}{}^{\nu }(\mathrm{\Phi })=\mathrm{diag}(\rho ,p_r,`$ $`p_{},p_{})`$ with
$`\rho `$ $`=`$ $`{\displaystyle \frac{1}{2}}(\omega ^2P^2e^\nu +P^2e^\lambda +U),`$ (8)
$`p_r`$ $`=`$ $`\rho U,`$ (9)
$`p_{}`$ $`=`$ $`p_rP^2e^\lambda .`$ (10)
This form is familiar from fluids, except that the radial and tangential pressure generated by the scalar field are in general different, i.e. $`p_rp_{}`$, due to the different sign of $`(P^{})^2`$ in these expressions.
In general, the resulting system of three coupled nonlinear equations for the radial parts of the scalar and the (strong) gravitational tensor field has to be solved numerically. (Exact solution of massless scalar fields or of the coupled Maxwell–Einstein–Klein–Gordon equation tend to be plagued with singularities.)
In order to specify the starting values for the ensuing numerical analysis, asymptotic solutions at the origin and at spatial infinity are instrumental. The resulting configuration turns out to be completely regular and does not exhibit an apparent event horizon, cf. .
The stress–energy tensor of a BS, unlike a classical fluid, is in general anisotropic . In contrast to neutron stars , where the ideal fluid approximation demands an isotropic symmetry for the pressure, for spherically symmetric boson stars there are different stresses $`p_r`$ and $`p_{}`$ in radial or tangential directions, respectively. The fractional anisotropy $`a_f:=(p_rp_{})/p_r=P^2e^\lambda /(\rho U)`$ depends essentially on the self-interaction; cf. Ref. . So, the perfect fluid approximation is inadequate for boson stars.
There exists a decisive difference between self–gravitating objects made of fermions or bosons: For a many fermion system the Pauli exclusion principle forces the typical fermion into a state with very high quantum number, whereas many bosons can coexist all in the same ground state (Bose–Einstein condensation). This also reflects itself in the critical particle number $`N:=_0^{\mathrm{}}j^0𝑑r`$ of stable configurations:
* $`N_{\mathrm{crit}}(M_{\mathrm{Pl}}/m)^3`$ for fermions
* $`N_{\mathrm{crit}}(M_{\mathrm{Pl}}/m)^2`$ for massive bosons without self–interaction.
Cold mixed boson–fermion stars have been studied by Henrique et al. and Jetzer .
### B Critical masses of boson stars
Since boson stars are macroscopic quantum states, below a certain critical mass $`M_{\mathrm{crit}}`$ they are prevented from complete gravitational collapse by the Heisenberg uncertainty principle $`\mathrm{\Delta }r\mathrm{\Delta }p\pi \mathrm{}`$, cf. Ref.. This provides us also with crude mass estimates: For a boson to be confined within the star of effective radius $`R_{\mathrm{eff}}:=(1/N)_0^{\mathrm{}}j^0r𝑑r`$, the Compton wavelength of the collective boson has to satisfy $`\lambda _\mathrm{\Phi }=(2\pi \mathrm{}/mc)2R_{\mathrm{eff}}`$. On the other hand, the star’s radius should be of the order of the last stable Kepler orbit $`3R_\mathrm{S}`$ around a black hole of Schwarzschild radius $`R_\mathrm{S}:=2GM/c^2`$ in order to avoid an instability against complete gravitational collapse.
For a mini-boson star, i.e. a massive boson model with merely the mass term $`U(|\mathrm{\Phi }|)=m^2|\mathrm{\Phi }|^2`$ as self-interaction, of an effective radius $`R_{\mathrm{eff}}(\pi /2)^2R_\mathrm{S}`$ close to the last stable Kepler orbit of a black hole, one obtains the estimate
$$M_{\mathrm{crit}}(2/\pi )M_{\mathrm{Pl}}^2/m0.633M_{\mathrm{Pl}}^2/m=M_{\mathrm{Kaup}},$$
(11)
cf. Ref. . This provides us with a rather good upper bound on the so-called Kaup limit. The correct value in the second expression was found numerically as a limit of the maximal or critical mass of a stable mini–BS. Here $`M_{\mathrm{Pl}}:=\sqrt{\mathrm{}c/G}`$ is the Planck mass and $`m`$ the mass of a bosonic particle.
For the likely mass $`m_\mathrm{h}=100`$ GeV/c<sup>2</sup> of the Higgs particle, e.g., one can estimate the total mass of this mini-boson star to be $`M10^{10}`$ kg and its radius $`R_{\mathrm{eff}}10^{18}`$ m yielding a density 10<sup>45</sup> times that of a neutron star.
A boson star is an extremely dense object, since non-interacting scalar matter is very “soft”. However, these properties are changed considerably by considering a repulsive self-interaction
$$U(|\mathrm{\Phi }|)=m^2|\mathrm{\Phi }|^2\left(1+\frac{1}{8}\mathrm{\Lambda }(|\mathrm{\Phi }|^2)\right)=m_{\mathrm{ren}}^2|\mathrm{\Phi }_{\mathrm{ren}}|^2.$$
(12)
where $`\mathrm{\Lambda }(|\mathrm{\Phi }|^2)`$ denotes an arbitrary nonlinear self-interaction. The choice $`\mathrm{\Lambda }(|\mathrm{\Phi }|^2)=(4\lambda /m^2)|\mathrm{\Phi }|^2`$ would lead us back to the quartic self-interaction of Colpi et al. . If we adopt the value $`|\mathrm{\Phi }_0|M_{\mathrm{Pl}}/\sqrt{16\pi }`$ inside the boson star, one finds for the energy density
$$\rho m^2|\mathrm{\Phi }_0|^2\left(1+\mathrm{\Lambda }/8\right),$$
(13)
where $`\mathrm{\Lambda }:=\mathrm{\Lambda }(M_{\mathrm{Pl}}^2/16\pi )`$ is a dimensionless coupling constant such that we would recover $`\mathrm{\Lambda }:=(\lambda M_{\mathrm{Pl}}^2/4\pi m^2)`$ for the quartic interaction. The self-interaction becomes dominating for $`\mathrm{\Lambda }8`$ or $`\lambda 32\pi (m/M_{\mathrm{Pl}})^2`$. Thus, even a rather tiny coupling constant $`\lambda `$ could have drastic effects on a BS.
Formally, this corresponds to a star formed from non-interacting bosons $`\mathrm{\Phi }_{\mathrm{ren}}=\mathrm{\Phi }(1+\mathrm{\Lambda }/8)`$ with a lower renormalized mass $`mm_{\mathrm{ren}}:=m/\sqrt{1+\mathrm{\Lambda }/8}`$ but larger Compton wave length $`\lambda _{\mathrm{\Phi }(\mathrm{ren})}`$ and, consequently, a larger radius $`R_{\mathrm{eff}}`$. (A reverse rescaling of the mass, as was presumed in a recent preprint , leads to a smaller Compton wave length and other inconsistencies.) Consequently, we can apply again (11) for the Kaup limit and find that the maximal mass of a stable BS scales approximately as
$$M_{\mathrm{crit}}(2/\pi )M_{\mathrm{Pl}}^2/m_{\mathrm{ren}}=\frac{2}{\pi }\sqrt{1+\mathrm{\Lambda }/8}\frac{M_{\mathrm{Pl}}^2}{m}\frac{1}{\pi \sqrt{2}}\sqrt{\mathrm{\Lambda }}\frac{M_{\mathrm{Pl}}^2}{m}\mathrm{for}\mathrm{\Lambda }\mathrm{}.$$
(14)
For the quartic self-interaction, this accounts rather well for the numerical results of Colpi et al. . Our formula not only reproduces their asymptotic mass formula (11) for $`\mathrm{\Lambda }\mathrm{}`$, but, by construction, interpolates as well with the Kaup limit (11) for $`\mathrm{\Lambda }=0`$.
Compact Critical mass Particle Number Object $`M_{\mathrm{crit}}`$ $`N_{\mathrm{crit}}`$ Fermion Star: $`M_{\mathrm{Ch}}:=M_{\mathrm{Pl}}^3/m^2`$ $`(M_{\mathrm{Pl}}/m)^3`$ Mini–BS: $`M_{\mathrm{Kaup}}=0.633M_{\mathrm{Pl}}^2/m`$ $`0.653(M_{\mathrm{Pl}}/m)^2`$ Boson Star: $`(1/\sqrt{2\pi })^3\sqrt{\lambda }M_{\mathrm{Ch}}`$ $`(M_{\mathrm{Pl}}/m)^3`$ Soliton Star: $`10^2(M_{\mathrm{Pl}}^4/m\mathrm{\Phi }_0^2)`$ $`2\times 10^3(M_{\mathrm{Pl}}^5/m^2\mathrm{\Phi }_0^3)`$
The Chandrasekhar limit is $`M_{\mathrm{Ch}}:=M_{\mathrm{Pl}}^3/m^21.5(\mathrm{GeV}/\mathrm{mc}^2)^2`$ $`M_{}`$, where $`M_{}`$ denotes the solar mass. In astrophysical terms, the maximal BS mass is $`M_{\mathrm{crit}}0.06\sqrt{\lambda }M_{\mathrm{Pl}}^3/m^2`$ $`=0.1\sqrt{\lambda }`$ (GeV/mc$`{}_{}{}^{2})^2`$ $`M_{}`$ which for $`\lambda =1`$ and proton mass $`m1`$ GeV/c<sup>2</sup> is in the interesting mass range $`0.1`$ $`M_{}`$ of MACHO’s.
In a scale-invariant theory built from nonlinearly coupled dilatons $`\phi `$, there arise a conserved dilaton charge via Noether’s theorem from Weyl rescaling and thus will ensure the stability of the configuration. For a dilaton star with quadratic self-interaction , the same formula (14) applies, but the coupling constant $`\mathrm{\Lambda }:=(\lambda M/4\pi \omega )^2`$ will be $`\omega `$ dependent. For a very light dilaton $`\phi `$ of mass $`m_{\mathrm{dil}}=10^{11}`$ eV/c<sup>2</sup>, resembling a misaligned ‘universal’ axion at its lower mass bound, Gradwohl and Kälbermann found
$$M_{\mathrm{crit}}=7\sqrt{\overline{\lambda }}M_{},R_{\mathrm{crit}}=40\sqrt{\overline{\lambda }}\mathrm{km},$$
(15)
where $`\overline{\lambda }:=\lambda (M/\omega )^2`$ is the rescaled coupling constant of the $`\phi ^4`$ interaction.
To repeat, in building macroscopic boson a Higgs–type self-interaction $`U(|\mathrm{\Phi }|)`$ is crucial for accommodating a repulsive force besides gravity. This repulsion between the constituents is instrumental to blow up the boson star so that much more particles will have room in the confined region. Thus the maximal mass of a BS can reach or even extend the limiting mass of 3.23 $`M_{}`$ for neutron stars with realistic equations of state $`p=p(\rho )`$ for which the (phase) velocity of sound is $`v_s=\sqrt{dp/d\rho }c`$. However, this fact depends on the strength of the self–interaction.
Therefore, if scalar fields would exist in nature, such compact objects could even question the observational AGN black hole paradigm in astrophysics.
## III Excited boson stars
### A Gravitational atoms as boson stars
Ruffini and Bonazzola used the formalism of second quantization for the complex Klein–Gordon field and noticed an important feature: If all scalar particles are within the same ground state $`|\mathrm{\Phi }=(N,n,l,a)=(N,0,0,0)`$, which is possible because of Bose–Einstein statistics, then the semi–classical Klein–Gordon equation of Kaup can be recovered in the Hartree–Fock approximation for the second quantized two–body problem. In contrast to the Newtonian approximation, the full relativistic treatment avoids an unlimited increase of the particle number $`N`$ and negative energies, but induces critical masses and particle numbers with a global maximum.
Due to this Hartree–Fock approximation and while also neglecting effects of the quantized gravitational field, the same coupled Einstein–Klein–Gordon equations (2,3) apply. Therefore, a boson star is also called a gravitational atom . Since a free Klein–Gordon equation for a complex scalar field is a relativistic generalization of the Schrödinger equation, we consider for the ground state a generalization of the wave function
$`|N,n,l,a:\mathrm{\Phi }`$ $`=`$ $`R_a^n(r)Y_l^a(\theta ,\phi )e^{i(E_n/\mathrm{})t}`$ (16)
$`=`$ $`{\displaystyle \frac{1}{\sqrt{4\pi }}}R_a^n(r)P_l^a(\mathrm{cos}\theta )e^{ia\phi }e^{i(E_n/\mathrm{})t}`$ (17)
of the hydrogen atom. Here $`R_a^n(r)`$ is the radial distribution, $`Y_l^a(\theta ,\phi )`$ the spherical harmonics, $`P_l^a(\mathrm{cos}\theta )`$ are the normalized Legendre polynomials, and $`|a|l`$ are the quantum numbers of azimuthal and angular momentum.
Due to their inherent ‘gravitational confinement’ gravitational atoms represent coherent quantum states, which nevertheless can have macroscopic size and large masses. The gravitational field is self-generated via the energy–momentum tensor, but remains completely classical, whereas the complex scalar fields are treated to some extent as Schrödinger wave functions, which in quantum field theory are referred to as semi-classical.
Moreover, Feinblum and McKinley found eigensolutions with nodes corresponding to the principal quantum number $`n`$ of the H–atom. Motivated by Heisenberg’s non-linear spinor equation additional self–interacting terms describing the interaction between the bosonic particles in a “geon” type configuration were first considered by Mielke and Scherzer , where also solutions with nodes, i.e. “principal quantum number” $`n>1`$ and non-vanishing angular momentum $`l0`$ for a t’Hooft type monopole ansatz $`\mathrm{\Phi }^{I=a}R(r)P_l^{|a|}(\mathrm{cos}\theta )`$ were found. These highly interesting instances of a possible fine structure in the energy levels of gravitational atoms poses the question if quantum geons are capable of internal excitations? Recently, without reference to these earlier works, such “exited boson stars” were recovered and their stability properties corroborated in some numerical details. Moreover, Rosen reviewed his old idea of an elementary particle built out of scalar fields within the framework of the Klein–Gordon geon (or the mini–boson star, as they are christened today).
Several surveys summarize the present status of the non-rotating case.
### B Rotating boson stars
In the framework of Newtonian theory, boson stars with axisymmetry have been constructed by several groups. Static axisymmetric boson stars, in the Newtonian limit and in GR , show that one can distinguish two classes of boson stars by their parity transformation at the equator. In both approaches only the negative parity solutions reveal axisymmetry, while those with positive parity merely converged to solutions with spherical symmetry. The metric potentials and the components of the energy-momentum tensor are equatorially symmetric despite of the antisymmetry of the scalar field. In the Newtonian description, Silveira and de Sousa followed the approach of Ref. and constructed solutions which have no equatorial symmetry at all. Hence, in GR, we have to separate solutions with and without equatorial symmetry.
Kobayashi et al. tried to find slowly rotating states (near the spherically symmetric ones) of general relativistic boson stars, but they failed. The reason for that is a quantization of the total angular momentum
$$J=T_3{}_{}{}^{0}\sqrt{|g|}d^3x=aNa=0,\pm 1,\pm 2,\mathrm{}$$
(18)
of boson stars which is proportional to the particle number $`N`$ and vanishes if $`a=0`$. This relation between angular momentum and particle number was first derived by Mielke and Schunck .
In recent papers , we proved numerically that rapidly rotating boson stars with $`a0`$ exist in general relativity. Because of the finite velocity of light and the infinite range of the scalar matter within the boson star, our localized configuration can rotate only differentially, but not uniformly. This new axisymmetric solution of the coupled Einstein–Klein–Gordon equations represent the field-theoretical pendant of rotating neutron stars which have been studied numerically for various equations of state and different approximation schemes as well as for differentially rotating superfluids as a model for (millisecond) pulsars.
On the basis of Ref. , it has erroneously been claimed that “rapidly rotating boson stars cannot exist”. However, more recently the same Japanese group (as well as ) followed exactly our Ansatz and could verify all our earlier results , albeit of some extension to stronger gravitational fields, due to better computational facilities. Due to the anisotropy of the stress–energy tensor, our configuration is differentially rotating, see for more details.
Moreover, the energy density of our rotating boson star is concentrated in an effective mass torus . Thus this first nonsingular model of a rotating body in GR realizes to some extent the suggestion of Newman et al. to fill in the Kerr metric, in view of its ring singularity, with a toroidal rather than a spherical source. Toroidal structure occurs also in relativistic star systems with an accretion disk .
Since rotating BS have a toroidal structure, there seems to exist the more speculative possibility of knotted vortex like excitations, cf. Ref. . For an $`O(3)`$ Skyrme model, their existence has recently been demonstrated numerically .
### C Formation of (primordial) boson stars
The possible abundance of solitonic stars with astrophysical mass but microscopic size could have interesting implications for galaxy formation, the microwave background, and formation of protostars. The formation of non-gravitating non-topological solutions was already studied by Frieman et al. .) In comparison with primordial black holes, it is therefore an important question if boson stars can actually form from a primordial bosonic “cloud” .
Collisionless star systems are known to settle to a centrally denser system by sending some of their members to larger radius. Likewise, a bosonic cloud will settle to a unique boson star by ejecting part of the scalar matter. Since there is no viscous term in the KG equation (3), the ‘radiation of the scalar field is the only dissipationless relaxation process called gravitational cooling. Seidel and Suen demonstrated this numerically by starting with a spherically symmetric configuration with $`M_{\mathrm{initial}}M_{\mathrm{Kaup}}`$, i.e. which is more massive then the Kaup limit. Actually such oscillating and pulsating branches have been predicted earlier in the stability analysis of Kusmartsev, Mielke, and Schunck by using catastrophe theory. Oscillating soliton stars were constructed by using real scalar fields which are periodic in time . Without spherical symmetry, i.e. for $`\mathrm{\Phi }R_a(r)Y_l{}_{}{}^{a}(\theta ,\phi )`$, the emission of gravitational waves would also be necessary.
For the formation of primordial BSs, an important issue is the breaking of unified gauge (super–)symmetry at high temperature in order to yield a scalar-antiscalar asymmetry $`ϵ_S=N_\mathrm{\Phi }/N_\gamma `$, as in the case of baryon-antibaryon asymmetry, where $`ϵ_B=N_B/N_\gamma 10^{10}`$. Here, we recall the situation of collapsing homogeneous mini-BS clouds in the early Universe, cf. . Because the Jeans scale at decoupling time is greater than the horizon scale, a bosonic mass of $`M_{\mathrm{Pl}}^3/m^2`$ immediately collapses and since this is a factor $`M_{\mathrm{Pl}}/m`$ higher than the maximal mass allowed within the mini-BS model, only black holes can form. For an asymmetry factor of order $`ϵ_Sm/M_{\mathrm{Pl}}`$, however, the total mass remaining within the horizon is $`M_{\mathrm{Pl}}^2/m`$, hence, BSs could form, avoiding the final state of a black hole.
For a real (pseudo-)scalar field, like the axion $`a`$ of the broken Peccei–Quinn symmetry in QCD, the outcome is quite different. The axion has the tendency to form compact objects (oscillatons) in a short time scale. Due to its intrinsic oscillations it would be unstable, contrary to a BS. Since the field disperses to infinity, finite non-singular self–gravitating solitonic objects cannot be formed with a massless Klein–Gordon field . In Ref. a different mechanism for forming axion miniclusters and starlike configurations was proposed. The self–coupling relaxion time is compatible or larger as the age of the Universe. For fermionic soliton stars, there is a temperature dependence in the forming of cold configurations.
### D Gravitational waves
In the last stages of boson star formation, one expects that first a highly excited configuration forms in which the quantum numbers $`n`$, $`l`$ and $`a`$ of the gravitational atom, i.e. the number $`n1`$ of nodes, the angular momentum and the azimuthal angular dependence $`e^{ia\phi }`$ are non-zero.
In a simplified picture of BS formation, all initially high modes have eventually to decay into the ground state $`n=l=a=0`$ by a combined emission of scalar radiation and gravitational radiation.
In a Newtonian approximation of Ferrell and Gleiser , the energy released by scalar radiation from states with zero quadrupole moment can be estimated by
$$E_{\mathrm{rad}}(n1)M_{\mathrm{Pl}}^2/m.$$
(19)
This is accompanied by a loss of boson particles with the rate $`\mathrm{\Delta }N(n1)(M_{\mathrm{Pl}}/m)^2`$. For investigating the gravitational radiation of macroscopic boson stars with large self-interaction, a reduction of the differential equations can be taken into account .
The lowest BS mode which has quadrupole moment and therefore can radiate gravitational waves is the $`3d`$ state with $`n=3`$ and $`l=2`$. For $`\mathrm{\Delta }J=2`$ transitions, it will decay into the $`1s`$ ground state with $`n=1`$ and $`l=0`$ while preserving the particle number $`N`$. The radiated energy is quite large, i.e., $`E_{\mathrm{rad}}=2.9\times 10^{22}`$ (GeV/mc<sup>2</sup>) Ws. Thus the final phase of the BS formation would terminate in an outburst of gravitational radiation despite the smallness of the object.
### E Gravitational evolution
There would occur an evolution of boson stars if the external gravitational constant $`\kappa `$ changes its value with time . This can be outlined within the theory of Jordan–Brans–Dicke or a more general scalar tensor theory. The results show that the mass of the boson star decreases due to a space-depending gravitational constant, given through the Brans–Dicke scalar. The mass of a boson star with constant central density is influenced by a changing gravitational constant. Moreover, the possibility of a gravitational memory of boson stars or a formation effect upon their surrounding has been analyzed as well .
## IV Are MACHOs axion stars?
Direct observation of boson stars seems to be impossible also in the far future. We propose here several effects which could possibly give indirect evidence . In the asymptotic region, the rotation velocity of baryonic objects surrounding the boson star can reveal the star’s mass. Assuming that the scalar matter of the BS interacts mainly gravitationally, we would have a ‘transparent’ BS detecting a gravitational redshift up to values of $`z=0.68`$ observable by radiating matter moving in the strong gravitational potential. For further investigations of rotation curves, cf. Ref. , and of boson stars as gravitational lenses, cf. .
Solutions with an infinite range can be found where the mass increases linearly . In the context of the dark matter hypothesis, it may be speculated if such boson halos as well as excited BS states can be used to fit the observed rotation curves for dwarf and spiral galaxies . Boson halos have a finite radius if a positive cosmological constant exists as most recent results from supernovae reveal .
Moreover, BSs could be the solution for the MACHO problem, as we are going to analyze in more detail.
In effective string theories, the dilaton $`\phi `$, another moduli field $`\beta `$, and the ‘universal’ invisible axion $`\sigma `$ are predicted . This can be read off from the effective string Lagrangian
$$_{\mathrm{eff}}=\sqrt{g}e^\phi \left[R+g^{\mu \nu }\left(_\mu \phi _\nu \phi 6_\mu \beta _\nu \beta \frac{1}{2}e^{2\phi }_\mu \sigma _\nu \sigma \right)2\mathrm{\Lambda }\right].$$
(20)
This corresponds to Eq. (11) with $`\eta =2`$ of Ref. and allows to combine the axion and the dilaton into a single complex scalar field $`\mathrm{\Phi }:=\sigma +ie^\phi `$, the axidilaton. In the conformally related Einstein frame $`g_{\mu \nu }\stackrel{~}{g}_{\mu \nu }:=e^\phi g_{\mu \nu }`$ and for constant modulus $`\beta `$, our results on BSs can easily by transferred to this axidilaton content of strings.
### A Mass range of axion stars
As macroscopic quantum states, BSs are quite generally prevented from complete gravitational collapse below a critical total mass $`M_{\mathrm{crit}}`$ which, typically, depends inversely on the particle mass, see Eq. (11).
The numerical results are shown in Fig.1. The left figure exhibits the dependence of the mass $`M`$ and the particle number $`N`$ (rest mass) on the central density $`\rho _0`$. Stable axionic BSs exist at central densities lower than the maximum mass. The critical values are: $`M_{\mathrm{crit}}=0.846`$ $`M_{}`$, $`mN_{\mathrm{crit}}=0.873`$ $`M_{}`$ and $`\rho _\mathrm{c}=9.1\times \rho _{\mathrm{nucl}}`$, where $`\rho _{\mathrm{nucl}}=2.8\times 10^{17}`$ kg/m<sup>3</sup> is the average density of nuclei. Since non-interacting bosons are very “soft”, BSs are extremely dense objects with a critical density higher than for neutron or strange stars . The figure on the right hand side gives the mass depending on the radius (measured in km). For the mass–radius diagram, we have chosen as radius 99.9% of the total mass. This ensures that the exponentially decreasing ‘atmosphere’ of the BS has almost no influence on the asymptotic Schwarzschild spacetime.
The stable BSs or axion stars (ASs) have radii larger than the minimum at 20.5 km and a mass of 0.846 $`M_{}`$. In order to derive these values, we have assumed that the mass of the scalar field is $`10^{10}`$ eV close to the lower bound of axions, leading to an asymmetry factor of $`ϵ_S10^{38}`$. and that no self-interaction exists. We stress that the total mass of these relativistic ASs is just in the observed range of 0.3 to 0.8 $`M_{}`$ for MACHOs. One could also turn this argument around: By identfying the MACHOs with known gravitational mass of about 0.5 $`M_{}`$ with ASs, we are essentially ‘weighing”, via $`M_{\mathrm{Kaup}}/N_{\mathrm{crit}}m`$, the axion mass to $`m_\sigma 10^{10}`$ eV/c<sup>2</sup>. It is gratifying to note that such a low value is perfectly compatible with the constraints on the mass range of the Kalb–Ramond axion seeding the large-scale CMB anisotropy, cf. the recent results of Gasperini and Veneziano within low-energy string cosmology.
For the other option of dilaton $`\phi `$ being stabilized through the axion, a much smaller dilaton mass of $`m_\phi 10^6m_\mathrm{a}`$ could be generated non-perturbatively, such that the dilaton behaves very similar to misalignment produced Peccei–Quinn axion $`a`$. Our conclusion also with respect to the mass range of an axidilaton star will not changed much, if we use the full Brans-Dicke type interaction for the combined axidilatons.
Thus, for cosmologically relevant (invisible) axions as cosmological dark matter also an AS with a rather large mass of would be possible and stable.
Therefore, if such–string inspired scalar fields would exist in Nature, axions could not only solve the non-baryonic dark matter problem , but their gravitationally confined mini–clusters, the axion stars, would also represent the observed MACHOs in our Galaxy.
## V Outlook: Gravitationally confined Hawking radiation?
Commonly for the Bekenstein–Hawking radiation the spacetime geometry is treated as a given fixed background, e.g. the Schwarzschild solution. However, due to the universality of gravitational interaction, the evaporating quantum field, say a scalar field $`\widehat{\mathrm{\Phi }}`$, may have a “back-reaction” upon the spacetime geometry via the semi–classical Einstein equation
$$G_{\mu \nu }=\kappa 0|T_{\mu \nu }(\widehat{\mathrm{\Phi }})|0.$$
(21)
For instance, a ‘bouncing shell’ model with retarded time $`u`$ leads to $`0|T_{uu}|0\kappa ^2/48\pi `$, the standard Hawking result. The situation becomes, however, much more complicated by the fact that the vacuum expectation value $`0|0`$ of the energy–momentum tensor $`T_{\mu \nu }`$, for instance defined by the point-splitting prescription, is not unique. One ambiguity in $`0|T_{\mu \nu }(\widehat{\mathrm{\Phi }})|0`$ is of the type $`m^2G_{\mu \nu }`$, i.e. linear in the curvature, and can be readily absorbed in a redefinition of the ‘bare’ gravitational constant $`\kappa `$. However, the next order corrections are quadratic in the curvature and therefore of the same one–loop order arising from the notorious nonrenormalizability of perturbative quantum gravity, cf. Ref., p. 90. To some extent, the finite part of such higher order curvature counterterms in the Lagrangian can be simulated by a self-interaction potential $`U(\widehat{\mathrm{\Phi }})`$, cf. .
Already on the semiclassical level one could ask the question what happens to the (massive) particles associated with the second quantized field $`\widehat{\mathrm{\Phi }}`$ in a patch of some strong gravitational background field? Could the particles created by the Unruh effect instead of evaporating to infinity rather form a bound state within their self-consistently generated gravitational field? Moreover, could it be possible that the full back-reaction on the geometry is strong enough lead to a curved spacetime without horizon and singularities, similarly as in some exact solvable (2+1)–dimensional models? Actually some aspects of this issue were already answered by Ruffini and Bonazzola for a spherically symmetric self-gravitating configuration of N particles in a Hartree–Fock approximation. Thus the back–reaction (21) may lead us back exactly to some stable branch of boson stars where the particles are treated on the first quantization level. These type of stars have an exponentially decreasing energy density of the scalar field, an ‘exosphere’ of particles in the stable state of equilibrium of particle creation and annihilation. Moreover, for these type of compact objects with an effective radius close to the last stable Kepler orbit an event horizon is suppressed due to the back-reaction (21).
Below the Kaup limit, we have seen that such macroscopic quantum states are absolutely stable, at higher central densities the configuration becomes more and more unstable, and undergoes complete gravitational collapse.
So could it be that the picture of an evaporating black hole is just a first order semi-classical approximation; rather, below some mass limit, we may end up in a self-consistent state of a boson or fermion star with a gravitationally confined Hawking radiation, a quantum geon?
## Acknowledgments
We would like to thank John Barrow, Andrew Liddle, Alfredo Macías, and Diego Torres for useful discussions, literature hints, and support. Moreover, we are grateful to the Referee for pointing out Ref. and the compatibilty of our axion mass with the independent estimates of Ref. . This work was partially supported by CONACyT, grant No. 28339E, and the joint German–Mexican project DLR–Conacyt E130–2924 and MXI 009/98 INF. One of us (E.W.M.) thanks Noelia Méndez Córdova for encouragement. F.E.S. was supported by an European Union Marie Curie TMR fellowship. |
warning/0001/math0001152.html | ar5iv | text | # A Generalized Torelli Theorem
## . §1 Introduction
All curves in subsequent sections will be assumed to be smooth projective curves over $``$. The genus of $`C`$ will always be denoted by $`g`$. If $`C`$ is such a curve (with $`g>0`$) we will let $`J(C)`$ denote its Jacobian and
$$u:CJ(C)$$
will be the Abel-Jacobi map. We will let $`C^{(d)}`$ denote the $`d`$th symmetric power of $`C`$ and for $`1dg1`$, $`W^d`$ will be the image of $`C^{(d)}`$ inside the Jacobian under the Abel-Jacobi map. Since by a theorem of Riemann, the theta divisor is a translate of $`W^{g1}`$, Torelli’s theorem asserts that the pair $`(J(C),W^{g1})`$ determines the curve, meaning that if $`C^{}`$ is another curve such that there is an isomorphism $`J(C)J(C^{})`$ carrying theta divisors to theta divisors then the curves must be isomorphic. Our aim is to show that an analogous statement holds for each $`1d<g1`$. With this in mind we will assume in all following sections that $`g4`$, as smaller genera are covered by existing theorems. Our strategy is largely based on the strategy in .
As a corollary we have that two curves are isomorphic if and only if their $`d`$th symmeteric powers are isomorphic, where $`d`$ is an integer smaller than the genus of one (and hence both) of the curves.
This problem and the above mentioned Corollary was originally proposed by Prof. Donu Arapura. Thanks also to the particpants of the Working Algebraic Geometry Seminar at Purdue, in particular to Prof. Kenji Matsuki who pointed out a mistake in an earlier version.
## . §2 Preliminaries
The Jacobian of a curve $`C`$ is defined to be
$$J(C)=\mathrm{H}^0(C,\mathrm{\Omega }_C^1)^{}/\mathrm{H}_1(C,).$$
The Abel-Jacobi map is defined by
$$\begin{array}{cc}u:& CJ(C)\\ & p_{p_0}^p\end{array}$$
where $`p_0`$ is a fixed basepoint. Let $`C^{(d)}=C^d/S^d`$ be the $`d`$th symmetric power of $`C`$. We identify the points of $`C^{(d)}`$ with effective divisors of degree $`d`$ on $`C`$. The Abel-Jacobi map can be extended to a morphism
$$u:C^{(d)}J(C).$$
We have
###### (2.1) Theorem (Abel’s).
Let $`D,D^{}C^{(d)}`$. Then
$$DD^{}ifandonlyifu(D)=u(D^{})$$
where the relation $``$ is linear equivalence.
###### Proof.
See . ∎
We let $`W^d=u(C^{(d)})`$. By Abel’s Theorem $`W^d`$ parameterises complete linear systems of degree $`d`$ on $`C`$. Our aim is to reconstruct $`C`$ from the pair $`(J(C),W^d)`$ where $`0<dg1`$. The main tool in doing this will be the Gauss map, defined as follows. Take $`pW_{\mathrm{smooth}}^d`$ and let $`\mathrm{T}_p(W^d)`$ be its holomorphic tangent space. There is an automorphism, translation by $`p`$,
$$\begin{array}{cc}\tau _p:& J(C)J(C)\\ & xxp\end{array}$$
This allows us to canonically identify $`\mathrm{T}_p(W^d)`$ with a $`d`$-dimensional subspace of $`\mathrm{T}_0(J(C))\mathrm{H}^0(C,\mathrm{\Omega }_C^1)^{}`$. This defines the Gauss map
$$𝒢:W_{\mathrm{smooth}}^d𝔾(d1,g1),$$
where $`𝔾(d1,g1)`$ is the Grassmanian parameterizing $`d1`$ dimensional linear subvarieties of $`^{g1}`$ (or equivalently $`d`$-dimensional subspaces of $`^g`$. The result we need is:
###### (2.2) Theorem.
Let $`\varphi _K:C(^{g1})^{}`$ be the canonical morphism and let $`DC^{(d)}`$. Then $`u(D)W_{\mathrm{smooth}}^d`$ if and only if $`\mathrm{dim}|D|=0`$. If we denote by $`\overline{\varphi _K(D)}`$ the linear span of $`D`$ on the canonical curve then
$$𝒢(u(D))=\overline{\varphi _K(D)}.$$
###### Proof.
This result can be found in §2.7 of . ∎
Note that the linear span of a multiple of a point is the appropriate osculating plane to $`C`$ inside $`^{g1}`$. The condition that $`\mathrm{dim}|D|=0`$ forces $`\overline{\varphi _K(D)}`$ to be a $`d1`$ dimensional linear subvariety of $`^{g1}`$. This is by
###### (2.3) Theorem (Geometric Riemann-Roch).
For $`D`$ as in the above discussion we have $`\mathrm{dim}|D|=d1\mathrm{dim}\overline{\varphi _K(D)}.`$
###### Proof.
Again this can be found in . ∎
## . §3 Our Strategy
We first describe the idea behind the proof of the Torelli theorem for curves, due to A. Andreotti, see . The Gauss map
$$𝒢:W_{\mathrm{smooth}}^{g1}(^{g1})^{}$$
is a quasi-finite morphism of degree
$$\left(\genfrac{}{}{0pt}{}{2g2}{g1}\right).$$
To see this, a hyperplane $`H`$ intersects the image of a curve $`C`$ under its canonical morphism in $`2g2`$ points $`p_1,p_2,\mathrm{},p_{2g2}`$, which are in general posiion for a generic $`H`$. By Theorem (2.2) the fibre over $`H`$ consists of all images of divisors of the form $`u(p_{i_1}+p_{i_2}+\mathrm{}+p_{i_{g1}})`$ where $`i_j`$ range over $`\{1,2,\mathrm{},2g2\}`$. If $`C`$ is non-hyperelliptic then let $`C^{}`$ be the dual variety to $`C`$, that is the locus of all tangent hyperplanes to $`\varphi _K(C)`$ inside $`(^{g1})^{}`$. Now one would expect that the (closure of the) branch locus of $`𝒢`$ to be $`C^{}`$ since the fibre over a tangent hyperplane $`H`$ should have cardinality smaller than
$$\left(\genfrac{}{}{0pt}{}{2g2}{g1}\right).$$
(Since $`H.C=2p_1+\mathrm{}p_{2g3}`$, the first point is repeated and there are fewer choices for points in the fibre.) It is known how to recover $`C`$ from $`C^{}`$, for example see . In the case that $`C`$ is hyperelliptic the canonical morphism $`\varphi _K:C^{g1}`$ is branched at $`2g+2`$ points labelled $`b_1,\mathrm{},b_{2g+2}`$. We denote by $`C^{}`$ the dual variety to the rational normal curve $`\varphi _K(C)`$ and $`b_i^{}`$ denotes the locus of all hyperplanes passing through $`b_i`$. In the hyperelliptic case, by the same reasoning as in the non-hyperelliptic case, one would expect that the branch locus of $`𝒢`$ to be $`C^{}b_1^{}\mathrm{}b_{2g+2}^{}`$. It is known how to recover $`C`$ from this information.
We would like to try to apply this technique to our situation. Firstly, we may reduce to the case where $`(g1)/2<d<g1`$. To do this choose an integer $`n`$ so that $`(g1)/2<ndg1`$. Then
$$W^{nd}=\underset{n\mathrm{times}}{\underset{}{W^d+W^d+\mathrm{}+W^d}}.$$
The above addition is addition inside the Jacobian.
Fix $`^{g1}=(\mathrm{H}^0(C,\mathrm{\Omega }_C^1)^{})`$. Now consider the locus
$$\begin{array}{c}\mathrm{F}(d,g)=\{(V,W)𝔾(d1,^{g1})\times 𝔾(d1,^{g1})\\ \overline{V+W}^{g1}\}.\end{array}$$
The notation $`\overline{V+W}`$ means linear span of $`V`$ and $`W`$. So $`\mathrm{F}(d,g)`$ is the locus of all pairs of $`(d1)`$-dimensional linear subvarieties that are contained inside some hyperplane. There is a rational morphism
defined by $`(V,W)\overline{V+W}`$. We take $`\mathrm{E}(d,g)`$ to be the pullback of $`\mathrm{F}(d,g)`$ under
$$𝒢\times 𝒢:W_{\mathrm{smooth}}^d\times W_{\mathrm{smooth}}^d𝔾(d1,g1)\times 𝔾(d1,g1).$$
Now let $`\beta `$ be the composed rational morphism
Arguing as in the case $`d=g1`$ we see that the branch locus of $`\beta `$ contains enough information to recover $`C`$. Note that the hypothesis $`(g1)/2<d<g1`$ is required to insure that $`\mathrm{E}(d,g)`$ is not empty.
## . §4 Generic Determinental Varieties
Two identities that will be useful later are presented in this section.
In this section $`d`$ and $`g`$ will be non-negative integers with $`(g1)/2<d<g1`$. We will need the case $`g4`$ later. Let $`M`$ be the generic $`g\times 2d`$ matrix,
$$M=\left(\begin{array}{cccc}x_{11}& x_{12}& \mathrm{}& x_{1,2d}\\ x_{21}& x_{22}& \mathrm{}& x_{1,2d}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ x_{g1}& x_{g2}& \mathrm{}& x_{g,2d}\end{array}\right)$$
over the polynomial ring $`[x_{ij}]`$. We will let $`M_{(i_1,i_2,\mathrm{},i_g)},`$ where $`i_1<i_2<\mathrm{}<i_g,`$ be the following submatrix of $`M`$.
$$M_{(i_1,i_2,\mathrm{},i_g)}=\left(\begin{array}{cccc}x_{1,i_1}& x_{1,i_2}& \mathrm{}& x_{1,i_g}\\ x_{2,i_1}& x_{2,i_2}& \mathrm{}& x_{1,i_g}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ x_{g,i_1}& x_{g,i_2}& \mathrm{}& x_{g,i_g}\end{array}\right).$$
Also let
$$N=\left(\begin{array}{cccc}x_{11}& x_{12}& \mathrm{}& x_{1,g1}\\ x_{21}& x_{22}& \mathrm{}& x_{1,g1}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ x_{g1}& x_{g2}& \mathrm{}& x_{g,g1}\end{array}\right).$$
Let $`f`$ be the product of the $`(g1)\times (g1)`$ minors of $`N`$. Let $`R=[x_{ij}]_f`$. Let $`I`$ be the ideal generated by the $`g\times g`$ minors of $`M`$ in $`[x_{ij}]`$. Finally let $`J`$ be the ideal of $`[x_{ij}]`$ generated by the minors of the form $`\mathrm{det}(M_{(1,2,\mathrm{},g1,i)})`$, as $`i`$ ranges over, $`gi2d`$. We wish to prove
###### (4.1) Proposition.
Consider the ideals $`I_f,J_f`$ obtained by extending $`I`$ and $`J`$ to the ring $`R`$. We have $`I_f=J_f`$.
###### Proof.
It is clear that $`J_fI_f`$. We proceed by showing that $`\sqrt{J_f}=\sqrt{I_f}`$ and then showing that $`J_f`$ is equal to its radical.
We begin by showing $`\sqrt{Jf}=\sqrt{If}`$. Here $`f`$ is the ideal generated by $`f`$. To show the above it suffices to show that the two ideals have the same zero locus inside $`𝔸^{g\times 2d}`$. It is clear that
$$Z(Jf)=Z(J)Z(f)Z(I)Z(f)=Z(If).$$
Now take $`p=(p_{ij})`$ in the zero locus of $`Jf`$. We may assume $`p`$ is not in the zero locus of $`f`$, for otherwise we are done. Consider the matrix
$$M_p=\left(\begin{array}{cccc}p_{11}& p_{12}& \mathrm{}& p_{1,2d}\\ p_{21}& p_{22}& \mathrm{}& p_{2,2d}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ p_{g1}& p_{g2}& \mathrm{}& p_{g,2d}\end{array}\right).$$
Showing that $`pZ(I)`$ is equivalent to showing that $`\mathrm{rank}(M_p)g1`$. Since $`(p_{ij})Z(f)`$ the first $`g1`$ columns of $`M_p`$ are linearly independent. As $`(p_{ij})Z(J)`$,
$$\mathrm{det}((M_p))_{(1,2,\mathrm{},g1,i)}=0,$$
for $`gi2d`$. So the $`i`$th column is in the linear span of the first $`g1`$ columns and we are done. We have shown $`\sqrt{Jf}=\sqrt{If}`$. An elementary argument now shows that $`\sqrt{I_f}=\sqrt{J_f}`$.
Finally we need to show that $`J_f`$ is radical. Notice that $`J_f`$ is generated by polynomials of the form
$$\begin{array}{c}\mathrm{det}(M_{(1,2,\mathrm{},g1,i)})=\mathrm{det}(N_1)x_{1i}\mathrm{det}(N_2)x_{2i}+\mathrm{}\\ (1)^g\mathrm{det}(N_{g1})x_{g1,i}.\end{array}$$
Here $`N_j`$ is the submatrix of $`N`$ obtained by deleting the $`j`$th row. Each of the $`\mathrm{det}(N_j)`$ are units in our ring $`R`$. The result follows from the following lemma. ∎
###### (4.2) Lemma.
Let $`A`$ be a reduced ring and consider the polynomial ring $`B=A[x_{ij}]`$, where $`1in`$ and $`1jm`$. Consider elements
$$f_i=u_{i1}x_{i1}+u_{i2}x_{i2}+\mathrm{}+u_{im}x_{im}.$$
Form the ideal $`I=(f_1,f_2,\mathrm{},f_n)`$. If the $`u_{ij}`$ are units in $`A`$ then $`B/I`$ is reduced.
###### Proof.
Observe that $`B/IA[x_{ij}]`$ but with new index ranges $`2in`$ and $`2jm`$. ∎
Now let $`M`$ be the matrix
$$M=\left(\begin{array}{ccc}x_{11}& \mathrm{}& x_{1,2d}\\ \mathrm{}& & \mathrm{}\\ x_{g1}& \mathrm{}& x_{g,2d}\end{array}\right)$$
over the polynomial ring $`[x_{ij}]`$. Consider the submatrices
$$A=\left(\begin{array}{ccc}x_{11}& \mathrm{}& x_{1,d}\\ \mathrm{}& & \mathrm{}\\ x_{d,1}& \mathrm{}& x_{d,d}\end{array}\right)B=\left(\begin{array}{ccc}x_{1,d+1}& \mathrm{}& x_{1,2d}\\ \mathrm{}& & \mathrm{}\\ x_{d,d+1}& \mathrm{}& x_{d,2d}\end{array}\right).$$
Set $`f=\mathrm{det}(A)`$ and $`g=\mathrm{det}(B)`$. We will be interested in the following ideals of the ring $`[x_{ij}]_{fg}`$. Let $`I`$ be ideal of the $`g\times g`$ minors of $`M`$ and let $`J`$ be the ideal of the $`g\times g`$ minors of
$$N=M\left(\begin{array}{cc}A^1& 0\\ 0& B^1\end{array}\right).$$
###### (4.3) Lemma.
The ideals $`I`$ and $`J`$ of $`[x_{ij}]_{fg}`$ are equal.
###### Proof.
The subschemes of $`\mathrm{spec}([x_{ij}]_{fg})`$ defined by $`I`$ and $`J`$ are supported on the same closed subset. So it suffices to show that both $`I`$ and $`J`$ are reduced. The fact that $`I`$ is reduced is the fundamental theorem of invariant theory, see . To show that $`J`$ is reduced consider the $``$ algebra automorphism of $`[x_{ij}]_{fg}`$ defined by
$$x_{ij}y_{ij}$$
where
$$M\left(\begin{array}{cc}A& 0\\ 0& B\end{array}\right)=\left(\begin{array}{ccc}y_{11}& \mathrm{}& y_{1,2d}\\ \mathrm{}& & \mathrm{}\\ y_{g1}& \mathrm{}& y_{g,2d}\end{array}\right).$$
This automorphism carries $`J`$ to $`I`$ so we are done. ∎
## . §5 A Subvariety of $`𝔾(d1,g1)\times 𝔾(d1,g1)`$
We let $`𝔾(d1,g1)`$ denote the Grassmanian paramaterizing $`(d1)`$ dimensional linear subspaces of $`^{g1}`$. Let
$$\begin{array}{c}\mathrm{F}(d,g)=\{(V,W)𝔾(d1,g1)\times 𝔾(d1,g1)\\ VH,WH\mathrm{for}\mathrm{some}\mathrm{hyperplane}H^{g1}\}.\end{array}$$
In the above $`V`$ and $`W`$ are closed points of the Grassmanian. We wish to describe the reduced scheme structure on $`\mathrm{F}(d,g)`$. First we recall how to cover Grassmanian with open affines isomorphic to $`^{d(gd)}`$.
Let $`V𝔾(d1,g1)`$ be a closed point. So $`V`$ can be thought of as the column space of a $`g\times d`$ matrix $`A`$. Write
$$A=\left(\begin{array}{cccc}a_{11}& a_{12}& \mathrm{}& a_{1d}\\ a_{21}& a_{22}& \mathrm{}& a_{2d}\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ a_{g1}& a_{g2}& \mathrm{}& a_{gd}\end{array}\right).$$
This representation is unique upto the action of $`\mathrm{GL}(d,)`$.
Let $`I=(i_1,i_2,\mathrm{}i_d)`$, where $`i_j\{1,2,\mathrm{},g\}`$ and $`i_1<i_2<\mathrm{}<i_d`$. We will denote by $`A^I`$ the following $`d\times d`$ submatrix of $`A`$:<sup>1</sup><sup>1</sup>1In the preceeding section we defined $`A_I`$. In that section the submatrix $`A_I`$ of $`A`$ was obtained by choosing columns of $`A`$, while here we are choosing rows.
$$A^I=\left(\begin{array}{cccc}a_{i_11}& a_{i_12}& \mathrm{}& a_{i_1d}\\ a_{i_21}& a_{i_22}& \mathrm{}& a_{i_2d}\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ a_{i_g1}& a_{i_g2}& \mathrm{}& a_{i_gd}\end{array}\right).$$
Now since the rank of $`A`$ is $`d`$, the matrix $`A`$ has a non vanishing $`d\times d`$ minor. Let this minor be $`det(A^I)`$. The matrix $`A^{}=A(A^I)^1`$ also has column space equal to $`V`$, furthermore it is the unique representative with $`(A^{})^I=\mathrm{Id}_d`$. For each $`I=(i_1,i_2,\mathrm{}i_d)`$ as above, set
$$\begin{array}{c}U_I=\{V\mathrm{G}(d,g)\mathrm{the}I\mathrm{minor}\mathrm{of}\mathrm{a}\mathrm{matrix}\\ \mathrm{representative}\mathrm{of}V\mathrm{is}\mathrm{invertible}\}.\end{array}$$
There is a bijection $`U_I^{d.(gd)}`$, which is in fact an isomorphism. For further details see or .
It follows from the above that $`𝔾(d1,g1)\times 𝔾(d1,g1)`$ has an open affine cover constisting of opens of the form $`U_I\times U_J^{2d(gd)}`$. Now take $`(V,W)U_I\times U_J`$, with $`V`$ the column space of a matrix $`A`$ and $`W`$ the column space of a matrix $`B`$. The locus we are trying to describe, $`\mathrm{F}(d,g)`$, consists of those pairs $`(V,W)`$ such that $`\mathrm{rank}(A|B)<g`$. Here $`(AB)`$ is the matrix obtained by augmenting the matrix $`A`$ with the matrix $`B`$. Now the rank of $`(A|B)<g`$ if and only if the $`g\times g`$ minors of $`(AB)`$ vanish. The latter condition holds if and only if the $`g\times g`$ minors of the matrix $`(AB)C`$ vanish where
$$C=\left(\begin{array}{cc}A_I^1& 0\\ 0& B_J^1\end{array}\right).$$
The entries of the matrix $`(AB)C`$ determine the image of $`(V,W)`$ under the isomorphism $`U_I\times U_J^{d(gd)+d(gd)}.`$ So the ideal generated by the $`g\times g`$ minors of $`(AB)C`$ determines a scheme structure on $`\mathrm{F}(d,g)U_I\times U_J`$.It follows from pg. 71 that this scheme structure is reduced, being a specialization of the ideal $`I_k`$ defined there. Hence these ideal sheaves on $`U_I\times U_J`$ glue together to give an ideal sheaf for the reduced structure on $`\mathrm{F}(d,g)`$.
We let
$$U_\mathrm{F}=\{(V,W)\mathrm{F}(d,g)\mathrm{rank}(AB)=g1\}.$$
There is a morphism
It takes a closed point $`(V,W)`$ to the linear span of $`V`$ and $`W`$. We will denote $`\overline{U}_F`$ by $`\mathrm{F}(d,g)_{\mathrm{main}}`$.
## . §6 The construction of $`\mathrm{E}(d,g)`$
In this section let $`C`$ be a curve of genus $`g4`$. Let $`(g1)/2<d<g1`$. We have a morphism
$$𝒢\times 𝒢:W_{\mathrm{smooth}}^d\times W_{\mathrm{smooth}}^d𝔾(d1,g1)\times 𝔾(d1,g1).$$
Define $`\mathrm{E}(d,g)W_{\mathrm{smooth}}^d\times W_{\mathrm{smooth}}^d`$ to be the fibre over $`\mathrm{F}(d,g)`$. We take $`U_\mathrm{E}`$ to be the preimage of $`U_\mathrm{F}`$ and $`\mathrm{E}(d,g)_{\mathrm{main}}`$ to be the closure of $`U_\mathrm{E}`$.There is morphism
$$\beta :U_\mathrm{E}(^{g1})^{}.$$
We have, by theorem (2.2),
$$\beta (u(D),u(D^{}))=\overline{\varphi _K(D)\varphi _K(D^{})},$$
where $`(D,D^{})C^{(d)}\times C^{(d)}`$ are divisors whose image under the Abel-Jacobi map is in $`W_{\mathrm{smooth}}^d`$. Recall that $`\overline{A}`$ means linear span of some subset $`A`$ of $`^{g1}`$ in $`^{g1}`$. Notice that $`\overline{\varphi _K(D)\varphi _K(D^{})}`$ is a hyperplane in $`^{g1}`$, for the condition $`(u(D),u(D^{}))\mathrm{E}(d,g)`$ forces $`\overline{\varphi _K(D)\varphi _K(D^{})}`$ to be contained in a hyperplane and the condition $`(u(D),u(D^{}))U_\mathrm{E}`$ forces $`\overline{\varphi _K(D)\varphi _K(D^{})}`$ to be exactly a hyperplane.
A generic hyperplane $`H(^{g1})^{}`$ intersects $`C`$ in $`2g2`$ points that are in general position , see . So suppose that $`H.C=p_1+p_2+\mathrm{}+p_{2g2}`$. Then by (2.3), the pair
$$(u(p_1+p_2+\mathrm{}+p_d),u(p_{d+1}+p_{d+2}+\mathrm{}+p_{2d})),$$
(notice $`2d<2g2`$) is a closed point of $`W_{\mathrm{smooth}}^d\times W_{\mathrm{smooth}}^d`$. Furthermore the above pair, gives a point in $`U_E`$ mapping to $`H`$ under $`\beta `$. Hence $`\beta `$ is dominant. Since a hyperplane can only intersect $`C`$ in a finite number of points, the map $`\beta `$ is quasi-finite. It follows that $`U_E`$ has dimension $`g1`$.
We let $`C^{}`$ denote the dual variety to $`\varphi _K(C)`$.
###### (6.1) Lemma.
(a) Suppose that $`C`$ is a non-hyperelliptic curve. Let $`H(^{g1})^{}C^{}`$. If $`\beta ((u(D),u(D^{})))=H`$ then $`(u(D),u(D^{}))`$ lies on a component of $`\mathrm{E}(d,g)`$ of dimension $`g1`$ and is in the smooth locus of $`\mathrm{E}(d,g)_{\mathrm{main}}`$.
(b) Suppose that $`C`$ is hyperelliptic. Let $`H(^{g1})^{}C^{}`$ and assume also that $`H`$ does not pass through any of the branch points of the canonical map $`\varphi _K:C^{g1}`$. If $`\beta ((u(D),u(D^{})))=H`$ then $`(u(D),u(D^{}))`$ lies on a component of $`\mathrm{E}(d,g)`$ of dimension $`g1`$ and is in the smooth locus of $`\mathrm{E}(d,g)_{\mathrm{main}}`$.
###### Proof.
The following proof is for (a).
Write $`D=p_1+p_2+\mathrm{}+p_d`$ and $`D^{}=p_1^{}+p_2^{}+\mathrm{}+p_d^{}`$. We choose local coordinates $`z_i`$ and $`z_i^{}`$ on $`C`$ centred at $`p_i`$ and $`p_i^{}`$ respectively. Now as $`H`$ is not a tangent hyperplane $`C`$, we have $`p_ip_j`$ and $`p_i^{}p_j^{}`$ for $`ij`$. It follows that $`z_1,z_2\mathrm{},z_d`$ and $`z_1^{},z_2^{}\mathrm{},z_d^{}`$ descend to local co-ordinates on $`C^{(d)}\times C^{(d)}`$ centred at $`(D,D^{})`$. Furthermore, by (2.1), the Abel-Jacobi map is an isomorphism around $`(D,D^{})`$, since $`u(D),u(D^{})W_{\mathrm{smooth}}^d`$. So we have some local co-ordinates on $`W^d\times W^d`$ centred at $`(u(D),u(D^{}))`$. Let $`\omega _1,\mathrm{}\omega _g`$ be a basis for $`\mathrm{H}^0(\mathrm{\Omega }_C^1)`$. We write $`\omega _j`$ as $`\mathrm{\Omega }_{ji}(z_i)dz_i`$ in a neighbourhood of $`p_i`$ and as $`\mathrm{\Omega }_{ji}^{}(z_j^{})dz_j^{}`$ in a neighbourhood of $`p_j^{}`$. Let
$$M(z)=\left(\begin{array}{cccccc}\mathrm{\Omega }_{11}(z_1)& \mathrm{}& \mathrm{\Omega }_{1d}(z_d)& \mathrm{\Omega }_{11}^{}(z_1^{})& \mathrm{}& \mathrm{\Omega }_{1d}^{}(z_d^{})\\ \mathrm{\Omega }_{21}(z_1)& \mathrm{}& \mathrm{\Omega }_{2d}(z_d)& \mathrm{\Omega }_{21}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{2d}^{}(z_1^{})\\ \mathrm{}& & \mathrm{}& \mathrm{}& & \mathrm{}\\ \mathrm{\Omega }_{g1}(z_1)& \mathrm{}& \mathrm{\Omega }_{gd}(z_d)& \mathrm{\Omega }_{g1}^{}(z_1^{})& \mathrm{}& \mathrm{\Omega }_{gd}^{}(z_d^{})\end{array}\right).$$
In a neighbourhood of $`(u(D),u(D^{}))`$, $`\mathrm{E}(d,g)`$ is defined by the vanishing of the $`g\times g`$ minors of $`M(z)`$, by (4.3). Now by (2.3), $`\mathrm{dim}\overline{\varphi _K(D)}=d1`$, so in a neighbourhood of $`(u(D),u(D^{}))`$ the first $`d`$ columns of $`M(z)`$ are linearly independent. Since $`M(Z)`$ has rank $`g1`$ at the point $`(u(D),u(D^{}))`$ we may reindex the points of $`D^{}`$ so that the first $`g1`$ columns of $`M(z)`$ are linearly independent in a neighbourhood of $`(u(D),u(D^{}))`$. Set
$$f_i=\mathrm{det}M(z)_{(1,2,\mathrm{},g1,i)},$$
where $`g1<i2d`$. By (4.1), $`\mathrm{E}(d,g)`$ is defined by $`f_i`$ in a neighbourhood of $`(u(D),u(D^{}))`$. The assertion that $`(u(D),u(D^{}))`$ lies on a component of dimension $`g1`$ of $`\mathrm{E}(d,g)`$ follows.
By definition, $`f_j`$ is independent of the co-ordinates $`z_i^{}`$ for $`gdid`$ and $`ijd`$. So the Jacobian matrix is of the form
$$\left(\begin{array}{cccc}\frac{f_g}{z_1}& \frac{f_{g+1}}{z_1}& \mathrm{}& \frac{f_{2d}}{z_1}\\ \frac{f_g}{z_2}& \frac{f_{g+1}}{z_2}& \mathrm{}& \frac{f_{2d}}{z_2}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ \frac{f_g}{z_d}& \frac{f_{g+1}}{z_d}& \mathrm{}& \frac{f_{2d}}{z_d}\\ \frac{f_g}{z_1^{}}& \frac{f_{g+1}}{z_1^{}}& \mathrm{}& \frac{f_{2d}}{z_1^{}}\\ \frac{f_g}{z_2^{}}& \frac{f_{g+1}}{z_2^{}}& \mathrm{}& \frac{f_{2d}}{z_2^{}}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ \frac{f_g}{z_{gd1}^{}}& \frac{f_{g+1}}{z_{gd1}^{}}& \mathrm{}& \frac{f_{2d}}{z_{gd1}^{}}\\ \frac{f_g}{z_{gd}^{}}& 0& \mathrm{}& 0\\ 0& \frac{f_{g+1}}{z_{gd+1}^{}}& \mathrm{}& 0\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ 0& 0& \mathrm{}& \frac{f_{2d}}{z_d^{}}\end{array}\right)|_{(u(D),u(D^{}))}.$$
Suppose that $`(u(D),u(D^{}))`$ is a singular point of $`\mathrm{E}(d,g)`$. This is true if and only if the above matrix has rank smaller than $`2dg+1`$. It has rank smaller than $`2dg+1`$ if and only if
$$\frac{f_j}{z_j^{}}|_{(u(D),u(D^{}))}=0$$
for some $`j`$. Now
(5) $`0`$ $`=`$ $`{\displaystyle \frac{f_j}{z_j^{}}}|_{(u(D),u(D^{}))}`$
$`=`$ $`\left|\begin{array}{ccccccc}\mathrm{\Omega }_{11}(p_1)& \mathrm{}& \mathrm{\Omega }_{1d}(p_d)& \mathrm{\Omega }_{11}^{}(p_1^{})& \mathrm{}& \mathrm{\Omega }_{1,g1d}^{}(p_{g1d}^{})& \frac{\mathrm{\Omega }_{1j}^{}}{z_j^{}}|_{p_j}\\ \mathrm{\Omega }_{21}(p_1)& \mathrm{}& \mathrm{\Omega }_{2d}(p_d)& \mathrm{\Omega }_{21}^{}(p_1^{})& \mathrm{}& \mathrm{\Omega }_{2,g1d}^{}(p_{g1d}^{})& \frac{\mathrm{\Omega }_{2j}^{}}{z_j^{}}|_{p_j}\\ \mathrm{}& & \mathrm{}& \mathrm{}& & \mathrm{}& \mathrm{}\\ \mathrm{\Omega }_{g1}(p_1)& \mathrm{}& \mathrm{\Omega }_{gd}(p_d)& \mathrm{\Omega }_{g1}^{}(p_1^{})& \mathrm{}& \mathrm{\Omega }_{g,g1d}^{}(p_{g1d}^{})& \frac{\mathrm{\Omega }_{gj}^{}}{z_j^{}}|_{p_j}\end{array}\right|`$
The first $`g1`$ columns lie inside $`H`$. So it follows that the last column is contained in $`H`$. This implies the tangent line to $`p_j^{}`$ is in $`H`$, which in turn contradicts $`HC^{}`$.
A similar argument proves (b). ∎
## . §7 Generic Tangent Hyperplanes
Let $`C`$ be a curve with a fixed non-degenerate embedding $`\varphi :C^n`$, with $`n3`$. Recall that all curves are assumed to be smooth and projective. The genus of our curve will also be assume to be $`4`$. We will denote by $`C^{}`$ the dual variety to $`C`$ inside $`(^n)^{}`$. By forming the incidence correspondence
$$\mathrm{\Sigma }=\{(p,H)pC,H(^n)^{},\mathrm{T}_p(C)H\}$$
and using standard arguments we see that $`C^{}`$ is an irreducible hypersurface in $`(^n)^{}`$. We use the notation $`\mathrm{T}_p(C)`$ to denote the tangent line to $`C`$ at $`p`$ inside $`^n`$.
Let $`\varphi _2:C𝔾(2,n)`$ be the second associated curve to $`\varphi `$. So $`\varphi _2(p)`$ is the unique plane having intersection order at least $`3`$ with $`C`$ at $`p`$. (See , pg. 263). Let $`\mathrm{\Gamma }_2C\times 𝔾(2,n)`$ be the graph of $`\varphi _2`$. We form the incidence correspondence
$$\mathrm{\Sigma }^{\prime \prime }=\{(p,P,H)\mathrm{\Gamma }_2\times (^n)^{}(p,P)\mathrm{\Gamma }_2,\mathrm{and}PH\}.$$
Let $`\pi _C:`$ be the projection $`\pi _C:\mathrm{\Sigma }^{\prime \prime }C`$. The fibre over $`pC`$ is irreducible of dimension $`n3`$. It follows that $`\mathrm{\Sigma }^{\prime \prime }`$ is irreducible of dimension $`n2`$. The projection from $`\mathrm{\Sigma }^{\prime \prime }`$ to $`C^{}`$ is a finite morphism, hence the locus of hyperplanes having intersection at least $`3`$ at some point of $`C`$ is an irreducible closed subvariety of codimension $`1`$ inside $`C^{}`$.
###### (7.1) Lemma.
Let $`\varphi _K:C^{g1}`$ be the canonical morphism.
(a) Suppose that $`C`$ is a non-hyperelliptic curve so that $`\varphi _K`$ is an immersion. Then for a generic $`HC^{}(^{g1})^{}`$,
$$H.C=2p_1+p_2+p_3+\mathrm{}+p_{2g3}$$
where the $`p_i`$ are distinct.
(b) Suppose that $`C`$ is a hyperelliptic curve so that $`\varphi _K(C)`$ is a rational normal curve . Let $`C^{}`$ be the dual variety to $`\varphi _K(C)`$. Let $`b_1,\mathrm{},b_{2g+2}`$ be the branch points of $`\varphi _K`$. We denote by $`b_i^{}(^{g1})^{}`$ the dual variety to $`b_i`$, consisting of all hyperplanes through $`b_i`$. So $`b_i^{}`$ is a hyperplane in $`(^{g1})^{}`$. Then for a generic
$$HC^{}b_1^{}\mathrm{}b_{2g+2}^{}$$
we have that
$$H.C=2p_1+p_2+p_3+\mathrm{}+p_{2g3}$$
where the $`p_i`$ are distinct.
###### Proof.
(a) We have seen, in the discussion preceeding the lemma, that for a generic $`HC^{}`$, $`H.C`$ has no points of multiplicity $`3`$. So we need to show that a generic tangent hyperplane has only one point of multiplicity 2. Form the incidence correspondence
$$\begin{array}{c}\mathrm{\Sigma }^{}=\{(p,q,P,H)C\times C\times 𝔾(3,g1)\times (^{g1})^{}pq\\ \mathrm{T}_p(C),\mathrm{T}_q(C)PH\}.\end{array}$$
Note that $`\mathrm{\Sigma }^{}`$ is only locally closed in $`C\times C\times 𝔾(3,g1)\times (^{g1})^{}`$. Let
$$\mathrm{\Sigma }^{}=\mathrm{\Sigma }_1^{}\mathrm{\Sigma }_2^{}\mathrm{}\mathrm{\Sigma }_l^{}$$
be an irreducible decomposition for $`\mathrm{\Sigma }^{}`$. There is a projection $`\mathrm{\Sigma }^{}C\times C`$. From IV Theorem (3.10) there is a closed subset $`XC\times C`$ such that for each $`(p,q)D`$, the tangent lines $`\mathrm{T}_p(C)`$ and $`\mathrm{T}_q(C)`$ do not meet and $`X`$ has codimension $`1`$ in $`C\times C`$. Consider the restricted projection
$$\pi _i:\mathrm{\Sigma }_i^{}C\times C.$$
Now if there is a point $`(p,q)X`$, and in the image of $`\mathrm{\Sigma }_i^{}`$, the fibre over $`(p,q)`$ has dimension $`g5`$ as $`\mathrm{T}_p(C)`$ and $`\mathrm{T}_q(C)`$ span a 3-plane in $`^{g1}`$. Hence
$$\begin{array}{ccc}\mathrm{dim}\mathrm{\Sigma }_i^{}& & \mathrm{dim}C\times C+\mathrm{dim}(\mathrm{fibre})\\ & =& g3.\end{array}$$
(Note that if $`g=4`$, then there is no such $`(p,q)`$.) If there is no such $`(p,q)`$ then the projection can be factored as
$$\pi _i:\mathrm{\Sigma }_i^{}X.$$
Now the fibre over a point has dimension $`g4`$. So as above $`\mathrm{dim}\mathrm{\Sigma }_i^{}g3`$. Hence, for the closure $`\overline{\mathrm{\Sigma }^{}}`$, we have
$$\mathrm{dim}\overline{\mathrm{\Sigma }^{}}g3.$$
So the image of the projection $`\overline{\mathrm{\Sigma }^{}}C^{}`$ has smaller than dimension $`g2`$. Since $`C^{}`$ is a hypersurface, the result follows.
(b) First consider $`HC^{}`$.
By the remark proceeding the lemma, it suffices to show that for a generic $`HC^{}`$, $`H.C`$ has only one point of multiplicity two and $`H`$ does not pass through one of the $`b_i`$. The first assertion follows as in (a). For the second assertion notice that $`C^{},b_1^{},\mathrm{},b_{2g+2}^{}`$ are distinct hypersurfaces in $`(^{g1})^{}`$. The result follows.
This last remark also deals with the case $`Hb_{i}^{}{}_{}{}^{}`$. ∎
## . §8 Proof of the Generalized Torelli Theorem
We wish to prove
###### (8.1) Theorem.
Let $`C`$ be a smooth projective curve over $``$ of genus $`g1`$. If $`1dg1`$ is an integer then the pair $`(J(C),W^d)`$ determine the curve, that is if $`(J(C),W^d(C))(J(C^{}),W^d(C^{}))`$ for some other smooth projective curve $`C^{}`$ then $`C^{}C`$.
###### Proof.
We may assume $`g4`$ as the cases $`g=1,2,3`$ are covered by the regular Torelli theorem. Furthermore we may reduce to the case $`(g1)/2<d<g1`$ as follows. If $`d=g1`$ we are done by Torelli’s theorem. If $`d<(g1)/2`$ then choose $`n`$ so that $`(g1)/2<ndg1`$. Now we may replace $`W^d`$ by
$$W^{nd}=\underset{n\mathrm{times}}{\underset{}{W^d+W^d+\mathrm{}+W^d.}}$$
We will study the branch locus of the map
Note that we can recover the rational map $`\beta `$ from the information $`(J(C),W^d)`$. Now let $`U_\mathrm{E}\mathrm{E}(d,g)`$ be the open subset defined at the start of §5. We have a morphism $`\beta |_{U_\mathrm{E}}:U_\mathrm{E}(^{g1})^{}`$. Let $`B`$ be the branch locus of $`\beta `$. This is the image of the ramification locus inside $`(^{g1})^{}`$. A closed point $`p`$ is in the ramification locus if and only if $`\beta `$ fails to be a local analytic isomorphism at $`p`$. At this point we break the proof into two cases, the case where $`C`$ is non-hyperelliptic and the case where $`C`$ is hyperelliptic.
First we study the case where $`C`$ is non-hyperelliptic. We will show that $`\overline{B}=C^{}`$. Then $`C`$ can be recovered from this information, see .
First we show that $`\overline{B}C^{}`$. Let $`HC^{}`$. Then $`H.C=p_1+p_2+\mathrm{}+p_{2g2}`$ with the $`p_i`$ distinct. Let $`T(^{g1})^{}`$ be all the hyperplanes having transverse intersection with $`C`$, that is $`T=(^{g1})^{}C^{}`$. The incidence correspondece
$$I=\{(p,H)C\times Tp\mathrm{Supp}H.C\}T$$
is a $`(2g2)`$-sheeted covering space of $`T`$, pg.110. Given $`(u(D),u(D^{}))U_E`$ with $`\beta ((u(D),u(D^{})))=HT`$. It is claimed that there exists an open neighbourhood $`V`$ in the usual topology such that
$$\beta |_V:V\beta (V)$$
is an injection. To see this, first take $`HWT`$, with sheets $`W_1,W_2,\mathrm{},W_{2g2}`$. Let $`\mu _i`$ be the compostion $`WW_iC`$, which is holomorphic. Write $`D=p_1+\mathrm{}+p_d`$. The $`p_i`$ are distinct by choice of $`H`$, so we may find opens $`p_iU_i`$ such that
(1) $`U_iU_j=`$ for $`ij`$
(2) $`U_i\mu _j(W)`$ for some $`j`$.
Writing $`D^{}=p_1^{}+\mathrm{}+p_d^{}`$ we may find similar opens $`U_i^{}`$. Set $`U=U_1\times \mathrm{}\times U_d`$, $`U^{}=U_1^{}\times \mathrm{}\times U_d^{}`$. By condition (1), $`U\times U^{}`$ is an open neighbourhood of $`(p_1+\mathrm{}+p_d,p_1^{}+\mathrm{}+p_d^{})`$ on $`C^{(d)}\times C^{(d)}`$. As the Abel-Jaacobi map is an isomorphism near $`(p_1+\mathrm{}+p_d,p_1^{}+\mathrm{}+p_d^{})`$, as $`(u(D),u(D^{}))W^dd_{\mathrm{smooth}}\times W^dd_{\mathrm{smooth}}`$. We take $`V=\beta ^1(U\times U^{})U_E`$. It is easy to see that this works.
It follows from theorem 7.6, of , that $`\beta `$ is a local isomorphism at $`(u(D),u(D^{}))`$ since this point is in the smooth locus of $`\mathrm{E}(d,g)`$ by lemma (6.1). It remains to show that $`B`$ contains an open dense subset of $`C^{}`$.
By (7.1) there exists an open subset $`VC^{}`$ such that for each $`HV`$,
$$H.C=2p_1+p_2+\mathrm{}+p_{2g3},$$
with the $`p_1,\mathrm{},p_{2g3}`$ are distinct. Since $`g0`$ and
$$K2p_1+p_2+\mathrm{}+p_{2g3},$$
we have that $`H=\overline{\varphi _K(p_1+p_2+\mathrm{}+p_{2g3})}`$. (Notice that there is no $`2`$ in front of $`p_1`$ in the last statement.) After reindexing we may assume that $`p_1,p_2,\mathrm{},p_{g1}`$ span $`H`$ and the tangent line at $`p_1`$ to $`C`$ lies inside $`H`$. Let
$$D=q_1+q_2+\mathrm{}+q_d\mathrm{and}D^{}=q_1^{}+q_2^{}+\mathrm{}+q_d^{}$$
where $`q_i=p_i`$ for $`1id`$ and $`q_i^{}=p_{gi}`$ for $`1id`$. So $`(u(D),u(D^{}))U_\mathrm{E}`$. Let $`z_i`$ (resp. $`z_i^{}`$) be local coordinates centred at $`q_i`$ (resp. $`q_i^{}`$). Since $`q_iq_j`$ (resp. $`q_i^{}q_j^{}`$) for $`ij`$, we have local coordinates $`(z_1,z_2,\mathrm{},z_d)`$ (resp. $`(z_1^{},z_2^{},\mathrm{},z_d^{}`$) on $`C^{(d)}`$ centred at $`(q_1,q_2,\mathrm{},q_d)`$ (resp. $`(q_1^{},\mathrm{},q_d^{})`$). As $`u`$ is an isomorphism around $`D`$ (resp. $`D^{}`$), by (2.1) and (2.3) and as $`\mathrm{dim}\overline{\varphi _K(D)}=d1`$ (resp. $`\mathrm{dim}\overline{\varphi _K(D^{})}=d1`$), we have that $`((z_1,z_2,\mathrm{}z_d),(z_1^{},z_2^{},\mathrm{}z_d^{}))`$ descend to local coordinates on $`W^d\times W^d`$ centred at $`(u(D),u(D^{}))`$.
Choose a basis $`\omega _1,\mathrm{},\omega _g`$ for $`\mathrm{H}^0(C,\mathrm{\Omega }_C^1)`$ and write $`\omega _i=\mathrm{\Omega }_{ij}(z_j)dz_j`$ (resp. $`\omega _i=\mathrm{\Omega }_{ij}^{}(z_j^{})dz_j^{}`$). Let
$$\begin{array}{c}M(z)=\hfill \\ \left(\begin{array}{cccccccc}\mathrm{\Omega }_{11}(z_1)& \mathrm{\Omega }_{12}(z_2)& \mathrm{}& \mathrm{\Omega }_{1,d}(z_d)& \mathrm{\Omega }_{11}^{}(z_1^{})& \mathrm{\Omega }_{12}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{1,d}^{}(z_d^{})\\ \mathrm{\Omega }_{21}(z_1)& \mathrm{\Omega }_{22}(z_2)& \mathrm{}& \mathrm{\Omega }_{2,d}(z_d)& \mathrm{\Omega }_{21}^{}(z_1^{})& \mathrm{\Omega }_{22}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{2,d}^{}(z_2^{})\\ \mathrm{}& \mathrm{}& & \mathrm{}& \mathrm{}& \mathrm{}& & \mathrm{}\\ \mathrm{\Omega }_{g1}(z_1)& \mathrm{\Omega }_{g2}(z_2)& \mathrm{}& \mathrm{\Omega }_{g,d}(z_d)& \mathrm{\Omega }_{g1}^{}(z_1^{})& \mathrm{\Omega }_{g2}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{g,d}^{}(z_d^{})\end{array}\right)\hfill \end{array}$$
and let
$$\begin{array}{c}M^{}(z)=\hfill \\ \left(\begin{array}{cccccccc}\frac{\mathrm{\Omega }_{11}(z_1)}{z_1}& \mathrm{\Omega }_{12}(z_2)& \mathrm{}& \mathrm{\Omega }_{1,d}(z_d)& \mathrm{\Omega }_{11}^{}(z_1^{})& \mathrm{\Omega }_{12}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{1,d}^{}(z_d^{})z_1\\ \frac{\mathrm{\Omega }_{21}(z_1)}{z_1}& \mathrm{\Omega }_{22}(z_2)& \mathrm{}& \mathrm{\Omega }_{2,d}(z_d)& \mathrm{\Omega }_{21}^{}(z_1^{})& \mathrm{\Omega }_{22}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{2,d}^{}(z_2^{})\\ \mathrm{}& \mathrm{}& & \mathrm{}& \mathrm{}& \mathrm{}& & \mathrm{}\\ \frac{\mathrm{\Omega }_{g1}(z_1)}{z_1}& \mathrm{\Omega }_{g2}(z_2)& \mathrm{}& \mathrm{\Omega }_{g,d}(z_d)& \mathrm{\Omega }_{g1}^{}(z_1^{})& \mathrm{\Omega }_{g2}^{}(z_2^{})& \mathrm{}& \mathrm{\Omega }_{g,d}^{}(z_d^{})\end{array}\right)\hfill \end{array}$$
By definition of $`D`$ and $`D^{}`$ the first $`g1`$ columns of $`M(z)`$ are linearly independent in a neighbourhood of $`(u(D),u(D^{}))`$. So $`\mathrm{E}(d,g)`$ is defined by
$$f_i=\mathrm{det}(M(z)_{1,2,\mathrm{},g1,i}),$$
where $`gi2d`$, in a neighbourhood of $`(u(D),u(D^{}))`$. (To see this, use (4.1) as in (6.1)) Now since the tangent line to $`C`$ at $`p_1`$ is inside $`H`$ we have
$$\frac{f_i}{z_1}=\mathrm{det}(M^{}(z)_{1,2,\mathrm{},g1,i})|_{(u(D),u(D^{}))}=0.$$
So the Jacobian matrix, as in the proof of (6.1), reduces to
$$\left(\begin{array}{cccc}0& 0& \mathrm{}& 0\\ \frac{f_g}{z_2}& \frac{f_{g+1}}{z_2}& \mathrm{}& \frac{f_{2d}}{z_2}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ \frac{f_g}{z_d}& \frac{f_{g+1}}{z_d}& \mathrm{}& \frac{f_{2d}}{z_d}\\ \frac{f_g}{z_1^{}}& \frac{f_{g+1}}{z_1^{}}& \mathrm{}& \frac{f_{2d}}{z_1^{}}\\ \mathrm{}& \mathrm{}& & \mathrm{}\\ \frac{f_g}{z_{g1d}^{}}& \frac{f_{g+1}}{z_{g1d}^{}}& \mathrm{}& \frac{f_{2d}}{z_{g1d}^{}}\\ \frac{f_g}{z_{gd}^{}}& 0& \mathrm{}& 0\\ 0& \frac{f_{g+1}}{z_{gd+1}^{}}& \mathrm{}& 0\\ & & \mathrm{}& \\ 0& 0& \mathrm{}& \frac{f_2d}{z_d^{}}\end{array}\right)$$
Arguing as in (6.1) we find that $`(u(D),u(D^{}))`$ is a smooth point of $`\mathrm{E}(d,g)`$. We also see that $`\frac{}{z_1}|_{(u(D),u(D^{}))}`$ is in the null space of the above Jacobian. Hence $`\frac{}{z_1}|_{(u(D),u(D^{}))}`$ is in fact a tangential to $`\mathrm{E}(d,g)`$ at $`(u(D),u(D^{}))`$. In order to show that $`HB`$ it will suffice to show that $`\frac{}{z_1}|_{(u(D),u(D^{}))}`$ maps to zero under the morphism of tangent space induced by $`\beta `$. Let
$$N(z)=\left(\begin{array}{cccccc}\mathrm{\Omega }_{11}(z_1)& \mathrm{}& \mathrm{\Omega }_{1d}(z_d)& \mathrm{\Omega }_{11}^{}(z_1^{})& \mathrm{}& \mathrm{\Omega }_{1,g1d}(z_{g1d})\\ \mathrm{}& & \mathrm{}& \mathrm{}& & \mathrm{}\\ \mathrm{\Omega }_{g1}(z_1)& \mathrm{}& \mathrm{\Omega }_{gd}(z_d)& \mathrm{\Omega }_{g1}^{}(z_1^{})& \mathrm{}& \mathrm{\Omega }_{g,g1d}(z_{g1d})\end{array}\right).$$
So $`N(z)`$ is just the first $`g1`$ columns of $`M(z)`$. In a neighbourhood of $`(u(D),u(D^{}))`$ the morphism $`\beta :U(^{g1})^{}`$ is given by $`z\mathrm{col}.\mathrm{space}N(z)`$. Identify $`(^{g1})^{}(^{g1}^g)`$ we see that $`\beta `$ is the morphism
$$z[\mathrm{det}(N(z)_1):\mathrm{det}(N(z)_2):\mathrm{}:\mathrm{det}(N(z)_g)].$$
Recall that $`N(z)_i`$ is the submatrix of $`N(z)`$ obtained by deleting the $`i`$th row. We may assume that $`\mathrm{det}(N(z)_1)0`$. So we need to show that
$$\frac{}{z_1}|_{(u(D),u(D^{}))}(\frac{\mathrm{det}(N(z)_i)}{\mathrm{det}(N(z)_1)}=0.$$
That is
$$\frac{\mathrm{det}(N(z)_1)}{z_1}.\mathrm{det}(N(z)_i)=\frac{\mathrm{det}(N(z)_i}{z_1}.\mathrm{det}(N(z)_1)$$
after evaluation at $`(u(D),u(D^{}))`$. Let
$$\frac{N(z)}{z_1}$$
be the matrix obtained from $`N(z)`$ by differentiating the first column with respect $`z_1`$. Observe that
$$\mathrm{col}.\mathrm{space}\frac{N(z)}{z_1}|_{(u(D),u(D^{}))}\mathrm{col}.\mathrm{space}N(z)|_{(u(D),u(D^{}))}$$
as the tangent line at $`p_1`$ lies inside $`H`$. It is a general fact from linear algebra that given two $`g\times (g1)`$ matrices $`M,N`$ with $`\mathrm{col}.\mathrm{space}M\mathrm{col}.\mathrm{space}N`$ then for each $`i`$ $`j`$ in the range $`1i,jg`$ we have
$$\mathrm{det}(M_i)\mathrm{det}(N_j)=\mathrm{det}(M_j)\mathrm{det}(N_i).$$
We will include the proof of this statement at the end of this proof for completeness. This shows that $`\overline{B}=C^{}`$.
Now we treat the case that $`C`$ is a hyperelliptic curve. We show that $`\overline{B}=C^{}b_1^{}b_2^{}\mathrm{}b_{2g+2}^{}`$ where the $`b_i`$ are the branch points of the canonical morphism $`\varphi _K:C(^{g1})^{}`$. The proof is almost identical to the above. Here are a few details. The same proof as in the non-hyperelliptic case shows that $`\overline{B}C^{}b_1^{}b_2^{}\mathrm{}b_{2g+2}^{}`$, and similarly we show that $`\overline{B}C^{}`$. To show that $`\overline{B}b_i^{}`$ proceed as follows. From (7.1) we know that for a generic $`Hb_i^{}`$ that
$$H.C=2p_1+p_2+\mathrm{}+p_{2g3}$$
where the $`p_i`$ are distinct and $`p_1=b_1`$. As above we form, after appropriate reindexing,
$$D=q_1+q_2+\mathrm{}+q_d\mathrm{and}D^{}=q_1^{}+q_2^{}+\mathrm{}+q_d^{}.$$
Note, these two divisors are defined exactly as they were before. Also define, as before, $`z_i`$, $`z_i^{}`$, $`M(z)`$, $`M^{}(z)`$ and $`f_i`$. To see that
$$\frac{f_i}{z_1}|_{(u(D),u(D^{}))}=0,$$
first observe that since $`p_1`$ is a branch point, $`J(\varphi _K)|_{q_1}=0`$. Around $`q_1`$,
$$\varphi _K=[\mathrm{\Omega }_{11}(z_1):\mathrm{}:\mathrm{\Omega }_{g1}(z_1)].$$
We may assume that $`\mathrm{\Omega }_{11}(z_1)0`$. Since the Jacobian at $`q_1`$ vanishes we see that
$$\mathrm{\Omega }_{11}(q_1)\frac{\mathrm{\Omega }_{1j}(q_1)}{z_1}|_{q_1}=\frac{\mathrm{\Omega }_{11}(z_1)}{z_1}|_{q_1}\mathrm{\Omega }_{1j}(q_1),$$
which in turn implies
$$\varphi _K(q_1)=[\mathrm{\Omega }_{11}(q_1):\mathrm{}:\mathrm{\Omega }_{g1}(q_1)]=[\frac{\mathrm{\Omega }_{11}}{z_1}:\mathrm{}:\frac{\mathrm{\Omega }_{g1}}{z_1}]_{q_1}.$$
So
$$\frac{f_i}{z_1}|_{(u(D),u(D^{}))}=f_i|_{(u(D),u(D^{}))}=0.$$
Now proceed as before.
Here is the linear algebra result that was needed before.
###### (8.2) Lemma.
Let $`M`$, $`N`$ be two $`g\times (g1)`$ matrices over $``$. If
$$\mathrm{col}.\mathrm{space}M\mathrm{col}.\mathrm{spaceN}$$
then
(6)
$$\mathrm{det}M_i\mathrm{det}N_j=\mathrm{det}M_j\mathrm{det}N_i,$$
for each $`i,j`$ with $`1i,jg`$. Recall that $`M_i`$ is the submatrix of $`M`$ obtained by deleting the $`i`$th row.
###### Proof.
Firstly if $`\mathrm{rank}M<g1`$ then both sides of (6) vanish. So we may assume $`M`$, $`N`$ are of maximal rank and that there column spaces are equal. So $`N=M.H`$ for some $`H\mathrm{Gl}(g1,)`$. The result follows from the observation $`(M.H)_i=M_i.H`$. ∎
###### (8.3) Corollary.
Let $`C`$ and $`C^{}`$ be two smooth projective curves and let $`d`$ be an integer less than or equal to the genus of $`C`$. If $`C^{(d)}C_{}^{}{}_{}{}^{(d)}`$ then $`CC^{}`$.
###### Proof.
This is because the Albanese varaiety $`\mathrm{Alb}(C^{(d)})`$ is isomorphic to $`J(C)`$ and the image of $`C^{(d)}`$ under the Albanese map is $`W^d`$. ∎ |
warning/0001/cond-mat0001143.html | ar5iv | text | # Superconducting and Insulating Behavior in One-Dimensional Josephson Junction Arrays
## 1 INTRODUCTION
The Coulomb blockade of Cooper pair tunneling (CBCPT) is a remarkable phenomena, where a normally superconducting tunnel junction becomes insulating, behaving as a capacitor with a critical voltage for current flow. The CBCPT can be observed when the Josephson coupling energy, $`E_J`$ is the same order of magnitude as the charging energy, $`E_C=e^2/2C`$, where $`C`$ is the junction capacitance. The temperature must be low, $`k_BT<E_C,E_J`$, and most importantly, the junction must experience a high impedance electromagnetic environment. The electrodynamic environment controls the quantum fluctuations of the Josephson phase, and thereby determines whether superconducting (Josephson-like) or insulating (Coulomb blockade) behavior is observed.
When the electrodynamic environment is described by the impedance $`Z_e(\omega )`$, quantum fluctuations of the phase become large if $`Re[Z_e(\omega )]R_Q`$, where $`R_Q=h/4e^2=6.45`$k$`\mathrm{\Omega }`$ is the quantum resistance for Cooper pairs. Large quantum fluctuations of the phase mean that the number difference of Cooper pairs across the junction becomes a sharply defined quantum variable, as $`\mathrm{\Delta }N\mathrm{\Delta }\varphi =1/2`$ for a coherent state. By realizing this high impedance environment with Josephson tunneling, we realize a unique state of charged matter, where many Bosons (Cooper pairs) are condensed into a state with well specified number.
For a single Josephson junction in an arbitrary, linear electrodynamic environment, the effect of quantum fluctuations on Josephson tunneling is well described theoretically for the case $`E_JE_C`$ . However, when a single Josephson junction is placed in the environment of several other Josephson junctions, the theory is much less clear. Theoretical work on Josephson junctions arrays has primarily concentrated on two-dimensional (2D) arrays. A phase diagram is often calculated which maps out insulating and superconducting regions, depending on the various parameters of the junctions, $`E_J`$, $`E_C`$, the normal state tunneling resistance $`R_N`$, or the quasiparticle tunneling resistance, $`R_{qp}`$ (for a review, see ). The theory can be related to experimental work on granular thin films of superconducting metals (for a review see ) and 2D Josephson Junction Arrays fabricated with electron-beam lithography . Finite size effects, and the interplay between the island stray capacitance and the junction capacitance are generally not considered in the theoretical treatment of Josephson junction arrays. However, experiments on 2D arrays show a connection between the array size and the character of the superconductor-insulator (S-I) transition . The Delft group has also studied long, two-dimensional arrays, which approach a 1D parallel array , as well as short 1D series arrays
In this work we will examine experimental results on one-dimensional (1D) arrays of small capacitance Josephson junctions. We will describe the evolution from superconducting to insulating behavior in the arrays, which qualitatively is very similar to that observed in granular thin films and 2D Josephson junction arrays. We argue that the ratio of the junction capacitance to the island stray capacitance is an important parameter for the coupling to a dissipative environment, and therefore is crucial for observation of the S-I transition.
### 1.1 One Dimensional Arrays
Theoretical work on 1D arrays of small capacitance Josephson junctions has analyzed the charge dynamics of 1D arrays in the Coulomb blockade state . Other theoretical analysis has centered on the questions of determining a phase diagram, and for what values of the array parameters an insulating phase (Coulomb blockade) or Superconducting phase (Josephson effect) can be expected . Theoretical analysis of small capacitance 1D arrays usually begins with simplification of the charging energy. In many cases one takes only ”on site” charging, where only the stray capacitance of each electrode to ground, $`C_0`$ is considered. A better approximation to the experiments is to reduce the capacitance matrix to a bi-diagonal form, where $`C_0`$ and the junction capacitance, $`C`$ are taken into account. In experiments, typically $`C>C_0`$. In our experiments, we strive for $`CC_0`$ by making the junction capacitance fairly large ( 1 to 3 fF) and the spacing between junctions very small ($`0.2\mu `$m) which leads to a smaller $`C_0`$.
When $`C>C_0`$ the electrostatic screening length extends over several junctions. The potential due to one excess charge decays exponentially with the characteristic length $`\mathrm{\Lambda }=\sqrt{C/C_0}`$. This potential profile is known as a charge soliton because the profile moves with the tunneling charge. Reducing the full capacitance to only $`C`$ and $`C_0`$ is valid only when a ground plane is located closer to the array than the junction spacing. One sample (B6-22 series described below) had a ground plane located $`1.5\mu `$m from the array, and other samples had no ground plane. Thus the bi-diagonal capacitance matrix is only approximate. In the absence of a ground plane, the potential distribution will decay less steeply than exponential, and go as $`1/r`$ at large distance. The charge soliton actually spreads out, and a slightly more effective polarization of the array occurs over the same characteristic distance, $`\sqrt{C/C_0}`$ .
Not only the electrostatics, but also the electrodynamics of the array is effected by a large $`\sqrt{C/C_0}`$ . For large $`\sqrt{C/C_0}`$ the junctions become strongly coupled, and the dynamics is dominated by Josephson plasmons. To demonstrate this point we consider an infinite 1D array of Josephson junctions. For small phase gradient, when the current is less than the critical current, $`IE_J`$, the Josephson phase-current relation can be linearized, and each junction is an effective inductance, $`L_J=\mathrm{\Phi }_0/2\pi I_c`$, where $`\mathrm{\Phi }_0=h/2e`$ is the flux quantum. Replacing each Josephson junction by a parallel combination of $`L_J`$ and $`C`$, we may calculate the impedance of the 1D network shown in fig 1. The results of such a calculation are shown in fig. 2 for an infinite array, and a finite array with $`N=50`$ junctions, for typical values of $`I_c`$, $`C`$ and $`C_0`$ in our experiments.
For a finite array, we see a set of discrete resonances which are the Josephson plasmon modes. The width of these resonances is determined by the terminating impedance, $`Z_0`$. For the infinite array, these resonances form a dense set of modes, where the impedance is much larger than $`Z_0`$. At frequencies much less than the Josephson plasma frequency, $`\omega _p=\sqrt{8E_JE_C}`$ and for $`CC_0`$ the real impedance can be written as,
$$Z_A(\omega )=\sqrt{\frac{L_J}{C_0}}=R_Q\sqrt{\frac{4E_C}{E_J}}\sqrt{\frac{C}{C_0}},\omega <\omega _p=\frac{\sqrt{4E_JE_C}}{\mathrm{}}$$
To observe the CBCPT, it is necessary that $`E_CE_J`$ and both $`E_C,E_Jk_BT`$. If $`E_CE_J`$ the maximum Cooper pair current (”Zener current” ) is suppressed exponentially and becomes immeasurable, and if $`E_CE_J`$ the threshold voltage is exponentially suppressed and becomes immeasurable . Thus, $`Z_AR_Q`$ requires that $`CC_0`$. This impedance does not come from electromagnetic fields, as in a usual transmission line, but rather due to long wave-length Josephson plasmons involving the phases of many junctions.
The large ratio $`\sqrt{C/C_0}10`$ in our arrays not only results in a large impedance, and therefore a Coulomb blockade, but also influences the effect of random offset charges. Simulations show that the static background potential due to random offset charge is smoothed by the averaging effects of the large ratio $`C/C_0`$ . One would expect that pinning of the charge in some local energy minimum in the array could be practically eliminated. However, simulations show that incoherent charge transport is substantially effected by static random background charge, even in the limit $`C>C_0`$ . To our knowledge, the effect of random offset charges on coherent Cooper pair tunneling has not been investigated theoretically.
## 2 EXPERIMENTAL RESULTS
The arrays are actually serially connected SQUIDs (Superconducting Quantum Interference Devices). The SQUID geometry allows one to externally impose a phase shift between neighboring superconducting electrodes by application of a magnetic flux. This external phase shift can be considered as effectively tuning the Josephson coupling energy with an external magnetic field, $`E_J(B)`$. Figure 3 shows scanning electron micrographs of two different arrays. The electrodes were Al, deposited on an oxidized Si substrate, with an Al<sub>2</sub>O<sub>3</sub> tunnel barrier. At the right hand side of fig.3 we see how the arrays were terminated. Two different arrangements were used. Figure 3a shows a 2-point termination, where a wide strip connects to the first SQUID. This 2-point arrangement had the disadvantage that expelled magnetic flux from the wide strip focused more flux into the edge loops. Consequently the two edge junctions had a smaller period of modulation with magnetic field. Figure 3b shows the 4-point termination, where no edge effect with the magnetic field was observed. For the 4-point termination, the junctions in the leads are not SQUIDs, and thus do not have a tunable $`E_J`$.
Table 1 shows the main parameters of the arrays studied thus far. Samples with the same chip code (B6-##) were made simultaneously on the same chip. The samples are arranged with decreasing $`E_{J0}/E_C`$. The note indicates whether the array was terminated with a 2 point or 4 point lead configuration. The value of $`E_{J0}=(\mathrm{\Phi }_0/2\pi )I_c=R_Q\mathrm{\Delta }_0/2R_N`$ was determined from the measured normal state resistance $`R_N`$ and the superconducting energy gap, $`\mathrm{\Delta }_0`$. The charging energy $`E_C=e^2/2C`$ was calculated from the observed area of the junction and the specific capacitance $`c_s=45`$fF$`/\mu `$m<sup>2</sup>. Depending on the array parameters, different behavior of the arrays could be observed. We divide these into three different regions discussed below.
### 2.1 Superconducting State
For the samples in table 1 with a smaller ratio $`L_J/C_0`$, the current-voltage (I-V) characteristic exhibited Josephson-like behavior at zero magnetic field. Figure 4 shows the I-V curve of sample B6-21a. Here we see a nearly vertical ”super current” branch of the I-V curve near zero voltage, and a successive set of vertical quasi-particle branches separated in voltage by $`2\mathrm{\Delta }_0/e`$. The dc load in the bias circuit was such that we could not trace out the negative differential resistance between these vertical quasi-particle branches. At higher voltages the successive quasi-particle branches merge into one flat branch with a current independent of voltage. At even higher voltages, the I-V curve eventually switches to the normal tunneling branch.
The I-V curve shown in fig. 4 is not that expected for arrays of classical, non-interacting Josephson junctions. As the bias is increased in such arrays we would expect an increasing series of switching currents, starting from the lowest critical current in the array. For these small capacitance, strongly coupled Josephson junctions, we rather observe a decrease of the switching current as we move out in voltage, as well as an increase of the retrapping current. We have no quantitative explanation for this observed behavior. We speculate that it has to do with electromagnetic interaction of the junctions, which are strongly coupled due to the very small $`C_0`$.
As the magnetic field is increased, the critical current of all junctions is decreased. The slope of the ”critical current” branch also increases, and as $`E_J`$ is further suppressed, a distinct threshold voltage emerges as the Coulomb blockade of Cooper pair tunneling (CBCPT) becomes observable (see fig. 5)
### 2.2 Superconductor - Insulator Transition
Figure 5 shows three I-V curves at different magnetic fields. The nearly vertical line (I-V at $`B=57`$G) has a finite slope corresponding to about $`300`$ k$`\mathrm{\Omega }`$. The critical current and the hysteretic switching between the ”supercurrent branch” and the quasiparticle branch is off scale for this curve. As the magnetic field is increased the I-V curve develops a very high resistance state for voltages below a threshold voltage, which is the CBCPT. The curve labeled $`B=70`$G in fig 5 shows a hysteretic I-V curve in this Coulomb blockade state, meaning that ”back bending” or region of negative differential resistance exists at low current. We will discuss this hysteresis in the next section. The intermediate curve ($`B=66`$G)in fig. 5 shows remnants of both types of hysteresis.
Figure 5 shows that the crossover to the CBCPT is characterized by a set of very nonlinear I-V curves. We would like to know the linear response of the array as in undergoes this transition from Josephson-like to Coulomb blockade behavior. To this end we have devised a measurement of the zero-bias resistance, $`R_0lim_{V0}dV_{rms}/dI_{rms}`$ under the condition that the power dissipation by the array is constant. $`R_0`$ is measured by phase sensitive detection of a small signal excitation with two lock-in amplifiers. The excitation is adjusted so that the product of the two signals $`(dI)_{rms}(dV)_{rms}=10^{16}`$ Watt.
$`R_0`$ could be measured versus temperature and magnetic field. Figure 6 displays the results of such measurements. At low magnetic fields (bottom curves of fig. 6), $`R_0`$ decreases as the array is cooled below the transition temperature of the Al ($`T_C=1.2`$K) and eventually becomes flat, showing a temperature independent resistive state as $`T0`$. The finite resistance corresponds to the slope of the ”critical current” branch of the I-V curve (see curve $`B=57`$ in fig. 5). Experiments indicate that in longer arrays the resistance of this ”flat tail” decreases, indicating the non-zero resistance in this ”superconducting” state is due to the finite size of the array. When the magnetic field is increased, the resistance rises and the ”flat tail” remains. Further increasing the magnetic field causes a transition to a different kind of behavior, where $`R_0`$ increases as $`T0`$, indicating the development of a Coulomb blockade.
We have previously shown how length scaling of $`R_0`$ could be used to determine the critical point between superconducting and insulating behavior. Length scaling analysis allowed us to determine a $`T=0`$ critical point where $`R_0`$ is independent of length. The critical point separates those curves with a ”flat tail” in $`R_0(T)`$ from those with increasing $`R_0`$ as $`T0`$. In the experiment shown in fig. 6 we find that this transition occurs at $`R_0(T=0)h/4e^2=6.45`$k$`\mathrm{\Omega }`$ (see the arrow in fig. 6). The transition does not always occur at $`R_Q`$, and the measurements displayed in fig. 6 were special in that the array was terminated with a 4 point configuration. The 4 point termination should reduce any series measurement of dissipation arising at the edge of the array. However, one junction is probably not enough to effectively isolate the array from the environment. Further experiments are needed to determine if this transition at $`R_0=R_Q`$ is universal.
### 2.3 Insulating State
When the CBCPT is well developed in the arrays, $`R_0`$ is essentially infinite, and we have a zero current state below the threshold voltage. Figure 7 shows how the threshold voltage changes with magnetic field on the insulating side of the S-I transition. We typically find that $`V_t(B)`$ exhibits a sharp cusp as $`E_J0`$ when $`\mathrm{\Phi }_{ext}\mathrm{\Phi }_0/2`$ ($`B=78`$ G for this sample). For samples B6-22 and B6-21 where we could measure length effects, we found that $`V_t`$ was roughly proportional to length. The threshold voltage depends on the magnetic field in a periodic way, with period corresponding the flux quantum in each loop. The magnitude of the current is also periodic in the magnetic field. These observations prove that the current and the threshold voltage are due to Cooper pair tunneling. Although the Cooper pair tunneling is a coherent process, we measure dissipation (finite current and non-zero voltage) because the pair tunneling interacts with dissipative degrees of freedom.
As in the case of a single small capacitance Josephson junction, dissipation arises from excitation of the electromagnetic environment or quasi-particle tunneling . Quasi- particle tunneling becomes important when the voltage drop across any junction exceeds the gap voltage, $`2\mathrm{\Delta }/e=400\mu `$V. It is important to note that although the voltage over the entire array may be greater than $`2\mathrm{\Delta }/e`$, this voltage is distributed over many junctions. The charge soliton length for Cooper pair charge solitons is determined by an effective capacitance of the junction , $`\lambda _s=\sqrt{C_{eff}/C_0}=\sqrt{(2e/2\pi V_C)/C_0}`$. The critical voltage $`V_C(E_J/E_C)`$ goes to zero exponentially for $`E_JE_C`$ . Hence, $`\lambda _s\mathrm{}`$ at the S-I transition and near this transition the voltage is dropped uniformly over the entire array.
In the insulating state, a very interesting model for charge transport has been developed, which has a direct duality to flux transport in long Josephson junctions . A static version ($`I=0`$) of this model has been used to explain the magnetic field dependence of the threshold voltage . We have been able to extend the model to the time domain, calculating the dc I-V curve and fitting it to the data. This dynamic model is able to explain the hysteresis observed in the dc IV curve (see fig. 8 and the curve labeled $`B=70`$G in fig. 5).
In the insulating state, the Josephson junctions in the 1D chain are described in terms of the dimensionless quasi-charge $`\chi `$. If a phenomenological resistance $`R`$ and inductance $`L`$ are introduced together with the capacitance $`C_0`$, a sine-Gordon like model can be derived to describe the dynamics of $`\chi (x,t)`$. Soliton solutions exist for this model, and the inductive term is what gives the charge soliton mass, or inertia. We may simplify the full sine-Gordon-like model by dropping the second spatial derivative in order to capture the basics of the dynamic behavior. The array is then viewed as one lumped element, which is valid if the array is shorter than the soliton length, as is the case near the S-I transition. Alternatively, the second time and spatial derivatives may combined by assuming a traveling wave solution to the sine-Gordon-like model. Either approach results in the following equation :
$$\frac{V+RI_Z}{V_C}=\beta \ddot{\chi }+\dot{\chi }+saw(\chi )$$
where $`V`$ is the applied voltage, and the dc current is given by the time average $`<\dot{\chi }>`$. Here saw$`(\chi )`$ is a $`2\pi `$ periodic function describing the charging and discharging of the junctions capacitance by single Cooper pair tunneling events. The dot refers to differentiation with respect to dimensionless time, $`t/RC`$. For the lumped element case, the dimensionless parameter,
$$\beta =\frac{L}{R^2\left(\frac{2e}{2\pi V_C}\right)}$$
determines the amount of hysteresis observed in the dc I-V curve. We have modeled the dissipation arising from Zener transitions in the Bloch energy bands of the single junction Coulomb blockade . To make the model simple we assumed that the resistance is zero below the Zener threshold current, $`I_Z`$, and characterized by a linear resistance $`R`$ above $`I_Z`$.
The parameters $`V_C`$, $`R`$, and $`I_Z`$ are fixed for a particular magnetic field ($`E_J/E_C`$), and are determined from the observed I-V curve. The parameter $`\beta `$ or $`L`$ is then adjusted to fit the observed amount of hysteresis. Good fits to the data can be made, as seen in fig. 8. However, the inductance $`L`$ required to explain the observed hysteresis is extremely large, $`10^2`$ H. It is certainly not due to electromagnetic inductance ($`10^{14}`$ H) and too small to be due to Josephson inductance ($`N\mathrm{\Phi }_0/2\pi I_C10^4`$ H). We do observe that the value of $`L`$ obtained from these fits does diverge as $`I_C0`$ as expected for the Josephson inductance .
Although the origin of this large inductance is not understood, we emphasize that an inductive term is absolutely necessary if we are to explain the hysteresis with any dynamic model. We further stress that this hysteresis is clearly associated with the presence of Josephson Coupling in the arrays. Such hysteresis has not been observed in any normal tunnel junction arrays to the best of our knowledge. In our arrays, the normal state behavior shows no clear zero current state. The normal sate is achieved at large magnetic fields ($`B1000`$G) where the superconducting energy gap is suppressed. Only a weak remnant of the Coulomb blockade exists in the normal state because the normal tunnel junction resistance is the order of the quantum resistance.
## 3 CONCLUSION
We have given an overview of the measured transport characteristics of 1D small capacitance Josephson junction arrays. The interplay between Josephson and Charging energies is clearly seen as a transition between Josephson-like to Coulomb blockade I-V curves. Coulomb blockade is observed even when $`E_JE_C`$, depending on the length of the array. We point to the importance of the stray capacitance $`C_0`$ and the long wavelength Josephson plasmons in determining the coupling to dissipation. A quantum phase transition can be observed by measuring the temperature dependence of the zero bias resistance as the effective Josephson coupling is modulated in the arrays. In the insulating state, the hysteretic current voltage characteristics can be explained by a dynamic model which is dual to the usual RSJ model for Josephson junctions.
## ACKNOWLEDGMENTS
We wish to acknowledge support from the Swedish NFR, and the EU grant 22953 CHARGE and the EU grant SMT4-CT96-2049 SETamp. Samples were made at the Swedish Nanometer Laboratory. |
warning/0001/math0001090.html | ar5iv | text | # Superbridge index of composite knots
## 1. Introduction
Throughout this article a knot is a piecewise smooth simple closed curve embedded in the three dimensional Euclidean space $`^3`$. Two knots are equivalent if there is a piecewise smooth autohomeomorphism of $`^3`$ mapping one knot onto the other. The equivalence class of a knot $`K`$ will be called the knot type of $`K`$ and denoted by $`[K]`$.
The crookedness of a knot $`K`$ embedded in $`^3`$ with respect to a unit vector $`\stackrel{}{v}`$ is the number of connected components of the preimage of the set of local maximum values of the orthogonal projection $`K\stackrel{}{v}`$, denoted by $`b_\stackrel{}{v}(K)`$. Figure 1 illustrates an example. For any open subarc $`S`$ of a knot $`K`$, the crookedness of $`S`$ with respect to $`\stackrel{}{v}`$, denoted by $`b_\stackrel{}{v}(KS)`$, can be defined similarly using the projection $`S\stackrel{}{v}`$. The superbridge number and the superbridge index of $`K`$, denoted by $`s(K)`$ and $`s[K]`$, are defined to be “$`\mathrm{max}b_\stackrel{}{v}(K)`$” and “$`\mathrm{min}\mathrm{max}b_\stackrel{}{v}(K)`$”, respectively, where the maximum is taken over all unit vectors and the minimum taken over all equivalent embeddings of $`K`$. This invariant was introduced by Kuiper who computed the superbridge index for all torus knots.
###### Theorem A (Kuiper).
For any two coprime integers $`p`$ and $`q`$, satisfying $`2p<q`$, the torus knot of type $`(p,q)`$ has superbridge index $`\mathrm{min}\{2p,q\}`$.
The bridge index $`b[K]`$ can be defined in a similar way by “$`\mathrm{min}\mathrm{min}b_\stackrel{}{v}(K)`$.” One of the most well-known theorem about bridge index is
###### Theorem B (Schubert).
Given two knots $`K_1`$ and $`K_2`$, any connected sum<sup>1</sup><sup>1</sup>1Since $`K_1`$ and $`K_2`$ are not oriented, their (unoriented) connected sum may not be unique. $`K_1\mathrm{}K_2`$ satisfies
$$b[K_1\mathrm{}K_2]=b[K_1]+b[K_2]1.$$
This work is an attempt to find a similar formula for the superbridge index. A proof of Schubert’s theorem in a more generalized context can be found in .
Let $`\beta [K]`$ denote the braid index, i.e., the minimal number of strings among all braids whose closures are equivalent to $`K`$. According to Kuiper, the superbridge index of a nontrivial knot is always greater than the bridge index and not greater than twice the braid index .
###### Theorem C (Kuiper).
If $`K`$ is a nontrivial knot, then
$$b[K]<s[K]2\beta [K].$$
Kuiper used Milnor’s total curvature to prove the first inequality . The closed braid constructed by Kuiper used to prove the second inequality is discussed in Section 3. From Theorem B and Theorem C, we obtain
###### Corollary 1.
If $`K_1`$ and $`K_2`$ are nontrivial knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]4.$$
## 2. Theorems and Conjectures
###### Theorem 1.
If $`K_1`$ and $`K_2`$ are nontrivial knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]\mathrm{max}\{2\beta [K_1]+\beta [K_2],\beta [K_1]+2\beta [K_2]\}1.$$
###### Theorem 2.
If $`K_1,K_2`$ are torus knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]\mathrm{max}\{s[K_1]+b[K_2],b[K_1]+s[K_2]\}1.$$
The next corollary shows that the equality in Theorem 2 holds in infinitely many cases.
###### Corollary 2.
Let $`p_i2`$ and let $`K_i`$ be the torus knot of type $`(p_i,p_i+1)`$, for $`i=1,2`$. Then
$$s[K_1\mathrm{}K_2]=p_1+p_2.$$
Proof: By Theorem A, $`s[K_i]=p_i+1`$. Since $`b[K_i]=p_i`$, from Theorem B, Theorem C and Theorem 2, we obtain $`p_1+p_21<s[K_1\mathrm{}K_2]p_1+p_2`$.∎
Using the first inequality in Theorem C, we obtain the following generalization of \[7, Corollary 11\].
###### Corollary 3.
If $`K_1,K_2`$ are torus knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]s[K_1]+s[K_2]2.$$
The inequality in Theorem 2 is equivalent to
$$s[K_1]+s[K_2]s[K_1\mathrm{}K_2]\mathrm{min}\{s[K_1]b[K_1],s[K_2]b[K_2]\}+1.$$
If $`K_i`$ is a torus knot of type $`(p_i,q_i)`$ with $`2p_i<q_i`$, the right hand side of the above inequality is equal to $`\mathrm{min}\{p_1,p_2,q_1p_1,q_2p_2\}+1`$, which can be arbitrarily large. Therefore we have
###### Corollary 4.
The difference $`s[K_1]+s[K_2]s[K_1\mathrm{}K_2]`$ can be arbitrarily large.
We conjecture that Theorem 2 and Corollary 3 are true for any knots:
###### Conjecture 1.
Any connected sum of two knots $`K_1`$ and $`K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]\mathrm{max}\{s[K_1]+b[K_2],b[K_1]+s[K_2]\}1.$$
###### Conjecture 2.
If $`K_1`$ and $`K_2`$ are nontrivial knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]s[K_1]+s[K_2]2.$$
As Corollary 3 follows from Theorem 2, Conjecture 2 follows from Conjecture 1. The readers may wonder whether the inequality
$$s[K_1\mathrm{}K_2]\mathrm{max}\{s[K_1]+b[K_2],b[K_1]+s[K_2]\}1$$
would be true. So far no reasonable lower bound formula for $`s[K_1\mathrm{}K_2]`$ has been found. We do not even know if the following is true.
###### Conjecture 3.
If $`K_1`$ and $`K_2`$ are nontrivial knots, any connected sum $`K_1\mathrm{}K_2`$ satisfies the inequality
$$s[K_1\mathrm{}K_2]>\mathrm{max}\{s[K_1],s[K_2]\}.$$
In Table 1, the symbols used for factors of $`K`$ indicate the prime knots as in the knot tables of . The knots $`3_1,5_1,7_1,8_{19},9_1`$ are torus knots of type $`(2,3),(2,5),(2,7),(3,4),(2,9)`$, respectively. Theorem 2 is used to find upper bounds of superbridge index for the connected sums of pairs of these knots. There are three among them for which Corollary 3 also applies. For the others, we used the inequality
(1)
$$2s[K]p[K]$$
to find upper bounds, where $`p[K]`$ is the polygon index , i.e., the minimal number of straight edges required to present the knot type of $`K`$. Using the polygonal knots given in , we verified that the inequality
$$p[K_1\mathrm{}K_2]p[K_1]+p[K_2]4$$
of \[7, Theorem 8\] can be applied to find upper bounds of $`p[K]`$ as given in the table. The nine-edged polygonal knot<sup>2</sup><sup>2</sup>2 It has vertices at $`(30,0,10)`$, $`(10,20,30)`$, $`(27,35,70)`$, $`(0,30,10)`$, $`(0,40,10)`$, $`(4,7,8)`$, $`(16,6,21)`$, $`(18,32,36)`$, $`(30,0,10)`$. Figure 2 is its projection into the $`xy`$-plane. of Figure 2 is a connected sum of a trefoil knot and a figure eight knot. It has polygon index 9 because it does not appear in the list of containing all eight-edged knots.
The values marked with $``$ are conjectured using Theorem A, Conjecture 3 and \[5, Table 1\]. If Conjecture 3 is not true for any of them, the correct value will be one less than as given in the table. For all others, Theorem B and Theorem C are used to determine strict lower bounds.
The next two sections describe the constructions and their properties required to prove Theorem 1 and Theorem 2. Section 5 contains the proofs.
## 3. Closed braids
Let i, j, k denote the standard basis vectors of $`^3`$ and let $`\eta `$ be the trivial knot given by the embedding $`(x,y)(x,y,x^2)`$ of the circle $`x^2+y^2=1`$. By \[9, Lemma 4.1\], we know that $`s(\eta )=2`$. Therefore, for any unit vector $`\stackrel{}{v}`$, either $`b_\stackrel{}{v}(\eta )=1`$ or $`b_\stackrel{}{v}(\eta )=2`$. Let $`N=\{\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤S^2v_3>0,b_\stackrel{}{v}(\eta )=2\}.`$ This is an open subset of $`S^2`$ satisfying the condition that $`b_\stackrel{}{v}(\eta )=2`$ if and only if $`\stackrel{}{v}N(N)`$. Two projections of $`N`$ and $`N`$ are shown in Figure 3.
###### Lemma 1.
Let $`G_{\rho ,\alpha }(t)=\rho \mathrm{sin}(t\alpha )(1\rho ^2)^{1/2}\mathrm{sin}2t`$.
1. For any $`\alpha `$, there is a unique positive number $`\xi (\alpha )[1/\sqrt{2},2/\sqrt{5}]`$ such that the function $`G_{\xi (\alpha ),\alpha }(t)`$ has a multiple root.
2. $`N`$ has a parametrization $`\alpha (\xi (\alpha )\mathrm{cos}\alpha ,\xi (\alpha )\mathrm{sin}\alpha ,(1\xi (\alpha )^2)^{1/2})`$.
Proof: (a) If $`t_0`$ is a multiple root of $`G_{\rho ,\alpha }(t)`$, then
$`G_{\rho ,\alpha }(t_0)`$ $`=`$ $`\rho \mathrm{sin}(t_0\alpha )(1\rho ^2)^{1/2}\mathrm{sin}2t_0=0,`$
$`G_{\rho ,\alpha }^{}(t_0)`$ $`=`$ $`\rho \mathrm{cos}(t_0\alpha )2(1\rho ^2)^{1/2}\mathrm{cos}2t_0=0.`$
Eliminating $`\alpha `$, we get $`\rho =((1+3\mathrm{cos}^22t_0)/(2+3\mathrm{cos}^22t_0))^{1/2}`$. Therefore the inequality $`1/\sqrt{2}\rho 2/\sqrt{5}`$ holds.
Suppose $`1/\sqrt{2}<\rho <2/\sqrt{5}`$, then $`1/2<(1/\rho ^21)^{1/2}<1`$. As illustrated in Figure 4, there are eight distinct values of $`\alpha `$ modulo $`2\pi `$, such that the graphs of $`p(t)=\mathrm{sin}(t\alpha )`$ and $`q(t)=(1/\rho ^21)^{1/2}\mathrm{sin}2t`$ are tangent at some point. For these values of $`\alpha `$, the function $`G_{\rho ,\alpha }(t)`$ has double roots.
If $`\alpha =k\pi \pm \pi /4`$, $`k=0,1`$, then $`\rho =1/\sqrt{2}`$. In these cases, the graphs of $`p(t)`$ and $`q(t)`$ are tangent at $`t_0=\pi \alpha `$, where $`G_{\rho ,\alpha }(t)`$ has a double root. If $`\alpha =k\pi /2`$, $`k=0,1,2,3`$, then $`\rho =2/\sqrt{5}`$. In these cases, the graphs of $`p(t)`$ and $`q(t)`$ are tangent at $`t_0=\pi \alpha `$, where $`G_{\rho ,\alpha }(t)`$ has a triple root. This finishes the proof of part (a) except the uniqueness which we omit.
(b) For a unit vector $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$, the projection $`\eta \stackrel{}{v}`$ is parametrized by
(2)
$$f_\stackrel{}{v}(t)=v_1\mathrm{cos}t+v_2\mathrm{sin}t+v_3\mathrm{cos}^2t.$$
Suppose $`0<v_3<1`$, then there is a unique number $`\alpha _\stackrel{}{v}`$ modulo $`2\pi `$ such that $`\mathrm{cos}\alpha _\stackrel{}{v}=v_1(1v_3^2)^{1/2}`$ and $`\mathrm{sin}\alpha _\stackrel{}{v}=v_2(1v_3^2)^{1/2}`$. Substituting $`\rho =(1v_3^2)^{1/2}`$, we get
(3)
$$f_\stackrel{}{v}(t)=\rho \mathrm{cos}(t\alpha _\stackrel{}{v})+(1\rho ^2)^{1/2}\mathrm{cos}^2t.$$
If $`\stackrel{}{v}N`$, then $`f_\stackrel{}{v}^{}(t)=0`$ has a multiple root. Since $`f_\stackrel{}{v}^{}(t)=G_{\rho ,\alpha _\stackrel{}{v}}(t)`$, we know that $`N`$ has the required parametrization. The projection of $`N`$ into the $`xy`$-plane in Figure 3 is the graph of the polar equation $`\rho =\xi (\alpha )`$.∎
###### Lemma 2.
Let $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$ be a unit vector. Then
$$b_\stackrel{}{v}(\eta \eta _+)=\{\begin{array}{cc}1\hfill & \text{if }\stackrel{}{v}N\text{ or }\mathrm{min}\{v_1,v_3\}>0\hfill \\ 0\hfill & \text{if }\stackrel{}{v}N,v_1<0\text{ and }v_3>0,\hfill \end{array}$$
where $`\eta _+=\eta \{(x,y,z)x>0\}`$.
Proof: Again we use the parametrizations (2) and (3) for $`\eta `$. We have
$`f_\stackrel{}{v}^{}(t)`$ $`=`$ $`v_1\mathrm{sin}t+v_2\mathrm{cos}tv_3\mathrm{sin}2t`$
$`=`$ $`\rho \mathrm{sin}(t\alpha _\stackrel{}{v})(1\rho ^2)^{1/2}\mathrm{sin}2t.`$
Case 1. Suppose $`\stackrel{}{v}N`$ and $`v_1>0`$. Then $`f_\stackrel{}{v}^{}(\pi /2)=v_1<0<v_1=f_\stackrel{}{v}^{}(\pi /2)`$. Therefore $`b_\stackrel{}{v}(\eta \eta _+)=1`$.
Case 2. Suppose $`v_3>1/\sqrt{2}`$. Since $`0\rho <(1\rho ^2)^{1/2}`$, we have
(4)
$$f_\stackrel{}{v}^{}(\pi /4)<0<f_\stackrel{}{v}^{}(\pi /4)\text{ and }f_\stackrel{}{v}^{}(5\pi /4)<0<f_\stackrel{}{v}^{}(3\pi /4).$$
Therefore there are two local maximum points, one in each of the two intervals $`(\pi /4,\pi /4)`$ and $`(3\pi /4,5\pi /4)`$. Therefore $`\stackrel{}{v}N`$ and $`b_\stackrel{}{v}(\eta \eta _+)=1`$.
Case 3. Suppose that $`\stackrel{}{v}N`$ and $`v_3=1/\sqrt{2}`$, then $`\rho =(1\rho ^2)^{1/2}=1/\sqrt{2}`$ and $`\alpha _\stackrel{}{v}k\pi /2+\pi /4`$ for any integer $`k`$. Therefore condition (4) holds, and again we have $`b_\stackrel{}{v}(\eta \eta _+)=1`$.
Case 4. Suppose that $`\stackrel{}{v}N`$ and $`v_3<1/\sqrt{2}`$. Then $`1/\sqrt{5}<v_3<1/\sqrt{2}`$, hence $`1/\sqrt{2}<\rho <2/\sqrt{5}`$ and $`1/2<(1/\rho ^21)^{1/2}<1`$. The circle $`x^2+y^2=\rho ^2`$ on the unit sphere meets $`N`$ at eight distinct points as shown in Figure 5. Let $`\alpha _0`$ be the smallest positive number that $`G_{\rho ,\alpha _0}(t)`$ has double roots. Since $`\stackrel{}{v}N`$, it is on one of the four open arcs of the circle inside $`N`$. These arcs correspond to the four intervals for $`\alpha _\stackrel{}{v}`$ given in the table below.
| (a) | (b) | (c) | (d) |
| --- | --- | --- | --- |
| $`\left|\alpha _\stackrel{}{v}\right|<\alpha _0`$ | $`\left|\alpha _\stackrel{}{v}\pi /2\right|<\alpha _0`$ | $`\left|\alpha _\stackrel{}{v}\pi \right|<\alpha _0`$ | $`\left|\alpha _\stackrel{}{v}3\pi /2\right|<\alpha _0`$ |
The four pairs of $`p(t)`$’s in Figure 4 correspond to the endpoints of these intervals. From Figure 4, we easily see that the sign of $`f_\stackrel{}{v}^{}(t)=\rho (p(t)q(t))`$ changes from positive to negative once in each of the intervals $`(\pi /2,\pi /2)`$ and $`(\pi /2,3\pi /2)`$. Therefore $`b_\stackrel{}{v}(\eta \eta _+)=1`$.
Case 5. Suppose $`\stackrel{}{v}N`$ and $`v_10`$. If $`v_1=0`$, any local extremum of $`f_\stackrel{}{v}`$ occurs only at $`(0,1,0)`$ or $`(0,1,0)`$. If $`v_1<0`$, then $`f_\stackrel{}{v}^{}(3\pi /2)=v_1<0<v_1=f_\stackrel{}{v}^{}(\pi /2)`$. Therefore $`b_\stackrel{}{v}(\eta \eta _{})=1`$ where $`\eta _{}=\eta \{(x,y,z)x<0\}`$. Since $`b_\stackrel{}{v}(\eta )=1`$, we obtain $`b_\stackrel{}{v}(\eta \eta _+)=0`$.∎
Suppose $`n`$ is a positive integer and $`K`$ is a knot parametrized by
$$K(t)=((1+\lambda _1(t))\mathrm{cos}nt,(1+\lambda _1(t))\mathrm{sin}nt,\lambda _2(t)+\mathrm{cos}^2nt)$$
over any interval of length $`2\pi `$, for some smooth periodic functions $`\lambda _1`$ and $`\lambda _2`$ with period $`2\pi `$ satisfying the conditions
(5) $`\lambda _1(t)^2+\lambda _2(t)^2<1,`$
(6) $`\lambda _1(t)=\lambda _2(t)=0\text{if }|t|3\pi /4n,`$
$`\lambda _1(t),\lambda _2(t)\text{are locally constant and negative}`$
(7) $`\text{if }5\pi /4n|t|\pi \text{and }\mathrm{cos}nt1/\sqrt{2}.`$
For any $`\epsilon `$ with $`0\epsilon 1`$, we define
(8)
$$K^\epsilon (t)=((1+\epsilon \lambda _1(t))\mathrm{cos}nt,(1+\epsilon \lambda _1(t))\mathrm{sin}nt,\epsilon \lambda _2(t)+\mathrm{cos}^2nt).$$
Then $`K^\epsilon `$ is a knot isotopic to $`K`$ and is the closure of the $`n`$-braid $`K^\epsilon \{(x,y,z)xyx\}`$ when $`0<\epsilon 1`$. When $`\epsilon =0`$, $`K^\epsilon `$ is an $`n`$-fold covering of $`\eta `$. Since $`K_+^\epsilon =K^\epsilon \{(x,y,z)x>0\}`$ is the union of $`n`$ disjoint parallel copies of $`\eta _+`$ up to radial scaling about the $`z`$-axis, we have $`b_\stackrel{}{v}(K^\epsilon K_+^\epsilon )=nb_\stackrel{}{v}(\eta \eta _+),`$ hence by Lemma 2, we obtain
(9)
$$b_\stackrel{}{v}(K^\epsilon K_+^\epsilon )=\{\begin{array}{cc}n\hfill & \text{if }\stackrel{}{v}N\text{ or }\mathrm{min}\{v_1,v_3\}>0\hfill \\ 0\hfill & \text{if }\stackrel{}{v}N,v_1<0\text{ and }v_3>0\hfill \end{array}$$
for any unit vector $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$.
By , we know that there is a number $`\epsilon ^{}>0`$ such that $`s(K^\epsilon )=2n`$ whenever $`0<\epsilon \epsilon ^{}`$. Let $`0<\epsilon \epsilon ^{}`$ and let
$$N^\epsilon =\{\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤S^2v_3>0,b_\stackrel{}{v}(K^\epsilon )=2n\}.$$
For any $`\epsilon `$, $`N^\epsilon `$ is an open set intersecting $`N`$ in a neighborhood of $`𝐤`$. Since $`N^0=N`$ and is connected, $`N^\epsilon `$ is also connected whenever $`0<\epsilon \epsilon ^{\prime \prime }`$ for some $`\epsilon ^{\prime \prime }(0,\epsilon ^{}]`$.
Suppose $`N^\epsilon \{(x,y,z)x<0\}N\{(x,y,z)x<0\}`$. Then there exists a unit vector $`\stackrel{}{v}NN^\epsilon \{(x,y,z)x<0\}`$. Since the projection $`K^\epsilon \stackrel{}{v}`$ assumes no local maximum in $`K^\epsilon \{(x,y,z)x=0\}`$, and by the equation (9), we have
$$b_\stackrel{}{v}(K^\epsilon K_{}^\epsilon )=b_\stackrel{}{v}(K^\epsilon K_+^\epsilon )+b_\stackrel{}{v}(K^\epsilon K_{}^\epsilon )=b_\stackrel{}{v}(K^\epsilon )=2n.$$
There exists an open neighborhood $`V`$ of $`\stackrel{}{v}`$ contained in $`N^\epsilon \{(x,y,z)x<0\}`$ such that
(10)
$$b_\stackrel{}{u}(K^\epsilon K_{}^\epsilon )=2n$$
for any $`\stackrel{}{u}V`$. For any $`\stackrel{}{u}VN`$, we obtain the following contradiction from (9) and (10):
$$2n=b_\stackrel{}{u}(K^\epsilon )=b_\stackrel{}{u}(K^\epsilon K_+^\epsilon )+b_\stackrel{}{u}(K^\epsilon K_{}^\epsilon )=3n.$$
###### Proposition 1.
There exist positive numbers $`\epsilon _0`$ and $`\delta _0`$ such that the following conditions hold for any $`\epsilon (0,\epsilon _0]`$.
1. $`s(K^\epsilon )=2n`$,
2. $`N^\epsilon \{(x,y,z)x<0\}N\{(x,y,z)x<0\}`$,
3. $`b_\stackrel{}{v}(K^\epsilon )=n`$, for any unit vector $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$ with $`|v_3|<\delta _0`$.
Proof: It remains to prove the part (c). As Kuiper did to prove part (a), we investigate the number of real roots of the function $`t(d/dt)K^\epsilon (t)\stackrel{}{v}`$ for a unit vector $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$. Approximating $`\lambda _1(t)`$ and $`\lambda _2(t)`$ by finite linear combinations of powers of $`\mathrm{sin}t`$ and $`\mathrm{cos}t`$, we get a curve $`\stackrel{~}{K}^\epsilon `$ which is $`C^1`$-close to $`K^\epsilon `$. We then substitute
$$\mathrm{cos}t=\frac{2w}{1+w^2},\mathrm{sin}t=\frac{1w^2}{1+w^2}$$
to have
$$\frac{d}{dt}\stackrel{~}{K}^\epsilon (t)\stackrel{}{v}=\frac{A^{2n}(w)}{(1+w^2)^n}+v_3\frac{B^{4n}(w)}{(1+w^2)^{2n}}+\epsilon \frac{C^{2N}(w)}{(1+w^2)^N}$$
where $`A^{2n}`$, $`B^{4n}`$ and $`C^{2N}`$ are polynomials of degree $`2n`$, $`4n`$ and $`2N`$, respectively, for some possibly large $`N`$. The real roots of this function are the same as those of the polynomial
$$P(w)=A^{2n}(w)(1+w^2)^{Nn}+v_3B^{4n}(w)(1+w^2)^{N2n}+\epsilon C^{2N}(w).$$
Since $`A^{2n}(w)=nv_1\mathrm{sin}nt+nv_2\mathrm{cos}nt=n(v_1^2+v_2^2)^{1/2}\mathrm{sin}(nt\alpha ),`$ it has $`2n`$ real roots. If $`\epsilon =v_3=0`$, they are the real roots of $`P(w)=A^{2n}(w)(1+w^2)^{Nn},`$ each of which is at least one unit away from the remaining roots $`\pm \sqrt{1}`$ of multiplicity $`Nn`$. Since the roots of $`P(w)`$ depend continuously on $`\epsilon `$ and $`v_3`$, $`P(w)`$ has exactly $`2n`$ real roots, when $`\epsilon `$ and $`v_3`$ are sufficiently small. One half of them correspond to the local maxima of the projection $`\stackrel{~}{K}^\epsilon \stackrel{}{v}`$ and the other half to local minima. Since $`K^\epsilon `$ is $`C^1`$-close to $`\stackrel{~}{K}^\epsilon `$, part (c) is proved.∎
## 4. Deformations of knots
In this section, we describe two kinds of deformations which do not increase the superbridge number. One is a local deformation and the other is a global one.
###### Lemma 3.
Given a knot $`K`$, let $`\overline{K}`$ be a knot obtained by replacing a subarc of $`K`$ with a straight line segment joining the end points of the subarc. Then $`s(K)s(\overline{K})`$.
Proof: Given a unit vector $`\stackrel{}{v}`$, let $`g:(1,2)\stackrel{}{v}`$ be a parametrization of the orthogonal projection of an open neighborhood of the subarc into $`\stackrel{}{v}`$, where the subarc corresponds to the closed interval $`[0,1]`$. Then the projection of a neighborhood of the straight line segment in $`\overline{K}`$ can be parametrized by
$$\overline{g}(t)=\{\begin{array}{cc}(1t)g(0)+tg(1)\hfill & \text{if }t[0,1]\hfill \\ g(t)\hfill & \text{if }t(1,0][1,2).\hfill \end{array}$$
Since $`\overline{g}`$ has no more local maxima than $`g`$, we have $`b_\stackrel{}{v}(K)b_\stackrel{}{v}(\overline{K})`$ for any $`\stackrel{}{v}`$. Therefore $`s(K)s(\overline{K})`$.∎
For a unit vector $`\stackrel{}{v}`$ and a non-singular linear transformation$`\varphi :^3^3`$, let $`\stackrel{}{v}^\varphi `$ denote the unit vector contained in the one-dimensional subspace $`(\varphi (\stackrel{}{v}^{}))^{}`$ satisfying $`\varphi (\stackrel{}{v})\stackrel{}{v}^\varphi >0`$. For any subset $`AS^2`$, we define
$$A^\varphi =\{\stackrel{}{v}^\varphi \stackrel{}{v}A\}.$$
###### Lemma 4.
Given a unit vector $`\stackrel{}{v}^3`$ and a nonsingular linear transformation $`\varphi `$ of $`^3`$, the formulas
$`b_{\stackrel{}{v}^\varphi }(\varphi (K))=b_\stackrel{}{v}(K)`$
$`b_{\stackrel{}{v}^\varphi }(\varphi (K)\varphi (S))=b_\stackrel{}{v}(KS)`$
hold for any knot $`K`$ and any open subarc $`S`$ of $`K`$.
Proof: At each local maximum point $`P`$ of the projection $`S\stackrel{}{v}`$, there is an open disk $`d_P`$ perpendicular to $`\stackrel{}{v}`$ and tangent to $`S`$ at $`P`$. Then $`\varphi (d_P)`$ is tangent to $`\varphi (S)`$ at $`\varphi (P)`$ and is perpendicular to $`\stackrel{}{v}^\varphi `$. By the definition of $`\stackrel{}{v}^\varphi `$, $`\varphi (P)`$ is a local maximum point of the projection $`\varphi (S)\stackrel{}{v}^\varphi `$ and hence $`b_\stackrel{}{v}(KS)b_{\stackrel{}{v}^\varphi }(\varphi (K)\varphi (S))`$. Since $`(\stackrel{}{v}^\varphi )^{\varphi ^1}=\stackrel{}{v}^{(\varphi ^1\varphi )}=\stackrel{}{v}`$, we also get
$$b_\stackrel{}{v}(KS)=b_{(\stackrel{}{v}^\varphi )^{\varphi ^1}}(\varphi ^1(\varphi (K))\varphi ^1(\varphi (S)))b_{\stackrel{}{v}^\varphi }(\varphi (K)\varphi (S)).$$
This proves the second formula. Setting $`S=K`$, the first formula is obtained.∎
The next proposition easily follows from Lemma 4.
###### Proposition 2.
Given a knot $`K`$ and a nonsingular linear transformation $`\varphi `$ of $`^3`$, we have $`s(\varphi (K))=s(K)`$. In particular, if a knot $`K`$ and a unit vector $`\stackrel{}{v}`$ satisfy $`b_\stackrel{}{v}(K)=s(K)=s[K]`$, then $`b_{\stackrel{}{v}^\varphi }(\varphi (K))=s(\varphi (K))=s[K]`$.
## 5. Proofs
For any $`\lambda `$ with $`0<\lambda 1`$, let $`\varphi _\lambda `$, $`\psi _\lambda `$, $`\psi `$ be the autohomeomorphisms of $`^3`$ defined by
$`\varphi _\lambda (x,y,z)`$ $`=`$ $`(x,y,\lambda z)`$
$`\psi _\lambda (x,y,z)`$ $`=`$ $`(1+\lambda \lambda z,y,1+\lambda x)`$
$`\psi (x,y,z)`$ $`=`$ $`(z,y,x).`$
The map $`\psi `$ is the $`180^{}`$ rotations about the line $`\{(x,0,z)x+z=0\}`$ and the map $`\psi _\lambda `$ is the composite map $`\varphi _\lambda `$ followed by the $`180^{}`$ rotations about the line $`\{(x,0,z)x+z=1+\lambda \}`$.
For any locally one-to-one closed parametrized path $`\gamma :S^1^3`$, we extend the definition of the crookedness $`b_\stackrel{}{v}(\gamma )`$ by considering the parametrized projection $`t\gamma (t)\stackrel{}{v}:S^1\stackrel{}{v}`$ instead of the projection $`\gamma (S^1)\stackrel{}{v}`$. In this way we can consider the crookedness for finite-fold coverings of knots and singular knots.
### Proof of Theorem 1
Throughout this proof, $`\lambda `$ is a constant satisfying $`0<\lambda 1/4`$, $`\stackrel{}{v}=v_1𝐢+v_2𝐣+v_3𝐤`$ is a unit vector, and $`i=1`$ or $`2`$. We may assume that the knot $`K_i`$ is parametrized by
$$K_i(t)=((1+\lambda _{i1}(t))\mathrm{cos}n_it,(1+\lambda _{i1}(t))\mathrm{sin}n_it,\lambda _{i2}(t)+\mathrm{cos}^2n_it)$$
where $`\lambda _{i1}`$ and $`\lambda _{i2}`$ are smooth periodic functions with period $`2\pi `$ satisfying the conditions corresponding to (5), (6) and (7), for $`i=1,2`$. For any $`\epsilon `$ with $`0\epsilon 1`$, we define $`K_1^\epsilon `$ and $`K_2^\epsilon `$ as in (8). Then $`K_i^\epsilon `$ is a knot isotopic to $`K_i`$ and is the closure of the $`n_i`$-braid $`K_i^\epsilon \{(x,y,z)xyx\}`$ when $`0<\epsilon 1`$. When $`\epsilon =0`$, $`K_i^\epsilon `$ is an $`n_i`$-fold covering of $`\eta `$. Since the two knots $`\varphi _\lambda (K_1^\epsilon )`$ and $`\psi _\lambda (K_2^\epsilon )`$ are tangent at the point $`(1,0,\lambda )`$, their union $`K_\lambda `$ can be regarded as a singular knot parametrized by
$$K_\lambda (t)=\{\begin{array}{cc}\varphi _\lambda (K_1^\epsilon (2t))\hfill & \text{if }\pi t0\hfill \\ \psi _\lambda (K_2^\epsilon (2t))\hfill & \text{if }0t\pi .\hfill \end{array}$$
Then $`\overline{K}_\lambda =(K_\lambda \varphi _\lambda (\eta _+)\psi _\lambda (\eta _+))S_+S_{}`$ is a singular knot with only one singular point at $`((1+\lambda )/2,0,(1+\lambda )/2)`$ where
$$S_\pm =\{(0,1,0)+s(1+\lambda ,\pm 2,1+\lambda )0<s<1\}.$$
By Lemma 3, $`b_\stackrel{}{v}(\overline{K}_\lambda )b_\stackrel{}{v}(K_\lambda )`$. Since $`(1,0,\lambda )`$ is a local maximum point of the parametrized projection $`tK_\lambda (t)\stackrel{}{v}`$ only if both of the projections $`\varphi _\lambda (K_1^\epsilon )\stackrel{}{v}`$ and $`\psi _\lambda (K_2^\epsilon )\stackrel{}{v}`$ have local maximum at $`(1,0,\lambda )`$, we have
(11)
$$b_\stackrel{}{v}(\overline{K}_\lambda )b_\stackrel{}{v}(\varphi _\lambda (K_1^\epsilon ))+b_\stackrel{}{v}(\psi _\lambda (K_2^\epsilon )).$$
The vectors $`\stackrel{}{w}_\pm =\pm (1+\lambda )(𝐢+𝐤)+2𝐣`$, are parallel to the segments $`S_\pm `$, respectively. A computation shows that
$`\stackrel{}{w}_+\stackrel{}{v}=(1+\lambda )(v_1+v_3)+2v_2(10\sqrt{89})/\sqrt{80}>0,`$
$`\stackrel{}{w}_{}\stackrel{}{v}=(1+\lambda )(v_1+v_3)+2v_2(10\sqrt{89})/\sqrt{80}<0,`$
whenever $`v_3(4\lambda ^2+1)^{1/2}`$. Therefore there exists a number $`\delta (1/\sqrt{2},(4\lambda ^2+1)^{1/2})`$ such that $`\stackrel{}{w}_{}\stackrel{}{v}<0<\stackrel{}{w}_+\stackrel{}{v}\text{ whenever }v_3\delta `$. At the endpoints $`(1+\lambda ,\pm 1,1+\lambda )`$ of $`S_\pm `$, we have
$`\underset{t\frac{4n_21}{4n_2}\pi ^{}}{lim}{\displaystyle \frac{d}{dt}}K_\lambda (t)\stackrel{}{v}=2n_2v_3<0<2n_2v_3=\underset{t\frac{\pi }{4n_2}^+}{lim}{\displaystyle \frac{d}{dt}}K_\lambda (t)\stackrel{}{v}`$
if $`v_3>0`$. Therefore there exist open arcs $`\stackrel{~}{S}_\pm `$ of $`\overline{K}_\lambda `$, containing the closures of $`S_\pm `$, respectively, satisfying $`b_\stackrel{}{v}(\overline{K}_\lambda \stackrel{~}{S}_+\stackrel{~}{S}_{})=0`$ whenever $`v_3\delta `$. Similarly we also have $`b_\stackrel{}{v}(\overline{K}_\lambda \stackrel{~}{S}_+\stackrel{~}{S}_{})=0`$ whenever $`v_1\delta `$.
By Lemma 2, Lemma 4, and the last two conditions, we have<sup>5</sup><sup>5</sup>5$`\overline{\eta }_+`$ is the closure of $`\eta _+`$.
(12) $`b_\stackrel{}{v}(\overline{K}_\lambda )`$ $``$ $`b_\stackrel{}{v}(\varphi _\lambda (K_1^\epsilon )\varphi _\lambda (K_1^\epsilon \overline{\eta }_+))+b_\stackrel{}{v}(\overline{K}_\lambda \stackrel{~}{S}_+\stackrel{~}{S}_{})`$
$`+b_\stackrel{}{v}(\psi _\lambda (K_2^\epsilon )\psi _\lambda (K_2^\epsilon \overline{\eta }_+))`$
$``$ $`b_\stackrel{}{v}(\varphi _\lambda (K_1^\epsilon ))+b_\stackrel{}{v}(\psi _\lambda (K_2^\epsilon ))1`$
whenever $`\stackrel{}{v}N^{\varphi _\lambda }Q_\delta \psi (N^{\varphi _\lambda }Q_\delta )`$ where $`Q_\delta =\{(x,y,z)S^2x>0,z>\delta \}`$. By Proposition 1 (a)–(b), we may assume that
(13) $`s(K_i^\epsilon )=2n_i`$
(14) $`(N_i^\epsilon )^{\varphi _\lambda }N^{\varphi _\lambda }Q_\delta `$
where $`N_i^\epsilon =\{\stackrel{}{v}S^2v_3>0,b_\stackrel{}{v}(K_i^\epsilon )=2n_i\}.`$ Since
$`\stackrel{}{v}^{\varphi _\lambda }𝐤=v_3(\lambda ^2(1v_3^2)+v_3^2)^{1/2},`$
$`\stackrel{}{v}^{\psi _\lambda }𝐢=v_1(\lambda ^2(1v_1^2)+v_1^2)^{1/2},`$
Proposition 1 (c) implies that
(15)
$$\begin{array}{c}b_\stackrel{}{v}(\varphi _\lambda (K_1^\epsilon ))=n_1\text{ whenever }|v_3|1/\sqrt{2}\hfill \\ b_\stackrel{}{v}(\psi _\lambda (K_2^\epsilon ))=n_2\text{ whenever }|v_1|1/\sqrt{2}\hfill \end{array}$$
provided $`\lambda `$ is sufficiently small. By (11) and (15), we get
(16)
$$b_\stackrel{}{v}(\overline{K}_\lambda )n_1+n_2\text{ if }\mathrm{max}\{|v_1|,|v_3|\}1/\sqrt{2}.$$
By (11), (13) and (15), we get
(17)
$$b_\stackrel{}{v}(\overline{K}_\lambda )\{\begin{array}{cc}2n_1+n_21\hfill & \text{if }\pm \stackrel{}{v}(N_1^\epsilon )^{\varphi _\lambda },|v_3|>1/\sqrt{2}\hfill \\ n_1+2n_21\hfill & \text{if }\pm \stackrel{}{v}\psi ((N_2^\epsilon )^{\varphi _\lambda }),|v_1|>1/\sqrt{2}\hfill \end{array}$$
By (12), (13), (14) and (15), we get
(18)
$$b_\stackrel{}{v}(\overline{K}_\lambda )\{\begin{array}{cc}2n_1+n_21\hfill & \text{if }\pm \stackrel{}{v}(N_1^\epsilon )^{\varphi _\lambda }Q_\delta \hfill \\ n_1+2n_21\hfill & \text{if }\pm \stackrel{}{v}\psi ((N_2^\epsilon )^{\varphi _\lambda }Q_\delta )\hfill \end{array}$$
For the last two formulas, we used the fact $`b_\stackrel{}{v}(\overline{K}_\lambda )=b_\stackrel{}{v}(\overline{K}_\lambda )`$. For a very small positive number $`ϵ`$, let $`\overline{S}_+=S_+\{(\mathrm{cos}t,\mathrm{sin}t,\lambda \mathrm{cos}^2t)\pi /2ϵt\pi /2\}`$ and let $`\stackrel{ˇ}{S}_+`$ be the line segment joining the endpoints of $`\overline{S}_+`$. By the conditions (5), (6) and (7), the knot $`\stackrel{ˇ}{K}_\lambda =(\overline{K}_\lambda \overline{S}_+)\stackrel{ˇ}{S}_+`$ is a knot representing $`K_1\mathrm{}K_2`$. By Lemma 3, (16), (17) and (18), we have
$$b_\stackrel{}{v}(\stackrel{ˇ}{K}_\lambda )b_\stackrel{}{v}(\overline{K}_\lambda )\mathrm{max}\{2n_1+n_2,n_1+2n_2\}1.\mathit{}$$
### Proof of Theorem 2
Let $`K_i`$ be a torus knot of type $`(p_i,q_i)`$ where $`p_i`$ and $`q_i`$ are coprime integers satisfying $`2p_i<q_i`$, for $`i=1,2`$. This proof breaks into three cases.
Case 1. Suppose that the inequality $`2p_i<q_i/2`$ holds for $`i=1,2`$. In this case, we have $`\beta [K_i]=b[K_i]=s[K_i]/2=p_i`$. Therefore a direct application of Theorem 1 shows that $`s[K_1\mathrm{}K_2]\mathrm{max}\{2p_1+p_2,p_1+2p_2\}1`$.
Case 2. Suppose that the inequality $`2p_i<q_i<2p_i`$ holds for $`i=1,2`$. In this case, $`\beta [K_i]=b[K_i]=p_i`$ and $`s[K_i]=q_i`$. As shown in , $`K_i`$ can be represented by a polygonal knot $`\tau _i=\tau _i(\alpha _i)`$ of $`2q_i`$ edges embedded on the torus $`H_{\alpha _i}H_{\beta _i}`$ where
$$H_\theta =\{(x,y,z)x^2+y^2z^2\mathrm{sin}^2\frac{\theta }{2}=\mathrm{cos}^2\frac{\theta }{2},|z|1\},$$
$`\pi p_i/q_i<\alpha _i<2\pi p_i/q_i`$ and $`\alpha _i+\beta _i=\pi `$. The knot $`K_i`$ has $`2q_i`$ vertices; $`q_i`$ on each of the two unit circles $`\{(x,y,\pm 1)x^2+y^2=1\}`$. By (1), we know that $`s(K_i)=q_i`$. We may assume that $`K_i`$ has a vertex at $`(1,0,1)`$. We define
$`N_i`$ $`=`$ $`\{\stackrel{}{v}S^2\stackrel{}{v}𝐤>0,b_\stackrel{}{v}(K_i)=q_i\},`$
$`M_i`$ $`=`$ $`\{\stackrel{}{v}S^2\text{The projection }K_i\stackrel{}{v}\text{ has a local minimum at }(1,0,1)\}.`$
For any $`\stackrel{}{v}N_i`$, the $`q_i`$ vertices of $`K_i`$ on the circle $`\{(x,y,1)x^2+y^2=1\}`$ are local maximum points of the projection $`K_i\stackrel{}{v}`$. Let $`tK_i(t)`$ parametrize $`K_i`$ modulo $`2\pi `$ with $`K_i(0)=(1,0,1)`$, as a closed $`p_i`$-braid around the $`z`$-axis. The singular knot $`K_\lambda `$ given by the parametrization
$$K_\lambda (t)=\{\begin{array}{cc}\varphi _\lambda (K_1(2t))\hfill & \text{if }\pi t0\hfill \\ \psi _\lambda (K_2(2t))\hfill & \text{if }0t\pi \hfill \end{array}$$
has only one singular point at $`(1,0,\lambda )`$. Straightening an arc near the singular point, we get a knot representing $`K_1\mathrm{}K_2`$ whose crookedness is not bigger than that of $`K_\lambda `$ in any direction. As $`\lambda `$ approaches zero, $`N_1`$ shrinks to the north pole $`(0,0,1)`$ whereas $`M_1`$ approaches a region of positive area containing the point $`(1,0,0)`$. Therefore, for a sufficiently small $`\lambda `$, we have
(19)
$$(N_1)^{\varphi _\lambda }\psi ((M_2)^{\varphi _\lambda }),\text{ and }\psi ((N_2)^{\varphi _\lambda })(M_1)^{\varphi _\lambda },$$
and as in (15), we also have
(20)
$$\begin{array}{c}b_\stackrel{}{v}(\varphi _\lambda (K_1))=p_1\text{ whenever }|\stackrel{}{v}𝐤|1/\sqrt{2},\hfill \\ b_\stackrel{}{v}(\psi _\lambda (K_2))=p_2\text{ whenever }|\stackrel{}{v}𝐢|1/\sqrt{2}.\hfill \end{array}$$
By (19), if $`\pm \stackrel{}{v}(N_1)^{\varphi _\lambda }\psi ((N_2)^{\varphi _\lambda })`$, the point $`(1,0,\lambda )`$ in not a local maximum point of $`K_\lambda `$. Therefore we have
$$b_\stackrel{}{v}(K_\lambda )=\{\begin{array}{cc}q_1+p_21\hfill & \text{if }\pm \stackrel{}{v}(N_1)^{\varphi _\lambda }\hfill \\ p_1+q_21\hfill & \text{if }\pm \stackrel{}{v}\psi ((N_2)^{\varphi _\lambda }).\hfill \end{array}$$
By (19) and (20), we obtain
$$b_\stackrel{}{v}(K_\lambda )b_\stackrel{}{v}(\varphi _\lambda (K_1))+b_\stackrel{}{v}(\psi _\lambda (K_2))<\mathrm{max}\{q_1+p_2,p_1+q_2\},$$
if $`\pm \stackrel{}{v}(N_1)^{\varphi _\lambda }\psi ((N_2)^{\varphi _\lambda })`$. Therefore $`b_\stackrel{}{v}(K_\lambda )\mathrm{max}\{q_1+p_2,p_1+q_2\}1`$, for any unit vector $`\stackrel{}{v}`$.
Case 3. Suppose that the inequalities $`2p_1<q_1/2`$ and $`2p_2<q_2<2p_2`$ hold. Let $`K_1(=K_1^\epsilon )`$ and $`K_2`$ be embedded and parametrized as in the Proof of Theorem 1 and in Case 2, respectively. We consider the singular knot parametrized by
$$K_\lambda (t)=\{\begin{array}{cc}\varphi _\lambda (K_1^\epsilon (2t))\hfill & \text{if }\pi t0\hfill \\ \psi _\lambda (K_2(2t))\hfill & \text{if }0t\pi .\hfill \end{array}$$
We replace the arc $`\varphi _\lambda (\eta _+)`$ of $`\varphi _\lambda (K_1^\epsilon )`$ by the broken line joining the three points $`(0,1,0)`$, $`(1,0,\lambda )`$ and $`(0,1,0)`$, consecutively, to get a new singular knot $`\overline{K}_\lambda `$. The remaining argument will be very similar to that of Case 2.∎
Acknowledgement: This work was done while the author was visiting the University of British Columbia during the academic year 1998–99. He is grateful to the members of the Department of Mathematics of UBC, especially Dale Rolfsen with whom he had many helpful discussions. |
warning/0001/hep-ex0001008.html | ar5iv | text | # Highlights in Neutrino and Astroparticle Physics
## 1 INTRODUCTION
This was a timely workshop, because of the many new important experimental results on neutrino physics and astrophysics, and the renewed interest in astroparticle physics, both experimentally and theoretically.
The experimental evidences on neutrino oscillations are mounting, and many new experiments are being planned to definitely prove these indications and accurately measure the neutrino oscillation parameters.
The astrophysical $`\gamma `$-ray bursts seem to be a dominant phenomenon in our universe. New, larger and more sophisticated detectors are planned to study the higher energy $`\gamma `$-rays.
Neutrino astrophysics started in the 1960’s with the first detection of electron neutrinos from the sun, and was in a sense reborn in 1987 with the detection of electron antineutrinos from Supernova SN87A. Many detectors are now ready to study these two phenomena. High energy muon neutrino detectors of large volumes are entering the scene, possibly opening the field of high energy muon neutrino astronomy.
Many search experiments are trying to detect possible components of the Dark Matter (DM), or search for new particles predicted by the Standard Model (SM) of particle physics, and by theories which go beyond the SM.
The study of the highest energy cosmic rays is another field of interest, in particular for determinig the mechanisms responsible for their acceleration. Very large area detectors, above and below ground, are needed to study this field.
A Large effort is being made to develop instrumentation for the detection of gravitational waves, which should reveal some of the most violent phenomena occurring in the cosmos. The detectors include supercooled antennas at 0.1 K and very long interferometers (few km).
In this summary of the workshop I shall recall many of the papers presented, with special emphasis on neutrino oscillations and on higher energy phenomena; I shall not be able to cover in detail all subjects, nor quote all results. I apologize for this impossibility and for possible omissions.
During the workshop we were informed of the sudden death of Bianca Monteleoni Conforto, a colleague and a collaborator. I dedicate these notes to her memory.
## 2 NEUTRINO PHYSICS
Most of the interest in this field concentrated on experimental results relevant to neutrino oscillations.
### 2.1 Atmospheric neutrinos
The interest in atmospheric neutrinos has grown in the last year, after the Neutrino ’98 Conference in Takayama, Japan. New, higher statistics data have been presented by the Soudan 2 , MACRO , and SuperKamiokande (SK) collaborations. The measured flux of muons induced by atmospheric $`\nu _\mu `$ shows a reduction with respect to the expectation; the reduction depends on the neutrino energy and direction. For $`\nu _e`$ induced electrons there is no deviation from the prediction. The three experiments can explain the $`\nu _\mu `$ reduction in terms of $`\nu _\mu \nu _\tau `$ neutrino oscillations, with maximum mixing and $`\mathrm{\Delta }m^2`$ values of few times $`10^3eV^2`$.
In the simplest scenario of two flavor oscillations, the survival probability of a pure $`\nu _\mu `$ beam is
$$P(\nu _\mu \nu _\mu )=1sin^22\theta sin^2(\frac{1.27\mathrm{\Delta }m^2L}{E_\nu })$$
(1)
where $`\mathrm{\Delta }m^2=m_2^2m_1^2`$ is the mass difference of the two neutrino mass states, $`\theta `$ is the mixing angle, $`E_\nu `$ is the neutrino energy and $`L`$ is the path length from the $`\nu _\mu `$ production point to the detector. For atmospheric neutrinos $`L`$ can be estimated through the neutrino arrival direction $`\mathrm{\Theta }`$. For upgoing neutrinos, as the zenith angle $`\mathrm{\Theta }`$ changes, one has $`L2R_{}cos\mathrm{\Theta }`$ ($`R_{}`$ is the Earth radius), while $`L`$ is only few tens of kilometers for vertical downgoing neutrinos.
Atmospheric neutrinos are detected in the SuperKamiokande (SK) water Ĉerenkov detector via their interactions with p and $`{}_{}{}^{16}O`$ nuclei in the $`22500m^3`$ water fiducial volume ($`50.000m^3`$ total volume). Three different classes of events are defined (with increasing average energy of the parent neutrino): fully contained events (FC), partially contained events (PC) and upward-going muons. FC events are further subdivided into sub-GeV and multi-GeV. Electrons are identified in the FC sample. The zenith angle distribution for e-like sub-GeV and multi-GeV events are in reasonable agreement with the predictions assuming no-oscillations. Instead the $`\mu `$-like events deviate considerably from the prediction, see Fig. 1. The ratio of the measured numbers of muons to electrons normalized to the respective Monte Carlo predictions is affected by a smaller systematic error, and it enhances the anomaly .
Assuming two flavor oscillations Fig.2 shows the $`90\%`$ C.L. contours delimiting the accepted regions by Kamiokande and SK. The SK data favour $`\nu _\mu \nu _\tau `$ neutrino oscillations with $`\mathrm{\Delta }m^2=(1.56)10^3eV^2`$ and $`sin^22\theta >0.9`$. For more details on further data on upthroughgoing muons and stopping muons, see the paper presented by Kajita at this workshop , and . SK obtains also indications for an east-west asymmetry, which may be considered a confirmation of the flux calculations and of the experimental methods .
The Soudan 2 results support the oscillation hypothesis by measuring atmospheric $`\nu _\mu `$ and $`\nu _e`$ interactions at low energies, below $`1`$ GeV . A different detection technique (drift chamber calorimeter) is used in this case; the total mass of the detector is about $`1kt`$ and the total exposure is $`4.6kty`$. Fig. 3 shows, for the high-resolution contained data, vs Log (L/E<sub>ν</sub>) relative to the no-oscillation expectations. Still with low statistical significance, the data agree with a reduction of $`\nu _\mu `$ events compared to expectations, while the $`\nu _e`$ events agree with expectations.
The MACRO detector is a box of $`76.6m\times 12m\times 9.3m`$ located at the Gran Sasso Lab.; the detection elements are planes of limited streamer tubes for tracking and liquid scintillation counters for timing. The lower half of the detector is filled with trays of absorbers alternating with streamer tube planes, while the upper part is open. The angular resolution for muons achieved by the streamer tubes is $`<1^{}`$. The time resolution of the scintillators is $`0.5ns`$. Fig. 4 shows the measured topologies.
The up throughgoing muons come from $`\nu _\mu `$ with $`\overline{E}_{\nu _\mu }80GeV`$ interacting in the rock below the detector; their flight direction is determined by time-of-flight (t.o.f.). $`\nu _\mu `$ with $`\overline{E}_{\nu _\mu }4GeV`$ interact inside the lower apparatus; yielding upgoing muons (IU). The partially contained downgoing events (ID) and upward going stopping muons (UGS) are identified via topological constraints.
Monte Carlo simulations play a crucial role in atmospheric neutrino studies. Macro used the Bartol neutrino flux and the deep inelastic scattering (DIS) parton distribution of ref. for the neutrino cross-sections. The propagation of muons is done using the energy loss calculations of ref. in standard rock. The systematic uncertainty on the expected flux of muons is $`\pm 17\%`$; this is a scale factor, that changes little the shape of the angular distribution.
The ratio of the observed to expected number of upthroughgoing muons is $`0.74\pm 0.031_{stat}\pm 0.044_{sys}\pm 0.12_{theo}`$. Fig. 5 shows their zenith angle distribution compared to Monte Carlo expectation without neutrino oscillations (solid line); the dashed line is the fit to the data assuming $`\nu _\mu \nu _\tau `$ oscillations. The reduction in the detected number of events and the deformation of the zenith angle distribution may be due to $`\nu _\mu `$ disappearance: fewer events are expected near the vertical $`(cos\mathrm{\Theta }=1)`$ than near the horizontal $`(cos\mathrm{\Theta }=0)`$, due to the longer path length of neutrinos from production to observation. The maximum of the $`\chi ^2`$ probability corresponds to $`\mathrm{\Delta }m^2=2.5\times 10^3eV^2`$ and maximum mixing. The confidence region at the 90% C.L. in $`(\mathrm{sin}^22\theta ,\mathrm{\Delta }m^2)`$ for $`\nu _\mu \nu _\tau `$ oscillations agrees and is somewhat larger than that of SK . Notice the possible excess of events at $`cos\mathrm{\Theta }0.65`$ (also the up-throughgoing muons of SK have a similar hint); it is consistent with a statistical fluctuation, but it could be a hint for a more complex scenario.
Fig. 6 shows the zenith angle distribution of the semicontained (IU) and upstopping muons plus partially contained downgoing muons (UGS + ID). The data are within errors consistent with a constant deficit with respect to the MC expectations. The ratios of the number of observed to expected events are $`R_{ID+UGS}=(\frac{Data}{MC})_{ID+UGS}0.71`$ and $`R_{IU}0.57`$. The theoretical and systematic errors are largely reduced (from 25% to about 5%) if the ratio of ratios is considered, $`=R_{IU}/R_{ID+UGS}=0.80\pm 0.09_{stat}`$; for no oscillations one expects $`R=1`$. The reductions are consistent with $`\nu _\mu \nu _\tau `$ oscillations with maximum mixing and $`\mathrm{\Delta }m^210^3÷10^2eV^2`$.
Several theoretical papers tried to interpret the atmospheric neutrino data in terms of oscillations among three neutrino types; the differences with the simpler $`\nu _\mu \nu _\tau `$ possibility are small . Other authors considered $`\nu _\mu \nu _{sterile}`$, which is slightly disfavoured by the data. Others include $`\nu _e,\nu _\mu ,\nu _\tau ,\nu _{sterile}`$.
G. Battistoni discussed the effects of the approximations used in the Monte Carlo predictions. Present MCs use the collinear approach, which cannot be a good approximation at low energies. But 3-D effects are smeared out because of Fermi motion of the nucleons in nuclei. Uncertainties remain in the knowledge of primary cosmic ray spectra, secondary particle production and neutrino cross sections. Hopefully measurements of muons in the atmosphere could improve the predictions, though the sub-GeV range remains problematic.
Improved atmospheric neutrino detectors are under discussion .
### 2.2 Solar neutrinos
Solar $`\nu _e`$ come from a chain of nuclear reactions and decays in the centre of the sun. The three important components of the spectrum are: i) the energetic neutrinos from $`{}_{}{}^{8}B`$ decay; ii) the monoenergetic neutrinos from <sup>7</sup>Be+$`e^{}`$ <sup>7</sup>Li+$`\nu _e`$ $`(\text{E}_{\nu _e}=0.862`$ MeV) and iii) the low–energy part, the $`pp`$ neutrinos $`(\text{E}_{\nu _e}0.41`$ MeV) (most abundant) .
Experimental measurements of solar neutrinos have been performed by five experiments using three different reactions .
The first measurement of solar neutrinos used a radiochemical method via inverse $`\beta `$ decay, $`\nu _e+^{37}`$Cl$`^{37}`$Ar$`+e^{}`$ , which has a neutrino energy threshold $`\text{ E}_{\nu th}=814`$ keV. The experiment, sensitive to $`{}_{}{}^{7}Be`$ and $`{}_{}{}^{8}B`$ neutrinos yields a flux smaller than that predicted by the Standard Solar Model (SSM): $`(\varphi _{7_{Be}}+\varphi _{8_B})(\text{Cl})<(\varphi _{7_{Be}}+\varphi _{8_B})(\text{SSM})`$.
The second measurement was performed in the Kamiokande water Ĉerenkov detector using the reaction $`\nu _e+e^{}\nu _e+e^{}`$. They apply a cut at E$`{}_{\nu th}{}^{}7MeV`$ and are thus sensitive only to $`{}_{}{}^{8}B`$ neutrinos. The angular distribution is peaked in the direction of the sun and therefore confirms that the detected neutrinos come from the sun. They obtained a ratio $`expected/measured=0.417\pm 0.069`$. Combining this with the chlorine result, one has a discrepancy expressed as $`(\varphi _{7_{Be}}+\varphi _{8_B})(\text{Cl})<\varphi _{8_B}(\text{Ka})`$.
A third reaction is studied by radiochemical methods using $`{}_{}{}^{71}Ga`$ in metallic (SAGE) and in a hydrochloric water solution (GALLEX, GNO): $`\nu _e+^{71}`$Ga$`^{71}`$Ge$`+e^{}`$ , which has a threshold at E$`_{\nu _e}`$=233 keV. Thus one may measure the neutrinos coming from the $`pp`$ reactions, proving that the sun is a $`pp`$ nuclear fusion plant. The experiments yield values smaller than the SSM prediction; these low values and the comparison with the preceeding measurements, lead to $`\varphi _{7_{Be}}(meas)<\varphi _{7_{Be}}(\text{SSM})`$, which seems to be the main problem.
Superkamiokande presented at this workshop new results on solar neutrinos which further confirm the above statements. The data also confirm that the solar $`\nu _e`$ come from the sun , see Fig. 7. A day-night effect might have been observed by SK: this would be expected for the MSW effect. Also a seasonal variation due to the eccentricity of the earth orbit might have been observed.
The lack of observed Be solar neutrinos seems at present to be the essence of the solar neutrino problem. This could be due to a faulty experiment. Assuming that the experiments are correct, physicists looked for an astrophysical solution and at neutrino oscillations. Improvements have been made in the knowledge of the sun interior and it seems that one cannot explain the deficits. One is therefore left with the possibility of neutrino oscillations.
Possible solutions of the problem assuming neutrino oscillations in vacuum and possible solutions assuming neutrino oscillations in solar matter (the MSW effect) are indicated in the compilation of Fig. 8; more up to date compilations have been presented in -; the MSW low-mixing solution seems to be disfavoured.
Most solar neutrino experiments are relatively low–rate experiments. Superkamiokande and future experiments have higher rates and have more specific aims. The SNO detector in Sudbury, Canada is starting to take data with neutrino interactions in $`D_2O`$. At Gran Sasso the Borexino experiment plans to detect $`{}_{}{}^{7}Be`$ neutrinos via the reaction $`\nu _ee^{}\nu _ee^{}`$ in liquid scintillators ; ICARUS should detect $`\nu _e`$ interactions in an $`{}_{}{}^{40}Ar`$ TPC . There are discussions about the possible use of Li I (Eu) scintillation counters.
### 2.3 Accelerator and reactor experiments
We have heard numerous reports about short baseline accelerator experiments: LSND, KARMEN, NOMAD, CHORUS and the future MINIBOONE. LSND gave a possible signal for neutrino oscillations for $`\mathrm{\Delta }m^21eV^2`$ and $`10^3\stackrel{<}{_{}}sin^22\theta <`$ few $`10^2`$, see Fig. 8 . The other experiments gave limits, which are globally summarized in Fig. 8. Several technical improvements were made by these experiments; I shall only recall the revival of the emulsion technique for neutrino physics, specifically for $`\nu _\tau `$ detection, because of the exceptional space resolution ($`1\mu m`$) of the technique .
The results from the Chooz and Palo Verde reactor experiments exclude $`\nu _e\nu _\mu `$ oscillations for $`\mathrm{\Delta }m^2>10^3eV^2`$ and $`2\times 10^2<sin^22\theta <1`$, not shown properly in Fig. 8.
### 2.4 Long baseline experiments
Several long baseline experiments have been proposed; they will cover the region $`\mathrm{\Delta }m^2>10^3eV^2`$ and $`sin^22\theta >10^2`$ .
\- K2K: from KEK to SuperKamiokande ($`230km`$) is starting to take data; they also have a near detector .
\- MINOS: from the Fermilab main injector to the Soudan mine ($`730km`$) is under construction; it is a $`6kton`$ calorimeter detector; they will also have a near detector .
\- KAMLAND: $`\overline{\nu }_e`$ from nuclear reactors will be detected in the Kamiokande mine with a liquid scintillator detector .
\- ICARUS $``$ now ICANOE: from CERN-SPS to Gran Sasso ($`730km`$) ; ICARUS will be an ”electronic” bubble chamber; NOE a tracking calorimeter detector.
\- OPERA: from CERN to Gran Sasso ; it is basically a (large) emulsion detector.
### 2.5 Direct measurement of the $`\overline{\nu }_𝐞`$ mass
Tritium decay, $`t^3He+e^{}+\overline{\nu }_e`$ with Q $`=18.6keV,t_{1/2}=12.3y`$, has been used by many groups to obtain increasingly better limits on the $`\overline{\nu }_e`$ mass ($`\overline{\nu }_1`$ mass if there is small mixing). The techniques and the calculations have been constantly improved. The latest results with the improved Mainz set up give a $`95\%`$ C.L. upper limit $`m_{\overline{\nu }e}\stackrel{<}{_{}}2.8eV`$ .
### 2.6 Neutrinoless double beta decay
For even-even nuclei the chain decays
$`(A,Z)(A,Z+1)+e^{}+\overline{\nu }_e,(A,Z+1)(A,Z+2)+e^{}+\overline{\nu }_e`$ (2a) are forbidden by energy conservation; the decay may be possible in a single step:
$`(A,Z)(A,Z+2)+2e^{}+2\overline{\nu }_e,`$ (2b)
$`(A,Z)(A,Z+2)+2e^{}+x,`$ (2c)
$`(A,Z)(A,Z+2)+2e^{}`$ (2d)
The neutrinoless double beta decay, Eq. (2d), is forbidden by lepton number conservation; it would be allowed if $`\nu _e`$ and $`\overline{\nu }_e`$ were identical and if they had a non-zero mass. The energy spectrum for the sum of the energies of the two electrons, $`E=E_1+E_2`$, is different for each of the three cases: a line for (2d), a continuum peaked at low $`E`$ for (2b) and a continuum peaked at higher $`E`$ for (2c).
Most of the direct searches for neutrinoless double beta decays use materials which act both as source and detector, such as $`{}_{}{}^{76}Ge^{76}Se+2e^2`$ . Germanium detectors ranging from 1 to $`7kg`$ have been used. Normal germanium contains $`15\%{}_{}{}^{76}Ge`$. The Heidelberg-Moscow Collaboration uses several kilograms of enriched germanium containing $`85\%{}_{}{}^{76}Ge`$. From an exposure of $`24kgy`$ they quote $`t_{1/2}>6\times 10^{25}y(90\%CL)`$, which in certain models corresponds to $`m_{\nu _e}<0.2eV`$. Some groups use visual detectors, separating the spatial detection of the two electrons. The double beta decay $`{}_{}{}^{136}Xe^{136}Ba+2e^{}`$ has a favorable transition energy of $`2479keV`$.
Considerable work is going on in the development of cryogenic detectors for double beta decays and for dark matter searches . At low temperature the heat capacity is very small, and a small energy deposition implies a relatively large increase in temperature. Four cryogenic crystals of Te $`O_2`$, each about $`340g`$, were used by the Milano-Gran Sasso collaboration to study the double beta decay of $`{}_{}{}^{130}Te`$. Four sapphire detectors, each of $`262g`$, are used by the CRESST experiment in a search for dark matter WIMPs.
## 3 NEUTRINO ASTROPHYSICS
### 3.1 Neutrino Astronomy
One of the main interests in neutrino astronomy is connected with the great penetrating power of neutrinos, which allows us to look directly at their sources. The universe is filled with fossil low-energy neutrinos from the Big Bang. Low–energy neutrinos of $`1`$ MeV come continuously from the interior of stars like the sun; slightly–higher–energy neutrinos $`(12`$ MeV) come in bursts from supernovae explosions. High–energy neutrinos ($`>1`$ GeV), may come from non–thermal point sources. Neutrinos of $`>1`$ GeV may also come from the sun and the earth, where annihilations of WIMPs could take place.
#### 3.1.1 Neutrinos from stellar gravitational collapses
Massive stars, $`m>6m_{}`$, evolve as increasingly heavier nuclei are produced and then burnt at their centres in a chain of thermonuclear processes, ultimately leading to the formation of a core composed of iron and nickel. When the core mass exceeds the Chandrasekar limit, the core implodes in a time slightly longer than the freefall time and leads to the formation of a neutron star. The energy released during a stellar collapse is at least the gravitational binding energy of the residual neutron star, $`E3\times 10^{53}(m/m_{}^2)`$ (10 $`km`$/R) $`ergs`$, $`10^{53}ergs`$ $``$ 0.1 $`m_{}`$, mostly in the form of neutrinos with $`E_\nu 12MeV`$. About $`4\times 10^{57}`$ neutrinos of each species are emitted. Three stages of neutrino emission may be identified.
All types of neutrinos may be detected via neutral current interactions with electrons, $`\nu _ee^{}\nu _ee^{},\overline{\nu }_ee^{}\overline{\nu }_ee^{}`$, etc, with a cross section $`\sigma =1.7\times 10^{44}`$ E<sub>ν</sub> ($`MeVcm^2`$). The dominant reaction, $`\overline{\nu }_epne^+`$, with $`\sigma =7.5\times 10^{44}E_\nu ^2`$ ($`MeV^2cm^2`$), is energetically possible only on free protons, as in H<sub>2</sub>O and in C<sub>n</sub> H<sub>2n+2</sub> detectors. The positron produced annihilates immediately, $`e^+e^{}2\gamma `$, whilst the neutron is moderated and captured after a mean time of about 180 $`\mu `$s ($`npd\gamma `$, with $`E_\gamma 2.2`$ MeV). The SNO detector with $`D_2O`$ will also detect $`\nu _enpe^{}`$. Because of the dependence of the cross section on neutrino energy, the average $`e^+`$ energy is about 2 MeV larger than the average $`\overline{\nu }_e`$ energy.
Only Supernova SN1987A in the Large Magellanic Cloud was observed with neutrinos. No other burst of supernova neutrinos has been detected. Present and future neutrino detectors, will only be able to observe galactic supernovae. An optimistic estimate of the rate of type-II Supernovae in our galaxy is one every 10–20 years. Several detectors are kept alive all the time and a worldwide supernova watch is in operation.
#### 3.1.2 High-energy neutrino astronomy
High-energy muon neutrinos can be detected via their charged-current interactions inside a detector or in the rock surrounding the detector leading to upgoing muons. Upward–going muons can be seen directly in Ĉerenkov detectors and can be separated by time–of–flight from downward–going muons in scintillators. At very high energies the $`\nu _\mu \mu `$ angle is small and the effective target may be large. In order to observe celestial ”point” sources of high–energy neutrinos one should plot for each muon its declination versus right ascension. A celestial source would reveal itself as an excess of events (in a certain direction) above the atmospheric neutrino background. Now, people are also looking at time coincidences with $`\gamma `$ -ray bursts.
Several underground experiments performed searches for astrophysical sources of $`\nu _\mu `$ , with negative results. In order to establish a flux limit for a specific source one may consider an error circle corresponding to the resolution of the detector in that direction ( $`3^{}`$ for tracking detectors, considerably more for H<sub>2</sub>O Ĉerenkov detectors), determine the number of events in that circle, and subtract the corresponding number of events expected from atmospheric neutrinos. MACRO, with about 1000 muon events, quotes limits at the $`10^{14}cm^2s^1`$ level .
Neutrino telescopes. Much larger detectors, the so called Neutrino Telescopes, will be required to really attack the field of $`\nu _\mu `$ astronomy. Prototypes of neutrino telescopes may be considered the Ĉerenkov detectors NESTOR, ANTARES and NEMO under deep sea water, Baikal under lake water and AMANDA under ice at the South Pole . The final detector will be around $`1km^3`$ of water or ice.
### 3.2 Searches for WIMPs
Weakly Interactive Massive Particles (WIMPs) could be part of the galactic dark matter. WIMPs should be neutral particles which may form a dissipationless gas trapped in the gravitational field of our Galaxy. Suitable WIMP candidates should have lifetimes comparable to the age of the Universe. In SUSY models, like the MSSM and SUGRA, they may be identified with the lightest neutralino; it is ok if R parity is violated provided it leads to a long lifetime neutralino. WIMPs have been searched for by direct and indirect methods.
\- Direct searches. WIMPs may be searched for via their interactions in refined low energy detectors of $`10100kg`$ mass. The WIMPs scatter elastically with the nuclei of the detector, with cross sections of the order of the weak ones or smaller. A scattering leads to a recoil of few keV energy. The detectors must have low radioactivity, be well shielded and use electronics which reduces unwanted noise signals. The DAMA collaboration presented results obtained with a $`100kg`$ NaI (Tl) detector, looking for a signal modulated over a one year period. They find a probable signal which could correspond to a neutralino mass of $`5060GeV`$ .
\- Indirect methods. WIMPs could be intercepted by celestial bodies, slowed down and trapped in their centres. WIMPs and anti-WIMPs could annihilate and yield neutrinos of GeV or TeV energies. The neutrinos would travel and interact below the detector yielding high energy muons which can be detected. The search should be performed in small angular windows around the directions of the celestial bodies. The $`90\%`$ C.L. MACRO limit for the flux from the Earth centre is $`10^{14}cm^2s^1`$ for a $`10^{}`$ cone around the vertical, see Fig. 9 . For the same cone searched for around the Sun direction, the limit stands at $`1.4\times 10^{14}cm^2s^1`$.
## 4 HIGH ENERGY COSMIC RAYS
The all-particle spectrum of cosmic rays is shown in Fig. 10
### 4.1 Underground muons
Underground experiments detect a sizeable downward flux of high-energy muons, single and multiple, coming from high-energy cosmic rays. Muons reaching the Gran Sasso detectors traverse a minimum path length of 3100 m.w.e. and an average one of 3700 m.w.e.. A muon must therefore have an energy larger than $`1.3TeV`$ to reach the detectors. The muon distribution in local coordinates (azimuth $`\phi `$ and zenith $`\theta `$) reflects the shape of the mountain: it may be considered an x–ray photograph of the mountain. Experiments proved that the arrival time distribution of underground muons is random.
The vertical muon flux $`I(h)`$, where $`h`$ is the slant depth, was measured with increasing accuracy by several experiments. The flux may be represented by
$$I(h)=B\left(h_1/h\right)^2e^{h/h_1},(3)$$
with $`B=(1.81\pm 0.06)\times 10^6\text{cm}^2\text{s}^1\text{sr}^1`$, $`h_1=(1231\pm 1)\text{hg cm}^2`$. The muon surface flux, obtained from the measured underground muon flux, is $`dN_\mu /dEd\mathrm{\Omega }=AE^\gamma `$ with $`\gamma 2.78`$.
Seasonal variations. Selected muon data from several experiments were used to search for seasonal variations. The muon rate shows clear variations of about $`\pm 1.4\%`$ amplitude which repeats over the years. The muons come from pion and kaon decays in the upper atmosphere (at depths of less than $`200mbar`$); their intensity becomes greater when the atmosphere is warmer. A new measurement performed by AMANDA at the South pole exhibits a much larger effect ($`\pm 15\%`$) reflecting the 6 month darkness and 6 month light . From underground muons it is possible to measure the effective temperature of the higher atmosphere to about $`1^{}`$C!
Muon astronomy. In “muon astronomy” one assumes that high–energy muons remember the arrival direction of the parent high–energy particle with the hope that the parent particle has not deviated. Thus a search may be made for celestial point sources, d.c., periodic or episodic. The interest in muon astronomy started in 1985 with reports of an excess of underground muons from the direction of Cyg. X–3 and with the Cyg. X–3 periodicity. Some reports of muon excesses could be connected with intense radio flares. In order to exclude with certainty a variation of the muon flux from the direction of Cyg. X–3 one has to analize data over a long period of time. Upper limits for a d.c. signal were established for specific sources, Cyg. X–3, Her X1, 1E2259+59, and the Crab. The d.c. limits range from 3 to $`6\times 10^{13}cm^2s^1`$ . For Cyg. X–3, MACRO searched for a muon signal modulated by the 4.8 h X–ray period. The phase diagram does not exhibit any excess above background in any phase bin. The upper limit on a modulated sygnal is $`F_{mod}3\times 10^{13}cm^2s^1`$.
Multiple muons. Multiple muons carry infomation about the energy spectrum and the chemical composition of primary cosmic rays with energies $`50`$ TeV. The sensitivity to composition arises from the fact that heavy nuclei are more effective than protons in producing multiple muons. The measurable distributions are: i) The decoherence function (the distribution of the distance between two muons) . ii) The decorrelation function (the double–differential distribution of two muon relative angles). iii) The multiplicity distribution. iv) The muon group sub-structure.
The analyses require a model of the hadronic interactions at high energy (nucleon–nucleus and nucleus–nucleus), trial models of the energy variation of the composition of cosmic rays, simulation of the cascade in air and in the rock, and a simulation of the detector. In practice one uses iteration procedures with continuous improvements in models and simulations, and eventually a multiparameter fit of all avaible data.
A slow increase of the average primary mass is observed when going from $`10^3`$ to $`10^4TeV`$, i.e. when crossing the “knee” of the cosmic ray all–particle flux .
It has to be noted that the muons in the same bundle arrive at the same time to within few $`ns`$.
### 4.2 Cosmic rays of highest energies
The origin of high–energy cosmic rays is essentially unknown, and it is difficult to devise acceleration mechanisms for the highest energy cosmic rays. Recently magnetic monopoles of relatively low mass accelerated by the galactic magnetic field to high energies and high velocites have been proposed as possible sources of the highest energy cosmic rays. It should be remembered that cosmic ray nuclei with energies $`>4\times 10^{19}eV`$ cannot come from distances $`>50Mpc`$ because of the Greisen cut–off caused by the interaction of protons with the $`2.7K`$ photons of the cosmic backround radiation (at these energies the c.m. $`p\gamma `$ energy is above pion threshold, the cross section becomes large, and cosmic rays are soon degraded in energy).
It is clear that more data are needed and that this requires large Extensive Air Shower Arrays (EAS). We have heard reports from KASKADE and EASTOP . The largest new project is AUGER, with a first array in South America; a similar array will also be built in North America. Each Auger array will cover $`5000km^2`$, with different types of detectors (hybrid air shower detectors): sampling water tanks and improved fly’s Eyes detectors which detect the nitrogen luminescence in the atmosphere, thus measuring the shower profile .
Among the different types of proposed detectors for large arrays we have heard Sorel’s presentation about the possibility of using standard solar pannels connected in series/parallel to detect the Ĉerenkov light .
## 5 SCIENTIFIC EXPLORATION OF SPACE
At this meeting we had a number of reports on physics and astrophysics research performed with balloons, satellites and the space station. There is an increasing effort in this field.
Recently balloon experiments have been performed to measure cosmic ray muons in the atmosphere : these measurements are relevant for a more precise determination of the atmospheric neutrino flux.
The BeppoSAX satellite measured x-ray bursts, identifying 14 x-ray sources as the counterparts of $`\gamma `$-ray bursts (see Section 6) .
The AMS (Alpha Magnetic Spectrometer) experiment on the space station should make a thorough search for antimatter, measure the cosmic ray composition, and perform other searches. A test flight was successful and it has already provided important information on the flux of $`p,d,{}_{}{}^{3}He,{}_{}{}^{4}He`$, and limits on antimatter .
A variety of experiments are becoming realities. For instance PAMELA is measuring $`\overline{p},e^+`$ of 100-200 GeV and is making a search for $`\overline{He}`$ nuclei . AGILE should be operative in 2002 ; it has several $`\gamma `$ ray detectors optimized to cover different $`\gamma `$ ray energies above $`30MeV`$, preliminary results have been obtained by NINA, etc. .
## 6 $`\gamma `$ -RAY BURSTS
Since few years, the observation of $`\gamma `$-ray bursts (GRBs) poses one of the main misteries of astrophysics. The $`\gamma `$-ray observatory, on board the Compton satellite, observed every day, a new $`\gamma `$-ray burst of MeV energy. The burst durations are from 30 ms to 1000s (but this depends on the sensitivity and time resolution of the experiment: one may only see the tip of an iceberg). The bursts come from all directions of space; in almost all cases they represent a single episode. The rise time of the bursts is very fast, and this suggests that they could be connected with neutron stars. Measurements from the BeppoSAX satellite observe the GRBs at x-ray energies, Fig. 11, determining more accurately the position of the source; they see a tail in intensity (afterglow). When seeing a burst, BeppoSAX alerts the astronomical community; it was thus possible to observe the optical counterparts of x-ray emitters; the optical signal lasts a few days ; at least one appeared to be at the border of a far away galaxy. It should be stressed that, even if it is seen in x-ray and in the visible, most of the emitted energy of GRBs is in $`\gamma `$ rays, at MeV energies.
It would be interesting to observe the GRBs at higher energies (multi GeV) . It would even be more interesting to observe them with neutrinos; trials are being made, but probably one needs larger neutrino telescopes, with lower energy thresholds.
## 7 RARE PARTICLES AND PROTON DECAY SEARCHES
### 7.1 Magnetic Monopoles
Grand Unified Theories (GUTs) of electroweak and strong interactions predict the existence of magnetic monopoles (MMs) with large mass, larger than $`10^{16}`$ GeV, and magnetic charges $`g=ng_{Dirac}=nc/2e=n68e`$, with $`n=1,2,\mathrm{}`$ . These theories leave open the question of monopole abundance. MMs were probably produced in the early universe, at the end of the GUT era as point defects; others may have been produced in ultra–high energy collisions. Standard cosmology predicts too many monopoles, whereas models with inflation at the GUT phase transition predict very few. Several superstring models predict the existence of multiply–charged MMs $`(n=3)`$. In some models, the primordial monopoles appeared when the temperature of the universe reached relatively low values. These monopoles were probably not diluted by inflation in the early universe. The existence of large–scale magnetic fields, on the galactic scale, leads to an astrophysical constraint, the so–called Parker bound, with an upper limit on the monopole flux at the level of $`10^{15}cm^2s^1sr^1`$; an extended Parker bound leads to a flux limit almost an order of magnitude smaller.
Underground experiments have searched for MMs in the penetrating cosmic radiation using scintillators, gas tubes, and nuclear track detectors via $`dE/dx`$, time of flight and pulse shape analyses. At present there are only a few large experiments. They tested the sensitivity of their detectors to low velocity MMs.
New limits have been presented by MACRO and by AMANDA . The present limits on massive cosmic MMs are summarized in Fig. 12 for $`g=g_D`$ bare poles with mass $`>10^{16}`$ GeV and for catalysis cross sections smaller than few mb.
### 7.2 Dark Matter
Analyses of the rotation curves of stars in galaxies, and of galaxies in clusters of galaxies prove (assuming the validity of Newton’s law) that most of the matter is unseen: $`\mathrm{\Omega }_{vis}0.01,\mathrm{\Omega }_{halo}0.1,\mathrm{\Omega }_{TH}1,\text{where}\mathrm{\Omega }=\rho /\rho _c`$.
The unseen DM could be: a) baryonic, in the form of gas, planets like jupiter, brown dwarfs, nuclearites; b) non baryonic, i.e. a gas of particles. In the latter case there could be: i) hot DM, i. e. particles which were relativistic when in the early universe they decoupled from the rest of matter and radiation (an example could be neutrinos with a mass of a few eV); ii) cold DM, i.e. particles which were non–relativistic at decoupling (for example the WIMPs, see Section 3.2). These particles are probably located in the galactic halos; their abundance in the vicinity of the solar system could typically be $`0.3GeV/cm^3`$, and their velocity $`300km/s`$.
Nuclearites. The hypothesized stable phase of quark matter, called strange quark matter or nuclearites, formed by quarks $`u,d`$ and $`s`$, may be the true ground state of QCD. Nuclearites could have masses ranging from a few GeV to the mass of a neutron star. Because of this wide range, searches were performed using a variety of experimental techniques. At this meeting new results have been presented, using techniques developed for MM searches. Limits for nuclearites with masses larger than $`0.1g`$ (which can penetrate the earth) are at the level of the limits of MMs ; the limits are twice as large for nuclearites with m $`<0.1g`$, which cannot traverse the earth.
Other ”exotic” objects, like the Q-balls (aggregates of squarks, sleptons and Higgs fields) have also been discussed at this meeting .
### 7.3 Proton decay
GUTs place quarks and leptons in the same multiplets. Quark $``$ lepton transitions are thus possible and should be mediated by supermassive vector bosons $`X,Y`$ with $`m10^{14}`$ $`GeV`$. A free proton may decay as $`N\mathrm{}^++`$ meson(s) or $`N\overline{\nu }+`$ meson(s). Proton decay, with a predicted lifetime of the order of $`10^{31}y`$, motivated the construction of the first underground detectors with masses of the order of $`1000t`$. Present detectors are either water Ĉerenkov detectors (SK) or tracking calorimeters (Soudan 2). Water detectors have larger masses and more free protons and may detect the sense of the track direction. Tracking calorimeters have a higher spatial resolution and a better $`\pi /\mu `$ separation at energies of about $`200MeV`$. Technical developments are being made towards a TPC type liquid chamber (ICARUS). A $`3t`$ prototype works well and a $`600t`$ module is under construction. Superkamiokande presented the following limits
$`\{\begin{array}{cc}\tau BR(pe^+\pi ^{})\hfill & >1.6\times 10^{33}y\hfill \\ \tau BR(p\overline{\nu }K^+)\hfill & >6.7\times 10^{32}y\hfill \end{array}`$
at $`90\%`$ C.L. ; they rule out the simplest SU(5) GUT models.
## 8 SELECTED RESULTS FROM ACCELERATOR EXPERIMENTS
LEP. The four experiments at the LEP positron-electron collider provided new improved precision values of the $`Z^{}`$ lines shape. The $`Z^{}`$ mass is now known with a precision of two parts in $`10^5`$, and has acquired the status of one of the three basic inputs of the Standard Model (SM) of particle physics. An important quantity derived from the line shape parameters is the number of light neutrino species which is now
$`N_\nu =\left(\frac{\mathrm{\Gamma }_{inv}}{\mathrm{\Gamma }_l}\right)/\left(\frac{\mathrm{\Gamma }_\nu }{\mathrm{\Gamma }_l}\right)_{SM}=2.9835\pm 0.0083`$, see Fig. 13; a direct method (the neutrino counting method) confirms this result .
From this determination one may deduce the amount of helium expected in primordial nucleosynthesis: one expects $`24\%`$, in fair agreement with astrophysical data. The charged lepton universality is now established at the $`0.1\%`$ level; the muon and the $`tau`$ lepton appear more and more to be replicas of the electron. The increased energy of LEP (LEP2) allowed to study the reaction $`e^+e^{}Z^{}W^+W^{}`$, proving the existence of the triple boson vertex, $`Z^{}W^+W^{}`$, and a precise measurement of the $`W^\pm `$ mass, which very likely will become one of the three inputs of the SM. LEP1 allowed a detailed study of QCD properties, in particular a precise determination of the strong coupling constant and of its variation with energy (LEP2 is showing that the variation continues to higher energies). It may be worth remembering that precision measurements lead to the first determination, below threshold, of the mass of the quark top and now gives a hint of the mass of the Higgs boson. LEP2 with data up to $`\sqrt{s}=202GeV`$ yields direct limits on the S.M. Higgs boson, $`m_H^{}>103GeV`$, and on a variety of particles prediced by models beyond the SM.
HERA. At the asymmetric $`e^+p`$ collider ($`E_e=26.7GeV,Ep=820GeV`$) at Hamburg, two experiments are providing a wealth of information on CC deep inelastic scattering, in particular at very small values of $`x`$ and large values of $`Q^2`$. They also measure the neutral current (NC) cross section, see Fig. 14 ; notice that it is related to parton densities and takes into account effects of $`xF_3`$. A considerable part of the HERA program concerns the searches for particles predicted by models beyond the SM . In particular new more stringent limits on leptoquarks have been presented.
## 9 GRAVITATIONAL WAVES
The earth should be continously bombarded by gravitational waves produced by distant celestial bodies subject to ”strong” gravitational effects. The amplitude of the gravitational wave emitted by a celestial body is proportional to its mass, to its acceleration, and to the inhomogeneity in its mass distribution. Gravitational waves are emitted when the quadrupole moment of an object of large mass is subject to large and fast variations. Only large celestial bodies subject to unusual accelerations should produce sizeable gravitational radiation measurable on earth. These bodies may be binary systems of close-by stars (in particular when a neutron star is about to fall on the other); they yield a periodic emission of gravitational waves, with frequencies from few hundred Hz to 1 MHz. Asymmetric supernovae explosions may give bursts of gravitational waves, with frequencies of the order of 1 kHz over few ms. Also vibrating black holes, star accretions, galaxy formation, and the Big Bang may produce or have produced gravitational waves.
A gravitational wave is a transverse wave which travels at the speed of light. A gravitational wave should modify the distances between objects in the plane perpendicular to the direction of propagation of the wave. These deformations are expected to be extremely small. It has been estimated that a star collapse at the centre of our galaxy may produce a variation of the order of $`h10^{18}`$ metre per metre of separation of two objects on earth. The Supernova 1987A in the large Magellanic Cloud could probably have produced a distortion 10 times smaller. A collapse in the Virgo cluster (at MPc), should yield relative variations of $`10^{21}`$.
Very sensitive instruments are needed to observe gravitational waves. The two main lines developed until now are resonating bars at low temperatures and long laser interferometric systems. A major program is underway for both types of detectors, hoping to be able to detect gravitational star collapses up to the Virgo cluster, corresponding to $`h10^{21}`$. The supercooled (0.1 K) bars NAUTILUS and AURIGA are operating in Frascati and Legnaro (Padova), respectively. A pair of long interefometers (LIGO) are under construction in the US, while one long interferometer is under construction at Pisa (VIRGO) . Several detectors, in coincidence, are needed to ensure that the observed signal is not spurious.
The detection of gravitational waves would have far reaching consequences. It would prove the validity of the general theory of relativity; in astrophysics it would open up a new observational window related to violent phenomena in the universe.
## 10 CONCLUSIONS
We had an interesting and lively workshop on topical subjects. Many new interesting results were presented, as well as many new proposals: the field of astroparticle physics in general and of neutrinos in particular is very alive and we look forward to many new exciting results in the future.
I would like to thank Ms. Luisa De Angelis for typing the manuscript. I acknowledge the cooperation of all the colleagues at the workshop and the help of the colleagues in Bologna, in particular of Dr. M. Giorgini. |
warning/0001/math0001181.html | ar5iv | text | # Some boundedness results for Fano-like Moishezon manifolds
## Introduction
In this Note, we say that a compact complex manifold $`X`$ is a Fano-like manifold if it becomes Fano after a finite sequence of blow-ups along smooth connected centers, i.e if there exist a Fano manifold $`\stackrel{~}{X}`$ and a finite sequence of blow-ups along smooth connected centers $`\pi :\stackrel{~}{X}X`$. We say that a Fano-like manifold $`X`$ is simple if there exists a smooth submanifold $`Y`$ of $`X`$ ($`Y`$ may not be connected) such that the blow-up of $`X`$ along $`Y`$ is Fano. If $`Z`$ is a projective manifold, we call smooth blow-down of $`Z`$ (with an $`s`$-dimensional center) a map $`\pi `$ and a manifold $`Z^{}`$ such that $`\pi :ZZ^{}`$ is the blow-up of $`Z^{}`$ along a smooth connected submanifold (of dimension $`s`$). We say that a smooth blow-down of $`Z`$ is projective (resp. non projective) if $`Z^{}`$ is projective (resp. non projective).
It is well-known that any Moishezon manifold becomes projective after a finite sequence of blow-ups along smooth centers. Our aim is to bound the geometry of Moishezon manifolds becoming Fano after one blow-up along a smooth center, i.e the geometry of simple non projective Fano-like manifolds.
Our results in this direction are the following, the simple proof of Theorem 1 has been communicated to us by Daniel Huybrechts.
Theorem 1. Let $`Z`$ be a Fano manifold of dimension $`n`$. Then, there is only a finite number of smooth blow-downs of $`Z`$.
Let us recall here that the assumption $`Z`$ Fano is essential : there are projective smooth surfaces with infinitely many $`1`$ rational curves, hence with infinitely many smooth blow-downs.
Since there is only a finite number of deformation types of Fano manifolds of dimension $`n`$ (see \[KMM92\] and also \[Deb97\] for a recent survey on Fano manifolds) and since smooth blow-downs are stable under deformations \[Kod63\], we get the following corollary (see section 1 for a detailed proof) :
Corollary 1. There is only a finite number of deformation types of simple Fano-like manifolds of dimension $`n`$.
The next result is essentially due to Wiśniewski (\[Wis91\], prop. (3.4) and (3.5)). Before stating it, let us define
$$d_n=\mathrm{max}\{(K_Z)^n|Z\text{is a Fano manifold of dimension}n\}$$
and
$$\rho _n=\mathrm{max}\{\rho (Z):=\mathrm{rk}(\mathrm{Pic}(Z)/\mathrm{Pic}^0(Z))|Z\text{is a Fano manifold of dimension}n\}.$$
The number $`\rho _n`$ is well defined since there is only a finite number of deformation types of Fano manifolds of dimension $`n`$ and we refer to \[Deb97\] for an explicit bound for $`d_n`$.
Theorem 2. Let $`X`$ be an $`n`$-dimensional simple non projective Fano-like manifold, $`Y`$ a smooth submanifold such that the blow-up $`\pi :\stackrel{~}{X}X`$ of $`X`$ along $`Y`$ is Fano, and $`E`$ the exceptional divisor of $`\pi `$. Then
1. if each component of $`Y`$ has Picard number equal to one, then each component of $`Y`$ has ample conormal bundle in $`X`$ and is Fano. Moreover $`\mathrm{deg}_{K_{\stackrel{~}{X}}}(E)(\rho _n1)d_{n1}`$.
2. if $`Y`$ is a curve, then (each component of) $`Y`$ is a smooth rational curve with normal bundle $`𝒪_^1(1)^{n1}`$.
Finally, we prove here the following result :
Theorem 3. Let $`Z`$ be a Fano manifold of dimension $`n`$ and index $`r`$. Suppose there is a non projective smooth blow-down of $`Z`$ with an $`s`$-dimensional center. Then
$$r(n1)/2\text{ and }sr.$$
Moreover,
1. if $`r>(n1)/3`$, then $`s=n1r`$ ;
2. if $`r<(n1)/2`$ and $`s=r`$, then $`Y^r`$.
Recall that the index of a Fano manifold $`Z`$ is the largest integer $`m`$ such that $`K_Z=mL`$ for $`L`$ in the Picard group of $`Z`$.
Remarks.
1. For a Fano manifold $`X`$ of dimension $`n`$ and index $`r`$ with second Betti number greater than or equal to $`2`$, it is known that $`2rn+2`$ \[Wi91\], with equality if and only if $`X^{r1}\times ^{r1}`$.
2. Fano manifolds of even dimension (resp. odd dimension $`n`$) and middle index (resp. index $`(n+1)/2`$) with $`b_22`$ have been intensively studied, see for example \[Wis93\]. Our Theorem 3 shows that there are no non projective smooth blow-down of such a Fano manifold, without using any explicit classification.
3. The assumption that there is a non projective smooth blow-down of $`Z`$ is essential in Theorem 3 : the Fano manifold obtained by blowing-up $`^{2r1}`$ along a $`^{r1}`$ has index $`r`$.
## 1. Proof of Theorem 1 and Corollary 1. An example.
### 1.1. Proof of Theorem 1.
Thanks to D. Huybrechts for the following proof.
Let $`Z`$ be a Fano manifold and $`\pi :ZZ^{}`$ a smooth blow-down of $`Z`$ with an $`s`$-dimensional connected center. Let $`f`$ be a line contained in a non trivial fiber of $`\pi `$. Then, the Hilbert polynomial $`P_{K_Z}(m):=\chi (f,m(K_Z)_{|f})`$ is determined by $`s`$ and $`n`$ since $`K_Zf=ns1`$ and $`f`$ is a smooth rational curve. Since $`K_Z`$ is ample, the Hilbert scheme $`\mathrm{Hilb}_{K_Z}`$ of curves in $`Z`$ having $`P_{K_Z}`$ as Hilbert polynomial is a projective scheme, hence has a finite number of irreducible components. Since each curve being in the component $``$ of $`\mathrm{Hilb}_{K_Z}`$ containing $`f`$ is contracted by $`\pi `$, there is only a finite number of smooth blow-downs of $`Z`$ with an $`s`$-dimensional center.
### 1.2. Proof of Corollary 1.
Let us first recall (\[Deb97\] section 5.2) that there exists an integer $`\delta (n)`$ such that every Fano $`n`$-fold can be realized as a smooth submanifold of $`^{2n+1}`$ of degree at most $`\delta (n)`$. Let us denote by $`T`$ a closed irreducible subvariety of the disjoint union of Chow varieties of $`n`$-dimensional subvarieties of $`^{2n+1}`$ of degree at most $`\delta (n)`$, and by $`\pi :𝒳_TT`$ the universal family.
Step 1 : Stability of smooth blow-downs. Fix $`t_0`$ in the smooth locus $`T_{smooth}`$ of $`T`$ and suppose that $`X_{t_0}:=\pi ^1(t_0)`$ is a Fano $`n`$-fold and there exists a smooth blow-down of $`X_{t_0}`$ (denote by $`E_{t_0}`$ the exceptional divisor, $`P`$ its Hilbert polynomial with respect to $`𝒪_{^{2n+1}}(1)`$). Let $`S`$ be the component of the Hilbert scheme of $`(n1)`$-dimensional subschemes of $`^{2n+1}`$ with Hilbert polynomial $`P`$ and $`u:_SS`$ the universal family. Finally, let $`I`$ be the following subscheme of $`T\times S`$ :
$$I=\{(t,s)|u^1(s)X_t\}$$
and $`p:IT`$ the proper algebraic map induced by the first projection. Thanks to the analytic stability of smooth blow-downs due to Kodaira (see \[Kod63\], Theorem 5), the image $`p(I)`$ contains an analytic open neighbourhood of $`t_0`$ hence it also contains a Zariski neighbourhood of $`t_0`$. Moreover, since exceptional divisors are rigid, the fiber $`p^1(t)`$ is a single point for $`t`$ in a Zariski neighbourhood of $`t_0`$. Finally, we get algebraic stability of smooth blow-downs (the $`^r`$-fibered structure of exceptional divisor is also analytically stable - \[Kod63\], Theorem 4 - hence algebraically stable by the same kind of argument).
Step 2 : Stratification of $`T`$ by the number of smooth blow-downs. For any integer $`k0`$, let us define
$$\begin{array}{c}U_k(T)=\{tT_{smooth}|X_t\text{ is a Fano manifold and there exists at least}\\ \text{ }k\text{ smooth blow-downs of }X_t\};\end{array}$$
and $`U_1(T)=T_{smooth}`$. Thanks to Step 1, $`U_k(T)`$ is Zariski open in $`T`$, and thanks to Theorem 1,
$$\underset{k1}{}U_k(T)=\mathrm{}.$$
Since $`\{U_k(T)\}_{k1}`$ is a decreasing sequence of Zariski open sets, by noetherian induction, we get that there exists an integer $`k`$ such that $`U_k(T)=\mathrm{}`$ and we can thus define
$$k(T):=\mathrm{max}\{k1|U_k(T)\mathrm{}\},U(T):=U_{k(T)}(T).$$
Finally, we have proved that $`U(T)`$ is a non empty Zariski open set of $`T_{smooth}`$ such that for every $`tU(T)`$, $`Z_t`$ is a Fano $`n`$-fold with exactly $`k(T)`$ smooth blow-downs ($`k(T)=1`$ means that for every $`tT_{smooth}`$, $`X_t`$ is not a Fano manifold).
Now let $`T_0=T`$, and $`T_1`$ be any closed irreducible component of $`T_0U(T_0)`$. We get $`U(T_1)`$ as before and denote by $`T_2`$ any closed irreducible component of $`T_1U(T_1)`$, and so-on. Again by noetherian induction, this process terminates after finitely many steps and we get a finite stratification of $`T`$ such that each strata corresponds to an algebraic family of Fano $`n`$-folds with the same number of smooth blow-downs.
Step 3 : Conclusion. Since there is only a finite number of irreducible components in the Chow variety of Fano $`n`$-folds, each being finitely stratified by Step 2, we get a finite number of deformation types of simple Fano-like $`n`$-folds.
As it has been noticed by Kodaira, it is essential to consider only smooth blow-downs. A $`2`$ rational smooth curve on a surface is, in general, not stable under deformations of the surface.
### 1.3. An example.
Before going further, let us recall the following well known example. Let $`Z`$ be the projective $`3`$-fold obtained by blowing-up $`^3`$ along a smooth curve of type $`(3,3)`$ contained in a smooth quadric $`𝒬`$ of $`^3`$. Let $`\pi `$ denotes the blow-up $`Z^3`$. Then $`Z`$ is a Fano manifold of index one and there are at least three smooth blow-downs of $`Z`$ : $`\pi `$, which is projective, and two non projective smooth blow-downs consisting in contracting the strict transform $`𝒬^{}`$ of the quadric $`𝒬`$ along one of its two rulings (the normal bundle of $`𝒬^{}`$ in $`Z`$ is $`𝒪(1,1)`$).
Lemma 1. There are exactly three smooth blow-downs of $`Z`$.
Proof : the Mori cone $`\mathrm{NE}(Z)`$ is a $`2`$-dimensional closed cone, one of its two extremal rays being generated by the class of a line $`f_\pi `$ contained in a non trivial fiber of $`\pi `$, the other one, denoted by $`[R]`$, by the class of one of the two rulings of $`𝒬^{}`$ (the two rulings are numerically equivalent, the corresponding extremal contraction consists in contracting $`𝒬^{}`$ to a singular point in a projective variety, hence is not a smooth blow-down). If $`E`$ is the exceptional divisor of $`\pi `$, we have
$$E[f_\pi ]=1,E[R]=3,𝒬^{}[f_\pi ]=1,𝒬^{}[R]=1.$$
Now suppose there exists a smooth blow-down $`\tau `$ of $`Z`$ with a $`1`$-dimensional center, which is not one of the three previously described. Let $`L`$ be a line contained in a non trivial fiber of $`\tau `$, then since $`K_Z[L]=1`$, we have $`[L]=a[f_\pi ]+b[R]`$ for some strictly positive numbers such that $`a+b=1`$. Since we have moreover
$$𝒬^{}[L]=ab=2a1\text{and}E[L]=3ba=34a,$$
we get $`a=b=1/2`$. Therefore $`𝒬^{}[L]=0`$ hence $`L`$ is disjoint from $`𝒬^{}`$ (it can not be contained in $`𝒬^{}`$ since $`𝒬_{|𝒬^{}}^{}=𝒪(1,1)`$). It implies that there are two smooth blow-downs of $`Z`$ with disjoint exceptional divisors, which is impossible since $`\rho (Z)=2`$.
Finally, if there is a smooth blow-down $`\tau :ZZ^{}`$ of $`Z`$ with a $`0`$-dimensional center, then $`Z^{}`$ is projective and $`\tau `$ is a Mori extremal contraction, which is again impossible since we already met the two Mori extremal contractions on $`Z`$.
## 2. Non projective smooth blow-downs on a center with Picard number $`1`$. Proof of Theorem 2.
The proof of Theorem 2 we will give is close to Wiśniewski’s one but we give two intermediate results of independant interest.
### 2.1. On the normal bundle of the center.
Let us recall that a smooth submanifold $`A`$ in a complex manifold $`W`$ is contractible to a point (i.e. there exists a complex space $`W^{}`$ and a map $`\mu :WW^{}`$ which is an isomorphism outside $`A`$ and such that $`\mu (A)`$ is a point) if and only if $`N_{A/W}^{}`$ is ample (Grauert’s criterion \[Gra62\]).
The following proposition was proved by Campana \[Cam89\] in the case where $`Y`$ is a curve and $`dim(X)=3`$.
Proposition 1. Let $`X`$ be a non projective manifold, $`Y`$ a smooth submanifold of $`X`$ such that the blow-up $`\pi :\stackrel{~}{X}X`$ of $`X`$ along $`Y`$ is projective. Then, for each connected component $`Y^{}`$ of $`Y`$ with $`\rho (Y^{})=1`$, the conormal bundle $`N_{Y^{}/X}^{}`$ is ample.
Before the proof, let us remark that $`Y`$ is projective since the exceptional divisor of $`\pi `$ is.
Proof of Proposition 1 : (following Campana) we can suppose that $`Y`$ is connected. Let $`E`$ be the exceptional divisor of $`\pi `$ and $`f`$ a line contained in a non trivial fiber of $`\pi `$. Since $`Ef=1`$, there is an extremal ray $`R`$ of the Mori cone $`\overline{\mathrm{NE}}(\stackrel{~}{X})`$ such that $`ER<0`$. Since $`ER<0`$, $`R`$ defines an extremal ray of the Mori cone $`\overline{\mathrm{NE}}(E)`$ which we still denote by $`R`$ (even if $`\overline{\mathrm{NE}}(E)`$ is not a subcone of $`\overline{\mathrm{NE}}(\stackrel{~}{X})`$ in general !). Since $`\rho (Y)=1`$, we have $`\rho (E)=2`$, hence $`\overline{\mathrm{NE}}(E)`$ is a $`2`$-dimensional closed cone, one of its two extremal rays being generated by $`f`$. Then :
1. either $`R`$ is not generated by $`f`$ and $`E_{|E}`$ is strictly negative on $`\overline{\mathrm{NE}}(E)\{0\}`$. In that case, $`E_{|E}=𝒪_E(1)`$ is ample by Kleiman’s criterion, which means that $`N_{Y/X}^{}`$ is ample.
2. or, $`R`$ is generated by $`f`$. In that case, the Mori contraction $`\phi _R:\stackrel{~}{X}Z`$ factorize through $`\pi `$ :
where $`\psi :XZ`$ is an isomorphism outside $`Y`$. Since the variety $`Z`$ is projective and $`X`$ is not, $`\psi `$ is not an isomorphism and since $`\rho (Y)=1`$, $`Y`$ is contracted to a point by $`\psi `$, hence $`N_{Y/X}^{}`$ is ample by Grauert’s criterion.
Let us prove the following consequence of Proposition 1:
Proposition 2. Let $`X`$ be a non projective manifold, $`Y`$ a smooth submanifold of $`X`$ such that the blow-up $`\pi :\stackrel{~}{X}X`$ of $`X`$ along $`Y`$ is projective with $`K_{\stackrel{~}{X}}`$ numerically effective (nef). Then, each connected component $`Y^{}`$ of $`Y`$ with $`\rho (Y^{})=1`$ is a Fano manifold.
Proof : we can suppose that $`Y`$ is connected. Let $`E`$ be the exceptional divisor of $`\pi `$. Since $`E_{|E}`$ is ample by Proposition 1, the adjunction formula $`K_E=K_{\stackrel{~}{X}|E}E_{|E}`$ shows that $`K_E`$ is ample, hence $`E`$ is Fano. By a result of Szurek and Wiśniewski \[SzW90\], $`Y`$ is itself Fano.
### 2.2. Proof of Theorem 2.
For the first assertion, we only have to prove that
$$\mathrm{deg}_{K_{\stackrel{~}{X}}}(E)(\rho _n1)d_{n1}.$$
Let $`Y^{}`$ be a connected component of $`Y`$ and $`E^{}=\pi ^1(Y^{})`$. Then, since $`E_{|E^{}}`$ is ample :
$$\mathrm{deg}_{K_{\stackrel{~}{X}}}(E^{})=(K_{\stackrel{~}{X}|E^{}})^{n1}=(K_E^{}+E_{|E^{}})^{n1}(K_E^{})^{n1}d_{n1}.$$
Now, if $`m`$ is the number of connected components of $`Y`$, then
$$\rho (\stackrel{~}{X})=m+\rho (X)m+1.$$
Putting all together, we get
$$\mathrm{deg}_{K_{\stackrel{~}{X}}}(E)(\rho _n1)d_{n1},$$
which ends the proof of the first point.
We refer to \[Wis91\] prop. (3.5) for the second point.
## 3. On the dimension of the center of non projective smooth blow-downs. Proof of Theorem 3.
Theorem 3 is a by-product of the more precise following statement and of Proposition 3 below :
Theorem 4. Let $`Z`$ be a Fano manifold of dimension $`n`$ and index $`r`$, $`\pi :ZZ^{}`$ be a non projective smooth blow-down of $`Z`$, $`YZ^{}`$ the center of $`\pi `$. Let $`f`$ be a line contained in a non trivial fiber of $`\pi `$, then
1. if $`f`$ generates an extremal ray of $`\mathrm{NE}(Z)`$, then $`dim(Y)(n1)/2`$.
2. if $`f`$ does not generate an extremal ray of $`\mathrm{NE}(Z)`$, then $`dim(Y)r`$. Moreover, if $`dim(Y)=r`$, then $`Y`$ is isomorphic to $`^r`$.
In both cases (i) and (ii), $`Y`$ contains a rational curve.
The proof relies on Wiśniewski’s inequality (see \[Wis91\] and \[AnW95\]), which we recall now for the reader’s convenience : let $`\phi :XY`$ be a Fano-Mori contraction (i.e $`K_X`$ is $`\phi `$-ample) on a projective manifold $`X`$, $`\mathrm{Exc}(\phi )`$ its exceptional locus and
$$l(\phi ):=\mathrm{min}\{K_XC;C\text{rational curve contained in}\mathrm{Exc}(\phi )\}$$
its length, then for every non trivial fiber $`F`$ :
$$dim\mathrm{Exc}(\phi )+dim(F)dim(X)1+l(\phi ).$$
Proof of Theorem 4. The method of proof is taken from Andreatta’s recent paper \[And99\] (see also \[Bon96\]).
First case : suppose that a line $`f`$ contained in a non trivial fiber of $`\pi `$ generates an extremal ray $`R`$ of $`\mathrm{NE}(Z)`$. Then the Mori contraction $`\phi _R:ZW`$ factorizes through $`\pi `$ :
where $`\psi `$ is an isomorphism outside $`Y`$. In particular, the exceptional locus $`E`$ of $`\pi `$ is equal to the exceptional locus of the extremal contraction $`\phi _R`$.
Let us now denote by $`\psi _Y`$ the restriction of $`\psi `$ to $`Y`$, $`s=dim(Y)`$, $`\pi _E`$ and $`\phi _{R,E}`$ the restriction of $`\pi `$ and $`\phi _R`$ to $`E`$. Since $`Z^{}`$ is not projective, $`\psi _Y`$ is not a finite map. Since $`\phi _R`$ is birational, $`W`$ is $``$-Gorenstein, hence $`K_W`$ is $``$-Cartier and $`K_Z^{}=\psi ^{}K_W`$. Therefore, $`K_Z^{}`$ is $`\psi `$-trivial, hence $`K_Y+detN_{Y/Z^{}}^{}`$ is $`\psi _Y`$-trivial. Moreover, $`𝒪_E(1)=E_{|E}`$ is $`\phi _{R,E}`$-ample by Kleiman’s criterion, hence $`N_{Y/Z^{}}^{}`$ is $`\psi _Y`$-ample. Finally, $`\psi _Y`$ is a Fano-Mori contraction, of length greater or equal to $`ns=\mathrm{rk}(N_{Y/Z^{}}^{})`$. Together with Wiśniewski’s inequality applied on $`Y`$, we get that for every non trivial fiber $`F`$ of $`\psi _Y`$
$$2sdim(F)+dim\mathrm{Exc}(\psi _Y)ns+s1$$
hence $`2sn1`$. Moreover, $`\mathrm{Exc}(\psi _Y)`$ is covered by rational curves, hence $`Y`$ contains a rational curve.
Second case : suppose that a line $`f`$ contained in a non trivial fiber of $`\pi `$ does not generate an extremal ray $`R`$ of $`\mathrm{NE}(Z)`$. In that case, since $`Ef=1`$, there is an extremal ray $`R`$ of $`\mathrm{NE}(Z)`$ such that $`ER<0`$. In particular, the exceptional locus $`\mathrm{Exc}(R)`$ of the extremal contraction $`\phi _R`$ is contained in $`E`$, and since $`f`$ is not on $`R`$, we get for any fiber $`F`$ of $`\phi _R`$ :
$$dim(F)s=dim(Y).$$
By the adjunction formula, $`K_E=K_{Z|E}E_{|E}`$, the length $`l_E(R)`$ of $`R`$ as an extremal ray of $`E`$ satisfies
$$l_E(R)r+1,$$
where $`r`$ is the index of $`Z`$. Together with Wiśniewski’s inequality applied on $`E`$, we get :
$$r+1+(n1)1s+dim(\mathrm{Exc}(R))s+n1.$$
Finally, we get $`rs`$, and since the fibers of $`\phi _R`$ are covered by rational curves, there is a rational curve in $`Y`$. Suppose now (up to the end) that $`r=s`$. Then $`E`$ is the exceptional locus of the Mori extremal contraction $`\phi _R`$. Moreover, $`K_Z+r(E)`$ is a good supporting divisor for $`\phi _R`$, and since every non trivial fiber of $`\phi _R`$ has dimension $`r`$, $`\phi _R`$ is a smooth projective blow-down. In particular, the restriction of $`\pi `$ to a non trivial fiber $`F^r`$ induces a finite surjective map $`\pi :F^rY`$ hence $`Y^r`$ by a result of Lazarsfeld \[Laz83\].
This ends the proof of Theorem 4.
The proof of Theorem 4 does not use the hypothesis $`Z`$ Fano in the first case. We therefore have the following :
Corollary 2. Let $`Z`$ be a projective manifold of dimension $`n`$, $`\pi :ZZ^{}`$ be a non projective smooth blow-down of $`Z`$, $`YZ^{}`$ the center of $`\pi `$. Let $`f`$ be a line contained in a non trivial fiber of $`\pi `$ and suppose $`f`$ generates an extremal ray of $`\overline{\mathrm{NE}}(Z)`$. Then $`dim(Y)(n1)/2`$. Moreover, if $`dim(Y)=(n1)/2`$, then $`Y`$ is contractible on a point.
We finish this section by the following easy proposition, which combined with Theorem 4 implies Theorem 3 of the Introduction :
Proposition 3. Let $`Z`$ be a Fano manifold of dimension $`n`$ and index $`r`$, $`\pi :ZZ^{}`$ be a smooth blow-down of $`Z`$, $`YZ^{}`$ the center of $`\pi `$. Then $`n1dim(Y)`$ is a multiple of $`r`$.
Proof. Write
$$K_Z=rL\text{ and }K_Z=\pi ^{}K_Z^{}(n1dim(Y))E$$
where $`E`$ is the exceptional divisor of $`\pi `$. Let $`f`$ be a line contained in a fiber of $`\pi `$. Then $`rLf=n1dim(Y)`$, which ends the proof.
Proof of Theorem 3. Let $`Z`$ be a Fano manifold of dimension $`n`$ and index $`r`$ and suppose there is a non projective smooth blow-down of $`Z`$ with an $`s`$-dimensional center. By Proposition 3, there is a strictly positive integer $`k`$ such that $`n1kr=s`$. By Theorem 4, either $`n1kr(n1)/2`$ or $`n1krr`$. In both cases, it implies that $`r(n1)/2`$ and therefore $`sr`$. If $`r>(n1)/3`$, since $`n1(k+1)r>(k+1)(n1)/3`$, we get $`k=1`$ and $`s=n1r`$.
## 4. Rational curves on simple Moishezon manifolds.
The arguments of the previous section can be used to deal with the following well-known question : does every non projective Moishezon manifold contain a rational curve ? The answer is positive in dimension three (it is due to Peternell \[Pet86\], see also \[CKM88\] p. 49 for a proof using the completion of Mori’s program in dimension three).
Proposition 4. Let $`Z`$ be a projective manifold, $`\pi :ZZ^{}`$ be a non projective smooth blow-down of $`Z`$. Then $`Z^{}`$ contains a rational curve.
Proof. With the notations of the previous section, it is clear in the first case where a line $`f`$ contained in a non trivial fiber of $`\pi `$ generates an extremal ray $`R`$ of $`\overline{\mathrm{NE}}(Z)`$ (in that case, the center of $`\pi `$ contains a rational curve). In the second case, since $`f`$ is not extremal and $`K_Z`$ is not nef, there is a Mori contraction $`\phi `$ on $`Z`$ such that any rational curve contained in a fiber of $`\phi `$ is mapped by $`\pi `$ to a non constant rational curve in $`Z^{}`$.
———–
L.B. : Institut Fourier, UMR 5582, Université de Grenoble 1, BP 74. 38402 Saint Martin d’Hères. FRANCE
e-mail : bonavero@ujf-grenoble.fr
S.T. : Department of Mathematics, Graduate School of Science, Osaka University. Toyonaka, Osaka, 560-0043 JAPAN.
e-mail : taka@math.sci.osaka-u.ac.jp |
warning/0001/astro-ph0001166.html | ar5iv | text | # A Definitive Optical Detection of a Supercluster at 𝑧≈0.91
## 1 Introduction
Superclusters comprise the largest known systems of galaxies, containing 2 – 5 massive clusters and extending over 10 – 20 Mpc (e.g., Bahcall & Soneira 1984; Postman, Geller & Huchra 1988; Quintana et al. 1995; Small et al. 1998). Since the dynamical timescales of superclusters are comparable to the Hubble time, large scale structures observed today are cosmic fossils of conditions that existed in the early universe. As a result, studies of these systems can be used to measure the cosmological density parameter $`\mathrm{\Omega }_o`$, to constrain the large-scale variation of the mass-to-light ratio, and to test theories of the formation and evolution of galaxies and clusters (e.g., Hoffman et al. 1982; Shaya 1984; Peebles 1986; Cen 1994). Large scale structures have been studied at low redshift via the local, Shapley, and Corona Borealis superclusters (e.g., Davis et al. 1980; Postman et al. 1988; Quintana et al. 1995; Small et al. 1998). At higher redshifts of $`z0.4`$, weak lensing studies of MS0302+16 (Kaiser et al. 2000) provide a direct measure of the projected mass distribution on 10 Mpc scales. These studies indicate that supercluster masses are $`M10^{16}10^{17}h^1\mathrm{M}_{}`$, their mass-to-light ratios are $`M/L_B200600hM_{}/L_{}`$, and the density parameter measured on supercluster scales is $`\mathrm{\Omega }_o0.10.4`$. This work can be extended to higher redshift, but until recently no such systems were known.
Such a high-redshift system has been discovered in a study of nine candidate clusters at $`z0.6`$ by Oke, Postman & Lubin (1998). Two clusters, CL1604+4304 at $`z=0.897`$ and CL1604+4321 at $`z=0.924`$, were observed as part of this survey. They are separated by $`4300\mathrm{km}\mathrm{s}^1`$ in radial velocity and by 17 arcminutes on the sky. This implies a projected separation of only $`5h^1\mathrm{Mpc}`$. We have already analyzed the spectra of a nearly complete sample of galaxies with $`R23.5`$ in a $`2.2^{^{}}\times 7.2^{^{}}`$ field centered on each cluster. The top panel in Figure 1 shows the combined velocity histogram of the 63 confirmed members in the two clusters (21 in CL1604+4304 and 42 in CL1604+4321). CL1604+4304 and CL1604+4321 have velocity dispersions of $`1226_{154}^{+245}`$ and $`935_{91}^{+126}\mathrm{km}\mathrm{s}^1`$ and masses of $`3.1`$ and $`1.6\times 10^{15}h^1\mathrm{M}_{}`$, respectively (Postman, Lubin & Oke 1998, 2000). All of the observational data suggest that CL1604+4304 and CL1604+4321 are typical of Abell richness class 1 to 3 clusters (Lubin et al. 1998, 2000).
More interestingly, these two clusters may comprise an even larger system of galaxies. The lower panel of Figure 1 shows the north-south position of the confirmed cluster members versus redshift. There is a clear trend in which the redshift on the north side of CL1604+4304 approaches the redshift of CL1604+4321. The apparent alignment in redshift space and the physical proximity of the clusters indicate that this may be a high-redshift supercluster. The estimated mass of this structure is $`5\times 10^{15}\mathrm{M}_{}`$, and the spatial overdensity is $`40`$. These numbers imply that the system is bound and has likely reached turnaround for reasonable cosmologies (Small et al. 1998).
In this Letter, we provide new evidence from multi-band optical imaging which strongly favors the supercluster hypothesis. Unless otherwise noted, we use $`H_0=100h\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$ and $`q_0=0.1`$.
## 2 The Observations
All of the optical imaging was completed with the Carnegie Observatories Spectroscopic Multislit and Imaging Camera (COSMIC; Kells et al. 1998) at the 200-in Hale telescope at Palomar Observatories. We have used the instrument in direct imaging mode which provides a pixel scale of 0$`\stackrel{}{\mathrm{.}}`$285 per pixel and a field-of-view of $`9.7^{^{}}\times 9.7^{^{}}`$. Two individual pointings were made in order to cover the region between CL1604+4304 and CL1604+4321. Each pointing covered a portion of one cluster. The overlap between the pointings was $`35^{^{\prime \prime }}`$. The photometric survey was conducted in three broadband filters $`B`$, $`R`$, and Gunn $`i`$. The total integration times on each pointing were 2 hours in $`B`$ and 1 hour in $`R`$ and $`i`$. The data were calibrated to the standard Cousins-Bessell-Landolt system through exposures of Landolt standard-star fields (Landolt 1992). Variations about the nightly photometric transformations are 0.03 mag or less.
Source detection and photometry were performed using SExtractor version 2.1.0 (Bertin & Arnouts 1996). SExtractor was chosen for its ability to detect objects in one image and analyze the corresponding pixels in a separate image. When applied uniformly to multi-band data, this technique generates a matched aperture dataset. Our detection image was constructed from the $`BRi`$ images using a $`\chi ^2`$ process (Szalay, Connolly & Szokoly 1998). Briefly, this process involves convolving each input image with a Gaussian kernel matched to the seeing. The convolved images were squared and normalized so that the background had zero mean and unit variance. The three processed images (corresponding to the original $`BRi`$ images) were coadded, forming the $`\chi ^2`$ detection image. A histogram of the pixel distribution in the $`\chi ^2`$ image was created and compared to a $`\chi ^2`$ function with three degrees of freedom (which corresponds to the sky pixel distribution). The difference between the actual pixel distribution and the $`\chi ^2`$ function provides an optimal estimate for the actual object pixel distribution. The Bayesian detection threshold was set equivalent to the intersection of the “sky” and “object” distributions (i.e. where the object pixel flux becomes dominant). To convert this empirical threshold for use with SExtractor, we scale the threshold (which is a flux per pixel value) into a surface brightness threshold (which is in magnitudes per square arcsecond) by defining a detection zeropoint using the desired detection threshold and the pixel scale. Approximately 4800 galaxies were detected in the combined fields.
For the color analysis, we use the total magnitudes as calculated by SExtractor. These magnitudes are variable-diameter aperture magnitudes measured in an elliptical aperture of major axis radius $`2\times r_k`$, where $`r_k`$ is the Kron radius (see Bertin & Arnouts 1996). Because we have used a matched aperture analysis, the total magnitudes in the three bands are measured within the same physical radius for a given galaxy. The limiting magnitudes of our survey are $`B=25.8`$, $`R=24.6`$, and $`i=23.5`$ for a $`5\sigma `$ detection.
## 3 The Results
### 3.1 The Galaxy Colors
With the multi-band imaging, we have generated photometry on a complete sample of galaxies in a contiguous area of $`10.3^{^{}}\times 18.3^{^{}}`$ or $`3.1h^1\mathrm{Mpc}\times 5.5h^1\mathrm{Mpc}`$ at the supercluster redshift of $`z0.91`$. We show the resulting $`(BR)`$ and $`(Ri)`$ color-magnitude (CM) diagrams in Figure 2. In both diagrams, we see a well-defined color-magnitude sequence which is redder than the vast majority of galaxies which comprise the field population. This red sequence of galaxies is considerably tighter in the $`(Ri`$) CM diagram where it is observed at $`(Ri)1.4`$. In the $`(BR)`$ CM diagram, the larger color scatter in this red sequence is a result of the fact that many of these galaxies lie at or beyond the completeness limits of this survey. Figure 3 shows a histogram of $`(Ri)`$ colors. In this figure, the red sequence of galaxies can be clearly distinguished from the field population where it is a red peak superimposed on the large distribution of bluer field galaxies. Fitting two Gaussian functions to this distribution, we find that the standard deviation of the red peak is 0.15 mag.
A tight, red color-magnitude relation is typical of the central regions of massive clusters both in the local universe and at intermediate and high redshift (e.g., Dressler 1980; Butcher & Oemler 1984; Stanford, Dickinson & Eisenhardt 1995, 1997). The galaxies contained in this “red locus” are the elliptical and S0 galaxies which comprise the majority of the cluster population. The early-type galaxies are characterized by their red color and their small color scatter, typically less than 0.2 mag. Studies of clusters from $`z1`$ to the present epoch imply that the observed color trend in this red envelope of galaxies is consistent with passive evolution of an old stellar population formed in a relatively synchronized burst of star formation at $`z2`$ (e.g., Ellis et al. 1997; Stanford et al. 1995, 1997; Bower, Kodama & Terlevich 1998).
We have confirmed that the red sequence observed in our supercluster field corresponds to a population of old, early-type galaxies. Within this field, there are 21 galaxies which have an absorption spectrum which is typical of an early-type galaxy and are spectroscopically-confirmed members of either CL1604+4304 at $`z=0.896`$ or CL1604+4321 at $`z=0.924`$ (Postman et al. 1998, 2000). In Figure 4, we indicate those galaxies on the $`(Ri)`$ CM diagram. All of these galaxies fall directly on the observed red locus, confirming that it is comprised mainly of early-type galaxies at the supercluster redshift of $`z0.91`$. As discussed in Postman et al. (1998, 2000), the color of these galaxies are consistent with the passive evolution of a stellar population which formed at redshifts of $`z25`$. While the presence of a red sequence of early-type galaxies is typical of the central $`0.5h^1\mathrm{Mpc}`$ in massive clusters, the fact that we observe such a well-defined red sequence over the entire $`5.5h^1\mathrm{Mpc}`$ scale of the supercluster strongly supports the existence of a large scale structure at $`z0.91`$.
### 3.2 The Spatial Distribution
Based on the $`(Ri`$) color-magnitude sequence, we can select galaxies which are likely supercluster members. We have chosen all galaxies which form the red locus ($`1.2Ri1.7`$) and can be spectroscopically observed in a reasonable time on a 10-m class telescope ($`20i23`$). The resulting sample contains 418 galaxies which are shown on the composite $`i`$ band image in Figure 5. The clusters CL1604+4304 and CL1604+4321 are at the bottom and top, respectively, of this image. The spatial distribution of red galaxies clearly delineates the large scale system of galaxies which encompasses the two rich clusters. These galaxies are spread throughout the full field, being noticeably more concentrated near the cluster centers. In addition, we observe a strong concentration of red galaxies directly between CL1604+4304 and CL1604+4321 at $`160425.7+431444.7(\mathrm{J2000})`$. Based on a control field, we find the number density of red field galaxies, as defined by our color selection, is $`0.7\pm 0.1`$ galaxies per arcmin<sup>2</sup>. Within a radius of $`0.2h^1\mathrm{Mpc}`$, the new concentration is overdense in red galaxies by a factor of $`20`$ compared to this control field. This overdensity is equivalent to that observed in the two original clusters; CL1604+4304 and CL1604+4321 are overdense by a factor of 20 and 13, respectively. At half an Abell radius, the overdensity of red galaxies in the three clusters is a factor of $`56`$. These data clearly indicate that we have detected a third, massive cluster associated with this large scale structure. We also observe at least one other, although more marginal, overdensity in this field, suggesting that this supercluster may contain additional clusters beyond the three discussed here. The spatial distribution of red galaxies provides further confirmation of a large scale structure spanning at least $`5.5h^1`$ Mpc.
## 4 Conclusions
We have confirmed the existence of a supercluster at $`z=0.91`$ with deep, multiband imaging taken at the Palomar 200-in telescope. In the resulting color–magnitude diagrams, we find a relatively tight, color–magnitude sequence of red galaxies. The characteristic color of this sequence corresponds directly to the colors of the confirmed early-type galaxies in the two member clusters, CL1604+4304 and CL1604+4321. Therefore, we have identified a large population of early-type galaxies within the redshift range of $`z0.890.93`$. These red galaxies cover the entire $`5.5h^1\mathrm{Mpc}`$ region which separates the two member clusters. They delineate the full extent of the large scale structure, while clearly encompassing the two clusters. Based on the distribution of red galaxies over this field, we have identified another rich cluster associated with this system. It lies directly between CL1604+4304 and CL1604+4321, and its overdensity of red galaxies is approximately equal to that of the two originally identified clusters. The strong red sequence of galaxies and their distribution on the sky leave no doubt that there exists a large scale structure at $`z0.91`$. This supercluster is the first massive structure confirmed at such a high redshift. At higher redshift, there is a two cluster system at $`z=1.27`$ (Rosati et al. 1999). The nature of large scale structure surrounding this pair has not yet been explored.
In order to study this structure even further, we are planning to perform two additional observational studies. Firstly, we will complete an extensive spectroscopic survey using the multislit capability of the Low Resolution Imaging Spectrograph (LRIS; Oke et al. 1995) on the Keck 10-m telescope. Based on our accurate color selection, we expect to measure redshifts for over 400 supercluster members. This sample is comparable in size to the largest spectroscopic studies of the Shapley and Corona Borealis superclusters (Quintana et al. 1995; Small et al. 1998). Secondly, we will perform multi-band imaging of the entire supercluster region, including two flanking fields, using the Large Field Camera (LFC; Metzger et al. 2000) on the Palomar 200-in telescope. This survey will provide $`u^{}g^{}r^{}i^{}z^{}`$ photometry for over 10,000 galaxies which will be used to calculate photometric redshifts accurate to $`\mathrm{\Delta }z=0.06`$ (Brunner, Connolly & Szalay 1999). These data will allow us to study galaxy properties as a function of position and local density within the supercluster; to measure the mass distribution on intermediate scales and estimate $`\mathrm{\Omega }_o`$; and to examine the early stages of cluster formation through the accretion of matter in the cluster infall regions. As a result, this high-redshift structure will be one of the most well-studied superclusters.
We thank the anonymous referee for useful comments. LML is supported by NASA through Hubble Fellowship grant HF-01095.01-97A from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. |
warning/0001/hep-th0001037.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The theme of D-geometry is an interesting approach to D-branes on the curved space. In this paper, we consider D-branes on the Calabi-Yau spaces in Type II string theory by using the boundary states. The BPS condition for the brane configurations on the Calabi-Yau spaces implies that the cycles on which D-branes wrap must be the special Lagrangian submanifolds for the middle-dimensional cycles and holomorphic for the even-dimensional cycles . These cycles correspond to the A-type and the B-type boundary states . It is well-known that Gepner model based on the $`N=2`$ minimal models describes the exactly soluble string propagation on the Calabi-Yau space at the special symmetry enhanced point in the moduli space. The Gepner point is smoothly related to the geometric Calabi-Yau phase . Recknagel and Schomerus first considered D-branes in Gepner models. They used the Cardy’s construction for the bosonic subalgebra of the $`N=2`$ SCA and directly applied this to Gepner models with the diagonal modular invariants. Subsequently, the relation to the $`N=2`$ black holes was discovered . On the other hand, the Cardy’s construction for the $`A`$-$`D`$-$`E`$ modular invariant theories has been established in . We apply this formalism to the boundary states in Gepner models in section 2. We find that the boundary states satisfy the same supersymmetric condition as the diagonal cases. We also comment on the $`K3`$ compactification. Our construction includes all the known Gepner models classified in . Then for these states, we compute the Witten index in Ramond sector which is interpreted as the geometric intersection form encoding the quantization condition . We can use this in order to count the number of the massless modes for the given configuration.
Some interesting approach was made in in order to relate the boundary states (in particular, the B-type states) to the brane configurations on the quintic Calabi-Yau in the large volume limit. The monodromy matrices found in and the intersection form were used in order to determine the charge of the boundary states which represents the geometric brane configurations. The same procedure was performed for the two-parameter Calabi-Yau threefolds with elliptic or $`K3`$ fibrations . This approach might bring some insights on the study of D-branes on the Calabi-Yau spaces, thus in section 3 we pursue this procedure for the one-parameter Calabi-Yau threefolds by using the calculation in . We calculate the charge for all the boundary states and find that there are D0-branes in some models as opposed to the quintic. We also compute the number of the moduli at the Gepner point for these states. We find that the value seems to be larger than the degrees of freedom for the deformation of the bundles. It is known that there are in some sense the nonperturbative modes which exist only at the Gepner point. Thus this calculation would include some modes which are unknown in the four-dimensional low energy effective theories.
After we had obtained the results of the present paper, we received a paper which has some overlaps with section 3.
## 2 Boundary states in Gepner models
In this section, we construct the boundary states for all the Gepner models . Then we calculate the intersection form as the Witten index in Ramond sector.
### 2.1 Boundary states
We begin with the rational CFTs with the chiral algebra $`𝒜`$ $`(=𝒜_L=𝒜_R)`$ which have a finite set $``$ of the irreducible highest weight representation $`_j`$, for $`j`$. The Hilbert space of the bulk CFT is $`=_{(j,\overline{j})\mathrm{Spec}}_j\overline{}_{\overline{j}}`$ with the multiplicity $`N_{j\overline{j}}`$ of the left and right copies of $`𝒜`$. Associated with this chiral algebra, there exist the chiral fields $`W(z)`$, $`\overline{W}(\overline{z})`$ with the spin $`s_W`$. In general, if there is an automorphism $`\mathrm{\Omega }`$ that preserves the equal time commutators of the algebra, the boundary conditions can be written as $`W(z)=\mathrm{\Omega }(\overline{W})(\overline{z})|_{z=\overline{z}}`$. The action of $`\mathrm{\Omega }`$ relates an irreducible representation $`_j`$ to another irreducible representation $`_{\omega (j)}`$. These boundary conditions are equivalent to those of the boundary states satisfying $`(W_n(1)^{s_W}\mathrm{\Omega }(\overline{W}_n))|\alpha _\mathrm{\Omega }=0`$, where $`W_n`$ and $`W_n`$ are the generators of the left- and right-moving chiral algebra $`𝒜`$. In particular, $`(L_n\overline{L}_n)|\alpha =0`$. These conditions were solved by Ishibashi for the rational CFT. The explicit solution $`|j`$ for the irreducible representation $`_j`$ of the algebra is given by $`|j=_N|j,NU\mathrm{\Omega }|\stackrel{~}{j,N}`$ where the sum is over all the descendants of $`_j`$ and $`U`$ is an anti-unitary operator which acts only on the right-moving generators as $`U\overline{W}_nU^1=(1)^{h_W}\overline{W}_n`$. There may be some multiplicity $`N_{jj}`$ for each Ishibashi states, but we omit it in order to avoid the notational complications. This boundary state $`|j`$ couples to the Hilbert space $`_j_{\omega (j)}`$. Thus it is labeled by $`=\{j|(j,\overline{j}=\omega (j))\mathrm{Spec}\}`$, where $``$ is called the set of the exponents of the theory.
The Cardy’s construction of the boundary states for the non-diagonal modular invariant $`SU(2)`$ Wess-Zumino-Witten models has been discussed in and we explain it in the following (we adopt the notation of ). Let $`G`$ be a Dynkin diagram of the $`A`$-$`D`$-$`E`$ type with the Coxeter number $`g`$ which is related to the level $`k`$ by $`g=k+2`$. To each diagram, we associate an $`n\times n`$ adjacency matrix $`G_{\alpha \beta }`$ ($`G_{\alpha \beta }=`$ \# of the links between the node $`\alpha `$ and $`\beta `$), where $`n`$ is the number of the nodes of its Dynkin diagram.
$$\begin{array}{ccc}\mathrm{Diagram}G& g=k+2& j\mathrm{Exp}(G)\\ A_n& n+1& 1,2,3,\mathrm{},n\\ D_{2n+1}& 4n& 1,3,5,\mathrm{},4n1,2n\\ D_{2n}& 4n2& 1,3,5,\mathrm{},4n3,2n1\\ E_6& 12& 1,4,5,7,8,11\\ E_7& 18& 1,5,7,9,11,13,17\\ E_8& 30& 1,7,11,13,17,19,23,29\end{array}$$
(1)
This matrix is diagonalized in the orthonormal basis $`\psi _\alpha ^j`$ which is labeled by the node $`\alpha `$ and the exponent $`j`$. The table (1) is the list of the Coxeter number and the exponents to each diagram. For the diagonal cases, Cardy found the consistent boundary states for the rational CFTs with the help of the Verlinde formula $`N_{ij}^k=_{\mathrm{}}\frac{S_i\mathrm{}S_j\mathrm{}S_k\mathrm{}^{}}{S_1\mathrm{}}`$. He required that the partition function on the strip which is periodic in the time direction (topologically an annulus) with the boundary conditions $`\alpha `$ and $`\beta `$ is equivalent, under the modular transformation $`\tau 1/\tau `$, to that on the cylinder between the boundary states $`|\alpha `$ and $`|\beta `$. For the non-diagonal cases, the boundary states and their conjugate states are given by
$$|\alpha =\underset{j\mathrm{Exp}(G)}{}\frac{\psi _\alpha ^j}{\sqrt{S_{1j}}}|j,\beta |=\underset{j\mathrm{Exp}(G)}{}j|\frac{(\psi _\beta ^j)^{}}{\sqrt{S_{1j}}},$$
(2)
where the set $`\{\alpha \}`$ and $`\{\beta \}`$ label the boundary states which denote the nodes of a Dynkin diagram $`G`$. The inner products of the Ishibashi states are defined by $`j^{}|\stackrel{~}{q}^{\frac{1}{2}(L_0+\overline{L}_0\frac{c}{12})}|j=\delta _{jj^{}}\chi _j(\stackrel{~}{q})`$, where $`\stackrel{~}{q}=e^{2\pi i/\tau }`$ and $`\chi _j`$ is the character $`\chi _j(\stackrel{~}{q})=\mathrm{Tr}__j\stackrel{~}{q}^{L_0\frac{c}{24}}`$. Then the partition function on the cylinder becomes $`Z_{\alpha \beta }=\beta |q^{\frac{1}{2}(L_0+\overline{L}_0\frac{c}{12})}|\alpha =_{j\mathrm{Exp}(G)}\psi _\alpha ^j(\psi _\beta ^j)^{}\frac{\chi _j(\stackrel{~}{q})}{S_{1j}}`$. On the other hand, the partition function as a periodic time evolution on the strip with the boundary conditions $`\alpha `$ and $`\beta `$ becomes $`Z_{\alpha \beta }=_in_{\alpha \beta }^i\chi _i(q)`$, where $`q=e^{2\pi i\tau }`$ and $`1ig`$. Under the S-transformation $`\chi _j(\stackrel{~}{q})=_iS_{ji}\chi _i(q)`$ (and $`S=S^t`$), we obtain the Cardy’s equation
$$n_{\alpha \beta }^i=\underset{j\mathrm{Exp}(G)}{}\frac{S_{ij}}{S_{1j}}\psi _\alpha ^j(\psi _\beta ^j)^{}.$$
(3)
The fused adjacency matrices which are defined by $`(V_i)_\alpha ^\beta =n_{\alpha \beta }^i`$ satisfy the fusion algebra $`V_iV_j=_kN_{ij}^kV_k`$ with the fusion coefficients $`N_{ij}^k`$. This algebra determines the numerical values of $`n_{\alpha \beta }^i`$ recursively by $`V_1=I`$, $`V_2=G`$, $`V_i=V_2V_{i1}V_{i2}`$ for $`3ig`$ and we can see $`V_g=0`$.
Next we turn to Gepner models . Gepner used the so-called “$`\beta `$ method” to construct the supersymmetric Type II string compactifications to $`d+2`$ dimensions by using the tensor product of the $`r`$ $`N=2`$ minimal models at the levels $`k_j(j=1,\mathrm{},r)`$ with the internal central charge $`c=12\frac{3}{2}d=_{j=1}^r\frac{3k_j}{k_j+2}`$. In each $`N=2`$ minimal model, the primary fields are labeled by the three integers $`(\mathrm{},m,s)`$. The standard range of $`(\mathrm{},m,s)`$ is $`0\mathrm{}k`$, $`|ms|\mathrm{}`$, $`s\{1,0,1\}`$ and $`\mathrm{}+m+s2𝐙`$. The representations with $`s=0`$ belong to the NS sector, while those with $`s=\pm 1`$ to the Ramond sector. The conformal dimension $`h`$ and the charge $`q`$ of the primary fields $`\mathrm{\Phi }_{m,s}^{\mathrm{}}`$ are given by $`h_{m,s}^{\mathrm{}}=\frac{\mathrm{}(\mathrm{}+2)m^2}{4(k+2)}+\frac{s^2}{8}(\mathrm{mod}\mathrm{\hspace{0.17em}1}),q_{m,s}^{\mathrm{}}=\frac{m}{k+2}\frac{s}{2}(\mathrm{mod}\mathrm{\hspace{0.17em}2})`$. The chiral primary state and the anti-chiral primary state are labeled by $`(\mathrm{},\pm \mathrm{},0)`$ in the NS sector and related to the Ramond ground states $`(\mathrm{},\pm \mathrm{},\pm 1)`$ by the spectral flow. These minimal models are invariant under the $`𝐙_n\times 𝐙_2`$ discrete symmetry group with $`n=k+2`$ for $`A_n`$, $`D_{2n+1}`$, $`E_6`$ and $`n=(k+2)/2`$ for $`D_{2n}`$, $`E_7`$, $`E_8`$ . Their actions are $`g\mathrm{\Phi }_{m,s}^{\mathrm{}}=e^{2\pi i\frac{m}{n}}\mathrm{\Phi }_{m,s}^{\mathrm{}}`$, $`h\mathrm{\Phi }_{m,s}^{\mathrm{}}=e^{2\pi i\frac{s}{2}}\mathrm{\Phi }_{m,s}^{\mathrm{}}`$. This $`𝐙_2`$ symmetry is essentially the charge conjugation. Then, let us introduce the vectors $`\lambda =(\mathrm{}_1,\mathrm{},\mathrm{}_r)\mathrm{and}\mu =(s_0;m_1,\mathrm{},m_r;s_1,\mathrm{},s_r)`$ and the $`2r+1`$ dimensional vectors $`\beta _0=(1;1,\mathrm{},1;1,\mathrm{},1)`$, and $`\beta _j=(2;0,\mathrm{},0;0,\mathrm{},0,2,0,\mathrm{},0)`$ for $`j=1,\mathrm{},r`$ which has 2 in the first and $`(r+1+j)`$-th entries and is zero everywhere else. We use $`s_0\{1,0,1,2\}`$ in order to characterize the irreducible ($`s,o,c,v`$) representations of the $`SO(d)`$ current algebra generated by the $`d`$ external fermions. Then, define the inner product among the $`2r+1`$ dimensional vectors as $`\mu \stackrel{~}{\mu }=\frac{d}{8}s_0\stackrel{~}{s_0}_{j=1}^r\frac{s_j\stackrel{~}{s_j}}{4}+_{j=1}^r\frac{m_j\stackrel{~}{m_j}}{2(k_j+2)}`$. The $`\beta `$-constraints are realized as $`q_{\mathrm{tot}}=2\beta _0\mu 2𝐙+1,\beta _j\mu 𝐙`$. These are consistent only for $`d=2,6`$. For $`d=4`$, we have to replace $`d`$ with $`d+2`$ in the inner product in order to impose the consistent conditions. Note that the massless fields satisfy $`2\beta _0\mu =\pm 1`$. Using all the ingredients, we obtain the partition function in Gepner models describing a superstring compactified to $`d+2`$ dimensions as follows:
$$Z=\frac{1}{2}\frac{(\mathrm{Im}\tau )^d}{|\eta (q)|^{2d}}\underset{b_0=0}{\overset{2K1}{}}\underset{b_1,\mathrm{},b_r=0,1}{}\underset{\lambda ,\mu }{\overset{\mathrm{ev},\beta }{}}(1)^{b_0}M_{\lambda \lambda ^{}}\chi _\mu ^\lambda (q)\chi _{\mu +b_0\beta _0+_{j=1}^rb_j\beta _j}^\lambda ^{}(\overline{q}),$$
(4)
where $`K=\mathrm{lcm}(2,k_j+2)`$ and the character is $`\chi _\mu ^\lambda (q)=\chi _{s_0}(q)\chi _{m_1s_1}^\mathrm{}_1(q)\mathrm{}\chi _{m_rs_r}^\mathrm{}_r(q)`$. We denote by $`M_{\lambda \lambda ^{}}`$ the products of the Cappelli-Itzykson-Zuber matrices . The summation $`^{\mathrm{ev},\beta }`$ is taken that the indices satisfy the $`\beta `$-constraints and $`\mathrm{}_j+m_j+s_j2𝐙`$. We can check that eq.(4) is invariant under the S-transformation:
$`S_{s_0,s_0^{}}^\mathrm{f}`$ $`=`$ $`{\displaystyle \frac{1}{2}}e^{i\pi \frac{d}{2}\frac{s_0s_0^{}}{2}},`$ (5)
$`S_{(\mathrm{},m,s),(\mathrm{}^{},m^{},s^{})}^k`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}(k+2)}}\mathrm{sin}\pi {\displaystyle \frac{(\mathrm{}+1)(\mathrm{}^{}+1)}{k+2}}e^{i\pi \frac{mm^{}}{k+2}}e^{i\pi \frac{ss^{}}{2}}.`$ (6)
Note that eq.(5) is correct only for $`d=2,6`$ and we have to replace $`d`$ with $`d+2`$ in the case of $`d=4`$. For the $`D_{2n}`$, $`E_7`$, and $`E_8`$ case, the field identification exist independently for the holomorphic and the anti-holomorphic part, i.e., $`\mathrm{\Phi }_{m,s;\overline{m},\overline{s}}^{\mathrm{};\overline{\mathrm{}}}=\mathrm{\Phi }_{m+k+2,s+2;\overline{m},\overline{s}}^{k\mathrm{};\overline{\mathrm{}}}`$ and this amounts to the additional factor $`1/2`$ in front of the partition function . In the $`D_{2n}`$ case, this factor is canceled for the particular values $`\mathrm{}=\overline{\mathrm{}}=k/2`$. For all the cases, the discrete symmetry and the Hodge number for Gepner models agree with those of the geometric Calabi-Yau manifolds .
Then we construct the boundary states in Gepner models. In this case, the algebra $`𝒜`$ is generated by the $`N=2`$ SCA. There are the two boundary conditions for the $`U(1)`$ current $`J`$ and the superconformal generators $`G^\pm `$ which preserve half of the spacetime supersymmetries . They are called the A-type and the B-type boundary conditions. The A-type boundary states corresponding to D-branes wrapping on the middle-dimensional cycles satisfy
$$(J_n\overline{J}_n)|B=0,(G_r^++i\eta \overline{G}_r^{})|B=0,(G_r^{}+i\eta \overline{G}_r^+)|B=0,$$
(7)
and the B-type boundary states corresponding to D-branes wrapping on the even-dimensional cycles satisfy
$$(J_n+\overline{J}_n)|B=0,(G_r^++i\eta \overline{G}_r^+)|B=0,(G_r^{}+i\eta \overline{G}_r^{})|B=0.$$
(8)
The choice of $`\eta =\pm 1`$ corresponds to the choice of the spin structure. The A-type boundary states satisfy $`q=\overline{q}`$ and the B-type boundary states satisfy $`q=\overline{q}`$. Due to the property of the anti-unitary operator $`U`$ and the mirror automorphism $`\mathrm{\Omega }`$ , the Ishibashi states $`|j=_N|j,NU\stackrel{~}{|j,N}`$ satisfy the B-type boundary conditions, while the states $`|j=_N|j,NU\mathrm{\Omega }\stackrel{~}{|j,N}`$ satisfy the A-type boundary conditions, where the Ishibashi states are labeled by the quantum number for the $`N=2`$ primaries : $`(\mathrm{},m,s)`$. One should note that they label the irreducible representations of the bosonic sub-algebra of the $`N=2`$ algebras rather than the full $`N=2`$ representations. Before we describe the boundary states in Gepner models, it is convenient to change the notation of $`\mathrm{}`$ into $`\mathrm{}+1\mathrm{Exp}(G)`$ in order to fit the range of $`\mathrm{}`$ to start from zero. Then let $`|\lambda ,\mu _A`$ ($`|\lambda ,\mu _B`$) be the tensor products of $`r`$ A-type (B-type) Ishibashi states and the external part. Following the procedure of , we write down the boundary states in Gepner models as
$$|\alpha _\mathrm{\Omega }|S_0;(L_j,M_j,S_j)_{j=1}^r_\mathrm{\Omega }=\frac{1}{\kappa _\alpha ^\mathrm{\Omega }}\underset{\lambda +1\mathrm{Exp}(G),\mu }{\overset{\beta }{}}\delta _\mathrm{\Omega }B_\alpha ^{\lambda ,\mu }|\lambda ,\mu _\mathrm{\Omega },$$
(9)
where $`S_0,L_j,M_j,S_j`$ are integers and $`\mathrm{}_j+1\mathrm{Exp}(G_j)`$ for $`1jr`$, and the normalization constant $`\kappa _\alpha ^\mathrm{\Omega }`$ is determined later. The explicit form of $`B_\alpha ^{\lambda ,\mu }`$ is
$$B_\alpha ^{\lambda ,\mu }=(1)^{\frac{s_0^2}{2}}e^{i\pi \frac{d}{2}\frac{s_0S_0}{2}}\underset{j=1}{\overset{r}{}}\sqrt{\frac{k_j+2}{2}}\frac{\psi _{L_j}^\mathrm{}_j}{\mathrm{sin}^{\frac{1}{2}}\pi \frac{\mathrm{}_j+1}{k_j+2}}e^{i\pi \frac{m_jM_j}{k_j+2}}e^{i\pi \frac{s_jS_j}{2}}.$$
(10)
The symbol $`\delta _\mathrm{\Omega }`$ ($`\mathrm{\Omega }=A,B`$) denotes the constraint which the Ishibashi state $`|\lambda ,\mu _\mathrm{\Omega }`$ must appear in the partition function (4) . For the A-type boundary states, this requires no constraint. For the B-type boundary states, all the $`m_j`$ are the same modulo $`k_j+2`$.
We can evaluate the partition functions on the cylinder by using eq.(3) and the identification $`\chi _{m,s}^{\mathrm{}}(q)=\chi _{m+k+2,s+2}^k\mathrm{}(q)`$. The result for the A-type boundary states is
$`Z_{\alpha \stackrel{~}{\alpha }}^A(q)`$ $`=`$ $`{\displaystyle \frac{1}{C_A}}{\displaystyle \underset{\lambda ^{},\mu ^{}}{\overset{\mathrm{ev}}{}}}{\displaystyle \underset{\nu _0=0}{\overset{2K1}{}}}{\displaystyle \underset{\nu _1,\mathrm{},\nu _r=0,1}{}}(1)^{S_0\stackrel{~}{S_0}+s_0^{}}\delta _{S_0\stackrel{~}{S_0}+s_0^{}+\nu _0+22_{j=1}^r\nu _j}^{(4)}`$ (11)
$`\times `$ $`{\displaystyle \underset{j=1}{\overset{r}{}}}n_{L_j\stackrel{~}{L_j}}^\mathrm{}_j^{}\delta _{M_j\stackrel{~}{M_j}+m_j^{}+\nu _0}^{(2k_j+4)}\delta _{S_j\stackrel{~}{S_j}+s_j^{}+\nu _02\nu _j}^{(4)}\chi _\mu ^{}^\lambda ^{}(q),`$
and the result for the B-type boundary states is
$`Z_{\alpha \stackrel{~}{\alpha }}^B(q)`$ $`=`$ $`{\displaystyle \frac{1}{C_B}}{\displaystyle \underset{\lambda ^{},\mu ^{}}{\overset{\mathrm{ev}}{}}}{\displaystyle \underset{\nu _0=0}{\overset{2K1}{}}}{\displaystyle \underset{\nu _1,\mathrm{},\nu _r=0,1}{}}(1)^{S_0\stackrel{~}{S_0}+s_0^{}}\delta _{S_0\stackrel{~}{S_0}+s_0^{}+\nu _0+22_{j=1}^r\nu _j}^{(4)}`$ (12)
$`\times `$ $`\delta _{\frac{M\stackrel{~}{M}}{2}+_{j=1}^r\frac{K^{}(m_j^{}+\nu _0)}{2k_j+4}}^{(K^{})}{\displaystyle \underset{j=1}{\overset{r}{}}}n_{L_j\stackrel{~}{L_j}}^\mathrm{}_j^{}\delta _{M_j\stackrel{~}{M_j}+m_j^{}+\nu _0}^{(2)}\delta _{S_j\stackrel{~}{S_j}+s_j^{}+\nu _02\nu _j}^{(4)}\chi _\mu ^{}^\lambda ^{}(q),`$
where $``$ means that $`\mathrm{}_j^{}`$ runs from 0 to $`k_j`$; and $`K^{}=\mathrm{lcm}(k_j+2)`$, $`M=_{j=1}^r\frac{K^{}M_j}{k_j+2}`$. The overall factors in eq. (11), (12) are expressed as
$$\frac{1}{C_A}=\frac{2^{\frac{r}{2}}(k_j+2)}{K\kappa _\alpha ^A\kappa _{\stackrel{~}{\alpha }}^A},\frac{1}{C_B}=\frac{2^{\frac{r}{2}}}{\kappa _\alpha ^B\kappa _{\stackrel{~}{\alpha }}^B},$$
(13)
and we choose $`\kappa _\alpha ^A`$ and $`\kappa _\alpha ^B`$ to satisfy the Cardy’s condition. We find that for the non-diagonal cases, the supersymmetric conditions among two boundary states take the same form
$`S_0\stackrel{~}{S}_0`$ $``$ $`S_j\stackrel{~}{S}_j(\mathrm{mod}\mathrm{\hspace{0.33em}2})\mathrm{for}\mathrm{all}j=1,\mathrm{},r,`$ (14)
$`Q(\alpha \stackrel{~}{\alpha })`$ $`=`$ $`{\displaystyle \frac{d}{2}}{\displaystyle \frac{S_0\stackrel{~}{S}_0}{2}}{\displaystyle \underset{j=1}{\overset{r}{}}}{\displaystyle \frac{S_j\stackrel{~}{S}_j}{2}}+{\displaystyle \underset{j=1}{\overset{r}{}}}{\displaystyle \frac{M_j\stackrel{~}{M}_j}{k_j+2}}2𝐙,`$ (15)
as the diagonal cases. This condition guarantees that there exists no tachyon in the open string spectrum.
Next we discuss the $`K3`$ compactification <sup>3</sup><sup>3</sup>3In this case, the different method by using the spectral flow invariant orbits was proposed in .. It turns out that the construction is straightforward. The difference essentially lies on $`d4𝐙`$. Some parts in the $`\beta `$-constraints and the S-modular transformation of the external part (5) depend on $`d`$. Thus the boundary states in Gepner models for the $`K3`$ can be obtained only by replacing $`d`$ with $`d+2`$ and particularly substitute $`d=4`$ in the formula (10), (15). With all the things considered, we have constructed the boundary states in all the Gepner models classified in <sup>4</sup><sup>4</sup>4There are some Gepner models in Type I string theory . We thank A. Sagnotti for pointing out this. .
It is convenient to omit the external fermion index $`s_0`$ in order to consider only the internal CFT . In this case, we adopt the ansatz for the modular invariant A-type and B-type boundary states with respect to the internal part to have the same form as eq.(9) but without the external labels $`s_0,S_0`$, i.e. $`|\alpha _\mathrm{\Omega }=|(L_j,M_j,S_j)_{j=1}^r_\mathrm{\Omega }`$. Then we set the coefficients $`B_\alpha ^{\lambda ,\mu }`$ to be
$$B_\alpha ^{\lambda ,\mu }=\underset{j=1}{\overset{r}{}}\frac{1}{\sqrt{2\sqrt{2}}}\frac{\psi _{L_j}^\mathrm{}_j}{\mathrm{sin}^{\frac{1}{2}}\pi \frac{\mathrm{}_j+1}{k_j+2}}e^{i\pi \frac{m_jM_j}{k_j+2}}e^{i\pi \frac{s_jS_j}{2}},$$
(16)
and we have to set $`\mu `$, $`\beta _0`$ and $`\beta _i`$ to be the all $`2r`$-dimensional vectors without the external indices (the inner product should be trivially modified). Also in this case, the $`\beta `$-constraints mean that the internal $`U(1)`$ charge is not necessarily odd-integer. Then the construction is essentially the same. We find that the condition that the two boundary states $`|\alpha `$ and $`|\stackrel{~}{\alpha }`$, with the same external part, preserve the same supersymmetries is
$$Q(\alpha \stackrel{~}{\alpha }):=\frac{S\stackrel{~}{S}}{2}+\frac{M\stackrel{~}{M}}{K^{}}2𝐙,$$
(17)
where $`S=_{j=1}^rS_j`$ and $`M=_{j=1}^r\frac{K^{}M_j}{k_j+2}`$.
Before closing this subsection, we discuss the specific aspects about the diagonal cases to use in section 3 <sup>5</sup><sup>5</sup>5Of course, we can make the case-by-case comments for the other cases. . First, $`\psi _{L_j}^\mathrm{}_j`$ in eqs.(10),(16) is simply the modular S-matrix. From the Verlinde formula, $`n_{L\stackrel{~}{L}}^{\mathrm{}}`$ in eqs.(11),(12) becomes $`N_{L\stackrel{~}{L}}^{\mathrm{}}`$ which is the $`SU(2)_k`$ fusion rule coefficient : $`N_{L\stackrel{~}{L}}^{\mathrm{}}=1\mathrm{for}|L\stackrel{~}{L}|\mathrm{}\mathrm{min}\{L+\stackrel{~}{L},2kL\stackrel{~}{L}\}\mathrm{and}\mathrm{}+L+\stackrel{~}{L}2𝐙`$ and otherwise $`N_{L\stackrel{~}{L}}^{\mathrm{}}=0`$. Then it follows from eqs.(11),(12) that the integers $`(L_j,M_j,S_j)`$ satisfy the condition $`(\stackrel{~}{L}_jL_j)+(\stackrel{~}{M}_jM_j)+(\stackrel{~}{S}_jS_j)2𝐙`$ and the identification $`(L_j,M_j,S_j)(k_jL_j,M_j+k_j+2,S_j+2)`$. Second, the boundary states have the $`𝐙_{k_j+2}`$, $`𝐙_2`$ symmetries $`g_j:M_jM_j+2`$, $`h_j:S_jS_j+2`$ inherited from the discrete symmetry in the $`k_j`$-th minimal models. Because of the $`\beta `$-constraints, the physically inequivalent choices for $`S_j`$ are $`S=0,2`$ $`(\mathrm{mod}\mathrm{\hspace{0.33em}4})`$. Thus the A-type boundary states are labeled by $`|\{L_j\};\{M_j\};S`$ and their discrete symmetries are $`g_j`$’s satisfying $`_{j=1}^rg_j=1`$. For the B-type boundary states, in addition to $`S_j`$, the physically inequivalent choices for $`M_j`$ are also restricted and they are labeled by $`M=_{j=1}^r\frac{K^{}M_j}{k_j+2}`$. Then each discrete symmetry $`g_j`$ becomes $`g_j=g^{K^{}/(k_j+2)}`$ for $`g𝐙_K^{}`$ and the B-type boundary states are labeled by
$$g^{M/2}h^{S/2}|\{L_i\}=|\{L_i\};M;S.$$
(18)
We adopt the convention that $`(L_j,M_j,S_j)`$ satisfy $`L_j+M+S2𝐙`$.
### 2.2 The intersection form
The CFT version of the intersection form in the classical geometry should be calculated as the Witten index $`I_\mathrm{\Omega }=\mathrm{Tr}_R(1)^F`$ in the Ramond sector in the open string channel . When D-branes give rise to the particles in the macroscopic directions, this number is equivalent to the Dirac-Schwinger-Zwanziger symplectic inner product on their charges. This quantization condition was also proposed in .
The Witten index corresponds to the amplitude between the RR boundary states with a $`(1)^{F_L}`$ inserted in the closed string channel. Only the Ramond ground states $`\varphi _{\mathrm{}+1,1}^{\mathrm{}}\varphi _{k+\mathrm{}1,1}^k\mathrm{}`$ contribute to the Witten index, thus we can evaluate it explicitly. Due to the Verlinde formula and the S-matrices, we can extend the range of the superscript $`\mathrm{}`$ of $`n_{L\stackrel{~}{L}}^{\mathrm{}}`$ so that $`\mathrm{}`$ has a period of $`2k+4`$. Then we identify $`n_{L\stackrel{~}{L}}^\mathrm{}2=n_{L\stackrel{~}{L}}^{\mathrm{}}`$ and set $`n_{L\stackrel{~}{L}}^1=n_{L\stackrel{~}{L}}^{k+1}=0`$. In the following, we present the results for $`d=2,6`$, but we have only to interchange the conditions on $`\frac{d}{2}+r`$ in order to obtain the results for $`d=4`$.
The result for the A-type boundary states is given by
$$I_A=\frac{1}{\stackrel{~}{C}_A}(1)^{\frac{S\stackrel{~}{S}}{2}}\underset{\nu _0=0}{\overset{K1}{}}(1)^{(\frac{d}{2}+r)\nu _0}\underset{j=1}{\overset{r}{}}n_{L_j\stackrel{~}{L}_j}^{2\nu _0+M_j\stackrel{~}{M}_j}.$$
(19)
Then we consider the B-type boundary states. When $`\frac{d}{2}+r`$ is even, the result is given by
$$I_B=\frac{1}{\stackrel{~}{C}_B}(1)^{\frac{S\stackrel{~}{S}}{2}}\underset{m_j^{}=0}{\overset{2k_j+3}{}}\delta _{\frac{M\stackrel{~}{M}}{2}+{\scriptscriptstyle {\scriptscriptstyle \frac{K^{}(m_j^{}+1)}{2k_j+4}}}}^{(K^{})}\underset{j=1}{\overset{r}{}}n_{L_j\stackrel{~}{L}_j}^{m_j^{}1},$$
(20)
and when $`\frac{d}{2}+r`$ is odd, it becomes
$`I_B`$ $`=`$ $`{\displaystyle \frac{1}{\stackrel{~}{C}_B}}(1)^{\frac{S\stackrel{~}{S}}{2}}{\displaystyle \underset{m_j^{}=0}{\overset{2k_j+3}{}}}(1)^{\frac{M\stackrel{~}{M}}{K^{}}+{\scriptscriptstyle {\scriptscriptstyle \frac{(m_j^{}+1)}{k_j+2}}}}`$ (21)
$`\times `$ $`{\displaystyle \frac{1}{2}}\delta _{\frac{M\stackrel{~}{M}}{2}+{\scriptscriptstyle {\scriptscriptstyle \frac{K^{}(m_j^{}+1)}{2k_j+4}}}}^{(K^{}/2)}{\displaystyle \underset{j=1}{\overset{r}{}}}n_{L_j\stackrel{~}{L}_j}^{m_j^{}1}.`$
We choose the normalization $`\stackrel{~}{C}_A=\kappa _\alpha ^A\kappa _{\stackrel{~}{\alpha }}^AK`$ and $`\stackrel{~}{C}_B=\kappa _\alpha ^B\kappa _{\stackrel{~}{\alpha }}^B_{j=1}^r(k_j+2)/K`$ to satisfy the Cardy’s condition on the partition functions.
The $`𝐙_2`$ action $`SS+2`$ changes the orientation of D-branes. The results for B-type boundary states depend only on $`M\stackrel{~}{M}`$ expected from the discrete symmetry.
We can count the number of the moduli of these boundary states by using the intersection form. The procedure was explained in . We can deal with all the $`A`$-$`D`$-$`E`$ cases but we deal only with the diagonal case of the B-type boundary states for the later use.
In the diagonal case, $`n_{L\stackrel{~}{L}}^{\mathrm{}}`$ becomes the $`SU(2)_k`$ fusion coefficient $`N_{L\stackrel{~}{L}}^{\mathrm{}}`$ and the intersection form can be easily rewritten in terms of the generator of the discrete symmetry. This fusion coefficient is expressed as
$$𝐧_{L\stackrel{~}{L}}=g^{\frac{|L\stackrel{~}{L}|}{2}}+g^{\frac{|L\stackrel{~}{L}|}{2}+1}+\mathrm{}+g^{\frac{L+\stackrel{~}{L}}{2}}g^{1\frac{|L\stackrel{~}{L}|}{2}}\mathrm{}g^{1\frac{L+\stackrel{~}{L}}{2}},$$
(22)
or we can write $`𝐧_{L\stackrel{~}{L}}=t_L𝐧_{00}t_{\stackrel{~}{L}}^t`$ in terms of the linear transformation $`t_L=t_L^t=_{l=L/2}^{L/2}g^l`$ and $`𝐧_{00}=(1g^1)`$. When $`\frac{d}{2}+r`$ is even, the intersection form for the B-type boundary states is given by
$$I_B=\underset{j=1}{\overset{r}{}}𝐧_{L_j\stackrel{~}{L}_j}.$$
(23)
When $`\frac{d}{2}+r`$ is odd, we can not use the formula (23) naively because the period of $`M`$ is not $`2K^{}`$ but $`K^{}`$ from eq.(21). Thus we have to multiply the additional factor $`\frac{1}{2}(1g^{K^{}/2})`$, which comes from the trivial (0-th) minimal model in order to use the above expression . Then the number of the moduli is given by the diagonal part of
$$\frac{1}{2}\underset{j=1}{\overset{r}{}}\stackrel{~}{𝐧}_{L_j\stackrel{~}{L}_j}Pv,$$
(24)
where $`\stackrel{~}{𝐧}_{L\stackrel{~}{L}}=|𝐧_{L\stackrel{~}{L}}|`$ and $`P`$ is the matrix which comes from the trivial factor, $`P=1`$ when $`\frac{d}{2}+r`$ is even and $`P=\frac{1}{2}(1+g^{K^{}/2})`$ when $`\frac{d}{2}+r`$ is odd. We denote by $`v`$ the number of the vacuum states. When $`\frac{d}{2}+r`$ is odd, $`v=2^\nu `$ where $`\nu `$ is the number of $`L_j`$’s equal to $`k_j/2`$. When $`\frac{d}{2}+r`$ is even, $`v=2^{\nu 1}`$ for $`\nu >0`$ and $`v=1`$ for $`\nu =0`$.
## 3 Geometric interpretation of the boundary states
The boundary states constructed in the previous section represent D-branes in the stringy scale, therefore it is interesting to relate them with D-brane configurations in the large volume limit . We consider this procedure for the B-type D-branes on the one-parameter Calabi-Yau threefolds $`X`$ corresponding to the $`4^41,6^4,8^33`$ Gepner models with the diagonal modular invariants which we call $`k=6,8,10`$, respectively as in . They are given by
$`W_0^{k=6}`$ $`=`$ $`z_1^6+z_2^6+z_3^6+z_4^6+2z_5^3=0,`$
$`W_0^{k=8}`$ $`=`$ $`z_1^8+z_2^8+z_3^8+z_4^8+4z_5^2=0,`$ (25)
$`W_0^{k=10}`$ $`=`$ $`z_1^{10}+z_2^{10}+z_3^{10}+2z_4^5+5z_5^2=0,`$
in $`\mathrm{𝐖𝐂𝐏}_{(\nu _1,\mathrm{},\nu _5)}`$ where $`\nu _i=k/n_i`$ ($`n_i`$ appear in the form $`z_i^{n_i}`$) and $`h_{2,1}=103,149,145`$ for $`k=6,8,10`$, respectively. The Kähler moduli space and the prepotential were considered in following which computed the periods of the 3-cycles on the mirror $`\widehat{X}`$ and the mirror map. In order to obtain the mirror $`\widehat{X}`$, consider the polynomial
$$W=W_0k\psi z_1z_2z_3z_4z_5=0,$$
(26)
and then orbifoldize the resulting algebraic hypersurface by $`G=𝐙_3\times 𝐙_6^2,𝐙_8^2\times 𝐙_2,𝐙_{10}^2`$ for $`k=6,8,10`$, respectively . The complex structure moduli space of $`\widehat{X}`$ is a Riemann sphere with the three singularities and parameterized by $`\psi ^k`$. Basically, the BPS states wrapped on 3-cycles $`Q_i[\mathrm{\Sigma }_i]`$ in $`\widehat{X}`$ are characterized by the central charge $`Z=Q_i\mathrm{\Pi }_i`$ where the periods are given by $`\mathrm{\Pi }_i=_{\mathrm{\Sigma }_i}\mathrm{\Omega }`$ ($`\mathrm{\Omega }`$ is the holomorphic three-form in $`\widehat{X}`$). Each singularity in the moduli space gives the monodromy which characterizes the period.
The procedure in consists of the following steps:
(i) Interpret the boundary states at the Gepner point as the generic BPS states in the moduli space.
(ii) Use the monodromy matrices in order to relate the Gepner point and the large volume limit.
(iii) Translate the BPS charge vectors into the microscopic topological charges using the Chern-Simons couplings.
This procedure involves the analytic continuation between the two distinct regions in the moduli space, therefore the spectrum of the BPS states would be affected by the jumping and the marginal stability phenomena. But we are not able to treat the dynamical aspects such as the stability and the existence of bound states here. The precise form of this comparison depends on the choice of the path in the Kähler moduli space which goes from the Gepner point to the large volume limit through the complex moduli space of the mirror. This is characterized by the flat $`Sp(4,𝐙)`$ connection provided by the special geometry.
We begin with the step (ii). At the origin ($`\psi ^k=0`$), the modes has an additional $`𝐙_k`$ global symmetry. This is an orbifold singularity in the moduli space. The Gepner point corresponds to this point in the Kähler moduli space of $`X`$. Let $`A_G`$ be the monodromy matrix induced by $`\psi \alpha \psi `$ around $`\psi ^k=0`$ where $`\alpha =e^{2\pi i/k}`$. This matrix satisfies $`A_G^k=1`$. If we choose a solution $`\varpi _0(\psi )`$ of Picard-Fuchs equations analytic near $`\psi ^k=0`$, the set of solutions $`\varpi _i(\psi )=\varpi _0(\alpha ^i\psi )`$ will provide a basis of the period at the Gepner point
$$\varpi =\frac{(2\pi i)^3}{\mathrm{Ord}G}(\varpi _2,\varpi _1,\varpi _0,\varpi _{k1})^t,$$
(27)
where $`\mathrm{Ord}G=36^2,28^2,10^2`$ with the constraints $`\varpi _0+\varpi _2+\varpi _4=0,\varpi _1+\varpi _3+\varpi _5=0;\varpi _i+\varpi _{i+4}=0(i=0,1,2,3);\varpi _0+\varpi _2+\varpi _4+\varpi _6+\varpi _8=0,\varpi _i+\varpi _{i+5}=0(i=0,1,2,3,4)`$<sup>6</sup><sup>6</sup>6There is a typo in for $`k=6,8,10`$, respectively. Then let us consider the large volume limit which is mirror to the large complex structure limit ($`\psi ^k\mathrm{}`$). There we obtain $`(\gamma \psi )^ke^{2\pi i(B+iJ)}`$ where $`\gamma =k_{i=1}^5\nu _i^{\nu _i/k}`$, and $`B`$ is the NS $`B`$-field flux around the 2-cycle forming a basis of $`H_2(X)`$; and $`J`$ is the size of that 2-cycle. The large volume basis is determined by the asymptotics $`\psi ^k\mathrm{}`$ as in ,
$$=\left(\begin{array}{c}_6\\ _4\\ _2\\ _0\end{array}\right)=\left(\begin{array}{c}_2\\ _1\\ w_1\\ w_2\end{array}\right)w_2\left(\begin{array}{c}\frac{\kappa }{6}(B+iJ)^3\\ \frac{\kappa }{2}(B+iJ)^2\\ B+iJ\\ 1\end{array}\right),$$
(28)
where $`\kappa =3,2,1`$ for $`k=6,8,10`$, respectively. The coefficients give the classical volumes of the even-dimensional cycles. The central charge corresponding to an integral vector $`Q=(Q_6,Q_4,Q_2,Q_0)`$ is
$$Z=Q_6_6+Q_4_4+Q_2_2+Q_0_0.$$
(29)
We can relate $`\varpi `$ to $``$, i.e. $`Z=Q=(Q_GM^1)(M\varpi )`$. This change of the basis by $`M`$ is $`=M\varpi ,Q=Q_GM^1`$. The $`𝐙_k`$ monodromy matrix in the large volume limit is given by $`A=MA_GM^1`$. We obtain $`M`$’s as follows
$$M=\left(\begin{array}{cccc}0& 1& 1& 0\\ 1& 0& 3& 2\\ \frac{1}{3}& \frac{1}{3}& \frac{1}{3}& \frac{1}{3}\\ 0& 0& 1& 0\end{array}\right),\left(\begin{array}{cccc}0& 1& 1& 0\\ 1& 0& 3& 2\\ \frac{1}{2}& \frac{1}{2}& \frac{1}{2}& \frac{1}{2}\\ 0& 0& 1& 0\end{array}\right),\left(\begin{array}{cccc}0& 1& 1& 0\\ 0& 1& 1& 1\\ 1& 0& 0& 1\\ 0& 0& 1& 0\end{array}\right),$$
(30)
for $`k=6,8,10`$, respectively <sup>7</sup><sup>7</sup>7Of course, there is an undetermined $`Sp(4,𝐙)`$ ambiguity. We can resolve this ambiguity by the following criteria. At the conifold point ($`\psi ^k=1`$), the wrapped D3-brane becomes massless . We choose the state which is the mirror to the conifold point to be a D6-brane with the trivial gauge bundle .. We also obtain $`A`$’s as follows
$$A=\left(\begin{array}{cccc}3& 1& 6& 4\\ 3& 1& 3& 3\\ 1& 0& 1& 1\\ 1& 0& 0& 1\end{array}\right),\left(\begin{array}{cccc}3& 1& 4& 4\\ 2& 1& 2& 2\\ 1& 0& 1& 1\\ 1& 0& 0& 1\end{array}\right),\left(\begin{array}{cccc}2& 1& 1& 3\\ 0& 1& 1& 0\\ 1& 0& 1& 1\\ 1& 0& 0& 1\end{array}\right),$$
(31)
for $`k=6,8,10`$, respectively. Note that $`A^4=1`$ for $`k=8`$ and $`A^5=1`$ for $`k=10`$.
Next we briefly comment on the step (iii). The BPS charge lattice of the low energy effective theory is an integral symplectic lattice which can be identified with the middle cohomology lattice of the mirror manifold $`H^3(\widehat{X},𝐙)`$. On the other hand, in the large volume limit, the lattice of the microscopic D-brane charges is an integral quadratic lattice identified with the K theory lattice $`K(X)`$. We have to construct a map between the low energy charges $`Q`$ and the topological invariants of the K theory class $`\eta `$ by exploiting the exact form of the Chern-Simons couplings. The effective charges of D-branes are measured by the Mukai vector $`qH^{\mathrm{even}}(X)`$ given by $`q=ch(\eta )\sqrt{Td(X)}`$. The central charge associated with this state is then
$$Z(t)=\frac{t^3}{6}q^0\frac{t^2}{2}q^2+tq^4q^6.$$
(32)
The comparison of these central charges gives the relation between the low energy charges and the topological invariants of $`\eta `$. It is hard to see the behavior of each cycles in the case of one-parameter hypersurfaces. Thus we can not care about this mapping and have to content ourselves with determining only $`(Q_6,Q_4,Q_2,Q_0)`$.
Finally, we come to the step (i) which is the core of the method. In principle, the above central charges should be compared with those of the boundary states, but such a comparison seems to be difficult. It is useful to compute the interaction between the two D-branes in the open string channel. This is canonically normalized because it is a partition function. Therefore we consider the intersection form. We first express the known intersection form in the large volume limit in terms of a natural basis at the Gepner point which has the $`𝐙_k`$ symmetry and then compare this with that of the boundary states. Given the classical intersection form $`\eta `$ (of course, this is different from the above topological class) in the large volume limit, we can determine the intersection form in the Gepner basis, $`\eta _G=M^1\eta (M^1)^t`$ where $`\eta `$ is given by $`\mathrm{\Sigma }_6\mathrm{\Sigma }_0=+1`$ and $`\mathrm{\Sigma }_4\mathrm{\Sigma }_2=1`$ where $`\mathrm{\Sigma }_{2n}`$ are the even-dimensional cycles in $`X`$. The intersection form $`\eta _G`$ is not canonically normalized, but this does not concern us. We can write down the intersection form invariant under the $`𝐙_k`$ symmetry. The $`𝐙_k`$ generators act on the 3-cycles at the Gepner points $`\mathrm{\Sigma }_i^G(i=0,\mathrm{},k1)`$ as $`g\mathrm{\Sigma }_i^G=\mathrm{\Sigma }_{i+1}^G`$. The results are
$`I_G^{k=6}`$ $`=`$ $`g(1+g)(1g)^3,`$
$`I_G^{k=8}`$ $`=`$ $`g(1+g)(1g)^3(1+g^2),`$ (33)
$`I_G^{k=10}`$ $`=`$ $`g(1+g)(1g)^2(1g^5).`$
On the other hand, we obtain the intersection form (23) among the $`L_i=0`$ states as follows
$`I_B^{k=6}`$ $`=`$ $`(1g^4)(1g^5)^4,`$
$`I_B^{k=8}`$ $`=`$ $`(1g^7)^4(1g^4),`$ (34)
$`I_B^{k=10}`$ $`=`$ $`(1g^9)^3(1g^8)(1g^5),`$
where we have inserted the trivial factors in $`k=8,10`$ as explained before. We can identify the $`𝐙_k`$ generator on the both basis to obtain
$$I_B=(1g)I_G(1g)^t.$$
(35)
¿From this equation, we can relate the boundary states $`|\{0\};M;S`$ to the basis of the periods $`\varpi `$ at the Gepner point. Then the charges of the boundary states $`Q_B`$ is given by $`Q_G=Q_B(1g)`$.
We summarize the procedure as follows. First, we take $`Q_G`$ for $`|\{0\};0;0`$ to be $`(0,1,1,0)`$ and use the monodromy matrix in order to obtain the charge in the large volume limit, $`Q=Q_GM^1`$. Then the state $`|\{0\};0;0`$ corresponds to the pure six-brane $`(1,0,0,0)`$. We can obtain charges for different $`M`$ by acting $`A^1`$ which implements $`g`$ : $`MM+2`$. The action $`h`$ : $`SS+2`$ is also implemented by $`QQ`$. The charges of states with $`L=_iL_i>0`$ can be obtained by $`Q_G=Q_B_{j=1}^rt_{L_j}(1g)`$. When $`L_i`$ is odd, we also have to multiply the boundary states by $`g^{1/2}`$. By using the above procedure, we can determine the geometric charges for all the boundary states. We can also calculate the number of supersymmetry preserving moduli of the brane configuration which consists of the same boundary states at the Gepner point by using eq.(24). We find that it depends only on $`L_j`$.
It would be meaningless to array all the results, thus we give some interesting ones. First, the values ($`\mathrm{\Delta }M=M\stackrel{~}{M},\mathrm{\Delta }S=S\stackrel{~}{S}`$) which satisfy the supersymmetric condition (17) are $`(0,\mathrm{\hspace{0.17em}0})`$ and $`(k,\mathrm{\hspace{0.17em}2})`$. Since all the $`k`$’s are even integers, the configuration between the state with $`L_j=`$ even and the state with $`L_j=`$ odd can not be supersymmetric because of the condition $`\mathrm{\Delta }L+\mathrm{\Delta }M+\mathrm{\Delta }S2𝐙`$. The conditions $`A^4=1`$ for $`k`$=8 and $`A^5=1`$ for $`k`$=10 force the independent states to satisfy only $`\mathrm{\Delta }S=\mathrm{\Delta }M=0`$. Thus the number of independent boundary states are $`90,140,350`$ for $`k=6,8,10`$, respectively. But we found that some different boundary states have the same charge in the large volume limit. Then we give some results on the geometric charges and the number of moduli of boundary states in the following tables. On the left side, the boundary states $`|L_i2𝐙;0;0`$ which are supersymmetric with $`|\{0\};0;0`$ are included. On the right side, the boundary states $`|L_i2𝐙+1;1;0`$ which are supersymmetric with $`|10000;1;0`$ for $`k=6`$ or $`|1000;1;0`$ for $`k=8,10`$ are included.
$`𝐤=\mathrm{𝟔}`$
$$\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 00000& 1& 0& 0& 0& 0\\ 20000& 1& 0& 3& 0& 7\\ 11000& 0& 0& 3& 0& 8\\ 22000& 2& 0& 6& 0& 22\\ 11110& 3& 0& 12& 0& 40\\ 22200& 4& 0& 12& 0& 68\\ 22110& 6& 0& 18& 0& 94\\ 22220& 8& 0& 24& 0& 208\end{array}\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 10000& 2& 1& 2& 4& 1\\ 21000& 2& 1& 9& 1& 15\\ 11100& 0& 0& 6& 3& 19\\ 22100& 4& 2& 18& 2& 46\\ 21110& 6& 3& 27& 3& 63\\ 22210& 8& 4& 36& 4& 140\end{array}$$
(36)
$`𝐤=\mathrm{𝟖}`$
$$\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 0000& 1& 0& 0& 0& 0\\ 2000& 1& 0& 2& 0& 6\\ 1100& 0& 0& 2& 0& 7\\ 3100& 2& 0& 4& 0& 14\\ \mathrm{}& & & & & \\ 3300& 0& 0& 4& 0& 28\\ \mathrm{}& & & & & \\ 3333& 8& 0& 32& 0& 496\end{array}\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 1000& 2& 1& 2& 4& 3\\ 3000& 4& 2& 8& 6& 6\\ 2100& 2& 1& 6& 2& 11\\ 1110& 0& 0& 4& 2& 15\\ \mathrm{}& & & & & \\ 3310& 0& 0& 8& 4& 60\\ \mathrm{}& & & & & \\ 3332& 8& 4& 40& 0& 376\end{array}$$
(37)
$`𝐤=\mathrm{𝟏𝟎}`$
$$\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 0000& 1& 0& 0& 0& 0\\ 2000& 0& 0& 1& 0& 4\\ 1100& 1& 0& 1& 0& 5\\ 0001& 1& 0& 1& 0& 3\\ \mathrm{}& & & & & \\ 4400& 0& 0& 4& 0& 36\\ \mathrm{}& & & & & \\ 4441& 8& 0& 24& 0& 392\end{array}\begin{array}{cccccc}L& Q_6& Q_4& Q_2& Q_0& \mathrm{moduli}\\ 1000& 2& 1& 0& 3& 2\\ 3000& 2& 1& 2& 2& 6\\ 2100& 0& 0& 2& 1& 8\\ 1110& 2& 1& 2& 4& 12\\ \mathrm{}& & & & & \\ 3111& 2& 1& 8& 1& 50\\ \mathrm{}& & & & & \\ 4331& 8& 4& 24& 0& 316\end{array}$$
(38)
Only for $`k=6`$, there are the other supersymmetric states which are not listed in (36). On the left side, they are three independent states $`|00000;6;2`$, $`|11000;6;2`$, $`|11110;6;2`$ with the charges $`(2,0,3,0)`$, $`(3,0,6,0)`$, $`(6,0,15,0)`$, which are supersymmetric with $`|00000;0;0`$. One should notice that the large number of the moduli such as 496 in $`k=8`$ may be related to the unknown degrees of freedom. The interpretation of this extra degrees of freedom seems to be the most urgent problem of this approach.
We can also pick up some states whose charge have the only one component except for the state $`|00000;0;0`$. For $`k=6`$, there are no such states. For $`k=8`$, $`|3000;2;0`$ has the charge $`(0,0,0,2)`$. For $`k=10`$, $`|4000;6;0,|4200;6;0,|4440;6;0`$ have the charges $`(0,0,0,2),(0,2,0,0),(0,8,0,0)`$. Thus we found D0-branes in $`k=8,10`$ as opposed to quintic. But this viewpoint may not be appropriate because we can make the linear combination of the boundary states to produce the various branes with the only one component. Thus it would be important to consider the precise map between the two basis. We can also observe the phenomena that the non-BPS states in the large volume limit become the stable BPS states at the Gepner point. This may be related to the existence of the marginal stability lines in .
In these settings, the marginal operators at the Gepner point would play the important role. These boundary operator may have the superpotentials, with the flat directions corresponding to the truly marginal operators. In , some first steps was made in order to compute the superpotentials on D-branes on Calabi-Yau. We hope to return this problem in the future work.
## Acknowledgments
We are grateful to T. Eguchi, Y. Matsuo for helpful discussions. We also wish to thank J. Hashiba, M. Jinzenji, K. Sugiyama and Y. Satoh for discussions and encouragements. The work of M. Naka is supported by JSPS Research Fellowships for Young Scientists. |
warning/0001/hep-ph0001026.html | ar5iv | text | # Making Baryons Below the Electroweak Scale
## I Introduction
Over the past twenty to thirty years, a variety of microphysical explanations for the observed baryon asymmetry of the universe have been proposed. Initially. Grand Unified Theories (GUTs), which easily satisfy Sakharov’s three criteria for baryogenesis, were demonstrated to be able to account for the observed baryon to photon ratio in the Universe today . Although proton decay experiments soon ruled out the simplest theories, GUT baryogenesis remained a viable possibility in more complicated models. However, GUTs also lead to several cosmological problems. Since inflation erases any preexisting asymmetry, GUT baryogenesis is only possible if the reheating scale following inflation is large. It has been realized for some time that this raises the possibility of unacceptable defect and, in supersymmetric (SUSY) models, gravitino and moduli, production after inflation . However, as we’ve heard in a number of talks here , these concerns have become much more pressing recently, with the realization that, in the context of preheating, gravitino and moduli production can be so efficient as to constrain the reheat temperature to be less than $`10`$ GeV in some models.
An additional problem for GUT baryogenesis contained the seeds for potentially viable baryogenesis at the much lower electroweak scale ($`10^2`$ GeV). Coherent configurations of electroweak gauge and Higgs fields, first pointed out by ’t Hooft , can violate baryon number via non-perturbative physics. At zero temperature this effect is exponentially suppressed by the energy of a field configuration called the sphaleron, and is essentially irrelevant. However, as pointed out by Kuzmin, Rubakov and Shaposhnikov , and later discussed by Arnold and McLerran , at finite temperature, sphaleron production and decay can be rampant. This has the virtue of allowing copious baryon number violation, but can also be a curse. If the universe remains in thermal equilibrium until sphaleron production ceases, the net effect of these processes will be to drive the baryon number of the universe to zero, unless careful precautions are made to ensure either out of equilibrium sphaleron decay, or quantum number restrictions which forbid the elimination of the net baryon number. In the light of recent lattice and experimental data, this seems to require new fields at the weak scale, perhaps those predicted by SUSY.
Finally, once again, if the gravitino reheating constraint is sufficiently strong, we can not allow the universe to reheat to a sufficient temperature to allow even electroweak baryogenesis to take place.
Thus, thermal sphaleron production creates both challenges and opportunities for the generation of the baryon asymmetry. While it can wipe out any baryon number generated at the GUT scale, it offers the possibility of electroweak baryogenesis, although in practice this is quite difficult to achieve (for reviews see ).
As I alluded to earlier, reheating after cosmological inflation has been carefully rethought over the last few years. Studies of the inflaton dynamics have revealed the possibility of a period of parametric resonance, prior to the usual scenario of energy transfer from the inflaton to other fields. This phenomenon, which is characterized by large amplitude, non-thermal excitations in both the inflaton and coupled fields, has become known as preheating . Two particularly interesting consequences of this are the strict graviton and moduli constraints that result , and the idea that topological defects may be produced after inflation even when the final reheat temperature is lower than the symmetry breaking scale of the defects .
In this article, I describe how all these ideas can be combined to yield a viable and attractive model which obviates many of the problems with both standard GUT, and electroweak baryogenesis . In particular, I show how baryogenesis might still occur, even if inflation ends with reheating below the electroweak scale.
## II The Basic Mechanism
The fundamental idea is that, if topological defects can be produced non-thermally during preheating, then so can coherent configurations of gauge and Higgs fields, carrying nontrivial values of the Higgs winding number
$$N_H(t)=\frac{1}{24\pi ^2}d^3xϵ^{ijk}\text{Tr}[U^{}_iUU^{}_jUU^{}_kU].$$
(1)
In this parameterization, the $`SU(2)`$ Higgs field $`\mathrm{\Phi }`$ has been expressed as $`\mathrm{\Phi }=(\sigma /\sqrt{2})U`$, where $`\sigma ^2=2\left(\phi _1^{}\phi _1+\phi _2^{}\phi _2\right)=\mathrm{Tr}\mathrm{\Phi }^{}\mathrm{\Phi }`$, and $`U`$ is an $`SU(2)`$-valued matrix that is uniquely defined anywhere $`\sigma `$ is nonzero.
These winding configurations are not stable and evolve to a vacuum configuration plus radiation. In the process fermions may be anomalously produced. If the fields relax to the vacuum by changing the Higgs winding then there is no anomalous fermion number production. However, if there is no net change in Higgs winding during the evolution (for example $`\sigma `$ never vanishes) then there is anomalous fermion number production. Since winding configurations will be produced out of equilibrium (by the nature of preheating) and since CP-violation affects how they unwind, all the ingredients to produce a baryonic asymmetry are present (see for a detailed discussion of the dynamics of winding configurations).
If the final reheat temperature is lower than the electroweak scale, then then production of small-scale winding configurations by resonant effects is analogous to the production of local topological defects. In fact, the configurations that are of interest can be thought of as gauged textures.
Given this connection, a rough underestimate of the number density of winding configurations may be obtained by counting defects in recent numerical simulations of defect formation during preheating , while keeping in mind that the important case is when the symmetry breaking order parameter is not the inflaton itself, but is the electroweak $`SU(2)`$ Higgs field, and is coupled to the inflaton. The relevant quantity is the number density of defects directly after preheating, since winding-anti-winding pairs of configurations will not typically have time to find each other and annihilate before they decay. Finally, since the Higgs winding is the only non-trivial winding present at the electroweak scale, it is reasonable to assume that any estimates of defect production in general models can be quantitatively carried over to estimate of the relevant Higgs windings for preheating at the electroweak scale.
## III A (Too?) Simple Example
Before I make an estimate of the baryon asymmetry from this mechanism, I’ll provide an example of a toy model which satisfies all the relevant constraints.
Consider the potential
$$V(\varphi ,\chi )=\frac{1}{2}m^2\varphi ^2+\frac{1}{2}g^2\varphi ^2\chi ^2+\frac{1}{4}\lambda (\chi ^2\chi _0^2)^2,$$
(2)
for an inflaton $`\varphi `$, coupled to the electroweak Higgs field $`\chi `$. <sup>*</sup><sup>*</sup>*This model has also been independently proposed in a similar context in , and for a description of this see Misha Shaposhnikov’s contribution to these proceedings Here $`\chi _0=246`$ GeV is the electroweak symmetry breaking scale, $`m`$ is the (false vacuum) inflaton mass, and $`\lambda `$ (the Higgs self-coupling, here assumed to be of order unity) and $`g`$ are dimensionless constants.
The mechanism only works if parametric resonance into electroweak fields occurs in this model. The condition for this to happen is
$$q=\frac{g^2\varphi _0^2}{2m_\varphi ^2}>10^3,$$
(3)
where $`\varphi _0`$ is the value of $`\varphi `$ at the end of inflation. For the values quoted here, this condition yields $`g<10^2`$ (I’ll take $`g10^2`$). It is important that the temperature fluctuations in the cosmic microwave background (CMB), given by
$$\frac{\delta T}{T}g\frac{\chi _0^5}{M_p^3m^2},$$
(4)
for the values I’ve chosen here, are of the correct magnitude. Clearly this is satisfied by the choice $`m10^{21}`$ GeV. Finally, since the reheat temperature in this model is roughly bounded by $`T_{RH}(m\varphi _0)^{1/2}`$, the requirement that any baryons produced not be erased by equilibrium sphaleron processes is also satisfied.
This is not a particularly natural toy model, and in fact, it may develop problems if we go beyond tree level . However, the point of this example is merely to provide an existence proof which makes explicit the constraints on such a possibility.
## IV Calculating the Asymmetry
Consideration of topological defect production following inflation has been discussed by several authors . For definiteness, let us focus on the results of Khlebnikov et al.. These show that, for sufficiently low symmetry breaking scales, the initial number density of defects produced is very high. Here, by initial, I mean the number seen after copious symmetry-restoring transitions cease. One may perform an estimate from the first frame of Figure 6. of reference . The box size has physical size $`L_{phys}50\eta ^1`$ where $`\eta `$ is the symmetry breaking scale, and I’ve assumed couplings of order unity. In this box there are of order $`N=50`$ defects at early times. Thus, a rough estimate of the number density of winding configurations is
$$n_{\mathrm{configs}}\frac{N}{L_{phys}^3}4\times 10^4\eta ^3.$$
(5)
In order to make a simple estimate of the baryon number produced, it remains to show how CP-violation may bias the decays of these configurations to create a net baryon excess.
The effect of CP-violation on winding configurations can be very complicated, and in general depends strongly on the shapes of the configurations and the particular type of CP-violation. However, in general, the situation considered here, when out of equilibrium configurations are produced in a background low-temperature electroweak plasma most closely resembles local electroweak baryogenesis in the “thin-wall” regime. Winding configurations are produced when non-thermal oscillations take place in a region of space and restore the symmetry there. Since the reheat temperature is lower than the electroweak scale, as the region reverts rapidly to the low temperature phase, the winding configuration is left behind. In the absence of CP-violation in the coupling of the inflaton to the standard model fields, a CP-symmetric ensemble of configurations with $`N_H=+1`$ and $`N_H=1`$ will be produced. (i.e. the probability for finding a particular $`N_H=+1`$ configuration in the ensemble is equal to that for finding its CP-conjugate $`N_H=1`$ configuration.) Then, without electroweak CP- violation, for every $`N_H=+1`$ configuration which relaxes in a baryon producing fashion there is an $`N_H=1`$ configuration which produces anti-baryons, and no net baryogenesis occurs. However, with CP-violation there will be some configurations which produce baryons whose CP-conjugate configurations relax without violating baryon number.
While an analytic computation of the effect of CP-violation does not exist , there exist numerical simulations (e.g. ), from which one expects that the asymmetry in the number density of decaying winding configurations should be proportional to a dimensionless number, $`ϵ`$, parameterizing the strength of the source of CP-violation. Now, at the electroweak scale the entropy density is $`s2\pi ^2g_{}T^3/45`$, where $`g_{}100`$ is the effective number of massless degrees of freedom at that scale. Thus, the final baryon to entropy ratio generated is
$$\eta \frac{n_B}{s}ϵg_{}^1\frac{n_{\mathrm{configs}}}{T_{RH}^3}.$$
(6)
Plugging in the approximate numbers obtained earlier, this yields
$$\eta 10^6ϵ.$$
(7)
This is the final estimate.
This estimate is quite rough, and the explicit model presented is merely a toy model. However these suggest that the mechanism proposed here could viably result in a phenomenologically allowed value of $`\eta 10^{10}`$, with CP violating physics within the range predicted in SUSY models for example.
## V Conclusions
I have described a new mechanism for baryogenesis, that is effective below the electroweak scale. The primary advantages of such a mechanism are that no thermal sphaleron production subsequently takes place to wash out any baryon number that is produced, and that no excess production of gravitinos or monopoles occurs, evading a very strong (although model-dependent) constraint. A more complete analysis of the mechanism requires a numerical solution to the coupled $`SU(2)`$-inflaton equations of motion, in the presence of CP-violation.
## Acknowledgements
I would like to thank Lawrence Krauss for a stimulating and enjoyable collaboration on this project. I also thank Matthew Parry, Richard Easther and Lisa Randall for helpful discussions. Finally, many thanks to the organizers, particularly Goran Senjanovic and Rachel Jeannerot for all their hard work, for making the conference fun, and for the excellent coffee. This work was supported by the US Department of Energy. |
warning/0001/math-ph0001004.html | ar5iv | text | # Solving second order ordinary differential equations by extending the PS method
## 1 Introduction
The fundamental position of differential equations (DEs) in scientific progress has, over the last three centuries, led to a vigorous search for methods to solve them. The overwhelming majority of these methods are based on classification of the DE into types for which a method of solution is known, which has resulted in a gamut of methods that deal with specific classes of DEs. This scene changed somewhat at the end of the 19th century when Sophus Lie developed a general method to solve (or at least reduce the order of) ordinary differential equations (ODEs) given their symmetry transformations . Lie’s method is very powerful and highly general, but first requires that we find the symmetries of the differential equation, which may not be easy to do. Search methods have been developed to extract the symmetries of a given ODE, however these methods are heuristic and cannot guarantee that, if symmetries exist, they will be found.
On the other hand in 1983 Prelle and Singer (PS) presented a deductive method for solving first order ODEs (FOODE) that presents a solution in terms of elementary functions if such a solution exists . The attractiveness of the PS method lies not only in its basis on a totally different theoretical point of view but, also in the fact that, if the given FOODE has a solution in terms of elementary functions, the method guarantees that this solution will be found (though, in principle it can admittedly take an infinite amount of time to do so). The original PS method was built around a system of two autonomous FOODEs of the form $`\dot{x}=P(x,y)`$, $`\dot{y}=Q(x,y)`$ with $`P`$ and $`Q`$ in $`C[x,y]`$ or, equivalently, the form $`y^{}=R(x,y)`$, with $`R(x,y)`$ a rational function of its arguments. Here we propose a generalization that allows us to apply the techniques developed by Prelle and Singer to second order differential equations (SOODEs). The key idea is to focus not on the final solution of the equation, but rather its invariants.
This paper is organized as follows: in section 2, the reader is introduced to the PS procedure; section 3 addresses our approach extending the ideas of the PS procedure to the case of SOODEs and discusses how generally applicable the method is to such equations. Section 4 is dedicated to some examples solved via our procedure and, finally, conclusions are presented in section 5.
## 2 The Prelle-Singer Procedure
Despite its usefulness in solving FOODEs, the Prelle-Singer procedure is not very well known outside mathematical circles, and so we present a brief overview of the main ideas of the procedure.
Consider the class of FOODEs which can be written as
$$y^{}=\frac{dy}{dx}=\frac{M(x,y)}{N(x,y)}$$
(1)
where $`M(x,y)`$ and $`N(x,y)`$ are polynomials with coefficients in the complex field $`C`$.
In , Prelle and Singer proved that, if an elementary first integral of (1) exists, it is possible to find an integrating factor $`R`$ with $`R^nC`$ for some (possible non-integer) $`n`$, such that
$$\frac{(RN)}{x}+\frac{(RM)}{y}=0.$$
(2)
The ODE can then be solved by quadrature. From (2) we see that
$$N\frac{R}{x}+R\frac{N}{x}+M\frac{R}{y}+R\frac{M}{y}=0.$$
(3)
Thus
$$\frac{D[R]}{R}=\left(\frac{N}{x}+\frac{M}{y}\right),$$
(4)
where
$$DN\frac{}{x}+M\frac{}{y}.$$
(5)
Now let $`R=_if_i^{n_i}`$ where $`f_i`$ are irreducible polynomials and $`n_i`$ are non-zero integers. From (5), we have
$`{\displaystyle \frac{D[R]}{R}}`$ $`=`$ $`{\displaystyle \frac{D[_if_i^{n_i}]}{_if_k^{n_k}}}={\displaystyle \frac{_if_i^{n_i1}n_iD[f_i]_{ji}f_j^{n_j}}{_kf_k^{n_k}}}`$ (6)
$`=`$ $`{\displaystyle \underset{i}{}}{\displaystyle \frac{f_i^{n_i1}n_iD[f_i]}{f_i^{n_i}}}={\displaystyle \underset{i}{}}{\displaystyle \frac{n_iD[f_i]}{f_i}}.`$
From (4), plus the fact that $`M`$ and $`N`$ are polynomials, we conclude that $`D[R]/R`$ is a polynomial. Therefore, from (6), we see that $`f_i|D[f_i]`$.
We now have a criterion for choosing the possible $`f_i`$ (build all the possible divisors of $`D[f_i]`$) and, by using (4) and (6), we have
$$\underset{i}{}\frac{n_iD[f_i]}{f_i}=\left(\frac{N}{x}+\frac{M}{y}\right).$$
(7)
If we manage to solve (7) and thereby find $`n_i`$, we know the integrating factor for the FOODE and the problem is reduced to a quadrature. Risch’s algorithm can then be applied to this quadrature to determine whether a solution exists in terms of elementary functions.
## 3 Extending the Prelle-Singer Procedure
In the previous section, the main ideas and concepts used in the Prelle-Singer procedure were introduced. Here we present an extension of these ideas applicable to SOODEs. The main idea is to focus on the first order invariants of the ODE rather than on the solutions.
### 3.1 Introduction
Consider the SOODE
$$y^{\prime \prime }=\frac{d^2y}{dx^2}=\frac{M(x,y,y^{})}{N(x,y,y^{})},$$
(8)
where $`M(x,y,y^{})`$ and $`N(x,y,y^{})`$ are polynomials with coefficients in $`C`$. We assume that (8) has a solution in terms of elementary functions, in which case there are two independent elementary functions of $`x`$ $`y`$ and $`y^{}`$ which are constant on all solutions of (8), namely the first order invariants
$$I_i(x,y,y^{})=C_ii=1,2.$$
(9)
Without loss of generalization we consider one of these and, dropping the index on $`I_i`$ we have
$$\mathrm{d}I=\frac{I}{x}\mathrm{d}x+\frac{I}{y}\mathrm{d}y+\frac{I}{y^{}}\mathrm{d}y^{}=0.$$
(10)
Now, introducing the notation $`\frac{I}{u}I_u`$, we have
$$I_x+I_yy^{}+I_y^{}y^{\prime \prime }=0,$$
(11)
and so
$$y^{\prime \prime }=\frac{I_x+I_yy^{}}{I_y^{}},$$
(12)
which is (8) in terms of the differential invariant $`I`$. Rewriting (8) as
$$\frac{M}{N}\mathrm{d}x\mathrm{d}y^{}=0$$
(13)
and observing that
$$y^{}\mathrm{d}x=\mathrm{d}y,$$
(14)
we can add the identically null term $`S(x,y,y^{})y^{}\mathrm{d}xS(x,y,y^{})\mathrm{d}y`$ to (13) and obtain the 1-form
$$\left(\frac{M}{N}+Sy^{}\right)\mathrm{d}xS\mathrm{d}y\mathrm{d}y^{}=0.$$
(15)
Notice that the 1-form (16) must be proportional to the 1-form (10). So, since the 1-form (10) is exact, we can multiply (16) by the integrating factor $`R(x,y,y^{})`$ to obtain
$$\mathrm{d}I=R(\varphi +Sy^{})\mathrm{d}xRS\mathrm{d}yR\mathrm{d}y^{}=0,$$
(16)
where $`\varphi M/N`$.
Comparing equations (10) and (16),
$`I_x`$ $`=`$ $`R(\varphi +Sy^{}),`$
$`I_y`$ $`=`$ $`RS,`$
$`I_y^{}`$ $`=`$ $`R.`$ (17)
Now equations (3.1) must satisfy the compatibility conditions $`I_{xy}=I_{yx},I_{xy^{}}=I_{y^{}x}`$ and $`I_{yy^{}}=I_{y^{}y}`$. This implies that
$`D[S]`$ $`=`$ $`\varphi _y+S\varphi _y^{}+S^2,`$ (18)
$`D[R]`$ $`=`$ $`R(S+\varphi _y^{}),`$ (19)
$`R_y`$ $`=`$ $`R_y^{}S+S_y^{}R,`$ (20)
where the differential operator $`D`$ is defined as
$$D\frac{}{x}+y^{}\frac{}{y}+\varphi \frac{}{y^{}}.$$
(21)
Combining (18) and (19) we obtain
$$D[RS]=R\varphi _y.$$
(22)
So if the product of $`S`$ and the integrating factor $`R`$ is a rational function of $`x`$, $`y`$ and $`y^{}`$, then $`D[RS]`$ is too. Since $`\varphi `$ is rational (and so, therefore, is $`\varphi _y`$), equation (22) tells us that $`R`$ is rational. Using (19) and similar arguments we conclude that $`S`$ must be a rational function of $`x`$, $`y`$ and $`y^{}`$.
In summary, from (3.1) it follows that the supposition that $`RS`$ is rational can be equated to the existence of a first order invariant whose derivatives in relation to $`x`$, $`y`$ and $`y^{}`$ are rational functions. With this in mind we restate the original supposition in the form of a conjecture.
### 3.2 The Conjecture
We first state a result proved in :
Theorem:Let $`K`$ be a differential field of functions in $`n+1`$ variables and $`L`$ an elementary extension of $`K`$. Let $`f`$ be in $`K`$ and assume there exists a nonconstant $`g`$ in $`L`$ such that $`g`$ is constant on all solutions of $`y^{(n)}=f(x,y,y^{},y^{\prime \prime },\mathrm{},y^{(n1)})`$. Then there exist $`w_1,\mathrm{},w_m`$ algebraic over $`K`$ and constants $`c_i,\mathrm{},c_m`$ such that
$$w_0(x,y,y^{},y^{\prime \prime },..,y^{(n1)})+\underset{i}{}c_i\mathrm{log}(w_i(x,y,y^{},y^{\prime \prime },\mathrm{},y^{(n1)}))$$
(23)
is a constant on all solutions of $`y^{(n)}=f(x,y,y^{},y^{\prime \prime },\mathrm{},y^{(n1)})`$.
This result shows that for the particular case of SOODEs whose solutions are elementary, there are two independent first order invariants of the form
$$w_0(x,y,y^{})+\underset{i}{}c_i\mathrm{log}w_i(x,y,y^{}).$$
(24)
Our conjecture is that if these two first order invariants exist it is always possible to find a function of them (which will, therefore, itself be a first order invariant) of the form
$$z_0(x,y,y^{})+\underset{i}{}c_i\mathrm{log}[z_i(x,y,y^{})],$$
(25)
where $`z_i`$ are rational functions of $`x`$, $`y`$ and $`y^{}`$.
Conjecture:Let $`K`$ be a differential field of functions in three variables and $`L`$ an elementary extension of $`K`$. Let $`f`$ be in $`K`$ and assume there exist two independent nonconstant $`\{g_1,g_2\}`$ in $`L`$ such that $`g_i`$ are constant on all solutions of $`y^{\prime \prime }=f(x,y,y^{})`$. Then there exists at least one constant of the form
$$z_0(x,y,y^{})+\underset{i}{}c_i\mathrm{log}(z_i(x,y,y^{}))$$
(26)
where the $`z_i`$ are in $`K`$.
By the previous reasoning it can be seen that (26) implies that the product $`RS`$ is a rational function of $`x`$, $`y`$ and $`y^{}`$.
If this conjecture holds, then our extension of the PS method applies to all SOODEs of the form (8). Though we have not been able to prove our conjecture, extensive trials while developing this procedure has not revealed any counter example. Even if the conjecture is false, our experience with real test cases has shown that the method is, at least, applicable to the vast majority of SOODEs of the form (8).
### 3.3 Finding $`R`$ and $`S`$
Our conjecture implies that, if the SOODE to be solved has an elementary general solution, then $`S`$ is a rational function which we may write as
$$S=\frac{S_n}{S_d}=\frac{_{i,j,k}a_{ijk}x^iy^jy_{}^{}{}_{}{}^{k}}{_{i,j,k}b_{ijk}x^iy^jy_{}^{}{}_{}{}^{k}}.$$
(27)
We can also see that (18) does not involve $`R`$. So, given a degree bound on the polynomials $`S_n`$ and $`S_d`$, we may find a set of solutions to this equation which are then candidates to solve the system of equations (18)–(20).
From (19) we have
$$\frac{D[R]}{R}=(S+\varphi _y^{})=\frac{S_n}{S_d}\left(\frac{M}{N}\right)_y^{}=\frac{S_nN^2+S_d(NM_y^{}MN_y^{})}{S_dN^2}$$
(28)
which can be rewritten as
$$\frac{𝒟[R]}{R}=S_nN^2+S_d(NM_y^{}MN_y^{}),$$
(29)
where the differential operator $`𝒟`$ is defined as
$$𝒟(S_dN^2)D.$$
(30)
We keep in mind that
* $`S_n`$, $`S_d`$, $`N`$ and $`M`$ are polynomials in $`x`$, $`y`$ and $`y^{}`$;
* $`𝒟`$ is a linear differential operator whose coefficients of $`\frac{}{x}`$, $`\frac{}{y}`$ and $`\frac{}{y^{}}`$ are polynomials in $`x`$, $`y`$ and $`,y^{}`$;
* $`R`$ is a rational function of $`x`$, $`y`$ and $`y^{}`$, which we may write as
$$R=\frac{R_n}{R_d}=\frac{_{i,j,k}c_{ijk}x^iy^jy_{}^{}{}_{}{}^{k}}{_{i,j,k}d_{ijk}x^iy^jy_{}^{}{}_{}{}^{k}}.$$
(31)
If we have a theoretical limit on the degrees of $`R_m`$ and $`R_d`$ (a degree bound), we may use a procedure analogous to that described in section 2 to obtain candidates for the integrating factor $`R`$. We simply construct all polynomials in $`x`$, $`y`$ and $`y^{}`$ up to the degree bound.
### 3.4 Reduction of the SOODE
Once $`R`$ and $`S`$ have been determined using equations (3.1) we have all the partial first derivatives of the first order differential invariant, $`I(x,y,y^{})`$, which is constant on the solutions. This invariant can then be obtained as
$$I(x,y,y^{})=R\left(\varphi +Sy^{}\right)𝑑x$$
$$\text{[}RS+\frac{}{y}R\left(\varphi +Sy^{}\right)𝑑x\text{]}𝑑y$$
$$\left[R+\frac{}{y^{}}\left(R\left(\varphi +Sy^{}\right)𝑑x\text{[}RS+\frac{}{y}R\left(\varphi +Sy^{}\right)𝑑x\text{]}𝑑y\right)\right]𝑑y^{}.$$
(32)
The equation $`I(x,y,y^{})=C_1`$ can then be solved to obtain a FOODE for $`y^{}`$: the reduced ODE
$$y^{}=\phi (x,y,C1).$$
(33)
To obtain the general solution of the original ODE, we can apply the Prelle-Singer method in its original form to this reduced ODE. Thus, if our conjecture is correct, the method proposed here (for SOODEs of the form (8)) is as algorithmic as the original PS method for FOODEs. We note that the original PS method fails to be what is strictly an algorithm because no theoretical degree bound is yet known for the candidate polynomials which enter in the prospective solution, and so the procedure has no effective terminating condition for the case when an elementary solution does not exists. In practice, a terminating condition is put in by hand (it is found that polynomials of degree higher than 4 lead to computations which are overly complex for the average desktop computer). However, should such a degree bound be established, and our conjecture shown to be true, then the method proposed here would be an algorithm for deciding whether elementary solutions of SOODEs of the form (8) exist.
## 4 Examples
In this section we present examples of physically motivated SOODEs that are solved by our procedure<sup>1</sup><sup>1</sup>1We present only the reduction of the SOODEs since the integration of the resulting FOODE can be achieved by various methods, including the PS method itself.. As a simple illustrative example, we begin with the classical harmonic oscillator and then consider some nonlinear SOODEs which arise from astrophysics and general relativity.
Example 1: The Simple Harmonic Oscillator
In its simplest form, the equation for the simple harmonic oscillator is
$$y^{\prime \prime }=y.$$
(34)
For this ODE equations (18), (19) and (20) are
$`S_x+y^{}S_yyS_y^{}`$ $`=`$ $`1+S^2,`$ (35)
$`R_x+y^{}R_yyR_y^{}`$ $`=`$ $`RS,`$ (36)
$`R_yR_y^{}SS_y^{}R`$ $`=`$ $`0.`$ (37)
One possible solution to these equations is
$$S=\frac{y}{y^{}},R=y^{}.$$
(38)
From this, and using (32), we get the reduced ODE
$$C1=y^2+y^2,$$
(39)
which, of course, represents the energy conservation for the oscillator.
This example is very simple and leads to a form of $`\varphi `$ which is independent of $`x`$ and $`y^{}`$. And, as with all linear ODEs, alternative and more straightforward solution methods exist. The other examples illustrate the solution method at work for non-linear SOODEs which can be placed in the form (8).
Example 2: An Exact Solution in General Relativity
A rich source of non-linear DEs in physics are the highly non-linear equations of General Relativity. In general, Einstein’s equations are, of course, partial DEs, but there exist classes of equations where the symmetry imposed reduces these equations to ODEs in one independent variable. One such class is that of static, spherically symmetric solutions for stellar models, which depend only on the radial variable, $`r`$. The metric for a general statically spherically spacetime has two free functions, $`\lambda (r)`$ and $`\mu (r)`$ say. On imposing the condition that the fluid is a perfect fluid, Einstein’s equations reduce to two coupled ODEs for $`\lambda (r)`$ and $`\mu (r)`$. Specifying one of these functions reduces the problem to solving an ODE (of first or second order) for the other.
Following this procedure, Buchdahl obtained an exact solution for a relativistic fluid sphere by considering the so-called isotropic metric
$$\dot{s}^2=(1f)^2(1+f)^2\dot{t}^2(1+f)^4[\dot{r}^2+r^2(\dot{\theta }^2+\mathrm{sin}^2\theta \dot{\varphi }^2]$$
with $`f=f(r)`$. The field equations for $`f(r)`$ reduce to
$$ff^{\prime \prime }3f^2r^1ff^{}=0.$$
Changing notation with $`y(x)=f(r)`$, equations (18), (19) and (20) assume the form
$`S_x+y^{}S_y+{\displaystyle \frac{y^{}\left(3y^{}x+y\right)}{xy}}S_y^{}`$ $`=`$ $`{\displaystyle \frac{y^{}}{xy}}+{\displaystyle \frac{y^{}\left(3y^{}x+y\right)}{xy^2}}+`$ (40)
$`\left({\displaystyle \frac{3y^{}x+y}{xy}}+3{\displaystyle \frac{y^{}}{y}}\right)S+S^2,`$
$`R_x+y^{}R_y+{\displaystyle \frac{y^{}\left(3y^{}x+y\right)}{xy}}R_y^{}`$ $`=`$ $`R\left(S+{\displaystyle \frac{3y^{}x+y}{xy}}+3{\displaystyle \frac{y^{}}{y}}\right),`$ (41)
$`R_yR_y^{}SS_y^{}R`$ $`=`$ $`0.`$ (42)
One solution of those equations is
$$S=3\frac{y^{}}{y},R=\frac{1}{xy^3}$$
(43)
By using (32) we obtain the reduced FOODE:
$$C_1=y^{}/(y^3x)$$
(44)
which is separable and easily integrated to obtain the general solution
$$y(x)^2=\left(C_1x^2+C_\mathit{2}\right)^1.$$
(45)
Example 3: A Static Gaseous General-Relativistic Fluid Sphere
In a later paper , Buchdahl approaches the problem of the general relativistic fluid sphere using a different coordinate system from the previous example. For ease in comparison of the originals,
Substituting the $`\xi (r)`$ of the original Writing $`y(x)`$ instead of the $`\xi (r)`$ in the original, arrives at the equation
$$y^{\prime \prime }=\frac{x^2y_{}^{}{}_{}{}^{2}+y^21}{x^2y},$$
(46)
For this SOODE, eqs (18, 19 and 20) become:
$`S_x+y^{}S_y+{\displaystyle \frac{x^2y_{}^{}{}_{}{}^{2}+y^21}{x^2y}}S_y^{}`$ $`=`$ $`2x^2+{\displaystyle \frac{x^2y_{}^{}{}_{}{}^{2}+y^21}{y^2x^2}}`$ (47)
$`+2{\displaystyle \frac{y^{}S}{y}}+S^2`$
$`R_x+y^{}R_y+{\displaystyle \frac{x^2y_{}^{}{}_{}{}^{2}+y^21}{x^2y}}R_y^{}`$ $`=`$ $`R\left(S+2{\displaystyle \frac{y^{}}{y}}\right)`$ (48)
$`R_yR_y^{}SS_y^{}R`$ $`=`$ $`0`$ (49)
One solution to those equations is:
$$S=\frac{x^2y_{}^{}{}_{}{}^{2}xyy^{}+1}{xy^2+x^2yy^{}},R=\frac{y+xy^{}}{xy^2}.$$
(50)
From this, using eq. (32), we get the reduced FOODE:
$$C_1=\frac{2xyy^{}+y^2+x^2y_{}^{}{}_{}{}^{2}1}{2x^2y^2}.$$
(51)
which can be solved to:
$$y(x)^2=\frac{\mathrm{tan}(\sqrt{2}\sqrt{C_\mathit{1}}(C_\mathit{2}+x))^2}{(2C_\mathit{1}+2\mathrm{tan}(\sqrt{2}\sqrt{C_\mathit{1}}(C_\mathit{2}+x))^2C_\mathit{1})x^2}$$
(52)
This example has an extra feature: It is not solved by other solvers we have tried (mainly the Maple solver, in the version 5, that we believe to be the best). So, apart from the (already) very interesting fact that our approach is an algorithmic attempt to solve SOODEs, we have also this present fact, i.e., some SOODEs are solved via our method and “escape” from other very powerful solvers.
## 5 Conclusion
In this paper, we presented an approach that is an extension of the ideas developed by Prelle-Singer to tackle FOODEs. We believe it to be the first technique to address algorithmically the solution of SOODEs with elementary first integrals.
Here, we dealt with a restrict class of SOODEs (namely, the ones of the form (8)). However, we can use our method in solving SOODEs where $`\varphi (x,y,y^{})`$ depends on elementary functions of $`x,y,y^{}`$, following the developments for the Prelle-Singer approach for FOODEs . We are presently working on those ideas.
The generality of our approach is based on a conjecture (see section (3.2)) that we have already proved for many special cases. Even if the conjecture is proven false, our approach is a powerful tool in dealing with SOODEs since we have extensively tested it with many equations, both from mathematics and physical origin. In fact, since all the examples we have encounter have been solved by our approach, we are preparing a computational package implementing the Prelle-Singer procedure (and our present extension) to be submitted to Computer Physics Communications. |
warning/0001/astro-ph0001529.html | ar5iv | text | # The Properties of Young Clusters in M82
## 1. Introduction
Observations with the Hubble Space Telescope (HST) have revealed that hundreds of super star clusters (SSCs) are present in the nearby starburst galaxy M82 (O’Connell et al. 1995). SSCs appear to be frequently associated with starbursts, and it has often been suggested that they represent young globular clusters. One critical aspect is whether they have enough mass to survive over long time-scales. In this paper, we present a detailed investigation of the luminous cluster F (O’Connell & Mangano 1978) and the nearby, highly reddened, cluster L (Kronberg, Pritchet, & van den Bergh 1972). They are located 440 pc south-west of the nucleus of M82.
Our study is based on optical spectroscopy obtained at the 4.2 m William Herschel Telescope (WHT) at resolutions of 1.6 Å and 8 km s<sup>-1</sup>. We use the intermediate dispersion spectra to obtain ages for the two clusters. The brightness and compactness of SSC F make it an ideal candidate for measuring the line of sight velocity dispersion, and hence obtaining the dynamical mass of the cluster. We use our recently obtained high dispersion red spectra to derive the mass and discuss whether our derived parameters are consistent with SSC F being able to survive to become a globular cluster.
## 2. The Ages of Clusters F and L
Observations of clusters F and L were obtained in 1997 March with the WHT on La Palma, Canary Islands, the double beam spectrograph ISIS and $`1024\times 1024`$ TeK CCDs on the blue and red arms. The spectra, shown in Fig. 1, cover 3300–5500 Å and 5700–8800 Å for cluster F and 5700–8800 Å for cluster L. The resolution is 1.6 Å with a S/N of 30–40 for a total exposure time of 100 min for each arm. We derive a $`V`$ magnitude of 15.8 and (B$`V`$)=1.07 for cluster F.
The blue spectrum of cluster F is dominated by broad Balmer absorption lines and a strong Balmer jump, indicative of a mid-B spectral type. The red spectrum shows strong absorption due to the Ca II triplet and many weak features attributable to F and G supergiants. The red spectrum of cluster L is very similar to that of F.
The quality of the WHT spectra allows us to considerably improve previous age estimates for cluster F. We have used the PEGASE spectral synthesis code (Fioc & Rocca-Volmerange 1997) to compute synthetic spectra for ages of 20, 40, 60, 80 and 100 Myr, using a Salpeter IMF, the Geneva evolutionary tracks and the Jacoby, Hunter, & Christian (1984) spectral library. The ratio of the observed to model spectra suggests a reddening $`E(BV)1.5`$, although we find that a standard Galactic extinction law does not properly correct the data. The 40 Myr model gives the best fit to the H$`\beta `$, H$`\gamma `$ and H$`\delta `$ absorption lines but does not do as well for the Balmer jump region as the 60 and 80 Myr models. The 20 and 100 Myr models gave worse fits in all areas. We therefore adopt an age for cluster F of $`60\pm 20`$ Myr. The ratio between the observed spectrum of cluster F and the 60 Myr model is shown in Fig. 2.
For cluster L, we find that similarities in the strength of the Ca II triplet and the overall spectral appearance with cluster F suggest a similar age. We conclude that M82 experienced an episode of intense star formation $``$ 60 Myr ago in the mid-disc region. In the central region, the age of the starburst is younger, at 20–30 Myr (Rieke et al. 1993). More details of this analysis can be found in Gallagher & Smith (1999).
## 3. The Mass of Cluster F
Cluster F was observed in 1999 February with the WHT and the Utrecht echelle spectrograph (UES). The detector used was a $`2048\times 2048`$ SITe CCD and the wavelength range covered was 5760–9140 Å in a single exposure at a resolution of 8 km s<sup>-1</sup>. We achieved a per pixel S/N of 15–25 in a total integration time of 6.4 hr.
We chose this spectral region because, as discussed by Ho & Filippenko (1996a,b), and demonstrated by our ISIS spectra, the wavelength region longward of 5000 Å is dominated by the light of cool supergiants. Thus by cross-correlating the cluster spectrum with a suitable template spectrum of a cool supergiant, it is possible to recover the velocity dispersion of the cluster, and hence derive its mass by application of the virial theorem.
We obtained UES spectra of eight stars with spectral types from A7 III to M2 Iab for the purpose of providing a suitable template. Comparison of these spectra with that of cluster F shows that the best spectral match is between HR 6863 (F8 II) and HR 1529 (K1 III), as demonstrated in Fig. 3 for the region containing the Ca II triplet.
We have therefore used these two stars as templates and cross-correlated their spectra with that of cluster F. We chose four spectral regions which are free of telluric absorption lines: 6010–6275 Å, 6320–6530 Å, 7340–7590 Å, and 7705–8132 Å for the analysis. We ignored the region containing the Ca II triplet because these lines are saturated in the template spectra and broad Paschen absorption lines from early-type stars are present in the cluster F spectrum.
For each cross-correlation function, we measured the FWHM by fitting a Gaussian profile. The relationship between the FWHM and the velocity dispersion was empirically calibrated by broadening the template spectra with Gaussians of different dispersions and cross-correlating with the original spectra.
Using this approach, we find that relative to HR 6863 (F8 II) and HR 1529 (K1 III) the velocity dispersion of F is $`13.2\pm 1.9`$ and $`16.5\pm 2.4`$ km s<sup>-1</sup> respectively. The latter value is larger because the dispersion due to macroturbulence is smaller in the K1 giant. The difference in the two values is consistent with the velocity dispersion derived by cross-correlating the two template stars. We adopt the lowest value as representing the velocity dispersion of F because we expect the spectrum to be dominated by cool supergiants, and the work of Gray & Toner (1986, 1987) indicates that the macroturbulence dispersion in an F8 bright giant is similar to that of cool supergiants.
To derive the dynamical mass of cluster F, we need its half-light radius. O’Connell et al. (1995) derive a radius of 1.9 pc from HST images taken with WFPC. We have searched the HST archive for images of cluster F taken with WFPC2 and have found two suitable images. We measure a half-light radius of 4.4 pc, uncorrected for the PSF which is likely to reduce it by $``$ 10%. We assume that this radius represents a good upper bound to the half-mass radius, and that the cluster is spherically symmetric with an isotropic velocity distribution. Application of the virial theorem then gives a mass of $`1.8\pm 0.5\times 10^6`$ M.
## 4. Is Cluster F a Young Globular Cluster?
In Table 1, we compare our derived properties of cluster F with those derived for the young super star clusters NGC 1569-A and NGC 1705-1 (Ho & Filippenko 1996a,b; Sternberg 1998). Cluster F appears to be more massive and luminous than both these clusters. To derive the value of M<sub>V</sub> given in Table 1, we have measured a $`V`$ magnitude of 16.5 from the WFPC2 image, and assumed a reddening of $`E(BV)=1.5`$ (Sect. 2).
Mandushev et al. (1991) provide measurements for 32 Galactic globular clusters. The mass of cluster F is a factor of $`9`$ higher than their mean mass of $`1.9\times 10^5`$ M. From the PEGASE models, we expect cluster F to dim by 4.5 mag for an age of 15 Gyr, giving M$`{}_{V}{}^{}=11.3`$. This is higher than the mean value from Mandushev et al. (1991) of $`8.1\pm 1.2`$ although the reddening towards F is uncertain. If we use $`E(BV)=1.0`$ (O’Connell et al. 1995) instead, then the predicted M<sub>V</sub> at 15 Gyr of $`9.8`$ agrees better with the mean value of Mandushev et al. (1991). The predicted mass-to-light ratio $`(M/L_V)_{}`$ for cluster F of 0.6 or 2.5 (with M$`{}_{V}{}^{}=11.3`$ and $`9.8`$, respectively) agrees well with the range of 0.7–2.9 for Galactic globular clusters (Mandushev et al. 1991). This suggests that cluster F could survive to become a globular cluster. Although it would be considerably more massive than the average Galactic globular cluster, it still would have less mass than large globular clusters such as $`\omega `$ Cen (Merritt, Meylan, & Mayor 1997).
## References
Fioc, M., & Rocca-Volmerange, B. 1997, A&A, 326, 950
Gallagher, J.S., III, & Smith, L.J. 1999, MNRAS, 304, 540
Gray, D.F., & Toner, C.G. 1986, ApJ, 310, 277
Gray, D.F., & Toner, C.G. 1987, ApJ, 322, 360
Ho, L.C., & Filippenko, A.V. 1996a, ApJ, 466, L83
Ho, L.C., & Filippenko, A.V. 1996b, ApJ, 472, 600
Jacoby, G.H., Hunter, D.A., & Christian, C.A. 1984, ApJS, 56, 257
Kronberg, P.P., Pritchet, C.J., & van den Bergh, S. 1972, ApJ, 173, L47
Mandushev, G., Spassova, N., & Staneva, A. 1991, A&A, 252, 94
Merritt, D., Meylan, G., & Mayor, M. 1997, AJ, 114, 1074
O’Connell, R. W., & Mangano, J. J. 1978, ApJ, 221, 62
O’Connell, R.W., Gallagher, J.S., III, Hunter, D.A., & Colley, W.N. 1995, ApJ, 446, L1
Rieke, G.H., Loken, K., Rieke, M. J., & Tamblyn, P. 1993, ApJ, 412, 99
Sternberg, A. 1998, ApJ, 506, 721 |
warning/0001/math0001116.html | ar5iv | text | # Finite jet determination of holomorphic mappings at the boundary
## 0. Introduction
A classical theorem of H. Cartan (\[HCa\]) states that an automorphism $`f`$ of a bounded domain $`D^N`$ is completely determined by its $`1`$-jet, i.e. its value and derivatives of order one, at any point $`Z_0D`$. If $`D`$, in addition, is assumed to be smoothly ($`C^{\mathrm{}}`$) bounded and strictly pseudoconvex, then by Fefferman’s theorem \[Fe\] any such automorphism extends smoothly to the boundary $`D`$ as an automorphism $`DD`$. It is then natural to ask: is $`f`$ completely determined by a finite jet at a boundary point $`pD`$? An affirmative answer to this question, when $`D`$ is strictly pseudoconvex, follows from the work of Chern and Moser \[CM\] (see also E. Cartan \[ECa1–2\] for the case $`N=2`$, and Tanaka \[T1–2\]). Indeed, the following local version of Cartan’s theorem is a consequence of their work. Any holomorphic mapping which is defined locally on one side of a smooth, Levi nondegenerate real hypersurface $`M^N`$ and extends smoothly to $`M`$, sending $`M`$ diffeomorphically into another smooth real hypersurface $`M^{}^N`$, is completely determined by its $`2`$-jet at a point $`pM`$. Observe that the conclusion is nontrivial even in the strictly pseudoconcave case when the mapping extends holomorphically to a full neighborhood of $`p`$.
The main objective of the present paper is to extend the above mentioned local result to a more general class of real hypersurfaces (Theorem 1 below). We should point out that the result for Levi nondegenerate hypersurfaces follows from the construction of a unique Cartan connection on a certain principal $`G`$-bundle over such a hypersurface. There is no analogue of this construction in the more general situation considered in this paper.
Let $`M^N`$ be a smooth real hypersurface and assume that $`M`$ is defined locally near a point $`p_0M`$ by the equation $`\rho (z,\overline{z})=0`$, where $`\rho `$ is a smooth function with $`\rho (p_0,\overline{p}_0)=0`$ and $`d\rho (p_0,\overline{p}_0)0`$. Let $`L_{\overline{1}},\mathrm{},L_{\overline{n}}`$, with $`n=N1`$, be a basis for the CR vector fields on $`M`$. We shall say that $`M`$ is $`k_0`$-nondegenerate at $`p_0`$ if
$$\text{Span}\left\{L^{\overline{\alpha }}\rho _z(p_0,\overline{p}_0)|\alpha |k_0\right\}=^N,$$
$`0.1`$
where $`\rho _z:=(_{z_j}\rho )_{1jN}`$, $`_{z_j}:=/z_j`$, and standard multi-index notation for differential operators is used i.e. $`L^\alpha :=L_{\overline{1}}^{\overline{\alpha }_1}\mathrm{}L_{\overline{n}}^{\overline{\alpha }_n}`$. This nondegeneracy condition will be given in a different, but equivalent, form in terms of the intrinsic geometry of $`M`$ in the next section. The reader is referred to the book \[BER3\] for basic material on real submanifolds in complex space and CR structures, and further discussion of various nondegeneracy conditions (see also §1 of the present paper). We mention here only that Levi nondegeneracy at a point $`p_0M`$ is equivalent to $`1`$-nondegeneracy. Our main result is the following.
###### Theorem 1
Let $`M,M^{}^N`$ be smooth ($`C^{\mathrm{}}`$) real hypersurfaces. Let $`f,gU^N`$, where $`U^N`$ is an open connected subset with $`M`$ in its boundary, be holomorphic mappings which extend smoothly to $`M`$ and send $`M`$ diffeomorphically into $`M^{}`$. If $`M`$ is $`k_0`$-nondegenerate at a point $`p_0M`$ and
$$(_z^\alpha f)(p_0)=(_z^\alpha g)(p_0),\alpha _+^N|\alpha |2k_0,$$
$`0.2`$
then $`fg`$ in $`U`$.
Finite jet determination of holomorphic mappings sending one real submanifold into another has attracted much attention in recent years. We mention here the papers \[BER1–2, 4–5\], \[L\], \[Han1–2\], \[Hay\], \[Z\]. The reader is also referred to the survey article \[BER6\] for a more detailed history. However, in all the above mentioned papers, it is either assumed that $`M`$ and $`M^{}`$ are real-analytic (which will imply that all mappings $`f`$ extend holomorphically to some neighborhood of $`M`$), or the conclusion is that the formal power series of the mapping $`f`$ is determined by a finite jet (see \[BER4\], \[L\]). Theorem 1 appears to be, to the best of the author’s knowledge, the first finite determination result, since the work of Chern and Moser mentioned above, which applies to merely smooth hypersurfaces and smooth mappings. We should mention that if $`M`$ and $`M^{}`$ are real-analytic, then the conclusion of Theorem 1 was proved in \[BER2\] (cf. also \[Han1\] and \[Z\]). A related notion is that of unique continuation at the boundary for holomorphic mappings. A unique continuation principle is said to hold for a class of mappings at a point $`p`$ if any mapping from this class which agrees with the constant mapping to infinite order at $`p`$ is necessarily constant. (Observe that, due to the nonlinear nature of mapping problems, a unique continuation principle for a class of mappings into a manifold does not imply that two mappings, in this class, which agree to infinite order are necessarily the same.) We shall not address this problem further here. We mention the papers \[ABR\], \[BR\], \[BL\], \[CR\], \[E1\], \[HK\], and refer the interested reader to these papers for further information.
The proof of Theorem 1 is based on Theorem 2 below, and a result from \[BER4\], alluded to above, which asserts that, under the assumptions of Theorem 1, the jet of $`f`$ at $`p_0`$ of any order is completely determined by its $`2k_0`$-jet. The proof of Theorem 1 is given at the end of §3.
Our second result, which is the basis for Theorem 1 above, states, loosely speaking, that given two suitably nondegenerate real hypersurfaces, there is a system of differential equations, which is complete in a certain sense, such that any CR diffeomorphism $`fMM^{}`$ must satisfy this system. The idea to look for such a differential system goes back to the work of E. Cartan and Chern–Moser mentioned above. The approach was further developed in the work of Han. To formulate the result more precisely, we need to fix some notation. Let us denote by $`J^k(M,M^{})_{(p,p^{})}`$ the space of $`k`$-jets at $`pM`$ of smooth mappings $`fMM^{}`$ with $`f(p)=p^{}M^{}`$. Given coordinate systems $`x=(x_1,\mathrm{},x_{2N1})`$ and $`x^{}=(x_1^{},\mathrm{},x_{2N1}^{})`$ on $`M`$ and $`M^{}`$ near $`p`$ and $`p^{}`$, respectively, there are natural coordinates $`\lambda ^k:=(\lambda _i^\beta )`$, where $`1i2N1`$ and $`\beta _+^{2N1}`$ with $`1|\beta |k`$, on $`J^k(M,M^{})_{(p,p^{})}`$ in which the $`k`$-jet at $`p`$ of a smooth mapping $`fMM^{}`$ is given by $`\lambda _i^\beta =(_x^\beta f_i)(p)`$, $`1|\beta |k`$ and $`1i2N1`$.
###### Theorem 2
Let $`M,M^{}^N`$ be smooth ($`C^{\mathrm{}}`$) real hypersurfaces. Assume that $`M`$ is $`k_0`$-nondegenerate at a point $`p_0`$. Let $`f^0MM^{}`$ be a smooth CR diffeomorphism. Then, for any multi-index $`\alpha _+^{2N1}`$ with $`|\alpha |=k_0^3+k_0^2+k_0+2`$ and any $`j=1,\mathrm{},2N1`$, there are smooth functions $`r_j^\alpha (\lambda ^k;x^{})(x)`$ on $`U`$, where $`k:=k_0^3+k_0^2+k_0+1`$ and $`UJ^k(M,M^{})_{(p_0,p_0^{})}\times M\times M^{}`$ is an open neighborhood of $`((_x^\beta f^0)(0),f^0(p_0),p_0)`$, such that
$$_x^\alpha f_j=r_j^\alpha (_x^\beta f;f),|\alpha |=k_0^3+k_0^2+k_0+2,j=1\mathrm{},2N1,$$
$`0.3`$
where $`1|\beta |k`$, for every smooth CR diffeomorphism $`fVM^{}`$, where $`VM`$ is some open neighborhood of $`p_0`$, with $`((_x^\beta f)(0),f(p_0),p_0)U`$. Moreover, the functions $`r_j^\alpha `$ are rational in $`\lambda ^kJ^k(M,M^{})_{(p_0,p_0^{})}`$; here, $`x=(x_1,\mathrm{},x_{2N1})`$ and $`x^{}=(x_1^{},\mathrm{},x_{2N1}^{})`$ are any local coordinate systems on $`M`$ and $`M^{}`$ near $`p_0`$ and $`f^0(p_0)`$, respectively, and $`f_i:=fx_i^{}`$.
Similar results for real-analytic hypersurfaces can also be found in \[Han1–2\] and \[Hay\]. The idea behind the proof of Theorem 2 is to consider the tangent mapping $`dfTMTM^{}`$ and derive differential equations for $`df`$ using properties of a sequence of invariant tensors (generalized Levi forms) which were developed in the author’s paper \[E3\]. The proof of Theorem 2 is given in §3.
We conclude this introduction by giving two applications of Theorems 1 and 2. For this, we need some more notation. A smooth real vector field $`X`$ on $`M`$ is called an infinitesimal CR automorphism if the local 1-parameter group of diffeomorphisms, $`\mathrm{exp}tX`$, generated by $`X`$ is a local group of CR diffeomorphisms (see e.g. \[BER2\] or \[S1–2\]). The set of infinitesimal CR automorphisms, defined near $`pM`$, forms a vector space over $``$ denoted by $`\text{aut}(M,p)`$. We shall give a sufficient condition on $`M`$ at a point $`p`$ for $`\text{dim}_{}\text{aut}(M,p)<\mathrm{}`$. A smooth real hypersurface $`M^N`$ is called (formally) holomorphically degenerate at $`pM`$, if there exists a formal holomorphic vector field
$$Y=\underset{j=1}{\overset{N}{}}a_j(z)_{z_j},$$
$`0.4`$
where the $`a_j(z)`$ are formal power series in $`zp`$, which is tangent to $`M`$, i.e. such that the Taylor series at $`p`$ of a defining function $`\rho (z,\overline{z})`$ for $`M`$ divides $`(Y\rho )(z,\overline{z})`$ in the ring of formal power series in $`(zp,\overline{z}\overline{p})`$. Being holomorphically nondegenerate (i.e. the opposite of being degenerate) at a point is a strictly weaker condition than that of being $`k`$-nondegenerate for some integer $`k`$. (See \[BER3, Chapter XI\] for a more detailed description of the relationship between the two notions). Also, recall that $`M`$ is said to be minimal at $`pM`$ (in the sense of Tumanov and Trepreau) if $`M`$ does not contain a complex hypersurface through $`p`$.
###### Theorem 3
Let $`M^N`$ be a smooth ($`C^{\mathrm{}}`$) real hypersurface which is holomorphically nondegenerate and minimal at $`p_0`$. Then,
$$\text{dim}_{}\text{aut}(M,p_0)(2N1)\left(\genfrac{}{}{0pt}{}{4N3}{2N2}\right)$$
$`0.5`$
A real-analytic hypersurface $`M`$ is said to be holomorphically degenerate at $`pM`$ if there exists a holomorphic vector field, i.e. a vector field of the form 0.4 with the $`a_j(z)`$ holomorphic, tangent to $`M`$ near $`p`$. This definition turns out to be equivalent to the one given in the smooth category above (i.e. using formal vector fields) for a real-analytic hypersurface (see \[BER3, Proposition 11.7.4\]). Stanton \[S2\] proved that $`\text{dim}_{}\text{hol}(M,p)<\mathrm{}`$ for a real-analytic hypersurface $`M`$, where $`\text{hol}(M,p)`$ denotes the subspace of $`\text{aut}(M,p)`$ consisting of those infinitesimal CR automorphisms which are real-analytic, if and only if $`M`$ is holomorphically nondegenerate at $`p`$. The corresponding statement (as well as results for higher codimensional real-analytic submanifolds) for $`\text{aut}(M,p)`$, with $`M`$ real-analytic, was proved in \[BER2\]. In contrast to the real-analytic case, the condition of (formal) holomorphic nondegeneracy is not necessary in Theorem 3. A real smooth hypersurface $`M`$ in $`^2`$ which is holomorphically degenerate and minimal at $`0`$, but everywhere Levi nondegenerate outside $`0`$ is given in \[BER3, Example 11.7.29\]. The fact that $`M`$ is Levi nondegenerate outside $`0`$ can be seen to imply (see the concluding remarks in §4.2) that $`\text{dim}_{}\text{aut}(M,0)`$ satisfies the bound in 0.5. However, if there exists a vector field
$$Y=\underset{j=1}{\overset{N}{}}a_j(z,\overline{z})_{z_j},$$
$`0.6`$
where the $`a_j(z,\overline{z})`$ are smooth functions whose restrictions to $`M`$ are CR, tangent to $`M`$ near $`p`$, then the arguments in \[S2\] easily show that $`\text{dim}_{}\text{aut}(M,0)=\mathrm{}`$. This discrepancy is addressed further in §4.2. The proof of Theorem 3 is given in §3.
For our final result, we shall denote by $`\text{Aut}(M,p)`$ the stability group of $`M`$ at $`pM`$, i.e. the group of germs at $`p`$ of local CR diffeomorphisms $`fVM`$, where $`VM`$ is some open neighborhood of $`p`$, with $`f(p)=p`$. If $`M`$ is $`k_0`$-nondegenerate at $`p_0`$, then, by Theorem 1, the jet mapping $`j_p^{2k_0}`$ sends $`\text{Aut}(M,p_0)`$ injectively into the jet group $`G^{2k_0}(^N)_{p_0}J^{2k_0}(^N,^N)_{(p_0,p_0)}`$, which consists of those jets that are invertible at $`p_0`$. We shall show that the elements of $`\text{Aut}(M,p_0)`$ depend smoothly on their $`2k_0`$-jets at $`p_0`$. More precisely, we have the following result.
###### Theorem 4
Let $`M^N`$ be a smooth ($`C^{\mathrm{}}`$) real hypersurface which is $`k_0`$-nondegenerate at $`p_0M`$. Then, the jet mapping
$$j^{2k_0}\text{Aut}(M,p_0)G^{2k_0}(^N)_{p_0}$$
is injective and, for every $`f^0\text{Aut}(M,p_0)`$, there exist an open neighborhood $`U_0`$ of $`j_{p_0}^{2k_0}(f^0)`$ in $`G^{2k_0}(^N)_{p_0}`$, an open neighborhood $`V_0`$ of $`p_0`$ in $`M`$, and a smooth ($`C^{\mathrm{}}`$) mapping $`FU_0\times V_0M`$ such that
$$F(j_{p_0}^{2k_0}(f),)=f,$$
$`0.7`$
for every $`f\text{Aut}(M,p_0)`$ with $`j_{p_0}^{2k_0}(f)U_0`$.
For real-analytic hypersurfaces, the result in Theorem 4 (with real-analytic dependence) was proved in \[BER1\]. (See \[BER4\] for the higher codimensional case; cf. also \[Z\].).
###### Remark Acknowledgement
The author would like to thank B. Lamel and D. Zaitsev for many helpful comments and discussions on a preliminary version of this paper.
## 1. Preliminaries
A real hypersurface $`M^N`$ inherits a CR structure $`𝒱:=T^{0,1}^NTM`$ from the ambient complex space $`^N`$. (Here, $`T^{0,1}^N`$ denotes the usual bundle of $`(0,1)`$ vectors in $`^N`$.) In this section, we shall consider abstract, not necessarily embedded (or integrable), CR structures. At the end of this section, we shall again specialize to embedded hypersurfaces, which substantially simplifies some of the computations in subsequent sections. The reader is referred to the concluding remarks in §4 for a brief discussion of the abstract case.
Let $`M`$ be a smooth ($`C^{\mathrm{}}`$) manifold with a CR structure $`𝒱TM`$. Recall that this means that $`𝒱`$ is a formally integrable subbundle (the commutator of two sections of $`𝒱`$ is again a section of $`𝒱`$) such that $`𝒱_p\overline{𝒱}_p=\{0\}`$ for every $`pM`$. Sections of the CR bundle are called CR vector fields. We shall denote by $`n1`$ the CR dimension of the CR manifold $`M`$, which by definition is the complex fiber dimension of $`𝒱`$, and we shall assume that the CR structure is of hypersurface type, i.e. that $`\text{dim}_{}M=2n+1`$. The reader is referred to \[BER3\] for an introduction to CR structures.
We define two subbundles $`T^0MT^{}MT^{}M`$ as follows
$$T^0M:=(𝒱\overline{𝒱})^{},T^{}M=𝒱^{},$$
$`1.1`$
where $`A^{}T^{}M`$, for a subset $`ATM`$, denotes the union over $`pM`$ of the set of covectors at $`p`$ annihilating every vector in $`A_p`$. Real nonvanishing sections of $`T^0M`$ are called characteristic forms and sections of $`T^{}M`$ are called holomorphic forms. Thus, characteristic forms are in particular holomorphic forms.
We shall give an alternative definition of $`k_0`$-nondegeneracy, as defined in the introduction, in terms of the intrinsic geometry of $`M`$. This definition appeared in \[E2\]. For a holomorphic form $`\omega `$, the Lie derivative with respect to a CR vector field $`X`$ is given by
$$_X\omega =X\mathrm{}d\omega ,$$
$`1.2`$
where $`\mathrm{}`$ denotes the interior product, or contraction, and $`d`$ denotes exterior differentiation. For $`pM`$, define the subspaces
$$T_p^0M:=E_0(p)E_1(p)\mathrm{}E_k(p)\mathrm{}T_p^{}M$$
$`1.3`$
by letting $`E_k(p)`$ be the linear span (over $``$) of the holomorphic covectors
$$(_{X_k}\mathrm{}_{X_1}\theta )(p),$$
$`1.4`$
where $`X_1,\mathrm{},X_k`$ range over all CR vector fields and $`\theta `$ over all characteristic forms near $`p`$. $`M`$ is called finitely nondegenerate at $`pM`$ if $`E_k(p)=T_p^{}M`$ for some $`k`$. More precisely, we say that $`M`$ is $`k_0`$-nondegenerate at $`p`$ if
$$E_{k_01}(p)E_{k_0}(p)=T_p^{}M.$$
$`1.5`$
For an argument showing that this definition coincides with that given for embedded hypersurfaces in the introduction, the reader is referred to \[BER3\] (see also \[E2\]). For each $`k`$, set
$$F_k(p)=\overline{𝒱}_pE_k(p)^{}.$$
$`1.6`$
It was shown in \[E3\] that the mapping
$$(X_1,\mathrm{},X_k,Y,\theta )(_{X_k}\mathrm{}_{X_1}\theta )(p),Y(p),$$
$`1.7`$
defines a multi-linear mapping
$$𝒱_p\times \mathrm{}𝒱_p\times F_{k1}(p)\times T_p^0M.$$
$`1.8`$
which is symmetric in the first $`k`$ positions. The tensor so defined for $`k=1`$ coincides with the classical Levi form, and the space $`F_1(0)`$ is the Levi nullspace.
Let us fix a distinguished point on $`M`$ denoted by $`0M`$. We choose a basis $`L_1,\mathrm{},L_n`$ of the sections $`C^{\mathrm{}}(U,\overline{𝒱})`$, where $`UM`$ is some sufficiently small neighborhood of $`0`$, adapted to the filtration
$$\overline{𝒱}_0=F_0(0)F_1(0)\mathrm{}F_k(0)\mathrm{}\{0\}$$
$`1.9`$
in the following way. Observe that the sequence of subspaces $`F_k(0)`$ stabilizes at a smallest subspace $`F_{k_0}(0)`$, which equals $`\{0\}`$ if and only if $`M`$ is $`k_0`$-nondegenerate at $`0`$. Let $`r_k=n\text{dim}_{}F_k(0)`$ and choose $`L_1,\mathrm{},L_n`$ so that $`L_{r_k+1}(0),\mathrm{}L_n(0)`$ spans $`F_k(0)`$ for $`k=0,1,\mathrm{},k_0`$. We shall use the following conventions for indices. For $`j=1,2,\mathrm{}`$, Greek indices $`\alpha ^{(j)},\beta ^{(j)}`$, etc., will run over the set $`\{1,\mathrm{},r_{j1}\}`$ and small Roman indices $`a^{(j)},b^{(j)}`$, etc., over $`\{r_{j1}+1,\mathrm{},n\}`$. Capital Roman indices $`A,B`$, etc., will run over $`\{1,\mathrm{},n\}`$.
Now, choose also a characteristic form $`\theta `$ on $`M`$ near $`0`$. We write
$$h_{\overline{A}_1\mathrm{}\overline{A}_kB}:=_{\overline{A}_k}\mathrm{}_{\overline{A}_1}\theta ,L_B,$$
$`1.10`$
where $`_{\overline{A}}:=_{L_{\overline{A}}}`$ and $`L_{\overline{A}}:=\overline{L_A}`$. Note that $`(h_{\overline{A}_1\mathrm{}\overline{A}_ka^{(k)}}(0))`$ represents the tensor defined by 1.7 relative to the bases $`L_{\overline{A}}(0)`$, $`L_{a^{(k)}}(0)`$, and $`\theta (0)`$ of $`𝒱_0`$, $`F_k(0)`$, and $`T_0^0M`$, respectively.
Let $`T`$ be a vector field near $`0`$ such that $`T,L_A,L_{\overline{A}}`$ form a basis for $`C^{\mathrm{}}(U,TM)`$. Let $`\theta ,\theta ^A,\theta ^{\overline{A}}`$ be the dual basis for $`C^{\mathrm{}}(U,T^{}M)`$. Note that, for each $`k=1,\mathrm{},k_0`$, the covectors $`\theta (0),\theta ^{\alpha ^{(k)}}(0)`$ form a basis for $`E_k(0)`$. For brevity, we introduce the functions
$$h_{\overline{A}_1\mathrm{}\overline{A}_k}:=_{\overline{A}_k}\mathrm{}_{\overline{A}_1}\theta ,T,$$
$`1.11`$
and also
$`R_{\overline{A}B}^C:=d\theta ^C,L_{\overline{A}}L_B,`$ $`R_{DB}^C:=d\theta ^C,L_DL_B`$ $`1.12`$
$`R_{\overline{A}}^C:=d\theta ^C,L_{\overline{A}}T,`$ $`R_B^C:=d\theta ^C,TL_B.`$
The following identity is useful.
###### Lemma 1.13
For any nonnegative integer $`k`$, and indices $`A_1,\mathrm{},A_k,C`$, $`D\{1,\mathrm{},n\}`$, the following identity holds
$$h_{\overline{A}_1\mathrm{}\overline{A}_k\overline{C}D}=L_{\overline{C}}h_{\overline{A}_1\mathrm{}\overline{A}_kD}+h_{\overline{A}_1\mathrm{}\overline{A}_kB}R_{\overline{C}D}^B+h_{\overline{A}_1\mathrm{}\overline{A}_k}h_{\overline{C}D}.$$
$`1.14`$
###### Demonstration Proof
Recall that $`_{\overline{A}_k}\mathrm{}_{\overline{A}_1}\theta `$ is a holomorphic 1-form and, by the definitions 1.10–11,
$$_{\overline{A}_k}\mathrm{}_{\overline{A}_1}\theta =h_{\overline{A}_1\mathrm{}\overline{A}_kD}\theta ^D+h_{\overline{A}_1\mathrm{}\overline{A}_k}\theta .$$
$`1.15`$
Here, and for the remainder of this paper, we use the summation convention which states that an index appearing in both a sub- and superscript is summed over; e.g. $`h_D\theta ^D=_Dh_D\theta ^D`$. We also have, by the definition of the interior product,
$$h_{\overline{A}_1\mathrm{}\overline{A}_k\overline{C}D}=d_{\overline{A}_k}\mathrm{}_{\overline{A}_1}\theta ,L_{\overline{C}}L_D.$$
$`1.16`$
The identity 1.14 follows by applying the exterior derivative $`d`$ to 1.15 and substituting in 1.16. ∎
Define $`\mathrm{}_0`$ to be the smallest integer $`\mathrm{}`$ for which
$$\{\begin{array}{cc}& h_{\overline{A}_1\mathrm{}\overline{A}_rD}(0)=0,A_1,\mathrm{}A_r,D\{1,\mathrm{},n\},r<\mathrm{}\hfill \\ & h_{\overline{A}_1^0\mathrm{}\overline{A}_{\mathrm{}}^0D^0}(0)0,\text{for some }A_1^0,\mathrm{}A_r^0,D^0\{1,\mathrm{},n\}.\hfill \end{array}$$
$`1.17`$
If no such $`\mathrm{}`$ exists then we set $`\mathrm{}_0=\mathrm{}`$. Observe that if $`M`$ is $`k`$-nondegenerate at $`0`$ for some $`k`$, then $`\mathrm{}_0k`$, but $`\mathrm{}_0<\mathrm{}`$ does not imply finite nondegeneracy. Also, note that, for any $`r\mathrm{}_0`$, the subspace $`F_{r1}(0)=\overline{𝒱}_0`$ and, hence, the indices $`a^r`$, $`b^r`$, etc., introduced above run over the whole index set $`\{1,\mathrm{},n\}`$.
(Also note, by the fact that $`L_A`$ is adapted to the filtration 1.9, that if $`\mathrm{}_0<\mathrm{}`$ then we can take $`D^0=1`$ in 1.17.)
###### Lemma 1.18
For any integer $`r2`$ and any integer $`j0`$ such that $`j+r\mathrm{}_0`$ and indices $`A_1,\mathrm{},A_r,C_1\mathrm{},C_j,D\{1,\mathrm{},n\}`$, the following holds
$`h_{\overline{A}_1\mathrm{}\overline{A}_{r1}\overline{A}_rD}(0)=`$ $`\left(L_{\overline{A}_r}h_{\overline{A}_1\mathrm{}\overline{A}_{r1}D}\right)(0)`$ $`1.19`$
$`\mathrm{}`$
$`\left(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_j}h_{\overline{A}_1\mathrm{}\overline{A}_{r1}\overline{A}_rD}\right)(0)=`$ $`\left(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_j}L_{\overline{A}_r}h_{\overline{A}_1\mathrm{}\overline{A}_{r1}D}\right)(0).`$
In particular,
$$h_{\overline{A}_1\overline{A}_2\mathrm{}\overline{A}_\mathrm{}_0D}(0)=\left(L_{\overline{A}_\mathrm{}_0}\mathrm{}L_{\overline{A}_2}h_{\overline{A}_1D}\right)(0).$$
$`1.20`$
###### Demonstration Proof
The first identity in 1.19 follows immediately by evaluating 1.14 at $`0`$ and using the definition of $`\mathrm{}_0`$. In particular, it follows that
$$\left(L_{\overline{A}_r}h_{\overline{A}_1\mathrm{}\overline{A}_{r1}D}\right)(0)=0$$
$`1.21`$
for any $`2r\mathrm{}_0`$. Now, the second identity in 1.19 follows by applying $`L_{\overline{C}_1}`$ to 1.14 and using 1.21. The conclusion of Lemma 1.18 follows by induction. ∎
Recall that $`M`$ is said to be of finite type at $`0M`$ if $`L_A,L_{\overline{A}}`$ and all their repeated commutators
$$[X_m,[X_{m1},\mathrm{}[X_2,X_1]\mathrm{}]],X_1,\mathrm{},X_m\{L_1,\mathrm{},L_n,L_{\overline{1}},\mathrm{},L_{\overline{n}}\},$$
$`1.22`$
evaluated at $`0`$ span $`T_0M`$. The commutator in 1.22 is said to have length $`m`$. (A commutator of length one is simply one of the vector fields $`L_A`$, $`L_{\overline{A}}`$.) If $`M`$ is of finite type at $`0`$, then it is said to be of type $`m_0`$ if $`m_0`$ is the smallest integer for which all commutators of the form 1.22 of lengths $`m_0`$ span $`T_0M`$. Define $`\mathrm{}_1`$ to be the smallest integer $`\mathrm{}`$ for which
$$\{\begin{array}{cc}& \theta ,[L_{\overline{A}_r},\mathrm{}[L_{\overline{A}_1},L_D]\mathrm{}](0)=0,A_1,\mathrm{}A_r,D\{1,\mathrm{},n\},r<\mathrm{}\hfill \\ & \theta ,[L_{\overline{A}_{\mathrm{}}^0},\mathrm{}[L_{\overline{A}_1^0},L_{D^0}]\mathrm{}](0)0,\text{for some }A_1^0,\mathrm{}A_r^0,D^0\{1,\mathrm{},n\}.\hfill \end{array}$$
$`1.23`$
If no such $`\mathrm{}`$ exists then we set $`\mathrm{}_1=\mathrm{}`$. Observe that $`\mathrm{}_1<\mathrm{}`$ implies that $`M`$ is of finite type $`m_0\mathrm{}_1+1`$ at $`0`$, but the converse is not true, i.e. $`M`$ can be of finite type at $`0`$ while $`\mathrm{}_1=\mathrm{}`$ .
###### Proposition 1.24
If either of the two integers $`\mathrm{}_0,\mathrm{}_1`$ is finite, then they are equal. Indeed, for any $`r\mathrm{}_0`$, it holds that
$$\theta ,[L_{\overline{A}_r},\mathrm{}[L_{\overline{A}_1},L_D]\mathrm{}](0)=h_{\overline{A}_1\mathrm{}\overline{A}_rD}(0),$$
$`1.25`$
for all $`A_1,\mathrm{}A_r,D\{1,\mathrm{},n\}`$. In particular, if $`M`$ is $`k`$-nondegenerate at $`0`$, then it is also of finite type $`k+1`$.
###### Demonstration Proof
Note that the first part of Proposition 1.24 clearly follows from 1.25. Hence, we shall only prove 1.25. For any $`1`$-form $`\xi `$ and vector fields $`X`$, $`Y`$, we have the following well known identity (see e.g. \[He\])
$$d\xi ,XY=X\xi ,YY\xi ,X\xi ,[X,Y].$$
$`1.26`$
Thus, for a holomorphic $`1`$-form $`\omega `$ on $`M`$, we obtain
$$\omega ,[L_{\overline{A}},L_d]=L_{\overline{A}}\omega ,L_D_{\overline{A}}\omega ,L_D.$$
$`1.27`$
By applying 1.27 with $`\omega =\theta `$, we deduce that
$$\theta ,[L_{\overline{A}_1},L_D]=h_{\overline{A}_1D}.$$
By Lemma 1.18 and the symmetry of the tensors $`h_{\overline{A}_1\mathrm{}\overline{A}_rD}(0)`$, we then deduce that
$$\left(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_s}\theta ,[L_{\overline{A}_1},L_D]\right)(0)=h_{\overline{C}_1\mathrm{}\overline{C}_s\overline{A}_1D}(0),\mathrm{\hspace{0.17em}0}s\mathrm{}_01,$$
$`1.28`$
where $`s=0`$ in 1.28 means $`\theta ,[L_{\overline{A}_1},L_D](0)=h_{\overline{A}_1D}(0)`$. By applying 1.27 with $`\omega =_{\overline{B}_j}\mathrm{}_{\overline{B}_1}\theta `$, we obtain
$$_{\overline{B}_j}\mathrm{}_{\overline{B}_1}\theta ,[L_{\overline{A}_1},L_D]=L_{\overline{A}_1}h_{\overline{B}_1\mathrm{}\overline{B}_jD}h_{\overline{A}_1\overline{B}_1\mathrm{}\overline{B}_jD}.$$
$`1.29`$
Hence, it follows from Lemma 1.18 and the symmetry of the $`h_{\overline{A}_1\mathrm{}\overline{A}_rD}(0)`$ that
$$\left(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_s}_{\overline{B}_j}\mathrm{}_{\overline{B}_1}\theta ,[L_{\overline{A}_1},L_D]\right)(0)=0,1j+s\mathrm{}_01.$$
$`1.30`$
Now, assume that
$$\begin{array}{c}(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_s}\theta ,[L_{\overline{A}_r},\mathrm{}[L_{\overline{A}_1},L_D])(0)=\hfill \\ \hfill h_{\overline{C}_1\mathrm{}\overline{C}_s\overline{A}_1\mathrm{}\overline{A}_rD}(0),\mathrm{\hspace{0.17em}1}s+r\mathrm{}_0,\end{array}$$
$`1.31`$
where $`s0`$ and the meaning for $`s=0`$ is analogous to 1.28, and
$$\begin{array}{c}\left(L_{\overline{C}_1}\mathrm{}L_{\overline{C}_s}_{\overline{B}_j}\mathrm{}_{\overline{B}_1}\theta ,[L_{\overline{A}_r},\mathrm{}[L_{\overline{A}_1},L_D]\mathrm{}]\right)(0)=0,\hfill \\ \hfill \mathrm{\hspace{0.17em}1}j+s+r\mathrm{}_0,\end{array}$$
$`1.32`$
where $`j,s0`$, for $`r=1,\mathrm{}R`$. Observe that we have proved this for $`R=1`$. Now, if $`R<\mathrm{}_0`$, then the 1.31 and 1.32 follows for all $`r=1,\mathrm{},R+1`$ by applying 1.27 and Lemma 1.18. The verification of this is straightforward and left to the reader. By induction, we deduce that 1.31 and 1.32 hold for $`r=1,\mathrm{},\mathrm{}_0`$. In particular, 1.25 holds for any $`r=1,\mathrm{},\mathrm{}_0`$. This completes the proof of Proposition 1.24.∎
So far, everything has been done with an arbitrary choice of basis $`T,L_A,L_{\overline{A}}`$, except that we chose the $`L_A`$ to be adapted to the filtration in 1.9 as explained above. We shall now use the fact that $`M`$ is embedded in $`^N`$ and choose a particular basis.
###### Lemma 1.33
Let $`M^N`$ be a smooth real hypersurface. Then, there is a basis $`T,L_A,L_{\overline{A}}`$ such that $`T`$ is real, the $`L_A`$ adapted to the filtration 1.9 as explained above, and
$$R_{\overline{A}B}^CR_{DB}^CR_{\overline{A}}^CR_B^C0,$$
$`1.34`$
for all indices $`A,B,C,D\{1,\mathrm{},n\}`$.
###### Remark Remark
Clearly, the conditions 1.34 are equivalent to $`d\theta ^C=0`$. Based on this observation, an alternative proof of Lemma 1.33 can be given by pulling back suitable coordinate functions from the ambient space.
###### Demonstration Proof
By making use of the identity 1.27, we conclude that
$`R_{\overline{A}B}^C=\theta ^C,[L_{\overline{A}},L_B],`$ $`R_{AB}^C=\theta ^C,[L_A,L_B],`$ $`1.35`$
$`R_{\overline{A}}^C=\theta ^C,[L_{\overline{A}},T],`$ $`R_A^C=\theta ^C,[T,L_A].`$
Hence, to prove the lemma, it is equivalent to showing that there is a basis $`T,L_A,L_{\overline{A}}`$ with $`T`$ real and $`L_A`$ adapted to the filtration 1.9 such that the $`L_A`$ commute, and $`[L_{\overline{A}},L_B]`$ and $`[L_A,T]`$ are multiples of $`T`$. The existence of such a basis, disregarding the adaption of the $`L_A`$ to the filtration, is well known (see e.g. \[BER3, Proposition 1.6.9\]). Since the adaption of the $`L_A`$ is a condition only at the point $`0`$, we may achieve this by a linear transformation with constant coefficients to any basis $`L_A`$. Such a transformation does not affect any commutator relations and, hence, the lemma follows.∎
In what follows, we shall assume that 1.34 holds.
## 2. A reflection identity for CR diffeomorphisms
Let $`M`$ be a smooth CR manifold as in the preceeding section, and let $`\widehat{M}`$ be another smooth CR manifold of the same dimension and CR dimension, with distinguished point $`\widehat{0}\widehat{M}`$. We shall denote corresponding objects on $`\widehat{M}`$ by using $`\widehat{}`$; e.g. $`\widehat{𝒱}T\widehat{M}`$ denotes the CR bundle on $`\widehat{M}`$, $`\widehat{T},\widehat{L}_A,\widehat{L}_{\overline{A}}`$ is a basis for $`C^{\mathrm{}}(\widehat{U},T\widehat{M})`$, where $`\widehat{U}`$ is some sufficiently small neighborhood of $`\widehat{0}\widehat{M}`$. We shall assume that both $`M`$ and $`\widehat{M}`$ are embeddable, locally near $`0M`$ and $`\widehat{0}\widehat{M}`$, as real hypersurfaces in $`^N`$. Hence, 1.35 holds on $`M`$ and analogous identities on $`\widehat{M}`$.
Assume that $`fM\widehat{M}`$ is a smooth CR diffeomorphism defined near $`0`$ in $`M`$ such that $`f(0)=\widehat{0}`$. Recall that a smooth mapping $`fM\widehat{M}`$ is called CR if $`f_{}(𝒱_p)\widehat{𝒱}_{f(p)}`$, where $`f_{}TMT\widehat{M}`$ denotes the tangent mapping or push forward, for every $`pM`$; a CR diffeomorphism is a diffeomorphism which is CR and whose inverse is also CR. In particular, if $`f`$ is a CR diffeomorphism then, for every $`pM`$ near $`0`$, $`f_{}(𝒱_p)=\widehat{𝒱}_{f(p)}`$. We introduce the smooth $`GL(^n)`$-valued function $`(\gamma _B^A)`$, and real-valued functions $`\xi ,\eta ^A`$ so that
$$f_{}(L_B)=\gamma _B^A\widehat{L}_A,f_{}(L_{\overline{B}})=\overline{\gamma _B^A}\widehat{L}_{\overline{A}},f_{}(T)=\xi \widehat{T}+\eta ^A\widehat{L}_A+\overline{\eta ^A}\widehat{L}_{\overline{A}}.$$
$`2.1`$
We can write 2.1 using matrix notation as
$$f_{}(T,L_B,L_{\overline{B}})=(\widehat{T},\widehat{L}_A,\widehat{L}_{\overline{A}})\left(\begin{array}{ccc}\xi & 0& 0\\ \eta ^A& \gamma _B^A& 0\\ \overline{\eta ^A}& 0& \overline{\gamma _B^A}\end{array}\right).$$
$`2.2`$
By duality, we then have
$$f^{}\left(\begin{array}{c}\widehat{\theta }\\ \widehat{\theta }^A\\ \widehat{\theta }^{\overline{A}}\end{array}\right)=\left(\begin{array}{ccc}\xi & 0& 0\\ \eta ^A& \gamma _B^A& 0\\ \overline{\eta ^A}& 0& \overline{\gamma _B^A}\end{array}\right)\left(\begin{array}{c}\theta \\ \theta ^B\\ \theta ^{\overline{B}}\end{array}\right).$$
$`2.3`$
The main technical result in this section is the following, which can be viewed as reflection identities for $`\gamma _E^D`$ and $`\eta ^D`$.
###### Theorem 2.4
If $`\widehat{M}`$ is $`k_0`$-nondegenerate at $`\widehat{0}\widehat{M}`$, then the following identities holds for any indices $`D,E\{1,\mathrm{},n\}`$,
$`\gamma _E^D=`$ $`r_E^D(\overline{L^J\gamma _A^C},\overline{L^I\xi };f),`$ $`2.5`$$`2.6`$
$`\eta ^D=`$ $`s^D(\overline{L^J\gamma _A^C},\overline{L^I\xi };f)`$
where
$`r_E^D(\overline{L^J\gamma _A^C},\overline{L^I\xi };q)(p),s^D(\overline{L^J\gamma _A^C},\overline{L^I\xi };q)(p)`$ $`2.7`$
are smooth functions which are rational in $`\overline{L^J\gamma _A^C}`$ and polynomial in $`\overline{L^I\xi }`$, the indices $`A`$, $`C`$ run over the set $`\{1,\mathrm{},n\}`$, and $`J`$, $`I`$ over all multi-indices with $`|J|k_01`$ and $`|I|k_0`$; here, $`(p,q)M\times \widehat{M}`$. Moreover, the functions in 2.7 depend only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$).
For the proof of Theorem 2.4, we shall make use of the following identity
$$df^{}\widehat{\omega },XY=d\widehat{\omega },f_{}Xf_{}Y,$$
$`2.8`$
which holds for any 1-form $`\widehat{\omega }`$ on $`\widehat{M}`$ and vector fields $`X`$, $`Y`$ on $`M`$. First, letting $`\widehat{\omega }=\widehat{\theta }`$, $`X=L_{\overline{A}}`$, and $`Y=L_B`$, we obtain
$$\xi h_{\overline{A}B}=\gamma _B^D\overline{\gamma _A^C}\widehat{h}_{\overline{C}D}.$$
$`2.9`$
Here, and in what follows, we abuse the notation in the following way. For a function $`\widehat{c}`$ defined on $`\widehat{M}`$, we use the notation $`\widehat{c}`$ to denote both the function $`\widehat{c}f`$ on $`M`$ and the function $`\widehat{c}`$ on $`\widehat{M}`$. It should be clear from the context which of the two functions is meant. For instance, in 2.9, we must have $`\widehat{h}_{\overline{C}D}=\widehat{h}_{\overline{C}D}f`$. By letting $`\widehat{\omega }=\widehat{\theta }^E`$, $`X=L_{\overline{A}}`$, and $`Y=L_B`$ in 3.1, we obtain
$$L_{\overline{A}}\gamma _B^E+\eta ^Eh_{\overline{A}B}=0.$$
$`2.10`$
Applying 2.8 with $`X=L_{\overline{A}}`$, $`Y=T`$, and $`\widehat{\omega }=\widehat{\theta }`$, we obtain
$$L_{\overline{A}}\xi +\xi h_{\overline{A}}=\xi \overline{\gamma _A^C}\widehat{h}_{\overline{C}}+\overline{\gamma _A^C}\eta ^D\widehat{h}_{\overline{C}D},$$
$`2.11`$
and with $`\widehat{\omega }=\widehat{\theta }^C`$, we obtain
$$L_{\overline{A}}\eta ^C+\eta ^Ch_{\overline{A}}=0.$$
$`2.12`$
To obtain 2.11 and 2.12, we have used the fact
$`d\widehat{\omega },\widehat{L}_{\overline{C}}\widehat{L}_{\overline{D}}=`$ $`\widehat{L}_{\overline{C}}\widehat{\omega },\widehat{L}_{\overline{D}}L_{\overline{D}}\widehat{\omega },\widehat{L}_{\overline{C}}\widehat{\omega },[\widehat{L}_{\overline{C}},\widehat{L}_{\overline{D}}]`$
$`=`$ $`\mathrm{\hspace{0.17em}0},`$
which holds for any holomorphic 1-form $`\widehat{\omega }`$ on $`\widehat{M}`$ by the formal integrability of the CR bundle $`\widehat{𝒱}`$. We apply 2.8 one last time, with $`\widehat{\omega }=\widehat{\theta }^C`$, $`X=T`$, and $`Y=L_A`$, to obtain
$$T\gamma _A^CL_A\eta ^C\eta ^C\overline{h_{\overline{A}}}=0.$$
$`2.13`$
###### Lemma 2.14
For any nonnegative integer $`k`$, and indices $`A_1,\mathrm{},A_k,B,C\{1,\mathrm{},n\}`$, the following identities hold
$$\begin{array}{c}L_{\overline{C}}\left(\gamma _B^D\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}\right)=\gamma _B^H\overline{\gamma _C^I}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_k\overline{I}H}\hfill \\ \hfill \gamma _B^H\overline{\gamma _C^I}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_k}\widehat{h}_{\overline{I}H}\eta ^H\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kH}h_{\overline{C}B}\end{array}$$
$`2.15`$
and
$$\begin{array}{c}L_{\overline{C}}\left(\eta ^D\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}\right)=\eta ^H\overline{\gamma _C^I}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_k\overline{I}H}\eta ^H\overline{\gamma _C^I}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_k}\widehat{h}_{\overline{I}H}\hfill \\ \hfill \eta ^H\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kH}h_{\overline{C}}.\end{array}$$
$`2.16`$
###### Demonstration Proof
We shall prove 2.15. Recall that $`\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}`$ in 2.15 denotes $`\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}f`$ by the convention introduced in §1. Hence,
$$L_{\overline{C}}(\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD})=\overline{\gamma _C^I}(\widehat{L}_{\overline{I}}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}),$$
$`2.17`$
where according to our convention $`\widehat{L}_{\overline{I}}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}=(\widehat{L}_{\overline{I}}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD})f`$, and we obtain
$$L_{\overline{C}}\left(\gamma _B^D\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}\right)=(L_{\overline{C}}\gamma _B^D)\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}+\gamma _B^D\overline{\gamma _C^I}(\widehat{L}_{\overline{I}}\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}).$$
$`2.18`$
Let us rewrite 2.10 as
$$L_{\overline{A}}\gamma _B^E=\eta ^Eh_{\overline{A}B}.$$
$`2.19`$
The identity 2.15 follows by substituting 2.19 in 2.18 and then applying Lemma 1.13.
The proof of the identity 2.16 is completely analogous. Expand the left hand side by the chain rule, and then substitute for the derivatives of $`\eta ^D`$ by using 2.12, and for the derivatives of $`\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}`$ by using Lemma 1.13. The details are left to the reader. ∎
The following two lemmas will be important in establishing Theorem 2.4.
###### Lemma 2.20
For any integer $`k0`$, and indices $`A_1,\mathrm{},A_k,B\{1,\mathrm{},n\}`$, the following identity holds
$$\begin{array}{c}\gamma _B^D\overline{\gamma _{A_1}^{C_1}}\mathrm{}\overline{\gamma _{A_k}^{C_k}}\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_kD}=r_{\overline{A}_1\mathrm{}\overline{A}_kB}(\overline{L^J\gamma _A^C};f)+\xi s_{\overline{A}_1\mathrm{}\overline{A}_kB}(\overline{L^J\gamma _A^C};f)\hfill \\ \hfill +\underset{l=1}{\overset{k1}{}}\gamma _E^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}t_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};f)+\underset{l=1}{\overset{k1}{}}\eta ^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}u_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f),\end{array}$$
$`2.21`$
where
$`r_{\overline{A}_1\mathrm{}\overline{A}_kB}(\overline{L^J\gamma _A^C};q)(p),s_{\overline{A}_1\mathrm{}\overline{A}_kB}(\overline{L^J\gamma _A^C};q)(p),`$ $`2.22`$
$`t_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};q)(p),u_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};q)(p)`$
are polynomials in $`\overline{L^J\gamma _A^C}`$, where $`A`$, $`C`$ run over the indices $`\{1,\mathrm{},n\}`$ and $`J=(J_1,\mathrm{},J_t)\{1,\mathrm{}n\}^t`$ for $`tk1`$, whose coefficients are smooth functions of $`(p,q)M\times \widehat{M}`$; here, we have used the notation $`L^J=L_{J_1}\mathrm{}L_{J_t}`$. Moreover, the functions in 2.22 depend only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$).
###### Demonstration Proof
We observe that 2.9 satisfies the conclusion of Lemma 2.20 for $`k=1`$. Assume that the conclusion of Lemma 2.20 holds for all integers $`k=1,\mathrm{}j1`$. Fix indices $`A_1,\mathrm{},A_{j1},B\{1,\mathrm{},n\}`$, choose an index $`A_j\{1,\mathrm{},n\}`$, and apply $`L_{\overline{A}_j}`$ to 2.21 with $`k=j1`$. The statement of the proposition for $`k=j`$ now follows by applying Lemma 2.14 and substituting for $`L_{\overline{A}_j}\xi `$ using 2.11. The proof is completed by induction on $`k`$. ∎
###### Remark Remark
In what follows, we shall use the notation $`r`$, $`s`$, $`t`$, and $`u`$ with varying sets of sub- and superscripts for “generic” functions which may be different from time to time.
###### Lemma 2.23
For any integer $`k0`$, and indices $`A_1,\mathrm{},A_k\{1,\mathrm{},n\}`$, the following identity holds
$$\begin{array}{c}\eta ^D\overline{\gamma _{A_1}^{C_1}}\mathrm{}\overline{\gamma _{A_k}^{C_k}}\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_kD}=r_{\overline{A}_1\mathrm{}\overline{A}_k}(\overline{L^J\gamma _A^C},L^{\overline{I}}\xi ;f)\hfill \\ \hfill +\underset{l=1}{\overset{k1}{}}\gamma _E^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}t_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};f)+\underset{l=1}{\overset{k1}{}}\eta ^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}u_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f),\end{array}$$
$`2.24`$
where
$`r_{\overline{A}_1\mathrm{}\overline{A}_k}(\overline{L^J\gamma _A^C},L^{\overline{I}}\xi ;q)(p),u_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};q)(p)`$ $`2.25`$
are polynomials in $`\overline{L^J\gamma _A^C}`$ and $`L^{\overline{I}}\xi `$ in the former case and in $`\overline{L^J\gamma _A^C}`$ in the latter, where $`A`$, $`C`$ run over the indices $`\{1,\mathrm{},n\}`$ and $`J=(J_1,\mathrm{},J_t)`$, $`I=(I_1,\mathrm{},I_{t+1})`$, with $`I_i,J_j\{1,\mathrm{}n\}`$, for $`tk1`$, whose coefficients are smooth functions of $`(p,q)M\times \widehat{M}`$; here, we have used the notation $`L^J=L_{J_1}\mathrm{},L_{J_t}`$ and $`L^{\overline{J}}=L_{\overline{J}_1}\mathrm{},L_{\overline{J}_t}`$. Moreover, the functions in 3.18 depend only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$).
###### Demonstration Proof
We start with equation 2.11 and proceed as in the proof of Lemma 2.20. We leave the details to the reader.∎
We are ready to prove Theorem 2.4.
###### Demonstration Proof of Theorem $`2.4`$
Using the fact that the matrices $`\overline{\left(\gamma _A^C\right)}`$ are invertible, we rewrite 2.21 and 2.24 as follows
$$\begin{array}{c}\gamma _B^D\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}+\underset{l=1}{\overset{k1}{}}\gamma _E^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}{}_{}{}^{}t_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};f)+\hfill \\ \hfill \underset{l=1}{\overset{k1}{}}\eta ^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}{}_{}{}^{}u_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f)={}_{}{}^{}r_{\overline{A}_1\mathrm{}\overline{A}_kB}^{}(\overline{L^J\gamma _A^C};f)+\\ \hfill {}_{}{}^{}s_{\overline{A}_1\mathrm{}\overline{A}_kB}^{}(\overline{L^J\gamma _A^C};f)\xi ,\end{array}$$
$`2.26`$
and
$$\begin{array}{c}\eta ^D\left(\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}+\underset{l=1}{\overset{k1}{}}\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}{}_{}{}^{}u_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f)\right)+\hfill \\ \hfill \underset{l=1}{\overset{k1}{}}\gamma _F^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}{}_{}{}^{}t_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_lF}(\overline{L^J\gamma _A^C};f)={}_{}{}^{}r_{\overline{A}_1\mathrm{}\overline{A}_k}^{}(\overline{L^J\gamma _A^C},\overline{L^I\xi };f).\end{array}$$
$`2.27`$
To prove the theorem, we must show that there are $`n`$ choices $`\underset{¯}{A}^j`$, where $`\underset{¯}{A}^j=A_1^j\mathrm{}A_{l_j}^j`$ with $`l_jk_0`$, such that the linear equations 2.26–2.27, with $`\underset{¯}{A}=\underset{¯}{A}^j`$ for $`j=1,\mathrm{},n`$ and $`B=1,\mathrm{}n`$, can be solved uniquely for $`\gamma _B^D`$ and $`\eta ^D`$ near $`0`$. For this it suffices to show that if, for some $`(v_B^D,v^D)^{n^2+n}`$,
$$\begin{array}{c}v_B^D\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}(0)+\underset{l=1}{\overset{k1}{}}v_E^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}(0){}_{}{}^{}t_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_lE}(\overline{L^J\gamma _A^C};f)(0)+\hfill \\ \hfill \underset{l=1}{\overset{k1}{}}v^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}(0){}_{}{}^{}u_{\overline{A}_1\mathrm{}\overline{A}_kB}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f)(0)=0\end{array}$$
$`2.28`$
and
$$\begin{array}{c}v^D\left(\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_kD}(0)+\underset{l=1}{\overset{k1}{}}\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}(0){}_{}{}^{}u_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_l}(\overline{L^J\gamma _A^C};f)(0)\right)+\hfill \\ \hfill \underset{l=1}{\overset{k1}{}}v_F^D\widehat{h}_{\overline{C}_1\mathrm{}\overline{C}_lD}(0){}_{}{}^{}t_{\overline{A}_1\mathrm{}\overline{A}_k}^{\overline{C}_1\mathrm{}\overline{C}_lF}(\overline{L^J\gamma _A^C};f)(0)=0,\end{array}$$
$`2.29`$
for all $`A_1,\mathrm{}A_k,B\{1,\mathrm{},n\}`$ and all $`kk_0`$, then $`v_B^D=v^D=0`$. To see this, note that 2.28–2.29, for $`k=1`$, implies directly that $`v^{\alpha ^{(2)}}=v_B^{\alpha ^{(2)}}=0`$; recall the convention introduced in §1 that the indices $`\alpha ^{(k+1)}`$ run over $`\{1,\mathrm{},r_k\}`$, where $`r_k=n\text{dim}F_k(0)`$ as introduced in §1, and the indices $`a^{(k+1})`$ run over the set $`\{r_k+1,\mathrm{}n\}`$. Thus, since $`\widehat{h}_{Aa^{(2)}}(0)=0`$, the equations 2.28–2.29 for $`k=2`$ reduce to
$$v_B^{a^{(2)}}\widehat{h}_{\overline{A}_1\overline{A}_2a^{(2)}}(0)=0,v^{a^{(2)}}\widehat{h}_{\overline{A}_1\overline{A}_2a^{(2)}}(0)=0,$$
$`2.30`$
which in turn implies $`v^{\alpha ^{(3)}}=v_B^{\alpha ^{(3)}}=0`$. Proceeding inductively, using at each step the fact that for any integers $`1j<kk_0`$,
$$\widehat{h}_{\overline{A}_1\mathrm{}\overline{A}_ja^{(k)}}(0)=0,$$
$`2.31`$
we conclude that the equations 2.28–2.29, for $`kk_0`$, imply $`v^{\alpha ^{(k_0+1)}}=v_B^{\alpha ^{(k_0+1)}}=0`$, which is equivalent to $`v^D=v_B^D=0`$ since $`r_{k_0+1}=n`$ for a $`k_0`$-nondegenerate CR manifold. This completes the proof of Theorem 2.4. ∎
## 3. Proofs of Theorems 1, 2, and 3
We begin with the proof of Theorem 2. For this proof, we shall need the following two lemmas. We shall keep the notation introduced in previous sections.
###### Lemma 3.1
For any indices $`D,E,F\{1,\mathrm{},n\}`$, multi-index $`J`$, and nonnegative integer $`k`$, we have the following
$`L_E\overline{L^J\gamma _F^D}=`$ $`r_{\overline{F}E}^{\overline{D}\overline{J}}\left(\overline{L^I\eta ^C}\right),`$ $`3.2`$$`3.3`$
$`L_E\overline{L^JT^k\eta ^D}=`$ $`s_E^{\overline{D}\overline{J}k}(\overline{L^IT^m\eta ^C},\overline{L^KT^{m+1}\eta ^C}),`$
where the functions in 3.2–3 are smooth functions which are rational in the arguments appearing inside the parentheses. The indices $`A`$, $`C`$ run over the set $`\{1,\mathrm{},n\}`$, and $`I`$, $`K`$, over all multi-indices with $`|I||J|`$, $`|K||J|1`$; the integer $`m`$ runs from $`0`$ to $`k`$. Moreover, the functions in 3.2–3 depend only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$).
###### Demonstration Proof
We shall use the following fact, which is an easy consequence of the commutator relations established in the proof of Lemma 1.32. For any vector field $`X\{L_A,L_{\overline{A}},T\}`$ and any multi-index $`J=(J_1,\mathrm{}J_{|J|})`$, we have
$$XL^J=L^JX+\underset{|K||J|1}{}c_KL^KT,$$
$`3.4`$
where the $`c_K`$ are smooth functions on $`M`$ (which depend on $`X`$ and $`J`$). To prove 3.2, we observe that, in view of 3.4, we have
$`L_E\overline{L^J\gamma _F^D}=`$ $`\overline{L_{\overline{E}}L^J\gamma _F^D}`$ $`3.5`$
$`=`$ $`\overline{L^JL_{\overline{E}}\gamma _F^D}+{\displaystyle \underset{|K||J|1}{}}c_K\overline{L^KT\gamma _F^D}.`$
The identity 3.2 follows from 2.10 and 2.13. The proof of 3.3 is similar, and left to the reader.∎
###### Lemma 3.6
For any index $`E\{1,\mathrm{},n\}`$, multi-indices $`J`$ and any nonnegative integer $`k`$, we have the following
$$L_E\overline{L^JT^k\xi }=s_E^{\overline{J}k}(\overline{L^K\gamma _A^C},\overline{L^IT^m\eta ^C},\overline{L^IT^m\xi },\overline{L^KT^{m+1}\xi },T^m\gamma _A^C,T^m\eta ^C;f)$$
$`3.7`$
where the function in 3.7 is a smooth functions which are rational in the arguments preceding the ;. The indices $`A`$, $`C`$ run over the set $`\{1,\mathrm{},n\}`$, and $`I`$, $`K`$ over all multi-indices with $`|I||J|`$, $`|K||J|1`$; the integer $`m`$ runs from $`0`$ to $`k`$. Moreover, the function in 3.7 depends only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$).
###### Demonstration Proof
We apply 3.4 as in the proof of Lemma 3.1 to deduce that that $`L_E\overline{L^JT^k\xi }`$ is linear in $`\overline{L^IT^m\xi }`$, $`\overline{L^KT^{m+1}\xi }`$, and $`\overline{L^IT^mL_{\overline{E}}\xi }`$. To evaluate the latter term, we make use of 2.11 and 2.13 to deduce that $`\overline{L^JT^kL_{\overline{E}}\xi }`$ is polynomial in $`\overline{L^IT^m\xi }`$, $`\overline{L^IT^m\eta ^C}`$, $`\overline{L^K\gamma _A^C}`$, and $`L^{\overline{I}}T^m\gamma _A^C`$. Finally, we commute $`L^{\overline{I}}`$ and $`T^m`$ using 3.4, and then use 2.10 to conclude that $`L^{\overline{I}}T^m\gamma _A^C`$ is a linear function of $`T^m\gamma _A^C`$ and $`T^m\eta ^C`$. Summing up, we obtain 3.7. This completes the proof of Lemma 3.6.∎
The following argument is inspired by the paper \[Han2\]. We shall say, for a function $`u`$ on $`M`$, that $`uC_p^a`$ if
$$u=r(\overline{L^I\gamma _A^C},\overline{L^IT^m\eta ^C},\overline{L^IT^m\xi },L^NT^n\gamma _A^C,L^NT^n\eta ^C;f),$$
$`3.8`$
where the function in 3.8 is a smooth functions which is rational in the arguments preceeding the ;. The indices $`A`$, $`C`$ run over the set $`\{1,\mathrm{},n\}`$. The multi-indices $`I`$, $`N`$ and the nonnegative integers $`m`$, $`n`$ run over all multi-indices with $`|I|+mp`$, $`|N|+na`$. Moreover, the function in 3.8 should depend only on $`M`$ and $`\widehat{M}`$ (and not on the mapping $`f`$). Similarly, we shall say that $`uC_{p,q}^{a,b}`$ if 3.8 holds with the additional condition that $`mq`$, $`nb`$. Observe that by Lemma 2.30 (and the reality of $`\xi `$), we have
$$\gamma _F^D,\eta ^C,\xi C_{k_0,0}^{1,1},$$
$`3.9`$
where the negative ones in the superscript signify that no terms involving $`L^N\gamma _A^C`$ or $`L^N\eta ^C`$ appear. Recall that $`k_0`$ is the order of nondegeneracy of $`\widehat{M}`$. By 2.10–12, we obtain
$$L^{\overline{J}}\gamma _F^D,L^{\overline{J}}\eta ^CC_{k_0,0}^{1,1}$$
$`3.10`$
for any multi-index $`J`$. By applying $`L_E`$ to e.g. the equations for $`L^{\overline{J}}\gamma _F^D`$ and using Lemmas 3.1 and 3.6, we conclude that
$$L_EL^{\overline{J}}\gamma _F^D=r_E^{\overline{S}}(\overline{L^K\gamma _A^C},\overline{L^IT^m\eta ^C},\overline{L^IT^m\xi },\gamma _A^C,\eta ^C;f),$$
$`3.11`$
where $`|K|k_01`$, $`|I|+mk_0`$, and $`m1`$. By substituting for $`\gamma _A^C`$ and $`\eta ^C`$ in 3.11 using the equations provided by 3.10, we conclude that $`L_E\gamma _F^DC_{k_0,1}^{1,1}`$. We obtain a similar equation for $`L_EL^{\overline{J}}\eta ^C`$. Hence, we obtain
$$L_EL^{\overline{J}}\gamma _F^D,L_EL^{\overline{J}}\eta ^CC_{k_0,1}^{1,1}.$$
$`3.12`$
Next, by applying $`L_{\overline{F}}`$ to the equations for $`L_EL^{\overline{J}}\gamma _F^D`$ and $`L_EL^{\overline{J}}\eta ^C`$, provided by 3.12, we obtain
$$L_{\overline{F}}L_EL^{\overline{J}}\gamma _F^D,L_{\overline{F}}L_EL^{\overline{J}}\eta ^CC_{k_0+1,1}^{1,1}.$$
$`3.13`$
Similarly, repeated application of $`L_{\overline{F}_1}`$, $`L_{\overline{F}_2}`$, $`\mathrm{},L_{\overline{F}_k}`$ yields
$$L_{\overline{F}_k}\mathrm{}L_{\overline{F}_1}L_EL^{\overline{J}}\gamma _G^D,L_{\overline{F}_k}\mathrm{}L_{\overline{F}_1}L_EL^{\overline{J}}\eta ^CC_{k_0+k,1}^{1,1}.$$
$`3.14`$
Hence, by taking linear combinations of $`L_{\overline{F}_k}\mathrm{}L_{\overline{F}_i}L_EL_{\overline{F}_{i+1}}\mathrm{}L_{\overline{F}_k}L^{\overline{J}}`$, we deduce that
$$[\mathrm{}[L_E,L_{\overline{F}_1}],\mathrm{},L_{\overline{F}_k}]L^{\overline{J}}\gamma _G^D,[\mathrm{}[L_E,L_{\overline{F}_1}],\mathrm{},L_{\overline{F}_k}]L^{\overline{J}}\eta ^CC_{k_0+k,1}^{1,1}.$$
$`3.15`$
Let $`\mathrm{}_0k_0`$ be the integer provided by Lemma 1.24 for which
$$[\mathrm{}[L_E,L_{\overline{F}_1}],\mathrm{},L_{\overline{F}_k}]=aT,$$
$`3.16`$
for some function $`a`$ with $`a(0)0`$. Then, 3.15 implies, in particular, that
$$T\gamma _G^C,T\eta ^CC_{k_0+\mathrm{}_0,1}^{1,1}.$$
$`3.17`$
Before proceeding, we shall need the following result on commutators of differential operators, which seems to be of independent interest.
###### Proposition 3.18
Let $`\mathrm{}_0`$ be the smallest integer for which 1.23 holds. Then, for any multi-index $`J`$, integer $`k1`$, and index $`F\{1,\mathrm{},n\}`$ there exist smooth functions $`a^{E_1\mathrm{}E_m\overline{F}_1\mathrm{}\overline{F}_s}`$, $`b_s^{E_1\mathrm{}E_m}`$ such that
$$\begin{array}{c}\underset{m=1}{\overset{|J|+k}{}}\underset{s=0}{\overset{m\mathrm{}_0}{}}a^{E_1\mathrm{}E_m\overline{F}_1\mathrm{}\overline{F}_s}[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_s}]=L^JT^k.\hfill \end{array}$$
$`3.19`$
and
$$\underset{m=1}{\overset{|J|+k}{}}\underset{s=0}{\overset{k}{}}b_s^{E_1\mathrm{}E_m}\underset{\text{length }s\text{ }}{\underset{}{[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}}],L_{\overline{F}}]\mathrm{},L_{\overline{F}}]}}=(h_{\overline{F}1})^pL^JT^k,$$
$`3.20`$
where $`p=k+|J||J|_1+1`$ and $`|J|_1`$ denotes the number of occurences of the index $`1`$ in the multi-index $`J`$; here, the length of the commutator $`[\mathrm{}[X,Y_1],Y_2]\mathrm{},Y_s]`$ is $`s`$.
###### Demonstration Proof
In this proof, we shall use the following notation to simplify the notation,
$$C_{E_1\mathrm{}E_m,\overline{F}_1\mathrm{}\overline{F}_s}:=[\mathrm{}[[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_s}],$$
where $`C_{E_1\mathrm{}E_m}`$ is understood to mean $`L_{E_1}\mathrm{}L_{E_m}`$. Using bilinearity of the commutator and the identity
$$[AB,C]=A[B,C]+[A,C]B,$$
$`3.21`$
a straightforward induction shows that
$$\begin{array}{c}[\mathrm{}[[L_{E_1}L_{E_2}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_s}]=\hfill \\ \hfill \underset{(\underset{¯}{i},\underset{¯}{j})P_2(s)}{}C_{E_1,\overline{F}_{i_1}\mathrm{}L_{\overline{F}_{i_{sl}}}}C_{E_2\mathrm{}E_m,\overline{F}_{j_1}\mathrm{}\overline{F}_{j_l}},\end{array}$$
$`3.22`$
where $`P_2(s)`$ denotes the set of all partitions of $`\{1,\mathrm{},s\}`$ into two disjoint increasing sequences $`\underset{¯}{i}=(i_1,\mathrm{},i_{sl})`$, $`1i_1<\mathrm{}i_{sl}s`$, and $`\underset{¯}{j}=(j_1,\mathrm{},j_l)`$, $`1j_1<\mathrm{}<j_ls`$ for $`l=0,\mathrm{}s`$. (Of course, for e.g. $`l=0`$ the partition is understood to be the trivial one $`\underset{¯}{i}=(1,\mathrm{},s)`$ and $`\underset{¯}{j}=\mathrm{}`$.) Similarly, if we denote by $`P_m(s)`$ the set of all partitions of $`\{1,\mathrm{},s\}`$ into $`m`$ disjoint increasing sequences $`\underset{¯}{i^t}=(i_1^t,\mathrm{},i_{s_t}^t)`$, $`1i_1^t<\mathrm{}i_{s_t}^ts`$, $`t=1,\mathrm{}m`$, and $`s_t=s`$ (allowing empty sequences), then we have
$$\begin{array}{c}[\mathrm{}[[L_{E_1}L_{E_2}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_s}]=\hfill \\ \hfill \underset{(\underset{¯}{i^1},\mathrm{},\underset{¯}{i^m})P_m(s)}{}C_{E_1,\overline{F}_{i_1^1}\mathrm{}L_{\overline{F}_{i_{s_1}^1}}}\mathrm{}C_{E_m,\overline{F}_{i_1^m}\mathrm{}\overline{F}_{i_{s_m}^m}},\end{array}$$
$`3.23`$
Observe that $`C_{E,\overline{F}_1\mathrm{}\overline{F}_s}=a_{E,\overline{F}_1\mathrm{}\overline{F}_s}T`$, for some function $`a_{E,\overline{F}_1\mathrm{}\overline{F}_s}`$ such that
$$a_{E,\overline{F}_1\mathrm{}\overline{F}_s}(0)=0,s<\mathrm{}_0,\text{and}a_{E,\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0}(0)=h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0E}(0)0,$$
for some choice of $`F_1,\mathrm{},F_\mathrm{}_0`$. Hence, with $`s=\mathrm{}_0`$ we obtain, by 3.23 and Lemma 1.24,
$$\begin{array}{c}[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_\mathrm{}_0}]=_{l=1}^m(h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0E_l}+o(1))L_{E_1}\mathrm{}\widehat{L_{E_l}}\mathrm{}L_{E_m}T\hfill \\ \hfill +\underset{\genfrac{}{}{0pt}{}{|K|+p=m}{|K|m2}}{}b_{Kp}L^KT^p+\underset{|K|m2}{}c_KL^KT,\end{array}$$
$`3.24`$
where $`b_{kp}(0)=0`$, $`\widehat{L_{E_l}}`$ means that factor is omitted, and $`o(1)`$ denotes a function vanishing at $`0`$; the last sum in 3.24 arises from arranging (by commuting) so that the vector field $`T`$ comes last in the first sum. Recall, from §1, that for each index $`\alpha ^{(\mathrm{}_0)}\{1,\mathrm{},r_\mathrm{}_0\}`$ there exist $`F_1,\mathrm{},F_\mathrm{}_0`$ so that $`h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0\alpha ^{(\mathrm{}_0)}}(0)0`$. For this argument, we only need the fact that there exist $`F_1,\mathrm{},F_\mathrm{}_0`$ so that $`h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_01}(0)0`$. We choose $`F_1,\mathrm{},F_\mathrm{}_0`$ to be minimal, in the lexicographical ordering ($`A_1\mathrm{}A_s<B_1\mathrm{}B_s`$ if, for some $`rs`$, $`A_iB_i`$ for $`i<r`$, and $`A_r<B_r`$), with this property. Setting $`E_1=\mathrm{}E_m=1`$, we observe that we can solve for $`L_1^{m1}T`$ in 3.24. Setting $`E_1=\mathrm{}E_{m1}=1`$ and $`L_{E_m}=L_E`$ with $`E2`$, we can then solve for $`L_1^{m1}L_ET`$. Proceeding inductively, we see that we can solve for any $`L^JT`$, with $`|J|=m1`$, in terms of
$$[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_\mathrm{}_0}],b_{Kp}L^KT^p,L^KT,$$
where $`K`$ runs over multi-indices with $`|K|m2`$, and $`p`$ over positive integers such that $`|K|+p=m`$, and each $`b_{kp}(0)=0`$. Next, letting $`s=2\mathrm{}_0`$ and $`F_{\mathrm{}_0+l}=F_l`$ for $`l=1,\mathrm{}\mathrm{}_0`$, we obtain (by also using that $`h_{\overline{F}_1\mathrm{}\overline{F}_sE}`$ is symmetric in the first $`s`$ indices $`F_1,\mathrm{}F_s`$)
$$\begin{array}{c}[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_{2\mathrm{}_0}}]=_{l=1}^ma_{E_j\overline{F}_1\mathrm{}\overline{F}_{2\mathrm{}_0}}L_{E_1}\mathrm{}\widehat{L_{E_l}}\mathrm{}L_{E_m}T\hfill \\ \hfill +c_{\overline{F}}\underset{1l_1<l_2m}{}(h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0E_{l_1}}h_{\overline{F}_1\mathrm{}\overline{F}_\mathrm{}_0E_{l_2}}+o(1))L_{E_1}\mathrm{}\widehat{L_{E_{l_1}}}\mathrm{}\widehat{L_{E_{l_2}}}\mathrm{}L_{E_m}T^2\\ \hfill +\underset{\genfrac{}{}{0pt}{}{|K|+p=m}{|K|m3}}{}o(1)L^KT^p+\underset{\genfrac{}{}{0pt}{}{|K|+pm1}{p=1,2}}{}c_{Kp}L^KT^p,\end{array}$$
$`3.25`$
where $`c_{\overline{F}}`$ is some combinatorial factor ($`>0`$) which depends on the minimal set of indices $`F_1,\mathrm{},F_\mathrm{}_0`$. Using the fact that we have already solved for the $`L^JT`$, $`|J|=m1`$, in terms of $`b_{K2}L^KT^2`$ where $`b_{K2}(0)=0`$, a similar argument to the one used above shows that we can solve for each $`L^JT^2`$, $`|J|=m2`$, in terms of
$$[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_{2\mathrm{}_0}}],o(1)L^KT^p,L^QT,L^KT^2$$
where $`K`$, $`Q`$, runs over multi-indices with $`|K|m3`$, $`|Q|m2`$, and $`p`$ over positive integers such that $`|K|+p=m`$. Proceeding by induction over $`k`$ (with the total order $`m`$ fixed), we eventually find that we can solve completely for $`T^m`$ in terms of
$$[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_{m\mathrm{}_0}}],L^KT^p,$$
with $`|K|+pm1`$. Substituting back, we obtain
$$\begin{array}{c}\underset{n=0}{\overset{k\mathrm{}_0}{}}a^{E_1\mathrm{}E_m\overline{F}_1\mathrm{}\overline{F}_n}[\mathrm{}[L_{E_1}\mathrm{}L_{E_m},L_{\overline{F}_1}],L_{\overline{F}_2}]\mathrm{},L_{\overline{F}_n}]=L^JT^k\hfill \\ \hfill +\underset{|K|+pm1}{}c_{Kp}L^KT^p,\end{array}$$
where $`m=|J|+k`$. The proof of 3.19 is completed by a simple induction on the total degree $`m`$.
For the proof of 3.20, we proceed analogously by first setting $`s=1`$ and $`E_1=\mathrm{}=E_m=1`$ in 3.23. We find that
$$mh_{\overline{F}1}L_1^{m1}T=C_{E_1\mathrm{}E_m,\overline{F}}+\underset{|K|m2}{}c_KL^KT.$$
$`3.26`$
Next, with $`E_1=\mathrm{}=E_{m1}=1`$ and $`E_m=E`$, we obtain
$$(m1)h_{\overline{F}1}L_1^{m2}L_ET+h_{\overline{F}E}L_1^{m1}T=C_{E_1\mathrm{}E_m,\overline{F}}+\underset{|K|m2}{}c_K^{}L^KT.$$
$`3.27`$
Thus, multiplying by $`h_{\overline{F}1}`$ and using 3.26, we obtain 3.20 for a multi-index $`J=(1,\mathrm{},1,E)\{1,\mathrm{},n\}^{m1}`$ and $`k=1`$. Similarly, we obtain 3.20 for arbitrary multi-indices $`J`$ and $`k=1`$. Proceeding inductively, setting $`s=2,3,\mathrm{}k`$, using 3.23, and multiplying through by a suitable power of $`h_{\overline{F}1}`$ to apply the results obtained in previous steps, we arrive at 3.20. The details are left to the reader. This completes the proof of Proposition 3.18. ∎
To complete the proof of Theorem 2, let us observe the following schematic diagram which describes the action of applying the operators $`L_E`$ to elements in $`C_{q+k,q}^{1,1}`$
$$\begin{array}{c}C_{q+k,q}^{1,1}\mathrm{@}>L_{E_1}>>C_{q+k,q+1}^{q,q}\mathrm{@}>L_{E_2}>>C_{q+k,q+2}^{q+1,q+1}\mathrm{@}>L_{E_3}>>\mathrm{}\mathrm{@}>L_{E_k}>>\hfill \\ \hfill \mathrm{@}>L_{E_k}>>C_{q+k,q+k}^{q+k1,q+k1}\mathrm{@}>L_{E_{k+1}}>>C_{q+k,q+k}^{q+k,q+k}\mathrm{@}>L_{E_{k+2}}>>C_{q+k,q+k}^{q+k+1,q+k}\mathrm{@}>L_{E_{k+3}}>>\mathrm{}\end{array}$$
$`3.28`$
The verification of the diagram is straightforward using Lemmas 3.1 and 3.6, and the details are left to the reader. Similarly, we have
$$C_{p,q}^{a+k,a}\mathrm{@}>L_{\overline{F}_1}>>C_{p+1,q}^{a+k,a+1}\mathrm{@}>L_{\overline{F}_2}>>\mathrm{}\mathrm{@}>L_{\overline{F}_k}>>C_{p+k,q}^{a+k,a+k}\mathrm{@}>L_{\overline{F}_{k+1}}>>C_{p+k+1,q}^{a+k,a+k}\mathrm{@}>L_{\overline{F}_{k+2}}>>\mathrm{}$$
$`3.29`$
We claim that the following holds for any multi-index $`J`$ and nonnegative integer $`k`$,
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DC_{k_0+m\mathrm{}_0,\mathrm{min}(k_0,m)}^{1,1},$$
$`3.30`$
where $`m=|J|+k`$. Observe that 3.30 holds for $`m=1`$ by 3.12 and 3.17. We shall prove 3.30 by induction on $`m`$. Thus, assume that 3.30 holds for all $`mm_01`$. By Proposition 3.18, we can produce the differential operator $`L^JT^k`$ by taking linear combinations of operators of the form $`L^{\overline{P}}L^QL^{\overline{R}}`$, where $`|P|+|R|m_0\mathrm{}_0`$, $`|Q|m_0`$, and $`m_0=|J|+k`$. Applying first $`L^QL^{\overline{R}}`$ to e.g. $`\gamma _F^D`$ we conclude, using 3.10 and the diagram 3.28, that $`L^QL^{\overline{R}}\gamma _F^DC_{k_0,\mathrm{min}(k_0,m_0)}^{m_01,\mathrm{min}(k_0,m_01)}`$. By applying $`L^{\overline{P}}`$ to the equation for $`L^QL^{\overline{R}}\gamma _F^D`$ and using the diagram 3.29, we obtain $`L^JT^k\gamma _F^DC_{k_0+m_0\mathrm{}_0,\mathrm{min}(k_0,m_0)}^{m_01,m_01}`$. The conclusion $`L^JT^k\gamma _F^DC_{k_0+m_0\mathrm{}_0,\mathrm{min}(k_0,m_0)}^{1,1}`$ follows by using the induction hypothesis to substitute for the $`L^IT^m\gamma _A^C`$ and $`L^IT^m\eta ^C`$, with $`|I|+mm_01`$, that appear in the equation for $`L^JT^k\gamma _F^D`$. By applying the same argument to $`\eta ^D`$, we conclude that 3.30 holds for $`m=m_0`$ and, hence, for all $`m`$ by induction. This proves the claim. In particular, we then have
$$T^k\gamma _F^D,T^k\eta ^DC_{k_0+k\mathrm{}_0,\mathrm{min}(k_0,k)}^{1,1}.$$
$`3.31`$
By applying $`L^J`$ to these equations and using 3.28, we deduce that
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DC_{k_0+k\mathrm{}_0,q(J,k)}^{a(J,k),q^{}(J,k)},$$
$`3.32`$
where
$`a(J,k)=`$ $`\mathrm{min}(k_0,k)+|J|1`$ $`3.33`$$`3.34`$
$`q(J,k)=`$ $`\mathrm{min}(k_0+k\mathrm{}_0,\mathrm{min}(k_0,k)+|J|),`$
and
$$q^{}(J,k)=\mathrm{min}(k_0+k\mathrm{}_0,\mathrm{min}(k_0,k)+|J|1).$$
Observe that 3.32 implies
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DC_{k_0+k_0\mathrm{}_0,k_0+k_0\mathrm{}_0}^{|J|+k1,k_0+k_0\mathrm{}_0},k\mathrm{\hspace{0.25em}0}kk_0.$$
$`3.35`$
Now, from 3.4 it follows that $`L^RT^k\xi =T^kL^R\xi `$ modulo terms of the form $`T^kL^S\xi `$ with $`|S|<|R|`$. Hence, by applying 2.11–13, we conclude that $`L^RT^k\xi `$ is a polynomial in $`T^m\xi =\overline{T^m\xi }`$, $`\overline{T^m\eta ^C}`$, $`L^K\gamma _A^C`$, and $`L^IT^{m1}\eta ^C`$, where $`mk`$, $`|K||R|1`$, and $`|I||R|`$. It follows that we also have
$$L^JT^k\xi C_{k_0+k_0\mathrm{}_0,k_0+k_0\mathrm{}_0}^{|J|+k1,k_0+k_0\mathrm{}_0},k\mathrm{\hspace{0.25em}0}kk_0.$$
$`3.36`$
Let us introduce the new class $`D_{p,q}^{a,b}`$ consisting of functions $`u`$ for which there is an equation
$$u=r(\overline{L^I\gamma _A^C},\overline{L^IT^m\eta ^C},L^NT^n\xi ,L^NT^n\gamma _A^C,L^NT^n\eta ^C;f),$$
$`3.37`$
where $`|I|+mp`$, $`mq`$, $`|N|+na`$, $`nb`$, and the function $`r`$ is rational in the arguments preceeding the “;” and only depends on $`M`$ and $`\widehat{M}`$. By the above remarks concerning $`L^RT^k\xi `$ and 3.31, we have
$$T^k\gamma _F^D,T^k\eta ^DD_{k_0+k\mathrm{}_0,\mathrm{min}(k_0,k)}^{\mathrm{min}(k_0,k),\mathrm{min}(k_0,k)}.$$
$`3.38`$
Since equations of the form 3.37 do not involve terms of the form $`\overline{L^IT^m\xi }`$, we obtain a different diagram describing the action of $`L_E`$ on the classes $`D_{p,q}^{a,b}`$, namely
$$\begin{array}{c}D_{q+k,q}^{a,b}\mathrm{@}>L_{E_1}>>D_{q+k,q+1}^{a+1,b}\mathrm{@}>L_{E_2}>>\mathrm{}\mathrm{@}>L_{E_k}>>D_{q+k,q+k}^{a+k1,b}\mathrm{@}>L_{E_{k+1}}>>\hfill \\ \hfill \mathrm{@}>L_{E_{k+1}}>>D_{q+k,q+k}^{a+k,b}\mathrm{@}>L_{E_{k+2}}>>D_{q+k,q+k}^{a+k+1,b}\mathrm{@}>L_{E_{k+3}}>>\mathrm{}\end{array}$$
$`3.39`$
By 3.38–39, we deduce
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DD_{k_0+k\mathrm{}_0,\mathrm{min}(k_0+k\mathrm{}_0,k_0+|J|)}^{k_0+|J|,k_0},kkk_0+1.$$
$`3.40`$
We have the following technical, but important, lemma.
###### Lemma 3.41
For any multi-index $`J`$, and nonnegative integer $`kk_0`$, we have
$$L^JT^k\gamma _F^D,L^JT^k\eta ^D,L^JT^k\xi C_{k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0,k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0}^{1,1}.$$
$`3.42`$
###### Demonstration Proof
By 3.35–36, we have
$$L^JT^k\gamma _F^D,L^JT^k\eta ^D,L^JT^k\xi C_{k_0+k_0\mathrm{}_0,k_0+k_0\mathrm{}_0}^{|J|+k1,k_0+k_0\mathrm{}_0}.$$
$`3.43`$
Observe that 3.43 reduces the total order of the unconjugated terms by at least one. Now, in the equations given by 3.43, there may appear terms of the form $`L^{I^1}T^{k_1}\gamma _A^C`$, $`L^{I^1}T^{k_1}\eta ^C`$, where $`|I^1|+k_1|J|+k1`$, and $`k_1k_0+k_0\mathrm{}_0`$. For those term with $`k_0+1k_1k_0+k_0\mathrm{}_0`$, we may apply 3.40 to deduce that
$$L^{I^1}T^{k_1}\gamma _A^C,L^{I^1}T^{k_1}\eta ^CD_{k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0,\mathrm{min}(k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0,k_0+|I^1|)}^{k_0+|I^1|,k_0}.$$
$`3.44`$
Note that, since $`k_1k_0+1`$ and $`|I^1|+k_1|J|+k1`$, we have $`k_0+|I^1||J|+k2`$. For those terms $`L^{I^1}T^{k_1}\gamma _A^C`$, $`L^{I^1}T^{k_1}\eta ^C`$ with $`k_1k_0`$, we may apply 3.35 again. In any case, we have reduced the total order of the unconjugated terms by two. In the equations given by 3.44, there may appear terms of the form $`L^{I^2}T^{k_2}\gamma _A^C`$, $`L^{I^2}T^{k_2}\eta ^C`$, and also $`L^{I^2}T^{k_2}\xi `$, where $`|I^2|+k_2k_0+|I_1||J|+k2`$, and $`k_2k_0`$. We again substitute for these terms, using the equations given by 3.35–36. This reduces the total order of the unconjugated terms another step. Proceeding in this way, alternately substituting using either 3.35–36 or 3.40, we eliminate all the unconjugated terms (in a finite number of steps). At each step we introduce new conjugated terms, but in view of 3.35–36 and 3.40, the total order of these never exceed $`k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0`$. This completes the proof of Lemma 3.41.∎
By substituting, using Lemma 3.41, for the conjugated terms that appear in the equations provided by 3.30, we conclude that for any multi-index $`J`$ and nonnegative integer $`k`$, we have
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DD_{1,1}^{k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0,k_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0}.$$
$`3.45`$
By using 2.10 and 2.12, we conclude that for any multi-indices $`R`$ and $`S`$, any nonnegative integer $`k`$, and any indices $`D,F\{1,\mathrm{}n\}`$, there are smooth functions, which are rational in their arguments preceeding the “;”, such that
$`L^RT^kL^{\overline{S}}\gamma _F^D=`$ $`r^{R\overline{S}k}(L^IT^j\gamma _A^C,L^IT^j\eta ^C,L^IT^j\xi ;f),`$ $`3.46`$
$`L^RT^kL^{\overline{S}}\eta _F^D=`$ $`s^{R\overline{S}k}(L^IT^j\gamma _A^C,L^IT^j\eta ^C,L^IT^j\xi ;f),`$
where $`|I|+jk_0+(k_0+k_0\mathrm{}_0)\mathrm{}_0`$. Finally, by using 2.11, its complex conjugate, and Proposition 3.18, it is not difficult to see that $`L^RT^kL^{\overline{S}}\xi `$ can be expressed in terms of $`\xi `$ and derivatives of $`\gamma _A^C`$, $`\overline{\gamma _A^C}`$, $`\eta ^C`$, and $`\overline{\eta ^C}`$. Thus, in view of 3.46, we also have, for any $`R`$, $`S`$, and $`k`$,
$$L^RT^kL^{\overline{S}}\xi =t^{R\overline{S}k}(L^IT^j\gamma _A^C,L^IT^j\eta ^C,L^IT^j\xi ,\overline{L^IT^j\gamma _A^C},\overline{L^IT^j\eta ^C},\overline{L^IT^j\xi };f),$$
$`3.47`$
where $`I`$ and $`j`$ run over the same indices as in 3.46. Now, recall that $`\mathrm{}_0k_0`$. The conclusion of Theorem 2 follows by writing 3.46–47, for all $`R`$, $`S`$, $`k`$ such that
$$|R|+|S|+k=k_0+(k_0+k_0^2)k_0+1$$
in any coordinate systems $`x=(x_1,\mathrm{},x_{2n+1})`$ and $`\widehat{x}=(\widehat{x}_1,\mathrm{}\widehat{x}_{2n+1})`$ for $`M`$ and $`\widehat{M}`$ near the points $`0M`$ and $`\widehat{0}\widehat{M}`$, respectively, and observing that the same system of differential equations holds for any CR mapping $`f`$ sending a neighborhood of $`0`$ in $`M`$ into $`\widehat{M}`$ with $`f(0)`$ sufficiently close to $`\widehat{0}`$. This completes the proof of Theorem 2.∎
###### Remark Remark
We would like to point out that a much simpler conclusion of the proof of Theorem 2 can be given in the case $`\mathrm{}_0=1`$, i.e. when the Levi form of $`M`$ has at least one nonzero eigenvalue at $`0`$. We can then use the commutator identity 3.20 instead of 3.19 to conclude
$$L^JT^k\gamma _F^D,L^JT^k\eta ^DC_{k_0+k,\mathrm{min}(k_0,m)}^{1,1},$$
$`3.48`$
instead of 3.30. By substituting for conjugated terms, using only 3.48, we obtain directly equations of the form 3.46 in which $`|I|+j2k_0`$. We invite the reader to carry out the details in this case. Observe that the system of differential equations obtained for $`f`$ using this argument is of order $`2k_0+2`$ rather than $`k_0^3+k_0^2+2`$ as given by Theorem 2 (or $`k_0+(k_0+k_0\mathrm{}_01)\mathrm{}_0+2=3k_0+1`$ for $`\mathrm{}_0=1`$, which is the order that actually follows from the proof of Theorem 2 presented above).
A similar simpification of the proof in the general case would be possible if one could prove that it suffices to take the sum over $`s`$ in 3.19 to run from $`s=0`$ to $`s=k\mathrm{}_0`$ instead of all the way up to $`s=m\mathrm{}_0`$.
###### Demonstration Proof of Theorem $`1`$
The system of differential equations 0.3 is a so-called complete system of order $`k_0^3+k_0^2+k_0+2`$. In particular, any solution is completely determined by its $`(k_0^3+k_0^2+k_0+1)`$-jet at $`0M`$ (see e.g. \[BCG<sup>3</sup>\]. cf. also \[Han2, Proposition 2.2\]). On the other hand, if $`xZ(x)`$ is an embedding of $`M`$ into $`^N`$ sending $`p_0M`$ to $`0^N`$ and $`x^{}Z^{}(x)`$ is an embedding of $`M^{}`$ sending $`p_0^{}`$ to $`0^N`$, then for any smooth CR mapping $`fM\widehat{M}`$, with $`f(p_0)=p_0^{}`$, there exists (see e.g. \[BER3, Proposition 1.7.14\]) a formal power series mapping $`Z^{}=F(Z)`$, with $`F(0)=0`$, sending $`M`$ into $`M^{}`$ (i.e. $`\rho (Z,\overline{Z})`$ divides $`\rho ^{}(F(Z),\overline{F(Z)})`$ in the ring of formal power series in $`Z,\overline{Z}`$; cf. e.g. \[BER4\]) such that
$$Z^{}(f(x))F(Z(x)),$$
$`3.49`$
where $``$ denotes equality as formal power series. Also, by \[BER4, Theorem 2.1.1\], the $`2k_0`$-jet at $`0`$ of any invertible formal mapping $`Z^{}=F(Z)`$, with $`F(0)=(0)`$, sending $`M`$ into $`M^{}`$ determines the series $`F(Z)`$ completely. In particular, it follows from 3.49 that the $`2k_0`$-jet at $`p_0`$ of a CR diffeomorphism $`fMM^{}`$, with $`f(p_0)=p_0^{}`$, determines its $`(k_0^3+k_0^2+k_0+1)`$-jet at $`p_0`$. Hence, the conclusion of Theorem 1 follows from Theorem 2. ∎
###### Demonstration Proof of Theorem $`3`$
We shall need the following proposition.
###### Proposition 3.50
If a smooth real hypersurface $`M^N`$ is holomorphically nondegenerate at $`p_0M`$, then there exists an open neighborhood $`U`$ of $`p_0M`$ and a dense open subset $`U^{}U`$ such that $`M`$ is $`(N1)`$-nondegenerate at every $`pU^{}`$.
###### Demonstration Proof
The statement that, under the hypotheses in the proposition, there exists an open neighborhood $`U`$ of $`p_0`$ such that $`M`$ is finitely nondegenerate on a dense open subset $`U^{\prime \prime }U`$ is a consequence of \[BER3, Theorem 11.7.5 (iii)\]. To prove Proposition 3.50, it suffices to show that if $`M`$ is not $`k`$-nondegenerate, for any $`kN1`$, on an open set $`V`$, then $`M`$ is in fact not finitely nondegenerate at any $`pV`$. Recall the subspaces $`E_j(p)T_p^{}M`$ defined for $`j=0,1,\mathrm{}`$ in §1. Assume that $`E_{N1}(p)`$ is a proper subspace of $`T_p^{}M`$ for every $`pV`$, i.e. $`M`$ is not $`k`$-nondegenerate, for any $`kN1`$, in $`V`$. Since $`\text{dim}_{}T_p^{}M=N`$, we conclude, by 1.3, that there must be an open subset $`V^{}V`$ and an integer $`1\mathrm{}N1`$ such that
$$E_\mathrm{}1(q)=E_{\mathrm{}}(q),qV^{}.$$
$`3.51`$
We claim that if $`E_\mathrm{}1(q)=E_{\mathrm{}}(q)`$ for all $`q`$ in some open sufficiently small set $`V^{}M`$, then in fact $`E_\mathrm{}1(q)=E_k(q)`$ for all $`k\mathrm{}`$ and all $`qV^{}`$. To see this, observe that 3.51 implies that, for every $`A_1,\mathrm{},A_{\mathrm{}}\{1,\mathrm{},N\}`$, there are smooth functions $`a_{\overline{A}_1\mathrm{}\overline{A}_{\mathrm{}}}^{\overline{C}_1\mathrm{}\overline{C}_j}`$ such that
$$_{\overline{A}_{\mathrm{}}}\mathrm{}_{\overline{A}_1}\theta =\underset{j=0}{\overset{\mathrm{}1}{}}a_{\overline{A}_1\mathrm{}\overline{A}_{\mathrm{}}}^{\overline{C}_1\mathrm{}\overline{C}_j}_{\overline{C}_j}\mathrm{}_{\overline{C}_1}\theta $$
$`3.52`$
in $`V^{}`$. 3.52 implies that $`E_{\mathrm{}+1}(q)=E_{\mathrm{}}(q)`$ for $`qV^{}`$, and the claim follows by induction. We conclude that $`M`$ is not finitely nondegenerate in $`V^{}`$. A simple connectedness argument applied to each component of $`V`$ proves that $`M`$ cannot be finitely nondegenerate at any point in $`V`$. This completes the proof of Proposition 3.50.∎
We return to the proof of Theorem 3. The fact that $`M`$ is minimal at $`p_0`$ implies, by a theorem of Trepreau (see e.g. \[BER3, Theorem 8.1.1\]; the analogous result in higher codimensions was proved by Tumanov), that for any open neighborhood $`U`$ of $`p_0`$ in $`M`$, there exists an open connected set $`\mathrm{\Omega }^N`$ (on “one side of $`M`$”) such that $`U^{}:=\overline{\mathrm{\Omega }}MU`$ is an open neighborhood of $`p_0`$ and every smooth CR function in $`U`$ is the smooth boundary value of a holomorphic function in $`\mathrm{\Omega }`$. We deduce by the uniqueness of boundary values of holomorphic functions, Proposition 3.50, and Theorem 1 that there exists $`p_1U^{}`$ such that if $`f_1,f_2UM^{}`$ are smooth CR diffeomorphisms into some smooth real hypersurface $`M^{}^N`$ and $`^\alpha f_1(p_1)=^\alpha f_2(p_1)`$, for all $`|\alpha |2(N1)`$, then $`f_1=f_2`$ in $`U^{}`$. Using this fact, the proof of Theorem 3 is completed exactly as the proof of \[BER2, Theorem 2\]. Choose $`Y_1,\mathrm{},Y_m\text{aut}(M,p_0)`$ which are linearly independent over $``$, and denote by $`F(x,y)`$, where $`x=(x_1,\mathrm{},x_{2N1})`$ is some local coordinate system on $`M`$ near $`p_0`$ and $`y=(y_1,\mathrm{},y_m)^m`$, the time-one map of the flow $`\mathrm{exp}t(y_1Y_1+\mathrm{}+y_mY_m)`$, for $`y`$ in some sufficiently small neighborhood $`V`$ of the origin $`^m`$. The arguments in \[BER2\] combined with the uniqueness result stated above, for a suitably chosen open neighborhood of $`U`$ of $`p_0`$ in $`M`$, shows that the mapping $`VJ^{2(N1)}(M,M^{})_{p_1}`$, given by
$$y(_x^\alpha F(p_1,y))_{|\alpha |2(N1)},$$
$`3.53`$
is smooth and injective. Hence, $`m\text{dim}_{}J^{2(N1)}(M,M^{})_{p_1}`$ which proves Theorem 3. ∎
###### Demonstration Proof of Theorem $`4`$
The conclusion of Theorem 4 is a direct consequence of Theorem 2 and Proposition 3.54 below. We shall use the notation $`J^k(^q,^m)_0`$ for the space of $`k`$-jets at $`0^q`$ of smooth mappings $`f^q^m`$, and $`\lambda ^k=(\lambda _i^\beta )`$, $`|\beta |k`$ and $`i=1,\mathrm{},m`$, for the natural coordinates on this space in which the $`k`$-jet of $`f`$ is given by $`\lambda _i^\beta =_x^\beta f_i(0)`$.
###### Proposition 3.54
Let $`UJ^k(^q,^m)_0\times ^q`$ be an open domain. Let $`r_j^\alpha (\lambda ^k)(x)`$, for any multi-index $`\alpha _+^m`$ with $`|\alpha |=k+1`$ and any $`j=1,\mathrm{},m`$, be smooth ($`C^{\mathrm{}}`$) functions on $`U`$. Then, for any $`\lambda _0^kJ^k(^q,^m)_0`$ such that $`(\lambda _0^k,0)U`$, there exists a uniquely determined smooth function $`FU_0\times V_0^m`$, where $`U_0`$ is an open neighborhood of $`\lambda _0^kJ^k(^q,^m)_0`$ and $`V_0`$ is an open neighborhood of $`0^q`$, such that if $`f=(f_1,\mathrm{},f_m)`$ solves
$$_x^\alpha f_j=r_j^\alpha (_x^\beta f),|\alpha |=k+1,j=1,\mathrm{},m,$$
$`3.55`$
near $`0^q`$ and $`j_0^k(f)U_0`$, then
$$F(j_0^k(f),)=f.$$
$`3.56`$
###### Remark Remark $`3.57`$
Observe that we do not claim that $`F(\lambda ^k,)`$ solves 3.55 for any initial value $`\lambda ^k`$, but only that if there is a solution with this initial condition then it coincides with $`F(\lambda ^k,)`$. The idea for Proposition 3.54 was suggested to the author by D. Zaitsev.
###### Demonstration Proof of Proposition $`3.54`$
By a standard argument (considering the derivatives $`_x^\beta f`$, $`|\beta |k`$, as new unknowns), it suffices to prove Proposition 3.54 with $`k=1`$. Thus, we may assume that the system 3.55 is of the form
$$_{x_j}f_i=r_{ij}(f),i=1,\mathrm{},m;j=1,\mathrm{},q.$$
$`3.58`$
Fix $`\lambda _0^1`$ as in the theorem. Write $`x=(t,x^{})\times ^{q1}`$ and consider the initial value problem for a system of ordinary differential equations
$$_tf_i(t,0)=r_{i1}(f(t,0))(t,0),f(0,0)=\lambda ^1,$$
$`3.59`$
for $`\lambda ^1`$ in some sufficiently small neighborhood of $`\lambda _0^1`$. By a classical result (see \[CL, Chapter 1.7\], Theorem 7.5 and the following remarks), there is a smooth function $`F^1U_1\times V_1^m`$, where $`U_1`$ is an open neighborhood of $`\lambda _0^1J^1(^q,^m)_0`$ and $`V_1`$ is an open neighborhood of $`0`$, such that $`tF^1(\lambda ^1,t)`$ is the unique solution of 3.59. Next, write $`x=(x_1,t,x^{\prime \prime })\times \times ^{q2}`$ and consider for each $`x_1U_1`$ the initial value problem
$$_tf_i(x_1,t,0)=r_{i1}(f(x_1,t,0))(x_1,t,0),f(x_1,0,0)=F^1(\lambda ^1,x_1).$$
$`3.60`$
Again by \[CL, Chapter 1.7\] (Theorem 7.5), there is a smooth function $`F^2U_2\times V_2^m`$, where $`U_2`$ is an open neighborhood of $`\lambda _0^1J^1(^q,^m)_0`$ and $`V_2`$ is an open neighborhood of $`(0,0)\times `$, such that $`tF^2(\lambda ^1,x_1,t)`$ is the unique solution of 3.60. Proceeding inductively in this way, we obtain the desired function $`F`$ after the $`q`$:th step. The straightforward details are left to the reader. We emphasize however that the function so obtained need not be a solution of the system 3.58, but it satisfies $`F(j_0^1(f),)=f`$ whenever $`f`$ is a solution. This completes the proof of Proposition 3.54.∎
The proof of Theorem 4 follows by applying Proposition 3.54 to the system of differential equations provided by Theorem 2. ∎
## 4.1. Concluding remarks
### 4.1. Abstract CR manifolds
In this paper, we have considered embedded real hypersurfaces as abstract manifolds with a(n integrable) CR structure. We have used the fact that the CR manifolds are embeddable (i.e. the CR structure is integrable) to choose a basis for the sections of $`TM`$ that satisfy certain commutation relations (Lemma 1.33). The author felt that the resulting equations 1.34 simplified the computations in the proofs to an extent which, by far, outweighed the loss of generality in assuming that the manifolds are embeddable. Without the use of the equations 1.34, the key equations 2.10-13 take the following form
$`L_{\overline{A}}\gamma _B^E+\gamma _D^ER_{\overline{A}B}^D+\eta ^Eh_{\overline{A}B}=\gamma _B^D\overline{\gamma _A^C}\widehat{R}_{\overline{C}D}^E,`$ $`\mathrm{4.1.1}`$
$`L_{\overline{A}}\xi +\xi h_{\overline{A}}=\xi \overline{\gamma _A^C}\widehat{h}_{\overline{C}}+\overline{\gamma _A^C}\eta ^D\widehat{h}_{\overline{C}D},`$
$`L_{\overline{A}}\eta ^C+\eta ^Ch_{\overline{A}}+\gamma _D^CR_{\overline{A}}^D=\xi \overline{\gamma _A^E}\widehat{R}_{\overline{E}}^C+\overline{\gamma _A^E}\eta ^F\widehat{R}_{\overline{E}F}^C,`$
$`T\gamma _A^CL_A\eta ^C+\gamma _B^CR_A^B\eta ^C\overline{h_{\overline{A}}}=\xi \gamma _A^B\widehat{R}_B^C+\gamma _A^B\eta ^D\widehat{R}_{DB}^C+\gamma _A^B\overline{\eta ^E}\widehat{R}_{\overline{E}B}^C.`$
Analogous reflection formulae to those in Theorem 2.4, as well as analogues of the crucial Lemmas 3.1, 3.6, and Proposition 3.18, can be established (with considerably more work than above). The author is confident that a proof of Theorem 2 for abstract CR manifolds (of hypersurface type) $`M`$ and $`M^{}`$ of the same dimension can be produced from these ingredients, but he has not had the patience to check the details.
### 4.2. Infinitesimal CR automorphisms
It is clear from the proof of Theorem 3 above that in order for the estimate 0.5 to hold, it suffices that $`M`$ is minimal at $`p_0`$ and that there exists an open subset $`UM`$ with $`p_0`$ in its boundary such that $`M`$ is finitely nondegenerate on $`U`$. The latter holds, in particular, if $`M`$ is holomorphically nondegenerate at $`p_0`$ (and, in the real-analytic case, only if), but may hold, in the case of merely smooth manifolds, even if $`M`$ is holomorphically degenerate at $`p_0`$ (see \[BER3, Example 11.7.29\]). On the other hand, as is mentioned in the introduction, if there exists a vector field $`Y`$ of the form 0.4, where the restrictions of the $`a_j`$ to $`M`$ are CR functions, which is tangent to $`M`$ near $`p_0`$, then $`\text{dim}_{}\text{aut}(M,p_0)=\mathrm{}`$. Let us call the restriction to $`M`$ of such a vector field $`Y`$ a CR holomorphic vector field. Thus, one is led to the following question. Suppose $`M`$ is not finitely nondegenerate at any point in an open neighboorhood of $`p_0`$. Does there then exist a CR holomorphic vector field on $`M`$ near $`p_0`$? The author does not know the answer to this question in general, but it seems to be related to the range of the tangential Cauchy-Riemann operator $`\overline{}_b`$ (see e.g. \[B\] for the definition). We conclude this paper by briefly outlining this connection.
First, observe that a vector field $`Y`$ is CR holomorphic if and only if $`Y`$ is a section of $`\overline{𝒱}`$ and $`[\overline{L},Y]`$ is a CR vector field (i.e. a section of $`𝒱`$) for every CR vector field $`\overline{L}`$. Now, suppose that there is an open set $`UM`$ in which $`M`$ is not finitely nondegenerate at any point. We claim that there exists a (non-vanishing) CR holomorphic vector field $`Y`$ near $`pU`$ if (i) $`\text{dim}_{}E_{N1}(q)`$ (which is $`<N`$ for $`q`$ in $`U`$ by assumption) is maximal at $`q=p`$, and (ii) $`\overline{}_bu=f`$ is solvable at $`p`$ for every $`(0,1)`$-form $`f`$ with $`\overline{}_bf=0`$. For simplicity, we shall indicate the proof of this only in the case $`\text{dim}_{}E_{N1}(p)=N1`$. We choose a smooth nonvanishing section $`X`$ of $`\overline{𝒱}`$ near $`p`$ such that $`X(q)E_k(q)^{}`$ for all $`k`$ and all $`q`$ in an open neighborhood of $`p`$. (This can be done by assumption (i) above.) We denote by $`L_{\overline{1}},\mathrm{},L_{\overline{n}}`$ a basis of the CR vector fields on $`M`$ near $`p`$, where $`L_n:=\overline{L}_{\overline{n}}=X`$. We choose a tranversal vector field $`T`$, as in §1, and denote by $`\theta ,\theta ^A,\overline{\theta }^{\overline{A}}`$, with notation and conventions as in §1, a dual basis of $`T,L_A,L_{\overline{A}}`$. The fact that $`L_n`$ is valued in $`E_k^{}`$, for every $`k`$, implies that $`[L_{\overline{A}},L_n]=b_{\overline{A}}L_n`$ modulo the CR vector fields. We shall look for a CR holomorphic vector field $`Y`$ of the form $`uL_n`$, where $`u`$ is a function to be determined. It is easy to see that $`[L_{\overline{A}},Y]`$ is a CR vector field if and only if
$$L_{\overline{A}}u+ub_{\overline{A}}=0$$
$`\mathrm{4.2.1}`$
and, hence, $`Y`$ is CR holomorphic if and only if 4.2.1 is satisfied for every $`A\{1,\mathrm{},n\}`$. If we could solve
$$\overline{}_bv=f,$$
$`\mathrm{4.2.2}`$
where $`f=b_{\overline{A}}\theta ^{\overline{A}}`$, then $`u=e^v`$ would solve 4.2.1. The $`(0,1)`$-form $`f`$ coincides with the form $`L_n\mathrm{}\overline{}_b\theta ^n`$, as the reader can verify. From this observation, one can check that $`f`$ satisfies the necessary compatibility condition for solvability,
$$\overline{}_bf=\overline{}_b(L_n\mathrm{}\overline{}_b\theta ^n)=0.$$
$`\mathrm{4.2.3}`$
Hence, if the tangential Cauchy-Riemann complex is solvable at level $`(0,1)`$ at $`p`$, then we can solve 4.2.2 and obtain a CR holomorphic vector field $`Y=uL_n`$ near $`p`$. However, the author knows of no results on solvability which apply in this situation (unless, of course, $`M`$ is real-analytic). |
warning/0001/hep-th0001089.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Recently it has been shown that the noncommutative gauge theories can arise as the low energy effective open string theory in the presence of D-branes with non-zero NSNS two form, B-field, on them .
The action for the noncommutative gauge theories is obtained from the usual (commutative ) Yang-Mills theories by replacing the ordinary product of the fields by the \*-product (for a review see ). From now on we call them NCYM theories. It has been shown that the noncommutative $`U(1)`$ theory can be understood as the commutative gauge theory with a deformed gauge group, so that the dynamics of this new gauge theory is governed by a series of dipoles in addition to the usual photons . Similar statements can be made for NCYM with any gauge group. It has been discussed that NCYM theories are renormalizable . Besides the pure gauge theories, the noncommutative scalar field theories have also been considered and it has been argued that these noncommutative gauge theories are renormalizable if and only if their commutative limit (noncommutative parameter, $`\theta `$, equal to zero) is renormalizable .
From the string theory side, the branes with constant B-field on them preserve 16 supercharges, half of the supercharges of the type II theories, similar to the usual D-branes, and hence when we study the low energy effective theory of such branes we actually deal with a NC (Super)YM (NCSYM) theory with 16 supercharges. Although the perturbative fermionic fields in a noncommutative background have not been well elaborated, since the noncommutative field theories can be understood as a commutative field theory with definite interaction terms, it is believed that these theories should have the usual feature of the gauge theories with 16 supercharges. As an example, it has been shown that the large N limit of the NCSYM can be studied through the gauge theory/gravity on definite background correspondence . The conceptual new point of these NCSYM theories is that, despite being maximally supersymmetric, unlike the usual SYM (with 16 supercharges), they are not conformal invariant and this is because the deformation parameter, $`\theta `$, is a dimensionful parameter with dimensions of (length)<sup>2</sup>.
For the $`𝒩=4,D=4`$ SYM theories large amount of supersymmetries kill all the interesting dynamics. In this paper we try to study more interesting theories with less supersymmetries, namely $`𝒩=2,D=4`$ NCSYM. After the work of we learnt how we can study the four dimensional $`𝒩=2`$ SYM theories, their BPS spectrum, their moduli space of vacua and its singularities, and how we can extract all of these physics from a proper elliptic curve. In another paper, , it was shown how the curve itself, and hence all of the above mentioned physics of $`𝒩=2,D=4`$ SYM theories, can be obtained from the definite type IIA string and M theories brane configuration In that paper only the $`_{\alpha =1}^NSU(k_\alpha `$) gauge group was considered, later the same procedure generalized to SP and SO gauge theories in extensive papers, which we are not going to list them here..
In this paper, we show how the brane configuration method should be extended and modified for studying the $`𝒩=2,D=4`$ NCSYM with $`SU(N)`$ gauge group. In this way we study some of the physical aspects of the deformed SYM theories, and discuss that most of the physics expected in the commutative $`𝒩=2,D=4`$ SYM hold in the noncommutative case too. Intuitively this should be related to the fact that the number of degrees of freedom of noncommutative gauge theories, at least for the planar diagrams, are the same as the commutative gauge theories .
The paper is organized as follows. In the next section we fix our conventions and notations and review the preliminaries we need. In section 3, we build the proper brane configuration from type IIA NS5-brane and (D4-D2)-brane bound states so that the brane system preserves 8 supercharges. In section 4 we relate the (pure) $`D=4`$ NCSYM to the brane configuration of the section 3 and identify the parameters of gauge theory in terms of the brane model parameters. In this way we find the one loop beta function of the theory. By lifting the brane configuration to M-theory we study the Coulomb branch of the theory. Moreover we discuss different phases of our noncommutative theory in different energies. The last section is devoted to concluding remarks and open questions.
## 2 Conventions and Preliminaries
a)String Theory
It has been shown that turning on B-fields polarized parallel to a $`D_p`$-brane, is equivalent to building various bound states of $`p,p2,p4,\mathrm{}`$ branes, depending on the rank of the B-field . For the B-field of rank one, these bound states are composed of a p-brane and a uniform distribution of (p-2)-branes on it. So that if the worldvolume of the $`p`$-brane is located along $`012\mathrm{}p`$ directions and $`B_{p1,p}`$ is the non-zero component, then the (p-2)-branes span $`012\mathrm{}(p2)`$ directions. This argument can be simply generalized to any higher rank B-fields. In our notations $`B`$ is a dimensionless parameter and $`l_s^2B`$ is the (p-2)-branes density. The mass density of the bound state in this notation is $`(l_s^{(p+1)}g_s)^1\sqrt{1+B^2}`$ .
The perturbative dynamics of these bound states, similar to the individual D-branes, is governed by the open strings ending on them, but this time because of the B-field, they should satisfy mixed boundary conditions. Studying such open strings we learn that the brane worldvolume is in fact a noncommutative plane:
$$[x^\mu ,x^\nu ]=il_s^2(\frac{B}{1B^2})^{\mu \nu }i\theta ^{\mu \nu }.$$
(2.1)
As we see $`\theta `$ is a parameter with dimensions of (length)<sup>2</sup>.
b)Noncommutative Field Theories
Field theories on a noncommutative plane are obtained from their commutative version by replacing the product of the fields by the \*-product defined as follows
$$(fg)(x)=exp(\frac{i}{2}\theta _{\mu \nu }_{x_\mu }_{y_\nu })f(x)g(y)|_{x=y}.$$
(2.2)
Besides the field products, to couple the noncommutative theories to a NCYM, we should replace the ordinary derivative with a ”covariant derivative”:
$$_\mu _\mu +\{A_\mu ,\}_{M.B.},$$
(2.3)
where the ”Moyal Bracket” in the above relation is $`\{f,g\}_{M.B.}=fggf`$. As we see for the slowly varying fields ($`|\frac{f}{f}|^2<<|\theta |^1`$), these noncommutative field theories effectively behave like the commutative theories.
It is worth noting that for the case of noncommutative SU(N) gauge theory, when $`A_\mu `$ take values in SU(N) algebra, the field strength $`F_{\mu \nu }`$ defined by (2.3) is not algebra valued unlike the usual gauge theories (it is sitting in the SU(N) group).
## 3 The Model with IIA branes
In order to reduce the number of supersymmetries, as it has been shown and discussed in , we consider the brane intersections.
Since we are interested in the four dimensional theories, like , we consider the IIA theory and its NS5- and (D4-D2)- branes. Let us consider the following brane configuration:
(N+1) NS5-branes labelled by $`\alpha =0,1,\mathrm{},N`$, spanning $`012345`$ directions. These fivebranes are located at different values of $`x^6`$ direction, and
a number of D4-branes with their worldvolume along $`01236`$ so that there are $`k_\alpha `$ number of them between the $`(\alpha 1)^{th}`$ and $`\alpha ^{th}`$ NS5-branes. As we see the intersection of these branes is a 3+1 dimensional space, $`0123`$, on which we have our $`𝒩=2`$ gauge theory.
To find a NCSYM, we turn on B-field along the D4-branes. We should recall that turning on a B-field will not change the number of conserved supercharges (see the Appendix for some explicit calculations). From the superalgebra point of view, turning on the B-field corresponds to choosing another central extension of the algebra (different from that of the individual brane). In other words, when we turn on the B-field, we again find a BPS solution which preserves half of supersymmetries.
There are various choices for the B-field, it can be of rank one or two, and also it can have different polarizations. In this paper we present calculations for the rank one case, with non-zero $`B_{23}=B`$, which is the most interesting case. We will discuss the $`B_{36}`$ and rank two cases briefly in the discussion section. With a non-zero $`B_{23}`$, our brane configuration consists of NS5-branes along $`012345`$ and D4-branes along $`01236`$ and a distribution of D2-branes having their worldvolume along $`016`$. These D2-branes are open D2-branes as discussed in .
In order to find a system with finite energy we should compactify the $`x^6`$ direction, or equivalently we should identify the $`(N+\alpha )^{th}`$ fivebrane with $`\alpha ^{th}`$.
Along the lines of , the $`x^6`$ coordinate as a function of $`vx^4+ix^5`$, is obtained by minimizing the total fivebrane worldvolume. For the large $`v`$ equation of $`x^6`$ reduces to a source free Laplace equation,
$$^2x^6=0.$$
(3.1)
Since $`x^6`$ is only a function of the directions normal to the brane bound state,
$$x^6=𝒦ln|v|+constant.$$
(3.2)
The parameter $`𝒦`$ is actually the ratio of the tensions (or mass densities) of two intersecting branes:
$$𝒦\frac{(\mathrm{D4}\mathrm{D2})\mathrm{mass}\mathrm{density}}{\mathrm{NS5}\mathrm{brane}\mathrm{mass}\mathrm{density}}\frac{l_s^5g_s^1\sqrt{1+B^2}}{l_s^6g_s^2}=l_sg_s\sqrt{1+B^2}.$$
(3.3)
For the case with $`k`$ number of these bound states on top of each other, obviously our $`𝒦`$ factor should be multiplied by a factor of $`k`$. It is worth noting that (3.3) is the $`B=0`$ result with $`g_s`$ replaced with the open string coupling and, it reduces to the results of for the $`B=0`$. Since the (D4-D2)-branes ending on the fivebrane on its left pulls it in the opposite direction compared to those ending on its right, the $`𝒦`$ factor for them should have different relative sign. If we have $`q_L`$ (and $`q_R`$) (D4-D2)-branes ending on the fivebrane on its left (and right), which are located at $`a_i,i=1,\mathrm{}.,q_L`$ (and $`b_j,j=1,\mathrm{}.,q_R`$), then the asymptotic form of $`x^6`$ is
$$x^6=𝒦(\underset{i=1}{\overset{q_L}{}}ln|va_i|\underset{j=1}{\overset{q_R}{}}ln|vb_j|)+constant.$$
(3.4)
A well-defined $`x^6`$ for $`v\mathrm{}`$ is obtained if and only if $`q_L=q_R`$.
To study the gauge theory dynamics we also need to consider the moving (D4-D2)-branes. This motion is realized by letting $`a_i`$ and $`b_j`$ become functions of the space-time coordinates ($`0123`$). This motion as explained in , corresponds to the question of IR divergences of the NCSYMAs we will discuss in the next chapter, the IR behaviour of the NCSYM is the same as the commutative SYM.. This motion contributes to the fivebrane kinetic energy as $`d^4xd^2v_\mu x^6^\mu x^6,\mu =0,1,2,3`$. We should note that the metric on our space-time is the open string metric. With $`x^6`$ given by (3.4), it becomes
$$d^4xd^2v|\mathrm{Re}((\underset{i}{}_\mu a_i(\frac{1}{va_i})\underset{j}{}_\mu b_j(\frac{1}{vb_j}))|^2.$$
(3.5)
The integral converges if and only if
$$\underset{i}{}a_i\underset{j}{}b_j=q_\alpha ,$$
(3.6)
where $`q_\alpha `$ is a constant, and is a characteristic of the $`\alpha ^{th}`$ fivebrane. The $`q_\alpha `$’s are determined by the separation between (D4-D2) branes, and hence we expect them to be related to the ”bare mass” of the gauge theory hypermultiplets.
The quantum mechanical treatment of our noncommutative gauge theory corresponds to the M-theory limit of our brane model. In that limit type IIA on $`𝐑^{10}`$ is actually the M-theory on $`𝐑^{10}\times S^1`$, with radius $`R=l_sg_s`$. But, this is only the compactification radius for closed strings, i.e. $`R`$ is the radius which relates the 10 and 11 dimensional supergravities. As we discuss in the following, the open string compactification radius is different<sup>§</sup><sup>§</sup>§The fact that open string compactification radius, $`R_o`$, is effectively different, in the work of Connes, Douglas and Schwarz , which was the light-like compactification of M-theory with C-field background, is reflected in their dim$``$=Tr1(the dimension of Schwartz space), and in our case is related to $`\frac{R_o}{R}`$..
In the M-theory limit, our (D4-D2)-bound state is lifted to a (M5-M2)-bound state. Such bound states as discussed in are formed by turning on a non-zero C-field background of 11 dimensional theory on the M5-brane. This corresponds to turning on a self-dual three form of the six-dimensional M5-branes worldvolume theory. In other words, in the M-theory limit $`B_{23}`$ will be replaced by $`C_{23(10)}`$ and $`C_{016}`$, and these fields are equal because of the self-duality condition. From the 10 dimensional point of view, this self-duality is related to the fact that, D2-branes are the sources for the $`RR`$ three form of type IIA theory, and in our case their density is given by $`B_{23}`$. Having this in mind, we learn that the actual compactification radius viewed from the fivebrane worldvolume theory, and hence from the open M2-branes point of view, is not $`R`$, but $`R\sqrt{1+C^2}`$. Since the eleven dimensional C-field and the ten dimensional B-field are supergravity fields they should be related by closed string parameters, which in our conventions are
$$l_p^3RC=l_s^2B.$$
(3.7)
So, the effective compactification radius for the open strings, is
$$R_o=R\sqrt{1+B^2}.$$
(3.8)
If we denote the eleventh dimension by $`x^{10}`$, which is periodic with periods $`2\pi R`$, we have
$$x^6+ix^{10}=R\sqrt{1+B^2}\left\{(\underset{i=1}{\overset{q_L}{}}ln|va_i|\underset{j=1}{\overset{q_R}{}}ln|vb_j|)+i\varphi \right\},$$
(3.9)
where $`0\varphi =\frac{x^{10}}{R_o}<2\pi `$, and hence
$$s\frac{x^6+ix^{10}}{R}=\sqrt{1+B^2}\{\underset{i=1}{\overset{q_L}{}}ln(va_i)\underset{j=1}{\overset{q_R}{}}ln(vb_j)\}+constant.$$
(3.10)
The fact that $`s`$ is holomorphic in $`v`$, is expected from the supersymmetry. Let us concentrate on the imaginary part of (3.10). It states that circling around one of $`a_i`$ and $`b_j`$’s in the $`v`$ complex plane, $`x^{10}`$ jumps by $`\pm 2\pi R_o`$. From the fivebrane worldvolume effective theory, the end points of (D4-D2)-brane bound states are viewed as ”dyonic vortices”. In the T-dual version, where our brane configuration is composed of IIB NS5-branes (spanning $`012345`$) and (D3-D1) bound state (with D3-branes along $`0126`$ and D-strings along $`06`$), the end point of the (D3-D1)-brane looks like ”dyons” of fivebrane effective theory. With our conventions in definition of $`𝒦`$, (3.3), these dyons carry one unit of magnetic charge and $`l_s^2BV_2`$ units of electric charge ($`V_2`$ is the volume of $`12`$ plane).
## 4 Four Dimensional Noncommutative Gauge Theory
Now let us return to our main question: what can we learn about the $`D=4`$ NCSYM from the above brane configuration.
Since we have considered the $`B_{23}0`$, according to (2.1) we deal with a NCSYM with non-zero $`\theta _{23}`$,
$$\theta \theta _{23}=l_s^2\frac{B}{1+B^2}.$$
(4.1)
If we consider the brane configuration we built in the previous section, because of (3.4), we deal with $`_{\alpha =1}^NSU(k_\alpha )`$ NCSYM theory.
Now we should identify our gauge theory coupling constant through the string theory parameters. The naive answer to this question can be understood in light of the commutative /noncommutative gauge theory correspondence proposed in , by expanding the Born-Infeld action for the D4-branes with non-zero B-field, up to the first order in $`\alpha ^{}`$. Then the gauge coupling $`g_\alpha `$ of the $`SU(k_\alpha )`$ should be given by
$$\frac{1}{g_\alpha ^2(v)}=\frac{x_\alpha ^6(v)x_{\alpha 1}^6(v)}{l_sg_s}.$$
(4.2)
Replacing $`x^6`$ from (3.4) we find that the above relation is exactly the same as its commutative counter-part, with $`g_s`$ which is the closed string coupling, replaced with open string coupling.
According to (4.2) we propose that $`g_\alpha ^2`$ for the NCSYM to be logarithmic divergent for large $`v`$. This is the familiar asymptotic behaviour of the commutative theory at high energies. This proposal is expected since we believe that the noncommutative gauge theories can be mapped into a commutative gauge theory through the Fourier expansion and also, we know that the planar degrees of freedom of the NCSYM is the same as its commutative counter-part . Moreover we propose that this logarithmic divergence corresponds to the one loop beta function of our four dimensional NCSYM. This is in exact agreement with perturbative results of .
The theta angle of the $`SU(k_\alpha )`$ theory, $`\theta _\alpha `$, is then naturally related to the imaginary part of (3.10) as$`\theta _\alpha `$ should not be confused with $`\theta `$, the noncommutativity parameter.
$$\theta _\alpha =\frac{x_\alpha ^{10}x_{\alpha 1}^{10}}{R}.$$
(4.3)
Hence the
$$\tau _\alpha (v)=\frac{\theta _\alpha }{2\pi }+\frac{4\pi i}{g_\alpha ^2}=i(s_\alpha (v)s_{\alpha 1}(v)).$$
(4.4)
In the large $`v`$ limit:
$$\tau _\alpha (v)=i\sqrt{1+B^2}(2k_\alpha k_{\alpha 1}+k_{\alpha +1})lnv.$$
(4.5)
As we see the only difference of (4.5) with the commutative version is the coefficient in front.
Comparing the standard one loop asymptotic freedom formula, $`\tau =ib_0lnv`$, we find the important result that the one loop beta function for NCSYM theories with the above mentioned gauge group is proportional to the one loop beta function of the commutative counter-part.
Analogous to the commutative case, the open strings having their end points on the open (D4-D2)-bound states on the opposite sides of fivebranes form the hypermultiplets of our NCSYM theory, obviously in the $`(k_1,\overline{k}_2)(k_2,\overline{k}_3)\mathrm{}.(k_{N1},\overline{k}_N)`$ representation. The only point which one should note here is that the mass of these hypermultiplets should be calculated by squaring the momentum with the open string metric . These hypermultiplets become massless classically when (D4-D2)-bound states on the right of a fivebrane end on the same point as the (D4-D2)-branes on the left. Like the commutative case, when one these hypermultiplets becomes massless, namely $`(k_i,\overline{k}_{i+1})`$, for the $`k_i=k_{i+1}`$ we expect the related factor of the gauge group to become a $`SU(k_i)`$ instead of $`SU(k_i)\times SU(k_{i+1})`$.
Up to here we have identified and related the parameters of NCSYM with those of our brane configuration model and briefly discussed the similarities of noncommutative and commutative theories at one loop. Now we want to study different phases of the noncommutative gauge theories.
As we mentioned the noncommutativity parameter, $`\theta `$, is a dimensionful parameter and hence there are two parameters of energy dimensions, $`\theta ^{1/2}`$ and $`R_o^1`$ ( the ”open string” eleven dimensional radius), with which we should compare our energy scales. We distinguish two different phases:
a) $`\theta /R_o^2\stackrel{<}{_{}}1`$
In this case our noncommutative theory effectively behaves as a commutative SYM because, before the energies that noncommutativity effects show themselves, $`\theta E^21`$, we should uplift our theory to M-theory, where we only have (M5-M2)-bound state effective theory which is not a noncommutative one.
If we replace the $`\theta `$ and $`R_o`$ by their string theoretic values we have
$$\frac{\theta }{R_o^2}\frac{l_s^2}{l_s^2g_s^2}\frac{B}{(1+B^2)^2}.$$
(4.6)
Since $`\frac{B}{1+B^2}1/2`$, for $`\theta /R_o^2\stackrel{<}{_{}}1`$ cases we have
$$g_{\mathrm{open}\mathrm{string}}\stackrel{>}{_{}}1,$$
(4.7)
and hence this case for generic B, corresponds to the strong coupling of our noncommutative gauge theory (which as we argued above) is not a noncommutative theory. For very large B, (4.7) reads as $`g_{YM}B\stackrel{>}{_{}}1`$ which corresponds to the commutative gauge theory weak coupling.
b) $`\theta /R_o^2\stackrel{>}{_{}}1`$
This is the more interesting case. From (4.6) we see that for generic B, we are actually dealing with weakly coupled gauge theory, and the theory shows different behaviours at different energies:
b-1) IR limit ($`\theta E^2<<1`$): In this regime the noncommutative effects can totally be neglected and we effectively see the commutative SYM theory. So, all the arguments of IR limit of commutative SYM holds for the NCSYM too.
b-2) Intermediate energies ($`\frac{1}{\sqrt{\theta }}\stackrel{<}{_{}}E<<\frac{1}{R}`$): This is the phase which we actually deal with a NCSYM theory.
b-3) UV limit ($`E\stackrel{>}{_{}}\frac{1}{R_o}`$): In this case we should uplift our brane configuration to the M-theory, where there is no noncommutative description.
M-Theory Interpretation and Seiberg-Witten Curves
In order to study the quantum structure of NCSYM, we need to lift the brane configuration we built in IIA theory (which describes the classical regime) to the M-theory. In the M-theory, as we stated earlier the (D4-D2)-bound state coincides with the (M5-M2)-bound state in the M-theory and NS5-branes are viewed as M5-branes. The (M5-M2)-brane, is actually a fivebrane with a non-zero self-dual three form on it. Hence, in the M-theory limit our IIA brane configuration is reinterpreted as a single fivebrane with a definite C-field on it, whose worldvolume sweeps $`R^4\times \mathrm{\Sigma }`$ where $`\mathrm{\Sigma }`$ is a hyper-surface given by a holomorphic function in the $`v`$ and $`s`$ complex planes. Since $`s`$ is not single valued, it is more convenient to introduce the $`t=exp(s/\sqrt{1+B^2})`$ instead. Analogue of the Seiberg-Witten elliptic curve, is exactly the curve defining $`\mathrm{\Sigma }`$ in the $`t`$ and $`v`$ plane, namely, $`F(t,v)=0`$. Compared to the commutative case the only difference is in the definition of $`t`$ with respect to $`s`$. However, the powers in $`v`$ will remain the same as the commutative case.
One can study the low energy four dimensional noncommutative theory from the six dimensional effective theory of M5-branes, if we denote the field strength of the two-form field living on the M5-brane by $`T`$, we know that $`T`$ is self-dual and its equation of motion is $`dT=0`$. One can decompose $`T`$ as follows:
$$T=F\mathrm{\Lambda }+^{}F^{}\mathrm{\Lambda }+C,$$
(4.8)
where $`C`$ is the self-dual constant background and in our case has non-zero components in $`016`$ (and $`23(10)`$). $`\mathrm{\Lambda }`$ is a harmonic one form defined on the $`\mathrm{\Sigma }`$, satisfying, $`d\mathrm{\Lambda }=d^{}\mathrm{\Lambda }=0`$. As discussed in , to see the reasonable low energy theory we need to define $`\mathrm{\Lambda }`$ on a point-wise compactified version of $`\mathrm{\Sigma }`$, $`\overline{\mathrm{\Sigma }}`$, which is obtained by adding (N+1) points to the $`\mathrm{\Sigma }`$. $`\overline{\mathrm{\Sigma }}`$ is a surface of genus $`g=_{\alpha =1}^N(k_\alpha 1)`$. $`F`$ is a two form and noncommutative field defined on $`R^4`$, obeying $`DF=D^{}F=0`$, where $`D`$’s are the noncommutative ”covariant derivative”. The low energy theory for the $`g`$ dimensional $`\mathrm{\Lambda }`$, the Coulomb branch of our theory, is (NC $`U(1)`$)<sup>g</sup>. Apart from the changes mentioned above, all the other results and discussions of are valid in our case too.
## 5 Concluding Remarks and Discussions
In this paper we studied the $`𝒩=2,D=4`$ NCSYM theories with gauge group $`_{\alpha =1}^NSU(k_\alpha )`$ through the brane configuration method. In order to find the NCSYM gauge theory, we considered the D4-branes with a B-field background. We should emphasize that actually the noncommutative gauge theories we are discussing are formulated on a space with open string metric . By means of our brane system, we found the one loop beta function of these noncommutative theories to be logarithmically divergent. Moreover, we argued that one can still find a holomorphic curve which describes the moduli space of our theory. For the case at hand, the pure gauge theory, we briefly discussed the behaviour of the theory in different energy regimes and its Coulomb branch.
In this paper we discussed the rank one B-field along $`23`$ directions. For the $`B_{16}`$ case, since the noncommutative coordinates are 3 and 6 directions, our four dimensional space-time ($`0123`$ directions) is not noncommutative and hence we have our usual commutative field theory. In addition, since instead of NS5-branes we have (NS5-D2)-brane bound states, we expect the equation of motion for $`x^6`$ not to be altered compared to the commutative case. For the same reasons for the case of rank two B-field, i.e. $`B_{23},B_{16}0`$, we expect to see the physics similar to $`B_{23}0`$. However, the $`B_{23}=B_{16}`$ seems to be an interesting special case to be discussed.
Another interesting extension of this work is adding matter fields and the Higgs branch of NCSYM. We believe that like the commutative version this can be done by adding D6-brane to the brane configuration. We hope to come back to this question in future works.
There are many other interesting problems which can be addressed, e.g. the generalization of Hanany-Witten work, and (2+1) dimensional NCSYM, the noncommutative two dimensional $`𝒩=(4,4)`$ theories, theories with lower supersymmetries (four supercharges) which are obtained by rotating one of the NS5-branes.
Acknowledgements
I would like to thank M. Alishahiha for many fruitful discussions and careful reading of the manuscript. I would like to thank H. Arfaei for discussions.
This research was partly supported by the EC contract no. ERBFMRX-CT 96-0090.
Appendix
The explicit form of the supercharges for a (D4-D2) bound state, resulting from a non-zero $`B_{23}=B`$ component, can be written as a linear combination of the type IIA supercharges, $`Q_R`$ and $`Q_L`$ ($`\mathrm{\Gamma }^{01\mathrm{}9}Q_R=Q_R`$ and $`\mathrm{\Gamma }^{01\mathrm{}9}Q_L=Q_L`$) :
$$Q=ϵ^RQ_R+ϵ^LQ_L,$$
where $`ϵ^R`$ and $`ϵ^L`$ are 16 component spinors satisfying the following conditions:
$$(I)\{\begin{array}{cc}(\frac{1}{\sqrt{1+B^2}}+\frac{B}{\sqrt{1+B^2}}\mathrm{\Gamma }^{23})\mathrm{\Gamma }^{016}ϵ^R=ϵ^L& \\ \mathrm{\Gamma }^{45789}ϵ^R=ϵ^L& \end{array}$$
The above conditions can be obtained from the T-dual version of the rotation matrices in $`23`$ plane. The $`\mathrm{\Gamma }^2`$ is rotated by $`R(\varphi )\mathrm{\Gamma }^2`$, with $`R(\varphi )=exp(\varphi \mathrm{\Gamma }^{23})=\mathrm{cos}\varphi +\mathrm{\Gamma }^{23}\mathrm{sin}\varphi `$ . By T-duality $`\mathrm{tan}\varphi `$ is replaced by $`B`$. It can easily be shown that relations (I) allow 16 supersymmetries. Moreover one can check that for the $`B=0`$ and $`B\mathrm{}`$ these conditions coincide with the supersymmetries preserved by a D4- and D2- brane respectively. When we also consider the NS5-branes as well as (I), we should consider
$$(II)\{\begin{array}{cc}\mathrm{\Gamma }^{012345}ϵ^R=ϵ^R& \\ \mathrm{\Gamma }^{012345}ϵ^L=ϵ^L& \end{array}$$
The simultaneous solutions of (I) and (II) give the allowed 8 supercharges. |
warning/0001/physics0001030.html | ar5iv | text | # Untitled Document
Simple quantum systems in the momentum representation
H. N. Núñez-Yépez† E-mail: nyhn@xanum.uam.mx
Departamento de Física, Universidad Autónoma Metropolitana-Iztapalapa, Apartado Postal 55-534, Iztapalapa 09340 D.F. México,
E. Guillaumín-España, R. P. Martínez-y-Romero,‡ On leave from Fac. de Ciencias, UNAM. E-mail: rodolfo@dirac.fciencias.unam.mx A. L. Salas-Brito¶ E-mail: asb@hp9000a1.uam.mx or asb@data.net.mx
Laboratorio de Sistemas Dinámicos, Departamento de Ciencias Básicas, Universidad Autónoma Metropolitana-Azcapotzalco, Apartado Postal 21-726, Coyoacan 04000 D. F. México
Abstract
The momentum representation is seldom used in quantum mechanics courses. Some students are thence surprised by the change in viewpoint when, in doing advanced work, they have to use the momentum rather than the coordinate representation. In this work, we give an introduction to quantum mechanics in momentum space, where the Schrödinger equation becomes an integral equation. To this end we discuss standard problems, namely, the free particle, the quantum motion under a constant potential, a particle interacting with a potential step, and the motion of a particle under a harmonic potential. What is not so standard is that they are all conceived from momentum space and hence they, with the exception of the free particle, are not equivalent to the coordinate space ones with the same names. All the problems are solved within the momentum representation making no reference to the systems they correspond to in the coordinate representation.
1. Introduction
In quantum mechanics the spatial position, $`\widehat{x}`$, and the linear momentum, $`\widehat{p}`$, operators play very symmetrical roles, as it must be obvious from the fundamental commutation relation
$$[\widehat{p},\widehat{x}]=\widehat{p}\widehat{x}\widehat{x}\widehat{p}=i\mathrm{}$$
$`(1)`$
where, apart from a minus sign, the roles of $`\widehat{x}`$ and $`\widehat{p}`$ are the same. Notice that a hat over the symbols has been used to identify operators. However, this fundamental symmetry is not apparent to many students of quantum mechanics because an excessive emphasis is put on the coordinate representation in lectures and in textbooks. Some students are even lead to think of the coordinate space wave function $`\psi (x)`$ as more fundamental, in a certain way, than its momentum space counterpart $`\varphi (p)`$; for, even in one of those rare cases where the Schrödinger equation is solved in momentum space, as is often the case with the linear potential $`x`$ (Constantinescu and Magyari 1978), many students feel that the quantum solution is somewhat not complete until the coordinate space wave function has been found. This is a pity since a great deal of physical consequences are understood better and many physical effects are more readily evaluated in the momentum rather than in the coordinate representation; as an example just think of scattering processes and form factors of every kind (Taylor 1972, Ch 3; Frauenfelder and Henley 1974; Bransden and Joachaim 1983; Griffiths 1987). To give another interesting example, let us remember the one-dimensional hydrogen atom, an apparently simple system whose properties were finally understood, after thirty years of controversy (Loudon 1959, Elliot and Loudon 1960, Haines and Roberts 1969, Andrews 1976, Imbo and Sukhatme 1985, Boya et al. 1988, Núñez-Yépez et al. 1988, 1989, Martínez-y-Romero et al. 1989a,b,c) only after an analysis was carried out in the momentum representation (Núñez-Yépez et al. 1987, 1988, Davtyan et al. 1987).
But, besides particular ocurrences, the advantages of an early introduction to the momentum representation in quantum mechanics are manyfold: a) to emphasize the basic symmetry between the representations, b) to introduce from the beginning and in a rather natural manner, distributions—pinpointing that the eigenfunctions are best regarded as generalized rather than ordinary functions—, non-local operators and integral equations, c) to help clarify the different nature of operators in both representations, for example, in the momentum representation a free particle (vanishing potential in any representation) cannot be considered as equivalent to a particle acted by a constant (in momentum space) potential, since this last system admits a bound state. According to us the problems discussed in this work make clear, using rather simple examples, the distinct advantages and perhaps some of the disadvantages of working in the momentum representation.
2. Quantum mechanics in momentum space.
For calculating the basic properties and the stationary states of quantum mechanics systems, the fundamental equation is the time-independent Scrhödinger equation which, in the coordinate representation, can be written as the differential eigenvalue equation
$$\frac{\mathrm{}^2}{2m}\frac{d^2\psi (x)}{dx^2}+U(x)\psi (x)=E\psi (x).$$
$`(2)`$
This can be obtained using, in the classical Hamiltonian $`H=p^2/2m+U(x)`$, the operator correspondence $`\widehat{p}i\mathrm{}d/dx`$, $`\widehat{x}x`$ \[operators complying with (1)\]. It is equally possible the use of the alternative operator correspondence $`\widehat{p}p`$ and $`\widehat{x}i\mathrm{}d/dp`$ \[also complying with (1)\] that can be shown to lead —though not as straightforwardly as in the previous case— to the integral Schrödinger equation in the momentum representation
$$\frac{p^2}{2m}\varphi (p)+𝑑p^{}U(pp^{})\varphi (p^{})=E\varphi (p),$$
$`(3)`$
where $`U(p)`$ is the Fourier transform of the potential energy function in the coordinate representation:
$$U(p)=\frac{1}{\sqrt{2\pi \mathrm{}}}_{\mathrm{}}^+\mathrm{}\mathrm{exp}(ipx/\mathrm{})U(x)𝑑x.$$
$`(4)`$
As it is obvious from (3), in this representation the potential energy becomes an integral operator (hence, usually non-local) in momentum space. Equations (2) and (3) are, in fact, Fourier transforms of each other; therefore the relationship between the coordinate and the momentum space wave functions is
$$\begin{array}{cc}\hfill \varphi (p)=& \frac{1}{2\pi \mathrm{}}_{\mathrm{}}^+\mathrm{}\mathrm{exp}(ipx/\mathrm{})\psi (x)𝑑x,\text{and}\hfill \\ & \\ \hfill \psi (x)=& _{\mathrm{}}^+\mathrm{}\mathrm{exp}(+ipx/\mathrm{})\varphi (p)𝑑p.\hfill \end{array}$$
$`(5)`$
Both functions, $`\psi (x)`$ and $`\varphi (p)`$, characterize completely and independently the state of the system in question; although they differ slightly in interpretation: whereas $`\psi (x)`$ is the probability amplitude that the a measurement of position gives a value in the interval $`[x,x+dx]`$, $`\varphi (p)`$ is the probability amplitude that a measurement of momentum gives a value in the interval $`[p,p+dp]`$. It should be remarked that the normalization of a function in momentum space does not guarantee the normalization of the corresponding function in coordinate space, for this to be true you should multiply the transform of $`\varphi (p)`$ times $`\sqrt{2\pi \mathrm{}}`$.
In spite of their complete equivalence, the momentum representation could throw light in certain features that may remain hidden in the coordinate representation; very good examples of this are the SO(4) symmetry of the hydrogen atom first uncovered by Fock (1935) using his masterly treatment of the problem in the momentum representation; or the treatment of resonant Gamow states in the momentum representation where they were found to be, contrary to what happens in the coordinate representation, square integrable solutions to a homogeneous Lippmann-Schwinger equation (Hernández and Mondragón 1984).
In this work we calculate the bound energy eigenstates and corresponding quantum levels of the simplest one-dimensional potential problems in the momentum representation. We choose to present them in order of increasing complexity, as it is usually done in basic quantum mechanics:
1) The free particle with $`U(p)=0`$.
2) Particle in a constant potential: $`U(p)=U_0`$ ($`U_0>0`$).
3) Particle interacting with the potential step
$$U(p)=\{\begin{array}{cc}0,\hfill & \text{if }p>0\text{,}\hfill \\ & \\ i\alpha \text{ (}\alpha \text{ a positive constant)},\hfill & \text{if }p0\text{;}\hfill \end{array}$$
$`(6)`$
please notice that as we assume $`\alpha `$ to be a real number, the $`i`$ factor is necessary to assure the Hermiticity of the potential energy operator.
4) Motion in the harmonic potential
$$U(p)=f_0\mathrm{cos}(ap),$$
$`(7)`$
where $`f_0>0`$ and $`a`$ are real numbers.
As we intend to illustrate in this contribution, in many instances the eigenfunctions are easier to calculate in momentum space than in the coordinate space representation. We have to recognize though that the momentum space eigenstates are best understood as generalized functions or distributions —to which the Riemann interpretation of integrals does not apply; this is explicitly illustrated by examples A, B, and D below. The energy eigenvalues are calculated, in most cases discussed here (A, B, and D), as consistence conditions on the eigenfunctions, and in the remaining one, C, from the univaluedness of the eigenfunctions.
3. The examples.
We want to point out that albeit we are addressing the same type of systems that are used to introduce quantum mechanics, here we employ the same notion of simplicity but with problems posed in momentum space (making them very different from the coordinate space ones). Please be aware that we use atomic units wherever it is convenient in the rest of the paper: $`\mathrm{}=e=m=1`$.
A. The free particle.
In the case of the free particle, as in the coordinate representation, $`U(p)=0`$ everywhere, so the Schrödinger equation (3) is simply
$$\left(\frac{p^2}{2}E\right)\varphi (p)=0;$$
$`(8)`$
this deceptively simple equation has as its basic solutions
$$\varphi _{p_E}(p)=A\delta (pp_E)$$
$`(9)`$
where $`p_E`$ is a solution of $`p^2=2E`$ and $`A`$ is a constant. This is so since, according to (8), the wave function vanishes excepting when the energy takes its “on shell” value $`E=p^2/2`$; furthermore as $`\varphi (p)`$ cannot vanish everywhere, equation (9) follows. The energy eigenfunctions (9) are also simultaneously eigenstates of the linear momentum,
$$\widehat{p}\varphi _{p_E}(p)=p\delta (pp_E)=p_E\varphi _{p_E},$$
$`(10)`$
and form a generalized basis —i.e. formed by Dirac improper vectors— for the states of a free particle with well defined energy and linear momentum (Böhm 1979, Sakurai 1985); for such a free particle the most general stationary momentum-space solution is then
$$\mathrm{\Phi }(p)=A_+\delta (p+|p_E|)+A_{}\delta (p|p_E|)$$
$`(11)`$
where the $`A_\pm `$ are complex normalization constants; this solution represent a particle traveling to the right with momentum $`|p_E|`$ and to the left with momentum $`|p_E|`$. The basic solutions (9) can be “orthonormalized” according to (Sakurai 1985)
$$_{\mathrm{}}^+\mathrm{}\varphi _{p_E}^{}(p)\varphi _{p_E^{}}(p)𝑑p=\delta (p_Ep_E^{})$$
$`(12)`$
which requires $`A=1`$ in (9). The possible energy values are constrained only by the classical dispersion relation $`E=p_E^2/2m`$ hence they form a continuum and the eigenstates cannot be bound.
It is to be noted that for describing the eigenstates of a free particle, quantum mechanics uses generalized functions for which the probability densities $`|\varphi _{p_E}(p)|^2`$ are not well defined! What it is well defined is their action on any square integrable function, hence on any physical state; therefore the eigenstates have to be regarded as linear functionals acting on $`L^2(R)`$, the set of all square integrable functions. The only physically meaningful way of dealing with free particles requires thus the use of wave packets as follows
$$\begin{array}{cc}\hfill \mathrm{\Phi }(p)=& _{\mathrm{}}^+\mathrm{}F(p^{})\delta (pp^{})𝑑p^{}\hfill \\ \hfill =& F(p),\hfill \end{array}$$
$`(13)`$
where $`F(p)`$ is any square integrable function of $`p`$. According to their properties then, improper vectors, like those in (9), though very useful for formal manipulations can never strictly represent physically realizable states (Taylor 1972, section 1a).
B. Motion under a constant potential
Substitution of the constant value $`U_0<0`$ into (3), gives us
$$\left(\frac{p^2}{2}E\right)\varphi (p)=U_0_{\mathrm{}}^+\mathrm{}\varphi (p^{})𝑑p^{};$$
$`(14)`$
to solve (14), let us define the number $`\stackrel{ˇ}{\phi }`$ as
$$\stackrel{ˇ}{\phi }_{\mathrm{}}^+\mathrm{}\varphi (p^{})𝑑p^{};$$
$`(15)`$
with this definition, the momentum representation Schrödinger equation (14) reduces to a purely algebraic equation for $`\varphi (p)`$,
$$\left(\frac{p^2}{2}E\right)\varphi (p)=U_0\stackrel{ˇ}{\phi };$$
$`(16)`$
let us now define $`k_0^2=2E>0`$, then the eigenfunctions are easily seen to be
$$\varphi (p)=\frac{2U_0\stackrel{ˇ}{\phi }}{p^2+k_0^2}.$$
$`(17)`$
To determine the energy eigenvalues we integrate both sides of (17) from $`\mathrm{}`$ to $`+\mathrm{}`$ to get
$$\stackrel{ˇ}{\phi }=2\pi \frac{U_0\stackrel{ˇ}{\phi }}{k_0}\text{ or }E=2\pi ^2U_0^2;$$
$`(18)`$
the system has a single energy eigenstate with the energy eigenvalue given in (18). The associated normalized eigenfunction is then
$$\varphi (p)=\sqrt{\frac{2}{\pi }}\frac{k_0^{^{\frac{3}{2}}}}{p^2+k_0^2}.$$
$`(19)`$
It is important to emphasize what we have shown: a constant potential in momentum space admits a bound state. Obviously then in this representation we have not the freedom of changing the origin of the potential energy by adding a constant. In momentum space the potential energy is undetermined not up to a constant value but up to a Dirac-delta function potential; that is, if you take an arbitrary potential $`U(p)`$ in momentum space, the physics of the problem is not changed when you consider instead the modified potential $`U^{}(p)=U(p)+\gamma \delta (p)`$ with $`\gamma `$ an arbitrary constant, whereas the change $`U^{\prime \prime }(p)=U(p)+\gamma U^{}(p)`$ does indeed change the physics. The reader can prove by herself this elementary fact.
This discussion is going forward apparently with no trouble; we have to acknowledge though that for getting to the condition (18), we quickly passed over a very important point, the integral of the right hand side of (17) does not exist in the ordinary Riemann sense. To obtain our result you need to do it instead in the distribution sense, regarding the momentum space function $`\varphi (p)`$ as a linear functional acting upon square integrable functions, as corresponds to possible state functions of a quantum system. Such idea is also behind the usefulness of the delta functions as generalized basis for the free particle states in example A.
To particularize to the present situation, this amounts to make $`\widehat{\phi }`$ convergent (hence meaningful) when acting on any state function (Richtmyer 1978, Böhm 1979). A direct way of accomplishing this is, as usually done in theoretical physics (Taylor 1972, Frauenfelder and Henley 1974, Griffiths 1987), to get the mentioned integral come into existence in a principal value sense (Mathews and Walker 1970). To this end first multiply the right hand side of (17) times an $`\mathrm{exp}(iϵp)`$ complex factor, then perform the integral using contour integration in the complex plane and, at the very end, take the limit $`ϵ0`$. With such provisos considered, it is not difficult getting the result (18). However, this means that the functions involved in our discussion have to be considered as linear functionals or generalized functions, as can be done—perhaps it would be better to say: should be done—for every wave function of a quantum system (Messiah 1976, Böhm 1979); forgetting this fact can produce erroneous results as it is exemplified by the case discussed in (Núñez-Yépez and Salas-Brito 1987).
It is to be noted that the free particle potential acts as a confining potential in momentum space; it allows, for each—out of a nonnegative continuum—energy value, just two choices for the momentum: $`|p_E|`$ and $`|p_E|`$; such extreme restriction is also reflected in the wave functions, they are Dirac delta functions which peak at the just quoted values of $`p`$. On the other hand, the constant potential, which does not restrict the possible values of the momentum in the severe way of the zero potential, is not as confining in momentum space and allows a single energy eigenvalue whose associated eigenstate requires a very wide range of momenta \[given in (19)\] to exist. At this point we invite the reader to try to solve the problem of a particle inside an infinite potential box—in momentum space. This is a simple and nice exercise to test the intuition on the differences between the momentum and the coordinate representation; it is not difficult to conclude that, in this case, the eigenfunctions are also Dirac delta functions with a lightly but subtly modified relation linking energy and momentum.
C. Motion in a potential step
In this case $`U(p)`$ is given in (6). Using such potential, the Schrödinger equation becomes a simple Volterra integral equation
$$\left(\frac{p^2}{2}E\right)\varphi (p)+i\alpha _{\mathrm{}}^p\varphi (p^{})𝑑p^{}.$$
$`(20)`$
To solve this equation, we derive both members and, using $`k_0^22E`$, we obtain a very simple differential equation
$$\frac{d\varphi (p)}{dp}=2\frac{pi\alpha }{p^2+k_0^2}\varphi (p),$$
$`(21)`$
whose solution is
$$\varphi _{k_0}(p)=\frac{A}{p^2+k_0^2}\left[\frac{k_0ip}{k_0+ip}\right]^{\alpha /k_0}$$
$`(22)`$
with $`A`$ an integration constant.
The energy eigenvalues follow, not from a consistency condition as in the last example, B, but from the requirement that the eigenfunctions be single valued. This is only possible if $`\alpha /k_0`$ takes nonnegative integer values (Churchill 1960), i.e. if $`k_0=\alpha /n`$, $`n=1,2,\mathrm{}`$, the value $`n=0`$ is not allowed for $`\varphi (p)`$ would vanish identically in that case (Núñez-Yépez et al. 1987). Thus, the system has an infinite number of bound energy eigenstates with energies given by
$$E_n=\frac{\alpha ^2}{2n^2},n=1,2,\mathrm{};$$
$`(23)`$
the normalization of the eigenfunctions requires that $`A=(2\alpha ^3/n^3\pi )^{1/2}`$ in equation (22).
A very important property of the eigenfunctions is
$$_{\mathrm{}}^+\mathrm{}\varphi (p)𝑑p=0,$$
$`(24)`$
this is required to guarantee the Hermiticity of the Hamiltonian operator of the problem (Andrews 1976, Núñez-Yépez et al. 1987, Salas-Brito 1990). We pinpoint that the potential step in momentum space is particularly interesting because it is closely related to the study of the momentum space behaviour of electrons interacting with the surface of liquid helium, with the properties of the an hydrogen atom in superstrong magnetic fields, and with certain supersymmetric problems (Cole and Cohen 1969, Imbo and Sukhatme 1985, Núñez-Yépez et al. 1987, Martínez-y-Romero et al. 1989c, Salas-Brito 1990).
D. Motion under a harmonic potential
Let us, as our final example, study the motion of a particle under the harmonic potential $`U(p)=f_0\mathrm{cos}(ap)`$, where $`a`$ and $`f_0>0`$ are real constants. The Schrödinger equation is then
$$\frac{p^2}{2}\varphi (p)f_0_{\mathrm{}}^+\mathrm{}\mathrm{cos}[a(pp^{})]\varphi (p^{})𝑑p^{}=E\varphi (p).$$
$`(25)`$
By changing $`p`$ for $`p`$ in (25) we can show that the Hamiltonian commutes with the parity operator, thus its eigenfunctions can be chosen as even or odd functions, i.e. as parity eigenstates.
For solving (25), let us define $`k^22E`$ and, using the identity $`2\mathrm{cos}x=\mathrm{exp}(ix)+\mathrm{exp}(ix)`$, we easily obtain the eigenfunctions as
$$\varphi (p)=\frac{2f_0}{p^2+k^2}\left[\stackrel{ˇ}{\phi }_+e^{iap}+\stackrel{ˇ}{\phi }_{}e^{+iap}\right],$$
$`(26)`$
where the numbers $`\stackrel{ˇ}{\phi }_\pm `$ are defined by
$$\stackrel{ˇ}{\phi }_\pm _{\mathrm{}}^+\mathrm{}e^{\pm iap^{}}\varphi (p^{})𝑑p^{}.$$
$`(27)`$
As in the constant potential (example B), the energy eigenvalues follow from using the definitions (27) back in the eigenfunctions (26)—please remember that we require the functions (26) be regarded as in example B, for doing the integrals (27). The integrals gives us the following two conditions
$$\begin{array}{cc}\hfill \stackrel{ˇ}{\phi }_+=& \frac{2f_0\pi }{k}[\stackrel{ˇ}{\phi }_++\stackrel{ˇ}{\phi }_{}\mathrm{exp}(2ak)],\hfill \\ \hfill \stackrel{ˇ}{\phi }_{}=& \frac{2f_0\pi }{k}[\stackrel{ˇ}{\phi }_{}+\stackrel{ˇ}{\phi }_+\mathrm{exp}(+2ak)].\hfill \end{array}$$
$`(28)`$
As the eigenfunctions can be chosen as even or odd, we can select the numbers $`\stackrel{ˇ}{\phi }_\pm `$ to comply with $`\stackrel{ˇ}{\phi }_+=\pm \stackrel{ˇ}{\phi }_{}`$, so the eigenfunctions are
$$\begin{array}{cc}\hfill \varphi _+(p)=& \frac{A_+}{p^2+k^2}\mathrm{cos}(ap),\hfill \\ \hfill \varphi _{}(p)=& \frac{A_{}}{p^2+k^2}\mathrm{sin}(ap);\hfill \end{array}$$
$`(29)`$
which correspond to the complete set of eigenfunctions of the problem.
From (28) we also get the two equations determining the energy eigenvalues
$$\frac{k}{2f_0\pi }1=\pm e^{2ak}.$$
$`(30)`$
one for the odd and the other for the even state. As can be seen in Figure 1, in general equations (30) admits two solutions, let us call them $`k_+`$ (for the even state) and $`k_{}`$ (for the odd state). Therefore the system has a maximum of two eigenvalues $`E_+=k_+^2/2`$ and $`E_{}=k_{}^2/2`$; the ground state is always even and the excited (odd) state exist only if $`f_0f_0^{\text{crit}}=1/4a\pi 0.0796`$.
The analysis is easily done using the graph shown in Figure 1, where we plot together $`\alpha k+1`$, $`\alpha k1`$ and $`\mathrm{exp}(2ak)`$ \[using $`\alpha 1/(2f_0\pi )=1`$\] against $`k`$, for illustrating the roots of (30). In the plot, we have used the values $`a=1`$, $`f_0=1/2\pi 0.1592`$, corresponding to the roots $`k_{}=0.7968`$ (the leftmost root) and $`k_+1.109`$ (the rightmost root); thus the energy eigenvalues are $`E_+0.6148`$ (the ground state) and $`E_{}0.3175`$ (the excited state). The criterion for the existence of the excited state and the value for $`f_0^{\text{crit}}`$ follows from Figure 1, by noting that such critical value stems from the equality of the slopes of the two curves meeting at the point $`(0,1)`$ in the plot. Notice also that the results previously obtained for the constant potential (example B) can be recovered as a limiting case of the harmonic potential if we let $`a0`$.
4. Concluding remarks
We have discussed four instructive one-dimensional examples in quantum mechanics from the point of view of momentum space. Purportedly we have not made any reference to the problems they represent in the coordinate representation. We expect to contribute with our approach to the development of physical insight for problems posed in the momentum representation and, furthermore, to help students to understand the different features of operators, as opposed to classical variables, in different representations. We also expect to made clear that sometimes it is better to treat a problem from the momentum space point of view since the solution can be simplified. The point at hand is the simple form in which the momentum space eigenfunctions are obtained in the problems discussed here; though these have to be regarded as distributions for obtaining the associated energy eigenvalues.
With goals as the mentioned in mind and to point out other advantages of the momentum representation, in a formal set of lectures and depending on the level of the students, it may be also convenient to discuss more complex problems: as scattering and dispersion relations (Taylor 1972), or the study of resonant states as solutions of a Lippmann-Schwinger equation in momentum space (Hernández and Mondragón 1984), or the 3D hydrogen atom, whose solution using Fock’s method is nicely exposed in (Bransden and Joachaim 1983).
Just in the case you are wondering and have not found the time for doing the transformations yourself, let us say that the problems, save the free particle, we have posed and solved in this paper are known, in the coordinate representation, as 1) the attractive delta potential (Example B), 2) quantum motion under the (quasi-Coulomb) potential $`1/x`$ (Example C) and, finally, the problem of two equal (intensity: $`A=\pi f_0/\sqrt{2}`$), symmetrically placed, attractive delta function potentials, which are displaced by $`2a`$ from one another (Example D).
Acknowledgements
This paper was partially supported by PAPIIT-UNAM (grant IN–122498). We want to thank Q Chiornaya, M Sieriy, K Hryoltiy, C Srida, M Mati, Ch Cori, F Cucho, S Mahui, R Sammi, and F C Bonito for their encouragement. ALSB also wants to thank the attendants of his UAM-A lectures on quantum mechanics (Física Moderna, 99-O term), especially Arturo Velázquez-Estrada and Elías Servín-Hernández, whose participation was relevant for testing the ideas contained in this work.
References
Andrews M 1976 Am. J. Phys. 44 1064
Böhm A 1979 Quantum Mechanics (New York: Springer)
Boya J, Kmiecik M, and Böhm A 1988 Phys. Rev. A 37 3567
Bransden B H and Joachaim C J 1983 Physics of Atoms and Molecules (London: Longman) Ch 2
Cole M W and Cohen M H 1969 Phys. Rev. Lett. 23 1238
Churchill R V 1960 Complex Variables and Applications (New York: McGraw-Hill) pp 59–60
Constantinescu F and Magyari E 1978 Problems in Quantum Mechanics (London: Pergamon) Ch V problem 118
Davtyan L S, Pogosian G S, Sissakian A N and Ter-Antonyan V M 1987 J. Phys. A: Math. Gen. 20 2765
Elliot R J and Loudon R 1960 J. Phys. Chem. Solids 15 196
Fock V A 1935 Z. Phys. 98 145
Frauenfelder H and Henley E M 1974 Subatomic Physics (New Jersey: Prentice-Hall) Ch 6
Griffiths D 1987 Elementary Particles (Singapore: Wiley) Ch 8
Haines L K and Roberts D H 1969 Am. J. Phys. 37 1145
Hernández E and Mondragón A 1984 Phys. Rev. C 29 722
Imbo T D and Sukhatme U P 1985 Phys. Rev. Lett. 54 2184
Loudon R 1959 Am. J. Phys. 27 649
Martínez-y-Romero R P, Núñez-Yépez H N, Vargas C A and Salas-Brito A L 1989a Rev. Mex. Fis. 35 617.
Martínez-y-Romero R P, Núñez-Yépez H N, Vargas C A and Salas-Brito A L 1989b Phys. Rev. A 39 4306.
Martínez-y-Romero R P, Núñez-Yépez H N and Salas-Brito A L 1989c Phys. Lett. A 142 318
Messiah A 1976 Quantum Mechanics Vol I (Amsterdam: North Holland) p 463
Mathews J and Walker R L 1970 Mathematical Methods of Physics (New York: Benjamin) Ch 11 and Appendix 2
Núñez-Yépez H N and Salas-Brito A L 1987 Eur. J. Phys. 8 307
Núñez-Yépez H N, Vargas C A and Salas-Brito A L 1987 Eur. J. Phys. 8 189
Núñez-Yépez H N, Vargas C A and Salas-Brito A L 1988 J. Phys. A: Math. Gen. 21 L651
Núñez-Yépez H N, Vargas C A and Salas-Brito A L 1989 Phys. Rev. A 39 4306
Richtmyer R D 1978 Principles of Advanced Mathematical Physics Vol I (New York: Springer) Ch 2
Sakurai J J 1985 Modern Quantum Mechanics (Reading: Addison-Wesley)
Salas-Brito A L 1990 Átomo de Hidrógeno en un Campo Magnético Infinito: Un Modelo con Regla de Superselección, Tesis Doctoral, Facultad de Ciencias, Universidad Nacional Autónoma de México (in Spanish)
Taylor J R 1972 Scattering Theory (New York: Wiley)
Figure Caption
Figure 1
The figure illustrates the solution to equations (30) determining the energy eigenvalues under the harmonic potential (7). We here plot $`\mathrm{exp}(2ak)`$, $`\alpha k1`$ and $`\alpha k+1`$ against $`k`$ all in the same graph. Just for illustration purposes, we have used the specific values $`a=1`$, $`f_0=1/2\alpha \pi 0.1592`$, (we have defined $`\alpha =1/2f_0\pi `$ and used $`\alpha =1)`$. The critical value of $`f_0`$, giving birth to the excited state, is $`f_0^{\text{crit}}=(4a\pi )^1`$ as can be obtained from the equality of the slopes of the two curves meeting at the point $`(0,1)`$ in the graph. In the situation exemplified by this figure, $`f_0^{\text{crit}}0.0796`$ and the roots of equations (30) are $`k_+1.109`$ and $`k_{}0.7968`$. |
warning/0001/hep-th0001070.html | ar5iv | text | # References
Magnetic Catalysis of Chiral Symmetry Breaking in Gauge Theories.
V. P. Gusynin
Bogolyubov Institute for Theoretical Physics, Kiev-143, 03143 Ukraine
## Abstract
Non-perturbative effect of the formation of a chiral symmetry breaking condensate $`\overline{\psi }\psi `$ and of a dynamically generated fermion mass in QED in the presence of an external magnetic field is considered. The dynamical mass of a fermion (energy gap in the fermion spectrum) is shown to depend essentially nonanalytically on the renormalized coupling constant $`\alpha `$ in a strong magnetic field. Possible applications of this effect are discussed.
The dynamics of fermions in a strong external magnetic field has been attracting much attention during last years. Perhaps, the brightest example has been the discovery and theoretical explanation of the fractional Hall effect leading to the 1998 Nobel Prize award (see Nobel lectures by Laughlin, Stormer and Tsui in ) for ”discovery that electrons acting together in strong magnetic fields can form new types of ”particles”, with charges that are fractions of electron charges ”, as is said in press release of the Royal Swedish Academy of Sciencies. Thus, strong magnetic fields can drastically affect the ground state of a system leading to new types of excitations.
In this talk, I will describe one more phenomenon in an external magnetic field: dynamical breaking of chiral symmetry induced by such a field, hence the name magnetic catalysis.(The talk is based on a series of recent papers with V. Miransky and I. Shovkovy.) This effect has been established as a universal phenomenon in $`2+1`$ and $`3+1`$ dimensions: a constant magnetic field leads to the generation of a fermion dynamical mass at the weakest attractive interaction between fermions . The essence of this effect is that electrons behave effectively as (1+1)-dimensional ones when their energy is much less than the Landau gap $`\sqrt{|eB|}`$ ($`B`$ is the magnitude of the magnetic field). The lowest Landau level (LLL) plays here the role similar to that of the Fermi surface in the BCS theory of superconductivity, leading to the dimensional reduction $`DD2`$ in the dynamics of fermion pairing in a magnetic field and to the formation of a chiral condensate at weak coupling. The effect may have interesting applications in cosmology and in condensed matter physics , as will be discussed below.
The effect of magnetic catalysis was studied in Nambu-Jona-Lasino (NJL) models in 2+1 and 3+1 dimensions , it was extended to the case of external non-abelian chromomagnetic fields , finite temperatures and chemical potential , curved spacetime , confirming the universality of the phenomenon. In particular, this phenomenon was considered in ($`3+1`$)-dimensional QED .
We emphasize that we will consider the conventional, weak coupling, phase of QED since the dynamics of the LLL is long-range (infrared), and the QED coupling constant is weak in the infrared region, therefore, the treatment of the nonperturbative dynamics is reliable there. Note that chiral symmetry breaking is not manifested in the weak coupling phase of QED in the absense of a magnetic field, even if it is treated nonperturbatively . We will show that a constant magnetic field $`B`$ changes the situation drastically, namely, it leads to dynamical chiral symmetry breaking in QED for any arbitrary weak interaction.
The Lagrangian density of QED in a magnetic field is
$$=\frac{1}{4}F^{\mu \nu }F_{\mu \nu }+\frac{1}{2}[\overline{\psi },(i\gamma ^\mu D_\mu m_0)\psi ],$$
(1)
where the covariant derivative $`D_\mu `$ is
$$D_\mu =_\mu ie(A_\mu ^{ext}+A_\mu ),A_\mu ^{ext}=(0,\frac{B}{2}x_2,\frac{B}{2}x_1,0),$$
(2)
i.e. we use the so-called symmetric gauge for $`A_\mu ^{ext}`$. Besides the Dirac index, the fermion field carries an additional flavor index $`a=1,2,\mathrm{},N`$. When the bare mass $`m_0=0`$ the Lagrangian density (1) is invariant under the chiral $`SU_L(N)\times SU_R(N)\times U_V(1)`$ symmetry (we will not discuss the dynamics related to the anomalous symmetry $`U_A(1)`$).
We consider first the problem of free relativistic fermions in a magnetic field in $`3+1`$ dimensions and compare it with the same problem in $`2+1`$ dimensions. We will see that the roots of the fact that a magnetic field is a catalyst of chiral symmetry breaking are actually in this dynamics.
The energy spectrum of fermions in a constant magnetic field is:
$$E_n(p_3)=\pm \sqrt{m_0^2+2|eB|n+p_3^2},n=0,1,2\mathrm{}$$
(3)
(the Landau levels). Each Landau level is infinitely degenerate. As the fermion mass $`m_0`$ goes to zero, there is no energy gap between the vacuum and the lowest Landau level with $`n=0`$. The density of states of fermions on the energy surface with $`E_0=0`$ is given by
$$\nu _0=\frac{|eB|N}{4\pi ^2},3+1\text{dimensions,\hspace{0.17em} and}\nu _0=\frac{|eB|N}{2\pi },2+1\text{dimensions}.$$
(4)
The dynamics of the LLL plays the crucial role in catalyzing spontaneous breaking of chiral symmetry. In particular, the density $`\nu _0`$ plays the same role here as the density of states on the Fermi surface $`\nu _F`$ in the theory of superconductivity. The next important point is that the dynamics of the LLL is essentially ($`1+1`$)-dimensional. Indeed, let us consider the fermion propagator in a magnetic field which was calculated by Schwinger long ago and has the following form (in the chosen gauge):
$$S(x,y)=\mathrm{exp}\left(\frac{ie}{2}(xy)^\mu A_\mu ^{ext}(x+y)\right)\stackrel{~}{S}(xy),$$
(5)
where the Fourier transform of $`\stackrel{~}{S}`$ is
$`\stackrel{~}{S}(p)={\displaystyle \underset{0}{\overset{\mathrm{}}{}}}𝑑s\mathrm{exp}\left[is\left(p_0^2p_3^2p_{}^2{\displaystyle \frac{\mathrm{tan}(eBs)}{eBs}}m_0^2\right)\right]`$
$`\left[(p^0\gamma ^0p^3\gamma ^3+m_0)(1+\gamma ^1\gamma ^2\mathrm{tan}(eBs))p_{}\gamma _{}(1+\mathrm{tan}^2(eBs))\right].`$ (6)
Here $`p_{}=(p_1,p_2)`$, $`\gamma _{}=(\gamma _1,\gamma _2)`$ (to get an expression in $`2+1`$ dimensions we should put $`p_3=0`$ in (6)). The propagator $`\stackrel{~}{S}(p)`$ can be decomposed over the Landau level poles as follows :
$$\stackrel{~}{S}(p)=i\mathrm{exp}\left(\frac{p_{}^2}{|eB|}\right)\underset{n=0}{\overset{\mathrm{}}{}}(1)^n\frac{D_n(eB,p)}{p_0^2p_3^2m_0^22|eB|n}$$
(7)
with
$`D_n(eB,p)`$ $`=`$ $`(\widehat{p}_{}+m_0)[(1i\gamma ^1\gamma ^2\mathrm{sign}(eB))L_n\left(2{\displaystyle \frac{p_{}^2}{|eB|}}\right)`$
$``$ $`(1+i\gamma ^1\gamma ^2\mathrm{sign}(eB))L_{n1}\left(2{\displaystyle \frac{p_{}^2}{|eB|}}\right)]+4\stackrel{}{p}_{}\stackrel{}{\gamma }_{}L_{n1}^1(2{\displaystyle \frac{p_{}^2}{|eB|}}),`$
where $`L_n(x)`$ are the generalized Laguerre polynomials ($`L_nL_n^0`$, $`L_1^\alpha (x)=0`$). Eq.(7) implies that at $`p_{}^2,m_0^2\sqrt{|eB|}`$, the LLL with $`n=0`$ dominates and we can write
$`\stackrel{~}{S}(p)2i\mathrm{exp}({\displaystyle \frac{p_{}^2}{|eB|}}){\displaystyle \frac{\widehat{p}_{}+m_0}{p_{}^2m_0^2}}O^{()},`$ (8)
where $`\widehat{p}_{}=p^0\gamma ^0p^3\gamma ^3`$ and $`\widehat{p}_{}^2=(p^0)^2(p^3)^2`$. The matrix $`O^{()}(1i\gamma ^1\gamma ^2\mathrm{sign}(eB))/2`$ is the projection operator on the fermion states with the spin polarized along the magnetic field. This point and Eq. (8) clearly demonstrate the (1+1)-dimensional character of the dynamics of fermions in the LLL. This property is preserved also in the case when the fermion mass is generated dynamically. Since at $`m_0^2,p_{}^2,p_{}^2|eB|`$ the LLL pole dominates in the fermion propagator, one concludes that the dimensional reduction ($`DD2`$) takes place for the infrared dynamics in a strong ($`|eB|>>m_0^2`$) magnetic field. Such a dimensional reduction reflects the fact that the motion of charged particles is restricted in directions perpendicular to the magnetic field.
Let us first calculate the chiral condensate in $`2+1`$ dimensions for free four-component fermions:
$`0|\overline{\psi }\psi |0=\underset{xy}{lim}\mathrm{tr}S(x,y)={\displaystyle \frac{i}{(2\pi )^3}}\mathrm{tr}{\displaystyle d^3p\stackrel{~}{S}_E(p)}`$
$`={\displaystyle \frac{4m_0N}{(2\pi )^3}}{\displaystyle d^3p_{1/\mathrm{\Lambda }^2}^{\mathrm{}}𝑑s\mathrm{exp}\left[s\left(m_0^2+p_3^2+\stackrel{}{p}_{}^2\frac{\mathrm{tanh}(eBs)}{eBs}\right)\right]}`$
$`=m_0{\displaystyle \frac{|eB|N}{2\pi ^{3/2}}}{\displaystyle _{1/\mathrm{\Lambda }^2}^{\mathrm{}}}{\displaystyle \frac{ds}{s^{1/2}}}e^{sm_0^2}\mathrm{coth}\left(|eBs|\right){\displaystyle \frac{|eB|N}{2\pi }},m_00,`$ (9)
where $`\mathrm{\Lambda }`$ is an ultraviolet cutoff in Euclidean space and, for concretness, we consider $`m_00`$. Thus, as $`m_00`$, the condensate $`0|\overline{\psi }\psi |0`$ remains non-zero due to the LLL. Note that the expression (9) is nothing else as the Banks-Casher formula relating the fermion condensate to the level density of the Dirac operator at zero eigenvalue . The appearance of the condensate in the chiral (flavor) limit, $`m_00`$, signals the spontaneous breakdown of the chiral (flavor) symmetry even for free fermions in a magnetic field at $`D=2+1`$ .
Repeating the same calculation of the chiral condensate in $`3+1`$ dimensions we would get
$$0|\overline{\psi }\psi |0m_0\frac{|eB|N}{4\pi ^2}\left(\mathrm{ln}\frac{\mathrm{\Lambda }^2}{m_0^2}+O(1)\right),m_00,$$
(10)
i.e. the condensate is zero and there is no chiral symmetry breaking. Note, however, the appearance of logarithmic singularity in (10) due to the LLL dynamics. As we will see below, switching on even a weak attraction between fermions leads to the formation of chiral condensate in ($`3+1`$)-dimensional case.
The above consideration suggests that there is a universal mechanism for enhancing the generation of fermion masses by a strong magnetic field: the fermion pairing takes place essentially for fermions at the LLL and this pairing dynamics is $`(1+1)`$-dimensional in the infrared region.This is the main reason why in a magnetic field spontaneous chiral symmetry breaking takes place even at the weakest attractive interaction between fermions in $`3+1`$ dimensions .
Now we shall consider QED in $`3+1`$ dimensions whose Lagrangian is given by Eq. (1). To study chiral symmetry breaking one has to solve the Schwinger-Dyson (SD) equation for the dynamical fermion mass. The SD equation for the fermion propagator $`G(x,y)`$ in an external field has the form
$`G^1(x,y)=S^1(x,y)+\mathrm{\Sigma }(x,y),`$ (11)
$`\mathrm{\Sigma }(x,y)=4\pi \alpha \gamma ^\mu {\displaystyle G(x,z)\mathrm{\Gamma }^\nu (z,y,z^{})𝒟_{\nu \mu }(z^{},x)d^4zd^4z^{}}.`$ (12)
Here $`S(x,y)`$ is the bare fermion propagator (5) in the external field $`A_\mu ^{ext}`$, $`\mathrm{\Sigma }(x,y)`$ is the fermion mass operator, and $`𝒟_{\mu \nu }(x,y)`$, $`\mathrm{\Gamma }^\nu (x,y,z)`$ are the full photon propagator and the full amputated vertex. The full photon propagator satisfies the equations
$`𝒟_{\mu \nu }^1(x,y)`$ $`=`$ $`D_{\mu \nu }^1(xy)+\mathrm{\Pi }_{\mu \nu }(x,y),`$ (13)
$`\mathrm{\Pi }_{\mu \nu }(x,y)`$ $`=`$ $`4\pi \alpha \text{tr}\gamma _\mu {\displaystyle d^4ud^4zG(x,u)\mathrm{\Gamma }_\nu (u,z,y)G(z,x)},`$ (14)
where $`D_{\mu \nu }(xy)`$ is the free photon propagator and $`\mathrm{\Pi }_{\mu \nu }(x,y)`$ is the polarization operator.
It is not difficult to show directly from the SD equations (11), (12), (13) and (14) that substitutions
$`G(x,y)=\mathrm{exp}\left(iex^\mu A_\mu ^{ext}(y)\right)\stackrel{~}{G}(xy),`$ (15)
$`\mathrm{\Gamma }(x,y,z)=\mathrm{exp}\left(iex^\mu A_\mu ^{ext}(y)\right)\stackrel{~}{\mathrm{\Gamma }}(xz,yz),`$ (16)
$`𝒟_{\mu \nu }(x,y)=\stackrel{~}{𝒟}_{\mu \nu }(xy),`$ (17)
$`\mathrm{\Pi }_{\mu \nu }(x,y)=\stackrel{~}{\mathrm{\Pi }}_{\mu \nu }(xy)`$ (18)
lead to equations for translation invariant parts of Green’s functions. In other words, in a constant magnetic field, the Schwinger phase is universal for Green’s functions containing one fermion field, one antifermion field, and any number of photon fields, and the full photon propagator is translation invariant.
We solve the SD equation for the fermion propagator in the so-called ladder approximation when the full vertex and full photon propagator are replaced by their bare ones. We have
$$\stackrel{~}{G}(x)=\stackrel{~}{S}(x)4\pi \alpha d^4x_1d^4y_1e^{ixA(x_1)+ix_1A(y_1)}\stackrel{~}{S}(xx_1)\gamma ^\mu \stackrel{~}{G}(x_1y_1)\gamma ^\nu \stackrel{~}{G}(y_1)𝒟_{\mu \nu }(x_1y_1),$$
(19)
where the shorthand $`xA^{ext}(y)`$ stands for $`x^\mu A_\mu ^{ext}(y)`$.
First, let us show that the solution to the above equation, $`\stackrel{~}{G}(x)`$, allows the factorization of the dependence on the parallel and perpendicular coordinates,
$$\stackrel{~}{G}(x)=\frac{i}{2\pi l^2}\mathrm{exp}\left(\frac{x_{}^2}{4l^2}\right)g\left(x_{}\right)O^{()}.$$
(20)
Notice that this form for $`\stackrel{~}{G}(x)`$ is suggested by a similar expression for the bare propagator,
$$\stackrel{~}{S}(x)=\frac{i}{2\pi l^2}\mathrm{exp}\left(\frac{x_{}^2}{4l^2}\right)s\left(x_{}\right)O^{()},$$
(21)
with
$$s\left(x_{}\right)=\frac{d^2k_{}}{(2\pi )^2}e^{ik_{}x_{}}\frac{\widehat{k}_{}+m}{k_{}^2m^2},$$
(22)
taken in the LLL approximation (here $`l=|eB|^{1/2}`$ is the magnetic length). Performing the integrations we arrive at
$`g\left(x_{}\right)=s\left(x_{}\right)`$ $`+`$ $`4\pi \alpha {\displaystyle \frac{d^4q}{(2\pi )^4}d^2x_1^{}d^2y_1^{}\mathrm{exp}\left(\frac{(q_{}l)^2}{2}iq_{}(x_1^{}y_1^{})\right)s(x^{}x_1^{})}`$ (23)
$`\times `$ $`\gamma _{}^\mu g(x_1^{}y_1^{})\gamma _{}^\nu g(y_1^{})𝒟_{\mu \nu }(q_{},q_{}).`$
Regarding this equation, it is necessary to emphasize that the “perpendicular” components of the $`\gamma `$-matrices are absent in it. Indeed, because of the identity $`O^{()}\gamma _{}^\mu O^{()}=0`$, all those components are killed by the projection operators coming from the fermion propagators. By switching to the momentum space, we obtain
$`g^1\left(p_{}\right)=s^1\left(p_{}\right)4\pi \alpha {\displaystyle \frac{d^4q}{(2\pi )^4}\mathrm{exp}\left(\frac{(q_{}l)^2}{2}\right)\gamma _{}^\mu g(p^{}q^{})\gamma _{}^\nu 𝒟_{\mu \nu }(q_{},q_{})}.`$ (24)
The general solution to this equation is given by the ansatz,
$$g\left(p_{}\right)=\frac{A\widehat{p}_{}+B}{A^2p_{}^2B^2},$$
(25)
where $`A`$ and $`B`$ are functions of $`p_{}^2`$. Making use of this as well as of the explicit form of the photon propagator in the Feynman gauge, we get that the function $`A=1`$ while for the mass function we get the following integral equation
$$B(p^2)=m_0+\frac{\alpha }{2\pi ^2}\frac{d^2qB(q^2)}{q^2+B^2(q^2)}\underset{0}{\overset{\mathrm{}}{}}\frac{dx\mathrm{exp}(xl^2/2)}{x+(qp)^2}$$
(26)
(henceforth we will omit the symbol $``$ from $`p`$ and $`q`$). Thus the SD equation has been reduced to a two–dimensional integral equation. Of course, this fact reflects the two–dimensional character of the dynamics of electrons from LLL.
Analytical and numerical analysis of this equation were performed in for the case $`m_0=0`$ and in for nonzero bare mass. The numerical analysis showed that the so called linearized approximation, with $`B(q^2)`$ replaced by the total mass $`m_{\mathrm{tot}}B(0)`$ in the denominator of Eq. (26), is an excellent approximation. Then we get
$$B(p^2)=m_0+\frac{\alpha }{2\pi ^2}\frac{d^2qB(q^2)}{q^2+m_{tot}^2}\underset{0}{\overset{\mathrm{}}{}}\frac{dx\mathrm{exp}(xl^2/2)}{x+(qp)^2}.$$
(27)
As was shown in (see Appendix C), in the case of weak coupling $`\alpha `$ and for $`m_0=0`$, the function $`B(p)`$ remains almost constant in the range of momenta $`0<p^2<1/l^2`$ and decays like $`1/p^2`$ outside that region. To get an estimate for $`m_{dyn}B(0)`$ at $`\alpha <<1`$, we set the external momentum to be zero and notice that the main contribution in the integral on the right hand side of Eq.(27) is formed in the infrared region with $`q^2<1/l^2`$. The latter validates in its turn the substitution $`B(q)B(0)`$ in the integrand of (26), and we finally come to the following gap equation:
$$B(0)\frac{\alpha }{2\pi ^2}B(0)\frac{d^2q}{q^2+m_{dyn}^2}\underset{0}{\overset{\mathrm{}}{}}\frac{dx\mathrm{exp}(l^2x/2)}{q^2+x},$$
(28)
which gives the expression for the dynamical fermion mass (energy gap in the fermion spectrum):
$$m_{dyn}C\sqrt{eB}\mathrm{exp}\left[\sqrt{\frac{\pi }{\alpha }}\right],$$
(29)
where the constant $`C`$ is of order one and $`\alpha =e^2/4\pi `$ is the renormalized coupling constant related to the scale $`\sqrt{eB}`$. The exponential factor in $`m_{dyn}`$ displays the nonperturbative nature of this result. It can be shown also that the expression (29) for the dynamical mass is gauge invariant .
A more accurate analysis , which takes into account the momentum dependence of the mass function, leads to the result
$$m_{dyn}C\sqrt{|eB|}\mathrm{exp}\left[\frac{\pi }{2}\sqrt{\frac{\pi }{2\alpha }}\right].$$
(30)
The ratio of the powers of this exponent and that in Eq.(29) is $`\pi /2\sqrt{2}1.1`$, thus the approximation used above is rather reliable.
We note that $`m_{dyn}`$ has rather unusual $`1/\sqrt{\alpha }`$ behavior of the exponents in (29) and (30). Similar dependence was found recently in QCD for a quark gap arising at high densities (color superconductivity) . The reason for such a behavior in both cases is the same: the presence of long-range interactions.
To study chiral symmetry breaking in an external field at nonzero temperature we use the imaginary time formalism. Now the analogue of the equation (27) (with $`m_0=0`$ and the replacement $`m_{dyn}m^2(T)`$ in the denominator) reads
$$B(\omega _n^{},p)=\frac{\alpha }{\pi }T\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\underset{\mathrm{}}{\overset{\mathrm{}}{}}\frac{dkB(\omega _n,k)}{\omega _n^2+k^2+m^2(T)}\underset{0}{\overset{\mathrm{}}{}}\frac{dx\mathrm{exp}(l^2x/2)}{(\omega _n\omega _n^{})^2+(kp)^2+x},$$
(31)
where $`\omega _n=\pi T(2n+1)`$ are Matsubara frequencies.
If we now take $`n^{}=0,p=0`$ in the left hand side of Eq.(31) and put $`B(\omega _n,k)B(\omega _0,0)=const`$ in the integrand, we come to the equation
$$1=\frac{\alpha }{\pi }T\underset{n=\mathrm{}}{\overset{\mathrm{}}{}}\underset{\mathrm{}}{\overset{\mathrm{}}{}}\frac{dk}{\omega _n^2+k^2+m^2(T)}\underset{0}{\overset{\mathrm{}}{}}\frac{dx\mathrm{exp}(l^2x/2)}{(\omega _n\omega _0)^2+k^2+x}.$$
(32)
The equation for the critical temperature is obtained putting $`m(T_c)=0`$ and this determines the critical temperature
$`T_c\sqrt{|eB|}\mathrm{exp}\left[\sqrt{{\displaystyle \frac{\pi }{\alpha }}}\right]m_{dyn}(T=0),`$ (33)
where $`m_{dyn}`$ is given by (29). The relationship $`T_cm_{dyn}`$ between the critical temperature and the zero temperature fermion mass was obtained also in NJL model in (2+1)- and (3+1)-dimensions ( and second paper in Ref.). The constant $`C`$, in the relation $`T_c=Cm_{dyn}`$, is of order one and can be calculated numerically. We note that the photon thermal mass, which is of the order of $`\sqrt{\alpha }T`$ , cannot change our result for the critical temperature.
Taking into account the non-zero bare electron mass we come to the equation for the total mass $`m`$:
$$m\mathrm{cos}\left(\sqrt{\frac{\alpha }{2\pi }}\mathrm{log}\frac{|eB|}{m^2}\right)=m_0.$$
(34)
It can be shown that the itterative solution of last equation reproduces all leading double logarithmic terms in perturbation theory:
$$m=m_0[1+\frac{\alpha }{4\pi }\mathrm{log}^2\frac{|eB|}{m_0^2}+\frac{5}{24}\left(\frac{\alpha }{2\pi }\mathrm{log}^2\frac{|eB|}{m_0^2}\right)^2+\frac{61}{720}\left(\frac{\alpha }{2\pi }\mathrm{log}^2\frac{|eB|}{m_0^2}\right)^3+\mathrm{}.]$$
(35)
From Eq.(34) we can estimate the dynamical mass due to a magnetic field. For fields of the order of $`10^{14}G`$ which are realized on surfaces of neutron stars we get $`(mm_0)/m10\%`$. In the real QED the expansion parameter $`\eta \frac{\alpha }{2\pi }\mathrm{log}^2(|eB|/m_0^2)1`$ in (35) explores the transition between the perturbative regime $`\eta 1`$ and the nonperturbative massless QED regime $`\eta 1`$. The value of the parameter $`\eta 1`$ is reached at fields of the order $`10^{26}G`$. We recall that strong magnetic fields ($`B10^{24}G`$) might have been generated during the electroweak phase transition . It has been speculated in Refs. that the character of electroweak phase transition could be affected by a generation of a dynamical electron mass under such strong fields. The nonperturbative regime becomes prevailing over the perturbative one for values of $`\eta `$ of the order of $`2.35`$ what corresponds to magnetic fields $`10^{32}G`$. Ambjørn and Olesen have claimed that even larger fields, $`10^{33}G`$, would be necessary at early stages of the Universe to explain the observed large-scale galactic magnetic fields.
Since the induced fermion dynamical mass contains an exponential factor (see (29), (30) ) it is quite small at all reasonable values of the coupling $`\alpha `$, therefore, there are tiny chances to find implications of the magnetic catalysis phenomenon in real experiments. However, it was shown recently that the Yukawa coupling and scalar-scalar interaction can considerably enhance the fermion dynamical mass (according to the dynamical mass is estimated to be $`m_{dyn}0.6\sqrt{|eB|}`$). The most immediate physical implication would be then in the electroweak theory.
Another interesting application of the magnetic catalysis phenomenon is found in ($`2+1`$)-dimensional condensed matter systems , given the suggestions that high-temperature superconductors can be described effectively by ($`2+1`$) relativistic field theories like NJL or QED (the relativistic (Dirac) nature of the fermion fields is related to the fact that they describe the quasi-particle excitations about the nodes of a $`d`$-wave superconducting gap). According to recent experiments , at temperatures significantly lower than $`T_c`$ of superconductivity, the thermal conductivity, as a function of a magnetic field perpendicularly applied to the cuprate planes, displays a sharp break in its slope at a transition field $`B_\kappa `$, followed then by a plateaux region in which it ceases to change with increasing field. The critical temperature for appearance of the kink-like behavior scales with the magnetic field as $`T_\kappa \sqrt{e|B|}`$. This phenomenon may indicate the opening of a second gap, at the nodes of the $`d`$-wave superconducting gap, that depends on the strength of the applied magnetic field . Indeed, as we saw, in ($`2+1`$)-dimensional systems the chiral condensate appears even in absence of interaction between fermions. The dynamically generated fermion mass scales with a magnetic field like $`m_{dyn}\sqrt{e|B|}`$ in $`2+1`$ NJL model , and $`m_{dyn}\alpha \mathrm{log}(\sqrt{|eB|}/\alpha )`$ in QED3 . The critical temperature for vanishing of the dynamical mass is determined by the dynamical mass at zero temperature (see Eq.(33)) and scales with a magnetic field in a way quite similar to the scaling law found in experiments.
In conclusion, we discuss very breifly the role of higher order radiative corrections in the magnetic catalysis problem. As was shown in Ref., because of the (1+1)-dimensional form of the fermion propagator of the LLL fermions, there are relevant higher order contributions. In particular, considering this problem in the improved rainbow approximation (with the bare vertex in the Schwinger-Dyson equations for both the fermion propagator and the polarization operator ), it was shown that the fermion mass is given by Eq. (30) but with $`\alpha \alpha /2`$. Recently we have shown that there exists a special (non-local) gauge in which the SD equations written in the improved rainbow approximation are reliable: in other words, in that gauge there exists a consistent truncation of the Schwinger-Dyson equations for this non-perturbative problem. The expression for $`m_{dyn}`$ takes the following form,
$$m_{dyn}=\stackrel{~}{C}\sqrt{|eB|}F(\alpha )\mathrm{exp}\left[\frac{\pi }{\alpha \mathrm{ln}\left(C_1/N\alpha \right)}\right],$$
(36)
where $`N`$ is the number of fermion flavors, $`F(\alpha )(N\alpha )^{1/3}`$, $`C_11.82`$ and the constant $`\stackrel{~}{C}`$ is of order one.
Thus, the magnetic catalysis of chiral symmetry breaking in QED yields a (first, to the best of our knowledge) example in which there exists a consistent truncation of the Schwinger-Dyson equations in the problem of dynamical symmetry breaking in a (3+1)-dimensional gauge theory without fundamental scalar fields. I am grateful to the members of the Institute for Theoretical Physics of the University of Heidelberg, especially Prof. M.G. Schmidt, for their hospitality during my stay there. This research has been supported in part by Deutscher Academischer Austauschdienst (DAAD) grant and by the National Science Foundation (USA) under grant No. PHY-9722059. |
warning/0001/hep-ph0001160.html | ar5iv | text | # The KARMEN anomaly, light neutralinos and type II supernovae
## Abstract:
The KARMEN experiment observes a time anomaly in events induced by pion decay at rest. This anomaly can be ascribed to the production of a new weakly interacting particle $`X`$ with mass $`m_X34`$ MeV. We show that a recently proposed identification of the $`X`$ particle with the lightest neutralino $`\chi `$ in the frame work of the MSSM with broken R parity is in contradiction to optical observations of type II supernovae.
Supersymmetric models, decays of $`\pi `$ mesons, hypothetical particles preprint: CERN-TH 2000-014
Introduction—The experiment KARMEN has investigated a variety of neutral and charged current neutrino interactions finding excellent agreement between measured cross-sections and the predictions of the standard model. However, the analysis of the time distribution of events induced by neutrinos from $`\pi ^+`$ and $`\mu ^+`$ decays at rest has revealed an anomaly: the measured distribution for subsequent events differs substantially from the expected exponential distribution with a time constant equal to the muon lifetime of 2.2 $`\mu `$s . As a possible explanation for this anomaly, the KARMEN collaboration proposed that their signal is a superposition of a Gaussian distribution centered at 3.4 $`\mu `$s and the exponential distribution describing muon decay. The Gaussian distribution is interpreted as time-of-flight signature of a hypothetical particle $`X`$ produced at the spallation target, passing through 7 m steel shielding and then decaying in the detector. A maximum likelihood analysis of this hypothesis showed that the probability that the Gaussian signal is a statistical fluctuation is only $`10^4`$. The best-fit values for the mass of the particle $`X`$ are $`m_X33.9`$ MeV, while the branching ratio BR$`(\pi ^+\mu ^++X)`$ and the decay rate $`\mathrm{\Gamma }_{\mathrm{vis}}`$ of $`X`$ into photons and electrons have to fulfill the relation BR$`(\pi ^+\mu ^++X)\mathrm{\Gamma }_{\mathrm{vis}}310^{11}`$s<sup>-1</sup>. Furthermore, the KARMEN data disfavor two-body decays of the $`X`$ particles, because no peak at 17 MeV has been seen in the energy spectra of the anomalous events.
There have been several theoretical works discussing candidates for the new particle $`X`$ . Proposed candidates are an active or sterile neutrino , a scalar boson and a light neutralino . While an active neutrino seems to be excluded by the new experimental limit for BR$`(\pi ^+\mu ^++X)`$, a sterile neutrino was found to be, within stringent limits on its mixing parameters, a viable candidate for the $`X`$ particle. In Ref. , a scalar boson with mass $`104`$ MeV was proposed as $`X`$ particle. Since the energy deposited by the decaying $`X`$ particle in the calorimeter of the KARMEN experiment is only between 11 and 35 MeV, additional light scalars have to be invoked to dilute the energy via cascade decays. Although the model seems to be compatible with laboratory constraints, it looks somewhat artificial. In Ref. , the $`X`$ particle was identified with the lightest neutralino $`\chi `$ in a supersymmetric model with broken R parity. However, the proposed decay mode of the neutralino was a two-body decay, $`\chi \gamma +\nu _\mu `$, and is therefore now disfavored by the KARMEN data. Recently, the authors of Ref. reconsidered this proposal. They introduced two R parity violating operators instead of one, explaining the anomalous pion decay by $`L_2Q_1D_1^c`$ and the neutralino decay $`\chi e^+e^{}\nu _{\mu ,\tau }`$ by $`L_eL_\mu E_e^c`$ or $`L_eL_\tau E_e^c`$. They found that this scenario is consistent with accelerator bounds provided that the neutralino is dominantly a bino and has a lifetime $`\tau _\chi =0.24100`$ s. Moreover, they showed that regions exist in the MSSM parameter space which allow such a bino without excessive fine-tuning. The same conclusion was obtained already earlier in Ref. .
The possibility that KARMEN has observed already over several years a supersymmetric particle is intriguing. In particular, the scenario of Ref. with its rather constrained parameter space would have impact on searches for supersymmetry at accelerators as well as on searches for dark matter. It is therefore of interest to check this model against all possible constraints. In this brief note, we discuss the influence of neutralino emission on core collaps supernovae (SN). We find that the production of neutralinos with $`m_\chi 34`$ MeV is practically not suppressed in the SN core. The energy deposited in the SN envelope by decaying neutralinos is for all allowed squark masses larger than the value allowed by observations of the light curves of type II supernovae<sup>1</sup><sup>1</sup>1This argument was first used by S.W. Falk and D.N. Schramm in Ref. to restrict radiative decays of neutrinos.. We conclude that a light neutralino with lifetime $`\tau _\chi =0.24100`$ s is in contradiction to observations.
Neutralino production—In the SN core, neutralinos are mainly produced in nucleon-nucleon bremsstrahlung $`NNNN\chi \chi `$ and in $`e^+e^{}`$ annihilations. The spin-averaged squared matrix element of the latter process is for a bino-like neutralino, degenerated selectron masses $`M_{\stackrel{~}{e}}`$ and $`m_e=0`$
$$|\overline{M}|^2=\frac{g^{\mathrm{\hspace{0.25em}4}}}{2}(Y_L^4+Y_R^4)\left\{\left(\frac{tM_\chi ^2}{tM_e^2}\right)^2+\left(\frac{uM_\chi ^2}{uM_e^2}\right)^2\frac{2sM_\chi ^2}{(tM_e^2)(uM_e^2)}\right\}.$$
(1)
Neglecting the Pauli-blocking factors of the neutralinos, the cross-section times the relative velocity is
$$v\sigma =\frac{1}{(2\pi )^2}\frac{1}{8EE^{}}\frac{d^3k}{2\omega }\frac{d^3k^{}}{2\omega ^{}}|\overline{M}|^2\delta ^{(4)}(p+p^{}kk^{}),$$
(2)
where $`p=(E,\stackrel{}{p})`$, $`p^{}=(E^{},\stackrel{}{p}^{})`$ are the four-momenta of the electron and positron and $`k=(\omega ,\stackrel{}{k})`$, $`k^{}=(\omega ^{},\stackrel{}{k}^{})`$ of the neutralinos, respectively. The emissivity $`\epsilon (e^+e^{}\chi \chi )`$ follows as
$$\epsilon (e^+e^{}\chi \chi )=\frac{1}{2\pi ^4}𝑑p𝑑p^{}𝑑\vartheta \mathrm{sin}\vartheta \frac{p^2}{e^{(E\mu )/T}+1}\frac{p^{}_{}{}^{}2}{e^{(E^{}+\mu )/T}+1}(E+E^{})v\sigma ,$$
(3)
where $`T`$ is the temperature and $`\mu `$ the chemical potential of the positron-electron plasma. The integration over the angle $`\vartheta =\mathrm{}(\stackrel{}{p},\stackrel{}{p}^{})`$ can be performed analytically, but results in lengthy expressions. Therefore, we have preferred to evaluate directly Eq. (3). In Fig. 1, we show the ratio $`R=\epsilon (x)/\epsilon (0)`$ as function of $`x=m_\chi /T`$ for different values of the degeneracy parameter $`\eta =\mu /T`$. Typical values found in simulations for the temperature and the electron degeneracy inside the SN core are $`T=1040`$ MeV and $`\eta =1030`$ . For an average temperature of $`T=20`$ MeV, i.e. $`x=1.7`$, the neutralino production is even for the low value $`\eta =10`$ only reduced by 40%. A significant suppression of the neutralino production requires $`x>10`$, i.e. temperatures well below those believed to exist in SN cores.
It is natural to assume that the bremsstrahlung process $`NNNN\chi \chi `$ is suppressed roughly by the same factor $`f0.5`$. This assumption allows us to obtain the emissivity simply from the one for the related processes with neutrinos , e.g.,
$$\frac{\epsilon (nnnn\chi \chi )}{\epsilon (nnnn\nu \overline{\nu })}f\frac{16m_W^4\mathrm{tan}^4\vartheta _W}{M_{\stackrel{~}{q}}^4}\left(\frac{_{i=u,d,s}(Y_{L,i}^2+Y_{R,i}^2)\mathrm{\Delta }q_i}{\mathrm{\Delta }u\mathrm{\Delta }d\mathrm{\Delta }s}\right)^2,$$
(4)
where $`\mathrm{\Delta }q_i`$ denotes the spin fraction carried by the quark $`i`$. Using for $`\epsilon (NNNN\nu \overline{\nu })`$ the expression given in Ref. for non-degenerate nucleons and the spin fractions of Ref. , the total energy $`E_\chi `$ emitted into neutralinos from the SN in the free-streaming regime is given by
$$E_\chi 2.710^{22}\frac{\mathrm{erg}}{\mathrm{g}\mathrm{s}}f\left(\frac{250\mathrm{G}\mathrm{e}\mathrm{V}}{M_{\stackrel{~}{q}}}\right)^4\left(\frac{T}{25\mathrm{M}\mathrm{e}\mathrm{V}}\right)^{5.5}\frac{\rho }{2\rho _0}\tau _{\mathrm{burst}}M_{\mathrm{core}}.$$
(5)
With $`f=0.5`$, $`T=25`$ MeV, $`\rho =2\rho _0610^{14}`$ g/cm<sup>3</sup>, a duration of the neutrino and neutralino burst of $`\tau _{\mathrm{burst}}=10`$ s and $`M_{\mathrm{core}}=1.5M_{}`$, we find $`E_\chi =4.010^{56}\mathrm{erg}(250\mathrm{GeV}/M_{\stackrel{~}{q}})^4`$ in the free-streaming regime.
Energy deposition in the SN envelope—An emitted neutralino with lifetime $`\tau _\chi `$ has the probability $`P1\mathrm{exp}(R/\tau _\chi \gamma )`$ to decay inside the SN progenitor with radius $`R`$, where $`\gamma =E_\chi /m_\chi `$. We will use conservatively as radius for the progenitor stars of type-II supernovae the value $`R1000`$ s . Thus for all allowed lifetimes $`\tau _\chi =0.24100`$ s a large fraction of the neutralinos decays inside the envelope, depositing its energy there during the first 10 s after core collapse. There are several possibilities how this energy deposition can be used to restrict or exclude a light, decaying neutralino. First, the electromagnetic displaying of the SN which is expected to start 3 hours after the neutrino burst, when the shock wave reaches the photosphere, should start much earlier. Second, model calculation for SN1987A find that the total energy of the shock is $`E_{\mathrm{sh}}310^{51}`$ erg. This number can be used as upper limit for the energy $`E_{\mathrm{en}}`$ released by the neutralinos in the envelope.
Assuming that the neutrino carries away one third of the energy, neutralinos deposit the energy $`E_{\mathrm{en}}=\frac{2}{3}E_\chi P`$ in the SN envelope. With $`P>0.63`$, we can exclude those parts of the parameter space of the model which give rise to $`E_\chi >710^{51}`$ erg. Thus, squark masses smaller than $`4`$ TeV are excluded in the free-streaming regime.
For very light squarks, nucleon-neutralino interactions become strong enough so that neutralinos are efficiently trapped and $`E_\chi `$ falls below $`710^{51}`$ erg. The results of Ref. suggest, that this possibility is already excluded by direct accelerator searches. Nevertheless, we will reconsider below the trapping regime for a bino like neutralino.
Neutralino opacity—In the calculation of the free-mean path of the neutralino in the SN core, we should take into account the thermal distribution functions of the electrons and nucleons, respectively. A detailed calculation performed in Ref. for massless photinos showed that the thermal cross-section of $`\chi e^{}\chi e^{}`$ can be rather well approximated by the vacuum one. Moreover, the thermal effects can be factored out in neutralino-nucleon scattering, when recoil effects are neglected.
If selectrons are not considerable lighter than the lightest squarks, the dominant opacity source for neutralinos is scattering on nucleons. We estimate the free-mean path $`l_\chi `$ of the neutralino as
$$l_\chi ^1(\chi p)4.210^6\mathrm{cm}^1\left(\frac{250\mathrm{G}\mathrm{e}\mathrm{V}}{M_{\stackrel{~}{q}}}\right)^4\frac{(1Y_n)\rho }{2\rho _0},$$
(6)
$$l_\chi ^1(\chi n)1.010^6\mathrm{cm}^1\left(\frac{250\mathrm{G}\mathrm{e}\mathrm{V}}{M_{\stackrel{~}{q}}}\right)^4\frac{Y_n\rho }{2\rho _0},$$
(7)
where we used as average energy of the neutralino $`E_\chi 3T75`$ MeV and a thermal suppression factor $`0.7`$ (cf. ). Requiring $`l_\chi ^1R=10`$ as trapping criteria with $`\rho =2\rho _0`$, $`Y_n=0.8`$, and $`R=10`$ km as size of the core, we find $`M_{\stackrel{~}{q}}160`$ GeV as borderline between the diffusion and the free-streaming regime.
In the diffusion regime, we estimate the neutralino luminosity $`_\chi `$ assuming blackbody surface emission, $`_\chi =(\pi ^3/15)R_\chi ^2T_\chi ^4`$, from a $`\chi `$-sphere with radius $`R_\chi `$. The position of this sphere is calculated from the optical depth $`\tau `$, $`\tau =_{R_\chi }^{\mathrm{}}𝑑rl_\chi ^1(r)=2/3`$, where we use as density profile $`\rho (r)=\rho _R(R/r)^n`$, as temperature profile $`T(r)=T_R(\rho (r)/\rho _R)^{1/3}`$ and $`Y_n=0.5`$ . We choose density and temperature at the edge of the core as $`T_R=10`$ MeV and $`\rho _R=10^{14}`$g/cm<sup>3</sup>, respectively. The exponent $`n`$ of the profiles is rather model dependent, $`n37`$. Using the scaling relation
$$L_\chi (r)=L_\chi (R)\left(\frac{R}{r}\right)^{\frac{4n}{3}2}$$
(8)
and $`n=5`$, the neutralino luminosity is small enough for $`R_\chi >2.8R`$. This is consistent with $`\tau (R_\chi )=2/3`$ only if the squark masses would be below 20 GeV.
Summary—The production of neutralinos with the mass $`m_\chi 34`$ MeV is practically unsuppressed in a SN core. The upper bound on the sfermion masses in the model of Ref. , $`M_{\stackrel{~}{f}}<1`$ TeV, together with the lower limit $`M_{\stackrel{~}{f}}>100`$ GeV from LEPII leaves no room for a decaying neutralino with the required lifetime $`\tau _\chi =0.24100`$ s: For all allowed values of $`M_{\stackrel{~}{f}}`$, the energy deposited in the SN envelope by decaying neutralinos is in contradiction to the optical observations of type II supernovae.
Note added: The preprint hep-ph/9912465 by I. Goldman, R. Mohapatra and S. Nussinov discusses also the consistency of a light, neutral, decaying fermion $`n^0`$ with SN 1987A. Its discussion is model-independent, using however the assumption that the $`n^0`$-nucleon cross-section is at least a factor of 10 smaller than the neutrino-nucleon cross-section. This assumption does not hold necessarily in the case of the neutralino. |
warning/0001/quant-ph0001061.html | ar5iv | text | # Causality and probabilistic interpretation of quantum mechanicsSubmitted to ”Theoretical and Mathematical Physics”
( )
## Abstract
It is shown that the probabilistic treatment of quantum mechanics can be coordinated with causality of all physical processes. The physical interpretation of quantum-mechanical phenomena such as process of measurement and collapse of quantum state is given.
The purpose of the present paper is the substantiation of a thesis that the probabilistic interpretation of quantum mechanics is quite compatible with the supposition about unique causality of all physical processes. The consideration is performed within the framework of the binary model (see papers ) of quantum mechanics in which it is supposed that there are separate material carriers of corpuscular and wave properties in quantum objects.
Following the basic idea of the papers we shall consider that any quantum object has dynamical and phase degrees of freedom. Carriers of the dynamical degrees of freedom are local. Further they will refer to as nucleuses of the quantum object (to not confuse to atomic nucleuses). Carriers of the phase degrees of freedom are fields, which further will refer to as an information fields or a shell of a quantum object. These fields are spread in the space.
An elementary quantum object consists of one nucleus and a shell (information field) which are coherent each other. It is possible to assume, though it is not necessary, that nucleuses exist in pulsatory (in the time and the space) regime like some splashes of the information field. In this case in microscopic scale a nucleus will not have continuous trajectory in the space - time but in macroscopic scale the trajectory will exist.
Action onto a quantum object can be dynamical and phase. The dynamical action is accompanied by transmission of dynamical quantities. It acts onto the dynamical degrees of freedom, i.e. onto the nucleuses. Respectively, the nucleuses are responsible for corpuscular properties of the quantum object and they contain the information about observables quantities in the latent shape. The phase action is not accompanied by transmission of dynamical quantities (or by very small transmission) and it acts on the phase degrees of freedom. A classical measuring device reacts to the dynamical action of the quantum object. Therefore the device reacts to an elementary quantum object as onto one aggregate. But the concrete result of this reaction is defined by structure of the shell of the quantum object. This structure depends on state of the information field. Further we shall name it physical state of the quantum object and we shall designate by symbol $`\phi `$.
Now we try to formalize these physical ideas about the quantum object. In order to take into account the latency of the information about observables quantities we postulate, that the observables $`\widehat{B}`$ are Hermite elements of noncommutative involute algebra (\*-algebras) A. Let’s consider that to each physical state $`\phi `$ of a quantum object univalently there corresponds a functional on the algebra A. This functional we shall designate by the same symbol $`\phi `$. The term ”the physical state” will denote structure of the information field, and the functional, corresponding to this structure.
Thus, if $`\widehat{B}\text{A}`$ and $`\widehat{B}^{}=\widehat{B}`$ then $`\phi (\widehat{B})=B`$ is a real number (value of the observable $`\widehat{B}`$) which is obtained in the concrete measurement.
Let $`\{\widehat{Q}\}`$ is some maximal set of mutually commuting Hermite elements of the algebra A, i.e. $`\{\widehat{Q}\}\text{A}`$ and
$$\begin{array}{c}\text{ if }\widehat{Q}_i,\widehat{Q}_j\{\widehat{Q}\}\text{ then }[\widehat{Q}_i,\widehat{Q}_j]=0;\hfill \\ \text{ if }\widehat{Q}_i\{\widehat{Q}\},\widehat{Q}_j\text{A}\text{ and }[\widehat{Q}_i,\widehat{Q}_j]=0\text{ then }\widehat{Q}_j\{\widehat{Q}\};\hfill \\ \text{ if }\widehat{Q}_i\{\widehat{Q}\},\widehat{Q}_j\{\widehat{Q}_j\}\text{ then }[\widehat{Q}_i,\widehat{Q}_j]0.\hfill \end{array}$$
The functional $`\phi `$ maps the set $`\{\widehat{Q}\}`$ in a set of real numbers
$$\{\widehat{Q}\}\stackrel{\phi }{}\{Q=\phi (\widehat{Q})\}.$$
For the different functionals $`\phi _i`$, $`\phi _j`$ the sets $`\{\phi _i(\widehat{Q})\}`$, $`\{\phi _j(\widehat{Q})\}`$ can differ and can coincide. If for all $`\widehat{Q}\{\widehat{Q}\}`$ is valid $`\phi _i(\widehat{Q})=\phi _j(\widehat{Q})=Q`$ then we shall term the physical states $`\phi _i`$ and $`\phi _j`$ as $`\{Q\}`$-equivalent. Let’s denote by $`\{\phi \}_Q`$ set of all $`\{Q\}`$-equivalent physical states. Let’s term the set $`\{\phi \}_Q`$ as a quantum state and we shall designate $`\mathrm{\Psi }_Q`$.
We shall take one fundamental supposition touching the physical state: each physical state is unique, i.e. there are no two identical states in the world and the physical states never repeat. It is possible to consider the physical state is determined by all previous history of the concrete physical object and this history for each object is individual. In particular, in two different experiments we necessarily deal with two different physical states. Different physical states $`\phi _i`$, $`\phi _j`$ correspond to different functionals $`\phi _i()`$, $`\phi _j()`$, i.e. always there will be such observable $`\widehat{B}`$, that $`\phi _i(\widehat{B})\phi _j(\widehat{B})`$. From this supposition follows, that each physical state $`\phi _i`$ can be exhibited in form of the functional $`\phi _i()`$ no more, than in one experiment.
For each observable $`\widehat{A}`$ we shall introduce concept ”an actual state”. It is a physical state, in which the observable $`\widehat{A}`$ was measured or will be measured. A set of such states we shall denote by $`[\phi ]^{\widehat{A}}`$. A set $`\{Q\}`$ of equivalent states, actual for an observable $`\widehat{A}`$, we shall designate $`\{\phi \}_Q^{\widehat{A}}`$. Following the standard quantum mechanics, we shall consider that only mutually commuting observables can be measured in one experiment.
The functional $`\phi `$ is not linear. However we shall require, that on Hermite elements of algebra A a functional $`\phi ()`$, corresponding to actual states, would satisfy to the following postulates:
$`1)`$ $`\phi (\lambda \widehat{I})=\lambda ,\widehat{I}\text{ is unity of algebra }\text{A},\lambda \text{ is real number};`$
$`2)`$ $`\phi (\widehat{A}+\widehat{B})=\phi (\widehat{A})+\phi (\widehat{B}),\phi (\widehat{A}\widehat{B})=\phi (\widehat{A})\phi (\widehat{B}),\text{ if }[\widehat{A},\widehat{B}]=0;`$
$`3)`$ $`\underset{\phi }{sup}\phi (\widehat{A}^{}\widehat{A})>0,\text{ if}\widehat{A}0;\phi (\widehat{A}^{}\widehat{A})=0,\text{ if }\widehat{A}=0;`$
$`4)`$ $`\text{for each set }\{Q\}\text{ and each }\widehat{A}\text{A}`$
$`\text{there is }\underset{n\mathrm{}}{lim}{\displaystyle \frac{1}{n}}{\displaystyle \underset{i=1}{\overset{n}{}}}\phi _i(\widehat{A})\mathrm{\Psi }_Q(\widehat{A}),`$
$`\text{where }\{\phi _1,\mathrm{},\phi _n\}\text{ is a random sample of the set}\{\phi \}_Q^{\widehat{A}};`$
$`5)`$ $`\text{for every }\widehat{A},\widehat{B}\text{A}\mathrm{\Psi }_Q(\widehat{A}+\widehat{B})=\mathrm{\Psi }(\widehat{A})+\mathrm{\Psi }(\widehat{B}).`$
We shall extend the functionals $`\phi ()`$ onto anti-Hermitian elements of the algebra A with the help of the equality $`\phi (i\widehat{A})=i\phi (\widehat{A})`$. The functional $`\mathrm{\Psi }_Q()`$, appearing in the fourth postulate, has a meaning of a functional $`\phi ()`$ which is averaged over all $`\{Q\}`$-equivalent actual states. Symbolically it can be represented in the form of Monte-Carlo integral over the actual states:
$$\mathrm{\Psi }_Q(\widehat{A})=_{\phi \{\phi \}_Q^{\widehat{A}}}𝑑\mu (\phi )\phi (\widehat{A}).$$
We shall note that
(2)
$$_{\phi \{\phi \}_Q^{\widehat{A}}}𝑑\mu (\phi )=1.$$
Let us connect the functional $`\mathrm{\Psi }_Q()`$ with each quantum state $`\mathrm{\Psi }_Q`$. The fourth postulate assumes that this functional does not depend on a concrete random sample. Further the term ”a quantum state $`\mathrm{\Psi }_Q`$” we shall use both for the set $`\{\phi \}_Q`$ of physical states and for the corresponding functional $`\mathrm{\Psi }_Q()`$ (the quantum average).
Let $`\widehat{A}^{}\widehat{A}=\widehat{B}\{\widehat{Q}\}`$. If $`\phi \{\phi \}_Q`$, then $`\phi (\widehat{A}^{}\widehat{A})=B`$, where $`B\{Q\}`$. Therefore
$$\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{A})=B=\phi (\widehat{A}^{}\widehat{A})|_{\phi \{\phi \}_Q}.$$
From here
$$\widehat{A}^2\underset{Q}{sup}\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{A})=\underset{\phi [\phi ]^{\widehat{A}}}{sup}\phi (\widehat{A}^{}\widehat{A})>0,\text{ if }\widehat{A}0.$$
As $`\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{B})`$ is a linear (in $`\widehat{A}^{}\widehat{B}`$) positive semidefinite functional then the Cauchy-Bunkyakovsky-Schwarz inequality is valid (see for example )
$$\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{B})\mathrm{\Psi }_Q(\widehat{B}^{}\widehat{A})\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{A})\mathrm{\Psi }_Q(\widehat{B}^{}\widehat{B}).$$
Therefore for $`\mathrm{\Psi }_Q(\widehat{A}^{}\widehat{A})`$ the postulates for square of seminorm of the element $`\widehat{A}`$ are fulfilled. Respectively it is possible to accept $`\widehat{A}`$ for the norm of $`\widehat{A}`$. The algebra A becomes $`C^{}`$ -algebra at such definition of the norm . According to the Gelfand-Naumark theorem (see ) every $`C^{}`$-algebra can be realized as algebra of linear operators in some Hilbert space. Thus, the proposed here construction of quantum mechanics permit the standard realization.
As against usual scheme of quantum mechanics in the proposed construction one additional element is present. It is the physical state $`\phi `$ and the corresponding nonlinear functional $`\phi ()`$. The functional $`\phi ()`$ describes results of individual measurement in a concrete experiment, and the functional $`\mathrm{\Psi }_Q()`$ describes average value of the observable in a series of experiments, performed in identical conditions from the point of view of the observer, i.e. at the same quantum state. The fact of existence of the functionals $`\phi ()`$ and $`\mathrm{\Psi }_Q()`$, satisfying to the enumerated postulates, is proved by all complex of quantum experiments. The standard quantum mechanics is busy in problems describing by the functionals $`\mathrm{\Psi }_Q()`$. However now single quantum phenomena take a great meaning. For example, such phenomena underlie an operation of quantum computer. Therefore it is desirable to supplement the formalism of the standard quantum mechanics by positions, which would allow considering the single quantum phenomena. On the other hand, it is extremely desirable, that such expanded formalism did not give rise to deductions, which would not agree with deductions of the standard quantum mechanics.
Namely for sufficing this condition the postulate of uniqueness of each physical state is accepted. It follows this postulate that the physical state can not be univalently fixed. Really, to fix the functional $`\phi ()`$, we should know its value $`\phi (\widehat{B})`$ for all independent observables $`\widehat{B}`$. Physically it is not realizable. In one experiment we can find $`\phi (\widehat{B}_i)`$ only for mutually commuting observables $`\widehat{B}_i`$, and in different experiments we necessarily deal with different functionals $`\phi ()`$.
The most hard fixing of the functional $`\phi ()`$, which can physically be realized, consists in reference it to some set $`\{\phi \}_Q`$, i.e. to a certain quantum state. For this purpose it is enough to perform measurements of mutually commuting observables. It is possible to be restricted only by independent measurements. In principle it can be done in one experiment. Thus we can not have the complete information about a physical state of a concrete physical object in essence. The maximal observable and controllable information on the physical object is concentrated in the quantum state.
At the same time, the quantum states have some subjective element. From the standard quantum mechanics it is well known that any quantum state can be represented in form of superposition of quantum states which are fixed by one maximal set $`\{\widehat{Q}\}`$ of mutually commuting observables. In the proposed here construction it corresponds to the fact that one physical state can belong to the different quantum states $`\{\phi \}_Q`$ and $`\{\phi \}_R`$. I.e. $`\phi \{\phi \}_Q\{\phi \}_Q`$, where the states $`\{\phi \}_Q`$ are classified by values of the set $`\{\widehat{Q}\}`$ of mutually commuting observables $`\widehat{Q}`$, and $`\{\phi \}_R`$ are classified by values of observables $`\widehat{R}\{\widehat{R}\}`$. The observables $`\widehat{Q}`$ and $`\widehat{R}`$ do not commute among themselves. Then depending on what set ($`\{\widehat{Q}\}`$ or $`\{\widehat{R}\}`$) we shall choose for classification the physical state $`\phi `$ will be referred either to the quantum state $`\{\phi \}_Q`$ or to the quantum state $`\{\phi \}_R`$.
For example, let a spin-free particle decays into two particles $`A`$ and $`B`$ with spins 1/2 which scatter at large distance. Let’s measure a projection of spin onto the axis $`z`$ for the particle $`A`$. Let result will be $`S_z(A)`$. Then using the conservation law, we can state that for the particle $`B`$ with absolute probability the projection of spin onto the axis $`z`$ is equal $`S_z(B)=S_z(A)`$. It denotes that the quantum state of the particle $`B`$ corresponds to such value of the projection of spin onto axis $`z`$. However for the particle $`A`$ we could measure the projection of spin onto axis $`x`$. Let result would be $`S_x(A)`$. Then we could state that the particle $`B`$ is in the quantum state, which corresponds to the value $`S_x(B)=S_x(A)`$ of the observable $`\widehat{S}_x(B)`$.
As any physical operations with the particle $`B`$ are not fulfilled, the physical state in both cases will be same $`\phi \{\phi \}_{S_z(A)}\{\phi \}_{S_x(A)}`$. The different quantum states of the particle $`B`$ are related only to our subjective choice of the device for measurement of the physical state of the particle $`A`$.
This example is the description of the experiment proposed by Bohm for demonstrating of the Einstein-Podolsky-Rosen paradox . In proposed here treatment any paradox is absent. The physical state $`\phi `$ particles $`B`$, which is an objective reality, does not depend on our manipulations with the remote particle $`A`$. Any transmission of an action on distance is absent. The described experiment is an example of indirect measurement at which the information about state of the quantum object is obtained without physical action onto it. Usually interpretation of the indirect experiment arouses the greatest difficulties, first of all, bound with concept of collapse of quantum state (reduction of wave function). In adduced example by our wish we ”channelize” the particle $`B`$ into the quantum state $`\{\phi \}_{S_z(A)}`$, or into the quantum state $`\{\phi \}_{S_x(A)}`$, physically not acting onto the particle $`B`$. Such collapse can be named subjective (or passive) in contrast to objective (active) collapse, which is related to the actual physical action onto the quantum object. About the objective collapse we shall talk later. Here we shall note that the subjective collapse is related to physical impossibility for us to receive the complete objective information about the quantum object (the complete information about the physical state). We should be content by the partial information (quantum state). It depends on our desire what concrete part we prefer to be satisfied by.
It is possible also to use this experiment for demonstrating that, strictly speaking, it would be possible to receive larger the information about state of quantum object, than ascertaining of its membership to this or that quantum state. In the experiment we can measure the projection of spin onto the axis $`z`$ for the particle $`A`$ and the projection onto the axis $`x`$ for the particle $`B`$. In this case we can establish that the physical state $`\phi `$ of the particles $`B`$ belongs to intersection of the corresponding quantum states
(3)
$$\phi \{\phi \}_{S_z(A)}\{\phi \}_{S_x(B)}.$$
But this information has specific character. It refer to the past, more precisely, to the restricted interval in the past from the moment of decay of the spin-free particle to the moment of measurement of the projection of spin of the particle $`B`$. In this moment the information described by the equation (3) will be garbled and will be useless for the further monitoring (or control) of the particle $`B`$.
Now we shall discuss dynamics and temporal evolution of quantum object. As was already spoken, the action onto quantum object can be dynamical and phase. By hypothesis elementary quantum object is nonlocal due to the shell. The different parts of the quantum object can undergo different exterior action and lose mutual coherence. In turn, it should result in disintegration of the quantum object, since all constituent parts of the elementary quantum object must be coherent by hypothesis. As the elementary quantum objects are rather stable structures, there should be a cause, which hinders loss of the coherence. Let’s assume that such cause is a strong phase interaction within the elementary quantum object. It recovers coherence. Let’s assume also that at the microscopic level the direct phase action onto quantum object is much feebler than dynamical action.
At the direct phase action the exterior objects act onto phase degrees of freedom of quantum object directly. However indirect phase action is possible. It is carried out as follows. The exterior objects dynamically act onto dynamical degrees of freedom (onto nucleus) of the quantum object. Further this action is transmitted to phase degrees of freedom through strong interior phase interaction. It brings to reorganization of the shell of the quantum object, i.e. to modification of its physical state. Such phase action is not weak as against dynamical.
Neglecting the direct phase action, we shall consider that the evolution of a physical state of quantum object is determined by dynamical action, and it is controlled by a dynamical equation of motion, which is coordinated with usual quantum-mechanical equation of motion. Namely we shall assume that the physical state $`\phi `$ evolves in the time so, that the functional, corresponding to this state, $`\phi (\widehat{A})(\widehat{A}\text{A})`$ varies as follows:
(4)
$$\phi _0(\widehat{A})\phi _t(\widehat{A})\phi _0(\widehat{A}(t)),$$
where
(5)
$$\frac{d\widehat{A}(t)}{dt}=\frac{i}{\mathrm{}}[\widehat{H},\widehat{A}(t)],\widehat{A}(0)=\widehat{A}.$$
Here $`\widehat{H}`$ is a usual Hamiltonian (considered as an element of the algebra A) of the quantum object.
The equations (4) and (5) quite unequivocally describe temporal evolution of the physical state. Therefore when only dynamical action is accounted (it corresponds to that that von Neumann names as action of the second type) physical processes are strictly determinable. Other matter, that with the help of observations we can determine the initial value $`\phi _0(\widehat{A})`$ of the functional (physical state) only to within its membership to some quantum state $`\{\phi \}_Q`$. Therefore majority of our predictions about the further observed dates for the considered quantum object can be only probabilistic.
Now we will turn to consideration of direct phase action. In the previous reasoning we considered that they can be neglected. A situation however is possible when it cannot be done. This situation is realized when the very large number of exterior objects act the quantum object. As the nucleus is local, it feels action of small number of the exterior objects if the long-range action is absent. As opposed to this, the shell having nonlocal structure feels action of the large number of the exterior objects. The mass character of the action can cancel weakness of the separate action. It happens only in that case when the separate weak actions do not cancel each other. It is possible to assume that exactly such situation is realized at action of a classical measuring device onto a quantum object. I.e. distinctive feature of a measuring device is that its separate microscopic elements exerts synchronous direct phase action onto the quantum object.
The typical classical measuring device comprises analyzer and detector. The analyzer is a classical device with one inlet and several exits. In the analyzer due to of direct phase action the united shell of quantum object decomposes onto several coherent constituents. Symbolically we shall figure it so:
$$\phi \phi _1\phi _2\mathrm{}\phi _i\mathrm{}.$$
I.e. the united structure (the physical state $`\phi `$) decomposes onto the direct (coherent) sum of constituents $`\phi _i`$. Inside the analyzer everyone evolves somehow , but all constituents preserve a mutual coherence. Therefore, if in the further the constituents will incorporate they will be able to interfere among themselves.
The constituent $`\phi _i`$ quits the analyzer through $`i`$-th exit. Everyone ($`i`$-th) exit corresponds to a particular value $`A_i`$ of certain observable $`\widehat{A}`$ (or of several mutually commuting observables). Thus, the analyzer is a point of branching of the initial physical state $`\phi `$. The part of the shell, falling into the $`i`$-th branch, belongs to a set $`[\phi ]^{A_i}`$, which contains all information fields $`\phi ^{}`$ such, that the corresponding functionals satisfy to equality $`\phi ^{}(\widehat{A})=A_i`$.
If only the dynamical action onto the quantum object is taken into account point of branching of the shell is a point of bifurcation for motion of the nucleus. In this point the interaction of the nucleus with the information field plays a role of ”random” force, which guides the nucleus along one of the branch. Let’s consider that the nucleus is retracted into the branch, for which
(6)
$$\phi [\phi ]^{A_i},$$
where $`\phi `$ is information field of the quantum object $`before`$ points of branching.
Such motion through the analyzer is admissible for the shell not changing structure. Let’s consider that the nucleus should be in a resonance with neighbouring part of the shell. Then such motion is admissible for nucleus for which the resonant condition does not vary. Actually the structure of the shell varies at the analyzer. Therefore equation (6) is necessary to consider, as the requirement of an invariance of the resonant condition for the nucleus at the point of bifurcation. The equation (6) is certain condition of continuity for the motion of the nucleus. This equation guarantees that at the point of bifurcation the evolution of the quantum object is uniquely determinated by its physical state $`\phi `$.
The observable evolution has probabilistic character. Firstly, it is not controlled by a dynamical equation of motion (the bifurcation point). Secondly, the physical state $`\phi `$ before the bifurcation point can be fixed only to within membership $`\phi `$ to certain quantum state $`\{\phi \}_Q`$. Due to the equation (6) the probability $`W_i`$ of falling into the $`i`$-th branch for the nucleus is determined by the equality
(7)
$$W_i=_{\phi \{\phi \}_Q[\phi ]^{A_i}}𝑑\mu (\phi ).$$
Now we shall discuss the detector. It is a classical object, which has strong dynamical and phase interaction with quantum object. The detector is in a macroscopically unstable state. As a result of dynamical action of nucleus of the quantum object it goes out equilibrium. A catastrophic process, which makes macroscopically observable result, develops in it.
The phase action of quantum object onto the detector is proportioned onto large number of microscopical constituents of the detector and does not give macroscopically observable effect. Thus, the detector macroscopically reacts only to nucleus of the quantum object, i.e. it reacts to the quantum object as onto one aggregate.
If the detector is located at the $`i`$-th branch then it works with probability $`W_i`$ (formula (7)). The nucleus of quantum object falls into the $`i`$-th branch with such probability. If the detector has worked, it denotes $`\phi [\phi ]^{A_i}`$, i.e. $`\phi (\widehat{A})=A_i`$. Thus, on the one hand, value of the functional $`\phi ()`$ really characterizes result of individual measurement. On the other hand, using formula (7) for average value $`<A>`$ of the observable $`\widehat{A}`$ we can receive
$$<A>=\underset{i}{}W_iA_i=\underset{i}{}_{\phi \{\phi \}_Q[\phi ]^{A_i}}𝑑\mu (\phi )\phi (\widehat{A})=_{\phi \{\phi \}_Q}𝑑\mu (\phi )\phi (\widehat{A})=\mathrm{\Psi }_Q(\widehat{A}).$$
It agrees with deductions of the standard quantum mechanics and with property (Causality and probabilistic interpretation of quantum mechanicsthanks: Submitted to ”Theoretical and Mathematical Physics”) of the functional $`\mathrm{\Psi }_Q()`$.
The inverse action of the detector onto the quantum object can go along two scenarios. The first scenario is realized when the nucleus is at the branch where there is the detector. The detector dynamically acts onto the nucleus and strongly (directly and indirectly) onto that part of the shell of the quantum object, which has fallen into the $`i`$-th branch. Due to this action these parts of the quantum object lose coherence with those parts of the shell, which have fallen into other branches. As a result, firstly, they lose possibility to interfere with the part of the shell, which has passed through the $`i`$-th branch. Secondly, only this part of the shell remains in the structure of the quantum object, as only it does not lose coherence with the nucleus due to strong interior phase interaction.
Thus, there is sharp reorganization of the shell of the quantum object (of its physical state). In standard quantum mechanics this phenomenon is treated as a collapse of the quantum state. Here this phenomenon can be named objective or active collapse. By the exterior displays the objective collapse is quite similar to the earlier described passive collapse, but the physical essences of these phenomena are completely different.
It is necessary to note, that the modification of the physical state and, as the consequence, its quantum state happens due to actual modification of the part of the quantum object, which is at the detector. But not as a result of vanishing (of reduction) of those parts, which are not at the detector. Nothing happens with them. Nevertheless, they cease to be constituents of the quantum object. In this case the action of the detector onto the quantum object is dynamical and phase. Either first type, or the second type of interaction can predominate. If the overwhelming contribution gives the dynamical action then this contribution can be described by the dynamical equations (4), (5).
If overwhelming or the essential contribution gives direct phase action then this contribution can not be described by the dynamical equations. By terminology of von Neumann it is interaction of the first type. However in this case (as opposed to the von Neumann’s opinion ) the physical evolution of the quantum object is uniquely determined by structure of that part of the shell, which has hit the detector. Other matter, that we have only the information, which there is in the equation $`\phi _i\{\phi \}_Q\{\phi \}^{A_i}`$ for this part $`\phi _i`$ of the shell. Therefore we can do only probabilistic predictions. Most probably, if the direct phase action plays main role then the modified part $`\phi _i^{}`$ of the shell will belong to the set $`\{\phi \}^{A_i}`$, but it will cease to belong to the set $`\{\phi \}_Q`$. In favour of such supposition speaks experiment, as a particular quantum state is practically prepared usually so.
Let’s consider now second scenario, when the nucleus falls into that branch, in which the detector is absent. For simplicity of reasoning we shall consider that the analyzer has only two branches. The detector is located in the second branch, and the nucleus falls into the first branch. In this case the detector does not work (negative experiment). However the standard quantum mechanics states that there is a collapse of a quantum state also. Let’s look, how it can be justified within the framework of considered here model.
In this case at the analyzer the field $`\phi `$ decomposes onto coherent constituents $`\phi _1`$ and $`\phi _2`$: $`\phi \phi _1\phi _2`$. The field $`\phi _2`$ falls into operative zone of the detector. There this part of the shell undergoes strong direct action of the detector. As a result of this action $`\phi _2`$ loses a coherence with the nucleus and $`\phi _1`$. Therefore $`\phi _2`$ ceases to be a part of the shell of the quantum object. Now the quantum object will have physical state $`\phi _1`$. There is a modification of the physical state of the quantum object. This modification is not controlled by the dynamical equations (objective collapse of a quantum state). In this case we quite definitely can state, that $`\phi _1[\phi ]^{A_1}`$.
Let’s note that both at the first and second scenario, on the one hand, as a result of action of the detector there is a (objective) collapse of the quantum state of the quantum object, on the other hand, a long-range action of the detector is absent.
The proposed scheme of quantum mechanics gives obvious and almost classical explanation of the most fundamental quantum phenomena. The scheme is free from any paradoxes. However there is a problem, maybe this scheme one of variants of scheme with hidden parameters. In a certain measure it is so, but the reasonings, over which schemes with the hidden parameters are rejected, is not correct in this case. The famous proof of von Neumann about impossibility of the hidden parameters in quantum mechanics essentially founds on linearity of quantum mechanics. One of main elements, functional $`\phi ()`$, is not linear in the scheme, proposed here. Therefore proof of von Neumann does not concern the present case.
Other, not less famous, argument against schemes with the hidden parameters is Bell inequality . We shall reproduce a typical deduction of this inequality. Let a quantum object $`Q`$ (particle with spin 0 in the elementary variant of experiment) decays into two objects $`A`$ and $`B`$ (particles with spins 1/2). The objects $`A`$ and $`B`$ scatter on large distance and hit detectors $`D(A)`$ and $`D(B)`$, respectively, in which the measurements are independent. The object $`A`$ has a set of observables $`\widehat{A}_a`$ (double projection of spin onto the direction $`a`$). The observables corresponding to different values of an index $`a`$, are not simultaneously measurable. Each of observables can take two values $`\pm 1`$. In a concrete experiment the device $`D(A)`$ measures an observable $`\widehat{A}_a`$ with a particular index $`a`$. For the object $`B`$ everything is similar.
Let’s assume that a quantum object $`Q`$ has a hidden parameter $`\lambda `$. In each individual event the parameter $`\lambda `$ has a particular value. The distribution of events according to the parameter $`\lambda `$ is characterized by a measure $`\mu (\lambda )`$ with usual properties $`\mu (\lambda )0`$, $`𝑑\mu (\lambda )=1`$.
All magnitudes, connected with individual event, depend on the parameter $`\lambda `$. In particular, the values of observables $`\widehat{A}_a`$ and $`\widehat{B}_b`$, obtained in a concrete experiment, are functions $`A_a`$, $`B_b`$ of the parameter $`\lambda `$. For individual event the correlation of observables $`\widehat{A}_a`$ and $`\widehat{B}_b`$ is characterized by the magnitude $`A_a(\lambda )B_b(\lambda )`$. The average value of the magnitude is referred to as correlation function $`E(a,b)`$:
$$E(a,b)=𝑑\mu (\lambda )A_a(\lambda )B_b(\lambda ).$$
Giving various values to the indexes $`a`$ and $`b`$ and taking into account that
(8)
$$A_a(\lambda )=\pm 1,B_b(\lambda )=\pm 1,$$
we shall obtain the following inequality
(9)
$$|E(a,b)E(a,b^{})|+|E(a^{},b)+E(a^{},b^{})|$$
$$𝑑\mu (\lambda )[|A_a(\lambda )||B_b(\lambda )B_b^{}(\lambda )|+|A_a^{}(\lambda )||B_b(\lambda )+B_b^{}(\lambda )|]=$$
$$=𝑑\mu (\lambda )[|B_b(\lambda )B_b^{}(\lambda )|+|B_b(\lambda )+B_b^{}(\lambda )|].$$
In the right-hand side of the formula (9), due to equalities (8), one of the expressions
(10)
$$|B_b(\lambda )B_b^{}(\lambda )|,|B_b(\lambda )+B_b^{}(\lambda )|$$
is equal to zero, and another is equal to two for each value of $`\lambda `$. From here we obtain the Bell inequality
(11)
$$|E(a,b)E(a,b^{})|+|E(a^{},b)+E(a^{},b^{})|2.$$
The correlation function $`E(a,b)`$ is easily calculated within the framework of the standard quantum mechanics. In particular, when $`A`$ and $`B`$ are particles with spin 1/2
(12)
$$E(a,b)=\mathrm{cos}\theta _{ab},\theta _{ab}\text{ is an angle between }a\text{ and }b.$$
It is easy to verify that there are directions $`a,b,a^{},b^{}`$, for which formulas (11) and (12) contradict each other.
It would seem it is possible to repeat this derivation in the proposed here model, having made replacements of type $`A(\lambda )\phi (\widehat{A})`$, $`B(\lambda )\phi (\widehat{B})`$, $`𝑑\mu (\lambda )\mathrm{}_{\phi \{\phi \}_Q^{\widehat{A}\widehat{B}}}𝑑\mu (\phi )\phi (\mathrm{})`$. However this opinion is erroneous. For derivation of the Bell inequality it is essential, that in the left-hand side of the inequality (9) it is possible to represent all terms in the form of united integral over one parameter $`\lambda `$. It is not valid for the quantum average substituting this integral, as it is necessary to integrate over actual states in it . The elements, appearing in different correlation functions, $`\widehat{A}_a\widehat{B}_b`$, $`\widehat{A}_a^{}\widehat{B}_b`$, $`\widehat{A}_a\widehat{B}_b^{}`$, $`\widehat{A}_a^{}\widehat{B}_b^{}`$, do not commute among themselves. Therefore sets of actual states, corresponding to these operators, do not intersect. In derivation of the inequality (11) we tacitly supposed that expression (10) exist for each $`\lambda `$. However there is no physical state $`\phi `$, which would be by actual state both for the observable $`\widehat{B}_b`$ and for the observable $`\widehat{B}_b^{}`$.
In summary it is necessary especially to note, that the present paper is not at all attempt to formulate a new rival theory to quantum mechanics. The proposed scheme does not contradict any statement of the standard quantum mechanics. Maybe some theses gain slightly other physical interpretation. All deductions of the standard quantum mechanics are valid in the described scheme. At the same time there are additional elements (for example, functional $`\phi ()`$, equality (Causality and probabilistic interpretation of quantum mechanicsthanks: Submitted to ”Theoretical and Mathematical Physics”)) in this scheme. They allow to include individual events in the domain of its application.
Strictly speaking, the formalism of the standard quantum mechanics assumes that the classical relations are reproduced only for average values of quantum observables. Meantime the practice shows that such relations are reproduced in each individual experiment. Of course, it concerns only those observables, what can be measured in one experiment. In the proposed approach this fact is consequence of the equalities (Causality and probabilistic interpretation of quantum mechanicsthanks: Submitted to ”Theoretical and Mathematical Physics”). |
warning/0001/hep-ph0001330.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Single-spin asymmetry study has recently received much attention, both experimentally$`^{\text{References}\text{References}}`$ and theoretically$`^{\text{References}\text{References}}`$. In fact, already in 1978, it was recognized<sup>?</sup> that such experiments, in which one of the colliding hadrons is transversely polarized, can be very useful in studying the properties of strong interaction in general and in testing Quantum Chromodynamics (QCD) in particular. These studies have been of particular interests in the Spin Community in recent years for the following reasons:
* The experiments are conceptually very simple.
* The observed effects are very striking.
* Theoretical expectations based on the pQCD parton model deviate drastically from the data.
* Information on transverse spin distribution and that on its flavor dependence can be obtained from such experiments.
A large amount of data is now available$`^{\text{References}\text{References}}`$. Besides the well known analyzing power in $`pp`$-elastic scattering<sup>?</sup>, we have now data$`^{\text{References}\text{References}}`$ on left-right asymmetries $`A_N`$ in single-spin inclusive processes for the production of various particles, such as different mesons, $`\mathrm{\Lambda }`$ hyperons and direct photons, measured in experiments using various transversely polarized projectiles such as protons and antiprotons. Compared to elastic scattering, there are not only more data at higher energies but also more theoretical discussions. We will therefore focus on the inclusive processes in the following.
In single-spin single particle inclusive production experiments, one uses transversely polarized (or unpolarized) projectile to collide with unpolarized (or transversely polarized) target, and measure the inclusive cross section (or production rate) for a given type of particle (or particle system) which enters the detector located, e.g., on the left-hand-side looking down stream. The asymmetry $`A_N`$ is defined as$`^{\text{References}\text{References}}`$,
$$A_N(x_F,p_{}|h,s)\frac{N_L(x_F,p_{},h|s,)N_L(x_F,p_{},h|s,)}{N_L(x_F,p_{},h|s,)+N_L(x_F,p_{},h|s,)}.$$
(1)
Here, $`h`$ denotes the produced particle or particle system; $`N_L(x_F,p_{},h|s,)`$ is the number-density of $`h`$’s which moves perpendicular to the polarization direction of incoming hadron and enters the detector when the colliding hadron is upwards polarized; $`N_L(x_F,p_{},h|s,)`$ is the corresponding density for such $`h`$’s in the downwards polarized case; the subscript $`L`$ denotes that the detector is located on the left-hand-side looking down stream. $`\sqrt{s}`$ is the total center of mass (c.m.) energy of the colliding hadron system; $`x_F2p_{}/\sqrt{s}`$, $`p_{}`$ is the longitudinal component of the momentum of $`h`$ in the c.m. frame and $`p_{}`$ is the magnitude of the transverse component. Since $`N_L(x_F,p_{};h|s,)=N_R(x_F,p_{};h|s,)`$, the above mentioned definition of $`A_N`$ can also be written as,
$$A_N(x_F,p_{}|h,s)\frac{N_L(x_F,p_{},h|s,)N_R(x_F,p_{},h|s,)}{N_L(x_F,p_{},h|s,)+N_R(x_F,p_{},h|s,)}.$$
(2)
That is why the asymmetry is usually referred to as left-right asymmetry. Obviously, the statistics would be very poor if one measured only those particles (or particle systems) which move perpendicularly to the polarization direction (strictly left or right). In practice, one makes use of particles produced in different transverse directions since the following is valid,
$$A_N(x_F,p_{}|h,s)=\frac{1}{\mathrm{cos}\phi }\frac{N(x_F,p_{},\phi ;h|s,)N(x_F,p_{},\pi \phi ;h|s,)}{N(x_F,p_{},\phi ;h|s,)+N(x_F,p_{},\pi \phi ;h|s,)}.$$
(3)
where $`\phi `$ is the azimuthal angle between the normal of the production plane and the polarization direction of the colliding hadron. Eq.(3) is a direct consequence of space quantization for collision processes with spin-1/2 particles.
It should be mentioned that $`A_N`$ is the only parity-conserving asymmetry which can exist in single-spin single particle inclusive processes. This can be seen most clearly in the rest frame of the polarized colliding hadron. Here, the S-matrix can only be a function of the following three vectors: the polarization vector of the hadron $`\stackrel{}{S}`$, the incident momentum of the other colliding hadron $`\stackrel{}{p}_{inc}`$ and the momentum $`\stackrel{}{p}`$ of the observed particle or particle system. Due to parity conservation, the S-matrix must be a scalar and the only spin-dependent scalar we can construct using these three vectors is $`\stackrel{}{S}(\stackrel{}{p}_{inc}\times \stackrel{}{p}_h)`$. We see that it is nonzero only if the transverse component of $`\stackrel{}{S}`$ is different from zero.
It has been observed that $`A_N`$ is up to $`40\%`$ in the beam fragmentation region, whereas the theoretical expectation<sup>?</sup> based on perturbative QCD parton model calculations were $`A_N0`$. Different new theoretical approaches have been made in the last few years to understand such large discrepancies. New experiments are now underway and/or planned. It is therefore timely to summarize the presently available experimental results, the main ideas and results of different models in order to get a deep insight into the physics behind these phenomena and make full use of the wealth of the new experiments. Brief summaries in this direction have been made by Meng<sup>?</sup> in the “Workshop on the Prospects of Spin Physics at HERA” held in August 1995 in Zeuthen, and by one of us<sup>?,?</sup> in the “Adriatico Research Conference on Trends in Collider Spin Physics” held in December 1995 in Trieste, and in the “XIII International Seminar on High Energy Physics Problems: Relativistic Nuclear Physics and Quantum Chromodynamics” held in Dubna September 1996. The characteristics of the available data and main ideas and results of different models have been briefly summarized in these talks. This is an extended version of those short summaries<sup>?,?,?</sup>. Part of it is also based on our doctoral theses<sup>?,?</sup> at FU Berlin. The text in the following is arranged as follows: After this introduction, we will briefly summarize the characteristics of the existing data, the different approaches based on the pQCD hard scattering model, the main ideas and results for $`A_N`$ of a non-perturbative approach of the Berliner group. They are given in section 2, 3 and 4 respectively. In section 5, we discuss the possibilities to differentiate these different models by performing suitable further experiments. In section 6, we discuss the connection of $`A_N`$ to hyperon polarization in unpolarized reactions. An example for the possible applications of such striking experimental results is given in section 7.
## 2 Characteristics of the data
Experiments on single-spin inclusive hadron production was first carried out<sup>?</sup> in 1976 using the polarized proton beam at the Argonne Zero-Gradient Synchrotron (ZGS) to collide with unpolarized proton for $`\pi ^\pm `$ production, and later for $`K`$ and $`\mathrm{\Lambda }`$ productions<sup>?</sup>. Striking asymmetries have been observed, although the incident momentum of the beam was quite low (6 and 11.75 GeV/c). Subsequently, similar experiments were carried at CERN<sup>?</sup> and at Brookhaven Alternating Gradient Synchrotron (AGS) at a bit higher energies and for the production of different hadrons<sup>?,?,?</sup>. At Serpukhov, experiments was carried out<sup>?</sup> using 40 GeV/c pion-beam to collide with transversely polarized proton or deuteron targets. More recently, a high energy (200GeV/c) transversely polarized proton and antiproton beam was obtained at Fermilab from the parity-violating decay of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$. Using these beams, high energy single-spin experiments have been carried out$`^{\text{References}\text{References}}`$ for the production of different kinds of mesons and $`\mathrm{\Lambda }`$. Very striking asymmetries have been observed$`^{\text{References}\text{References}}`$. Data are now available at $`x_F0`$ but rather high transverse momentum $`p_{}`$ ($`14`$ GeV/c) for $`p()+p(0)\pi ^0+X`$ and at large $`x_F`$ but moderately large $`p_{}`$ ($`0.22`$ GeV/c) for production of different kinds of hadrons and using proton and antiproton beams. Here, as well as in the following of this review, we use the following notations: The first particle in a reaction denotes the projectile, the second denotes the target; $`()`$ means that the particle is transversely polarized, $`(0)`$ means that it is unpolarized.
Not only the $`x_F`$\- but also the $`u`$-dependence have been studied in the lower energy experiments<sup>?,?</sup>. \[Here, as well as in the following, $`s,t,u`$ are the usual Mandelstam variables.\] Very striking features have been observed for the asymmetries at this energy. It has, in particular, been observed that the $`x_F`$-dependences for the asymmetries at different regions of $`u`$ are quite different from each other. E.g., it has been observed that $`A_N`$ for $`\pi ^{}`$ is positive but small at $`x_F0.5`$ for both $`u=0.2`$(GeV/c)<sup>2</sup> and $`u=0.2`$(GeV/c)<sup>2</sup> but it increases monotonically with increasing $`x_F`$ for $`u=0.2`$(GeV/c)<sup>2</sup> and reaches more than 30% at $`x_F0.8`$ whereas decreases monotonically with increasing $`x_F`$ for $`u=0.2`$(GeV/c)<sup>2</sup> and is even negative for $`x_F>0.6`$ and reaches -0.16 at $`x_F0.8`$. This is very interesting especially if we look at, in stead of $`u`$, the transverse momenta of the produced pion in the two cases, they are not very much different from each other: those in case of $`u=0.2`$(GeV/c)<sup>2</sup> are around $`00.3`$GeV/c and those in the case of $`u=0.2`$(GeV/c)<sup>2</sup> are around 0.5GeV/c. The existence of $`A_N`$ at these energies together with such striking features is still a puzzle for the theoretians. In fact, little theoretical discussion has yet been made in this connection. Whether the asymmetry in this energy region and that at the higher energies are of the same origin, whether the striking features observed here still exist in the higher energy experiments are questions which are still open. We recall that for small transverse momenta, that is small positive $`u`$, diffractive dissociation may play an important role in particular at low energies. It is therefore also interesting and instructive to study these processes to see whether diffractive scattering play an important role and whether/how one can obtain useful information in connection with the understanding of the mechanism for diffractive scattering. This is an interesting topic which deserves much effort in the future. In the following, we will concentrate on the higher energy region because almost all the theoretical discussions now available are in this region.
To show the main features of these data at the high energy, we show in Figs. 1a, 1b, 1c and 1d the $`A_N`$’s for $`p()+p(0)\pi +X`$, $`\overline{p}()+p(0)\pi +X`$, $`p()+p(0)\mathrm{\Lambda }+X`$ as functions of $`x_F`$ at moderate $`p_{}`$, and in Fig. 1e $`A_N`$’s for $`p()+p(0)\pi ^0+X`$ and $`\overline{p}()+p(0)\pi ^0+X`$ as functions of $`p_{}`$ at $`x_F0`$. These data show the following characteristics:
* $`A_N`$ is significant in, and only in, the fragmentation region of the polarized colliding object and for moderate transverse momenta:
It can be seen that $`A_N`$ is approximately equal to zero for $`x_F0`$ and different $`p_{}`$. For moderate $`p_{}`$, its magnitude increases monotonically with $`x_F`$ and reaches, e.g., 40% for $`p()+p(0)\pi ^++X`$ at $`x_F0.8`$.
* $`A_N`$ depends on the flavor quantum number of the produced hadrons:
It can be seen that $`A_N`$ in $`p()+p(0)\pi +X`$ is positive for $`\pi ^+`$ and $`\pi ^0`$ but negative for $`\pi ^{}`$, and that the magnitude of $`A_N`$ for $`\pi ^+`$ and that for $`\pi ^{}`$ are approximately equal but larger than that for $`\pi ^0`$.
* $`A_N`$ depends also on the flavor quantum number of the polarized projectile:
By using $`\overline{p}()`$ projectile instead of $`p()`$, one observed that while the asymmetry for $`\pi ^0`$ remains almost the same, those for $`\pi ^+`$ and $`\pi ^{}`$ exchange their roles.
* $`A_N0`$ in the beam fragmentation region in $`\pi ^{}+p()\pi ^0`$ or $`\eta +X`$:
Measurement has also been made<sup>?</sup> in the fragmentation of the pseudoscalar meson $`\pi ^{}`$-beams (not shown in Figs.1a-e), the results show that in this region the asymmetry is consistent with zero.
New experiments are underway and/or planned. Single-spin asymmetries for hadron and lepton-pair production will be carried out at RHIC<sup>?</sup> at $`\sqrt{s}=200`$GeV and at Serpukhov by RAMPEX Collaboration<sup>?</sup> at $`p_{inc}=70`$ GeV/c. Plans for similar experiments at HERA<sup>?</sup> have also been discussed. These experiments will certainly provided new insights into the origin of the observed large single-spin asymmetries.
## 3 PQCD based hard scattering models
Already in 1978, is was realized<sup>?</sup> that single-spin asymmetries can be very useful in studying the properties of hadron-hadron interactions in general and in testing the validity of pQCD calculations in particular. Using a straightforward generalization of the pQCD based hard scattering model to polarized case, Kane, Pumplin and Repko obtained<sup>?</sup> that, to the leading order in pQCD, $`A_N0`$ at high energies. This result is in clear disagreement with the data$`^{\text{References}\text{References}}`$ mentioned in the last section. Since pQCD calculations was very successful in describing unpolarized cross section even for transverse momentum of a few GeV (see, e.g. \[References\]), it was therefore a great surprise to see how large the discrepancy between the theoretical prediction<sup>?</sup> and the corresponding data$`^{\text{References}\text{References}}`$ is, and it is often referred as a challenge to the theoretians to understand this discrepancy.
A number of mechanisms$`^{\text{References}\text{References}}`$ have been proposed in last few years which can give rise to non-zero $`A_N`$’s in the framework of QCD and quark or quark-parton models. As has been discussed in last section, the single-spin left-right asymmetries have been observed only for hadrons in the beam fragmentation regions and with moderate transverse momenta. As is well know, hadrons with large transverse momenta are products of processes with large momentum transfer and such processes are called “hard processes”. For hard processes, impulse approximation can be used hence the constituents of the colliding hadrons can be treated as “free particles” and perturbative QCD calculations for the scattering amplitudes are valid. It is also in such processes that the pQCD based hard scattering models were expected to, and indeed, work well<sup>?</sup>. In contrast, hadrons with small transverse momenta are dominated by the products of processes with small momentum transfer. Such processes are called “soft processes”. For soft processes, perturbative QCD cannot be used and collective effects of the constituents in the colliding hadrons and other non-perturbative effects are important. Phenomenological models have to be used in describing these effects. In studying the origin of the observed single-spin left-right asymmetries, we work in a kinematic region between the typical regions of the above mentioned two extreme cases. It is unclear whether perturbative methods are useful and whether typical “soft effects” play a role in this region. This characteristics of the problem makes the investigation even more interesting since it is a suitable place to study the interplay between the “soft” and “hard” interactions in hadronic collision processes. It determines also that the theoretical approaches are divided into the following two categories: One of them starts from the “hard” aspects and try to include some of the influences from the “soft” aspects in some unknown factors. The other starts from the “soft” side and neglects most of the influences from the “hard” aspects. The former kind of approaches is usually referred as perturbative QCD based hard scattering models and one of the well-known example of the latter is non-perturbative orbiting valence quark model. The former have been discussed by many authors in literature$`^{\text{References}\text{References}}`$, and the latter is mainly pursued by the Berliner group$`^{\text{References}\text{References}}`$ and is sometimes referred as “Berliner Model” or “Berliner Relativistic Quark Model (BRQM)” for single-spin asymmetries. We review the main ideas and results of these two approaches in the following. We start with pQCD based hard scattering models in this section and will deal with the second kind of models in the next section.
### 3.1 PQCD based hard scattering model for high-$`p_{}`$ hadron production in unpolarized hadron-hadron collisions
It is well known for a long time that in unpolarized hadron-hadron collisions, pQCD calculations can be applied for the production of high $`p_{}`$ jets and/or high $`p_{}`$ particles. It has been shown that the inclusive cross section for hadron production can be factorized and thus be expressed as a convolution of the following three factors: the momentum distribution functions of the constituents (quarks, antiquarks or gluons) in the colliding hadrons; the cross section for the elementary hard scattering between the constituents of the two colliding hadrons; and the fragmentation function of the scattered constituent. E.g., for the inclusive process $`A(0)+B(0)C+X`$, this is illustrated in Fig. 2, and one has<sup>?</sup>,
$$E\frac{d\sigma }{d^3p}[A(0)B(0)C+X]=\underset{abcd}{}dx_adx_bdz_C\varphi _{a/A}(x_a)\varphi _{b/B}(x_b)$$
$$\frac{\widehat{s}}{z_C^2\pi }\frac{d\sigma }{d\widehat{t}}(abcd)\delta (\widehat{s}+\widehat{t}+\widehat{u})D_F^{C/c}(z_C),$$
(4)
Here, $`p`$ is the four momentum of $`C`$, whose longitudinal and transverse components are usually denoted by $`p_{}=x_F\sqrt{s}/2`$ and $`\stackrel{}{p}_{}`$ respectively; $`\varphi _{a/A}(x_a)`$ and $`\varphi _{b/B}(x_b)`$ are the distribution function of the constituent $`a`$ in hadron $`A`$ and that of $`b`$ in $`B`$, where $`x`$ denotes the momentum fraction of hadron carried by the constituent; $`\frac{d\sigma }{d\widehat{t}}(abcd)`$ is the cross section for the elementary scattering process $`abcd`$, where $`\widehat{s},\widehat{t}`$ and $`\widehat{u}`$ are the usual Mandelstam variables for the process; $`D_F^{C/c}(z_C)`$ is the fragmentation function describing the hadronization of $`c`$ into the hadron $`C`$ and anything else, where $`z_C`$ is the momentum fraction of the parton $`c`$ obtained by the hadron $`C`$. While the distribution functions $`\varphi _{a/A}(x_a)`$ and $`\varphi _{b/B}(x_b)`$ can be obtained by parameterizing the data from deeply inelastic lepton hadron scattering and other experiments and the fragmentation function $`D_F^{C/c}(z_C)`$ from phenomenological models, $`\frac{d\sigma }{d\widehat{t}}(abcd)`$ is the only factor which can be calculated using perturbation theory. From these calculations, one obtained not only the production rates but also the transverse momentum (or transverse energy) distributions of the high $`p_{}`$ hadrons and/or jets.
We note in particular here that, in the above mentioned simple version of the pQCD based hard scattering model, no intrinsic transverse momentum $`\stackrel{}{k}_{}`$ of the constituents, $`a`$ and $`b`$, inside the hadrons, $`A`$ and $`B`$, is taken into account, and that the produced hadron in the fragmentation process is assumed to be in the same direction as the initial parton (i.e. $`p=z_Cp_c`$). We note also that<sup>?</sup> the model is very successful in describing the cross section for the production of hadrons or jets with large transverse momentum $`p_{}`$ (say, $`>5`$GeV/c). The model can also be and has been applied to hadron or jet production in the region where $`p_{}`$ is of the order of several GeV/c (say, 1 to 5 GeV/c), but in this region, effect of $`k_{}`$-smearing, which is an effect due to the intrinsic transverse momentum of the constituents inside the hadron, is very significant and should be taken into account. This means that, to describe hadron production in this region, one should take a transverse momentum distribution for the constituents $`a`$ and $`b`$ in $`A`$ and $`B`$ into account, and make the following replacement in Eq.(4),
$$𝑑x_a𝑑x_b\varphi _{a/A}(x_a)\varphi _{b/B}(x_b)𝑑x_a𝑑x_b𝑑\stackrel{}{k}_a𝑑\stackrel{}{k}_b\varphi _{a/A}(x_a,\stackrel{}{k}_a)\varphi _{b/B}(x_b,\stackrel{}{k}_b).$$
(5)
It is usually assumed<sup>?</sup> that the transverse distribution can be factorized from the longitudinal part, and a Gaussionian for $`f(\stackrel{}{k}_{})`$ was used to fit the data, i.e.,
$$\varphi (x,\stackrel{}{k}_{})=\varphi (x)f(\stackrel{}{k}_{}),$$
(6)
$$f(\stackrel{}{k}_a)=\frac{e^{k_{}^2/<k_{}^2>}}{\pi <k_{}^2>}.$$
(7)
It has been obtained<sup>?</sup> that $`<k_{}^2>0.95`$ (GeV/c)<sup>2</sup> which corresponds approximately to $`<k_{}>0.86`$ GeV/c.
### 3.2 Direct extension to polarized case
The above mentioned pQCD based hard scattering models has been extended$`^{\text{References}\text{References}}`$ in a straightforward manner to single-spin hadron-hadron collision processes. In this way, one obtains the following expression for the inclusive cross section for e.g. $`A()+B(0)C+X`$,
$$E\frac{d\sigma }{d^3p}[A()B(0)C+X]=\underset{abcd,s_a,s_c}{}dx_adx_bdz_c\varphi _{a/A()}(x_a,s_a)\varphi _{b/B}(x_b)$$
$$\frac{\widehat{s}}{z_C^2\pi }\frac{d\sigma }{d\widehat{t}}[a(s_a)bc(s_c)d]\delta (\widehat{s}+\widehat{t}+\widehat{u})D_F^{C/c}(z_C;s_c),$$
(8)
Here, $`s_a`$ and $`s_c`$ denote the spin of $`a`$ and that of $`c`$ respectively.
Now, if we assume that the fragmentation function is independent of the spin of the scattered constituent $`c`$, i.e.,
$$D_F^{C/c}(z_C;s_c)=D_F^{C/c}(z_C;s_c)D_F^{C/c}(z_C),$$
(9)
we obtain the following expression for $`E\mathrm{\Delta }\frac{d\sigma }{d^3p}[A(tr)B(0)C+X]`$ the difference between $`E\frac{d\sigma }{d^3p}[A()B(0)C+X]`$ and $`E\frac{d\sigma }{d^3p}[A()B(0)C+X]`$ as
$$E\mathrm{\Delta }\frac{d\sigma }{d^3p}[A(tr)B(0)C+X]=\underset{abcd}{}dx_adx_bdz_C\mathrm{\Delta }\varphi _{a/A(tr)}(x_a)\varphi _{b/B}(x_b)$$
$$\frac{\widehat{s}}{z_C^2\pi }\mathrm{\Delta }\frac{d\sigma }{d\widehat{t}}[a(tr)bcd]\delta (\widehat{s}+\widehat{t}+\widehat{u})D_F^{C/c}(z_C).$$
(10)
Here $`\mathrm{\Delta }\varphi _{a/A(tr)}(x_a)\varphi _{a/A()}(x_a,+)\varphi _{a/A()}(x_a,),`$ where $`+`$ means $`a`$ is polarized in the same direction as $`A()`$, $``$ means in the opposite direction; and,
$$\mathrm{\Delta }\frac{d\sigma }{d\widehat{t}}[a(tr)bcd]\frac{d\sigma }{d\widehat{t}}[a()bcd]\frac{d\sigma }{d\widehat{t}}[a()bcd]$$
(11)
which is usually written as,
$$\mathrm{\Delta }\frac{d\sigma }{d\widehat{t}}[a(tr)bcd]=a_N[a()bcd]\frac{d\sigma }{d\widehat{t}}(abcd).$$
(12)
Here, $`a_N`$ is the single-spin asymmetry for the elementary process $`abcd`$,
$$a_N[a()bcd]\frac{\mathrm{\Delta }\frac{d\sigma }{d\widehat{t}}[a(tr)bcd]}{\frac{d\sigma }{d\widehat{t}}(abcd)}$$
(13)
The cross section, or the scattering matrix $``$, for the elementary hard scattering $`abcd`$ can be calculated using pQCD. These calculations are most conveniently performed using the helicity basis. Take $`q_1q_2q_1q_2`$ as an example, and we obtain (See, e.g., \[References\])
$$\mathrm{\Delta }\frac{d\sigma }{d\widehat{t}}[q_1()q_2q_1q_2]2Im[_{++,+}^{}(_{++,++}+_{+,+})$$
$$_{++,+}^{}(_{++,}_{+,+})],$$
(14)
where the subscripts of $``$ denote the helicities of the particles in the initial and final states. We see clearly from Eq.(14) that $`a_N`$ is nonzero only if the helicity-flip amplitude $`_{++,+}`$ is nonzero and has a phase difference with the helicity conserving amplitude(s) $`_{++,++}`$ and/or $`_{+,+}`$, or the helicity flip amplitudes $`_{++,+}`$ and $`_{++,}`$ are nonzero and have a phase difference with each other or $`_{++,+}`$ and $`_{+,+}`$ are nonzero and have phase difference with each other. In either case, one needs at least one non-zero helicity flip amplitude to get a nonzero $`\mathrm{\Delta }`$ thus a nonzero $`a_N`$. But, it is a well-known and remarkable property of perturbative QCD as well as perturbative QED that, in the limit of zero quark mass, helicity is conserved. That is, in this limit, all the helicity flip amplitudes vanish hence $`a_N[q_1()q_2q_1q_2]=0`$.
Taking the quark mass $`m_q`$ into account, one obtains that, to the first order of pQCD, the helicity flip amplitude is nonzero but proportional to $`m_q`$. Using this, one gets<sup>?</sup> $`a_N`$ for the above mentioned QCD elementary process $`q_1()+q_2(0)q_1+q_2`$ is proportional to $`m_q/\sqrt{s}`$, which is negligibly small at high energies. Hence, we obtain that $`A_N0`$ at high energy $`\sqrt{s}`$.
The situation will not be changed much if we take an intrinsic transverse momentum distribution of the quarks in proton into account and assume that it is symmetric in the azimuthal direction. Under such circumstances, we obtain $`A_N0`$ for hadron production.
### 3.3 Comparison with data: What conclusion can we draw now?
We now compare the above mentioned result of the pQCD based hard scattering model with the available data$`^{\text{References}\text{References}}`$. We are particularly interested in the following questions: Do the available data contradict QCD? Do the available data confirm QCD?
The answer to the first question is: No! This is because pQCD works well only in the case in which the momentum transfer is large. Hence we expect that the above mentioned result is true only for the production of hadrons with high $`p_{}`$ (say, $`p_{}5`$GeV/c). The only piece of data which is now available for high $`p_{}`$ is that from E704 Collaboration for the production of $`\pi ^0`$ and at $`x_F0`$. This data show that $`A_N`$ is indeed consistent with zero. (C.f. Fig.1e).
The answer to the second question is unfortunately also: No! This is because the only piece of data with which the prediction of pQCD can be compared is that for the production of $`\pi ^0`$ and at $`x_F0`$. There may be different reasons for the vanishing of $`A_N`$ in this case. E.g., it can be simply because the light sea quarks (antiquarks) which are responsible for the production of $`\pi ^0`$ in this region are not polarized. It can be also because the polarizations of the $`u`$, $`\overline{u}`$, $`d`$, and $`\overline{d}`$ sea quarks (antiquarks) compensate with each other and therefore lead to zero $`A_N`$ for $`\pi ^0`$ at $`x_F0`$. It does not necessary mean that the asymmetry from the elementary hard scattering process is zero, which is the prediction of QCD based on the perturbation theory. To test this prediction, one needs to measure $`A_N`$ for large $`p_{}`$, large $`x_F`$ and at high energy. Large $`p_{}`$ and high energy are necessary to guarantee the validity of the parton picture and that of the pQCD calculation, and large $`x_F`$ to guarantee that the quarks which contribute to the production of such hadrons are predominately the valence quarks of the polarized projectile and are transversely polarized before the interactions take place. Such experiments can be carried out in the future. E.g., at RHIC<sup>?</sup>, the energy is already much higher than E704 experiment, hence, according to the above mentioned prediction of the pQCD parton model, the asymmetry $`A_N`$ at large $`x_F`$ high $`p_{}`$ should be substantially smaller than that observed by E704.
One thing is however clear. That is, in the moderate $`p_{}`$ and large $`x_F`$ regions, the above mentioned straightforward extended version of pQCD hard scattering model contradicts the data. This means that improvements of the model in this region are necessary.
### 3.4 New improvements
The answer to the question of how to improve the model in order to describe the striking $`A_N`$’s observed in the moderate $`p_{}`$ and large $`x_F`$ regions is in fact not difficult to find. We recall that, in the model, the cross section is a convolution of three different factors, and the result $`A_N0`$ was obtained under the following approximations and/or assumptions:
* the left-right asymmetry $`a_N`$ for the elementary process was calculated using pQCD to the leading order;
* the distribution functions of the constituents in nucleon was assumed to be symmetric in intrinsic transverse motion;
* the fragmentation function was assumed to be independent of the spin of the quarks.
It is therefore also clear, to get a large asymmetry $`A_N`$, one can make use of one or more of the following three possibilities:
* Look for higher order effects in the elementary processes which lead to larger asymmetries;
* Introduce asymmetric intrinsic transverse momentum distributions for the transversely polarized quarks in a transversely polarized nucleon;
* Introduce asymmetric transverse momentum distributions in the fragmentation functions for the transversely polarized quarks.
All these three possibilities have been discussed in the literature$`^{\text{References}\text{References}}`$.
It is clear that under the condition that pQCD is indeed applicable for the description of such processes, it should (at least in principle) be possible to find out how significantly the effects mentioned in (i) contribute to $`A_N`$ by performing the necessary calculations. In fact the calculations to the next leading order are not difficult to be carried out. The resulted asymmetry $`a_N`$ is still negligibly small at high energies.
In contrast to this, the asymmetric momentum distributions mentioned in (ii) and (iii) can in principle be large. But, these asymmetries have to be introduced by hand in the models. Whether such asymmetries indeed exist, and how large they are if they exist, are questions which can only be answered by performing suitable experiments. In the following, we review the discussions of these different possibilities in more details.
#### 3.4.1 Asymmetric quark transverse momentum distribution?
Possibility (ii) was first discussed by Sivers<sup>?</sup>. He argued that, to describe single transverse spin asymmetries, it is important to take the intrinsic transverse momentum $`k_{}`$ of the quarks in nucleon into account and to write the quark distribution functions as $`\varphi _{a/A}(x_a,\stackrel{}{k}_{})`$. He assumed that, in a transversely polarized hadron $`A()`$, $`\varphi _{a/A()}(x_a,\stackrel{}{k}_{},s_a)`$ can be asymmetric, i.e.
$$\varphi _{a/A()}(x_a,\stackrel{}{k}_{},\pm )\varphi _{a/A()}(x_a,\stackrel{}{k}_{},\pm ),$$
(15)
where $`\pm `$ means $`a`$ is polarized in the same ($`+`$) or the opposite ($``$) direction as the hadron $`A`$. In this case, even if $`a_N=0`$, one obtains,
$$E\mathrm{\Delta }\frac{d\sigma }{d^3p}[A()B(0)C+X]=\underset{abcd}{}dx_adx_bd\stackrel{}{k}_{}dz_C\mathrm{\Delta }^N\varphi _{a/A()}(x_a,\stackrel{}{k}_{})\varphi _{b/B}(x_b)$$
$$\frac{\widehat{s}}{z_C^2\pi }\frac{d\sigma }{d\widehat{t}}(abcd)\delta (\widehat{s}+\widehat{t}+\widehat{u})D_F^{C/c}(z_C),$$
(16)
which can lead to $`A_N`$ that is significantly different from zero if
$$\mathrm{\Delta }^N\varphi _{a/A()}(x_a,\stackrel{}{k}_{})\underset{s_a}{}[\varphi _{a/A()}(x_a,\stackrel{}{k}_{},s_a)\varphi _{a/A()}(x_a,\stackrel{}{k}_{},s_a)],$$
(17)
is significantly different from zero.
We recall that the effect of $`k_{}`$-smearing plays an important role in describing the cross section of hadron production in unpolarized hadron-hadron collisions in the region of transverse momentum of several GeV/c, and that left-right asymmetry in single spin reactions have been observed$`^{\text{References}\text{References}}`$ mainly in the region of moderately large transverse momentum. It is therefore quite natural and in fact even instructive to examine whether such effects due to the intrinsic transverse momenta of the quarks make significant contribution to these asymmetries. We note in particular that, the asymmetries have been observed$`^{\text{References}\text{References}}`$ for $`p_{}1`$GeV/c and that the average intrinsic transverse momentum of the quarks has been determined<sup>?</sup> as $`<k_{}>0.86`$ GeV/c. This implies that $`k_{}`$ has to contribute to the $`p_{}`$ significantly. Hence, it would be completely not surprising if it turns out that the $`k_{}`$-distribution plays an important role in describing these asymmetry data$`^{\text{References}\text{References}}`$. As mentioned in the last subsection, it can easily be shown that a symmetric intrinsic transverse momentum has little influence on the asymmetry. On the other hand, it is also clear that an asymmetric intrinsic transverse momentum distribution of the quarks in polarized hadron can manifest itself in the final state hadrons. It should lead to an asymmetric transverse momentum distribution for the produced hadrons if it indeed exists. It is also obvious that one can get a good fit to the data if one can choose the form and the magnitude of the asymmetric distribution freely. It is therefore a natural and good question to ask whether the such asymmetric distribution indeed exists and if yes whether it is the major source for the asymmetries observed in the single-spin experiments$`^{\text{References}\text{References}}`$.
However, it has been shown by Collins<sup>?</sup> that asymmetric intrinsic quark distribution cannot exist at leading twist, i.e. twist-2. At leading twist, the quark distribution is given by,
$$\varphi _{a/A}(x,\stackrel{}{k}_{})\frac{dy^{}d^2\stackrel{}{y}_{}}{(2\pi )^3}\mathrm{exp}(ixP^+y^{}+i\stackrel{}{k}_{}\stackrel{}{y}_{})P|\overline{\psi }_a(0,y^{},\stackrel{}{y}_{})\frac{\gamma ^+}{2}\psi _a(0)|P.$$
(18)
Here, $`P`$ is the four momentum of the hadron $`A`$; $`P^+\frac{1}{\sqrt{2}}(P^0+P^3)`$, $`\gamma ^+\frac{1}{\sqrt{2}}(\gamma ^0+\gamma ^3)`$, and $`y^{}\frac{1}{\sqrt{2}}(y^0y^3)`$ are light-cone variables. Since the transversely polarized state, e.g. $`|`$, can be expressed in terms of the left- and right-handed helicity states $`|L`$ and $`|R`$,
$$|=\frac{1}{\sqrt{2}}[|L+|R],$$
(19)
we have,
$$\varphi _{a/A()}(x,\stackrel{}{k}_{})=\frac{1}{2}[\varphi _{a/A}(x,\stackrel{}{k}_{};LL)+\varphi _{a/A}(x,\stackrel{}{k}_{};RR)+\varphi _{a/A}(x,\stackrel{}{k}_{};LR)+\varphi _{a/A}(x,\stackrel{}{k}_{};RL)].$$
(20)
Here,
$$\varphi _{a/A}(x,\stackrel{}{k}_{};h,h^{})\frac{dy^{}d^2\stackrel{}{y}_{}}{(2\pi )^3}\mathrm{exp}(ixP^+y^{}+i\stackrel{}{k}_{}\stackrel{}{y}_{})P,h|\overline{\psi }_a(0,y^{},\stackrel{}{y}_{})\frac{\gamma ^+}{2}\psi _a(0)|P,h^{},$$
(21)
where $`h`$ or $`h^{}`$ denotes the helicity ($`L`$ or $`R`$) of the hadron $`A`$. It is clear that $`\varphi _{a/A}(x,\stackrel{}{k}_{};LL)`$ and $`\varphi _{a/A}(x,\stackrel{}{k}_{};RR)`$ can only depend on the magnitude of $`\stackrel{}{k}_{}`$ but not on the direction since no particular transverse direction is specified in this case. We obtain in particular that,
$$\varphi _{a/A}(x,\stackrel{}{k}_{};hh)=\varphi _{a/A}(x,\stackrel{}{k}_{};hh),$$
(22)
for $`h=L`$ or $`R`$. Furthermore, using time reversal and parity inversion invariance, Collins<sup>?</sup> obtained that
$$\varphi _{a/A}(x,\stackrel{}{k}_{};LR)=\varphi _{a/A}(x,\stackrel{}{k}_{};RL)=0.$$
(23)
It follows from Eqs.(20), (22) and (23) that,
$$\varphi _{a/A()}(x,\stackrel{}{k}_{})=\varphi _{a/A()}(x,\stackrel{}{k}_{})$$
(24)
which contradicts Eq.(15). This shows explicitly that time reversal and parity inversion invariances of strong interaction forbids the existence of such asymmetry at twist two. This is similar to that discussed by Christ and Lee more than twenty years ago<sup>?</sup> for deeply inelastic lepton-nucleon scattering, where these invariances forbid the existence of single-spin asymmetry.
However, recently, Anselmino, Boglione, and Murgia<sup>?</sup> pointed out that the proof of Collins<sup>?</sup> mentioned above is valid at and only at leading twist, where the initial state interactions between the constituents of the two colliding hadrons are neglected. Asymmetric distributions are in principle allowed to exist at higher twists where the initial state interactions are taken into account. The suggestion of Anselmino et al. <sup>?</sup> is that these higher twist effects can be lumped into an effective asymmetric transverse momentum distribution in parton distribution functions. It is however unclear whether such asymmetric distribution indeed exists and, if yes, how large it is. These are questions which cannot be answered in the model. The asymmetry has to be introduced by hand and can only be studied beyond the model or by experiments. They therefore suggested<sup>?</sup> to measure these asymmetric distributions experimentally by measuring $`A_N`$ under the assumption that such asymmetric distributions are indeed responsible for the observed $`A_N`$. Obviously, such measurements make sense only in the case that one has tested such asymmetric distributions indeed exist and are responsible for the observed left-right asymmetries. The results would be useful if one has proved that they are universal in the sense that they can be used in different reactions. Since they are not intrinsic in the sense that they depend on the interaction of the hadron with the other, they cannot be measured in deeply inelastic lepton-nucleon scatterings using transversely polarized nucleon similar to that for unpolarized parton distributions. The “measurement” of the asymmetric quark distributions itself will provide unfortunately no deep insight into the problem. Furthermore, it is not clear whether factorization is valid at higher twists and whether expression (16) is a reasonable approximation to the full calculation. Therefore, one should analyze the scattering process by including higher twist contributions in the calculation right from the beginning.
#### 3.4.2 Higher twist parton distributions?
Higher twist effects in connection with single-spin asymmetries were discussed first by Efremov and Teryaev<sup>?,?</sup> and later by Qiu and Sterman<sup>?,?</sup>. It has been argued<sup>?</sup> that, the results of the calculations by Kane, Pumplin and Repko<sup>?</sup>, which show that $`A_N`$ is proportional to the quark mass $`m_q`$, indicate already that $`A_N`$ is a higher twist effect. Qiu and Sterman have therefore analyzed<sup>?</sup> the possible contributions of the next to leading twist terms, i.e. the twist-3 contributions. Since the asymmetries are observed mainly in the fragmentation region, where contributions of the valence quarks of the projectile dominate, Qiu and Sterman used a “valence-quark-soft-gluon” approximation where only interaction of valence-quarks from one colliding hadron with gluons from the other are taken into account. In this way they simplified the problem in a great deal. They argued that the leading contributions should come from the term containing the following twist-3 parton distribution,
$$T_{Fa}^{(V)}(x_1,x_2)=\frac{dy_1^{}dy_2^{}}{4\pi }e^{ix_1P^+y_1^{}+i(x_2x_1)P^+y_2^{}}P,\stackrel{}{s}_{}|\psi _a(0)\gamma ^+[\epsilon ^{\rho s_{}n\overline{n}}F_\sigma ^+(y_2^{})]\psi _a(y_1^{})|P,\stackrel{}{s}_{}$$
(25)
where $`\epsilon ^{\rho s_{}n\overline{n}}\epsilon ^{\rho \sigma \mu \nu }\stackrel{}{s}_\sigma n_\mu \overline{n}_\nu `$, $`n(n^+,n^{},\stackrel{}{n}_{})=(0,1,\stackrel{}{0}_{})`$, $`\overline{n}(\overline{n}^+,\overline{n}^{},\stackrel{}{\overline{n}}_{})=(1,0,\stackrel{}{0}_{})`$. They showed that the existence of such twist-3 parton distribution and the exchanged terms of the leading and next-to-leading graphs can lead to an significant left-right asymmetry for the produced particle.
Compared with the definition of the twist-2 parton distributions in Eq.(18), we see that while there are two field operators in Eq.(18) which can thus be interpreted as probability distribution, there are three filed operators in Eq.(25). One cannot interpret $`T_{Fa}^{(V)}(x_1,x_2)`$ as probability distribution. Since it depends on two fractional momentum $`x_1`$ and $`x_2`$, it is called correlation function.
It is usually expected that a twist-3 contribution to $`A_N`$ should be proportional to $`\mu /p_{}`$ and thus decreases with increasing $`p_{}`$ (where $`\mu `$ is a non-perturbative scale for the twist-3 matrix element). However, based on the above mentioned approximation, Qiu and Sterman showed that the $`\mu /p_{}`$ term is not the dominating term in the twist-3 contribution to the $`A_N`$ in the E704 kinematic region. From dimensional analysis alone, one would expect that there are two types of twist-3 contributions, one of which is proportional to $`\mu p_{}/(u)`$ and the other is proportional to $`p_{}\mu /(t)`$. We note that, for high energy $`\sqrt{s}`$, moderately large $`p_{}`$ and large $`x_F`$, $`tp_{}^2`$, $`us`$. Thus the two terms are proportional to $`\mu p_{}/s`$ and $`\mu /p_{}`$ respectively. We see that while the former vanishes at $`s\mathrm{}`$, the latter is suppressed at high $`p_{}`$. For not very large energy $`\sqrt{s}`$, the former, which is proportional to $`p_{}`$, can be larger than the latter in particular in the large $`p_{}`$ region. The results of Qiu and Sterman showed that at the FNAL E704 energy, the first type of contribution dominates at high $`x_F`$ moderate $`p_{}`$ and that $`A_N`$ changes rather smoothly with increasing $`p_{}`$. They found also that $`A_N`$ is mainly proportional to $`T_{Fa}^{(V)}(x_1,x_2=x_1)`$.
A simple model for $`T_{Fa}^{(V)}(x,x)`$ was proposed<sup>?</sup>. Since, for $`x_1=x_2=x`$, one has,
$$T_{Fa}^{(V)}(x,x)=\frac{dy_1^{}dy_2^{}}{4\pi }e^{ixP^+y_1^{}}P,\stackrel{}{s}_{}|\psi _a(0)\gamma ^+[\epsilon ^{\rho s_{}n\overline{n}}F_\sigma ^+(y_2^{})]\psi _a(y_1^{})|P,\stackrel{}{s}_{},$$
(26)
which is similar to the twist-2 parton distribution \[Eq.(18)\] with an extra factor $`[\epsilon ^{\rho s_{}n\overline{n}}F_\sigma ^+(y_2^{})]`$. Qiu and Sterman argued<sup>?</sup> this extra factor leads to no extra $`x`$ dependence, thus they assume,
$$T_{Fa}^{(V)}(x,x)=\kappa _a\lambda q_a(x),$$
(27)
where $`\kappa _a`$ denotes the sign of $`T_{Fa}^{(V)}`$ and $`\lambda `$ is a constant describing its magnitude. From the signs of data$`^{\text{References}\text{References}}`$ of $`A_N`$ for $`\pi ^+`$ and $`\pi ^{}`$, they determined that $`\kappa _u=1`$ and $`\kappa _d=1`$. The magnitude of $`A_N`$ can also be roughly reproduced by taking $`\lambda =0.080`$GeV. By using this simple model for $`T_{Fa}^V(x,x)`$, they demonstrated also how the twist-3 contribution to $`A_N`$ increases with increasing $`x_F`$.
This result is rather interesting and attractive for the following reasons: First, it directly extends the pQCD based hard scattering model to twist-3. If it turns out to work well, this opens another field to test the applicability of pQCD. Second, the results show that twist-3 effects at the energy region like that of the FNAL E704 experiments do not vanish at high $`p_{}`$. This provides us a good opportunity to study twist-3 distributions, which have poorly studied yet.
On the hand, this result is still somewhat unsatisfactory in particular from the phenomenological point of view, since it does not supply a physical picture for the single spin asymmetries. The correlation functions, which characterize the soft-hadron properties responsible for the asymmetries, are to be extracted from the experiments rather than to be given by the theory. The two parameters, $`\lambda `$ and $`\kappa _a`$, which determine respectively the sign and the magnitude of $`T_{Fa}^{(V)}`$, are free parameters in the model which are determined by the $`A_N`$ data. It is therefore unclear whether they are indeed the major source for the observed $`A_N`$. Furthermore, there seems to be an inconsistency in the restriction to twist three, at least as far as the understanding of the present data is concerned. We can easily see that, if we can obtain from the twist-3 contribution an $`A_N`$ as large as $`30\%`$ to $`40\%`$, the twist-3 contribution to the cross section should at least half or $`60\%`$ as large as the leading twist-2 contribution. This is then not a small correction any more. How about the twist-4 and even higher twist contributions? According to the above mentioned dimensional analysis, do they vanish or even increase as $`p_{}^2`$ with increasing $`p_{}`$? While we expect pQCD calculations with higher twist effects included to work for larger $`p_{}`$ values, it seems that a modeling of the soft-hadronic properties is necessary to understand the physical origin of (the present) data.
#### 3.4.3 Spin dependent fragmentation function?
Since fragmentation of quark is a basically soft process containing contribution from higher twists, it is conceivable that fragmentation effects could account for at least part of the single-spin asymmetries.
After having shown that the quark intrinsic transverse momentum cannot be asymmetric, Collins, Heppelmann and Ladinsky have argued that<sup>?,?</sup> the origin for the observed left-right asymmetry can only be the fragmentation, i.e. possibility (iii) mentioned above should be true. They suggested that transverse momentum for hadron obtained in the fragmentation process with respect to (w.r.t.) the initial parton is important in the description of single-spin asymmetries and that the fragmentation function depends on the spin of the quark. It can be asymmetric in transverse momentum distribution if the quark is transversely polarized. More precisely, one should replace the $`D_F^{C/c}(z_C;s_c)`$ in Eq.(8) by
$$D_F^{C/c}(z_C,\stackrel{}{k}_F;s_c)=D_F^{C/c}(z_C,\stackrel{}{k}_F)(1+\alpha \stackrel{}{s}_c\frac{\stackrel{}{p}_c\times \stackrel{}{k}_F}{|\stackrel{}{p}_c\times \stackrel{}{k}_F|}),$$
(28)
where $`\stackrel{}{k}_F`$ is the transverse momentum of $`C`$ w.r.t. the momentum of the parton $`c`$. Based on this assumption, the authors of \[References\] have constructed a simple model based on the string fragmentation model<sup>?</sup> and have calculated $`A_N`$ for meson production.
We compare this possibility with possibility (ii) mentioned in the last subsection, and we see the following. First, just as the assumption of the existence of the asymmetric quark transverse momentum distribution, the existence of spin dependence of the fragmentation function cannot be derived from the model, it has to be introduced by hand. Whether it indeed exists and, if yes, how large it is, has to be studied beyond the model and/or by experiments. Second, compared with the possibility (ii), it has however a great advantage that this assumption can be tested directly by performing suitable experiments! It is possible to have processes where only fragmentation effects play a role. Studying such processes can provide direct information on the spin effects in fragmentation processes. Presently, there exist already a number different measurements which directly or indirectly suggest that this effect cannot be large. In this connection, we have the following.
(a) That the fragmentation function can be dependent of the spin of the quark was first discussed for longitudinally polarized case in 1977 by Nachmann<sup>?</sup>, in 1978 by Efremov<sup>?</sup> and more recently by Efremov, Mankiewicz, Tornqvist<sup>?</sup>. These authors argued that the fragmentation can be dependent of the spin of the fragmented quark and introduced a quantity which they called “handedness” of the jet produced by the fragmentation of the quark. They suggested that the jet handedness should be significantly different from zero if quark fragmentation is indeed dependent of the helicity of the quark.
We recall that, jet handedness $`H`$ is defined as$`^{\text{References}\text{References}}`$,
$$H\frac{N_{\mathrm{\Omega }<0}N_{\mathrm{\Omega }>0}}{N_{\mathrm{\Omega }<0}+N_{\mathrm{\Omega }>0}},$$
(29)
where $`\mathrm{\Omega }\stackrel{}{e}(\stackrel{}{p}_1\times \stackrel{}{p}_2)`$, $`\stackrel{}{e}`$ is the unit vector along the jet axis, $`\stackrel{}{p}_1`$ and $`\stackrel{}{p}_2`$ are the momenta of two particles in the jet chosen in a charge independent way such as $`|\stackrel{}{p}_1|>|\stackrel{}{p}_2|`$. This quantity $`H`$ has been measured recently by SLD Collaboration at SLAC<sup>?</sup>. They found an upper limit of 0.063 for $`H`$ at 95% c.l. This result shows that the dependence of fragmentation on the spin of the quark in the longitudinally polarized case is very weak. If there is such a dependence at all, the corresponding asymmetry that one obtains in the transversely polarized case should be less than 10%, which is far from enough to account for the left-right asymmetries observed in single-spin hadron-hadron collisions$`^{\text{References}\text{References}}`$.
(b) The suggestion<sup>?</sup> that the fragmentation results depend on the spin of the initial quark in the transversely polarized case has in fact a model realization i.e. the LUND model discussed by Andersson, Gustafson and Ingelman<sup>?</sup> in 1979 in connection with hyperon polarization<sup>?</sup> in unpolarized hadron-hadron collisions. Here, a semi-classical picture was proposed to explain the striking $`\mathrm{\Lambda }`$ polarization observed in unpolarized hadron-hadron or hadron-nucleus collisions. We note that hyperon polarizations in unpolarized hadron-hadron collisions have been observed mainly in the fragmentation regions of the colliding hadrons and that $`\mathrm{\Lambda }`$’s in fragmentation regions in $`pp`$ or $`pA`$ collisions are predominately fragmentation products of spin-zero $`ud`$-diquarks. In Lund model, it was argued that, if a quark-antiquark pair $`s\overline{s}`$ is produced in the fragmentation of a spin zero $`ud`$-diquark with the $`s`$-quark has a transverse momentum to the left, there has to be a space separation between the $`s`$ and $`\overline{s}`$ since a piece of string is needed to generate the mass and transverse energy of the $`s\overline{s}`$. Hence, the $`s`$ and $`\overline{s}`$ should have a relative orbital angular momentum pointing upwards. According to angular momentum conservation, the spins of the $`s`$ and $`\overline{s}`$ should have a large probability to point downwards to compensate this relative orbital angular momentum. One expects therefore to see a downwards polarized $`\mathrm{\Lambda }`$ since the spin of $`\mathrm{\Lambda }`$ is completely determined by its $`s`$-quark. This qualitative result is indeed in agreement with the data<sup>?</sup> for $`\mathrm{\Lambda }`$ production in $`pp`$ or or $`pA`$ collisions.
We now apply the arguments to fragmentation of an upwards polarized quark. We note that, to produce a pseudoscalar meson such as a pion with a moderately large transverse momentum, one needs an antiquark which is downwards polarized. The transverse momentum of this antiquark should therefore have a large probability to the left so that the relative angular momentum between the produced $`\overline{q}`$ and $`q`$ can compensate their spins. This leads to an asymmetric transverse momentum distribution for the produced hadrons. We see that such a simple picture indeed leads to a left-right asymmetry for the hadrons produced in the fragmentation of a transversely polarized quark.
Calculations based on LUND model for $`\mathrm{\Lambda }`$ polarization in unpolarized $`pp`$ collisions have been made<sup>?</sup>. The results are substantially smaller than data<sup>?</sup>. Furthermore, if this is indeed the origin of the $`\mathrm{\Lambda }`$ polarization observed in hadron-hadron collisions, one would expect no $`\mathrm{\Lambda }`$ polarization in the beam fragmentation region of $`K^{}+p(0)\mathrm{\Lambda }+X`$ where $`s\mathrm{\Lambda }+X`$ gives the dominate contribution. This is in contradiction with the data<sup>?</sup> which shows that $`\mathrm{\Lambda }`$’s here are positively polarized and the magnitude of the polarization is quite large. Furthermore, measurements of $`\mathrm{\Lambda }`$ transverse polarization in $`e^+e^{}`$ annihilation have also been carried out by TASSO at DESY<sup>?</sup> and ALEPH at CERN<sup>?</sup>. Here, only fragmentation effects contribute. The results<sup>?,?</sup> of both groups show that transverse $`\mathrm{\Lambda }`$ polarization is consistent with zero in $`e^+e^{}`$ annihilation into hadrons.
All these different experiments suggest that spin dependence of fragmentation function is, if at all, quite weak and can not be the major source of the left-right asymmetries observed$`^{\text{References}\text{References}}`$ in high energy singly polarized hadron-hadron collisions. It would also be difficult to understand why $`A_N`$ for $`p()+p(0)\mathrm{\Lambda }+X`$ is<sup>?</sup> quite large in magnitude for large $`x_F`$, if fragmentation is the major source for the existence of such asymmetries. This is because here fragmentation of spin zero $`(u_vd_v)_{0,0}`$ diquarks dominates. Since this is a spin zero object, there should be no left-right asymmetry for the produced $`\mathrm{\Lambda }`$ w.r.t. the moving direction of this $`ud`$-diquark. Hence, if fragmentation is the major source of $`A_N`$, one would expect that $`A_N`$ for $`\mathrm{\Lambda }`$ in the very large $`x_F`$ region is very small, which contradicts the E704 data<sup>?</sup>. Further experimental tests can and will be carried out. Experiments by HERMES at HERA will be able test this assumption directly. Further suggestions to test this can also be found in section 5 below.
## 4 A non-perturbative approach
As has been seen in the last section, the “pQCD based hard scattering models” start from the “hard” side and try to include the influences from the “soft” aspects in some of the unknown factors. The model itself cannot determine whether these “soft” effects indeed exist, and if yes, how large they are. It is also not clear which effect plays the most important role in describing single-spin asymmetries. It is therefore useful and necessary to have phenomenological model studies in order to get deeper insight into the physics behind these possible effects and to find out which effect plays the dominating role. Several approaches$`^{\text{References}\text{References}}`$ of this kind have been proposed recently. These approaches start directly from the “soft” side and thus offers the possibility to study the origin of the “soft effects” in an explicit manner. The most successful one is perhaps the “orbiting valence quark model” proposed by the Berliner group$`^{\text{References}\text{References}}`$. This model is sometimes also referred as “Berliner Model” or “Berliner Relativistic Quark Model (BRQM)” for single-spin asymmetries. In this section, we will concentrate ourselves on this approach. The materials we presented here are based on the publications$`^{\text{References}\text{References}}`$ of the group in this connection and the two doctoral theses<sup>?,?</sup> from us at FU Berlin.
### 4.1 Orbiting valence quarks in polarized nucleon
The existence of orbital motion of quarks in nucleon have been studied by many author in different connections$`^{\text{References}\text{References},\text{References}\text{References}}`$. We note that, various experimental facts show that valence quarks in a light hadron should be treated as Dirac particles moving in a confining field created by other constituents of the same hadron. The masses of these valence quarks are much smaller than the proton mass. This is very much different from what one expected<sup>?</sup> in 1960s that the masses of the quarks in nucleon should be of the order of several GeV and is larger than the proton mass so that non-relativistic approximation can be used in describing the motion of these quarks in the nucleon. In contrast, it is now a well established fact that the masses of the valence quarks of the nucleon are much smaller than the proton mass and is also negligbly small compared to their kinetic energies in nucleon. Hence, to describe their motion inside a nucleon, it is the ultra-relativistic limit rather than the non-relativistic approximation should be used. It is known that Dirac particles confined in a limited space have a number of remarkable properties, and one of these features (which is almost trivial but very important) is the following: Independent of the details of the confining system, the orbital angular momentum of this particle is not a good quantum number — except in the non-relativistic limit. That is, except for those cases in which the ratio between the kinetic energy and the mass of the quark is much less than unity, the eigenstates of the quarks cannot be characterized by their orbital angular momentum quantum numbers. This implies that one can never simply say that the valence quark has a definite orbital angular momentum $`l=0`$. Hence, orbital motion is always involved for the valence quarks even when they are in their ground states.
A demonstrating example, in which the confining potential is taken as central, is given in \[References\] and \[References\]. In this case, stationary states should be characterized by the following set of quantum numbers $`\epsilon ,j,m`$ and $`𝒫`$ which are respectively the eigenvalues of the operators $`\widehat{H}`$ (the Hamiltonian), $`\widehat{\stackrel{}{j}}^2,\widehat{j}_z`$ (the total angular momentum and its $`z`$-component) and $`\widehat{𝒫}`$ (the parity), and the ground state, which is characterized by $`\epsilon =\epsilon _0,`$ $`j=1/2,m=\pm 1/2`$ and $`𝒫=+`$, is given by,
$$\psi _{\epsilon _0\frac{1}{2}m+}(r,\theta ,\varphi )=\left(\begin{array}{cc}& f_{00}(r)\mathrm{\Omega }_0^{\frac{1}{2}m}(\theta ,\varphi )\\ \\ & g_{01}(r)\mathrm{\Omega }_1^{\frac{1}{2}m}(\theta ,\varphi )\end{array}\right),$$
(30)
where,
$$\mathrm{\Omega }_0^{\frac{1}{2}m}(\theta ,\varphi )=Y_{00}(\theta ,\varphi )\xi \left(m\right),$$
(31)
$$\mathrm{\Omega }_1^{\frac{1}{2}m}(\theta ,\varphi )=\sqrt{\frac{32m}{6}}Y_{1m\frac{1}{2}}(\theta ,\varphi )\xi \left(\frac{1}{2}\right)+\sqrt{\frac{3+2m}{6}}Y_{1m+\frac{1}{2}}(\theta ,\varphi )\xi \left(\frac{1}{2}\right).$$
(32)
Here, $`\xi (\pm \frac{1}{2})`$ stand for the eigenfunctions for the spin-operator $`\widehat{\sigma }_z`$ with eigenvalues $`\pm 1`$, and $`Y_\mathrm{}\mathrm{}_z(\theta ,\varphi )`$ for the spherical harmonics which form a standard basis for the orbital angular momentum operators $`(\widehat{\stackrel{}{\mathrm{}}}^2,\widehat{\mathrm{}}_z)`$. The radial part, $`f_{00}(r)`$ and $`g_{01}(r)`$, of the two-spinors are determined by the Dirac equation for given potentials. Taking the $`j_zm=+\frac{1}{2}`$ state as example, we obtain,
$$\widehat{l}_z(\epsilon _0,\frac{1}{2},\frac{1}{2},+)=\frac{2}{3}_0^{\mathrm{}}g_{01}^2(r)r^2𝑑r>0,$$
(33)
$$\widehat{l}^2(\epsilon _0,\frac{1}{2},\frac{1}{2},+)=2_0^{\mathrm{}}g_{01}^2(r)r^2𝑑r>0.$$
(34)
Also, the current density $`J^\mu =(\rho ,\stackrel{}{J})=\overline{\psi }\gamma ^\mu \psi `$ is given by,
$$\rho (\epsilon _0,\frac{1}{2},\frac{1}{2},+|\stackrel{}{r})=\frac{1}{4\pi }[f_{00}^2(r)+g_{01}^2(r)],$$
(35)
$$J_x(\epsilon _0,\frac{1}{2},\frac{1}{2},+|\stackrel{}{r})=+\frac{y}{2\pi r}f_{00}(r)g_{01}(r),$$
(36)
$$J_y(\epsilon _0,\frac{1}{2},\frac{1}{2},+|\stackrel{}{r})=\frac{x}{2\pi r}f_{00}(r)g_{01}(r),$$
(37)
$$J_z(\epsilon _0,\frac{1}{2},\frac{1}{2},+|\stackrel{}{r})=0.$$
(38)
Here, we explicitly see that, even in the ground state, $`\widehat{l}_z0`$, $`\stackrel{}{J}(\stackrel{}{r})0`$. This implies that orbital motion is always involved for the valence quarks in the nucleon. We see also that $`l_z>0`$ implies that the momentum distributions of these valence quarks and their spatial coordinates inside the hadron must be correlated in a given manner. This can be seen by examining at the $`p_y`$-distribution of this valence quark at different $`x`$, namely in different plane parallel to the $`oyz`$-plane. It has been demonstrated that<sup>?</sup> for $`x>0`$, $`p_y`$ has an extra component in the positive $`y`$-direction i.e. $`p_y>0`$; but for $`x<0`$, $`p_y`$ has an extra component in the negative $`y`$-direction i.e. $`p_y<0`$.
These features of the intrinsic motion of the quarks inside nucleon, in particular the distribution of transverse momentum mentioned above, are very interesting and they should be able to manifest themselves in hadron production in hadron-hadron collisions. To study this, we need to know how the quarks are polarized in a polarized nucleon. This is determined by the wavefunction of the nucleon. It has also been shown<sup>?,?</sup> that the wave function of the proton in such a relativistic quark model can be obtained simply by replacing the Pauli spinors in the static quark models by the corresponding Dirac spinors. Both of them are direct consequence of Pauli principle and the assumption that baryon is in the color singlet state hence the color degree is completely antisymmetric.
The wave function has two direct consequences: First, it can be used to calculate the baryons’ magnetic moments $`\mu _B`$ in terms of the magnetic moments $`\mu _u`$, $`\mu _d`$ and $`\mu _s`$ of the valence quarks. An explicit expression of $`\mu _B`$ in terms of $`\mu _u`$, $`\mu _d`$ and $`\mu _s`$ has been obtained in \[References\]. The results show that this expression is exactly the same as that in the static quark model and it is independent of the confining potentials. The only difference is that $`\mu _u`$, $`\mu _d`$ and $`\mu _s`$ are free parameters in the static models, they are constants depending on the choices of the confining potentials in the relativistic models. This is very interesting since it shows that such relativistic models are as good (or as bad) as the static models in describing the baryons’ magnetic moments data. It explains also why the static models give a reasonable description of baryons’ magnetic moments although we know that the quark masses are not as large as one thought in the 1960s to justify the use of non-relativistic limit. It explains also why the obtained values for $`\mu _u`$, $`\mu _d`$ and $`\mu _s`$ in the static models by fitting the data are not simply $`1/(2m_q)`$. Second, the polarization of the valence quarks is also determined by the wave function of the nucleon. This implies that, for proton, $`5/3`$ of the 2 $`u`$ valence quarks are polarized in the same, and $`1/3`$ in the opposite, direction as the proton. For $`d`$, they are $`1/3`$ and $`2/3`$ respectively. We see that both of them are polarized and the polarization is flavor dependent.
### 4.2 Production mechanism for hadrons in the fragmentation region in hadron-hadron collisions
In order to study the asymmetries observed in single-spin reactions, we need to know the production mechanism of the hadron in such reactions. Since the asymmetries are observed mainly in the fragmentation region, we will also concentrate on the production of hadrons in this region.
We recall that, already in 1970s, it has become well-known (See, e.g., \[References, References\] and the papers cited there.) that inclusive longitudinal momentum distribution of a given type of hadron which has a valence quark in common with one of the colliding hadrons reflects directly the momentum distribution of this valence quark in that colliding hadron. Different experiments (See, e.g., \[References-References\]) have shown that the longitudinal momentum distributions of the produced hadrons in the fragmentation region are very much similar to those of the corresponding valence quarks in the colliding hadrons. More precisely, it has been observed$`^{\text{References}\text{References}}`$ that, e.g., the number density of $`\pi ^+`$ or that of $`K^+`$ in $`pp`$-collisions is proportional to the $`u`$-valence-quark distribution in proton; that of $`\pi ^{}`$ is proportional to that of $`d`$-valence quark. But that for $`K^{}`$, which does not share a valence quark with the colliding hadron, drops much more fast and earlier than that for $`K^+`$ in the large $`x_F`$ region. These experimental facts suggest that, if there is a scattering process at all which takes place between these valence quarks and other constituents of the colliding hadrons before they hadronize into hadrons, this scattering should not destroy the momentum distribution of the valence quarks. In other words, the momentum transfer in the scattering cannot be large. This implies that there should be no hard scattering in these processes between such valence quarks and other constituents of the colliding hadrons.
Having these in mind, one is naturally led to the following picture$`^{\text{References}\text{References}}`$ for the production of mesons in the fragmentation regions: A valence quark of one of the colliding hadrons picks up an anti-sea-quark associated with the other and combine with it to form a meson which can be observed experimentally, e.g. $`q_v^P+\overline{q}_s^TM`$. \[Here, the subscript of a quark, $`v`$ or $`s`$, denotes whether it is for valence or sea quarks; the superscript, $`P`$ or $`T`$, denotes whether it is from the projectile or the target.\] This mechanism for hadron production is referred as “direct formation” or “direct fusion”.
We note that the “direct fusion” mechanism mentioned above is very similar to the recombination model proposed by Das and Hwa<sup>?</sup> several years ago. These are models which aim to describe the production of hadrons in the fragmentation regions. This is also the simplest model which reproduces the data for hadron production in fragmentation regions of unpolarized hadron-hadron collisions. There exist surely many other hadronization models in the literature, some of which can also reproduce most of the data for hadron production in the unpolarized hadronic reactions. Since our purpose here is to investigate the origin of the left-right asymmetries observed in the fragmentation region in single-spin hadron-hadron collisions, in particular the contribution of orbital motion of the valence quarks to these asymmetries, it is our intention to simplify the other factors as much as we are allowed in order to show the effects from the key factors. We therefore chose this model for the production of hadrons in the fragmentation region. In the following of this subsection, we will briefly review the picture and its main success in describing the unpolarized data to show that it can indeed describe the main properties of the hadrons in fragmentation regions. For a review of the recombination models and/or comparison of this model to other hadronization models, see e.g. \[References,References\].
For the sake of definiteness, we now consider the process $`p(0)+p(0)M+X`$. It is clear that, in this picture, the number density $`N(x_F,M|s)`$ of the produced mesons should be given by,
$$N(x_F,M|s)=N_0(x_F,M|s)+D(x_F,M|s),$$
(39)
where $`N_0(x_F,M|s)`$ represents the contribution from non-direct formation, which comes from the interaction of the seas (sea quarks, sea antiquarks and gluons) of the two colliding hadrons; and $`D(x_F,M|s)`$ is the number density of the meson produced through the direct formation process $`q_v^P+\overline{q}_s^TM`$. Obviously, the $`x_F`$-dependence of $`D`$ can be obtained from the following integrals,
$$D(x_F,M|s)=\underset{q_v,\overline{q}_s}{}𝑑x^P𝑑x^Tq_v(x^P)\overline{q}_s(x^T)K(x^P,q_v;x^T,\overline{q}_s|x_F,M,s).$$
(40)
Here $`q_v(x)`$ is the distribution of the valence quarks in the projectile proton, and $`\overline{q}_s(x)`$ is the sea-quark distribution in the target; $`K(x^P,q_v;x^T,\overline{q}_s|x_F,M,s)`$ is the probability density for a valence quark of flavor $`q_v`$ with fractional momentum $`x^P`$ to combine directly with an anti-sea-quark of flavor $`\overline{q}_s`$ with fractional momentum $`x^T`$ to form a meson $`M`$ with fractional momentum $`x_F`$. We note that whether, if yes how much, the $`K`$-function depends on the dynamical details is something we do not know a priori. But, what we do know is that this function has to guarantee the validity of all the relevant conservation laws. Hence, the simplest choice of the corresponding $`K`$-function for e.g. $`\pi ^+=(u\overline{d})`$ is,
$$K(x^P,q_v;x^T,\overline{q}_s|x_F,\pi ^+,s)=\kappa _\pi \delta _{q_v,u}\delta _{\overline{q}_s,\overline{d}}\delta (x^Px_F)\delta (x^T\frac{x_0}{x_F}),$$
(41)
where, $`\kappa _\pi `$ is a constant; the two Dirac-$`\delta `$-functions come from the energy and momentum conservation which requires $`x^Px_F`$ and $`x^Tx_0/x_F`$ where $`x_0=m^2/s`$ ($`m`$ is the mass of the produced meson). In this way, we obtain,
$$D(x_F,M|s)=\kappa q_v(x_F)\overline{q}_s(\frac{x_0}{x_F}),$$
(42)
It can easily be seen that such a picture is consistent with the above mentioned experimental observations, namely in the fragmentation region,
$$N(x_F,M|s)D(x_F,M|s)q_v(x_F).$$
(43)
It is also consistent with the existence <sup>?,?</sup> of a limiting behavior in the fragmentation region at high energies. This is because, at extremely high energy, $`x^T`$ is very small and $`\overline{q}_s(x^T)`$ is very large. This implies that there exists a tremendously large number of antiseaquarks which are suitable to combine with the valence quarks $`q_v^P`$ of the projectile to form the mesons. Since we have only three valence quarks, further increasing of the energy, which means further increasing the number of the antiseaquarks, will bring nothing more, thus the distribution of the mesons produced in the fragmentation region remains the same.
Since $`N_0(x_F,M|s)`$ comes from the interaction of the sea of the colliding hadrons, it is expected that $`N_0`$ should be isospin invariant and should be the same for particle and antiparticle. This was first checked<sup>?</sup> using data<sup>?,?</sup> for pion production. Here, one can obtain $`N_0(x_F,M|s)`$ by subtracting $`D(x_F,M|s)`$ from the corresponding data for $`N(x_F,M|s)`$, where $`D(x_F,M|s)`$ can be calculated by using the parameterizations for the quark distribution functions (See, e.g., \[References,References\]). The results for such a subtraction is shown in Fig.3. In Fig.4, we see the cross-sections<sup>?,?</sup> as the sums of two parts: the direct-formation part and the non-direct-formation part.
Form these results, we not only see that the non-direct-formation part is indeed isospin-independent but also explicitly see the existence of a transition region in the inclusive cross-sections near $`x_F=0.40.5`$. In this region, $`N_0(x_F,\pi |s)`$ and $`D(x_F,\pi |s)`$ are expected to switch their roles: While the former is the main contribution for $`x_F<0.40.5`$, the latter begins to dominate for larger values of $`x_F`$.
There is an advantage to use $`K`$-production in $`p(0)+p(0)K+X`$ in testing this picture. Here, direct fusion of the valence quarks with suitable anti-sea-quarks does not contribute to the production of $`K^{}`$ and $`\overline{K}^0`$. This means, for such kaons, we have only the contributions from the non-direct-formation parts $`N_0(x_F,K|s)`$, which should be the same for $`K^\pm ,K^0`$ and $`\overline{K}^0`$. Hence, $`N_0(x_F,K|s)`$ can be determined unambiguously by experiments. We thus obtain from the following relations between the number densities (or the corresponding differential cross sections) for $`K`$ produced in $`p(0)+p(0)K+X`$:
$$N(x_F,K^{}|s)=N(x_F,\overline{K}^0|s)=N_0(x_F,K|s),$$
(44)
$$N(x_F,K^+|s)=N(x_F,K^{}|s)+\kappa _Ku_v(x_F)\overline{s}_s(\frac{x_0}{x_F}),$$
(45)
$$N(x_F,K^0|s)=N(x_F,K^{}|s)+\kappa _Kd_v(x_F)\overline{s}_s(\frac{x_0}{x_F}).$$
(46)
These are direct consequences of the picture and can be used to test the picture in a quantitative manner. A comparison of Eq.(45) with the ISR data<sup>?,?</sup> was made in \[References\] and is shown in Fig.5. We see that the agreement is indeed very good.
The picture has also been applied<sup>?</sup> to $`\mathrm{\Lambda }`$ production. Here, we have the following three possibilities<sup>?</sup> for direct formations of $`\mathrm{\Lambda }`$ in $`p(0)+p(0)\mathrm{\Lambda }+X`$:
(a) A $`(u_vd_v)`$-valence-diquark from the projectile $`P`$ picks up a $`s_s`$-sea-quark associated with the target $`T`$ and forms a $`\mathrm{\Lambda }`$: $`(u_vd_v)^P+s_s^T\mathrm{\Lambda }`$.
(b) A $`u_v`$-valence-quark from the projectile $`P`$ picks up a $`(d_ss_s)`$-sea-diquark associated with the target $`T`$ and forms a $`\mathrm{\Lambda }`$: $`u_v^P+(d_ss_s)^T\mathrm{\Lambda }`$.
(c) A $`d_v`$-valence-quark from the projectile $`P`$ picks up a $`(u_ss_s)`$-sea-diquark associated with the target $`T`$ and forms a $`\mathrm{\Lambda }`$: $`d_v^P+(u_ss_s)^T\mathrm{\Lambda }`$.
Just as that for meson, the corresponding number densities are given by,
$$D_a(x_F,\mathrm{\Lambda }|s)=\kappa _\mathrm{\Lambda }^df_D(x^P|u_vd_v)s_s(x^T),$$
(47)
$$D_b(x_F,\mathrm{\Lambda }|s)=\kappa _\mathrm{\Lambda }u_v(x^P)f_D(x^T|d_ss_s),$$
(48)
$$D_c(x_F,\mathrm{\Lambda }|s)=\kappa _\mathrm{\Lambda }d_v(x^P)f_D(x^T|u_ss_s),$$
(49)
respectively. Here, $`x^Px_F`$ and $`x^Tm_\mathrm{\Lambda }^2/(sx_F)`$, followed from energy-momentum conservation. $`f_D(x|q_iq_j)`$ is the diquark distribution functions, where $`q_iq_j`$ denotes the flavor and whether they are valence or sea quarks. $`\kappa _\mathrm{\Lambda }^d`$ and $`\kappa _\mathrm{\Lambda }`$ are two constants. The number density for $`\mathrm{\Lambda }`$ is given by:
$$N(x_F,\mathrm{\Lambda }|s)=N_0(x_F,\mathrm{\Lambda }|s)+\underset{i=a,b,c}{}D_i(x_F,\mathrm{\Lambda }|s).$$
(50)
where $`N_0(x_F,\mathrm{\Lambda }|s)`$ is the non-direct-formation part.
$`D_i(x_F,\mathrm{\Lambda }|s)`$ has been calculated<sup>?</sup> using the parameterizations of the quark and diquark distribution (See, e.g., \[References,References\] and \[References\]). The two constants $`\kappa _\mathrm{\Lambda }`$ and $`\kappa _\mathrm{\Lambda }^d`$ were fixed by fitting two data points in the large $`x_F`$-region. The results are compared with the data<sup>?</sup> in Fig.6. We see that the data can indeed be fitted very well in the fragmentation region. In fact, compared with those for pions (See, e.g., \[References\] and \[References\]), the characteristic feature of the data for $`\mathrm{\Lambda }`$ (See, e.g., \[References\]) is that it is much broader than the latter in the large $`x_F`$ region, and this is just a direct consequence of the contribution from the valence diquarks through the process (a) given above. We see also that the whole $`0<x_F<1`$ region can be divided into three parts: In the large $`x_F`$-region (say, $`x_F>0.6`$), the direct process (a) plays the dominating role; and for small $`x_F`$-values ($`x_F<0.3`$, say), the non-direct-formation part $`N_0(x_F,\mathrm{\Lambda }|s)`$ dominates, while in the middle (that is, in the neighborhood of $`x_F0.40.5`$), the direct formation processes (b) and (c) provide the largest contributions.
### 4.3 Left-right asymmetry for meson production
Having seen that $`A_N`$ is significant only in the fragmentation region, that hadron production in this region can be well described in terms of the abovementioned direct fusion of the valence quarks with suitable antiseaquarks, and that one of the most remarkable properties of the valence quarks in polarized nucleon is that they have to perform orbital motions even when they are in their ground state, it is natural to ask: Can we describe the observed asymmetry data if we take the orbital motion of the valence quarks into account? In particular, we have seen that the longitudinal momentum distribution of the produced hadrons in the fragmentation regions directly reflects the longitudinal momentum distributions of the valence quarks. Does the transverse momentum distribution of these hadrons reflect also the intrinsic transverse motion of the valence quarks in hadrons?
It has been shown$`^{\text{References}\text{References}}`$ that these questions should be answered in the affirmative. The existence of the left-right asymmetries in the fragmentation regions should be considered as a strong evidence for the existence of orbital motion of the valence quarks in a transversely polarized nucleon and the hadronic surface effects in single-spin hadron-hadron collisions. The arguments for the existence of the surface effect caused by the “initial state interactions” of the constituents of the two colliding hadrons when they are overlap in space are summarized in the following<sup>?,?</sup>.
Hadrons — the constituents of which are known to be quarks and gluons — are spatially extended objects. Inside a hadron, the constituents interact with one another through color forces, and only color-singlets can leave the color-fields. As spatially extended objects, the geometrical aspect of hadrons plays an important role in describing unpolarized high-energy hadron-hadron collisions. In this connection, it is useful to recall the following: Without specifying the details about the constituents and their interactions, the geometrical picture in which the colliding hadrons are simply envisaged<sup>?</sup> to “go through each other with attenuation thereby in general get excited and break up” is already able to describe the characteristic limiting behaviors of the final state particles in the fragmentation regions. Having these in mind, we now examine the role played by geometry in direct-meson-formation processes in the fragmentation regions when one of the colliding hadrons is transversely polarized
For the sake of convenience and definiteness, we consider the process in the rest frame of the projectile $`P`$, in which $`P`$ is an upwards transversely polarized proton, and the target proton $`T`$ is unpolarized. We concentrate our attention on the transverse momentum distribution of the directly formed mesons observed in the projectile fragmentation region, and ask the following questions: (A) What do we expect to see if the observed meson is formed through fusion of a $`u`$-valence-quark, which was upwards polarized before the collision takes place, of $`P`$ near its front surface (towards the target proton $`T`$) and a suitable anti-sea-quark of $`T`$? (B) What do we expect to see when this direct fusion takes place elsewhere in $`P`$ — in particular near $`P`$’s back surface?
To answer these questions, we recall that, in the transversely polarized $`P`$, the valence quarks are performing orbital motion and the direction of such orbital motion is determined by the polarization. In particular, an upwards polarized $`u`$-valence quark is “going-left” when looked near the front surface of $`P`$. Hence, because of momentum conservation, the answer to (A) is the following: Due to orbital motion of the valence quarks, it is more probable that this meson acquires an extra transverse momentum going to the left. This is to be compared with the direct fusion processes which take place when $`T`$ has already entered $`P`$ after some time, in particular when $`T`$ would be already near the back surface and about to leave $`P`$ if both $`P`$ and $`T`$ would remain undestroyed by the color interactions between their constituents. We note: before $`T`$ enters $`P`$, axial symmetry w.r.t. the polarization axis requires that the upwards $`u`$-valence quark is “right-going” near the back surface of $`P`$. Hence, if such a valence $`u`$-quark could retain its transverse momentum until $`T`$ reaches $`P`$’s back surface, the produced meson would have the same probability to go right as that for the meson formed near the front surface to go left. We would obtain that, in total, the chance for a meson formed by fusion of an upwards polarized $`u`$ valence quark with suitable anti-sea-quark to go left and that to go right equal to each other. But, since color forces between the constituents in the $`PT`$ system become effective as soon as $`P`$ and $`T`$ begin to overlap, and such interactions (often referred as “initial state interaction”) in general cause changes in the intrinsic motion of the constituents inside the hadrons $`P`$ and $`T`$. The above-mentioned $`u`$-valence-quark in $`P`$ would be able to retain its initial momentum only if the time needed for $`T`$ to travel through $`P`$ would be less than the time needed for the color interaction to propagate the same distance. Hence, in a picture — in fact in any picture — in which $`T`$ cannot go through $`P`$ without any resistance/attenuation, while color gluons propagate with the velocity of light, it is more likely that the contrary is true. This means, the answer to Question (B) should be: If the direct fusion takes place near the back surface of $`P`$, it is very likely that the $`u`$ valence quark has already lost the information of polarization which it had in the initial state hadron thus the produced hadron can go left or right with equal probabilities. We thus obtain that meson formed through direct fusion of an upwards polarized valence quark with a suitable anti-sea-quark has a large probability to go left.
Together with the surface effect, the orbital motion of a polarized valence quark in a polarized nucleon leads to a correlation between the polarization of the valence quarks and the direction of transverse motion of the mesons produced through the direct fusion of the valence quarks and suitable antiquarks. More precisely, we obtain that mesons produced through the direct formation of upwards transversely polarized valence quarks of the projectile with suitable antiseaquarks associated with the target have large probability to go left and vice versa. Hence, once we know the polarization of the projectile, we can use the baryon wave function to determine the polarization of its valence quarks and the signs of $`A_N`$’s for the produced hadrons. E.g., for $`p()+p(0)\pi +X`$, we obtain the results in table 1. Here, in the table, the following coordinate is used: The projectile is moving in $`+z`$ direction, polarization up is $`+x`$ direction; therefore moving to the left means $`p_y<0`$ (denoted by $``$) and moving to the right means $`p_y>0`$ (denoted by $``$). From this table we see clearly the following:
* $`A_N[p()+p(0)\pi ^++X]>0,`$ $`A_N[p()+p(0)\pi ^{}+X]<0,`$
* $`0<A_N[p()+p(0)\pi ^0+X]<A_N[p()+p(0)\pi ^++X]`$,
* The magnitudes of all these $`A_N`$’s increase with increasing $`x_F`$.
All these qualitative features are in good agreement with the data$`^{\text{References}\text{References}}`$.
Similar tables can also be constructed in a straightforward manner for the production of other mesons and/or in processes using other projectiles and/or targets. In this way, we obtained many other direct associations of the picture which can be tested directly by experiments. E.g.:
(A) If we use, instead of $`p()`$, $`\overline{p}()`$-projectile, we obtain table 2. Compare the results in this table with those in table 1, we see that, while $`A_N`$ for $`\pi ^0`$ remains the same, those for $`\pi ^+`$ and $`\pi ^{}`$ change their roles. More precisely, $`A_N`$ for $`\overline{p}()+p(0)\pi ^{}+X`$ should be approximately the same as that for $`p()+p(0)\pi ^++X`$; that for $`\overline{p}()+p(0)\pi ^++X`$ should be approximately the same as that for $`p()+p(0)\pi ^{}+X`$.
(B) It is also clear that the asymmetries are expect to be significant in the fragmentation of the polarized nucleon but should be absent in the fragmentation of the unpolarized hadron. This implies in particular that there should be no asymmetry in the beam fragmentation region of $`\pi ^{}+p()\pi +X`$.
(A) is a prediction of \[References, References\] and \[References\] and has been confirmed by the E704 data<sup>?,?</sup>. (B) was also predicted in \[References\] and \[References\] and has been confirmed by the data<sup>?</sup>.
Since, as have seen in the last section, the $`x_F`$-dependence for the number density of mesons for direct formation can be calculated easily, we can also calculate the $`x_F`$-dependence for $`A_N`$ in a quantitative manner. Such calculations have been carried out in \[ReferencesReferences\]. We summarize them in the following. For the sake of explicity, we consider $`p()+p(0)M+X`$. We recall that \[see Eqs.(1-3)\] $`A_N(x_F,M|s)`$ is defined as the ratio of the difference and the sum of $`N_L(x_F,M|s,)`$ and $`N_L(x_F,M|s,)`$. We denote by $`D(x_F,M,+|s,)`$ the number density of $`M`$’s produced through direct fusion of the valence quarks which are polarized in the same direction as the transversely polarized proton and $`D(x_F,M,|s,)`$ the corresponding number density for $`M`$’s formed by the valence quarks polarized in the opposite direction as the projectile proton. Then the $`N`$’s can be written as:
$$N_L(x_F,M|s,)=\alpha D(x_F,M,+|s,)+(1\alpha )D(x_F,M,|s,)+N_{0L}(x_F,M|s),$$
(51)
$$N_L(x_F,M|s,)=(1\alpha )D(x_F,M,+|s,)+\alpha D(x_F,M,|s,)+N_{0L}(x_F,M|s).$$
(52)
Here, $`\alpha `$ stands for the probability for a meson produced by the direct fusion of an upwards polarized valence quark with a suitable anti-sea-quark to go left. It follows from Eqs.(51) and (52),
$$N_L(x_F,M|s,)N_L(x_F,M|s,)=C[D(x_F,M,+|s,tr)D(x_F,M,|s,tr)],$$
(53)
where $`C2\alpha 1`$, $`D(x_F,M,\pm |s,tr)D(x_F,M,\pm |s,)=D(x_F,M,\pm |s,)`$. Hence,
$$A_N(x_F,M|s)=\frac{C\mathrm{\Delta }D(x_F,M|s,tr)}{2N_{0L}(x_F,M|s)+D(x_F,M|s)},$$
(54)
Since mesons directly formed through fusion of upwards polarized valence-quark of the projectile and anti-sea-quarks of the target have a larger probability to go left, we have $`1/2<\alpha <1`$ which implies $`0<C<1`$. It can easily be seen that $`N_0(x_F,M|s)[=2N_{0L}(x_F,M|s)]`$ — especially the interplay between this quantity and the corresponding $`D(x_F,M|s)`$ — plays a key role in understanding the $`x_F`$-dependence of $`A_N(x_F,M|s)`$.
As we have shown in last section, $`N_0(x_F,M|s)`$ can be obtained from the unpolarized data. The $`D`$’s can be obtained from the following integrals,
$$D(x_F,M,\pm |s,tr)=\underset{q_v,\overline{q}_s}{}dx^Pdx^Tq_v^\pm (x^P|tr)\overline{q}_s(x^T)K(x^P,q_v;x^T,\overline{q}_s|x_F,M,s).$$
(55)
Here $`q_v^\pm (x|tr)`$ is the distribution of the valence quarks polarized in the same or in the opposite direction of the transversely polarized proton. Using the $`K(x^P,q_v;x^T,\overline{q}_s|x_F,M,s)`$’s as that for $`\pi ^+`$ given in Eq.(41), we obtain the $`D(x_F,M,\pm |s,tr)`$’s and thus,
$$A_N(x_F,\pi ^+|s)=\frac{C\kappa _\pi \mathrm{\Delta }u_v(x_F|tr)\overline{d}_s(\frac{x_0}{x_F})}{N_0(x_F,\pi |s)+\kappa _\pi u_v(x_F)\overline{d}_s(\frac{x_0}{x_F})},$$
(56)
$$A_N(x_F,\pi ^{}|s)=\frac{C\kappa _\pi \mathrm{\Delta }d_v(x_F|tr)\overline{u}_s(\frac{x_0}{x_F})}{N_0(x_F,\pi |s)+\kappa _\pi d_v(x_F)\overline{u}_s(\frac{x_0}{x_F})},$$
(57)
$$A_N(x_F,\pi ^0|s)=\frac{C\kappa _\pi [\mathrm{\Delta }u_v(x_F|tr)\overline{u}_s(\frac{x_0}{x_F})+\mathrm{\Delta }d_v(x_F|tr)\overline{d}_s(\frac{x_0}{x_F})]}{2N_0(x_F,\pi |s)+\kappa _\pi [u_v(x_F)\overline{u}_s(\frac{x_0}{x_F})+d_v(x_F)\overline{d}_s(\frac{x_0}{x_F})]},$$
(58)
$$A_N(x_F,K^+|s)=\frac{C\kappa _K\mathrm{\Delta }u_v(x_F|tr)\overline{s}_s(\frac{x_0}{x_F})}{N(x_F,K^{}|s)+\kappa _Ku_v(x_F)\overline{s}_s(\frac{x_0}{x_F})},$$
(59)
$$A_N(x_F,K^0|s)=\frac{C\kappa _K\mathrm{\Delta }d_v(x_F|tr)\overline{s}_s(\frac{x_0}{x_F})}{N(x_F,K^{}|s)+\kappa _Kd_v(x_F)\overline{s}_s(\frac{x_0}{x_F})},$$
(60)
and $`A_N(x_F,K^{}|s)=A_N(x_F,\overline{K}^0|s)=0`$.
It should be emphasized that the $`q_v^\pm (x|tr)`$ defined above is quite different from the $`q^\pm (x|l)`$ defined in the parton model: First, while the former refers to the transverse polarization, the latter refers to the longitudinal polarization. Second, the $`\pm `$ in the former refers to the third component of the total angular momenta of the valence quarks, that in the latter refers to the helicities of the quarks. $`q_v^\pm (x|tr)`$ has not yet been determined experimentally. The only thing we know theoretically is that they have to satisfy the following constraints:
$$𝑑xu_v^+(x|tr)=5/3,𝑑xu_v^{}(x|tr)=1/3,$$
(61)
$$𝑑xd_v^+(x|tr)=1/3,𝑑xd_v^{}(x|tr)=2/3,$$
(62)
Hence, the simplest ansatz for these $`q_v^\pm (x|tr)`$ is the following:
$$u_v^+(x|tr)=(5/6)u_v(x),u_v^{}(x|tr)=(1/6)u_v(x),$$
(63)
$$d_v^+(x|tr)=(1/3)d_v(x),d_v^{}(x|tr)=(2/3)d_v(x),$$
(64)
This was used in \[References-References\] to make rough estimation of $`A_N`$.
#### 4.3.1 $`A_N`$ for $`\pi `$ production
A rough estimation for $`A_N`$ for $`p()+p(0)\pi +X`$ was made in \[References\] using Eqs.(56)–(58) and Eqs.(63)–(64). The results are shown in Fig.7. We see that all the qualitative features of the data are well reproduced.
It has also been found<sup>?</sup> that there exist many simple relations between the $`A_N`$’s for hadron production in reactions using different projectile-target combinations. E.g.,
$$A_N^{\overline{p}()p(0)}(x_F,\pi ^+|s)A_N^{p()p(0)}(x_F,\pi ^{}|s),$$
(65)
$$A_N^{\overline{p}()p(0)}(x_F,\pi ^{}|s)A_N^{p()p(0)}(x_F,\pi ^+|s),$$
(66)
$$A_N^{\overline{p}()p(0)}(x_F,\pi ^0|s)A_N^{p()p(0)}(x_F,\pi ^0|s),$$
(67)
$$A_N^{p(0)p()}(x_F,\pi ^+|s)=A_N^{p(0)n()}(x_F,\pi ^{}|s)A_N^{\pi p()}(x_F,\pi ^+|s),$$
(68)
$$A_N^{p(0)p()}(x_F,\pi ^{}|s)=A_N^{p(0)n()}(x_F,\pi ^+|s)A_N^{\pi p()}(x_F,\pi ^{}|s),$$
(69)
$$A_N^{p(0)p()}(x_F,\pi ^0|s)=A_N^{p(0)n()}(x_F,\pi ^0|s)A_N^{\pi p()}(x_F,\pi ^0|s),$$
(70)
$$A_N^{p(0)D()}(x_F,\pi ^+|s)=A_N^{p(0)D()}(x_F,\pi ^{}|s)=A_N^{p(0)D()}(x_F,\pi ^0|s)=A_N^{p(0)p()}(x_F,\pi ^0|s).$$
(71)
Here, the superscript of $`A_N`$ denotes the projectile and the target of the reaction. Eqs.(65)-(67) are consistent with the data$`^{\text{References}\text{References}}`$. In fact, Eqs.(65) and (66) were predicted<sup>?,?</sup> before the corresponding were available. The others can be tested by future experiments.
#### 4.3.2 $`K`$ production
Comparison of Eq.(59) with Eq.(56), and Eq.(60) with Eq.(57) explicitly shows<sup>?,?</sup> that $`A_N(x_F,K^+|s)`$ should be similar to $`A_N(x_F,\pi ^+|s)`$, and that $`A_N(x_F,K^0|s)`$ should be similar to $`A_N(x_F,\pi ^{}|s)`$. Since both the $`K_S^0`$ and $`K_L^0`$ are linear combinations of $`K^0`$ and $`\overline{K}^0`$, the left-right asymmetry should be the same for both of them, and it is given by<sup>?</sup>
$$A_N(x_F,K_S^0|s)=\frac{C\kappa _K\mathrm{\Delta }d_v(x_F|tr)\overline{s}_s(\frac{x_0}{x_F})}{2N(x_F,K^{}|s)+\kappa _Kd_v(x_F)\overline{s}_s(\frac{x_0}{x_F})}.$$
(72)
Hence $`A_N(x_F,K_S^0|s)`$ should have the same sign as $`A_N(x_F,\pi ^{}|s)`$. This has been confirmed by the preliminary E704 data<sup>?</sup>. A quantitative estimation of the $`A_N`$’s for Kaons have also been made<sup>?</sup>. The results are shown in Fig.8.
In this connection, it may be particularly interesting to note the following. If we use, instead of transversely polarized proton beam, transversely polarized anti-proton beam, we have,
$$A_N^{\overline{p}()p}(x_F,K^{}|s)=\frac{C\kappa _K\mathrm{\Delta }\overline{u}_v^{\overline{p}}(x_F|tr)s_s(\frac{x_0}{x_F})}{N(x_F,K^{}|s)+\kappa _K\overline{u}_v^{\overline{p}}(x_F)s_s(\frac{x_0}{x_F})},$$
(73)
$$A_N^{\overline{p}()p}(x_F,K_S^0|s)=\frac{C\kappa _K\mathrm{\Delta }\overline{d}_v^{\overline{p}}(x_F|tr)s_s(\frac{x_0}{x_F})}{2N(x_F,K^{}|s)+\kappa _K\overline{d}_v^{\overline{p}}(x_F)s_s(\frac{x_0}{x_F})},$$
(74)
and $`A_N^{\overline{p}()p}(x_F,K^+|s)=0`$. Using $`\mathrm{\Delta }\overline{q}_v^{\overline{p}}(x|tr)=\mathrm{\Delta }q_v(x|tr)`$ and $`\overline{q}_v^{\overline{p}}(x)=q_v(x)`$, we obtain that
$$A_N^{\overline{p}()p}(x_F,K^{}|s)=A_N(x_F,K^+|s),$$
(75)
$$A_N^{\overline{p}()p}(x_F,K_S^0|s)=A_N(x_F,K_S^0|s).$$
(76)
The latter has also been confirmed by the preliminary E704 data<sup>?</sup>.
### 4.4 $`A_N`$ for lepton-pair production
Not only mesons but also lepton pairs should<sup>?</sup> exhibit left-right asymmetry in single-spin processes. This is particularly interesting because the production mechanism — the well known Drell-Yan mechanism<sup>?</sup> — is very clear, and there is no fragmentation hence no contribution from fragmentation. The measurements of such asymmetry provide a crucial test of the model and is also very helpful to study the origin of the observed $`A_N`$ in general. This will be discussed in more detail in Section 5.
The calculation for $`A_N`$ for lepton-pair is straightforward. Here, we can calculate not only the numerator in a similar way as that for hadron production, but also the denominator by using the Drell-Yan mechanism. Such calculations have been done in \[References\] and \[References\]. Here, we give some examples of the results in Figs. 7, 9 and 10.
### 4.5 $`A_N`$ for hyperon production
Striking $`A_N`$ have also been observed<sup>?</sup> for $`\mathrm{\Lambda }`$. Compared with those for pions<sup>?</sup>, $`A_N`$ for $`\mathrm{\Lambda }`$ as a function of $`x_F`$ shows the following (C.f. Fig.1):
(1) Similar to those for pions, $`A_N(x_F,\mathrm{\Lambda }|s)`$ is significant in the large $`x_F`$ region ($`x_F>0.6`$, say), but small in the central region.
(2) $`A_N(x_F,\mathrm{\Lambda }|s)`$ is negative in the large $`x_F`$ region, and it behaves similarly as $`A_N(x_F,\pi ^+|s)`$ does in this region.
(3) $`A_N(x_F,\mathrm{\Lambda }|s)`$ begins to rise up later than $`A_N(x_F,\pi ^\pm |s)`$ does. For $`x_F`$ in the neighborhood of $`x_F0.40.5`$, it is even positive. In the region $`0<x_F<0.5`$, its behavior is similar to that of $`A_N(x_F,\pi ^0|s)`$.
At the first sight, these results, in particular those at large $`x_F`$, are rather surprising. This is because, as mentioned in section 4.2, the large $`x_F`$ region is dominated by the direct fusion process (a) $`(u_vd_v)^P+s_s^T\mathrm{\Lambda }`$, where, according to the wave function of $`\mathrm{\Lambda }`$, the $`(u_vd_v)^P`$ diquark from the projectile has to be in a spin-zero state. How can a spin zero object transfer the information of polarization to the produced $`\mathrm{\Lambda }`$? This question has been discussed in \[References\] and a solution has been suggested where associated production plays an important role. It has been pointed out<sup>?</sup> that the direct fusion process (a) should be predominately associated with the production of a meson directly formed through fusion of the remaining $`(u_v^a)^P`$ valence quark of the projectile with a suitable anti-sea-quark of the target. This $`(u_v^a)^P`$ carries the information of polarization of the projectile thus the associatively produced meson ($`K^+`$ or $`\pi ^+`$) which contains this $`(u_v^a)^P`$ should have a large probability to obtain an extra transverse momentum to the left if the projectile is upwards polarized. This is caused by the the intrinsic transverse motion of the $`u`$-valence quark and the surface effect. According to momentum conservation, the intrinsic transverse momentum of the valence quark should be approximately compensated by that of the other valence quarks. Hence, the left valence-diquark thus the $`\mathrm{\Lambda }`$ produced through the above mentioned process (a) should have a large probability to obtain an extra transverse momentum in the opposite direction as the associatively produced meson, i.e. to the right. This implies that $`(a)`$ contributes negatively to $`A_N`$, opposite to that of the associatively produced meson. Taking the contributions from the direct fusion processes (b) $`u_v^P+(d_ss_s)^T\mathrm{\Lambda }`$ and (c) $`d_v^P+(u_ss_s)^T\mathrm{\Lambda }`$ mentioned in section 4.2 and that from $`N_0`$ into account, we can obtain the $`A_N`$ for $`\mathrm{\Lambda }`$ in different $`x_F`$ region. In fact, as shown in \[References\], the above-mentioned differences and similarities between $`A_N(x_F,\mathrm{\Lambda }|s)`$ and $`A_N(x_F,\pi |s)`$ directly reflect the interplay between the different direct fusion processes and the non-direct-formation part: In the region $`x_F<0.40.5`$, (b), (c) and non-direct-formation part $`N_0`$ dominates. The situation is very much the same as that for $`\pi ^0`$, hence we obtain that $`A_N(x_F,\mathrm{\Lambda }|s)`$ in this region is similar to $`A_N(x_F,\pi ^0|s)`$. For $`x_F>0.40.5`$, (a) dominates thus $`A_N(x_F,\mathrm{\Lambda }|s)`$ is similar to $`A_N(x_F,\pi ^+|s)`$.
The $`x_F`$-dependence of $`A_N(x_F,\mathrm{\Lambda }|s)`$ has also been calculated<sup>?</sup>. Now, the difference $`\mathrm{\Delta }N(x_F,\mathrm{\Lambda }|s)N(x_F,\mathrm{\Lambda }|s,)N(x_F,\mathrm{\Lambda }|s,)`$ contains contributions from (a), (b) and (c). The contributions from (b) and (c) are simply proportional to $`\mathrm{\Delta }D_{b,c}(x_F,\mathrm{\Lambda }|s)D_{b,c}(x_F,\mathrm{\Lambda }|s,+)D_{b,c}(x_F,\mathrm{\Lambda }|s,)`$; where $`D_{b,c}(x_F,\mathrm{\Lambda }|s,\pm )`$ are the number densities for $`\mathrm{\Lambda }`$’s formed the corresponding direct processes of upwards $`(+)`$ or downwards $`()`$ polarized valence quark \[$`u_v`$ in case (b) and $`d_v`$ in case (c)\] with suitable sea-diquarks. The contribution from (a) is opposite in sign to that of the associatively produced meson and is proportional to $`r(x|u_v,tr)\mathrm{\Delta }u_v(x|tr)/u_v(x)`$, where $`x`$ is the fractional momentum of the $`u_v`$ valence quark. We have therefore,
$$\mathrm{\Delta }N(x_F,\mathrm{\Lambda }|s)=C\left[r(x|u_v,tr)D_a(x_F,\mathrm{\Lambda }|s)+\mathrm{\Delta }D_b(x_F,\mathrm{\Lambda }|s)+\mathrm{\Delta }D_c(x_F,\mathrm{\Lambda }|s)\right],$$
(77)
where $`D_a(x_F,\mathrm{\Lambda }|s)`$ is given by Eq.(47); $`\mathrm{\Delta }D_{b,c}(x_F,\mathrm{\Lambda }|s,\pm )`$ are given by,
$$\mathrm{\Delta }D_b(x_F,\mathrm{\Lambda }|s)=\kappa _\mathrm{\Lambda }\mathrm{\Delta }u_v(x^P|tr)f_D(x^T|d_ss_s),$$
(78)
$$\mathrm{\Delta }D_b(x_F,\mathrm{\Lambda }|s)=\kappa _\mathrm{\Lambda }\mathrm{\Delta }d_v(x^P|tr)f_D(x^T|u_ss_s).$$
(79)
$`A_N(x_F,\mathrm{\Lambda }|s)`$ can now be calculated using the ansatz for $`q_v^\pm (x|tr)`$ in Eqs.(63)-(64). In this case $`r(x|u_v,tr)=2/3`$ is a constant. The results of this calculation are compared with the data<sup>?</sup> in Fig. 11. We see that all the qualitative features of the data<sup>?</sup> are well reproduced.
Similar analysis can be and have been made for other hyperons in a straight-forward way. Qualitative results have been obtained in \[References\], which can be tested by future experiments.
### 4.6 Further tests and developments
At the end of this section, we would like to add a few comments concerning further tests and future developments and/or applications of this model.
From the discussions in the last subsections, we see that the model has an advantage that the picture is very clear and the model is simple. In fact, in order to demonstrate the role played by the key factors, many other effects which are not essential in understanding the left-right asymmetries observed in single-spin reactions are simplified and/or neglected. Semi-classical arguments are used in some places in order to make the picture as clear as possible. Thanks to this advantage, the model has a rather strong prediction power. In fact, as can be seen from the last few subsections, a number of predictions have been made in different connections. The model can therefore be tested easily. Till now, four of the predictions have already been confirmed by the experiments performed in the last years, many others can be tested by future experiments. Here, we would like to add another testable characteristics of the model, i.e., according to the model, $`A_N`$’s in different reactions should have the following in common:
($`\alpha `$) $`A_N`$ for hadron production is expected to be significant in and only in the fragmentation region of the polarized colliding object. It should be zero in the fragmentation region of the unpolarized one.
($`\beta `$) $`A_N`$ for hadron production in the fragmentation regions of the polarized colliding objects depends little on what kinds of unpolarized colliding objects are used.
($`\gamma `$) In contrast, $`A_N`$ for lepton-pair production depends not only on the polarized colliding object but also on the unpolarized one.
These common characteristics of $`A_N`$ for the production of different hadrons are consistent with the available data<sup>?,?</sup> and can be checked further by future experiments<sup>?,?</sup>.
Despite of its successes in describing the available data, the model has its limitations. At the end, it is a phenomenological model where the physical picture is based on arguments some of which are semi-classical. A field theoretical description is still lacking. This limits the prediction power and its applicability to other processes. In fact, there are many related questions which are still awaiting for answer. E.g.: Can we calculate the constant $`C`$? Does it depends on transverse momentum $`p_{}`$? How is the energy dependence of $`A_N`$? How is the transverse momentum dependence of $`A_N`$? These are questions which should be investigated. Furthermore, if it turns out that the model is right which implies that orbital motion of valence quarks and “surface effect” are important in describing spin effects in hadron-hadron collisions, is it possible, if yes, how to take them into account in the description of such reactions using perturbative QCD parton model?
## 5 Testing different models by performing further experiments
In the last two sections, we have discussed two different approaches. We have seen that both types of models have the potential to give non-vanishing values for the left-right asymmetries in high energy single-spin hadron-hadron collisions. It is then natural to ask: “Can we tell which approach is the more appropriate one for the description of the above mentioned data$`^{\text{References}\text{References}}`$?” The answer to this question is unfortunately “No!”. The reason is that each of these models are based on a set of special assumptions. In particular, to obtain significant asymmetries in Type One models we have three possibilities, (i), (ii) and (iii); but what has been measured$`^{\text{References}\text{References}}`$ is the convolution of all three factors. Hence it is difficult to find out which one is asymmetric simply through comparisons between the model and the existing data. Is it possible to have experiments with which the basic assumptions of these models can be tested individually ?
This question has been discussed by different authors in different occasions (See, in particular, \[References\] and the references given there.) It has been shown that the question should at least partly be answered in the affirmative for the following reasons:
First, if we can determine the direction of motion of the polarized quark before it fragments into hadrons and measure the left-right asymmetry of the produced hadrons w.r.t. this (known as the jet-) direction, we can directly see whether the products of the quark hadronization process are asymmetric w.r.t. this jet axis. Such measurements should be able to tell us whether the corresponding quark fragmentation function is asymmetric.
Second, there is no hadronization (i.e. no fragmentation) of the struck quarks in inclusive lepton-pair or $`W^\pm `$-production processes. Hence, measurements of the left-right asymmetries in such processes should yield useful information on the properties of the factors other than the quark-fragmentation function.
Third, while surface effect and/or “initial state interactions” may play a significant role in hadron-hadron collision processes, they do not exist in deep inelastic lepton-hadron scattering in large $`Q^2`$ and large $`x_B`$ region where the exchanged virtual photons are considered as “bare photons” <sup>?</sup>. Hence, comparisons between these two processes can yield useful information on the role played by surface effects.
Four experiments are listed in \[References\]. They are the following:
(A) Perform $`l+p()l+\pi +X`$ for large $`x_B`$ ($`>0.1`$, say) and large $`Q^2`$ ($`>10`$ GeV<sup>2</sup>, say) and measure the left-right asymmetry in the current fragmentation region w.r.t. the jet axis (See, also, \[References\]). Here, $`l`$ stands for charged lepton $`e^{}`$ or $`\mu ^{}`$; $`x_BQ^2/(2Pq)`$ is the usual Bjorken-$`x`$, $`Q^2q^2`$, and $`P,k,k^{},qkk^{}`$ are the four momenta of the proton, incoming lepton, outgoing lepton and the exchanged virtual photon respectively. The $`x_B`$ and $`Q^2`$ are chosen in the abovementioned kinematic region in order to be sure that the following is true: (1) The exchanged virtual photon can be treated as a bare photon<sup>?</sup>; (2) This bare photon will mainly be absorbed by a valence quark which has significant polarization in a transversely polarized proton.
In this reaction, a valence quark is knocked out by the virtual photon $`\gamma ^{}`$ and fragments into the hadrons observed in the current jet. The jet direction is approximately the moving direction of the struck quark before its hadronization; and the struck quark has a given probability to be polarized transversely to this jet axis The transverse momenta of the produced hadrons w.r.t. this axis come solely from the fragmentation of the quark. Hence, by measuring this transverse momentum distribution, we can directly find out whether the fragmentation function of this polarized quark is asymmetric.
(B) Perform the same kind of experiments as that mentioned in (A) and measure the left-right asymmetry of the produced pions in the current fragmentation region w.r.t. the photon direction in the rest frame of the proton, and examine those events where the lepton plane is perpendicular to the polarization axis of the proton. In such events, the obtained asymmetry should contain the contributions from the intrinsic transverse motion of quarks in the polarized proton and those from the fragmentation of polarized quarks, provided that they indeed exist. That is, we expect to see significant asymmetries, if and only if one of the abovementioned effects is indeed responsible for the asymmetries observed for pion production in single-spin hadron-hadron collisions.
(C) Perform the same kind of experiments as that in (A), but measure the left-right asymmetry in the target fragmentation region w.r.t. the moving direction of the proton in the collider (e.g. HERA) laboratory frame. By doing so, we are looking at the fragmentation products of “the rest of the proton” complementary to the struck quark (from the proton). Since there is no contribution from the elementary hard scattering processes and there is no hadronic surface effect, $`A_N`$ should be zero if the existence of left-right asymmetries is due to such effects. But, if such asymmetries originate from the fragmentation and/or from the intrinsic transverse motion of the quarks in the polarized proton, we should also be able to see them here.
(D) Measure the left-right asymmetry $`A_N`$ for $`l\overline{l}`$ and/or that for $`W^\pm `$ in $`p()+p(0)l\overline{l}\text{or }W^\pm +X`$. Here, if the observed $`A_N`$ for hadron production indeed originates from the quark fragmentation, we should see no left-right asymmetry in such processes. This is because there can be no contribution from the quark fragmentation here. Hence, non-zero values for $`A_N`$ in such processes can only originate from asymmetric quark distributions — including those due to orbital motion of valence quarks and surface effect.
The qualitative results are summarized in table 3. We emphasize here that the results of these experiments will not only shed light on the origin of the single-spin asymmetries for hadron production in hadron-hadron collisions but also yield extremely useful information on the widely used quark distribution functions in transversely polarized nucleon and the fragmentation functions. They should have strong impact on the study of spin distributions in nucleon and on the study of spin-dependent hadronic interactions.
## 6 Orbiting valence quarks and hyperon polarization in inclusive production processes at high energies
In a recent Letter<sup>?</sup>, it has been pointed out that there should be a close relation between the above discussed left-right asymmetries $`A_N`$ observed in single-spin hadron-hadron collisions and hyperon polarization $`P_H`$ observed$`^{\text{References}\text{References}}`$ in unpolarized hadron-hadron collisions. There exist a large number of experimental indications and theoretical arguments which show that these two spin phenomena may have the same dynamical origin. We recall that the striking hyperon polarization in inclusive production processes at high energies has been discovered<sup>?,?</sup> already in 1970s. There has been a lot of interest in studying the origin of this effect, both experimentally$`^{\text{References}\text{References}}`$ and theoretically$`^{\text{References}\text{References},\text{References}}`$. The pQCD parton model predicts zero polarization<sup>?</sup>. It is now a well-known experimental fact$`^{\text{References}\text{References}}`$ that hyperons produced in high energy hadron-hadron or hadron-nucleus collisions are polarized transversely to the production plane, although the projectiles and the targets are unpolarized. Experimental results$`^{\text{References}\text{References}}`$ have now been obtained for production of different hyperons in reactions using different projectile and/or different targets at different energies. But, theoretically, the origin still remains a puzzle. In fact, it has been considered as a standing challenge for the throretians to understand it for all these years. Hence, the result of \[References\] is rather interesting since if it turns out to be true, it should certainly provide some clue in the searching of the origin of $`P_H`$. We review the main ideas and results in the following. For the sake of simplicity, we use the model for $`A_N`$ discussed in section 4 as an example. We emphasize here that the discussion in \[References\] depends in fact little on the model. We use this model in the following since we think such an example may be helpful in understanding the underlying physics of the discussions in \[References\]. The readers are referred to the original paper for a more model independent discussion.
As has been mentioned above, there exist a large amount of data$`^{\text{References}\text{References}}`$ on hyperon polarization in unpolarized hadron-hadron collisions. These data show a number of striking characteristics very similar to those of $`A_N`$. In fact, we can simply replace $`A_N`$ by $`P_H`$ in (1) through (3) in section 2. These similarities already suggest that both phenomena have the same origin(s). Furthermore, we note: $`A_N0`$ implies that the direction of transverse motion of the produced hadron depends on the polarization of the projectile. $`P_H0`$ means that there exists a correlation between the direction of transverse motion of the produced hyperon and the polarization of this hyperon. That is, both phenomena show the existence of correlation between transverse motion and transverse polarization. Hence, unless we (for some reason) insist on assuming that the polarization of the produced hyperons observed in the projectile-fragmentation region is independent of the projectile — which would in particular contradict the empirical fact recently observed by E704 Collaboration<sup>?</sup> for $`\mathrm{\Lambda }`$ production — we are practically forced to accept the conclusion that $`A_N`$ and $`P_H`$ are closely related to each other.
This close relation has been studied in \[References\] by considering the following questions: Can we understand the existence of $`P_H`$ and reproduce the main characteristics of the data if we use the data of $`A_N`$ as input? Do we need further input(s)? Two different cases, i.e. production of hyperon which has only one (or two) valence quark(s) in common with the projectile, have been considered separately in \[References\]. In the following, we review these discussions using the model for $`A_N`$ as an example.
In the model discussed in section 4, $`A_N`$ comes from the orbital motion of the valence quarks and the surface effect in hadron-hadron collisions. More precisely, the $`A_N`$ data, both those for meson and those for $`\mathrm{\Lambda }`$ production, can be described using the following two points:
(I) Mesons $`(M)`$ and baryons $`(B)`$ produced through $`q_v^P+\overline{q}_s^TM`$ and $`q_v^P+(q_sq_s)^TB`$ have large probability to go left (w.r.t. the collision axis looking downstream) if $`q_v^P`$ is upwards polarized (w.r.t. the production plane).
(II) Baryons produced through $`(q_vq_v)^P+q_s^TB`$ are associated with $`(q_v^a)^P+\overline{q}_s^TM`$ and have large probability to move in the opposite transverse direction w.r.t. the collision axis as $`M`$ does.
We note that these two points are direct implications of the picture in section 4. They can also be considered as direct implications<sup>?</sup> of the available $`A_N`$ data for mesons and $`\mathrm{\Lambda }`$. We now use these two points as input to see if we can understand $`P_H`$. To do this, we encountered the following two questions:
(1) Will the polarization of the quarks be retained in fragmentation?
(2) What kind of picture for the spin structure of hadron shall we use in obtaining the polarization of the hadrons produced in fragmentation processes from the polarization of the quark(s) contained in the hadron: the one from the static quark model or that from polarized deep-inelastic lepton-nucleon scattering (DIS) data?
These questions have been discussed in \[References\]. We found out that $`\mathrm{\Lambda }`$ is an ideal place to study these questions: First, the polarization of $`\mathrm{\Lambda }`$ can easily be determined in experiments by measuring the angular distribution of its decay products. Second, $`\mathrm{\Lambda }`$ has a very particular spin structure in the static quark model, i.e., $`|\mathrm{\Lambda }^{}=(ud)_{0,0}s^{}`$ (here the subscripts of $`ud`$ denote its total spin and the third component), where the spin of $`\mathrm{\Lambda }`$ is completely carried by its $`s`$-valence quark. The results of the measurements<sup>?,?</sup> of longitudinal $`\mathrm{\Lambda }`$ polarization in $`e^+e^{}Z\mathrm{\Lambda }+X`$ at LEP provided much insight into answering these questions. Since, according to the standard model for electroweak interactions, $`s`$-quarks from $`Z`$ decay are longitudinally polarized before hadronization. This longitudinal polarization can be transferred to the produced $`\mathrm{\Lambda }`$, and the maximal value of longitudinal $`\mathrm{\Lambda }`$ polarization is expected if the following two conditions are true:
(1) Quark polarization is not destroyed in fragmentation.
(2) The SU(6) wavefunction from the static quark model is used to obtain the polarization of the produced hadron from that of the quark(s).
This maximal expectation has been estimated in \[References\]. The obtained result<sup>?</sup> is indeed much larger than that obtained<sup>?</sup> using the DIS picture, and is in good agreement with the ALEPH and OPAL data<sup>?,?</sup>. This strongly suggests that (1) and (2) are true.
We now take points (I), (II) and (1) and (2) as inputs and check whether we can understand the $`P_H`$-data based on these points. We consider the first case discussed in \[References\], i.e. the production of hyperons which have only one valence quark in common with the projectile. In this case, hyperons in the fragmentation region are dominated by those containing this common valence quark from the projectile and a sea diquark from the target, i.e., they are mainly from the type of direct fusion process mentioned in (I). We ask the question whether we can understand $`P_H`$ in this case if we assume (I) is true. It can easily be shown that this question should be answered in the affirmative: $`P_H`$ in this case can be determined uniquely by (I) and the wave function of the hyperon. To see this, we recall that $`P_H`$ is defined w.r.t. the production plane. Hence, we need only to consider e.g. those hyperons which are going left and check whether they are upwards (or downwards) polarized. (C.f. Fig.12.) According to (I), if the hyperon is going left, $`q_v^P`$ should have a large probability to be upwards polarized. This means, by choosing those hyperons which are going left, we obtain a subsample of hyperons which are formed by $`q_v^P`$’s that are upwards polarized with suitable sea diquarks. This information, together with the wave functions of the hyperons, determines whether the hyperon is polarized and, if yes, how large the polarization is. To demonstrate this, we consider $`p+p\mathrm{\Sigma }^{}+X`$. Here, the contributing direct fusion is $`d_v^P+(d_ss_s)^T\mathrm{\Sigma }^{}`$ and the wavefunction of $`\mathrm{\Sigma }^{}`$ is: $`|\mathrm{\Sigma }^{}=\frac{1}{2\sqrt{3}}[3d^{}(ds)_{0,0}+d^{}(ds)_{1,0}\sqrt{2}d^{}(ds)_{1,1}]`$, where the subscripts of the diquarks are their total angular momenta and the third components. We see that if $`d_v^P`$ is upwards polarized, $`\mathrm{\Sigma }^{}`$ has a probability of $`2/3`$ ($`1/6`$) to be upwards (downwards) polarized. Hence, we obtain that the $`\mathrm{\Sigma }^{}`$ from this direct formation process is positively polarized and the polarization is $`(2/3)C`$, \[where $`C`$ is the positive constant mentioned in last section which describes the probability for $`B`$ from $`q_v^P+(q_sq_s)^TB`$ to go left if $`q_v^P`$ is upwards polarized.\] Similar analysis can also be done for other hyperons and, e.g., we obtained that both $`\mathrm{\Xi }^{}`$ and $`\mathrm{\Xi }^0`$ produced in $`pp`$-collisions are negatively polarized and the polarization is $`C/3`$. These results show that $`P_\mathrm{\Sigma }^{}`$ is positive and its magnitude is large while $`P_\mathrm{\Xi }`$ is negative and its magnitude is smaller. Measurements of both $`P_\mathrm{\Sigma }^{}`$ and $`P_\mathrm{\Xi }`$ in $`pp`$-collisions have been carried out$`^{\text{References}\text{References}}`$. The results are consistent with the above mentioned expectations.
We see that these results follow directly from (I) together with (1) and (2) without any further input. There are also many other direct associations, e.g. the following:
(A) $`P_\mathrm{\Lambda }`$ in the beam fragmentation region of $`K^{}+p\mathrm{\Lambda }+X`$ is large and is, in contrast to that in $`pp`$-collisions, positive in sign. This is because, according to the wave function, $`|\mathrm{\Lambda }^{}=s^{}(ud)_{0,0}`$, the polarization of $`\mathrm{\Lambda }`$ is entirely determined by the $`s`$ quark. Here, the only contributing direct formation is $`s_v^P+(u_sd_s)^T\mathrm{\Lambda }`$ and, according to (I), $`s_v^P`$ should have large probability to be upwards polarized if $`\mathrm{\Lambda }`$ is going left.
(B) $`P_\mathrm{\Lambda }`$ in $`\pi ^\pm +p\mathrm{\Lambda }+X`$ should be negative and the magnitude should be very small in the $`\pi `$ fragmentation region. This is because, in this region, the only contributing direct formation process is $`u_v^P+(d_ss_s)^T\mathrm{\Lambda }`$ \[or $`d_v^P+(u_ss_s)^T\mathrm{\Lambda }`$\] and the $`\mathrm{\Lambda }`$ directly produced in this process is unpolarized. A small polarization is expected only from the decay of $`\mathrm{\Sigma }^0`$.
(C) Not only the produced hyperons but also the produced vector mesons are expected to be transversely polarized in the projectile fragmentation region of hadron-hadron or hadron-nucleus collisions. E.g., vector mesons such as $`\rho ^\pm ,\rho ^0,K^+`$ produced in $`p+p`$ collisions are expected to be positively polarized in the proton fragmentation region. This is because, if such mesons are going left, the valence quarks which combine with suitable anti-sea-quarks to form such mesons should have large probability to be upwards polarized. This upwards polarized valence quark combine with a suitable (unpolarized) anti-sea-quark and leads to a vector meson which has large probability to be upwards polarized.
(D) There should be no significant polarization transverse to the production plane for hyperons produced in $`e^+e^{}`$-annihilations. This is because, in such processes there is neither hadronic surface effect nor orbital motion of the initial state quarks.
(E) There should be no significant polarization transverse to the production plane for hyperons produced in deep inelastic lepton-hadron collision process in the large $`Q^2`$ and large $`x_B`$ region. This is because, in this kinematic region, the exchanged virtual photons are “bare photons”. They are point-like objects and hence there is no hadronic surface effect in the photon-proton reactions.
Such direct associations can be readily checked by experiments. There are already data for the processes mentioned in (A), (B) and (D) $`^{\text{References}\text{References}}`$, and all of them are in agreement with these associations. More precisely, we see the following: Experiments using $`K^{}`$ beams have been carried out and the results<sup>?</sup> show that $`P_\mathrm{\Lambda }`$ in the beam fragmentation region is indeed positive and large. There are also experiments using $`\pi `$ beam and the results show that $`P_\mathrm{\Lambda }`$ in beam fragmentation is indeed much smaller than that in $`p+p\mathrm{\Lambda }+X`$. $`\mathrm{\Lambda }`$-polarization in processes $`e^+e^{}`$ hadrons has also been studied by TASSO Collaboration at PETRA DESY<sup>?</sup> and ALEPH at LEP CERN<sup>?</sup>. The data from both Collaborations show that $`P_\mathrm{\Lambda }`$ is much less significant than that in $`p+A`$-collisions. This is consistent with (A). (C) and (E) can be checked by future experiments.<sup>*</sup><sup>*</sup>*We note here, since vector mesons such as $`\rho `$ and $`K^{}`$ decay into two hadrons via strong interactions, it is impossible to determine whether they are upwards or downwards polarized w.r.t. the production plane by measuring these decay products. This means to test (C) is rather academic according to the present technic of measuring the polarization of the produced particles. However, one can determine whether they are transversely or longitudinally polarized by measuring the angular distributions of these decay products. \[See, K. Schiling and G. Wolf, Nucl. Phys. B61, 381 (1973).\] This means that (C) can at least be tested partly using the technology we know presently.
These results are rather encouraging. Hence, we continue to consider the second case, i.e., the production of hyperon which has two valence quarks in common with the projectile. In this case hyperon containing a valence diquark of the projectile with a suitable seaquark of the target dominates the beam fragmentation region. The most well known process of this type is $`p+p\mathrm{\Lambda }+X`$. In this process, the dominating contribution in the fragmentation is from the direct fusion process (a) $`(u_vd_v)^P+s_s^T\mathrm{\Lambda }`$ mentioned in sections 4.2 and 4.5. This direct fusion process (a) is mainly associated with $`(u_v^a)^P+\overline{s}_s^TK^+`$; and, according to (II), if the $`\mathrm{\Lambda }`$ is going left, the associatively produced $`K^+`$ should have a large probability to go right. This implies that $`(u_v^a)^P`$ has a large probability to be downwards polarized. Since $`K`$ is a pseudoscalar meson and thus a spin-zero object, $`\overline{s}_s^T`$ should be upwards polarized. Hence, the corresponding $`s_s^T`$ should be downwards polarized, provided that the sea quark-anti-quark pair is not transversely polarized.
We see that, to explain the existence of $`P_\mathrm{\Lambda }`$ in this case, we need, besides (II), a further assumption, i.e., the sea quark-antiquark pair $`s\overline{s}`$ is not transversely polarized. As has been mentioned in \[References\] that this can be considered as a further implication of the existence of $`P_\mathrm{\Lambda }`$ for the structure of nucleon in the framework of the picture described in section 4. Whether it is indeed true can and should be checked by further experiments. It is therefore important to consider the direct consequences of this assumption and compare them with the available data. For this reason, the following have been made in \[References\].
First, a similar analysis has been made for the production of other hyperons of this type. Qualitative results for their $`P_H`$’s have been obtained and they are all consistent with the available data<sup>?</sup>.
Second, a quantitative estimation of $`P_\mathrm{\Lambda }`$ in $`p+p\mathrm{\Lambda }+X`$ as a function of $`x_F`$ has been made. Since all the different contributions \[i.e. those from the direct fusion (a), (b) or (c), or the non-direct formation part $`N_0`$\] and also the constant $`C`$ are known (see sections 4.2 and 4.5), this estimation can be made without any free parameter. The results<sup>?</sup> are compared with the data$`^{\text{References}\text{References}}`$ in Fig.13.We note that, in this estimation, only associated production of $`\mathrm{\Lambda }`$ and $`K^+`$ is considered while that of $`\mathrm{\Lambda }`$ and $`K^+`$ is neglected. It is clear that the above mentioned $`\mathrm{\Lambda }`$ polarization can be obtained in the former case but completely destroyed in the latter. It is also clear that associated production of $`\mathrm{\Lambda }`$ and $`K^+`$ contributes also significantly to $`pp\mathrm{\Lambda }X`$. Inclusion of this contribution should reduce $`|P_\mathrm{\Lambda }|`$ in the large $`x_F`$ region. In this sense, the curve in Fig.13 in the large $`x_F`$ region represents only the maximal expectation of $`|P_\mathrm{\Lambda }|`$ from the model. From the figure, we see also that $`|P_\mathrm{\Lambda }|`$ in this region is indeed higher than the data, which implies that there is room for including such influences. However, a detailed calculation in which all effects such as associated production of $`\mathrm{\Lambda }`$ and $`K^{}`$ are taken into account needs a rather detailed hadronization model which describes the production of different hadrons in the fragmentation region. Although there exist several fragmentation models on the market, it is unclear whether they can describe vector meson production, in particular vector to pseudoscalar ratio, in the fragmentation regions since no such study has been made yet. This implies that such a detailed calculation would involve rather high theoretical uncertainties. Hence, we choose not to do such a calculation but seek for further direct tests of the picture. See in particular point ($`\gamma `$) in the following and our discussion of $`P_\mathrm{\Lambda }`$ in $`pp\mathrm{\Lambda }K^+p`$ at the end of this section.
Third, a number of other consequences have been derived. The following are three examples which are closely related to the assumption that the $`s`$ and $`\overline{s}`$ which take part in the associated production are opposite in transverse spins.
($`\alpha `$) The polarization of the projectile and that of $`\mathrm{\Lambda }`$ in the fragmentation region of $`p+p\mathrm{\Lambda }+X`$ should be closely related to each other. In other words, the spin transfer $`D_{NN}`$ (which is defined as the probability for the produced $`\mathrm{\Lambda }`$ to be upwards polarized in the case that the projectile proton is upwards polarized) is expected to be positive and large for large $`x_F`$. It is true that the $`ud`$-diquark which forms the $`\mathrm{\Lambda }`$ is in a spin-zero state thus carries no information of polarization. But, according to the mechanism of associated production, the polarization of the left-over $`u_v^P`$ determines the polarization of the projectile and that of the $`s_s`$ quark which combines with the $`ud`$-diquark to form the $`\mathrm{\Lambda }`$. Hence, we expect to see a strong correlation between the polarization of the proton and that of the $`\mathrm{\Lambda }`$. A quantitative estimation of $`D_{NN}`$ as a function of $`x_F`$ is made. The result is shown in Fig.14.
($`\beta `$) $`P_\mathrm{\Lambda }`$ in the beam fragmentation region of $`\mathrm{\Sigma }^{}+A\mathrm{\Lambda }+X`$ should be negative and much less significant than that in $`p+p\mathrm{\Lambda }+X`$. Here, the dominating contributions are the $`\mathrm{\Lambda }`$’s which consist of $`(d_vs_v)^P`$ and $`u_s^T`$, $`d_v^P`$ and $`(u_ss_s)^T`$, or $`s_v^P`$ and $`(u_sd_s)^T`$. Exactly the same analysis as that mentioned above for $`p+p\mathrm{\Lambda }+X`$ show that the $`\mathrm{\Lambda }`$’s of the first two kinds are unpolarized; and those of the third kind are positively polarized. Since the first kind dominates in the large $`x_F`$ ($`x_F>0.6`$, say, this should be the same as that for $`p+p\mathrm{\Lambda }+X`$, see section 4.2), the second and the third dominate the middle $`x_F`$ ($`x_F0.30.4`$) region, we expect to see the following: If we exclude the contribution from $`\mathrm{\Sigma }^0`$ and $`\mathrm{\Sigma }^0`$ decay, $`P_\mathrm{\Lambda }`$ should be approximately zero for large $`x_F`$ and should be small but positive in the middle $`x_F`$ region. Taking $`\mathrm{\Sigma }^0`$ and $`\mathrm{\Sigma }^{}`$ decay into account, we expect a small negative $`P_\mathrm{\Lambda }`$ for large $`x_F`$.
($`\gamma `$) Hyperon polarization in processes in which a vector meson is associatively produced should be very much different from that in processes in which a pseudoscalar meson is associatively produced. E.g., $`P_\mathrm{\Lambda }`$ in the fragmentation region of $`p+p\mathrm{\Lambda }+K^++X`$ should be negative and its magnitude should be large, but $`P_\mathrm{\Lambda }`$ in the fragmentation region of $`p+p\mathrm{\Lambda }+K^++X`$ should be positive and its magnitude should be much smaller. This is because, in the latter case, using the same arguments as we used in the former case, we still obtain that $`(u_v^a)^P`$ (contained in $`K^+`$) has a large probability to be downwards polarized if $`\mathrm{\Lambda }`$ is going left. But, in contrast to the former case, the $`\overline{s}_s^T`$ here in the $`K^+`$ can be upwards or downwards polarized since $`K^+`$ is a spin-1 object. If $`\overline{s}_s^T`$ is upwards polarized, the produced meson can either be a $`K^{}`$ or a $`K`$, and the corresponding $`\mathrm{\Lambda }`$ should be downwards polarized. But if $`\overline{s}_s^T`$ is downwards polarized, the produced meson can only be a $`K^+`$ and the corresponding $`\mathrm{\Lambda }`$ should be upwards polarized, i.e. $`P_\mathrm{\Lambda }>0`$.
Presently, there are data available for the processes mentioned in ($`\alpha `$) and ($`\beta `$) <sup>?,?</sup>, and both of them are in agreement with the above expectations. More precisely, we have the following: E704 Collaboration has recently observed<sup>?</sup> a strong correlation between the polarization of the proton projectile and that for $`\mathrm{\Lambda }`$ in the fragmentation region. This seems, at the first sight, difficult to be understood since the $`ud`$-diquark which is common for $`\mathrm{\Lambda }`$ and proton should be in the spin zero state thus carries no information of polarization, but is a natural consequence \[see ($`\alpha `$)\] of the proposed picture. WA89 Collaboration has recently found out<sup>?</sup> that $`P_\mathrm{\Lambda }`$ in the beam fragmentation region in $`\mathrm{\Sigma }^{}+A\mathrm{\Lambda }+X`$ is negative in sign but indeed much less significant than that in $`p+A\mathrm{\Lambda }+X`$, which is consistent with ($`\beta `$). The prediction mentioned in ($`\gamma `$) is another characteristic property of the model and can be used as a crisp test of the picture.
We note that<sup>?</sup> a characteristic property of the proposed picture is that $`\mathrm{\Lambda }`$ polarization in $`pp\mathrm{\Lambda }X`$ comes predominately from those $`\mathrm{\Lambda }`$’s each of which contains a valence diquark $`(ud)_{0,0}`$ of the colliding proton and is associated with a spin-zero meson such as $`K^+`$ which contains the other valence $`u_v`$ of that proton. Hence, if these conditions are guaranteed in a particular channel for $`pp\mathrm{\Lambda }X`$, $`|P_\mathrm{\Lambda }|`$ should take its maximum in this channel and the maximal value is $`C`$. There exists indeed such a channel, i.e. $`pp\mathrm{\Lambda }K^+p`$. In fact this is the only channel in which the above-mentioned two conditions are guaranteed. $`\mathrm{\Lambda }`$ polarization in $`pp\mathrm{\Lambda }K^+p`$ has been studied<sup>?</sup> by R608 Collaboration at CERN. They obtained that $`P_\mathrm{\Lambda }=0.62\pm 0.04`$. This is in excellent agreement with the above expectation $`|P_\mathrm{\Lambda }|=C=0.6`$. This is strong support of the picture.
## 7 Probing dissociation of space-like photons in deep-inelastic lepton-nucleon scattering
As have been seen in the previous sections, although the origin is still in debate, it is now a well know fact that striking left-right asymmetry $`A_N`$ exists in single-spin hadron-hadron collisions and striking transverse polarization $`P_H`$ exists for hyperon production in unpolarized hadron-hadron collisions. It is therefore natural to think about using such striking spin effects as a tool to study the properties of the particles and/or of strong interactions. A well-known example for the experimentalists is the use of the existence of $`A_N`$ as a polarimeter which has been discussed in different occasions. Another example has been discussed in \[References\]. We make a brief introduction of the latter here.
It is known already for a long time that hadronic dissociation of space-like photons may play a significant role in deep-inelastic lepton-hadron scattering — especially in diffractive processes<sup>?,?</sup>. People seem to agree (for a list of references, see \[References\]) that, viewed from the hadron- or nucleus-target, not only real, but also space-like photons $`(Q^2q^2>0`$, where $`q`$ is the four-momentum of such a photon) may exhibit hadronic dissociation. In the small $`x_B`$ region such as that at HERA, the coherent length of such hadronic dissociation may be as long as $`10^3`$ fm, much much longer than the typical size of a hadron. What does this mean for the interaction of such virtual photon $`\gamma ^{}`$ with hadrons? Does this imply that, viewed from the hadron, such photon behaves always like a hadron so that we have always a hadron-hadron collision in inelastic lepton-nucleon scattering in the small $`x_B`$ region?
In the above sections, we summarized that a characteristic property of hadron-hadron collisions is the existence of left-right asymmetry $`A_N`$ in the fragmentation region in single-spin processes and the existence of hyperon polarization in the fragmentation of unpolarized collision processes. Although the origins of such striking spin effects are still in debate, both experimental results and theoretical arguments seem to suggest the following: there is no such spin effects in processes where the hadronic surface effect or “initial state interaction” does not exist. Hadronic surface or initial state interaction exists in any hadron-hadron collision but should not exist in $`\gamma ^{}`$-hadron collision if $`\gamma ^{}`$ can be treated as pointlike. Hence, we can use the existence of such spin effects as a sensor to test whether the virtual photon behaves like a hadron.
To be more precise, it has been proposed<sup>?</sup> that one can perform single-particle inclusive measurements in the proton fragmentation region at HERA for small-$`x_B`$ and different $`Q^2`$ in reactions using transversely polarized or unpolarized protons, and compare the results with those in the corresponding hadron-hadron collisions. Two extreme cases have been discussed<sup>?</sup>. In case one, $`\gamma ^{}`$ is assumed to behave always like a hadron in the same $`x_B`$ region independent of $`Q^2`$. In this case, one expects to see $`A_N`$ for $`\pi `$’s or $`K`$’s in reactions using transversely polarized protons, or transverse polarization $`P_\mathrm{\Lambda }`$ for $`\mathrm{\Lambda }`$ in reactions using unpolarized protons. The $`A_N`$’s and $`P_\mathrm{\Lambda }`$ should be the same as those observed in the corresponding hadron-hadron collisions and they should be independent of $`Q^2`$. In the second case, the virtual photon $`\gamma ^{}`$ is assumed to behave like a hadron only in the small $`Q^2`$ region, where vector meson dominance plays a role, but behaves like a pointlike object for large $`Q^2`$ independent of $`x_B`$. If this is true, one should see a strong $`Q^2`$-dependence for $`A_N`$ and $`P_\mathrm{\Lambda }`$. They should be approximately the same as those in the corresponding hadron-hadron collisions if $`Q^2`$ is small, but tends to zero for large $`Q^2`$. This has been demonstrated<sup>?</sup> in a quantitative manner by separating $`F_2^p(x_B,Q^2)`$-data<sup>?,?</sup> in the small-$`x_B`$ region into the well-known vector-dominance contribution (See e.g. \[References\] and the references cited there) from “the rest” which should be identified as the contribution from “pointlike” photon in this case. This separation is shown in Fig.15. In this way, the $`Q^2`$ dependence of $`A_N`$ and that of $`P_\mathrm{\Lambda }`$ have been obtained and are shown in Figs.16 and 17.
We see from the figures that the differences between the results obtained in the two cases are significantly large. We therefore expect that they can indeed be used as a good probe for the hadronic dissociation of the photon. From this example, we also explicitly see the following: Single-spin asymmetry study is interesting not only because the understanding of its origin can provide us a great deal of useful information on the structure of hadron and on the properties of hadronic interactions but also because it can be used as an useful tool to study the properties of different elementary particles and those of high energy reactions in general.
Acknowledgment
It is a great pleasure for both of us to express our sincere thanks to our common supervisor, Professor Meng Ta-chung, for bringing us to this interesting field, for the continuous guidance, the inspiring discussions and encouragements at every step of our works. We thank also R. Rittel for collaborations and discussions. Our sincere thank also goes to Professors M. Anselmino, D.H.E. Groß, K. Heller, S. Nurushev, J. Qiu and A. Yokosawa for helpful discussions in different occasions. This work was supported in part by National Natural Science Foundation of China (NSFC), by the State Education Commission of China, by the Australian Research Council and by Deutsche Forschungsgemeinschaft (DFG Me470-2). |
warning/0001/cond-mat0001395.html | ar5iv | text | # Numerical Study on Aging Dynamics in Ising Spin-Glass Models: Temperature-Change Protocols
## 1 Introduction
Aging phenomena in spin glasses have been extensively studied in recent years. Experimentally, many elaborated protocols have been adopted to reveal various aspects of the phenomena. $`^{\text{?}\text{}\text{?}\text{}\text{?}\text{)}}`$ Theoretically, on the other hand, many ideas and models on spin glasses and related systems have been proposed and examined to get a proper understanding of aging dynamics in such glassy systems. $`^{\text{?}\text{}\text{?}\text{)}}`$ On the side of numerical study, a number of simulations, which could account for qualitative features of the aging effects, have been performed. $`^{\text{?}\text{}\text{?}\text{}\text{?}\text{)}}`$ In spite of the enormous efforts, however, a unified picture of aging dynamics in spin glasses has not been established yet. The difficulty is directly connected to nature of the low-temperature spin-glass (SG) phase which has been a controversial issue since more than a decade ago.
Among simulational results on the short-ranged Ising SG models, of particular importance is the growth of a mean domain size which is observed through the replica-overlap function, $`G(r,t)`$, in an isothermal aging process. $`^{\text{?}\text{}\text{?}\text{}\text{?}\text{}\text{?}\text{)}}`$ In the process a system is quenched instantaneously from temperature above its SG transition temperature $`T_\mathrm{c}`$ to $`T`$ below $`T_\mathrm{c}`$, and is isothermally aged at $`T`$. The function $`G(r,t)`$ is defined as
$$G(r,t)=\frac{1}{N}\underset{i=1}{\overset{N}{}}\left[S_i^{(\alpha )}(t)S_i^{(\beta )}(t)S_{i+r}^{(\alpha )}(t)S_{i+r}^{(\beta )}(t)\right]_{\mathrm{av}},$$
(1.1)
where $`\alpha `$ and $`\beta `$ are indices of two replicas which have different random spin configurations at $`t=0`$ (an instant of the quench), and are independently updated. Its correlation length $`R_T(t)`$ is found to grow almost in a power-law in the case of 3 dimensional (3D) models, $`^{\text{?, ?, ?)}}`$
$$R_T(t)L_0(t/\tau _0)^{1/z(T)},$$
(1.2)
where $`L_0`$ and $`\tau _0`$ are certain characteristic units of length and time, respectively, and the exponent $`1/z(T)`$ depends on $`T`$. In our previous work, $`^{\text{?)}}`$ hereafter referred to as I, we obtained
$$1/z(T)0.17T,$$
(1.3)
at $`T<0.7J`$ for the Gaussian model with zero mean and variance of interactions $`J`$. It is also fitted to a logarithmic functional form written as $`R_T(t)R_0+b(\mathrm{ln}t)^{1/\psi }`$ with $`R_0,b`$ and $`\psi `$ being constants. $`^{\text{?, ?, }\text{?}\text{)}}`$ This ambiguity is attributed to the insufficiency of the time window of simulations which can be performed by the presently available computers. Although there remains such an ambiguity in the explicit growth form of $`R_T(t)`$, we can naturally regard $`R_T(t)`$ as a characteristic length scale of the SG ordering in aging dynamics, i.e., a mean distance of domain walls separating different pure states.
It is natural to ask to what extent the mean domain size $`R_T(t)`$ is related to aging behaviour observed in various quantities simulated. The droplet theory, $`^{\text{?}\text{}\text{?}\text{}\text{?}\text{}\text{?}\text{)}}`$ particularly the one proposed by Fisher and Huse, $`^{\text{?)}}`$ contains scaling ansatz which relate various quantities to the characteristic length scales such as $`R_T(t)`$. We have been numerically studying aging phenomena in the 3D Ising SG model to clarify the validity of the ansatz. $`^{\text{?, ?, }\text{?}\text{)}}`$
Experimentally, various interesting phenomena have been also observed in aging processes in addition to the isothermal one. They are $`T`$-shift, $`T`$-cycling, and continuous $`T`$-change with an intermittent stop(s) protocols whose schedules of $`T`$-change are schematically shown in Fig. 1. The ac susceptibility has been continuously measured through these $`T`$-changes. In the case of a negative $`T`$-shift process with $`\mathrm{\Delta }TT_2T_1<0`$ in Fig. 1 and with a magnitude $`|\mathrm{\Delta }T|`$ larger than a certain value, the ac susceptibility is observed to increase discontinuously just after $`t=t_{\mathrm{w1}}`$ and then to relax again. $`^{\text{?}\text{}\text{?}\text{}\text{?}\text{)}}`$ Its behaviour looks as if the system restarts aging by the $`T`$-shift at $`t=t_{\mathrm{w1}}`$ forgetting its thermal history before $`t_{\mathrm{w1}}`$. It is called the rejuvenation, or chaos effect. Furthermore when the temperature is turned back to $`T_1`$ at $`t=t_{\mathrm{w1}}+t_{\mathrm{w2}}`$ ($`T`$-cycling protocol), the value of the ac susceptibility returns to that on the isothermal aging curve of $`T_1`$ at almost $`t=t_{\mathrm{w1}}`$, and then relaxes along the curve. This behaviour indicates that the system in fact remembers its history before $`t_{\mathrm{w1}}`$, and is called the memory effect. Similar coexistence of the paradoxical phenomena, rejuvenation and memory effect, has been recently observed also in a continuous $`T`$-change with an intermittent stop(s). $`^{\text{?}\text{}\text{?}\text{)}}`$ These experimental findings have stimulated theoretical interest. $`^{\text{?)}}`$
The main purpose of the present paper is to investigate aging phenomena in various $`T`$-change protocols by simulations on the 3D Gaussian Ising SG model. $`^{\text{?}\text{)}}`$ We concentrated mostly on the $`T`$-shift process, which is the most fundamental among the various $`T`$-change protocols, since the latter are considered as certain combinations of the former. We investigated the time evolution of the mean domain size and the relaxation of the energy density and the spin auto-correlation function during the $`T`$-shift process.
We have found, within the time window of our simulations, the following characteristic features of the aging dynamics. i) The mean domain size $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ in the $`T`$-shift process (see Fig. 1), with either a negative or positive $`\mathrm{\Delta }T`$, does continue to grow. Here and hereafter $`t`$ is a time measured from the first quench from above $`T_\mathrm{c}`$ to $`T_1`$, $`t_2=tt_{\mathrm{w1}}`$ a period that the system ages at $`T=T_2`$ after the $`T`$-shift at $`t=t_{\mathrm{w1}}`$. ii) The time evolution of the energy density and the spin auto-correlation function is related with the mean domain size $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ at sufficiently long time scales after the $`T`$-shift by the scaling relations which are the same as isothermal aging. iii) However, there is a transient time regime after the $`T`$-shift where there is extra contribution to the relaxations. Interestingly, the extra contribution also shows very slow, non-exponential relaxation. We interpret the latter as not due to the chaos effect predicted by the droplet theory $`^{\text{?, ?)}}`$ but as due to adjustment of the population of droplet excitations within each domain.
The memory effect observed experimentally is easily interpreted from the above-mentioned scenario, namely, it is attributed to the persistence of domains in a $`T`$-shift process. It is not yet conclusive, however, whether nature of the transient regime in the scenario is common to phenomena observed experimentally just after a $`T`$-shift. In this respect, the recent work on aging in the ferromagnetic fine particles system (FFPS) done by Mamiya et al. $`^{\text{?}\text{)}}`$ is of quite interest, since the time window of their observation in unit of the microscopic time ($`10^3`$sec for this system $`^{\text{?}\text{)}}`$) is rather closer to that of our simulations. Indeed, they have observed $`T`$-shift aging phenomena which can be well interpreted by our scenario. It seems, however, that the scenario is hard to explain the rejuvenation, or chaos effect observed experimentally in ordinary spin glasses, $`^{\text{?, ?, }\text{?}\text{}\text{?}\text{)}}`$ whose microscopic time is of the order of $`10^{13}`$sec.
The present paper is organized as follows. In the next section we briefly review the scaling properties of isothermal aging derived from the droplet theory, add further comments on them, and present the scenario on the $`T`$-shift process. In §3 we present the results of our simulations on various aspects of the $`T`$-shift, as well as the $`T`$-cycling and continuous $`T`$-change processes. In the final section implications of the results with the experimental observations are discussed.
## 2 Phenomenological Picture
### 2.1 Scaling properties of isothermal aging
Here we briefly review the scaling properties of isothermal aging derived by the droplet theory $`^{\text{?, ?)}}`$ due to Fisher and Huse (FH). In an isothermal aging process a system is quenched from above $`T_\mathrm{c}`$ to temperature $`T`$ below $`T_\mathrm{c}`$. At waiting time $`t_\mathrm{w}`$ after the quench, there are domain walls separating different pure states of the SG phase at a typical distance $`R_T(t_\mathrm{w})`$ from each other. The distance $`R_T(t_\mathrm{w})`$ can be measured in simulations for instance through the replica overlap function defined in eq.(1.1).
Except for places where the domain walls run, the bulk of the system is essentially in equilibrium. An important subtle feature inside the domains of a spin glass is that there can be large scale thermal fluctuations due to droplet excitations. In ideal equilibrium, a droplet excitation can be defined as a global flip from a ground state of a droplet (cluster) of spins within a distance $`L/2`$ from a certain given site $`i`$$`^{\text{?)}}`$ The latter can be considered as a simple two-state system with a free-energy excitation gap $`F_L(i)`$ and a barrier energy $`B_L(i)`$ of a thermal activation process.
The typical value $`F_L^{\mathrm{typ}}`$ of the gap $`F_L(i)`$ scales as,
$$F_L^{\mathrm{typ}}\mathrm{{\rm Y}}(L/L_0)^\theta ,$$
(2.1)
where $`\mathrm{{\rm Y}}`$ is the stiffness constant of domain walls on the boundary of the droplet. The gap is however broadly distributed and its distribution function $`\rho _L(F)`$ is considered to follow the scaling form,
$$\rho _L(F)=\frac{1}{F_L^{\mathrm{typ}}}\stackrel{~}{\rho }\left(\frac{F}{F_L^{\mathrm{typ}}}\right).$$
(2.2)
A very important property of the scaling function $`\stackrel{~}{\rho }(x)`$ is that it has finite intensity at $`x=0`$, $`\stackrel{~}{\rho }(0)>0`$, which allows large scale droplet excitations even at very low temperatures. The probability to have a thermally active droplet at a very low temperature $`T`$ is given by
$$p(L;T)\frac{k_\mathrm{B}T}{\mathrm{{\rm Y}}(L/L_0)^\theta }\stackrel{~}{\rho }(0).$$
(2.3)
The latter feature yields equilibrium properties which make spin glasses distinctly different from simple ferromagnets. $`^{\text{?)}}`$
The relaxation time of a droplet of size $`L`$ centered at site $`i`$ is given by the Arrhenius law
$$\tau _L(i)=\tau _0\mathrm{exp}\left(\frac{B_L(i)}{k_\mathrm{B}T}\right),$$
(2.4)
where $`B_L(i)`$ is the energy barrier of the droplet. The typical value $`B_L^{\mathrm{typ}}`$ of the energy barrier $`B_L(i)`$ scales as,
$$B_L^{\mathrm{typ}}\mathrm{\Delta }(L/L_0)^\psi ,$$
(2.5)
where $`\mathrm{\Delta }`$ is a characteristic unit for energy barriers associated with thermal activation of droplet excitations, and the exponent $`\psi `$ satisfies $`\theta \psi d1`$.
In the droplet theory, it is assumed that the growth of domains is also governed by droplet excitations. An obvious but very important consequence of eqs.(2.4) and (2.5) is that dynamical processes, including both the domain growth and droplet excitations, at different length scales are extremely widely separated. $`^{\text{?)}}`$ Hence within a time scale of small scale processes, large scale processes look as if they are almost frozen.
In isothermal aging, for example, the relaxation of the excessive energy per spin $`\delta e_T(t)`$ with respect to the equilibrium value is expected to follow the scaling form as,
$$\delta e_T(t)\stackrel{~}{\mathrm{{\rm Y}}}(R_T(t)/L_0)^\theta /(R_T(t)/L_0)^d,$$
(2.6)
where $`\stackrel{~}{\mathrm{{\rm Y}}}`$ is a characteristic energy scale, and $`d`$ is the dimension of space which is 3 here. This scaling form was confirmed, in our previous work I, with the energy exponent $`\theta =0.20\pm 0.03`$ which agrees with the result of the defect energy analysis at $`T=0`$$`^{\text{?}\text{)}}`$
The droplet theory also predicts important scaling properties of the spin auto-correlation function,
$$C_T(\tau ;t_\mathrm{w})=\overline{S_i(\tau +t_\mathrm{w})S_i(t_\mathrm{w})},$$
(2.7)
in isothermal aging. Here the over-line denotes the averages over sites and different realizations of interactions (samples), and the bracket denotes the average over thermal noises. In the so-called quasi-equilibrium regime defined by $`\tau t_\mathrm{w}`$, the typical size $`L_T(\tau )`$ of droplet excitations which take place in the time scale of $`\tau `$ is much smaller than the typical separation $`R_T(t_\mathrm{w})`$ of domain walls present after the waiting time $`t_\mathrm{w}`$. For this situation the droplet theory introduces a phenomenological concept called effective stiffness which characterizes a reduction of the excitation gap of small scale droplets due to the presence of the domain walls. The latter yields $`t_\mathrm{w}`$-dependence of $`C_T(\tau ;t_\mathrm{w})`$ which is written in terms of $`R_T(t_\mathrm{w})`$ and $`L_T(\tau )`$ as the following:
$$1C_T(\tau ;t_\mathrm{w})=\mathrm{\Delta }C_T(\tau ;t_\mathrm{w})+\alpha _T(\tau ),$$
(2.8)
with
$$\alpha _T(\tau )=1C_T^{(\mathrm{eq})}(\tau ),$$
(2.9)
and
$$\mathrm{\Delta }C_T(\tau ;t_\mathrm{w})\stackrel{~}{\rho }(0)\frac{T}{\mathrm{{\rm Y}}}\left(\frac{L_0}{L_T(\tau )}\right)^\theta \left(\frac{L_T(\tau )}{R_T(t_\mathrm{w})}\right)^{d\theta }.$$
(2.10)
Then the zero field cooled susceptibility $`\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})`$ which is often observed by experiments should also be understood within the same ansatz since the fluctuation-dissipation theorem (FDT)
$$\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})\frac{1}{T}[1C_T(\tau ;t_\mathrm{w})]\stackrel{~}{\chi }_T(\tau ;t_\mathrm{w}).$$
(2.11)
holds in the quasi-equilibrium regime. The FDT has been tested and confirmed by the careful experiments $`^{\text{?}\text{}\text{?}\text{)}}`$ even in the presence of weak violation of the time translational invariance.
### 2.2 Further comments on isothermal aging
In our previous work, $`^{\text{?)}}`$ hereafter referred to as II, we confirmed the scaling ansatz on $`C_T(\tau ;t_\mathrm{w})`$ described in the previous subsection. Here, we first make the comment that $`\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})`$ experimentally observed and analyzed through eq.(2.11) follow the scaling ansatz.
In Fig. 2, for example, we plot $`\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})`$ of a Ag:Mn 2.6at% spin glass against $`t_\mathrm{w}/\tau `$. It is evaluated from $`m_{\mathrm{TRM};T}(\tau ;t_\mathrm{w})`$ measured by the Saclay group $`^{\text{?)}}`$ as $`\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})=[m_{\mathrm{FC};T}m_{\mathrm{TRM};T}(\tau ;t_\mathrm{w})]/h`$ with $`m_{\mathrm{FC};T}`$ being the field-cooled magnetization under a field of strength $`h`$. We note that we plot here only a part of their data with $`t_\mathrm{w}>\tau `$, i.e., those in the quasi-equilibrium regime.
The data for different $`\tau `$’s lie on a universal curve when they are vertically shifted by suitable amounts depending on $`\tau `$ as shown in the inset of the figure. The feature is quite similar to $`1C_T(\tau ;t_\mathrm{w})`$ analyzed in II. \[Note that the factor $`(L_0/L_T(\tau ))^\theta `$ is practically constant in eq.(2.10) since $`\theta `$ is much smaller than $`d\theta `$, as was the case of $`1C_T(\tau ;t_\mathrm{w})`$ in II.\] This indicates that $`\chi _{\mathrm{ZFC};T}(\tau ;t_\mathrm{w})`$, via eq.(2.11), follows the scaling forms eqs.(2.8), (2.9) and (2.10) as well. The values of the exponents $`(d\theta )/z(T)`$ and $`\theta /z(T)`$ turn out to roughly agree with the estimates obtained by making use of our previous results in II, $`\theta 0.20`$ and $`1/z(T)0.16(T/T_\mathrm{c})`$, i.e., eq.(1.3) with $`T_\mathrm{c}=0.95\pm 0.04`$$`^{\text{?}\text{)}}`$ However, such semi-quantitative agreement is lost for the data at temperatures closer to $`T_\mathrm{c}`$.
An intriguing point is that the above result suggests that $`R_T(t)`$ in the time scale of experiments is also compatible with the estimate from our simulations. In this context it is worth pointing out that Joh et al. $`^{\text{?}\text{)}}`$ have recently extracted $`R_T(t)`$ of Cu:Mn 6at% and CdCr<sub>1.7</sub>In<sub>0.3</sub>S<sub>4</sub> spin glasses from the thermoremanent magnetization (TRM). The growth law they have obtained turns out to agree with the simulational result of eqs.(1.2) and (1.3) even quantitatively. The estimated $`R_T(t)`$ with $`t`$ of the order of laboratory time is at most only several tens of lattice distances. However, the experimental studies on the isothermal aging dynamics so far reported have not converged to a unified picture yet as we discussed in II.
Another comment is related to the characteristic length and time scales, $`L_0`$ and $`\tau _0`$ in eq.(1.2), which we did not fix explicitly in II. According to the droplet theory ,$`^{\text{?)}}`$ they are some microscopic scales at lower temperatures, but near below $`T_\mathrm{c}`$ they are expected to crossover to the SG coherence length $`\xi _{}`$ and the corresponding critical relaxation time, respectively. Indeed, Hukushima et al. $`^{\text{?}\text{)}}`$ have found recently such a crossover behaviour in aging dynamics in the 4D $`\pm J`$ Ising SG model. A similar analysis on the present 3D model certainly has to be done, which may clarify the disagreement between the experimental and simulational results close to $`T_\mathrm{c}`$ mentioned just above. Furthermore, as suggested in Ref. 34, it may also provide a useful key to resolve the apparent ambiguity of the growth law mentioned previously.
### 2.3 $`T`$-shift process within overlap-length
Now let us consider the $`T`$-shift process. After a system is first quenched from above $`T_\mathrm{c}`$ to a temperature $`T_1`$ below $`T_\mathrm{c}`$, domains of different pure states grow up. Their mean size reaches to $`R_{T_1}(t_{\mathrm{w1}})`$ after a given waiting time $`t=t_{\mathrm{w1}}`$. After the temperature is changed at $`t=t_{\mathrm{w1}}`$ to a new temperature $`T_2`$, we expect the domains continue to grow but with a rate specific to the new temperature $`T_2`$ (see eq.(3.1) below). It can be said that, at the instant of the $`T`$-shift, the system is aged as if it is aged for the effective waiting time $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ by isothermal aging at $`T_2`$, where $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ is determined by
$$R_{T_1}(t_{\mathrm{w1}})R_{T_2}(t_{\mathrm{w1}}^{(\mathrm{eff})}),$$
(2.12)
with $`R_T(t)`$ being the growth law of the mean domain size in isothermal aging at $`T`$.
Another important process which must take place after the $`T`$-shift is adjustment of the population of thermally active droplets within the domain. Since the equilibrium population is proportional to $`T`$ as given in eq.(2.3), there must be a change of the population of an amount proportional to $`\mathrm{\Delta }T=T_2T_1`$ at each length scale $`L`$. It is emphasized that this process itself is intrinsically very slow since adjustment of the droplets takes place only by activation processes described by eqs.(2.4) and (2.5). Thus we naturally expect that there is a new characteristic length $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ which slowly grows after the $`T`$-shift such that the population of droplets smaller than $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ is equilibrated to the new temperature $`T_2`$. However, the population of droplets larger than $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ is still equilibrated with respect to the old temperature $`T_1`$. We call the characteristic length $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ a size of quasi-domains.
The growth law of $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ is expected to be the same as the domain. When the system ages at $`T_2`$ for a period about $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ defined by eq.(2.12), $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ catches up $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ and the quasi-domains merge into domains in isothermal aging at $`T_2`$. Thus at large time scales $`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$, aging becomes essentially the same as isothermal aging at $`T_2`$. The time range $`t_2<t_{\mathrm{w1}}^{(\mathrm{eff})}`$, on the other hand, is regarded as a transient regime where adjustment of the population of droplets are taking place.
In the above argument we implicitly assumed that equilibrium states at different temperatures are the same except for the change of the population of droplets eq.(2.3). It should be remarked that the assumption makes sense only for length scales smaller than the so-called overlap length $`l_{\mathrm{\Delta }T}`$$`^{\text{?, }\text{?}\text{)}}`$ It is predicted that equilibrium states at different nearby temperatures $`T_1`$ and $`T_2`$ differ from each other at length scales larger than this length $`l_{\mathrm{\Delta }T}`$. This aspect is called chaotic nature of the SG phase. $`^{\text{?)}}`$ The overlap length $`l_{\mathrm{\Delta }T}`$ is expected to diverge when $`\mathrm{\Delta }T=T_2T_10`$ as
$$l_{\mathrm{\Delta }T}L_0\left(\frac{\mathrm{{\rm Y}}^{3/2}}{T^{1/2}|\mathrm{\Delta }T|}\right)^{2/(d_s2\theta )},$$
(2.13)
where, $`\theta `$ is the exponent in eq.(2.1) and $`d_s`$ the fractal dimension of surface of the droplets.
## 3 Results of Simulations
### 3.1 Model and method
We have carried out MC simulations on various aging processes in the same 3D Ising SG model as in our previous works, i.e., the one with Gaussian nearest-neighbor interactions with mean zero and variance $`J=1`$. The critical temperature has been obtained as $`T_\mathrm{c}=0.95\pm 0.04`$. .$`^{\text{?)}}`$ The heat-bath MC method we use here is also the same as in I and II. The data we will discuss below are obtained mostly at $`T=0.60.8`$ in systems with linear size $`L_\mathrm{s}=24`$ averaged over a few hundreds samples with one MC run for each sample. In I, it was confirmed that finite size effects do not appear for these values of $`T`$ and $`L_\mathrm{s}`$ within our time window ($`<2\times 10^5`$ MCS).
### 3.2 SG coherence length — domain growth
Let us begin with time evolution of $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$, a mean domain size in a negative $`T`$-shift protocol with $`\mathrm{\Delta }T=T_2T_1<0`$ (see Fig. 1). The simulated results with $`T_1=0.8,T_2=0.6`$ and $`t_{\mathrm{w1}}=1000`$ MCS are presented by solid circles in Fig. 3(a). They are extracted from the replica overlap function eq.(1.1) which is obtained by averaging over several thousands samples with $`L_s=32`$.
Clearly, $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ does grow continuously without any noticeable decrease by the $`T`$-shift. This implies that, just after the $`T`$-shift, the system looks as if it has aged isothermally at $`T=T_2`$ for a period of the effective waiting time $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ defined in eq.(2.12). Indeed, when $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ at $`t>t_{\mathrm{w1}}`$ is plotted against $`\stackrel{~}{t}=t_{\mathrm{w1}}^{(\mathrm{eff})}+t_2`$ (open triangles in the figure), it almost lies on the isothermal aging curve $`R_{T_2}(t)`$, i.e., the two curves are related as
$$R_{[T_2,T_1]}(t;t_{\mathrm{w1}})R_{T_2}(t_2+t_{\mathrm{w1}}^{(\mathrm{eff})}),\mathrm{at}t_2>0.$$
(3.1)
Here, a value of $`t_{\mathrm{w1}}^{(\mathrm{eff})}(6230)`$ has been chosen so as to obtain a good collapse of the two curves at time range $`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$. The chosen value turns out to satisfy eq.(2.12) combined with eqs.(1.2) and (1.3).
As shown in Fig. 3(b), $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ in a positive $`T`$-shift protocol is also described by eq.(3.1), but with $`t_{\mathrm{w1}}^{(\mathrm{eff})}(1400)<t_{\mathrm{w1}}(=10^4)`$ in this case. The continuous growth of domains in the $`T`$-shift process, either a negative or positive one, is one of the most important results observed in the present simulation.
### 3.3 Evolution of energy — Transient and isothermal regimes
In Fig. 4(a) we present the time evolution of energy per spin, $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$, in the same negative $`T`$-shift process as shown in Fig. 3(a). The original data drawn by the solid circles quickly deviate from the isothermal aging curve, $`e_{T_1}(t)`$, just after the $`T`$-shift at $`t=t_{\mathrm{w1}}`$, and then slowly crossover to $`e_{T_2}(t)`$ from below at longer times. As drawn by the open triangles in the figure, the part of $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ after the crossover can be laid upon $`e_{T_2}(t)`$ if it is shifted rightwards properly, i.e., the two curves are related as
$$e_{[T_2,T_1]}(t;t_{\mathrm{w1}})e_{T_2}(t_2+t_{\mathrm{w1}}^{(\mathrm{eff})})\mathrm{at}t_2>t_{\mathrm{w1}}^{(\mathrm{eff})},$$
(3.2)
with the same $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ as the one determined before through $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$.
Notice that the shifted $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ approaches $`e_{T_2}(t)`$ from above.
In the positive $`T`$-shift process, $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ behaves similarly and eq.(3.2) also holds as shown in Fig. 4(b). In this case the original (shifted) $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ merges into $`e_{T_2}(t)`$ from above (below).
From inspection of Figs. 3 and 4 we may divide the aging process after the $`T`$-shift into the following three regimes.
1. Quasi-reversal regime ($`t_2t_{\mathrm{rev}}1`$)
Each spin responds individually to changes in its local Boltzmann weights. This gives rise to a relatively large change in $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ in the first MC step after the $`T`$-shift. The process is expected almost reversible, though we have not explicitly confirmed it.
2. Transient regime ($`t_{\mathrm{rev}}<t_2<t_{\mathrm{w1}}^{(\mathrm{eff})}`$)
Judging from the behaviour of $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$, eq.(3.1), we may regard that the system has aged roughly for $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ at $`T=T_2`$ already at around $`t_2t_{\mathrm{rev}}`$. Actually, $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ is significantly smaller (larger) than $`e_{T_2}(t)`$ at $`t>t_{\mathrm{w1}}`$ in the negative (positive) $`T`$-shift process.
Another interesting comparison is between $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ as a function of $`t_2`$ and $`e_{T_2}(tt_{\mathrm{w1}}^{(\mathrm{eff})})`$ which is shown in the insets of Fig. 4. Now the former is larger (smaller) than the latter in the negative (positive) $`T`$-shift process, and it relaxes to the latter slowly, by no means exponentially. Within our scenario introduced in §2.3, these can be explained qualitatively as follows. Just after the negative (positive) $`T`$-shift, the population of active droplets within the domain is larger (smaller) than in equilibrium at $`T_2`$. Thus the energy density decreases (increases) toward the isothermal curves.
3. Isothermal regime ($`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$)
In this regime time evolution of the system is nothing but isothermal aging at $`T=T_2`$. The result eq.(3.2) combined with eq.(3.1) means that the mean domain size and the energy density are related by the scaling relation eq.(2.6). The latter implies that relaxation of the energy density in this regime is due to the coarsening of domain walls.
The existence of isothermal regime 3), to which the system crossovers from regime 2) at around $`t_2t_{\mathrm{w1}}^{(\mathrm{eff})}`$, is another important aspect of the aging dynamics observed by the present simulation. It is certainly a consequence of the persistence of domains through the $`T`$-shift process described before, and can be regarded as one of the memory effects. An interesting question is if the phenomena in transient regime 2) are related with the rejuvenation, or chaos effect observed experimentally.
### 3.4 Spin auto-correlation function
We define the spin auto-correlation function in the $`T`$-shift process as
$$C_{[T_2,T_1]}(\tau ;t_{\mathrm{w2}},t_{\mathrm{w1}})=\overline{S_i(\tau +t_{\mathrm{w1}}+t_{\mathrm{w2}})S_i(t_{\mathrm{w1}}+t_{\mathrm{w2}})}.$$
(3.3)
and measure relaxation of
$$\stackrel{~}{\chi }_{[T_2,T_1]}(\tau ;t;t_{\mathrm{w1}})\frac{1}{T}[1C_{[T_2,T_1]}(\tau ;t_2,t_{\mathrm{w1}})]$$
(3.4)
with increasing $`t_2`$ after the $`T`$-shift. At each measurement, a fixed value of the time separation $`\tau `$ is chosen as a parameter.
We naively expect that the FDT holds for time regime $`\tau <t_{\mathrm{w2}}`$ as in the case of isothermal aging (see eq.(2.11)). If it is the case the above $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau ;t;t_{\mathrm{w1}})`$ is identical to the ZFC susceptibility. In experiments, relaxation of the out-of-phase component of the ac susceptibility with a fixed frequency, say $`\omega `$, is most frequently measured also in the $`T`$-shift protocol. It has been numerically ascertained in II that the relaxation of the ac susceptibility at a fixed frequency $`\omega `$ is very similar to that of the ZFC susceptibility with a fixed $`\tau =2\pi /\omega \tau _\omega `$. Thus, in the present work, we regard that $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ is essentially equivalent to the ac susceptibility measured in experiments.
The function with $`t_{\mathrm{w2}}=0`$, denoted by $`C_{[T_2,T_1]}(\tau ;0,t_{\mathrm{w1}})`$, was examined in our previous work$`^{\text{?)}}`$ hereafter referred to as III. We found that at $`\tau t_{\mathrm{w1}}^{(\mathrm{eff})}`$ it follows similar scaling forms to those for $`C_T(\tau ;t_\mathrm{w})`$ in isothermal aging described in §2.1 if the factor $`L_T(\tau )/R_T(t_\mathrm{w})`$ in eq.(2.10) is replaced by $`L_{[T_2,T_1]}(\tau ;t_{\mathrm{w1}})/R_{T_1}(t_{\mathrm{w1}})`$. Here $`L_{[T_2,T_1]}(\tau ;t_{\mathrm{w1}})`$ is a mean size of the droplet excitations, or quasi-domains, which are in local equilibrium at the new temperature $`T_2`$ within time scale $`\tau `$ after the $`T`$-shift.
In Fig. 5 we show typical results of $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ simulated for a fixed $`\tau _\omega (=32)`$. Their behaviour is rather similar to that of $`e_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ shown in Fig. 4. Just after the $`T`$-shift, it rapidly decreases (increases) and then relaxes slowly to the isothermal curve $`\stackrel{~}{\chi }_{T_2}(\tau _\omega ;t)`$ from below (above) for $`\mathrm{\Delta }T<0(>0)`$. Furthermore, it also satisfies a relation similar to eq.(3.2): when the part of its curve at $`t_2>0`$ is plotted against $`\stackrel{~}{t}=t_2+t_{\mathrm{w1}}^{(\mathrm{eff})}`$ it now relaxes to the isothermal curve from above (below),
$$\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})\stackrel{~}{\chi }_{T_2}(\tau _\omega ;t_2+t_{\mathrm{w1}}^{(\mathrm{eff})})\mathrm{at}t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}.$$
(3.5)
Here, a value of $`t_{\mathrm{w1}}^{(\mathrm{eff})}(824)`$ has been chosen so as to obtain a good collapse of the two curves at the time range $`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$. The resultant $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ is again consistent with eq.(2.12) combined with eqs.(1.2) and (1.3) obtained in I.
The above feature in isothermal regime 3), $`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$, means that there the mean domain size and the correlation function are related by the scaling relations of eqs.(2.8), (2.9) and (2.10). Thus relaxation of $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ in this regime is due to increase of the effective stiffness due to the coarsening of domain walls.
Behaviour of $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ in transient regime 2) is considered to reflect nature associated with quasi-domains. Let us here focus on the negative $`T`$-shift shown in Fig. 5(a). As pointed out just above, $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ is definitely smaller than $`\stackrel{~}{\chi }_{T_2}(\tau _\omega ;t)`$ at $`t>t_{\mathrm{w1}}`$. This is different from the rejuvenation, or chaos effect observed in experiments if the latter means that the system after the $`T`$-shift looks almost as young as the system which has just started a new isothermal aging at $`T=T_2`$. If this is the case, $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ as a function of $`\stackrel{~}{t}`$ with $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ smaller than $`t_{\mathrm{w1}}`$ should coincide with $`\stackrel{~}{\chi }_{T_2}(\tau _\omega ;t)`$ at $`t_2>t_{\mathrm{rev}}`$$`^{\text{?)}}`$ The behaviour of $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ may suggest that, if the overlap length $`l_{\mathrm{\Delta }T}`$ introduced by eq.(2.13) in §2.3 exists in the present SG model, it should be larger than $`R_{T_1}(t_{\mathrm{w1}})`$ observed in the simulation.
Let us then compare $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ as a function of $`t_2`$ with $`\stackrel{~}{\chi }_{T_2}(\tau _\omega ;tt_{\mathrm{w1}}^{(\mathrm{eff})})`$. The two curves are shown in the inset of Fig. 5(a). Interestingly, the former is larger than the latter and the excessive part decays very slowly. This feature in transient regime 2) may be explained within the scenario sketched in §2.3 as the following. After $`t_{\mathrm{w2}}`$ from the $`T`$-shift, there remains excessive population of droplets of length scales larger than the size of the quasi-domain $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$. The excessive population is proportional to $`\mathrm{\Delta }T=T_2T_1`$. The boundary of the latter droplets may act as frozen-in domain walls and reduces the effective stiffness of droplets smaller than $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ just as in the case of the domain walls which separates different pure states. $`^{\text{?, ?)}}`$ This yields the extra contribution, which is proportional to $`\mathrm{\Delta }T`$, to the relaxation. The scaling analysis on the excessive part in terms of $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ and $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$ is now under investigation.
### 3.5 $`T`$-cycling process
A $`T`$-cycling process is a combination of two $`T`$-shift processes with the same $`|\mathrm{\Delta }T|`$ but with opposite signs. At a time $`t_2=t_{\mathrm{w2}}`$ after the $`T`$-shift from $`T_1`$ to $`T_2`$ the temperature is turned back to $`T_1`$, and the subsequent relaxation is observed. When $`t_{\mathrm{w2}}t_{\mathrm{w1}}^{(\mathrm{eff})}`$, the $`T`$-change back to $`T_1`$ is effectively the same as a simple $`T`$-shift process from an isothermally aged state at $`T_2`$ to $`T_1`$, as shown in Fig. 6(a). In this negative $`T`$-cycling process ($`T_2<T_1`$), the approximated ac susceptibility at $`t=t_3+t_{\mathrm{w2}}+t_{\mathrm{w1}}`$, now denoted simply by $`\stackrel{~}{\chi }_{T(t)}(\tau _\omega ;t)`$, merges into the isothermal curve $`\stackrel{~}{\chi }_{T_1}(\tau _\omega ;t)`$ from below when the former is plotted against $`\stackrel{~}{t}=t_3+t_\mathrm{w}^{(\mathrm{eff})}`$. Here $`t_\mathrm{w}^{(\mathrm{eff})}`$ is the effective waiting time given by $`t_{\mathrm{w2}}^{z(T_1)/z(T_2)}`$ (since $`t_{\mathrm{w2}}t_{\mathrm{w1}}^{(\mathrm{eff})}`$ in this case).
In the case that $`t_{\mathrm{w2}}`$ is small enough, we obtain a similar behaviour but with $`t_\mathrm{w}^{(\mathrm{eff})}t_{\mathrm{w1}}`$ (not shown). This means that aging at $`T_2`$ does not contribute at all to aging at $`T_1`$. This is nothing but the memory effect and has been observed by many experiments. $`^{\text{?, ?, ?)}}`$
In a positive $`T`$-cycling process with $`T_2>T_1`$ shown in Fig. 6(b), our simulated data still exhibit the memory effect in the sense that the curve at $`t_3>0`$ lies on the isothermal curve of $`T_1`$ if it is shifted by a proper amount, i.e., $`t_\mathrm{w}^{(\mathrm{eff})}`$ which depends not only $`t_{\mathrm{w2}}`$ but also on $`t_{\mathrm{w1}}`$. This type of memory retained over a positive $`T`$-cycling has been indeed observed in the ferromagnetic fine particle system. $`^{\text{?)}}`$
### 3.6 Continuous $`T`$-change with an intermittent stop
We have also performed simulations on a continuous $`T`$-change with an intermittent stop. Following the experimental protocol, $`^{\text{?)}}`$ we cool a system by a constant rate from above $`T_\mathrm{c}`$ to below $`T_\mathrm{c}`$ with an intermittent stop at a certain temperature $`T_\mathrm{I}`$ below $`T_\mathrm{c}`$, restart the cooling, and then heat it back by the same rate as in the cooling.
A typical result is shown in Fig. 7. In this analysis $`\stackrel{~}{\chi }_{T(t)}(\tau _\omega ;t)`$ has been directly measured as the response to an ac field $`h(t)=h_0\mathrm{sin}(\omega t)`$ with $`h_0=0.3`$ and $`2\pi /\omega =160`$ (MCS).
At the intermittent stop at $`T_\mathrm{I}`$, $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ relaxes downward and deviates from the reference curve of cooling. After restarting the cooling, $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ is seen to merge into the reference curve. It appears that the long time aging at $`T_\mathrm{I}`$ does not affect subsequent cooling at temperatures lower than $`T_\mathrm{I}`$ by a certain amount. Here we call this phenomenon a rejuvenation-like effect, since $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ merges with the reference curve but without a strong vertical increase in contrast to the rejuvenation seen in the experiment. $`^{\text{?)}}`$ By a detailed inspection of Fig. 7, one may notice that the curve $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ looks as if it increases just after the cooling is restarted. We consider that this behaviour does not indicate the rejuvenation but is due to the following reason: $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ during the stop initially decreases appreciably but finally its relaxation becomes invisible because of its statistical fluctuation. The curve looks as if it increases initially at the restart of cooling simply because the relaxation starts in the middle of the fluctuation.
When the system is heated back from the lowest temperature of cooling process, $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ exhibits a dip at around $`T=T_\mathrm{I}`$, though it is rather shallow. This means that the memory imprinted by the long time aging at $`T_\mathrm{I}`$ in the cooling process, in fact, has not been erased until the system comes back to $`T_\mathrm{I}`$ from the lower temperatures. To conclude the present SG model exhibits the rejuvenation-like and memory effects within a time-window of the present simulation.
## 4 Discussions
We have studied aging phenomena in various $`T`$-change protocols on the 3D Gaussian Ising SG model by Monte Carlo simulations. In particular, we have argued that, in a $`T`$-shift protocol, slow aging dynamics after the $`T`$-change from $`T_1`$ to $`T_2`$ at $`t_2=0`$ consists of two regimes: transient ($`t_2<t_{\mathrm{w1}}^{(\mathrm{eff})}`$) and isothermal ($`t_2>t_{\mathrm{w1}}^{(\mathrm{eff})}`$) regimes, where $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ is the effective waiting time specified by eq.(2.12). In the transient regime, re-distribution of thermal weights, from those at $`T_1`$ to $`T_2`$, of droplet excitations with the characteristic length scale $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ is taking place within each domain of the mean size $`R_{[T_2,T_1]}(t;t_{\mathrm{w1}})`$. We called the length scale $`L_{[T_2,T_1]}(t_2,t_{\mathrm{w1}})`$ as the size of quasi-domains.
Suppose, in a negative $`T`$-shift process, we further decrease temperature to $`T_3(<T_2)`$ at $`t_2=t_{\mathrm{w2}}(t_{\mathrm{w1}}^{(\mathrm{eff})})`$ and wait a certain period. Then there appear now quasi-domains locally equilibrated to $`T_3`$ within each quasi-domain of $`T_2`$. Because of the wide separation of time scales at different length scales emphasized in §2.1, a nest (or hierarchy) of quasi-domains can be realized dynamically when the system is cooled step-wise, or continuously. At low temperatures the ac field can only excite droplets of the size $`L_T(1/\omega )`$ corresponding to the frequency $`\omega `$. If the size is small enough, a new contribution to the ac response will appear every time a new frozen-in domain wall appears associated with the quasi-domain due to a step of cooling. This picture, quasi-domains within (quasi-) domains, is our interpretation of the rejuvenation-like effect seen in Fig. 7. The picture is somewhat similar to the one proposed before. $`^{\text{?, ?)}}`$ By our interpretation the observation of the rejuvenation-like effect crucially depends on time and length scales by which one looks at the system. It is the easier to be observed, the shorter and smaller are the time and length scales.
By means of our scenario the memory effect is rather easy to be understood. For example, the continuous $`T`$-change protocol shown in Fig. 7 is interpreted as follows. When the system is re-heated from the lowest temperature, the hierarchical structure of quasi-domains is erased from the lower levels. When the temperature comes back to $`T_\mathrm{I}`$, the large domains imprinted at the intermittent stop in the cooling process reappear and govern the response of the system, which causes the memory effect in this protocol. According to our scenario, a memory imprinted by aging at a certain temperature $`T_\mathrm{I}`$ is erased when the system is aged at different temperature, say $`T_\mathrm{d}`$, for such a long period that the mean size of quasi-domains of $`T_d`$ catches up the domains imprinted at $`T_\mathrm{I}`$, as what happens in regime 3) of the $`T`$-shift protocol. Even in that case, the memory of the previous thermal history is kept in the effective waiting time, i.e., the system looks more aged than the period that it is actually aged at $`T=T_\mathrm{d}`$ (see eq.(3.5)).
Let us now compare our simulated results with experimental observations. As pointed out already in §1, in the ferromagnetic fine particles system (FFPS), whose aging dynamics has been recently studied by Mamiya et al.$`^{\text{?)}}`$ the microscopic time, $`\tau _{\mathrm{mic}}`$, required for a magnetic moment of each fine particle to flip is of the order of $`10^3`$sec. $`^{\text{?)}}`$ Therefore the time window of their observation, $`110^3`$ ksec, in unit of $`\tau _{\mathrm{mic}}`$ is not much different from that of our MC simulation. It is also noted that in the FFPS each magnetic moment behaves as an Ising spin due to strong magnetic anisotropy.
Indeed in the $`T`$-shift protocol, Mamiya et al. have observed relaxation of $`\chi _{[T_2,T_1]}^{\prime \prime }(\omega ;t;t_{\mathrm{w1}})`$ which is qualitatively quite similar to our simulated $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ shown in Fig. 5. They have determined amount of the shift of $`\chi _{[T_2,T_1]}^{\prime \prime }(\omega ;t;t_{\mathrm{w1}})`$, or $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$, by the same method that we have used for analysis of $`\stackrel{~}{\chi }_{[T_2,T_1]}(\tau _\omega ;t;t_{\mathrm{w1}})`$ in Fig. 5. The effective waiting time they obtained is written as $`t_{\mathrm{w1}}^{(\mathrm{eff})}3\times (t_{\mathrm{w1}})^b`$ with $`b1.0`$. By contrast, $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ extracted from our simulation is given by eq.(2.12), which, with making use of eq.(1.2), is reduced to $`t_{\mathrm{w1}}^{(\mathrm{eff})}(t_{\mathrm{w1}})^b`$ with $`b=z(T_2)/z(T_1)(1.11(1.25)`$ for $`T_1=0.8`$ and $`T_2=0.7(0.6)`$). Also aging dynamics of the FFPS in the $`T`$-cycling protocol agrees qualitatively with the results of our simulation. In particular, the effective waiting time $`t_\mathrm{w}^{(\mathrm{eff})}`$ which depends on $`t_{\mathrm{w1}}`$ (memory effect) is observed even in the positive $`T`$-cycling process, though Mamiya et al. have claimed that the memory is partially lost because some particles with a larger size than the average are affected by the chaos effect associated with the temperature change.
For ordinary spin glasses with $`\tau _{\mathrm{mic}}10^{13}`$sec, there have been only a few experiments, in which $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ in the $`T`$-shift process is systematically studied. Djurberg et al. $`^{\text{?)}}`$ measured the ZFC magnetization of Cu:Mn 13.5at% spin glass with a fixed $`t_{\mathrm{w1}}`$ and varying $`\mathrm{\Delta }T`$. Via the FDT similar to eq.(2.11) their analysis is directly comparable with our simulation on $`C_{[T_2,T_1]}(\tau ;0,t_{\mathrm{w1}})`$ in III. The peak position of its relaxation rate with respect to ln$`t`$ may be assigned as $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ (crossover from regime 2) to regime 3)). The experimentally observed $`t_{\mathrm{w1}}^{(\mathrm{eff})}`$ becomes larger (smaller) than $`t_{\mathrm{w1}}`$ in the positive (negative) $`T`$-shift, which is qualitatively consistent with eq.(2.12) with monotonic increase of $`1/z(T)`$ on $`T`$. Quantitatively, however, the shift observed experimentally is much larger than what is expected from eq.(1.3).
Djurberg et al. $`^{\text{?)}}`$ furthermore claimed that they observed the chaos effect in $`T`$-shift processes with $`|\mathrm{\Delta }T|`$ larger than a certain value. The rejuvenation, or chaos effect in the $`T`$-shift and $`T`$-cycling protocols observed in $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ has been reported for CdCr<sub>1.7</sub>In<sub>0.3</sub>S<sub>4</sub> $`^{\text{?, ?)}}`$ and Cu:Mn 2at% $`^{\text{?)}}`$ spin glasses, as already mentioned in §1. The observed $`\chi _{T(t)}^{\prime \prime }(\omega ;t)`$ is claimed $`^{\text{?)}}`$ to indicate that the system is literally rejuvenated in the sense described in §3.4. Neither such a chaos effect nor literal rejuvenation have been seen in our present simulations (see also Kisker et al. $`^{\text{?)}}`$). At the moment it is not clear yet whether this discrepancy originates simply from the difference in time scales of the observations, or from something else.
To conclude, we have simulated various $`T`$-change protocols of aging in the 3D Ising SG model, and have proposed a scenario in terms of domains and quasi-domains: the domains ever grow in any aging process in the SG phase and the quasi-domains appear within the domains or within themselves. The results observed in our simulations are reasonably well interpreted by the scenario. Qualitatively, it is also consistent with many features observed experimentally, particularly, those related to the memory effect. For certain phenomena such as the growth law of $`R_T(t)`$, the consistency is even semi-quantitative. Concerned with the rejuvenation, or chaos effect, however, the scenario may not be applicable, and further studies are required.
## Acknowledgments
We would like to sincerely thank E. Vincent and M. Ocio for their fruitful discussions and for kindly allowing us to use their data on a AgMn spin glass. We also thank H. Mamiya for his helpful discussions on the experiments of his research group. Our thanks are also to J.P. Bouchaud, K. Hukushima and P. Nordblad for their useful discussions. Two of the present authors (T. K. and H. Y.) were supported by Fellowships of Japan Society for the Promotion of Science for Japanese Junior Scientists. This work is supported by a Grant-in-Aid for International Scientific Research Program, “Statistical Physics of Fluctuations in Glassy Systems” (#10044064) and by a Grant-in-Aid for Scientific Research Program (#10640362), from the Ministry of Education, Science and Culture. The present simulation has been performed on FACOM VPP-500/40 at the Supercomputer Center, Institute for Solid State Physics, the University of Tokyo. |
warning/0001/astro-ph0001296.html | ar5iv | text | # H2 emission as a diagnostic of physical processes in starforming galaxies
## 1 Introduction
Direct observations of $`\text{H}_2`$ emission from external galaxies have become standard practice in the past decade through the revolution in ground-based near-infrared instrumentation. As a result, the near-infrared $`\text{H}_2`$ rovibrational lines are now readily detectable throughout the local universe (e.g., Moorwood & Oliva 1988, 1990; Puxley et al. 1988, 1990; Goldader et al. 1995, 1997; ? (?). More recently, the Short Wavelength Spectrograph (SWS) on the Infrared Space Observatory (ISO) has for the first time allowed detection of the purely rotational $`\text{H}_2`$ lines in the mid-infrared spectral regime. For instance, the first detection (outside the solar system) of the $`\text{H}_2`$ S(0) line at $`28.21\mu \text{m}`$ was reported by ? (?) from the star forming nucleus of the nearby spiral galaxy $`\mathrm{NGC}\mathrm{\hspace{0.17em}6946}`$.
The interpretation of these data is, however, far from trivial. At typical molecular cloud temperatures ($`T20\text{K}`$) the upper levels of even the lowest $`\text{H}_2`$ transitions are essentially unpopulated, and hence $`\text{H}_2`$ emission is, unlike for instance CO $`J=10`$ emission, not a tracer of bulk molecular gas mass. Instead, an excitation mechanism capable of populating these energy levels is required. Furthermore, the excitation rate needs to be high enough to maintain an excited level population sufficient for producing detectable emission. Hence the $`\text{H}_2`$ line luminosities measure the rate of excitation; as such they provide unique and highly diagnostic information, that cannot be obtained in any other way. In addition, since the $`\text{H}_2`$ lines are forbidden quadrupole transitions with very small Einstein $`A`$ values, the lines are optically thin; hence there is, in contrast to the situation with most other molecular lines, no need to solve a complicated, geometry-dependent radiative transfer problem for a physical interpretation of the $`\text{H}_2`$ lines. The near-infrared $`\text{H}_2`$ lines will of course suffer from extinction by dust, but this effect can usually be quantified adequately by using suitable ratios of hydrogen recombination lines, \[Feii\] lines or $`\text{H}_2`$ rovibrational lines in the same spectral range, with accurately known intrinsic flux ratios.
The situation is complicated, however, by the fact that $`\text{H}_2`$ excitation can be brought about by a variety of mechanisms, which may all play a role. For instance, in a starburst galaxy, $`\text{H}_2`$ emission may be generated by UV-pumping in starforming regions or by shocks due to supernova remnants or outflows; moreover, these processes are expected to occur together in the same small volume occupied by the starburst, and thus a combination of excitation mechanisms may be expected at every position. Generally, this combination will be difficult to separate (e.g., $`\mathrm{NGC}\mathrm{\hspace{0.17em}253}`$ observations: ? (?; ? (?). In addition, shocks due to large-scale streaming motions in spiral arms, bars (favoured for the barred Seyfert 2 $`\mathrm{NGC}\mathrm{\hspace{0.17em}1068}`$ by ? (?) or merger-driven flows (e.g., in $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$, ? (?; see also § 3) may play a role. Furthermore, X-ray excitation may be produced by multiple supernova remnants, or, if present, by an active galactic nucleus (proposed for e.g., Cyg A by ? (? and for Cen A by ? (?) or a cooling flow ?. In the absence of sufficient spatial resolution for separating the various excitation mechanisms, spectral diagnostics must be used. Models for UV-pumped, shocked or X-ray excited $`\text{H}_2`$ emission have reached considerable predictive power; however, the densities of the emitting regions are often sufficiently high to thermalize the relevant level populations, quenching the typical signatures of non-thermal excitation processes. The only accessible diagnostics are then fluxes of lines with very high critical densities. However, these lines are intrinsically faint, and accurate fluxes for such lines require long integrations with large telescopes. The complexity of the problem is well illustrated by the $`\text{H}_2`$ emission of the merging system $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$, which has been attributed to X-ray excitation ?, UV-pumping ?, shocks ? and formation pumping ?. However, all of the non-thermal excitation processes relied on poorly measured fluxes of faint lines. More recent spectroscopy of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ by ? (?) showed that only shock excitation can account for the $`\text{H}_2`$ rovibrational spectrum (see also § 3).
Generally a combination of accurate multi-line spectroscopy and suitable additional information such as high resolution spatial information is needed for a proper analysis of the dominant excitation mechanism and hence a physical analysis of the $`\text{H}_2`$ emission. A complicating factor when combining $`\text{H}_2`$ rovibrational and purely rotational lines is the fact that the dominant excitation mechanisms of these lines may be different, because of the different excitation requirements of these transitions: while the lowest rotational lines require $`T>100\text{K}`$ and are excited at moderate densities ($`n_{\mathrm{H}_2}>10^2\text{cm}^3`$), the rovibrational lines require $`T>2000\text{K}`$ and $`n_{\mathrm{H}_2}>10^4\text{cm}^3`$. The following sections will therefore present two case studies. First an analysis of the extended purely rotational $`\text{H}_2`$ emission from the nearby edge-on spiral galaxy $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ is given (§ 2). Then, the more extreme conditions in the merger $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ are discussed, using new high resolution near-infrared data. The general implications of these two cases are discussed in § 4.
## 2 Extended H<sub>2</sub> rotational line emission in the disk of NGC 891
? (?) have observed seven positions in the disk of the nearby (distance $`D=9.5\text{Mpc}`$) edge-on spiral galaxy $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ with the ISO SWS. The positions observed include the nucleus and positions spaced along the disk to galactocentric distances $`R`$ of $`8\text{kpc}`$ south of the nucleus and $`11\text{kpc}`$ north of the nucleus. At the $`11\text{kpc}`$ north position, the $`{}_{}{}^{12}\text{CO}`$ $`J=10`$ ?; ?; ?, and dust emission ? are barely detected. With ISO, the $`\text{H}_2`$ S(0) ($`28.21\mu \text{m}`$) and S(1) ($`17.03\mu \text{m}`$) lines were detected at all positions. These are the first detections of these lines outside starburst or active nuclei, with the exception of the detection in the disk of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6946}`$ reported by ? (?).
The simplest analysis of these data would assume that the emission is fully thermalized (i.e., the high-density limit is assumed) and arises from an isothermal gas layer with an ortho-para ratio of three, in agreement with the statistical weights of the ortho- and para-varieties. Under these assumptions the S(1)/S(0) ratio yields the temperature of the emitting region, which can be combined with the observed fluxes to give the column density of emitting $`\text{H}_2`$, averaged over the SWS aperture. However, these simple assumptions lead to unacceptable results. For instance, at the position $`8\text{kpc}`$ north (a typical “disk” position), the S(0) and S(1) data then imply $`N(\text{H}_2)=2.710^{22}`$ at $`T=76\text{K}`$. This result is incompatible with the CO $`J=21`$/$`J=10`$ line ratio of 0.75, at this position ?, which strongly rules out a large mass of dense molecular gas at $`T=76\text{K}`$. This argument shows that the simple assumptions used above do not suffice.
The discrepancy can be removed by allowing lower densities, and therefore subthermal excitation of the $`\text{H}_2`$ $`J=3`$ level, giving rise to a higher implied temperature for a given S(1)/S(0) ratio. Because the temperatures involved are still much lower than the upper levels of the transitions involved, the implied $`\text{H}_2`$ mass will be extremely sensitive to the adopted temperature, and thus strongly decrease as lower densities are allowed. An additional advantage of this procedure is that the CO $`J=2`$ level will now also be subthermally populated, so that CO $`J=21`$/$`J=10`$ line ratios lower than the thermal value can be tolerated in the analysis. Indeed the multi-level multi-isotope CO data by ? (?) indicate dominant densities of $`n_{\mathrm{H}_2}10^3\text{cm}^3`$ in the nuclear region and $`n_{\mathrm{H}_2}200\text{cm}^3`$ throughout the disk of $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$. Adopting these densities, the ISO data imply $`\text{H}_2`$ column densities varying from 10 to 30% of the value derived from CO, at temperatures of 120 to $`130\text{K}`$, throughout the inner disk ($`R<8\text{kpc}`$) of $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$.
At the outer disk positions ($`R=811\text{kpc}`$) however, this solution is not adequate, since the S(0) and S(1) line profiles at these positions are significantly different. The SWS uses an aperture much larger than the diffraction beam and hence the observed line width depends on the extent of the emission region; the S(0) line shows the broad profile characteristic of aperture-filling emission, while the S(1) profile is narrow, indicating emission from a region much smaller than the aperture (Fig. 1). Thus the S(0) and S(1) lines in the outher disk of $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ must originate in separate regions: a warm component ($`T>130\text{K}`$) dominating the S(1) emission and located in isolated regions in the disk and a separate more pervasive component, dominating the S(0) line. The latter component may contain very significant mass, depending on its temperature, and the implications of this possibility for the mass budget at the outer positions have been explored by ? (?). However, a solution where the component dominating the S(0) line is cool ($`T90\text{K}`$) is problematic, since the heating required for maintaining a very large mass of $`\text{H}_2`$ at $`T90\text{K}`$ cannot be provided. In thermal equilibrium the heating rate should equal the cooling rate, which is dominated by \[Cii\] $`158\mu \text{m}`$ emission (which at $`T90\text{K}`$ is an extremely efficient cooler, since the upper level of the $`158\mu \text{m}`$ transition is at $`91\text{K}`$), with possible contributions from $`\text{H}_2`$ rotational lines, CO rotational lines, and \[Ci\] emission. The \[Cii\] $`158\mu \text{m}`$ emission from $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ has been observed by ? (?), and it is easily verified that in the outer disk the \[Cii\] emission dominates the $`\text{H}_2`$ emission (SWS data discussed here), CO emission (barely detected in the outer disk) and \[Ci\] emission ?. If all available carbon (assumed to have solar abundance) is in the form of C<sup>+</sup>, the \[Cii\] emission allows a column density of at most $`N(\text{H}_2)710^{21}\text{cm}^2`$ at $`T90\text{K}`$ in the outer disk, averaged over the SWS aperture, while $`N(\text{H}_2)10^{23}\text{cm}^2`$ would be required to produce the observed S(0) line flux.
This problem can be solved by relaxing the final assumption: that of an ortho-para of three. A lower ortho-para ratio raises the implied temperature and lowers the implied mass in the same way as a lower density does (as discussed above). Assuming an ortho-para ratio of unity, the implied mass is lowered sufficiently that the required heating can be accounted for.
This analysis thus favours an interpretation in which the $`\text{H}_2`$ rotational emission arises in low-density ($`n_{\mathrm{H}_2}200\text{cm}^3`$), warm ($`T120\text{K}`$) molecular gas with an ortho-para ratio of about unity, which pervades the disk of $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ at least to the end of the detectable CO disk and dominates the S(0) emission; it contains 10 to 30% of the mass implied by CO observations (lower fractions will result if an ortho-para ratio below unity is adopted); concentrations of warmer gas (plausibly identified with active star forming regions) are ubiquitous in the inner disk and rarer in the outer disk and give rise to the S(1) emission. Given the density and temperature of the extended warm gas, this component can most likely be identified with extended low-density photon-dominated regions (PDRs) which form the warm envelopes of giant molecular clouds, heated by the local interstellar radiation field. Observations of Galactic molecular cloud edges in the $`21\text{cm}`$ Hi line reveal the presence of such warm envelopes by a “limb brightening” in Hi ( ? (?; Van der Werf et al. 1988, 1989). Modeling of molecular cloud envelopes by ? (?) shows that the temperatures and densities estimated here are typical for such regions, especially in the zone where the transition from molecular to atomic hydrogen takes place. This interpretation of the present $`\text{H}_2`$ data in terms of extended diffuse PDRs in cloud envelopes is corroborated by the excellent numerical agreement with column density estimates based on \[Cii\] $`158\mu \text{m}`$, the principal coolant for such regions. In addition, the fact that the implied ortho-para ratio is significantly lower than three also points to a low-density PDR origin of this emission, since shocks or hot, high-density PDRs would produce an ortho-para ratio of three (the high temperature thermal value) by spin exchange reactions with H and H<sup>+</sup>. In summary, the $`\text{H}_2`$ rotational lines in the disk of $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ arise in warm, extended molecular cloud envelopes; these envelopes provide a physical link between the giant molecular clouds and the atomic medium in which they are embedded ?. The $`\text{H}_2`$ rotational lines provide unique diagnostics of these regions.
## 3 Vibrational H<sub>2</sub> emission in the nuclear region of the luminous merger NGC 6240
The vibrational $`\text{H}_2`$ emission of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ has attracted attention because of its extraordinary luminosity: $`710^7\text{L}_{}`$ in the $`\text{H}_2`$ $`v=10`$ S(1) line alone (for $`H_0=75\text{km}\text{s}^1\text{Mpc}^1`$ and with no correction for extinction). This line contains 0.012% of the bolometric luminosity of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$, which is considerably higher than any other galaxy ?. Together the vibrational lines may account for 0.1% of the total bolometric luminosity.
Imaging of the $`\text{H}_2`$ $`v=10`$ S(1) emission from $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ has shown that the $`\text{H}_2`$ emission peaks between the two remnant nuclei of the merging system ?. This morphology provides a unique constraint on the excititation mechanism, since it argues against any scenario where the excitation is dominated by the stellar component (e.g., UV-pumping, excitation by shocks or X-rays from supernova remnants). Instead, the favoured excitation mechanism is slow shocks in the nuclear gas component, which, as shown by recent high resolution interferometry in the CO $`J=21`$ line ?, also peaks between the nuclei of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$. A multi-line $`\text{H}_2`$ vibrational spectrum (Fig. 2) confirms this excitation mechanism ?, in agreement with more limited spectroscopy by ? (?). The interpretation in terms of slow shocks also naturally accounts for the high ratio of $`\text{H}_2`$ line emission to infrared continuum emission, which is a characteristic of such shocks (e.g., ? (?).
What is the role of these shocks? In the shocks mechanical energy is dissipated and radiated away, mostly in spectral lines (principally $`\text{H}_2`$, CO, $`\text{H}_2\text{O}`$ and \[Oi\] lines). This energy is radiated away at the expense of the orbital energy of the molecular clouds in the central potential well. Consequently, the dissipation of mechanical energy by the shocks will give rise to an infall of molecular gas to the centre of the potential well. Therefore, the $`\text{H}_2`$ vibrational lines measure the rate of infall of molecular gas into the central potential well in $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$. This conclusion can be quantified by writing
$$L_{\mathrm{rad}}=L_{\mathrm{dis}},$$
(3.1)
where $`L_{\mathrm{rad}}`$ is the total luminosity radiated by the shocks and $`L_{\mathrm{dis}}`$ the dissipation rate of mechanical energy, giving rise to a molecular gas infall rate $`\dot{M}_{\mathrm{H}_2}`$ given by
$$L_{\mathrm{dis}}=\frac{1}{2}\dot{M}_{\mathrm{H}_2}v^2,$$
(3.2)
where $`v`$ is the circular orbital velocity at the position where the shock occurs.
Using a $`K`$-band extinction of $`0^\mathrm{m}.15`$ ?, the total luminosity of $`\text{H}_2`$ vibrational lines from $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ becomes $`7.210^8\text{L}_{}`$; inclusion of the purely rotational lines observed with the ISO SWS approximately doubles this number, so that $`L_{\mathrm{rad}}=1.510^9\text{L}_{}`$.
In order to use this number to estimate $`\dot{M}_{\mathrm{H}_2}`$, it is necessary to establish more accurately the fraction of the $`\text{H}_2`$ emission that is due to infalling gas. Observations with NICMOS on the Hubble Space Telescope (HST) by ? (?) provide the required information (Fig. 3). The NICMOS image shows that the emission consists of a number of tails (presumably related to the superwind also observed in $`\text{H}\alpha `$ emission), and concentrations assocated with the two nuclei, and a further concentration approximately (but not precisely) between the two nuclei. The relative brightness of the $`\text{H}_2`$ emission from the southern nucleus is deceptive, since this nucleus is much better centred in the filter that was used for these observations than the other emission components, in particular the northern nucleus. Taking this effect into account, it is found that 32% of the total $`\text{H}_2`$ flux is associated with the southern nucleus, 16% is associated with the northern nucleus, and 12% with the component between the two nuclei, the remaining 40% being associated with extended emission. Using inclination-corrected circular velocities (from ? (?) of $`270`$ and $`360\text{km}\text{s}^1`$ for the southern en northern nucleus respectively, and of $`280\text{km}\text{s}^1`$ for the central component ?, the mass infall rates derived using Eqs. $`\left(\text{3.1}\right)`$ and $`\left(\text{3.2}\right)`$ are $`80\text{M}_{}\text{yr}^1`$ for the southern nucleus, $`22\text{M}_{}\text{yr}^1`$ for the northern nucleus and $`28\text{M}_{}\text{yr}^1`$ for the central component.
The derived molecular gas inflow rate to the two nuclei is remarkably close to the mass consumption rate by star formation of approximately $`60\text{M}_{}\text{yr}^1`$, indicating that the $`\text{H}_2`$ emission from the nuclei directly measures the fueling of the starbursts in these regions. The gas inflow towards the central component is more remarkable, since this component is not associated with a prominent stellar component. The gravitational potential at this position is therefore most likely dominated by the gas itself, which is also indicated by high resolution interferometry in the CO $`J=21`$ line ?. The absence of prominent star formation at this position shows that the central gas component is gravitationally stabilized, possibly by a high local shear. However, as the central gas column density increases, the shear must eventually be overcome and given the high gas density (and hence short free-fall time) an explosive starburst will result. In that stage $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ will become a true ultraluminous infrared galaxy.
## 4 General implications
The two case studies discussed above illustrate the unique diagnostic power of $`\text{H}_2`$ emission lines provided that the excitation mechanism can be determined. The extended PDR emission detected in $`\mathrm{NGC}\mathrm{\hspace{0.17em}891}`$ reveals the presence of a widespread fluorescent component, which is probably a common feature of star forming galaxies. For instance, ? (?) have detected extended diffuse $`\text{H}_2`$ vibrational emission from the inner Milky Way and argue that this emission is UV-excited; similarly, ? (?) argue for purely fluorescent extended $`\text{H}_2`$ emission in the moderate luminosity starburst galaxy $`\mathrm{NGC}\mathrm{\hspace{0.17em}253}`$.
The configuration of shocked $`\text{H}_2`$ emission from a pronounced molecular gas concentration between the two nuclei as found in $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$ is not unique either: it is also found in $`\mathrm{Arp}\mathrm{\hspace{0.17em}220}`$ ?, and several other mergers. Perhaps the best case in point is the well-known merging “Antennae” system ($`\mathrm{NGC}\mathrm{\hspace{0.17em}4038}4039`$) where pronounced CO emission is found in the interaction zone between the two nuclei ?, which is the site of vigorous obscured star formation ?. These geometries may be due to the fact that the gas components are dissipative, and thus merge on a shorter time scale than the dissipationless stellar components, which merge by the slower process of dynamical friction. The $`\text{H}_2`$ emission thus provides unique insight into the physics of these gas-rich mergers.
###### Acknowledgements.
I would like to thank my collaborators Frank Israel, Alan Moorwood, Tino Oliva, and Edwin Valentijn for discussions on this subject, and Guido Kosters for reducing the NICMOS data of $`\mathrm{NGC}\mathrm{\hspace{0.17em}6240}`$. |
warning/0001/quant-ph0001102.html | ar5iv | text | # Quantum Stochastic Resonance in Electron Shelving
## Abstract
Stochastic resonance shows that under some circumstances noise can enhance the response of a system to a periodic force. While this effect has been extensively investigated theoretically and demonstrated experimentally in classical systems, there is complete lack of experimental evidence within the purely quantum mechanical domain. Here we demonstrate that stochastic resonance can be exhibited in a single ion and would be experimentally observable using well mastered experimental techniques. We discuss the use of this scheme for the detection of the frequency difference of two lasers to demonstrate that stochastic resonance may have applications in precision measurements at the quantum limit.
Imagine that you are set the task of detecting a very weak periodic signal by means of its interaction with a suitable probe system. Given that the signal is weak, one intuitively may think that the optimal experimental set up should minimize any other interaction that the probe may undergo. However, this is not always the case and there are situations where noise can indeed play a constructive role in a high sensitivity detection. A clear illustration of this fact is provided by the phenomenon of stochastic resonance (SR) , where the response of a nonlinear system to external periodic driving is enhanced in the presence of noise. Typically , the signal-to-noise ratio (SNR) increases monotonically up to a maximum for certain optimal noise intensity, and then decreases gradually as randomization dominates over the cooperative effect between the coherent driving and the stochastic forces. The simplest system for describing the appearance of SR consists of a particle in a bistable potential subject to both thermal noise and a periodic forcing . However, many other scenarios have been proposed and recent research has shown that SR may also be observed in some monostable systems . Experimental research has confirmed that the phenomenon, if at first sight counter-intuitive, is rather ubiquitous. Since the first demonstration in a Schmitt trigger circuit , SR has been observed in a wide variety of physical systems, ranging from ring lasers to neuronal cells (see for a detailed review). Moreover, there is experimental evidence that certain complex living systems (such as crickets) make use of SR to improve the sensitivity of their sensory organs .
Recently the concept of SR has been extended to the quantum domain . However, experimental verifications at the level of individual quantum systems are difficult (see e.g. ) and it would be of great interest to find feasible experimental scenarios that allow the investigation of stochastic resonance in the quantum regime. In this letter we show that SR can be demonstrated in a conceptually simple truly microscopic quantum optical system. We first analyze the proposed system qualitatively, highlighting the key ideas behind our proposal and allowing for an intuitive understanding as to why SR arises in this scenario. Following this, we present a quantitative analysis of the phenomenon by means of exact numerical computation of the frequency response of the probe. We demonstrate that the SNR at the driving frequency is maximized at a certain noise level, an unambiguous signature of the occurrence of SR. Furthermore, to illustrate an application of SR in our proposed system, we discuss and simulate the noise-assisted precision measurement of the frequency difference of two coherent fields.
The scheme we present here is in principle easy to implement, as it relies entirely on techniques have been employed by experimentalists for more than 10 years. In Fig. 1 the system under consideration and the applied driving fields are shown. The probe consists of a four level atomic system subject to both coherent and incoherent radiation. The $`12`$ transition is driven by a resonant modulated coherent laser field of Rabi frequency $`\mathrm{\Omega }(t)`$ while the $`13`$ and $`24`$ transitions are driven by noisy fields with effective pump rates $`W_{33}`$ and $`W_{44}`$ respectively. The atomic level $`2`$ can decay with a rate $`2\mathrm{\Gamma }_{22}`$ and it is this radiation that will be detected. We assume that level $`4`$ is metastable and we neglect its spontaneous decay rate in the following. In addition, we assume that level $`3`$ can only decay, at a rate $`2\mathrm{\Gamma }_{33}`$, into level $`4`$. The fact that the coherent driving is modulated implies that the Rabi frequency and therefore the energy separation $`\mathrm{}\mathrm{\Omega }`$ between the dressed levels $`|\pm =1/\sqrt{2}(|1\pm |2)`$ are time dependent. Let us consider the simplest case where the modulated driving can be described by a step function, as illustrated in Fig. 2. Then the time dependence of the coherent Rabi frequency is given by
$$\mathrm{\Omega }(t)=\mathrm{\Omega }\pm \mathrm{\Delta }\mathrm{\Omega },$$
(1)
where $`\mathrm{\Delta }\mathrm{\Omega }\mathrm{\Omega }`$ (weak modulation), and where the $`(+)`$ sign holds for $`t[kt_m,(k+1/2)t_m]`$, k is a positive integer, and the $`()`$ sign corresponds to the remaining part of the modulation cycle, $`t_m`$ denoting the corresponding modulation period. Under these conditions, the relative position of the dressed states alternates between two possible configurations, as illustrated by the dashed (dotted) lines in Fig. 1. For the sake of clarity, we will refer to these two possible situations as strong laser (i. e. larger Rabi frequency; larger energy separation between dressed levels) and weak laser (i. e. smaller Rabi frequency; smaller energy separation between dressed levels) regimes.
We will now show in a qualitative way that the system we have described above may exhibit, when the broad band fields are suitably tuned, a dynamics of extended bright and dark periods in the resonance fluorescence intensity which are strongly synchronized with the modulated coherent driving field, as depicted in Fig. 2. However, synchronization is not the only signature of SR and we will show quantitatively later on that indeed the proposed scheme fulfills the necessary requirements for exhibiting SR.
Let us denote by $`\widehat{\omega }_i`$, $`(i=3,4)`$, the central frequency of each broad band field, while $`\omega _{i1}`$ refers to the corresponding natural frequency of the atomic transition involved. Let us now choose the detunings $`\mathrm{\Delta }_{i1}=\widehat{\omega }_i\omega _{i1}`$ as follows
$`\mathrm{\Delta }_{31}`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Omega }+\mathrm{\Delta }\mathrm{\Omega }}{2}}`$ (2)
$`\mathrm{\Delta }_{41}`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Omega }\mathrm{\Delta }\mathrm{\Omega }}{2}}.`$ (3)
This choice of detunings is schematically shown in Fig. 1. When the coherent driving operates in the weak mode, that is, when $`\mathrm{\Omega }(t)=\mathrm{\Omega }\mathrm{\Delta }\mathrm{\Omega }`$, the noisy field driving the $`13`$ transition is resonant with the dressed level $`|+`$ while the field driving the $`14`$ is off-resonant. Under these circumstances, the system is rapidly pumped from $`|1|3|4`$. It remains in level 4 and as a consequence no light is emitted from the atom, i.e. we are in a dark period, given that level $`|4`$ is metastable. On the other hand, when the coherent driving is operating in the strong mode, the noise field driving the $`14`$ becomes resonant with the dressed level $`|`$ while the driving $`13`$ becomes detuned. As a result, the system is pumped back and forth between level $`|4`$ and the dressed level $`|`$, with photons being emitted at a rate proportional to $`\mathrm{\Gamma }_{22}/2`$. Therefore, the system is in a bright period, i.e. the atom emits many photons. It can be understood quite easily how this synchronized dynamics depends on the noise intensity. If the effective pump rate $`W_{ii}`$, $`(i=3,4)`$, is too strong, e. g. $`W_{ii}\mathrm{\Delta }\mathrm{\Omega }`$, the pump rates become insensitive to the value of $`\mathrm{\Delta }\mathrm{\Omega }`$, synchronization is lost and the observed fluorescence would just show a constant intensity. A lack of synchronization should also be observable in the opposite regime, where $`W_{ii}`$ is too weak, e. g. $`W_{ii}<t_m^1`$, with the system exhibiting now extended bright and dark periods and an additional noise background in the fluorescence spectrum. Therefore, there seems to be an intermediate regime in which synchronization is optimal. We will demonstrate in the following that this optimal regime can be achieved and how in fact the system may exhibit stochastic resonance. To clearly identify SR we have to compute the autocorrelation function of the intensity emitted by the atom. We simplify it to a binary process, assuming value $`1`$ if the intensity exceeds a certain threshold (e.g. $`10\%`$ of the intensity expected in a bright period), and $`0`$ for intensities below the threshold. We then compute the power spectrum of this process which defines the spectral response of the system to a periodic perturbation. For that, we will have to compute explicitly the normalized spectrum and the corresponding SNR at the driving frequency. The starting point for evaluating these quantities is provided by the system’s master equation, whose relevant terms , under exact resonance for the $`12`$ transition, are as follow:
$`\dot{\rho }_{++}`$ $`=`$ $`(2W_{33}+{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}})\rho _{++}+{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _{}+2W_{33}\rho _{33}`$ (4)
$`\dot{\rho }_{}`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho ++(2W_{44}+{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}})\rho _{}+2W_{44}\rho _{44}`$ (5)
$`\dot{\rho }_+`$ $`=`$ $`\mathrm{\Gamma }_{22}(\rho _{++}+\rho _{})`$ (6)
$``$ $`(W_{33}+W_{44}{\displaystyle \frac{3\mathrm{\Gamma }_{22}}{2}}+i{\displaystyle \frac{\mathrm{\Omega }}{2}})\rho _+{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _+`$ (7)
$`\dot{\rho }_{33}`$ $`=`$ $`2W_{33}\rho _{++}(2W_{33}2\mathrm{\Gamma }_{33})\rho _{33}`$ (8)
$`\dot{\rho }_{44}`$ $`=`$ $`=\dot{\rho }_{++}\dot{\rho }_{}\dot{\rho }_{33},\dot{\rho }_+=\dot{\rho }_+^{}`$ (9)
The last line arises from the preservation of trace and the hermiticity of the density operator. It should be stressed that this master equation is valid for a certain range of parameters in which the broadband assumption for the noise fields is correct, i.e. we can replace the effect of the noise by a simple pump rate $`W_{ii}`$. This approximation is valid if the frequency bandwidth $`\mathrm{\Delta }\omega _i`$ $`(i=3,4)`$ of the noise field is larger than the spontaneous decay rates in the system and the detunings do not greatly exceed the bandwidth of the noise field. It should also be noted here that we are working in a dressed state picture, in which the incoherent pump rates are between the dressed level $`|\pm `$ and the level $`3`$ and $`4`$. This assumption is only valid if the Rabi frequency of the driving field on the $`12`$ transition is large compared to the bandwidth of the noise fields. This condition is intuitively clear, as in that case a noise field can selectively address only one dressed state. Therefore, our analysis applies provided that the following inequality holds
$$\mathrm{\Gamma }_{ii}\mathrm{\Delta }\omega _i\mathrm{\Omega },(i=3,4).$$
(10)
The master equation contains all the information about the dynamics of the system; however, if we are mainly interested in the behaviour of a single quantum system, e.g. a single ion in an ion trap, then it is more convenient to use the quantum jump approach ( see and references therein). The main idea behind the quantum jump approach is to determine the time evolution of the system under the condition that no photon has been emitted on the $`12`$ transition. This conditional time evolution is no longer trace preserving and the decreasing trace reflects the probability that no photon has been emitted in the time interval $`[0,t]`$. The conditional time evolution can easily be obtained either directly from the master equation (removing some terms) or by rederiving it under the constraint that no photon has been emitted . The key point is realizing that in order to obtain the conditional time evolution one has to remove from the ensemble all those systems that have emitted a photon. This can be done heuristically by removing the contribution $`2\mathrm{\Gamma }_{22}\rho _{22}`$ from the time evolution equation for the population of the ground state $`\rho _{11}`$. When using dressed states $`|\pm `$, the result is that we need to replace Eqs. (4,5,7) by
$`\dot{\rho }_{++}`$ $`=`$ $`(2W_{33}+\mathrm{\Gamma }_{22})\rho _{++}`$ (11)
$`+`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _++{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _++2W_{33}\rho _{33}`$ (12)
$`\dot{\rho }_{}`$ $`=`$ $`(2W_{44}+\mathrm{\Gamma }_{22})\rho _{}`$ (13)
$`+`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _++{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}\rho _++2W_{44}\rho _{44}`$ (14)
$`\dot{\rho }_+`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }_{22}}{2}}(\rho _{++}+\rho _{})`$ (15)
$``$ $`(W_{33}+W_{44}+\mathrm{\Gamma }_{22}+i{\displaystyle \frac{\mathrm{\Omega }}{2}})\rho _+`$ (16)
Given the conditional time evolution, the evaluation of the normalized power spectrum is a straightforward task. For the modulation described in Eq. (1) numerical results are presented in Fig. 3. This simulation shows the expected behaviour of a resonant-like process, with a sharp peak (a delta function within numerical precision) at the frequency of the modulation and subsequent weaker peaks at its odds harmonics. The peaks at even harmonics are suppressed due to the symmetry of the modulation. The emergence of delta peaks in the spectrum is a first indication of SR.
In order to establish unambiguously the occurrence of SR, one has to evaluate the signal-to-noise ratio. This quantity, defined as the ratio of the spectral peak to the spectral background at a given frequency, is a measure of the probe sensitivity to the periodic driving. Fig. 4 shows the SNR at the modulation frequency, i.e. the frequency at which the power spectrum exhibits a delta peak. As expected, the SNR exhibits a maximum at an optimal noise pump rate. If the noise intensity is increased beyond this value, the sensitivity of the probe is diminished. Although we have presented numerical results for a specific choice of parameters, extensive numerical simulations show that SR can be observed over a wide range of parameters.
The scheme discussed above may seem academic, but it allows us to exemplify how stochastic resonance could be observed at the level of a single quantum system using currently available experimental techniques. Moreover, we will now show that this phenomenon may find practical applications in precision measurements. To this end
we will discuss how the sensitivity for the detection of a frequency mismatch between two coherent fields can be enhanced using incoherent driving, i.e., employing SR.
Let us consider two coherent fields with Rabi frequencies $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$, whose oscillation frequencies differ by a small amount $`\delta \omega `$. As it is well known, the superposition of two laser fields in a running wave configuration results in a beating signal. The idea is to use this modulated signal as our driving field (giving rise to the dressed states $`|\pm `$) while tuning appropriately the additional incoherent driving fields (see Fig. 1 for illustration). Numerical simulations of the power spectrum for experimentally accessible parameters reveal again that the frequency response of the atomic system exhibits stochastic resonance. The signal to noise ratio for the first delta peak in the power spectrum is shown in Fig. 5. The optimal performance, that is, the largest SNR, is achieved for certain finite but non-zero noise intensity. This result implies that stochastic resonance can have applications for example in frequency measurements.
Summarizing, we have showed that the phenomenon of SR, a paradigm of the counter-intuitive role that noise may play in high sensitivity detection, can be demonstrated at the level of a single ion. As an illustration of the potential that this effect may have, we have discussed the use of a SR scheme for the detection of the frequency difference of two lasers. As the proposed experimental scenario relies on techniques well mastered by quantum opticians, quantum SR may be expected to open up new experimental possibilities in precision measurements at the quantum limit.
This work was supported by The Leverhulme Trust, the UK EPSRC and the European Union. We would also like to thank D.M. Segal and R.C. Thompson and their experimental ion trapping group at Imperial College for interesting discussions and P.L. Knight for helpful comments on the manuscript. |
warning/0001/hep-ph0001003.html | ar5iv | text | # THE LIFETIMES OF HEAVY FLAVOUR HADRONS - A CASE STUDY IN QUARK-HADRON DUALITY 11footnote 1Invited talk given at the 3rd International Conference on B Physics and CP Violation, Taipeh, Taiwan, Dec. 3 - 7, 1999
## 1 Introduction
The lifetime of a hadron represents an observable of fundamental as well as practical importance: (i) Its magnitude reveals whether the decay is driven by strong, electromagnetic or weak forces; (ii) it constitutes an essential engineering number for translating measured branching ratios into widths. Yet a strong motivation to measure a quantity does not necessarily imply a need for a precise theoretical description. Furthermore we all understand that nothing that is going to happen or not happen in the theory of weak lifetimes will make anybody abandon QCD as the theory of the strong interactions. After all, it is the only game in town after string theory has raised its ambition to become the theory of everything rather than merely the theory of the strong forces where it had first emerged.
The central theme of my talk will be that developing such a theory represents a forum for addressing the next frontier in QCD, namely quark-hadron duality or duality for short. The concept of duality constitutes an essential element in any QCD based description and it has been invoked since the early days of the quark model. For a long time little progress happened in this area; for a violation of duality can be discussed in a meaningful way only if one has a reliable theoretical treatment of nonperturbative effects.
Let me illustrate that through an example. A priori it would be quite reasonable to assume that relating the weak decay width of a heavy flavour hadron to the fifth power of its mass rather than the heavy quark mass – $`\mathrm{\Gamma }(H_Q)M^5(H_Q)`$ – would incorporate boundstate effects as the leading nonperturbative corrections (and that is indeed what we originally did ). Only after developing a consistent theory for the weak decays of such hadrons through the operator product expansion did we realize that such an ansatz would violate duality. For it leads to a large correction of order $`1/m_Q`$$`M^5(H_Q)(m_Q+\overline{\mathrm{\Lambda }})^5m_Q^5(1+5\overline{\mathrm{\Lambda }}/m_Q)`$ – which is anathema to the OPE !
This example already indicates that the study of heavy flavour decays had given new impetus to addressing duality: it has provided us with new theoretical tools, and it has re-emphasized the need to understand the limitations to duality since one aims at extracting fundamental KM parameters with high numerical accuracy from semileptonic decays.
Nonleptonic transitions provide a rich and multilayered lab to analyze duality and its limitations; they can act as a microscope exactly because they are thought of containing larger duality violations than semileptonic reactions.
## 2 Heavy Quark Expansions
The weak decay width for a heavy flavour hadron $`H_Q`$ into an inclusive final state $`f`$ can be expressed through an operator product expansion (OPE) :
$$\mathrm{\Gamma }(H_Qf)=\frac{G_F^2m_Q^5}{192\pi ^3}|V_{CKM}|^2[c_3^{(f)}H_Q|\overline{Q}Q|H_Q+c_5^{(f)}\frac{\mu _G^2(H_Q)}{m_Q^2}+$$
$$+\underset{i}{}c_{6,i}^{(f)}\frac{H_Q|(\overline{Q}\mathrm{\Gamma }_iq)(\overline{q}\mathrm{\Gamma }_iQ)|H_Q}{m_Q^3}+𝒪(1/m_Q^4)],$$
(1)
where $`\mu _G^2(H_Q)H_Q|\overline{Q}\frac{i}{2}\sigma GQ|H_Q`$. Eq.(1) exhibits the following important features:
* The expansion involves
+ c-number coefficients $`c_i^{(f)}`$ calculable within short-distance physics;
+ expectation values of local operators given by long distance physics; their values can be inferred from symmetry arguments, other observables, QCD sum rules, lattice studies and quark models;
+ inverse powers of the heavy quark mass $`m_Q`$ scaling with the known dimensions of the various operators.
The nonperturbative effects on the decay width – a dynamical quantity – can thus be expressed through expectation values and quark masses. Those being static quantities can be calculated with decent reliability.
* A crucial element of Wilson’s prescription for this expansion is that it allows a selfconsistent separation of short-distance dynamics that is lumped into the coefficients $`c_i^{(f)}`$ and long-distance dynamics that enters through the expectation values of local operators. This is achieved through the introduction of the auxiliary scale $`\mu `$ that enters both in the coefficients and the matrix elements. As a matter of principle observables have to be independant of $`\mu `$ since Nature cannot be sensitive to how we arrange our computational tasks. In practise, however, $`\mu `$ has to be chosen judiciously for those very tasks: on one hand one would like to choose $`\mu `$ as high as possible to obtain a reliable perturbative expression in powers of $`\alpha _S(\mu )`$; on the other hand one likes to have it as low as possible to evaluate the expectation values in powers of $`\mu /m_Q`$. This ‘Scylla and Charybdis’ dilemma can be tackled by choosing $`\mu 1\text{GeV}`$. For simplicity I will not state the dependance on $`\mu `$ explicitely although it is implied.
* The free quark model or spectator expression emerges asymptotically (for $`m_Q\mathrm{}`$) from the $`\overline{Q}Q`$ operator since $`H_Q|\overline{Q}Q|H_Q=1+𝒪(1/m_Q^2)`$.
* No $`𝒪(1/m_Q)`$ contribution can arise in the OPE since there is no independant dimension four operator (with colour described by a local gauge theory) <sup>2</sup><sup>2</sup>2The operator $`\overline{Q}i\mathit{}Q`$ can be reduced to the leading operator $`\overline{Q}Q`$ through the equation of motion.. This has two important consequences:
+ With the leading nonperturbative corrections arising at order $`1/m_Q^2`$, their size is typically around 5 % in beauty decays. They had not been anticipated in the phenomenological descriptions of the 1980’s.
+ A $`1/m_Q`$ contribution can arise only due to a massive duality violation. Thus one should set a rather high threshold before accepting such a statement.
* Pauli Interference (PI) , Weak Annihilation (WA) for mesons and W-scattering (WS) for baryons arise unambiguously and naturally in order $`1/m_Q^3`$ with WA being helicity suppressed . They mainly drive the differences in the lifetimes of the various hadrons of a given heavy flavour.
The expectation values of $`\overline{Q}Q`$ and $`\overline{Q}\frac{i}{2}\sigma GQ`$ are known with sufficient accuracy for the present purposes from the hyperfine splittings and the charm and beauty hadron masses .
The largest uncertainties enter in the expectation values for the dimension-six four-fermion operators in order $`1/m_Q^3`$. For mesons I will invoke approximate factorization at a low scale of around 1 GeV. One should note that factorizable contributions at a low scale $``$ 1 GeV will be partially nonfactorizable at the high scale $`m_Q`$!
For baryons there is no concept of factorization, and we have to rely on quark model results.
Below I will discuss mainly hadron-specific duality violations affecting the ratios between different hadrons of a given heavy flavour.
## 3 Lifetimes of Charm Hadrons
One rough measure for the numerical stability of the $`1/m_c`$ expansion is provided by $`\sqrt{\mu _G^2(D)/m_c^2}0.5`$ as an effectice expansion parameter which is not small compared to one. Obviously one can expect – at best – a semiquantitative description. The mesonic four-quark matrix elements are calibrated by $`f_D200`$ MeV and $`f_{D_s}/f_D1.11.2`$.
On general grounds one expects the following hierarchy in lifetimes :
$$\tau (D^+)>\tau (D^0)\tau (D_s^+)\tau (\mathrm{\Xi }_c^+)>\tau (\mathrm{\Lambda }_c^+)>\tau (\mathrm{\Xi }_c^0)>\tau (\mathrm{\Omega }_c)$$
(2)
Table 1 shows the predictions and data.
A few comments are in order here:
* You apply the $`1/m_c`$ expansion at your own risk. It is easy to list reasons why it should fail to reproduce even the qualitative pattern expressed in Eq.(2). However comparing the data with the expectations shows agreement even on the semiquantitative level. This could be accidental; yet I will explore the possibility that it is not. One should note that the longest and shortest lifetimes differ by a factor of about twenty!
* PI is the main engine driving the $`D^+D^0`$ lifetime difference as already anticipated in the ‘old’ analysis of Guberina et al. ; the main impact of the HQE for this point was to show that WA cannot constitute the leading effect and that $`BR_{SL}(D^0)7\%`$ is consistent with PI being the leading effect, see below <sup>3</sup><sup>3</sup>3$`\mathrm{\Gamma }(D^+)`$ is guaranteed to remain positive if the range in momentum over which PI can occur is properly evaluated. To put it differently: while one cannot count on obtaining a reliable value for $`\mathrm{\Gamma }(D^+)`$, a nonsensical result will arise only if one makes a mistake.. In quoting a lifetime ratio of $`2`$ I am well aware that the measured value is different from two. Yet that numerical difference is within the theoretical noise level: one could use $`f_D=220`$ MeV rather than 200 MeV and WA, which has been ignored here, could account for 10 - 20 % of the $`D^0`$ width.
* Since $`\tau (D_s)/\tau (D^0)1.07`$ can be generated without WA , the ‘old’ data on $`\tau (D_s)/\tau (D^0)`$ had provided an independant test for WA not being the leading source for $`\tau (D^0)\tau (D^+)`$; it actually allowed for it being quite irrelevant. The ‘new’ data reconfirm the first conclusion; at the same time they point to WA as a still significant process. A recent analysis of WA relying on QCD sum rules is not quite able to reproduce the observed lifetime ratio; further analysis along these lines is called for.
* The $`1/m_c^2`$ contribution controlled by $`\mu _G^2(D)`$ reduces the semileptonic width common to $`D^0`$ and $`D^+`$ mesons; this brings the absolute values observed for $`BR_{SL}(D^0)`$ and $`BR_{SL}(D^+)`$ in line with what is expected when it is mainly PI that generates the $`D^+D^0`$ lifetime difference.
* The description of the baryonic lifetimes is helped by the forgiving experimental errors. More accurate measurements of $`\tau (\mathrm{\Xi }_c^{+,0},\mathrm{\Omega }_c)`$ are needed. They might well exhibit deficiencies in the theoretical description.
* Nonuniversal semileptonic widths – $`\mathrm{\Gamma }_{SL}(D)\mathrm{\Gamma }_{SL}(\mathrm{\Lambda }_c)\mathrm{\Gamma }_{SL}(\mathrm{\Xi }_c)\mathrm{\Gamma }_{SL}(\mathrm{\Omega }_c)`$ – are predicted with the main effect being constructive PI in $`\mathrm{\Xi }_c`$ and $`\mathrm{\Omega }_c`$ decays; the lifetime ratios among the baryons will thus not get reflected in their semileptonic branching ratios; one estimates
$`BR_{SL}(\mathrm{\Xi }_c^0)BR_{SL}(\mathrm{\Lambda }_c)`$ $``$ $`\tau (\mathrm{\Xi }_c^0)0.5\tau (\mathrm{\Lambda }_c)`$ (3)
$`BR_{SL}(\mathrm{\Xi }_c^+)2.5BR_{SL}(\mathrm{\Lambda }_c)`$ $``$ $`\tau (\mathrm{\Xi }_c^+)1.3\tau (\mathrm{\Lambda }_c)`$ (4)
$`BR_{SL}(\mathrm{\Omega }_c)`$ $`<`$ $`15\%`$ (5)
* On general grounds one expects $`\mathrm{\Delta }\mathrm{\Gamma }(D^0)/\mathrm{\Gamma }(D^0)\mathrm{tg}^2\theta _CSU(3)_{Fl}`$ breaking $`𝒪(0.01)`$. If the data show that the lifetime difference for the two neutral $`D`$ mass eigenstates is significantly below this bound, one would have learned an intriguing lesson on duality.
## 4 Lifetimes of Beauty Hadrons
### 4.1 Orthodoxy
The numerical stability of the $`1/m_b`$ expansion is characterised by $`\sqrt{\mu _G^2(B)/m_b^2}`$ $`0.131`$; i.e. such an expansion should yield rather reliable numerical results. Merely reproducing the qualitative pattern would be quite unsatisfactory. I will also use $`f_B200`$ MeV and $`f_{B_s}/f_B1.11.2`$. Table 2 shows predictions and data.
Again some comments to elucidate these findings:
* The original predictions for the meson lifetimes, which had encountered theoretical criticism , are on the mark. (i) A recent lattice study finds a result quite consistent with the original work based on factorization :
$$\frac{\tau (B^+)}{\tau (B_d)}=1.03\pm 0.02\pm 0.03$$
(6)
(ii) Sceptics will argue that predicting lifetime ratios close to unity is not overly impressive. In response one should point out that the largest lifetime difference by far – $`\frac{\tau (B_c)}{\tau (B_d)}\frac{1}{3}`$ – has been predicted correctly and that the absence of contributions $`𝒪(1/m_Q)`$ had been crucial there!
* A serious challenge arises from the ‘short’ baryon lifetime. In terms of $`\mathrm{\Delta }1\tau (\mathrm{\Lambda }_b)/\tau (B_d)`$ the data can be expressed by
$$\mathrm{\Delta }_{\mathrm{experim}}=0.21\pm 0.05$$
(7)
A detailed analysis of quark model calculations finds however
$$\mathrm{\Delta }_{\mathrm{theor}.}=0.030.12$$
(8)
Reanalyses by other authors agree with Eq.(8) – as does a pilot lattice study : $`\mathrm{\Delta }_{\mathrm{lattice}}=(0.070.09)`$. A recent analysis based on QCD sum rules arrives at a significantly larger value: $`\mathrm{\Delta }_{\mathrm{QCDSR}}=0.130.21`$ . If true it would remove the problem. However, I would like to understand better how the sum rules analysis can differ so much from other studies given that it still uses the valence quark approximation.
* An essential question for future studies concerns the lifetimes of the beauty hyperons $`\mathrm{\Xi }_b^{,0}`$. On general grounds one expects $`\tau (\mathrm{\Xi }_b^{})>\tau (\mathrm{\Lambda }_b)`$, $`\tau (\mathrm{\Xi }_b^0)`$ . More specifically, using the observed charm hyperon lifetimes and $`SU(3)`$ symmetry a very sizeable effect has been predicted :
$$\frac{\tau (\mathrm{\Xi }_b^{})\tau (\mathrm{\Lambda }_b)}{\tau (\mathrm{\Lambda }_b)}0.140.21$$
(9)
### 4.2 Heresy
As said before, the ansatz $`\mathrm{\Gamma }(H_Q)M(H_Q)^5`$ which would yield $`\tau (\mathrm{\Lambda }_b)/\tau (B_d)`$ $`0.75`$ and therefore has been re-surrected is anathema to the OPE since it would imply the nonperturbative corrections to be of order $`1/m_Q`$! The $`B^{}B_d`$ lifetime difference is still a $`𝒪(1/m_b^3)`$ effect. <sup>4</sup><sup>4</sup>4 It has been shown (at least for semileptonic transitions) that duality is implemented as follows: ‘quark phase space + nonperturbative corrections $`\widehat{=}`$ hadronic phase space + boundstate effects’!
Notwithstanding my employer I am willing to consider heresy, though, since it makes some further prediction that differ from the OPE findings:
$$\frac{\overline{\tau }(B_s)}{\tau (B_d)}=\left(\frac{M(B_d)}{M(B_s)}\right)^50.94$$
(10)
$$\tau (\mathrm{\Lambda }_b)/\tau (\mathrm{\Xi }_b^0)/\tau (\mathrm{\Xi }_b^{})1/0.85/0.85;$$
(11)
the expectation $`M(\mathrm{\Xi }_b)M(\mathrm{\Lambda }_b)M(\mathrm{\Xi }_c)M(\mathrm{\Lambda }_c)`$ has been used in Eq.(11).
On the down side I do not see how such an ansatz can yield a correct prediction for $`\tau (B_c)`$ in a natural way.
One can also add that such an ansatz helps to understand neither the pattern nor the size of the lifetime differences in the charm sector. Agreement with the data can be enforced, though, by adjusting – in an ad-hoc fashion I would say – the contributions from PI, WA and WS.
## 5 On Quark-Hadron Duality
### 5.1 General Remarks
Duality has been an early and somewhat fuzzy element of quark model arguments. It can be expressed as follows: ”Rates evaluated on the parton level ‘approximate’ observable rates summed over a ‘sufficient number’ of hadronic channels.” It was never stated clearly how good an approximation it provided and what is meant by ‘sufficient number’; it was thought, though, that this number had to be larger than of order unity.
Heavy quark theory has opened up new theoretical tools as well as perspectives onto duality; it demonstrated that duality can hold even with one or two channels dominating – if an additional feature like heavy quark symmetry intervenes . This has been demonstrated for semileptonic $`bc`$ decays. The goal is to understand better the origins of limitations to duality and to develop some quantitative measures for it. The new tools that are being brought to bear on this problem are (a) the OPE; (b) the so-called small velocity sum rules and (c) the ’t Hooft model.
The results obtained so far show there are different categories of duality – local vs. global etc. duality – depending on the amount of averaging or ‘smearing’ that is involved and that duality can neither be universal nor exact.
Duality is typically based on dispersion relations expressing observable rates averaged over some kinematical variables through an OPE constructed in the Euclidean region. There are natural limitations to the accuracy of such an expansion; among other things it will have to be truncated. In any case, such a power expansion will exhibit no sensitivity to a term like $`exp(m_Q/\mathrm{\Lambda }_{QCD})`$. Yet upon analytical continuation from the Euclidean to the Minskowskian domain this exponentially suppressed term turns into sin$`(m_Q/\mathrm{\Lambda }_{QCD})`$ – which is not surpressed at all! I.e., the OPE cannot account for such terms that could become quite relevant in Minkowski space and duality violations can thus enter through these ‘oscillating’ terms; the opening of thresholds provides a model for such a scenario. Actually they will be surpressed somewhat like $`(1/m_Q^k)\mathrm{sin}(m_Q/\mathrm{\Lambda }_{QCD})`$ with the power $`k`$ depending on the dynamics in general and the reaction in particular. This could produce also a ‘heretical’ $`1/m_Q`$ contribution from a dimension-five operator:
$$\frac{1}{m_Q^2}\mathrm{sin}\left(\frac{m_Q}{\mathrm{\Lambda }_{QCD}}\right)=𝒪(1/m_Q)$$
(12)
The colour flow in semileptonic as well as nonleptonic spectator decays and in WA is such that duality can arise naturally; i.e., nature had to be malicious to create sizeable duality violations. Yet in PI – because it is an interference phenomenon – the situation is much more complex leading to serious concerns about the accuracy with which duality can apply here.
### 5.2 ’t Hooft Model
The most relevant features of the ’t Hooft model are: (1) QCD in 1+1 dimensions obviously confines. (2) It is solvable for $`N_C\mathrm{}`$: its spectrum of narrow resonances can be calculated as can their wavefunctions.
Duality can then be probed directly by comparing the width evaluated through the OPE with a sum over the ‘hadronic’ resonances appropriately smeared over the threshold region:
$$\mathrm{\Gamma }_{OPE}(H_Q)\underset{n}{}\mathrm{\Gamma }(H_Qf_n)$$
(13)
Such a program has been first pursued using numerical methods; it lead to claims that duality violations arise in the total width through a $`1/m_Q`$ term and quantitatively more massively in WA .
However an analytical study has shown that neither of these claims is correct: perfect matching of the OPE expression with the sum over the hadronic resonances was found through high order in $`1/m_Q`$ . The same result was obtained for the more intriguing case of PI .
## 6 Summary and Outlook
A mature formalism genuinely based on QCD has been developed for describing inclusive nonleptonic heavy flavour decays. It can tackle questions that could not be addressed before in a meaningful way; even failures can teach us valuable lessons on nonperturbative aspects of QCD, namely on limitations to duality.
A fairly successful semiquantitative picture has emerged for the lifetimes of charm hadrons considering the wide span characterised by $`\tau (D^+)/\tau (\mathrm{\Omega }_c)20`$; while this might be a coincidence, it should be noted:
* PI provides the leading effect driving the $`D^+D^0`$ lifetime difference; this conclusion is fully consistent with the absolute value for $`BR_{SL}(D^0)`$.
* This year’s precise new experimental result
$$\frac{\tau (D_s)}{\tau (D^0)}=1.211\pm 0.017$$
(14)
confirms this picture: WA is nonleading, though still significant. It represents an interesting challemge to find the footprint of WA in some classes of exclusive final states.
* More precise data on the $`\mathrm{\Xi }_c^{0,+}`$ and $`\mathrm{\Omega }_c`$ lifetimes are very much needed. Those might reveal serious deficiencies in the predictions. It should be noted that the semileptonic widths for baryons are not universal!
The scorecard for beauty lifetimes looks as follows:
* The predictions for $`\tau (B^{})/\tau (B_d)`$ and even more impressively for $`\tau (B_c)`$ appear to be borne out by experiment.
* The jury is still out on $`\tau (B_s)`$.
* The $`\mathrm{\Lambda }_b`$ lifetime provides a stiff challenge to theory. It should be noted that most authors have shown a remarkable lack of flexibility in accommodating $`\tau (\mathrm{\Lambda }_b)/\tau (B_d)<0.9`$, which is quite unusual in this line of work. Maybe experiment will show more flexibility.
* Accurate data on $`\tau (\mathrm{\Xi }_b^{,0})`$ will be essential to celebrate success or diagnose failure.
Having developed a theoretical framework for treating nonperturbative effects, we can address the issue of duality violations. While we have begun to understand better their origins, we have not (yet) found a model theory that could explain the short $`\mathrm{\Lambda }_b`$ lifetime as a duality limitation.
There are, of course, different layers of failure conceivable and the lessons one would have to draw:
* A refusal by the data to move up the value of $`\tau (\mathrm{\Lambda }_b)`$ could be interpreted as showing that the quark model provides a very inadequate tool to estimate baryonic expectation values.
* A low value of the average $`B_s`$ lifetime – say $`\overline{\tau }(B_s)<0.96\tau (B_d)`$ – had to be seen as a very significant limitation to duality.
Clearly there is a lot we will learn from future data on lifetimes and other inclusive rates – one way or the other.
Acknowledgements This truly inspiring and enjoyable meeting organized by Profs. H.-Y. Cheng and G. Hou clearly wetted the appetite of participants for the next incarnation of this series, namely BCP 4. I am grateful to my collaborators Profs. M. Shifman, N. Uraltsev and A. Vainshtein for generously sharing their insights with me. This work has been supported by the NSF under the grant PHY 96-0508. |
warning/0001/nucl-th0001036.html | ar5iv | text | # F-spin as a Partial Symmetry
## Abstract
We use the empirical evidence that F-spin multiplets exist in nuclei for only selected states as an indication that F-spin can be regarded as a partial symmetry. We show that there is a class of non-F-scalar IBM-2 Hamiltonians with partial F-spin symmetry, which reproduce the known systematics of collective bands in nuclei. These Hamiltonians predict that the scissors states have good F-spin and form F-spin multiplets, which is supported by the existing data.
The interacting boson model (IBM-2) describes collective low-lying states in even-even nuclei in terms of monople ($`s_\rho `$) and quadrupole ($`d_\rho `$) proton ($`\rho =\pi `$) and neutron ($`\rho =\nu `$) bosons. Microscopic, shell-model-based interpretation of the model suggests that the number of bosons of each type ($`N_\rho `$) is fixed and is taken as the sum of valence proton and neutron particle and hole pairs counted from the nearest closed shell. The proton-neutron degrees of freedom are naturally reflected in the IBM-2 via an SU(2) F-spin algebra with generators $`\widehat{F}_+=s_\pi ^{}s_\nu +d_\pi ^{}\stackrel{~}{d}_\nu `$, $`\widehat{F}_{}=(\widehat{F}_+)^{}`$, $`\widehat{F}_0=(\widehat{N}_\pi \widehat{N}_\nu )/2`$. The basic F-spin doublets are $`(s_\pi ^{},s_\nu ^{})`$, and $`(d_{\pi \mu }^{},d_{\nu \mu }^{})`$, with F-spin projection +1/2 ($``$1/2) for proton (neutron) bosons. In a given nucleus, with fixed $`N_\pi `$, $`N_\nu `$, all states have the same value of $`F_0=(N_\pi N_\nu )/2`$, while the allowed values of the F-spin quantum number F range from $`|F_0|`$ to $`F_{max}(N_\pi +N_\nu )/2N/2`$ in unit steps. F-spin characterizes the $`\pi `$-$`\nu `$ symmetry properties of IBM-2 states. States with maximal F-spin, $`FF_{max}`$, are fully symmetric and correspond to the IBM-1 states with only one type of bosons . There are several arguments, e.g., the empirical success of IBM-1, the identification of F-spin multiplets \[4-7\] (series of nuclei with constant $`F`$ and varying $`F_0`$ with nearly constant excitation energies), and weakness of M1 transitions, which lead to the belief that low lying collective states have predominantly $`F=F_{max}`$ . States with $`F<F_{max}`$, correspond to ‘mixed-symmetry’ states, most notably, the orbital magnetic dipole scissors mode has by now been established experimentally as a general phenomena in deformed even-even nuclei .
Various procedures have been proposed to estimate the F-spin purity of low lying states . These involve exploiting the data on M1 transitions (which should vanish between pure $`F=F_{max}`$ states), extracting the difference in proton-neutron deformations from pion charge exchange , using ratios of $`\gamma `$ and ground band magnetic moments and the experimental g-factors of $`2_1^+`$ states , and considering the excitation energy of mixed symmetry states. In the majority of analyses the F-spin admixtures in low lying states are found to be of a few percents ($`<10\%`$), typically $`2\%4\%`$ . In spite of its appeal, however, F-spin cannot be an exact symmetry of the Hamiltonian. The assumption of F-spin scalar Hamiltonians is at variance with the microscopic interpretation of the IBM-2, which necessitates different effective interactions between like and unlike nucleons. Furthermore, if F-spin was a symmetry of the Hamiltonian, then all states would have good F-spin and would be arranged in F-spin multiplets. Experimentally this is not case. As noted in an analysis of rare earth nuclei, the ground bands are in F-spin multiplets, whereas the vibrational $`\beta `$ bands and some $`\gamma `$ bands do not form good F-spin multiplets. The empirical situation in the deformed Dy-Os region is portrayed in Table I and Fig. 1. From Table I it is seen that, for $`F>13/2`$, the energies of the $`L=2^+`$ members of the $`\gamma `$ bands vary fast in the multiplet and not always monotonically. The variation in the energies of the $`\beta `$ bands is large and irregular. Thus both microscopic and empirical arguments rule out F-spin invariance of the Hamiltonian. F-spin can at best be an approximate quantum number which is good only for a selected set of states while other states are mixed. We are thus confronted with a situation of having ‘special states’ endowed with a good symmetry which does not arise from invariance of the Hamiltonian. These are precisely the characteristics of a “partial symmetry” for which a non-scalar Hamiltonian produces a subset of special (at times solvable) states with good symmetry. Such a symmetry notion was recently applied to nuclei , to molecules and to the study of mixed systems with coexisting regularity and chaos . Previously determined non-F-scalar Hamiltonians were shown to have solvable ground bands with good F-spin. It is the purpose of this Letter to analyze in detail these Hamiltonians and to show that their partial F-spin symmetry reproduces the known systematics of ground and excited bands. In particular, we find F-spin multiplets in only selected bands, and observe common collective signatures for the ground and scissors bands in deformed nuclei, e.g. the same F-spin purity and equal moments of inertia. We further test a prediction for the existence of F-spin multiplets of scissors states.
The ground band in the IBM-2 is represented by an intrinsic state which is a product of a proton condensate and a rotated neutron condensate with $`N_\pi `$ and $`N_\nu `$ bosons respectively . It depends on the quadrupole deformations, $`\beta _\rho ,\gamma _\rho `$, ($`\rho =\pi ,\nu `$) of the proton-neutron equilibrium shapes and on the relative orientation angles $`\mathrm{\Omega }`$ between them. For $`\beta _\rho >0`$, the intrinsic state is deformed and members of the rotational ground-state band are obtained from it by projection. It has been shown in that the intrinsic state will have a well defined F-spin, $`F=F_{max}`$, when the proton-neutron shapes are aligned and with equal deformations. The conditions ($`\beta _\pi =\beta _\nu `$,$`\gamma _\pi =\gamma _\nu `$, $`\mathrm{\Omega }=0`$) are weaker than the conditions for F-spin invariance, which makes it possible for a non-F-scalar IBM-2 Hamiltonian to have an equilibrium intrinsic state with pure F-spin. Since the angular momentum projection operator is an F-spin scalar, the projected states of good L will also have good $`F=F_{max}`$. A non-F-spin scalar Hamiltonian which has the above equilibrium condensate as an eigenstate is therefore guaranteed to have a ground band with good F-spin symmetry. Such explicit construction of an IBM-2 Hamiltonian with partial F-spin symmetry was presented in for the most likely situation, namely, aligned axially symmetric (prolate) deformed shapes ($`\beta _\rho =\beta `$, $`\gamma _\rho =\mathrm{\Omega }=0`$). In this case, the equilibrium deformed intrinsic state for the ground band with $`F=F_{max}`$ has the form
$`|c;K=0`$ $``$ $`|N_\pi ,N_\nu =(N_\pi !N_\nu !)^{1/2}(b_{c,\pi }^{})^{N_\pi }(b_{c,\nu }^{})^{N_\nu }|0,`$ (1)
$`b_{c,\rho }^{}`$ $`=`$ $`(1+\beta ^2)^{1/2}(s_\rho ^{}+\beta d_{\rho ,0}^{}),`$ (2)
where $`K`$ denotes the angular momentum projection on the symmetry axis. The relevant IBM-2 Hamiltonian with partial F-spin symmetry can be transcribed in the form
$`H`$ $`=`$ $`{\displaystyle \underset{i}{}}{\displaystyle \underset{L=0,2}{}}A_i^{(L)}R_{i,L}^{}\stackrel{~}{R}_{i,L}+{\displaystyle \underset{L=1,2,3}{}}B^{(L)}W_L^{}\stackrel{~}{W}_L+C^{(2)}[R_{(\pi \nu ),2}^{}\stackrel{~}{W}_2+H.c.],`$ (3)
where $`H.c.`$ means Hermitian conjugate and the dot implies a scalar product. The $`R_{i,L}^{}`$ ($`L=0,2`$) are boson pairs with $`F=1`$ and $`(F_0=1,0,1)(i=\pi ,(\pi \nu ),\nu )`$, and $`W_L^{}`$ $`(L=1,2,3)`$ are F-spin scalar ($`F=0`$) boson pairs defined as
$`\begin{array}{cc}R_{\rho ,0}^{}=d_\rho ^{}d_\rho ^{}\beta ^2(s_\rho ^{})^2,\hfill & R_{(\pi \nu ),0}^{}=\sqrt{2}(d_\pi ^{}d_\nu ^{}\beta ^2s_\pi ^{}s_\nu ^{})\hfill \\ R_{\rho ,2}^{}=\sqrt{2}\beta s_\rho ^{}d_\rho ^{}+\sqrt{7}(d_\rho ^{}d_\rho ^{})^{(2)},\hfill & R_{(\pi \nu ),2}^{}=\beta (s_\pi ^{}d_\nu ^{}+s_\nu ^{}d_\pi ^{})+\sqrt{14}(d_\pi ^{}d_\nu ^{})^{(2)}\hfill \\ W_L^{}=(d_\pi ^{}d_\nu ^{})^{(L)}(L=1,3),\hfill & W_2^{}=s_\pi ^{}d_\nu ^{}s_\nu ^{}d_\pi ^{}\hfill \end{array}`$ (7)
with $`\rho =\pi ,\nu `$ and $`\stackrel{~}{R}_{i,L,\mu }=(1)^\mu R_{i,L,\mu }`$, $`\stackrel{~}{W}_{L,\mu }=(1)^\mu W_{L,\mu }`$. The pair operators satisfy $`R_{i,L,\mu }|c=W_{L,\mu }|c=0`$ and consequently, the condensate is a zero energy eigenstate of H for any choice of parameters $`A_i^{(L)},B^{(L)},C^{(2)}`$ and any any $`N_\pi ,N_\nu `$. When $`A_i^{(L)},B^{(L)},A_{\pi \nu }^{(2)}B^{(2)}(C^{(2)})^20`$, the above Hamiltonian is positive-definite and hence $`|c`$ is its exact ground state with $`F=F_{max}`$. $`H`$, however, is an F-spin scalar only when $`A_\pi ^{(L)}=A_\nu ^{(L)}=A_{\pi \nu }^{(L)},(L=0,2)`$ and $`C^{(2)}=0`$. We thus have a non-F-spin scalar Hamiltonian with a solvable (degenerate) ground band with $`F=F_{max}`$. The degeneracy can be lifted by adding to the Hamiltonian (F-spin scalar) SO(3) rotation terms which produce $`L(L+1)`$ type splitting but do not affect the wave functions. States in other bands can be mixed with respect to F-spin, hence the F-spin symmetry of H is partial. $`H`$ trivially commutes with $`\widehat{F}_0`$ but not with $`\widehat{F}_\pm `$. However, $`[H,\widehat{F}_\pm ]|c=0`$ does hold and therefore $`H`$ will yield F-spin multiplets for members of ground bands. On the other hand, states in other bands can have F-spin admixtures and are not compelled to form F-spin multiplets. These features which arise from the partial F-spin symmetry of the Hamiltonian are in line with the empirical situation as discussed above and as depicted in Table I and Fig. 1. It should be noted that the partial F-spin symmetry of H holds for any choice of parameters in Eq. (3). In particular, one can incorporate realistic shell-model based constraints, by choosing the $`A_\rho ^{(2)}`$ ($`\rho =\pi ,\nu `$) terms (representing seniority-changing interactions between like nucleons), to be small. For the special choice $`A_i^{(2)}=C^{(2)}=0`$ and $`B^{(1)}=B^{(3)}`$, $`H`$ of Eq. (3) becomes SO(5) scalar which commutes, therefore, with the SO(5) projection operator and hence produces F-spin multiplets with good SO(5) symmetry. Such multiplets were reported in the Yb-Os region of $`\gamma `$-soft nuclei .
The same conditions ($`\beta _\rho =\beta `$, $`\gamma _\rho =\mathrm{\Omega }=0`$) which resulted in $`F=F_{max}`$ for the condensate of Eq. (2), ensure also $`F=F_{max}1`$ for the intrinsic state representing the scissors band
$`|sc;K=1`$ $`=`$ $`\mathrm{\Gamma }_{sc}^{}|N_\pi 1,N_\nu 1,`$ (8)
$`\mathrm{\Gamma }_{sc}^{}`$ $`=`$ $`b_{c,\pi }^{}d_{\nu ,1}^{}d_{\pi ,1}^{}b_{c,\nu }^{}.`$ (9)
Here $`\mathrm{\Gamma }_{sc}^{}`$ is a $`F=0`$ deformed boson pair whose action on the condensate with $`(N2)`$ bosons produces the scissors mode excitation. Furthermore, the scissors intrinsic state (9) is an exact eigenstate of the following Hamiltonian, obtained from Eq. (3) for the special choice $`C^{(2)}=0`$ and $`B^{(1)}=B^{(3)}=2B^{(2)}2B`$
$`H^{}`$ $`=`$ $`{\displaystyle \underset{i}{}}{\displaystyle \underset{L=0,2}{}}A_i^{(L)}R_{i,L}^{}\stackrel{~}{R}_{i,L}+B\widehat{}_{\pi \nu }`$ (10)
The last term in Eq. (10) is the Majorana operator , related to the total F-spin operator by $`\widehat{}_{\pi \nu }=[\widehat{N}(\widehat{N}+2)/4\widehat{F}^2]`$, with eigenvalues $`k(Nk+1)`$ for states with $`F=F_{max}k`$. The Hamiltonian $`H^{}`$ is non-F-scalar but is rotational invariant. If we add to it an SO(3) rotation term, $`H^{}+\lambda \widehat{L}^2`$, ($`\widehat{L}=\widehat{L}_\pi +\widehat{L}_\nu `$), the resulting Hamiltonian will have a subset of solvable states which form the $`K=0`$ ground band ($`L=0,2,4,\mathrm{}`$) with $`F=F_{max}`$, and the $`K=1`$ scissors band ($`L=1,2,3\mathrm{}`$) with $`F=F_{max}1`$. The resulting spectrum is
$`\begin{array}{cc}E_g(L)=\lambda L(L+1)\hfill & (F=F_{max})\hfill \\ E_{sc}(L)=BN+\lambda L(L+1)\hfill & (F=F_{max}1)\hfill \end{array}`$ (13)
where the Majorana coefficient $`B`$ may depend on the boson numbers and deformation . It follows that for such Hamiltonians, with partial F-spin symmetry, both the ground and scissors band have good F-spin and have the same moment of inertia. The latter derived property is in agreement with the conclusions of a recent comprehensive analysis of the scissors mode in heavy even-even nuclei , which concluded that, within the experimental precisions ($``$ 10%), the moment of inertia of the scissors mode are the same as that of the ground band. It is the partial F-symmetry of the Hamiltonian (10) which is responsible for the common signatures of collectivity in these two bands.
The Hamiltonian $`H^{}`$ of Eq. (10) is not F-spin invariant, however, $`[H^{},\stackrel{}{F}]|c;K=0=[H^{},\stackrel{}{F}]|sc;K=1=0`$. This implies that members of both the ground and scissors bands are expected to form F-spin multiplets. For ground bands such structures have been empirically established \[4-7\]. The prediction for F-spin multiplets of scissors states requires further elaboration. Although the mean energy of the scissors mode is at about 3 MeV , the observed fragmentation of the M1 strength among several $`1^+`$ states prohibits, unlike ground bands, the use of nearly constant excitation energies as a criteria to identify F-spin multiplets of scissors states. Instead, a more sensitive test of this suggestion comes from the summed ground to scissors B(M1) strength. The IBM-2 M1 operator $`(\widehat{L}_\pi \widehat{L}_\nu )`$ is an F-spin vector ($`F=1,F_0=0`$). Its matrix element between the ground state \[$`L=0_g^+,(F=F_{max},F_0)`$\] and scissors state \[$`L=1_{sc}^+,(F^{}=F1,F_0`$)\] is proportional to an F-spin Clebsch Gordan coefficient $`C_{F,F_0}=(F,F_0;1,0|F1,F_0)`$ times a reduced matrix element. It follows that the ratio $`B(M1;0_g^+1_{sc}^+)/(C_{F,F_0})^2`$ does not depend on $`F_0`$ and should be a constant in a given F-spin multiplet. In Table II we list all F-spin partners for which the summed B(M1) strength to the scissors mode has been measured todate . It is seen that within the experimental errors, the above ratio is fairly constant. The most noticeable discrepancy for <sup>172</sup>Yb (F=8), arises from its measured low value of summed B(M1) strength. The latter should be regarded as a lower limit due to experimental deficiencies (large background and strong fragmentation ). These observations strengthen the contention of high F-spin purity and formation of F-spin multiplets of scissors states.
As noted in and shown in Table I and Fig. 1, for nuclei with $`F=6,\mathrm{\hspace{0.17em}6.5},`$ also members of the $`\gamma `$ bands display constant excitation energies and seem to form good F-spin multiplets. This empirical observation has a natural explanation within the family of Hamiltonians with partial F-spin symmetry. For the choice $`\beta =\sqrt{2}`$ and $`A_\pi ^{(2)}=A_\nu ^{(2)}=A_{\pi \nu }^{(2)}`$ in Eq. (10), $`H^{}`$ will have both F-spin and SU(3) partial symmetries. In such circumstances, the ground ($`K=0`$), scissors ($`K=1`$), symmetric-$`\gamma `$ ($`K=2`$), and antisymmetric-$`\gamma `$ ($`K=2`$) bands are solvable and have good SU(3) and F-spin symmetries: $`[(\lambda ,\mu ),F]=[(2N,0),F_{max}]`$, $`[(2N2,1),F=F_{max}1]`$, $`[(2N4,2),F=F_{max}]`$ and $`[(2N4,2),F=F_{max}1]`$ respectively. The intrinsic states for the symmetric-$`\gamma `$ or antisymmetric-$`\gamma `$ bands are obtained by F-spin coupling the $`F=1`$ pair $`R_{i,2,\mu =2}^{}`$ to the ($`F=F_{max}1`$) condensate $`|N_\pi 1,N_\nu 1`$ with ($`N2`$) bosons to form a N-boson intrinsic state with $`F=F_{max}`$ or $`F=F_{max}1`$. Since, in this case, the commutator $`[H^{},\stackrel{}{F}]`$ vanishes when it acts on the solvable intrinsic states, the projected states are ensured to have good F-spin and form F-spin multiplets. At the same time, since the Hamiltonian is not F-spin scalar, the $`\beta `$ bands can have F-spin admixtures and need not form F-spin multiplets.
In summary, we have examined in detail IBM-2 Hamiltonians with partial F-spin symmetry. The latter are not F-spin scalars, yet have a subset of solvable eigenstates with good F-spin symmetry. In particular, the corresponding ground bands form F-spin multiplets with $`F=F_{max}`$, but excited bands can be mixed, which is in line with the empirically observed F-spin multiplets \[4-7\]. A class of IBM-2 Hamiltonians with partial F-spin symmetry predict the occurrence of F-spin multiplets of scissors states, with a moment of inertia equal to that of the ground band. This prediction is in agreement with recent analyses of the empirical systematics of excitation energy and M1 strength of the scissors mode in even-even nuclei . All the above findings illuminate the potential useful role of F-spin (and other) partial symmetries in nuclear spectroscopy and motivate their further study.
We acknowledge helpful discussions with N. Pietralla on the empirical data and thank the Institute for Nuclear Theory at the University of Washington for its hospitality. This work was was supported in part by the Israel Science Foundation, the Zevi Hermann Schapira research fund, and by the U.S. Department of Energy under contract W-7405-ENG-36. |
warning/0001/cond-mat0001067.html | ar5iv | text | # Optical Limiting in Single-walled Carbon Nanotube Suspensions
## I Introduction
Carbon nanotubes provide a unique class of nanostructured materials. Improved methods of synthesis, purification and functionalization have triggered many experiments exploring basic physics of mesoscopic objects as well as various possible applications \[1-4\]. Like in C<sub>60</sub> and other fullerene derivatives, optical limiting is an important application of nanotubes. Optical limiting in liquid suspensions of multi-walled carbon nanotubes ( MWNTs ) has been reported recently . The observed limiting has been compared with that of its well-studied cousin C<sub>60</sub> fullerene. Whereas optical limiting in C<sub>60</sub> solution occurs due to the reverse saturable absorption and subsequent nonlinear refraction and scattering , optical limiting in MWNT suspension is attributed to the nonlinear scattering arising from expanding microplasmas as in carbon black suspension (CBS) . Optical limiting investigations of carbon nanotubes become even more important since they combine the relative advantages of both CBS and fullerene solutions. As for CBS, the nanotube suspensions consist of relatively large size (length $`>`$ 100 nm ) constituents. However, unlike CBS but like fullerenes, their structure and electronic properties are well characterized and their optical response is susceptible to further improvements by molecular engineering. We have carried out a detailed investigation of the optical limiting of suspensions of well-characterised single-walled carbon nanotubes ( SWNTs ) in ethanol, water and ethylene glycol by carrying out optical limiting, z-scan and scattering measurements. The results demonstrate that optical limiting in SWNT suspensions occurs mainly due to absorption-induced scattering in the suspension. Furthermore, the limiting strongly depends on the host liquid. While this manuscript was in preparation, optical limiting in a water suspension of SWNTs has been reported by Vivien et al . However, the present report includes studies of three host liquids, an aspect relevant to optical limiting, and the results suggest that the optical limiting is due to nonlinear scattering from absorption-induced microbubbles.
## II Experimental
SWNTs were produced by the dc arc discharge method using a composite graphite rod containing Y<sub>2</sub>O<sub>3</sub> (1 at.%) and Ni (4.2 at.%) as anode and a graphite rod as cathode under a helium pressure of 660 torr with a current of 100 A and 30 V . The web produced from the arc-discharge contained SWNT bundles, amorphous carbon along with metal encapsulated carbon particles as seen from the high resolution electron microscope (HREM) image. It was heated in air at 300C for about 24 hours to remove the amorphous carbonaceous materials. The heat-treated material was stirred with concentrated nitric acid at 50C for about 12 h and washed with distilled water to remove the dissolved metal particles. The SWNT material so obtained was suspended in ethanol by using an ultrasonicator and filtered through a micropore filter paper (0.3 $`\mu `$m) from Millipore to remove polyhedral carbon nanoparticles present. The product was then dried at 50C for about 12 h. The SWNT content of the product was found to be 80% by thermogravimetric analysis. High-resolution electron microscopic examination (HREM) showed that the SWNTs with an average diameter of 1.4 nm were present as bundles of 10-50 nanotubes. About 5-10 mg of the purified SWNT was dispersed in 15 ml of water/ethanol by ultrasound sonication for 30 minutes. This dispersion was used for our study. All suspensions were stable for several hours after ultrasonication. Optical transmission spectra were found to remain unchanged during the experiment. Dynamic light scattering measurements were performed on the suspensions to characterise the average size of the scatterers. The light scattering experiments were done using $`\lambda `$ = 647.1nm radiation from a Kr<sup>+</sup> ion laser, a home-made spectrometer and a correlator (Malvern 7132CE 64 channel model). Fig. 1 shows the plot of the normalised intensity autocorrelation function g<sub>2</sub>(t) - 1= $`<`$I(0) I(t)$`>`$/$`<`$I(0)$`>`$<sup>2</sup> \- 1 versus time for different scattering angles $`\theta `$, which have been fitted to g<sub>2</sub>(t) - 1 = A exp (-$`\mathrm{\Gamma }`$ t ), where $`\mathrm{\Gamma }`$ = 2Dq<sup>2</sup>, q =$`\frac{4\pi n}{\lambda }`$ sin$`\frac{\theta }{2}`$ is the wavevector transfer, n = 1.33 is the refractive index of the solvent and D is the diffusion coefficient of the scatterer (nanotube). The inset of Fig. 1 shows a plot of $`\mathrm{\Gamma }`$ versus q<sup>2</sup>, and a linear fit gives D = 1.49\*10<sup>-8</sup> cm<sup>2</sup>s<sup>-1</sup>. Taking the average diameter of the nanotube bundles to be 40nm as suggested by electron microscopy and using the equations D = k<sub>B</sub>T \[6 - 0.5($`\gamma _{}`$\+ $`\gamma _{}`$)\]/3$`\pi \eta _{}`$L, $`\gamma _{}`$ = 1.27 - 7.4($`\frac{1}{\delta }`$ \- 0.34)<sup>2</sup>, $`\gamma _{}`$ = 0.19 - 4.2($`\frac{1}{\delta }`$ \- 0.39)<sup>2</sup> and $`\delta `$ = ln $`\frac{2L}{d}`$ , where L and d are the length and diameter of the cylindrical rods and $`\eta _{}`$ is the solvent viscosity, we obtain an average value for L $``$ 25$`\mu `$m. Experiments were performed using a frequency doubled Nd:YAG laser giving 532 nm, 15 ns laser pulses with pulse repetition every $``$3 seconds. The laser emission was focussed using a 50 cm lens such that 1/e<sup>2</sup> radius of the focussed beam was $``$ 50 $`\mu `$m at the focus. For optical limiting measurements the nanotubes sample was kept at the focus and the transmitted emission was passed through an aperture with $``$ 95% transmission at low fluences. Thus the aperture would block off-axis scattered emission, if any, from the sample at high fluences. The input energy was measured by taking a fraction of the input beam on a calibrated biplanar photodiode and output energy was measured by a calibrated PIN-photodiode kept after the aperture. The input and output fluences were estimated by measuring input and output energy and beam size at the sample position.
## III Results and Discussion
Fig. 2 shows the observed variation of output fluence with input fluence for a SWNTs suspension in water and a C<sub>60</sub> -toluene solution . The transmission spectrum of SWNT sample is shown in the inset. Both absorption and scattering may contribute to the loss in transmitted intensity. The SWNTs and C<sub>60</sub> samples were kept in two identical cuvettes and concentrations were adjusted so that the low fluence transmission through the cuvette was the same ( $``$ 55 % ) for both the samples. It is evident from the figure that optical limiting in C<sub>60</sub> solution is stronger than that in SWNTs suspension in water. On the other hand, stronger limiting in MWNTs suspension in ethanol than that in C<sub>60</sub> solution in toluene has been earlier reported by Chen et al . To know the contribution of nonlinear refraction to the observed optical limiting in nanotubes suspension, we performed z-scan measurements . In these measurements the aperture used in the limiting geometry was moved to the far-field region with $``$ 16% transmission without the nanotubes sample in the beam path. Fig. 3 shows the z-scan results. The transmission shown in this figure is the ratio of the aperture transmission measured with the sample in the beam path to that measured without the sample. Thus at large z values the transmission corresponds to the low intensity transmission through the sample. At smaller z values ( i.e higher fluences ) transmission reduces and reaches its minimum near z=0 . The absence of any peak in the z-scan data in Fig. 3 indicates that nonlinear scattering is much stronger than nonlinear refraction. The observed pronounced valley shown in this figure can occur due to nonlinear scattering as well as nonlinear absorption. We believe that the dominant contribution to the observed transmission valley in Fig. 3 is from nonlinear scattering because the optical limiting was found to be very sensitive to the size of the aperture. We note that optical limiting in multi-walled carbon nanotubes suspensions has also been attributed to absorption induced scattering in the suspensions . Z-scan measurement of SWNT suspension in water by Vivien et al showed negative lensing corresponding to thermally induced nonlinear refraction, in apparent variance with our observations. The difference could be due to much higher intensities and much smaller aperture transmission used by them. On the other hand, Chen et al found stronger optical limiting in MWNT suspension in ethanol compared to C<sub>60</sub> solution in toluene. Similarly, Vivien et al also reported stronger optical limiting in SWNT suspension in water compared to C<sub>60</sub> solution in toluene. We note that optical limiting due to nonlinear refraction of thermal origin would be stronger for longer pulses used in our experiments. This mechanism is expected to be more important in C<sub>60</sub> solution than in SWNT suspensions. To understand the nature of scattering, we measured the scattered light at different angles from the sample cell. Fig. 4 shows the results for an angle $``$ 0.8 from the beam axis. For these measurements, the energy of scattered light passing through a suitably positioned 3 mm diameter aperture was measured using a PIN photodiode in integrating mode. The y-axis in Fig. 4 shows the ratio of the signal from this PIN photodiode to the signal from another photodiode monitoring the input pulse energy. For linear scattering this ratio is expected to be constant for all values of the input fluence. The data in Fig. 4 clearly shows evidence of nonlinear scattering. The ratio initially rises and then falls as input fluence is increased which appears to be due to a change in angular distribution of scattered light. We have also measured optical limiting in nanotubes suspended in different liquids viz. ethylene glycol, water and ethanol. For this purpose, identical 10 mm path length cuvettes were used and low fluence transmission was kept at $``$ 42%. During measurements with the three samples, all the cells containing SWNTs suspension were accurately positioned at the same place and other experimental set-up was unaltered. Fig. 5 shows the observed limiting data for different suspensions. It is evident that optical limiting in ethanol suspension was strongest among the three suspensions. The threshold fluence (defined as the fluence at which the transmission reduces to half of its low fluence value ) in different liquids was also different. As estimated from Fig. 5, the lowest value was $``$ 1.0 J/cm<sup>2</sup> for ethanol suspension. These observations suggest that the extent of nonlinear scattering depends on the liquid host. The nonlinear scattering from suspensions of absorbing particles in liquids can result from several mechanisms . The scattering centres can be expanding microplasmas generated by vaporization of the particles , bubbles formed by vaporization of the liquid or transient refractive index inhomogeneities resulting from localised heating around absorbing particles . Our observation of strong solvent dependence of optical limiting rules out plasma formation as being an important mechanism. The change in refractive index of a liquid depends on its thermal figure of merit F = $`\frac{1}{c\rho }\frac{dn}{dT}`$ (where $`c`$ is the specific heat, $`\rho `$ is the density and $`\frac{dn}{dT}`$ is the thermo-optic coefficient). The values of F are 7.5, 8.0 and 1.0 10<sup>-4</sup>cm<sup>3</sup>calK<sup>-1</sup> for ethylene glycol, ethanol and water , respectively. We note that although ethylene glycol has much larger F, optical limiting in glycol suspension was much weaker than that in water suspension. This implies that absorption induced refractive index inhomogeneities did not make significant contribution to the nonlinear scattering. We thus believe that the observed nonlinear scattering in our experiments is mainly from microbubbles formed in the suspension. The strongest optical limiting in ethanol suspension is qualitatively consistent with its lowest boiling point of $``$ 78 C among the three liquids. This conclusion also agrees with the recent pump-probe investigation of CBS by Durand et al using 30 ps laser pulses. These authors found that while for the first few nanosecond after the pump pulse, probe attenuation was solvent independent suggesting scattering from microplasma, for larger delays the probe transmission showed strong solvent dependence implying scattering from bubbles or thermo-optic effects. In conclusion, we report optical limiting of visible ns laser pulses in suspensions of single-walled carbon nanotubes. The dominant mechanism for the observed limiting has been found to be absorption induced nonlinear scattering. Optical limiting has been found to be significantly different for different liquids indicating that scattering is probably due to bubble formation in the suspension, although other causes like sublimation of particles cannot be ruled out at present.
## IV Acknowledgements
AKS thanks Department of Science and Technology, New Delhi for financial assistance. Technical help of S. K. Tiwari and M. Laghate during experiments on optical limiting is gratefully acknowledged.
## V References
S. Ijima, Nature 354, 56 (1991).
T. W. Ebbesen, ”Synthesis and Characterization of Carbon Nanotubes” in Physics and Chemistry of Fullerenes, Ed., K. Prassides ( Kluwer Academic Publishers, Netherlands, 1994), T. W. Ebbesen, ”Carbon Nanotubes”, Physics Today ( June 1996, p-26 ).
C.N.R.Rao, A. Govindaraj, R.Sen and B. Satishkumar, Mater. Res. Innovat 2, 128 (1998) and references therein.
R. Saito, G. Dresselhaus and M. S. Dresselhaus, ”Physical Properties of Carbon Nanotubes” ( Imperial College Press, London, 1998 )
X. Sun, R. Q. Yu, G. Q. Xu, T. S. A. Hor and W.Ji , Appl. Phys. Lett. 73, 3632 (1998).
P. Chen, X. Wu, X. Sun, J. Lin, W. Ji and K. L. Tan, Phys. Rev. Lett. 82, 2548 (1999).
S. R. Mishra, H. S. Rawat, M.P. Joshi and S. C. Mehendale, J. Phys. B 27, L157 (1994).
L. Vivien, E. Anglaret, D. Riehl, F. Bacou, C. Journet, C. Goze, M. Andrieux, M. Brunet, F. Lafonta, P. Bernier, F. Hache, Chem. Phys. Lett. 307, 317 (1999).
Journet, C., Maser, W. K., Bernier, P., Loiseau, A., Lamy de la Chapelle, M., Lefrant, S., Denierd, P., Lee, R. and Fischer, J. E., Nature, 1997, 388, 756.
S. Broersma, J. Chem. Phys. 32, 1626 (1960).
M. Sheik-Bahae, A. A. Said, T. H. Wei, D. J. Hagen and E. W. Van Stryland, IEEE J.Quant. Electr. 26, 760 (1990).
K. Mansour, M. J. Soilcau and E. W. Van Stryland, J.Opt. Soc. Am. B 9, 1100 (1992).
T. F. Bogges, G. R. Allan, D. R. Labergerie, C. H. Venzke, A. L Smirl, L. W. Tutt, A. R. Kost, S. W. McCahahon, M. B. Klein, Opt. Engg. 32(5), 1063 (1993)
O. Durand, V. Grolier-Mazza and R. Frey, Opt. Lett. 23, 1471 (1998).
## VI Figure Captions
Fig. 1. Intensity autocorrelation functions measured at different scattering angles in dynamic light scattering experiments with SWNT suspension in water. The inset shows the linear fit to the plot of $`\mathrm{\Gamma }`$ versus q<sup>2</sup>.
Fig. 2. Optical limiting in SWNTs - water suspension and C<sub>60</sub>-toluene solution. Circles show output with SWNTs-water suspension and triangles show output with C<sub>60</sub> solution. The inset shows transmission loss spectrum of SWNTs-water suspension.
Fig. 3. Variation of transmission with z in a close-aperture z-scan measurements on SWNTs-water suspension.
Fig. 4. Measured variation of ratio of the scattered energy to the input fluence with input fluence at an angle of $``$ 0.8 in the forward direction from the beam axis.
Fig. 5. Measured variation of output fluence with input fluence in SWNTs suspensions in different solvents. Crosses, circles and triangles show output from SWNTs suspension in ethanol, water and ethylene glycol respectively.
Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5 |
warning/0001/hep-ph0001263.html | ar5iv | text | # References
As known, the dynamics of the neutral fermions with anomalous magnetic moments are described by Dirac equation with nonminiaml coupilngs of neutral fermions with electromagnetic field ,:
$$(\widehat{k}m_n+\mu _n(\stackrel{}{\mathrm{\Sigma }}\stackrel{}{B}i\stackrel{}{\alpha }\stackrel{}{E})+iq(\stackrel{}{E}\stackrel{}{B})\gamma _5)\psi (k)=0,$$
(1)
where $`\mu _n`$-is anomalous magnetic moment, $`\frac{1}{2}\stackrel{}{\mathrm{\Sigma }}`$ is spin operator,and defined by formula (21,21) of operator $`\stackrel{}{\alpha }`$ is defined by formula (21,20) of .
In this equation we include also term obtained in ref. :
$$L=iq(\stackrel{}{E}\stackrel{}{H})\overline{\psi }\gamma _5\psi $$
(2)
which appears e.g. in electroweak models at one-loop level in theories with $`P`$-parity violation(for electroweak thories and $`P`$-parity violation see e.g. and references therein).
In this article we consider bound states of particles with anomalous magnetic moments and with interaction (2) in the presence of monopole (see references in ). We obtained generalization of equations (13),(14) where has been considered bound states of particles with anomalous magnetic moments in arbitrary radial electric field.
In has been considered joint influence of the static radial electric field and magnetic fild.Although in this paper $`B`$ is not radial as has been shown that in some condition term $`\stackrel{}{\mathrm{\Sigma }}\stackrel{}{B}`$ is radial.However in contrast to monopole case in term $`\stackrel{}{\mathrm{\Sigma }}\stackrel{}{B}`$ is $`P`$-parity conserved.
The magnetic field (but not magnetic field of monopole) has been presented in equations (13),(14) of ref. besides radial electric field. During derivation of equations (12)-(15) below which defines energy levels of the fermions with anomalous magnetic moment in the presence of monopole the method of the has been used (because in case of radial magnetic field the term $`\stackrel{}{\mathrm{\Sigma }}\stackrel{}{B}\stackrel{}{\mathrm{\Sigma }}\stackrel{}{r}`$ violates $`P`$-parity) for angular variables separation, in accordance with this method it is necessary to present components of spinors as linear combination of spheric spinors $`\mathrm{\Omega }_{jlM}(\stackrel{}{n}),\mathrm{\Omega }_{jl^{}M}(\stackrel{}{n})`$ which have different $`P`$-parity. We also confirm the result of where has been stressed that in case of Coulomb electric field take place the fall down on the center takes place.
As known(see (see e.g. references in )) exist nontrivial solutions of to Yang-Mills theories which have in general both electric and magnetic fields:
$$\stackrel{}{B}=\stackrel{}{n}B(r)$$
(3)
$$\stackrel{}{E}=\stackrel{}{n}E(r)$$
(4)
where$`\stackrel{}{n}=\frac{\stackrel{}{r}}{r}`$.At large distances from the core of the monopole we have:
$$\stackrel{}{E}=\frac{e\stackrel{}{r}}{r^3}$$
(5)
$$\stackrel{}{B}=\frac{g\stackrel{}{r}}{r^3}$$
(6)
It must be noted that between between electric and magnetic charges there exist a relation like that between electric and magnetic charges in case of Dirac monopole :
$$eg=\frac{1}{2}n,n=0,\pm 1,\pm 2,\pm 3,\mathrm{}$$
(7)
Thus below we consider bound states of neutral fermions with anomalous magnetic moment in both radial magnetic and electric fields created by these objects.
Also it is of interest to note that in case of monopole $`\stackrel{}{E}\stackrel{}{B}=E(r)B(r)`$ and thus pseudoscalar interaction (2) is spherically symmetric (depends only on $`r`$) as well as the interaction connected with anomalous magnetic moment and separation of angular variable is possible as seen below.At large distances:
$$\stackrel{}{E}\stackrel{}{B}=E(r)B(r)=eg\frac{1}{r^4}$$
(8)
and thus we have radial interaction (2) which is attractive at appropriate sign of $`n`$ and lead to fall down on the center.The non-locality of $`L=iq(\stackrel{}{E}\stackrel{}{B})\overline{\psi }\gamma _5\psi `$ vertex and finite size of monopole prevent this fall down.Besides, the term:
$$q^2(\stackrel{}{E}\stackrel{}{B})^2=q^2\frac{e^2g^2}{r^8}$$
(9)
in effective potential is always repulsive and dominates at small distances.
It must be stressed that $`\stackrel{}{E}\stackrel{}{B}`$ depends only on $`r`$ because $`\stackrel{}{B}`$ is radial. Also radial is the interaction of particles with anomalous magnetic moment with electric and magnetic fields of monopole in Dirac equation for particles with anomalous magnetic moment :
$$(\widehat{k}M(r)+\mu (g\stackrel{}{\mathrm{\Sigma }}\stackrel{}{n}B(r)ie\stackrel{}{\alpha }\stackrel{}{n}E(r))+iqE(r)B(r)\gamma _5)\psi (k)=0,$$
(10)
Here $`M(r)=c\varphi (r)`$ ($`\varphi (r)`$\- is Higgs field which give mass to the fermion, $`m=M(\mathrm{})`$-is the observed mass of fermion).
Here $`P`$-violation presented (because interaction (2) is effectively pseudoscalar, $`\stackrel{}{\mathrm{\Sigma }}\stackrel{}{n}B(r)`$ also $`P`$-odd )and we find as in the solution as linear combinations of spheric spinors $`\mathrm{\Omega }_{jlM}(\stackrel{}{n}),\mathrm{\Omega }_{jl^{}M}(\stackrel{}{n})`$ which have different $`P`$-parity:
$`\psi ^T=(\varphi ,\chi ),`$
$`\varphi =f_1(r)\mathrm{\Omega }_{jlM}(\stackrel{}{n})+(1)^{\frac{1+ll^{}}{2}}f_2(r)\mathrm{\Omega }_{jl^{}M}(\stackrel{}{n}),`$
$`\chi =g_1(r)\mathrm{\Omega }_{jlM}(\stackrel{}{n})+(1)^{\frac{1+ll^{}}{2}}g_2(r)\mathrm{\Omega }_{jl^{}M}(\stackrel{}{n}))`$ (11)
After separation of angular variables we obtain the following set of equations for radial functions:
$$f_1^{}(r)+\frac{1+\kappa }{r}f_1(r)+\mu E(r)f_1(r)=(ϵ+M(r))g_1(r)i\mu B(r)g_2(r)+a_P(r)f_2(r)=0$$
(12)
$$f_2^{}(r)+\frac{1\kappa }{r}f_2(r)+\mu E(r)f_2(r)=(ϵ+M(r))g_2(r)i\mu B(r)g_1(r)a_Pf_1(r)=0$$
(13)
$$g_1^{}(r)+\frac{1\kappa }{r}g_1(r)\mu E(r)g_1(r)=(ϵM(r))f_1(r)+i\mu B(r)f_2(r)+a_Pg_2(r)=0$$
(14)
$$g_2^{}(r)+\frac{1+\kappa }{r}g_2(r)\mu E(r)g_2(r)=(ϵM(r))f_2(r)+i\mu B(r)f_1(r)a_Pg_1(r)=0$$
(15)
where
$$\kappa =l(l+1)j(j+1)\frac{1}{4},$$
(16)
$$a_P=iqE(r)B(r),$$
(17)
At large distances $`a_P=\frac{qeg}{2r^4}`$.
In pure electric field case i.e. at $`B=0,q=0`$ $`P`$-parity is conserved and we obtain two decoupled system of equations (13),(14) of the in which of course the magnetic field is also zero.
If only electric field is presented we obtain for radial functions the following equations:
$$(\stackrel{}{p}^2+m^2ϵ^2+\mu ^2E^2+4\mu \pi \rho \pm \frac{2\mu E(r)(1+\kappa )}{r})R_{1,2}(r)=0.$$
(18)
where $`\stackrel{}{p}^2=\frac{1}{r^2}\frac{d}{dr}r^2\frac{d}{dr}+\frac{l(l+1)}{r^2}`$ (we find the solution as $`\varphi =R_1(r)\mathrm{\Omega }_{jlM}(\stackrel{}{n}),\chi =R_2(r)\mathrm{\Omega }_{jl^{}M}(\stackrel{}{n})`$ ). During derivation of this formulas we take into account that $`div\stackrel{}{E}=4\pi \rho `$.
In it has been stressed (page 3 after formulas (13),(14)) that for particles with anomalous magnetic moment we have fall down on the center in case of Coulomb electric field which prevented by cut off of the potential, i.e. by taking into account charge distribution inside neutron.
Indeed, from equations (18)-(19)(as well as from equations (13)(14) of the ref.) it is seen that in case of Coulomb attraction ($`E(r)=\frac{Ze}{r}`$) we have fall down on the center due to the term $`\mu \frac{E(r)}{r}(1+\kappa )=\frac{\mu eZ(1+\kappa )}{r^3}`$in effective potential. It must be noted, however, that the term $`\mu ^2E^2=\frac{Z^2\mu ^2e^2}{r^4}`$ in potential is always repulsive and at small $`r`$ prevents fall down on the center. In case of pointlike charge term $`2\pi \rho =2\pi \mu Ze\delta (\stackrel{}{r})`$ may be considered as perturbation.
In case of Coulomb potential we can calculate by using quasiclassical methods energy levels which are defines by the equation :
$$\sqrt{2m(EV(r)}𝑑r=n$$
(19)
where
$$V(r)=\frac{(l+\frac{1}{2})^2}{2mr^2}+\frac{\mu E(1+\kappa )}{r}+\frac{1}{4}\mu ^2E^2$$
(20)
$$2mE=ϵ^2m^2$$
(21)
Appendix
Below we present system of equation for radial functions (19)-(22) of ref. which obtained after angular variables separation in electroweak Dirac equation:
$$(f_1^{}(r)+\frac{1+\kappa }{r}f_1(r))(E+M(r)V(r))g_1(r)V_+(r)f_2(r)=0$$
(22)
$$(f_2^{}(r)+\frac{1\kappa }{r}f_2(r))+(E+M(r)V(r))g_2(r)+V_+(r)f_1(r)=0$$
(23)
$$(g_1^{}(r)+\frac{1\kappa }{r}g_1(r))+(EM(r)V(r))f_1(r)+V_{}(r)g_2(r)=0$$
(24)
$$(g_2^{}(r)+\frac{1+\kappa }{r}g_2(r))(EM(r)V(r))f_2(r)V_{}(r)g_1(r)=0$$
(25)
where:
$$M(r)=ma_SV_S(r)$$
(26)
$$V(r)=eg_VZ_0(r)+eQA_0(r),$$
(27)
$$V_\pm (r)=eg_AZ_0(r)\pm a_PV_P(r),$$
(28)
The author express his sincere gratitude to E.B.Prokhorenko and Zh.K.Manucharyan for helpful discussions. |
warning/0001/gr-qc0001054.html | ar5iv | text | # Generalised scalar-tensor theory in the Bianchi type I model
## 1 Introduction
Scalar-tensor theories seem to be essential to describe gravitational interactions near the Plank scale : string theory, higher order theories in the Ricci scalar , extended inflation and many others theories imply scalar field.
The generalised scalar-tensor Lagrangian has the same form as the Brans-Dicke theory but with a coupling constant $`\omega `$ depending on the scalar field. Such a theory is interesting for many reasons. Hence, if we choose $`\omega `$ as a constant, the Lagrangian is identical to Brans-Dicke Lagrangian. This theory tends to General Relativity for large value of the coupling constant ($`\omega >500`$). But, if we choose $`\omega =1`$, the Brans-Dicke theory is identical to the string theory in the low-energy limit. Hence, the generalised scalar-tensor theory seems to be able to build a " bridge " between string theory and General Relativity. Other reasons, as inflation, can be put forward : such a theory with a varying coupling constant, may drive the scale factors to accelerate without potential or cosmological constant , i.e. called kinetic inflation.
The generalised scalar-tensor theory has been studied by many authors and the method we will use to find exact solutions has always been described in in the presence of matter in the Lagrangian. Here, we will consider the empty Bianchi type I Universe, which is spatially flat, and will use three different forms of the coupling constant $`\omega (\varphi )`$. The first form, $`2\omega (\varphi )+3=2\beta (1\varphi /\varphi _c)^\alpha `$, has been introduced by Garcia-Bellido and Quiros and studied by Barrow in the context of a FLRW flat model with vacuum or radiation. It has also been studied in , for a Bianchi type I model, where a solution is found in presence of matter. In this paper, we will write explicitly an exact solution and will study the dynamical behaviour of the metric functions which depends on the integration constant. We will cast light on interesting features such as kinetic inflation. The second form is a power law type, $`3+2\omega (\varphi )=\varphi _c^2\varphi ^{2m}`$. Here again, we will give explicitly an exact solution and study it. An interesting feature is the possibility of a non-singular Universe. The third form is an exponential law type, $`3+2\omega (\varphi )=e^{2\varphi _c\varphi }`$ and will be studied qualitatively. These two last laws seem interesting because power and exponential laws are very present in physics. They play a fundamental role for the metric functions of course, but also when we consider a potential $`V(\varphi )`$ giving birth to extended or chaotic inflation . Moreover, we will see how the power law form of the coupling constant is linked to minimally coupled and induced gravity for large or small values of the scalar field.
This paper is organised as follows. In section 2, we write field equations in both Brans-Dicke and conformal frame and explain how to proceed to solve them. In section 3, we derive solution for each of the three forms of $`\omega (\varphi )`$ and study them.
## 2 Field equations.
### 2.1 Field equations in the Brans-Dicke frame.
We work with the metric:
$$ds^2=dt^2+a(t)^2(\omega ^1)^2+b(t)^2(\omega ^2)^2+c(t)^2(\omega ^3)^2$$
(1)
$`a(t)`$, $`b(t)`$, $`c(t)`$ are the metric functions, $`\omega ^i`$ are the 1-forms of the Bianchi type I model and $`t`$, the proper time. We express the Lagrangian of the theory in the form:
$$L=\varphi R+\omega (\varphi )\varphi _{,\alpha }\varphi ^{,\alpha }\varphi ^1$$
(2)
One can also cast (2) on the form:
$$L=f(\mathrm{\Phi })R+1/2_\alpha \mathrm{\Phi }^\alpha \mathrm{\Phi }$$
(3)
with
$$\omega (\varphi )=1/2ff^{,2}$$
(4)
The corresponding field equations and Klein-Gordon equation are obtained by varying the action (2) with respect to the space-time metric and the scalar field. If we introduce the $`\tau `$ time through
$$abcd\tau =dt$$
(5)
then, denoting $`d/d\tau `$ by a prime, the field equations are:
$`{\displaystyle \frac{a^{,,}}{a}}{\displaystyle \frac{a^{,2}}{a^2}}+{\displaystyle \frac{a^,}{a}}{\displaystyle \frac{\varphi ^,}{\varphi }}{\displaystyle \frac{1}{2}}{\displaystyle \frac{\omega ^,}{3+2\omega }}{\displaystyle \frac{\varphi ^,}{\varphi }}`$ $`=`$ $`0`$
$`{\displaystyle \frac{b^{,,}}{b}}{\displaystyle \frac{b^{,2}}{b^2}}+{\displaystyle \frac{b^,}{b}}{\displaystyle \frac{\varphi ^,}{\varphi }}{\displaystyle \frac{1}{2}}{\displaystyle \frac{\omega ^,}{3+2\omega }}{\displaystyle \frac{\varphi ^,}{\varphi }}`$ $`=`$ $`0`$ (6)
$`{\displaystyle \frac{c^{,,}}{c}}{\displaystyle \frac{c^{,2}}{c^2}}+{\displaystyle \frac{c^,}{c}}{\displaystyle \frac{\varphi ^,}{\varphi }}{\displaystyle \frac{1}{2}}{\displaystyle \frac{\omega ^,}{3+2\omega }}{\displaystyle \frac{\varphi ^,}{\varphi }}`$ $`=`$ $`0`$
$`{\displaystyle \frac{a^,}{a}}{\displaystyle \frac{b^,}{b}}+{\displaystyle \frac{a^,}{a}}{\displaystyle \frac{c^,}{c}}+{\displaystyle \frac{b^,}{b}}{\displaystyle \frac{c^,}{c}}+{\displaystyle \frac{\varphi ^,}{\varphi }}({\displaystyle \frac{a^,}{a}}+{\displaystyle \frac{b^,}{b}}+{\displaystyle \frac{c^,}{c}}){\displaystyle \frac{\omega }{2}}({\displaystyle \frac{\varphi ^,}{\varphi }})^2`$ $`=`$ $`0`$ (7)
$`\varphi ^{,,}={\displaystyle \frac{\omega ^,\varphi ^,}{3+2\omega }}`$ (8)
We can integrate (8) to obtain the useful equation :
$$A\varphi ^,\sqrt{3+2\omega }=1$$
(9)
$`A`$ being an integration constant. Hence, we see that $`\omega >3/2`$.
### 2.2 Field equations in the conformal frame.
Now, we work with the conformal metric:
$$ds^2=d\stackrel{~}{t}^2+\stackrel{~}{a}(t)^2(\omega ^1)^2+\stackrel{~}{b}(t)^2(\omega ^2)^2+\stackrel{~}{c}(t)^2(\omega ^3)^2$$
(10)
By the conformal transformation the metric has been redefined as:
$$\stackrel{~}{g}_{\alpha \beta }=\varphi g_{\alpha \beta }$$
(11)
and the Lagrangian becomes :
$$L=R1/2(3+2\omega )\varphi _{,\alpha }\varphi ^{,\alpha }\varphi ^2$$
(12)
Hence, the generalised scalar-tensor theory is cast into Einstein gravity with a minimally coupled scalar field. In the $`\stackrel{~}{\tau }`$ time defined as :
$$\stackrel{~}{a}\stackrel{~}{b}\stackrel{~}{c}d\stackrel{~}{\tau }=d\stackrel{~}{t}$$
(13)
the field equations and the klein-Gordon equation become in the conformal frame
$`{\displaystyle \frac{\stackrel{~}{a}^{,,}}{\stackrel{~}{a}^,}}{\displaystyle \frac{\stackrel{~}{a}^,}{\stackrel{~}{a}}}`$ $`=`$ $`0`$
$`{\displaystyle \frac{\stackrel{~}{b}^{,,}}{\stackrel{~}{b}^,}}{\displaystyle \frac{\stackrel{~}{b}^,}{\stackrel{~}{b}}}`$ $`=`$ $`0`$ (14)
$`{\displaystyle \frac{\stackrel{~}{c}^{,,}}{\stackrel{~}{c}^,}}{\displaystyle \frac{\stackrel{~}{c}^,}{\stackrel{~}{c}}}`$ $`=`$ $`0`$
$`{\displaystyle \frac{\stackrel{~}{a}^,}{\stackrel{~}{a}}}{\displaystyle \frac{\stackrel{~}{b}^,}{\stackrel{~}{b}}}+{\displaystyle \frac{\stackrel{~}{a}^,}{\stackrel{~}{a}}}{\displaystyle \frac{\stackrel{~}{c}^,}{\stackrel{~}{c}}}+{\displaystyle \frac{\stackrel{~}{b}^,}{\stackrel{~}{b}}}{\displaystyle \frac{\stackrel{~}{c}^,}{\stackrel{~}{c}}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}(\omega +3/2)({\displaystyle \frac{\varphi ^,}{\varphi }})^2`$ (15)
$`{\displaystyle \frac{\varphi ^{,,}}{\varphi ^,}}{\displaystyle \frac{\varphi ^,}{\varphi }}`$ $`=`$ $`{\displaystyle \frac{\omega ^,}{3+2\omega }}`$ (16)
Equations (14) are exactly the same as in the Bianchi type I model in General Relativity. Only the constraint equation (15) is different. The solutions of the field equations are in the $`\stackrel{~}{t}`$ time the well-known Kasnerian solutions:
$$\stackrel{~}{a}=\stackrel{~}{t}^{p_1},\stackrel{~}{b}=\stackrel{~}{t}^{p_2},\stackrel{~}{c}=\stackrel{~}{t}^{p_3}$$
(17)
$`p_1`$, $`p_2`$, $`p_3`$ being the Kasner exponents with :
$$p_i=1$$
(18)
With the constraint equation, we obtain :
$$p_i^2=12\varphi _0^2$$
(19)
$`\varphi _0`$ being the integration constant of the scalar field. Hence, for all coupling constant $`\omega (\varphi )`$, in the conformal frame, there will always be one negative Kasner exponent or three positive Kasner exponents and then two or three decreasing metric functions. In the $`\stackrel{~}{\tau }`$ time, the solutions of (14) are:
$`\stackrel{~}{a}=e^{\alpha _1\stackrel{~}{\tau }+\alpha _0}`$
$`\stackrel{~}{b}=e^{\beta _1\stackrel{~}{\tau }+\beta _0}`$ (20)
$`\stackrel{~}{c}=e^{\gamma _1\stackrel{~}{\tau }+\gamma _0}`$
where $`\alpha _i`$, $`\beta _i`$, $`\gamma _i`$ are integration constants. We integrate the Klein-Gordon equation to obtain the important equation:
$$\stackrel{~}{\varphi }_0\varphi ^,\varphi ^1\sqrt{3+2\omega }=1$$
(21)
$`\stackrel{~}{\varphi }_0`$ being an integration constant (in fact $`\stackrel{~}{\varphi }_0=A`$). Hence, we deduce from the constraint equation that:
$$\alpha _1\beta _1+\alpha _1\gamma _1+\beta _1\gamma _1=1/4\stackrel{~}{\varphi }_0^2\text{}\omega (\varphi )$$
(22)
To find solutions to the field equations (2.1) we proceed as follow: first, we have to find solutions, for the scalar field, of the equations (9) and (21) so that we obtain respectively $`\varphi (\tau )`$ and $`\varphi (\stackrel{~}{\tau })`$. Second, we write $`\varphi (\tau )=\varphi (\stackrel{~}{\tau })`$ and reverse $`\varphi (\stackrel{~}{\tau })`$ to find $`\stackrel{~}{\tau }=\stackrel{~}{\tau }(\tau )`$. Third, using (11), we write :
$$a=\stackrel{~}{a}(\stackrel{~}{\tau }(\tau ))/\varphi (\tau ),b=\stackrel{~}{b}(\stackrel{~}{\tau }(\tau ))/\varphi (\tau ),c=\stackrel{~}{c}(\stackrel{~}{\tau }(\tau ))/\varphi (\tau )$$
(23)
Let us examine what are the relations between the quantities in the $`\tau `$ time and in the $`t`$ time. The amplitudes of the metric functions are the same in the both time since $`a(\tau )=a(\tau (t))=a(t)`$. The sign of the first derivatives are also the same : remember that $`d\tau /dt=1/abc`$ is positive since the metric functions are positive-definite. Hence, $`\tau `$ is an increasing function of $`t`$ and the sign of the first derivative of the metric functions will be the same in both $`\tau `$ time and $`t`$ time. The sign of the second derivatives in the $`t`$ time and $`\tau `$ time are different. If an overdot denotes differentiation with respect to $`t`$, the sign of $`\ddot{a}`$ will be that of $`a^{,,}a^,(a^,/a+b^,/b+c^,/c)`$. We will study both the sign of $`a^{,,}`$ and $`\ddot{a}`$ in the applications of section 3. Of course, the amplitudes of the derivatives are different in the $`t`$ and $`\tau `$ times. But we will not study them since we are mainly interested in their signs and therefore dynamical behaviour of the metric functions: whether they are increasing, decreasing or bouncing, and whether there is inflation.
Another difference between the two times comes from their asymptotic behaviours. For instance, the $`t`$ time could diverge at a finite value of the $`\tau `$ time. It depends mainly on $`dt/d\tau =abc=V`$, where $`V`$ is the volume of the Universe. In the cases we are going to study, the volume will always tend toward $`0`$ or infinity (we will show it for the two first theories of section 3). Then, if $`V0`$ when $`\tau `$ tends toward a constant or infinity, $`t`$ tends toward a constant. If $`V\mathrm{}`$ when $`\tau \mathrm{}`$, $`t\mathrm{}`$. If $`V\mathrm{}`$ when $`\tau `$ tends toward a constant, $`t`$ may tend toward infinity or a constant. In this last case, we need to integrate the volume $`abc`$ to make the asymptotic behaviour of the cosmic time $`t`$ precise. Unhappily, it will not be possible in the theories of section 3. We have studied the behaviour of the volume for the two first theories so that one can always get the asymptotic behaviour of $`t(\tau )`$ by using these rules (except the case $`\tau cte`$ and $`V\mathrm{}`$).
Concerning the presence of singularity, to ensure that a theory is non-singular, we will check that the Ricci curvature scalar $`R`$ is finite. The Ricci scalar can be writen:
$$R=(abc)^2\left[\omega (\varphi ^,\varphi ^1)^2+3\varphi ^1\omega ^,\varphi ^,(3+2\omega )^1\right]$$
(24)
## 3 Non-singular and accelerated behaviours.
To simplify the study of the metric functions, we will consider in what follows only an increasing function of the scalar field, which means the only positive constants are $`A`$ and $`\stackrel{~}{\varphi }_0`$.
### 3.1 The case $`3+2\omega =2\beta (1\varphi /\varphi _c)^\alpha `$
We use the form for the coupling constant $`3+2\omega =2\beta (1\varphi /\varphi _c)^\alpha `$ where $`\beta `$ is a positive constant, $`\alpha `$, $`\varphi _c`$ are constant. The case $`\alpha =0`$ corresponds to Brans-Dicke theory and the case $`\alpha =1`$ and $`\beta =1/2`$ to Barker’s theory . Barrow showed in his paper that the case $`\alpha =2`$ is representative of the behaviour of other cases with $`\alpha 2`$ in the neighbourhood of the singularity. Hence, we will consider only this case. From (9) we derive:
$$\varphi (\tau )=\varphi _c\left[1e^{(\tau +\tau _0)/(A\sqrt{2\beta }\varphi _c)}\right]$$
(25)
from (21) we deduce:
$$\varphi (\stackrel{~}{\tau })=\varphi _c(1+e^{(\stackrel{~}{\tau }_0+\stackrel{~}{\varphi }_0^1\stackrel{~}{\tau })/\sqrt{2\beta }})^1$$
(26)
Equating (25) and (26), we get:
$$\stackrel{~}{\tau }=\stackrel{~}{\varphi }_0\sqrt{2\beta }ln\left[e^{(\tau +\tau _0)/(A\sqrt{2\beta }\varphi _c)}1\right]\stackrel{~}{\varphi }_0\stackrel{~}{\tau }_0$$
(27)
$`\tau _0`$ being an integration constant. Hence, using (23), we write:
$$a(\tau )=\frac{e^{\stackrel{~}{\varphi }_0\stackrel{~}{\tau }_0\alpha _1+\alpha _0}}{\sqrt{\varphi _c}}(e^{\frac{\tau +\tau _0}{A\sqrt{2\beta }\varphi _c}}1)^{\sqrt{2\beta }\stackrel{~}{\varphi }_0\alpha _1}(1e^{\frac{\tau +\tau _0}{A\sqrt{2\beta }\varphi _c}})^{1/2}$$
(28)
and identical expressions for $`b(\tau )`$, $`c(\tau )`$ with $`\beta _0`$, $`\beta _1`$ and $`\gamma _0`$, $`\gamma _1`$ respectively. If we introduce:
$$u=(\tau +\tau _0)/(A\sqrt{2\beta }\varphi _c),\text{ }a_0=e^{\stackrel{~}{\varphi _0}\stackrel{~}{\tau _0}\alpha _1+\alpha _0}/\sqrt{\varphi _c}>0,\text{ }\alpha _1=\sqrt{2\beta }\stackrel{~}{\varphi }_0\alpha _1$$
(29)
the expression (28) becomes:
$$a(\tau )=a_0(e^u1)^{a_11/2}e^{u/2}$$
(30)
$`u`$ and the $`\tau `$ time vary in the same manner as long as $`A`$ and $`\varphi _c`$ are positive constants. The constraint equation (22) is rewritten as:
$$a_1b_1+a_1c_1+b_1c_1=\frac{1}{2}\beta $$
(31)
The metric function will be real for positive $`u`$. One can show that there is no non-singular behaviour for this theory in an anisotropic Universe. The Ricci curvature can be written as:
$$R=(e^u1)^{1+2(a_1+b_1+c_1)}(32\beta e^{2u}24\beta ^2e^{4u}+24\beta ^2e^{5u})(2a_0^2b_0^2c_0^2e^{3u})^1$$
(32)
We check that conditions to get finite $`R`$ for asymptotic times $`(u0,u\mathrm{})`$ are not compatible: for $`u0`$ we need $`a_1+b_1+c_1>1/2`$ whereas for $`u+\mathrm{}`$, we need $`a_1+b_1+c_1<3/2`$. So there is always a singularity for the Ricci curvature at small or/and large times.
The first derivative of (30) shows that the metric function $`a(\tau )`$ will have a minimum for $`u=ln(2a_1)`$ and $`a_1]0,1/2[`$. For small $`u`$, we have $`\varphi 0`$, $`\omega \beta 3/2`$ and:
$$aa_0(e^u1)^{a_11/2}$$
(33)
Hence, if $`a_1<3/2`$, $`da/d\tau `$ and $`a`$ tend to 0, if $`a_1[3/2,1/2]`$, $`da/d\tau `$ tends to infinity and $`a`$ tends to 0, if $`a_1>1/2`$, $`da/d\tau `$ and $`a`$ tends respectively to $`\mathrm{}`$ and $`+\mathrm{}`$. For large $`u`$, we have $`\varphi \varphi _c`$, $`\omega +\mathrm{}`$ if $`\alpha >0`$ and:
$$aa_0e^{a_1u}$$
(34)
Hence, if $`a_1<0`$, $`da/d\tau `$ and $`a`$ tend to infinity, if $`a_1>0`$, $`da/d\tau `$ and $`a`$ tend to 0. We see that the form of the metric function depends only on the parameter $`a_1`$:
* If $`a_1<3/2`$, the metric function is increasing (fig 1).
* If $`a_1[3/2,1/2]`$, it is increasing but with an inflexion point (fig 2). By studying the second derivative of $`a(\tau )`$, one can show that the condition to have an inflexion point is $`a_1[3/2,1/2]`$. In the other cases, the second derivative is always positive and the dynamic is always accelerated. Lets note that it is not inflation since for that we must have $`\ddot{a}>0`$ and not $`a^{,,}>0`$.
* If $`a_1[1/2,0]`$, the metric function has a minimum. Hence, if $`a_1`$, $`b_1`$ and $`c_1`$ belong to $`[1/2,0]`$, all the metric functions have a bounce. However that does not mean that the Universe is non-singular since in this case the Ricci scalar become infinite for large $`\tau `$.
* If $`a_1>0`$, the metric function is decreasing (fig 4).
Example of these four behaviours are illustrated on figures 1-4.
Now we examine the sign of the second derivative of the metric function $`a`$ in the $`t`$ time so that we can detect inflation. It is the same as the quantity $`2a_1(b_1+c_1)e^{2u}+(1b_1c_1)e^u1`$ which is a second degree equation for $`e^u`$. One finds two roots: $`e^{u_{1,2}}=(1b_1c_1\pm \sqrt{\mathrm{\Delta }})\left[4a_1(b_1+c_1)\right]^1`$ with $`\mathrm{\Delta }=(b_1+c_11)^28a_1(b_1+c_1)`$. If they are complex or inferior to 1, the sign of $`\ddot{a}`$ is the same as $`2a_1(b_1+c_1)`$. If they are superior to 1, there are two inflexion points: $`\ddot{a}`$ is first positive (negative), negative (positive) and then positive (negative) if $`2a_1(b_1+c_1)>0`$ ( respectively $`2a_1(b_1+c_1)<0`$). For the same reasons, if one of the roots is not real or inferior to 1, there is one inflexion point and $`\ddot{a}`$ is first positive (negative) and then negative (positive) if $`2a_1(b_1+c_1)>0`$ (respectively $`2a_1(b_1+c_1)<0`$). Here, $`\ddot{a}>0`$ can correspond to inflation when in the same time $`\dot{a}`$, or equivalently $`a^,`$, is positive. Hence, one see an example of kinetic inflation as described by Jana Levin in and . We remark also that inflation can end in a natural way.
If now we write the volume:
$$V=abc$$
(35)
For small (large) $`u`$, $`V`$ vanishes if $`a_1+b_1+c_1<3/2`$ ($`a_1+b_1+c_1>0`$) else it tends toward infinity. Another interesting feature of this model is that for $`\beta =1/2`$, we have $`\omega 1`$ for small value of $`u`$, $`\omega \mathrm{}`$ and $`\omega ^3(d\omega /d\varphi )0`$ if $`\alpha >1/2`$ for large value. That is the two value of the coupling constant that corresponds to String theory in the low-energy limit and to General Relativity (by General Relativity we means that the post-Newtonian parameters of General Relativity are recovered).
### 3.2 The case $`3+2\omega =\varphi _c^2\varphi ^{2m}`$.
Now, we consider the following form of the coupling constant:
$$3+2\omega =\varphi _c^2\varphi ^{2m}$$
(36)
where $`\varphi _c`$ and $`m`$ are real constants. Using the same process than before, from (9) we derive :
$$\varphi (\tau )=\left[(m+1)/(A\varphi _c)(\tau +\tau _0)\right]^{1/(m+1)}$$
(37)
and from (21) we get :
$$\varphi (\stackrel{~}{\tau })=\left[m(\stackrel{~}{\varphi }_0\varphi _c)^1(\stackrel{~}{\tau }+\stackrel{~}{\tau }_0)\right]^{1/m}$$
(38)
Equating (37) and (38) we have:
$$\stackrel{~}{\tau }=\frac{\stackrel{~}{\varphi }_0\varphi _c}{m}\left[\frac{m+1}{A\varphi _c}(\tau +\tau _0)\right]^{m/(m+1)}\stackrel{~}{\tau }_0$$
(39)
Then, with (23) we obtain :
$$a(\tau )=exp\{\frac{\alpha _1\stackrel{~}{\varphi }_0\varphi _c}{m}\left[\frac{m+1}{A\varphi _c}(\tau +\tau _0)\right]^{m/(m+1)}\alpha _1\stackrel{~}{\tau }_0+\alpha _0\}\left[\frac{m+1}{A\varphi _c}(\tau +\tau _0)\right]^{1/2(m+1)}$$
(40)
We introduce the variables :
$$a_0=e^{\alpha _1\stackrel{~}{\tau }_0+\alpha _0},\text{ }a_1=\alpha _1\stackrel{~}{\varphi }_0\varphi _c,\text{ }u=(\tau +\tau _0)/(A\varphi _c)$$
(41)
and (40) becomes :
$$a=a_0exp(a_1m^1\left[(m+1)u\right]^{m/(m+1)})\left[(m+1)u\right]^{1/2(m+1)}$$
(42)
We get the same type of expressions for $`b(\tau )`$ and $`c(\tau )`$. From the constraint equation (22) we deduce:
$$a_1b_1+a_1c_1+b_1c_1=\varphi _c^2/4$$
(43)
The expression (42) of the metric function shows that $`(m+1)u`$ must be positive. Hence, if $`m>1`$, $`u[0,+\mathrm{}[`$ and if $`m<1`$, $`u]\mathrm{},0]`$. $`u`$ and the $`\tau `$ time vary in the same manner as long as $`A`$ and $`\varphi _c=\sqrt{\varphi _c^2}`$ are two positive constants.
First, let us examine the Ricci scalar. It is written:
$`R=\left[(1+m)u\right]^{(12m)/(1+m)}\text{[}3\varphi _c^2\left[(m+1)u\right]^{2m/(m+1)}+`$
$`6m\varphi _c^4\left[(m+1)u\right]^{4m/(1+m)}\text{]}\left[2a_0^2b_0^2c_0^2(1+m)^2e^{2(a_1+b_1+c_1)\left[(m+1)u\right]^{m/(m+1)}/m}\right]^1`$ (44)
Only if $`m[0,1/2]`$ and $`a_1+b_1+c_1>0`$, is the Ricci scalar always finite at both small and large times, avoiding the singularity. Now we examine the dynamic of $`a`$ in the $`\tau `$ time. The first derivative of (42) vanishes for $`u=(2a_1)^{(m+1)/m}/(m+1)`$ and hence, $`a(\tau )`$ has an extremum for this value that exists only if $`a_1`$ is positive. The asymptotic study of (42) when $`u0`$ and $`u\pm \mathrm{}`$ gives the results summarised in table 1. We found eight different behaviours. The figures 5-12 show an example of each of them. To summarise the main characteristics of each case in the $`\tau `$ time:
* For $`a_1<0`$, the metric function is always decreasing and has an inflexion point when $`m<3/2`$.
* For $`a_1>0`$, the metric function has a minimum if $`m>0`$ and a maximum if $`m<0`$. Hence, only the case where $`a_1,b_1,c_1`$ and $`m`$ are positive, gives birth to a "bounce" Universe. It avoids the singularity if $`m[0,1/2]`$ and $`a_1+b_1+c_1>0`$ and will be today in expansion in all directions of space.
If we define the volume $`V`$ by $`V=abc`$, then it tends to vanish for small $`u`$ if $`m/(m+1)<0`$ and $`m(a_1+b_1+c_1)<0`$ or $`m/(m+1)>0`$ and $`3/\left[2(m+1)\right]>0`$. It becomes infinite if $`m/(m+1)<0`$ and $`m(a_1+b_1+c_1)>0`$ or $`m/(m+1)>0`$ and $`3/\left[2(m+1)\right]<0`$. For large $`u`$, it tends to vanish if $`m/(m+1)>0`$ and $`m(a_1+b_1+c_1)<0`$ or $`m/(m+1)<0`$ and $`3/\left[2(m+1)\right]<0`$. It becomes infinite if $`m/(m+1)>0`$ and $`m(a_1+b_1+c_1)>0`$ or $`m/(m+1)<0`$ and $`3/\left[2(m+1)\right]>0`$.
By examining the sign of $`a^{,,}`$, we can conclude that the dynamic of the metric function will always be accelerated (recall again that it is not inflation since it does not mean that $`\ddot{a}>0`$) if $`m>1/2`$ or $`m[3/2,1/2]`$ and $`a_1<0`$. If $`m<3/2`$ the dynamic is first accelerated and then decelerated. The same thing happens when $`m[0,1/2]`$ and $`a_1>0`$ whereas for $`m[3/2,0]`$ and $`a_1>0`$, the metric accelerates again.
We complete this study by examining the sign of the second derivative in the $`t`$ time. It is the same as $`m+(b_1+c_1)\left[\left[(m+1)u\right]^{m/(m+1)}2a_1\left[(m+1)u\right]^{2m/(m+1)}\right]`$. This is a second degree equation for $`\left[(m+1)u\right]^{m/(m+1)}`$. The two roots are
$$u_{1,2}=(m+1)^1\left[(b_1+c_1\pm \sqrt{\mathrm{\Delta }})(4a_1(b_1+c_1))^1\right]^{(m+1)/m}$$
with $`\mathrm{\Delta }=(b_1+c_1)(8a_1m+b_1+c_1)`$. If $`u_{1,2}`$ are not real, the sign of $`\ddot{a}`$ is the one of $`2a_1(b_1+c_1)`$. When the two roots are real, they always belong to the interval where $`u`$ varies since their sign is the same as $`m+1`$. Then $`\ddot{a}`$ has the same sign as $`2a_1(b_1+c_1)`$ if $`u`$ is out of $`[u_1,u_2]`$ or the opposite sign if $`u[u_1,u_2]`$. There are two inflexion points. Hence, we get the same type of behaviour for $`\ddot{a}`$ as in the previous subsection. In the same manner, if only one root is real, the dynamic of $`a`$ will be accelerated and then decelerated or vice-versa depending on the sign of $`2a_1(b_1+c_1)`$. So, there is one inflexion point. For this theory also, inflation can end naturally.
Concerning the coupling constant, we have for $`m+1>0`$: when $`\tau +\mathrm{}`$, $`\varphi +\mathrm{}`$, $`\omega \varphi _c^2\varphi ^{2m}/2+\mathrm{}`$ if $`m>0`$ and $`\omega 3/2`$ if $`m[1,0]`$. When $`\tau \tau _0`$, $`\varphi 0`$, $`\omega \varphi _c^2\varphi ^{2m}/2+\mathrm{}`$ if $`m[1,0]`$ and $`\omega 3/2`$ if $`m>0`$. Considering these last remarks and the relation (3), one can deduce that the asymptotic behaviours of the metric functions when $`\varphi 0`$, $`\omega \varphi _c^2\varphi ^{2m}/2+\mathrm{}`$ and $`m[1,0]`$ are the same as in the cases of a coupling function of type $`f(\mathrm{\Phi })=f_0e^{n\mathrm{\Phi }}`$ when $`\varphi _c^2=n^2`$ and $`m=1/2`$ and $`f(\mathrm{\Phi })=(f_0\mathrm{\Phi }+f_1)^n`$ when $`\varphi _c^2=(f_0n)^2`$ and $`2m=(2n)/n`$ with $`n[0,2]`$. Moreover, the asymptotic behaviour of the metric functions when $`\varphi +\mathrm{}`$, $`\omega \varphi _c^2\varphi ^{2m}/2+\mathrm{}`$ and $`m>0`$ are the same as in the previous case but with $`n[0,2]`$.
Hence the study of the metric functions when $`3+2\omega =\varphi _c^2\varphi ^{2m}`$, give us information on the asymptotic behaviours of two different couplings $`f(\mathrm{\Phi })`$, that is $`f(\mathrm{\Phi })=(f_0\mathrm{\Phi }+f_1)^n`$ and $`f(\mathrm{\Phi })=f_0e^{n\mathrm{\Phi }}`$. For the first of these functions, the minimally coupled theory is obtained for $`f_0=0`$ and $`f_1^n=1/2`$ , whereas the induced gravity is obtained for $`f_1=0`$, $`f_0=\sqrt{ϵ/2}`$ and $`n=2`$. We note that the study of one coupling constant $`\omega (\varphi )`$ permit us to get informations on two types of coupling $`f(\mathrm{\Phi })`$ because $`\omega (\varphi )`$ and $`f(\mathrm{\Phi })`$ are linked by the differential equation (4). Hence to one type of function $`\omega `$, having one or several free parameters, can correspond more than one type of functions $`f`$. What we say above comes from the fact that to a power or exponential law for $`f(\mathrm{\Phi })`$ correspond only a power law for $`\omega (\varphi )`$.
### 3.3 The case $`3+2\omega =e^{2\varphi _c\varphi }`$.
We take the form $`3+2\omega =e^{2\varphi _c\varphi }`$, $`\varphi _c`$ being a real constant. This is an interesting case because, as in the subsection 3.1, when the scalar field vanishes, the coupling constant tends towards -1, which is the low limit of the string theory, whereas when it becomes infinite, the coupling constant tends towards infinity and the theory towards General Relativity if $`\varphi _c>0`$.
Here, we can not integrate equation (21) in a closed convenient form. We rewrite the equations (9) and (21) in the following form:
$$H(\varphi )=\tau =Ae^{\varphi _c\varphi }𝑑\varphi \tau _0=A\varphi _c^1e^{\varphi _c\varphi }\tau _0$$
(45)
$$G(\varphi )=\stackrel{~}{\tau }=\stackrel{~}{\varphi }_0e^{\varphi _c\varphi }\varphi ^1𝑑\varphi \stackrel{~}{\tau }_0$$
(46)
That means we have $`\varphi (\tau )=H^{(1)}(\tau )`$ and $`\varphi (\stackrel{~}{\tau })=G^{(1)}(\stackrel{~}{\tau })`$. By equalling these last two expressions and reversing (45), we get :
$$\stackrel{~}{\tau }=G(H^{(1)})=G(\varphi )=G(\varphi _c^1ln\left[\varphi _cA^1(\tau +\tau _0)\right])$$
(47)
With (23), we can easily obtain the metric functions :
$$a=e^{\alpha _1G(\varphi _c^1ln\left[\varphi _cA^1(\tau +\tau _0)\right])+\alpha _0}/\sqrt{(A\varphi _c)^1ln\left[\varphi _c(\tau +\tau _0)\right]}$$
(48)
and the same form for $`b(\tau )`$ and $`c(\tau )`$ with their integration constants. The reality conditions for the metric functions will be $`\varphi _cA^1(\tau +\tau _0)>0`$ and $`\varphi _c^1ln\left[\varphi _cA^1(\tau +\tau _0)\right]>0`$.
Hence, if $`\varphi _c<0`$, the metric function will be real if $`\tau ]A\varphi _c^1\tau _0,\tau _0[`$, and if $`\varphi _c>0`$, we will have $`\tau ]A\varphi _c\tau _0,+\mathrm{}[`$. The first derivative of (48) will be of the sign of $`\alpha _1\stackrel{~}{\varphi }_0\varphi _cA^1(\tau +\tau _0)1/2`$. For all value of $`\varphi _c`$, when $`\tau =A(2\alpha _1\varphi _c\stackrel{~}{\varphi }_0)^1\tau _0`$, $`da/d\tau `$ vanishes in the following cases:
\- when $`\tau ]A\varphi _c^1\tau _0,\tau _0[`$, that means $`\varphi _c<0`$, if $`2\alpha _1\stackrel{~}{\varphi }_0>1`$,
\- when $`\tau ]A\varphi _c^1\tau _0,+\mathrm{}[`$, that means $`\varphi _c>0`$, if $`2\alpha _1\stackrel{~}{\varphi }_0[0,1]`$.
From these results and after a numerical study we can write that :
* If $`\varphi _c<0`$, $`\tau ]A\varphi _c\tau _0,\tau _0[`$:
+ If $`\alpha _1<(2\stackrel{~}{\varphi }_0)^1`$, the metric function is decreasing and tends to infinity, in a positive manner when $`\tau A\varphi _c^1\tau _0`$, and to zero when $`\tau \tau _0`$.
+ If $`\alpha _1>(2\stackrel{~}{\varphi }_0)^1`$, the metric function tends to zero for these two values of $`\tau `$ and has a maximum . So, if the three integration constants $`\alpha _1`$, $`\beta _1`$, $`\gamma _1`$ of each of the metric functions are such that they are all superior to $`(2\stackrel{~}{\varphi }_0)^1`$, we have a close Universe (for the time) which exists during a finite time in the $`\tau `$-time. Since $`dt/d\tau =abc`$, this quantity vanishes in $`\tau =A\varphi _c^1\tau _0`$ and $`\tau =\tau _0`$ and then $`t(\tau )`$ stays finite for these two values and the Universe also exists during a finite t time.
* If $`\varphi _c>0`$, $`\tau ]A\varphi _c\tau _0,+\mathrm{}[`$ :
+ If $`\alpha _1<0`$, the metric function decreases from infinity to zero.
+ If $`\alpha _1[0,(2\stackrel{~}{\varphi }_0)^1]`$, the metric function has a minimum and tends to $`+\mathrm{}`$ when $`\tau `$ tends to $`A\varphi _c^1\tau _0`$ or $`+\mathrm{}`$. If the three integration constants $`\alpha _1`$, $`\beta _1`$, $`\gamma _1`$ are all in the same interval, the Universe will have a bounce since each metric function has a minimum.
+ If $`\alpha _1>(2\stackrel{~}{\varphi }_0)^1`$, the metric function is increasing from zero to infinity with an infinite slope.
## 4 Conclusion
In the conformal frame, the scalar field is minimally coupled. Hence, the spatial components of the field equations are exactly the same as in General Relativity and their solutions for the Bianchi type I model are the kasnerian solutions . The Klein-Gordon equation and the constraint equation, that are different from General Relativity, impose that the sum of the square of the Kasner exponents is always inferior to unity. Their sum is equal to one. Hence, there are always two or three positive Kasner exponents.
To express the metric function in the Brans-Dicke frame, we have equated the expressions of the scalar field in both Brans-Dicke and conformal frames and then deduced the time $`\stackrel{~}{\tau }`$ of the conformal frame as a function of the time $`\tau `$ of the Brans-Dicke frame. Then it is easy to find the form of the metric functions in this last frame. The amplitude of the metric functions and the sign of their first derivative in the $`\tau `$ time of the Brans-Dicke frame are the same as in the $`t`$ time. This is not the case for the second derivative of the metric functions.
We have studied three forms of the coupling constant $`\omega (\varphi )`$ and found solutions for which the Universe could avoid the singularity. We have also detected kinetic inflation for the two first examples and notice that, under some conditions, it can end naturally. For small or large value of the $`\tau `$ time, the coupling constant can become infinite or constant. It is always interesting to find classes of coupling constant for which it tends naturally toward -1 or infinite for small or large value of $`\tau `$ because such a class of theories tends respectively toward string theory in the low-energy limit and General Relativity. It seems to be true in the special case $`3+2\omega =(1\varphi /\varphi _c)^2`$ and for $`3+2\omega =e^{2\varphi _c\varphi }`$.
Figures 1 to 4 : forms of the metric functions when $`3+2\omega =2\beta (1\varphi /\varphi _c)^2`$.
Figures 5 to 12 : forms of the metric functions when $`3+2\omega =\varphi _c^2\varphi ^{2m}`$. |
warning/0001/astro-ph0001459.html | ar5iv | text | # High–resolution radio observations of Seyfert galaxies in the extended 12–micron sample – I. The observations.
## 1 Introduction
Exceptional amounts of energy are being released at the centres of a few percent of all galaxies. These Active Galactic Nuclei (AGN) are among the most luminous objects in the Universe and may emit more radiation than an entire galaxy from a region thought to be around ten thousand times as small. Seyfert nuclei, which are usually found in nearby spiral galaxies, are an important class of AGN because of the high quality and wide variety of information we may obtain about them and their host galaxies: they are sufficiently close and sufficiently luminous to be observed with good linear resolution using a variety of important techniques. For these reasons, samples of Seyfert galaxies may be defined more selectively than other classes of AGN and permit detailed comparisons of a wider range of properties.
Radio studies of several Seyferts show highly–collimated structures similar to those found in radio galaxies e.g. Markarian 3 (?), Markarian 6 (?), Markarian 463 (?), NGC 1068 (?), and NGC 4151 (?). Hubble Space Telescope images have confirmed that small–scale radio structures are often associated with individual narrow–line–region features (?; ?), as previously suspected from ground–based spectroscopy (?; ?) and it is now clear that the outflows which cause collimated radio structures have a direct influence on the narrow–line emission we observe (see models by ? and ?). Despite this, the importance of collimated outflows from Seyfert nuclei is often overlooked. In order to incorporate the radio properties of Seyfert nuclei into models of their activity and provide the context for studies of well–known individual Seyfert galaxies, it is necessary to document their generic radio properties.
Early statistical studies of the radio properties of Seyfert galaxies were carried out by ? and ?. The radio properties of the following samples of Seyferts have been studied subsequently; Seyferts from the lists of Markarian (?; ?), distance–limited samples (?; ?; ?), an X–ray flux–limited sample (?), the CfA Seyfert sample (?; ?; ?), samples selected from the literature (?; ?), far–infrared–selected samples (?; ?; ?) and the mid–infrared–selected extended 12 $`\mu `$m sample (?). Large, unbiased samples are required to compare the properties of the two types of Seyfert galaxy, but the definition of such samples is problematic. Ideally, all physically–similar objects within a certain volume of space should be selected, however this is extremely difficult to achieve. In practice the selection criteria together with observational limitations tend to bias samples towards certain classes of object.
In this paper we present observations of the extended 12 $`\mu `$m AGN sample of ? using an observing technique which has been chosen to optimize sensitivity to small–scale radio structures i.e. the VLA in A–configuration at 8.4 GHz. This sample is one of the largest homogeneously–selected samples of Seyfert galaxies available and contains well–matched populations of type 1 and type 2 sources. The paper is organized as follows; in Section 2 we briefly describe the sample, in Section 3 we describe the observations and data reduction, in Section 4 we present the results and in Section 5 we explain how 5 radio–loud objects from the sample may be identified. An analysis of the radio properties of the sample will be carried out in subsequent papers.
A value of H = 75 km s<sup>-1</sup>Mpc<sup>-1</sup> is assumed throughout.
## 2 THE EXTENDED 12–MICRON SAMPLE
The extended 12 $`\mu `$m AGN sample of ? is an extension of the original 12 $`\mu `$m AGN sample of ? to fainter flux levels using the IRAS Faint Source Catalogue Version 2 (?). From an initial sample of 893 mid–infrared–bright sources, AGN catalogues were used to define a subsample of active galaxies which contains 118 objects, the majority of which are Seyfert galaxies. The sample was selected at 12 $`\mu `$m in order to minimize wavelength–dependent selection effects. ? proposed that this wavelength carries an approximately constant fraction, around 20%, of the bolometric flux for quasars and both types of Seyfert.
Previous radio observations of part of the extended 12 $`\mu `$m AGN sample have been made by ? who observed 16 sources with the same resolution as the observations presented in this paper and ? who presented lower resolution observations of the original 12 $`\mu `$m sample which includes 51 sources from the extended sample. The hard X–ray properties of the sample have been studied by ? and the host galaxies of the sample have been studied by ?.
## 3 THE OBSERVATIONS
We have made new observations of 87 sources and when combined with matched observations of 19 sources from the CfA Seyfert sample (?) these observations cover 91% of the AGN sample of ?. Of the 12 sources not observed, 10 are unobservable at the VLA due to their low declination and two are well–studied radio sources (3C 120 and 3C 273).
An almost identical observing strategy to that used by ? was followed for all observations. The Very Large Array<sup>1</sup><sup>1</sup>1Operated by Associated Universities Inc. under contract with the National Science Foundation. ? give instrumental details. (VLA) was used in A–configuration at 8.4 GHz in snapshot mode with approximately 16 minutes on each source and approximately 6 minutes on a corresponding phase calibrator. J2000 co–ordinates were used, with phase calibrators selected from the A–category NRAO list and the list of ?. In theory the positional accuracy of the final images is 0.005 arcsec, but to allow for atmospheric fluctuations we adopt a conservative estimate of 0.05 arcsec. Due to an on–line computer failure during the initial observing run, the 87 new sources observed were split between two separate observing runs. The final radio database will therefore contain data from 3 separate epochs; 15th July 1995 (63 sources), 25th November 1996 (24 sources) and June 1991 (19 sources from the CfA sample). The epoch of each new observation is indicated in Table 1.
Between the 1995 run and the 1996 run, the two default VLA observing frequencies changed from 8.415 and 8.464 GHz to 8.435 and 8.485 GHz to avoid interference. The mean 1–$`\sigma `$ noise level for all maps was 53$`\pm `$20 $`\mu `$Jy; 56$`\pm `$16 $`\mu `$Jy for the 1995 maps compared with 44$`\pm `$26 $`\mu `$Jy beam<sup>-1</sup> for the 1996 maps. Table 1 gives the noise levels for individual maps. Mean noise values have been calculated using only those thermal–noise–limited sources (peak flux $`<`$ 100 mJy beam<sup>-1</sup>) observed at high elevation (axial ratio of beam $`<`$ 5), errors represent the standard deviation of the mean.
All data processing, including calibration and mapping, was performed using the Astronomical Image Processing System (AIPS) in the standard way. The data were Fourier–transformed using a natural weighting scheme in order to maximize sensitivity. After CLEAN deconvolution, the maps were restored with a 0.25 arcsec FWHM Gaussian beam. The largest detectable angular size in this configuration is 3.5 arcsec. Seventeen strong sources, whose peak flux density exceeded 10 mJy, were subjected to several cycles of self–calibration. In the final maps the sensitivity approached thermal noise levels (1–$`\sigma `$ $`<`$ 120 $`\mu `$Jy beam<sup>-1</sup>) for all except 5 sources, these were either bright ($`>`$ 100 mJy beam<sup>-1</sup>) or observed at low elevation (axial ratio of beam $`>`$ 5).
## 4 RESULTS AND ANALYSIS
### 4.1 Observational results
Contour maps of all detected sources are shown in Figure 1. The ellipse in the lower left–hand corner shows the shape of the restoring beam at half power. Optical nuclear positions are marked by a cross where they are available from ?, ? and ?; the diameter of the cross shows the 2–$`\sigma `$ positional uncertainty.
Descriptions of the radio maps are shown in Table 1 which is arranged as follows; Column 1: Galaxy name, a dagger ($``$) indicates data from the July 1995 observing run and an asterisk ($``$) indicates data from the November 1996 observing run. Column 2: Beam major axis, $`\theta _{maj}`$. Column 3: Beam minor axis, $`\theta _{min}`$. Column 4: Beam position angle in degrees, PA. Column 5: Peak flux density, S<sub>peak</sub> (mJy/beam). Column 6: The root mean square noise, $`\sigma `$, measured at the edge of the field where no deconvolution techniques were applied ($`\mu `$Jy/beam). Column 7: Contour levels for each map ($`\mu `$Jy/beam). Where possible the base contour level was set at 3–$`\sigma `$ however in 28% of the maps instrumental ‘side–lobes’ were present and the base contour level was set at a level which excluded obvious spurious features. Side–lobes result from the incomplete coverage of the uv–plane of the interferometer. They are usually removed by deconvolution techniques such as CLEAN but can remain strong when the original data are of poor quality or badly calibrated, or when the target source is particularly bright or observed at low elevation.
### 4.2 Description of sources
A brief description of those sources with unusual radio morphologies is given below (all positions are given in J2000 co–ordinates).
NGC 34 (Markarian 938): This infrared–luminous galaxy is in the advanced stage of a merger with two nuclei separated by approximately 6 kpc. The weak \[O III\] emission and strong H<sub>α</sub> emission, which is distributed over the entire galaxy, indicate that this galaxy is more properly classified as a starburst galaxy rather than a Seyfert galaxy (?).
Note that NGC 34 contains a bright radio source more luminous than two–thirds of the Seyferts observed (P<sub>8.4GHz</sub>=1$`\times `$10<sup>22</sup> WHz<sup>-1</sup>) and slightly resolved with a deconvolved size of 0.4 arcsec (150 pc) and position angle of 140. High–resolution mid–infrared images show a double source separated by approximately 1.2 arcsec, with the fainter source at a position angle of around 180 from the brighter source (?). Radio observations by ? show a faint southern radio extension, not found in our map, which is consistent with the infrared structure.
NGC 526A: NGC 526A is strongly interacting with a galaxy to the east. The position of the slightly resolved radio source is closer to the peak in the optical continuum than to the apex of the putative emission–line wedge identified by ?; whereas the emission–line wedge has a position angle of 123, that of the radio source is 43.
Markarian 1034 (V Zw 233): Markarian 1034 is an interconnected pair of Seyfert 1 galaxies, MCG+05-06-035 (PGC 0009071) at $`\alpha `$=02<sup>h</sup> 23<sup>m</sup> 18.84<sup>s</sup>, $`\delta `$=+32 11 18.2<sup>′′</sup> and MCG+05-06-036 (PGC 0009074) at $`\alpha `$=02<sup>h</sup> 23<sup>m</sup> 21.99<sup>s</sup>, $`\delta `$=+32 11 49.6<sup>′′</sup>; positions are taken from the NASA/IPAC Extragalactic Database (NED) as described by ?. The new observations show MCG +05-06-036. This source has extended ‘wings’ of radio emission symmetrical about a central bright component.
NGC 1125: The radio source has a clear linear structure with 3 aligned compact components at a position angle of 120. The radio structure is not aligned with the slightly resolved \[O III\] emission at a position angle of 56 (?).
NGC 1365: There is no alignment between the four compact radio components observed and none is co–incident with the photographic position of the nucleus which lies at $`\alpha `$=03<sup>h</sup> 33<sup>m</sup> 35.57<sup>s</sup>, $`\delta `$=-36 08 22.9<sup>′′</sup> (NED), or either of the two infrared sources observed by ?, the brightest of which lies at $`\alpha `$=03<sup>h</sup> 33<sup>m</sup> 36.17<sup>s</sup>, $`\delta `$=-36 08 25.9<sup>′′</sup> (positional error of 1.5 arcsec).
IRAS F04385-0828: We detect another weak radio source to the west at a projected linear separation of 15.4 kpc, it has an integrated flux of 2.05 mJy and lies at $`\alpha `$=04<sup>h</sup> 40<sup>m</sup> 51.45<sup>s</sup>, $`\delta `$=-08 22 23.85<sup>′′</sup>. This weak source is well beyond the optical radius of the host galaxy, which extends westwards approximately 6 kpc, and is unlikely to be related to the Seyfert nucleus.
Markarian 6 (UGC 3547): This source contains a central, well–collimated radio structure. On larger scales, features suggestive of shells or bubbles are seen at varying position angles (?; ?).
Markarian 79 (UGC 3973): This large radio source shows a linear structure with 3 clearly aligned components. Lower resolution measurements, at lower frequency, show that the southern component is stronger than the northern component (?), whereas in our image these components are almost equally bright.
NGC 2639 (UGC 4544): The source has a bright core and symmetrical east–west ‘wings’. In our map, the core–to–wings brightness ratio is an order of magnitude higher than in the 6 cm map of ?; probably due to core variability. NGC 2639 displays the rare properties of H<sub>2</sub>O megamaser emission (?) and VLBI–scale radio emission (?). It has also been classified as a LINER.
NGC 2992 (Arp 245): This edge–on, interacting galaxy has unusual ‘loops’ of diffuse radio emission (?; ?).
NGC 4922A/B: NGC 4922 is a system of 3 galaxies situated at $`\alpha `$=13<sup>h</sup> 01<sup>m</sup> 24.50<sup>s</sup>, $`\delta `$=+29 18 29.9<sup>′′</sup> (Seyfert 2), $`\alpha `$=13<sup>h</sup> 01<sup>m</sup> 24.67<sup>s</sup>, $`\delta `$=+29 18 33.0<sup>′′</sup> and $`\alpha `$=13<sup>h</sup> 01<sup>m</sup> 25.26<sup>s</sup>, $`\delta `$=+29 18 49.58<sup>′′</sup> (PGC 044896/FIRST J130125.2+291849); positions are taken from NED. We have detected PGC 044896 (an unresolved 7.8 mJy source at $`\alpha `$=13<sup>h</sup> 01<sup>m</sup> 25.26<sup>s</sup>, $`\delta `$=29 18 49.53<sup>′′</sup>, shown in Figure 1), but not the nearby Seyfert nucleus.
NGC 5135: Despite the bright radio flux of the nucleus at lower resolutions and frequencies (?), no radio emission is observed at the nucleus in the current observations. ? have suggested that, “most of the radio emission of this Seyfert galaxy emanates from structures on either side of the nucleus rather than from the nucleus itself”. We detect a weak component with a flux density of 2.33 mJy at $`\alpha `$=13<sup>h</sup> 25<sup>m</sup> 44.9<sup>s</sup>, $`\delta `$=-29 50 16.17<sup>′′</sup>, but it is unlikely to be related to the active nucleus (see Section 4.3).
Markarian 273 (UGC 8696, I Zw 071): This double–nucleus source is one of the most ultra–luminous infrared galaxies known. The radio continuum shows two compact radio components separated by approximately 600 pc. The radio structure has been mapped previously by ?, who classify it as a compact starburst, and ?.
Markarian 463 (UGC 8850): Three aligned north–south radio components are detected in this well–studied double–nucleus galaxy. North–south radio structures have previously been observed from VLBI–scales (?; ?) to up to 18 kpc south of the nucleus (?). The new observations match the 6 cm radio structure observed by ? with the exception of the weak central radio component which appears to coincide with a bright knot in the aligned optical jet (?). This is the second most radio–luminous Seyfert in the extended 12 $`\mu `$m sample.
NGC 5506: This edge–on galaxy has a compact core surrounded by a diffuse halo. The ‘loop’ identified by ? is just traceable to the north–west of the core (see also ?).
UGC 9913 (Arp 220): The two closely separated radio components are the nuclei of this well–known double–nucleus galaxy. The nature of its activity is uncertain. Despite being the 7th most radio–luminous radio–quiet source in the extended 12 $`\mu `$m sample, recent VLBI observations by ? provide strong evidence that the radio emission from the north–western component originates in a compact nuclear starburst.
IRAS F22017+0319: If the weak northernmost component is included, this source is a linearly aligned triple radio source.
NGC 7314: This variable X–ray source is a north–south radio double.
MCG-03-58-007: A north–south radio double.
NGC 7582: The source has a diffuse radio structure consistent with the lower–resolution observations of ?.
### 4.3 Radio properties
Radio parameters for all observed sources from the extended 12 $`\mu `$m AGN sample are given in Table 2; radio parameters derived by ? for 19 sources which belong to the CfA sample have also been included. The table is organized as follows; Column 1: Galaxy name. An asterisk ($``$) is used to indicate those sources whose radio parameters are taken from ?. Column 2: Seyfert type. For this paper objects are classified simply as type 1 or type 2 following ? except where alternative classifications have been proposed by ? (9 sources) and ? (NGC 34). Reclassified sources are labeled using the following the abbreviations; Q for Quasar, L for LINER and sb for Starburst. For these sources, the classification given by Rush et al is given in parenthesis. Column 3: Redshift, as taken from ?. Columns 4 and 5: Right ascension, RA (h, m, s), and declination, Dec (deg, arcmin, arcsec), of each radio component in J2000 co–ordinates. Positions were determined by Gaussian fitting. For point–like sources the accuracy of fit, estimated by comparing the modelled and measured flux densities, was found to be around 5%. Column 6: Integrated flux density of each component, S (mJy/beam). The mean flux density of each component was measured directly from the map. Care was taken to ensure that the effects of non–zero background levels were taken into account. The uncertainty on each flux measurement may be estimated by combining the calibration error (? estimated a value of around 4%) and the map error for each source (around 4% for strong sources and 14% for weak sources). For those sources which were not detected an upper limit on their radio flux of 5–$`\sigma `$ has been assumed. In multiple component sources the component nearest the photographic position (taken from NED) is labeled with a dagger symbol ($``$). Ninety–three percent of the sources had at least one radio component within 2–$`\sigma `$ of the available photographic position; of the others, four show close alignment between the radio position and the galactic centre as judged from Palomar Sky Survey images (MCG-3-7-11, NGC 5194 = M51, NGC 5033 and NGC 5005) and 2 have been excluded from further analysis (NGC 4922A/B and NGC 5135). The sky density of sources above the average 3–$`\sigma `$ detection threshold of 159 $`\mu `$Jy at 8.4 GHz has been estimated by ? and implies a 2% probability of detecting an unrelated source within the 51 $`\times `$51 arcsec<sup>2</sup> field–of–view. Column 7: Maximum angular size of radio structure in arcseconds, $`\theta `$. For unresolved sources an upper limit to the angular size was taken as one third of the major axis of the beam. For slightly resolved sources the angular size was taken as the length of the major axis of the nominal deconvolution<sup>2</sup><sup>2</sup>2The nominal deconvolution is obtained when the Gaussian fit to a component is deconvolved from the CLEAN beam for single Gaussian fits, or the maximum separation of the peaks for multiple Gaussian fits. For all other sources the maximum size measured was either the maximum peak separation (for those sources with point–like components) or the maximum size at the lowest contour (for those sources with diffuse components). Column 8: Maximum linear size in parsecs, D, assuming H = 75 km s<sup>-1</sup>Mpc<sup>-1</sup>. Column 9: Type of radio structure, T<sub>rad</sub>, according to the notation used by ?: U for single unresolved sources, S for single slightly–resolved sources, A for sources with ambiguous structures, D for sources with diffuse structures and L for sources with possible linear structures (sources with two components, sources with three or more aligned components or sources with extended linear components). Gaussian fitting was used to distinguish between types U and S, type S sources had a signal to noise ratio greater than 20 and a nominal deconvolution size greater than one third of the beam at FWHM. Column 10: Position angle of radio structure, PA, measured North to East from 0 to 180. For partially resolved sources (type S) the position angle of the nominal deconvolution was used. The position angle of clearly resolved linear sources (type L) was measured directly from the map.
## 5 THE IDENTIFICATION OF RADIO-LOUD SOURCES
The distinction between ‘radio–loud’ and ‘radio–quiet’ AGN (?; ?) is widely accepted and usually thought to result from truly distinguishable physical processes (e.g. ?). In this section we describe the identification of five radio–loud objects in the extended 12 $`\mu `$m AGN sample (OJ 287, 3C 120, 3C 234, 3C 273 and 3C 445); these sources will be excluded from further statistical analysis and henceforth the remaining sources will be referred to as the extended 12 $`\mu `$m Seyfert sample.
We have chosen to use the radio to far–infrared luminosity ratio as the main diagnostic of ‘radio–loudness’. This has been done by using IRAS FSC 60 $`\mu `$m luminosities and our newly–measured 8.4 GHz (3.6 cm) radio luminosities; single–dish radio observations from ? were used for the two unobserved radio–loud objects, 3C 120 and 3C 273, and 3C 445 for which the majority of the flux is outside the field–of–view of the new observations. The radio to far–infrared luminosity ratio is probably a more useful indicator of radio–loudness than the radio luminosity alone given that the bolometric luminosities of the sources in the sample are likely to span several orders of magnitude. Figure 2 is a histogram showing the frequency of sources in the extended 12 $`\mu `$m AGN sample per logarithmic radio to far–infrared luminosity ratio interval $`\mathrm{\Delta }`$log(L<sub>3.6cm</sub>/L<sub>60μm</sub>) = 0.4; the L<sub>3.6cm</sub>/L<sub>60μm</sub> ratio for each detected source is given in Table 5. The 5 excluded sources have the highest radio to far–infrared luminosity ratios of the sample; all have log(L<sub>3.6cm</sub>/L<sub>60μm</sub>) ratios greater than -4.6, whereas log(L<sub>3.6cm</sub>/L<sub>60μm</sub>) ratios are less than -5.1 for the Seyferts. These values are in agreement with the radio to far–infrared luminosity ratios used by ? to identify 3 of the same radio–loud sources in the original 12 $`\mu `$m AGN sample (OJ287, 3C 120 and 3C 273). The mean ratio between the 8.4 GHz A–configuration flux to the 5 GHz D–configuration flux for those sources observed by ? is around 0.3; this ratio is consistent with a mean radio spectral index of $`\alpha `$ = -0.7 (?) and a A–configuration to D–configuration flux ratio of around 0.5. Using this flux ratio, the radio to far–infrared luminosity ratios used by ? translate to -4.7 $`<`$ log(L<sub>3.6cm</sub>/L<sub>60μm</sub>) $`<`$ -2.3 for radio–loud sources and -6.3 $`<`$ log(L<sub>3.6cm</sub>/L<sub>60μm</sub>) $`<`$ -4.8 for radio–quiet sources.
Note that there is no clear evidence for a bimodal distribution of the radio to far–infrared luminosity ratio, possibly because of the infrared flux-limit used to define the sample. As well as having the highest radio to far–infrared luminosity ratios, the radio–loud sources we have identified are also the 5 most radio luminous sources in the sample, being the only sources more luminous than L<sub>3.6cm</sub> $`>`$ 10<sup>24</sup> WHz. They are all well–known objects with powerful jets; three show super–luminal motions in their jet components (OJ287, 3C120 and 3C273, as cited by ?) and the other two are classical FR–II radio galaxies with radio structures hundreds of kiloparsecs in size (see ? and ? for images of 3C234 and 3C445 respectively). A strong reason for excluding these sources is that they are broad–line objects which, when grouped with the Seyfert 1 subsample, would systematically affect comparisons between the two Seyfert types.
## 6 SUMMARY
The maps presented in this paper reveal for the first time the sub–arcsecond radio structures of Seyferts contained in the extended 12 $`\mu `$m AGN sample. They provide a large and homogeneously–selected database for investigating the generic properties of compact radio cores in Seyfert nuclei.
Seventy–five of the 87 sources observed were detected; 36 contain single unresolved radio sources, 13 contain single slightly–resolved radio sources, 9 contain radio sources with diffuse or ambiguous structures, 8 contain radio sources with two distinguishable components and 9 contain radio sources with three or more linearly–aligned components or extended linear structures. Subsequent papers will discuss the statistical properties of the sample in detail, paying particular attention to comparisons of the radio powers and radio morphologies of the two Seyfert types.
## 7 ACKNOWLEDGMENTS
AHCT would like to acknowledge the receipt of a studentship from the Particle Physics and Astronomy Research Council and a visit funded by the STScI visitor program. Part of this research was supported by the European Commission, TMR Programme, Research Network Contract ERBFMRXCT96-0034 “CERES”. We have made use of NASA’s Astrophysics Data System Abstract Service, the NASA/IPAC Extragalactic database (NED), which is operated by the Jet Propulsion Laboratory. |
warning/0001/cond-mat0001095.html | ar5iv | text | # Spin twists, domain walls, and the cluster spin-glass phase of weakly doped cuprates
## I Introduction:
The now frequent experimental observations of spin and/or charge modulations in the cuprate superconductors and related doped transition metal oxides was predicted by the “frustrated phase separation” phenomenology of Emery and Kivelson. Their theoretical considerations involved the assertion that a doped Mott insulator phase separates as a consequence of the competition between the kinetic energy of mobile holes and the magnetic energies of an antiferromagnetic (AFM) phase with long-range correlations. While it seems unlikely that this scenario is correct in the strong correlation limit , when one adds the Coulomb interaction into this problem Emery and Kivelson argued that macroscopic phase separation became “frustrated”, and the resulting anomalous normal state possessed low-energy fluctuations corresponding to stripes, or domain walls . To connote that these entities correspond to metallic stripes, it is now common to refer to these structures as “rivers of charge” .
Recent neutron scattering studies of the single-layer La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub> (LSCO) system have revealed that (at least in experimental results to date) an elastic magnetic response associated with static stripe-like correlations are only found in (i) low $`x`$ systems ($`x0.06`$) at low temperatures such that the transport is that of a doped semiconductor , and (ii) in the famous $`x1/8`$ system. One could interpret both the strongly disordered, low temperature results, and the $`x1/8`$ data, as evidence that pinning effects are necessary, either from disorder or commensurability interactions, to produce static stripes. Such arguments are consistent with the successful approach of Tranquada and coworkers in producing static stripes that could then be observed in scattering experiments in a variety of systems .
The low $`x`$, low temperature region of the LSCO phase diagram in which the static magnetic stripe correlations are found corresponds to the spin-glass phase of LSCO. Early magnetic resonance work on this system suggested that it is appropriate to think of this phase as a cluster spin glass, so named because small clusters of spins achieve their own short-range AFM correlations, but the cluster-cluster ordering is spin-glass like. A numerical simulation of this phase, coupled with new crystals and new susceptibility data, lended support to this characterization .
In the latter paper , one conundrum associated with the mechanism behind the formation of the cluster spin glass phase was pointed out, and goes as follows: The frustrated phase separation phenomenology claims that support for such physics is found in the existence of the cluster spin glass phase . However, detailed analysis of the transport in this region of the LSCO phase diagram concludes that the transport is similar to that of a doped semiconductor . Thus, at least in the low-temperature cluster spin-glass phase the competition between kinetic and magnetic energies does not exist in the form proposed originally by Emery and Kivelson, and the question can then be asked, does the absence of the holes’ kinetic energy from extended states not eviscerate the frustrated phase separation phenomenology as a viable mechanism associated with the formation of the cluster spin glass phase? Put another way, the numerical simulations of Ref. that found evidence for the character of the spin-glass phase being like that of a cluster spin glass produced this spin texture with zero kinetic energy, and thus is the kinetic energy a necessary ingredient in the formation of stripe phases ?
This question becomes more important in view of recent experiments of Julien, et al. . These NMR/NQR results demonstrated the existence of a so-called charge glass at higher temperatures, followed by the appearance of the superconducting phase at lower temperatures, followed by the cluster spin glass phase at the lowest temperatures. The frustrated phase separation phenomenology predicts this sequence of charge glass/cluster spin glass phases, and thus unlike the above arguments, suggests that the cluster spin glass is stabilized by the pinning of the charge stripes by defects, followed by the subsequent freezing of the spin degrees of freedom within clusters defined by the pinned stripes of the charge glass phase. Unfortunately, again, the transport of this system at low temperatures is insulating (for, say, $`T<75K`$), so in the spin glass phase there are no rivers of charge which could “carve out” the domain walls of the cluster spin glass phase!
In this manuscript we present analytical results that formalizes the claim that via quenched disorder from the Sr impurities one can produce the topology of pinned stripes without any kinetic energy. To this end, we derive the spin distortion pattern produced by such quenched disorder (which frustrates the background AFM order), and then demonstrate that this theory successfully predicts the stability of localized clusters of AFM correlated spins produced in a situation with zero kinetic energy. In this case, the origin of the stripes associated with the domain walls comes from the spin twists between the clusters, the clusters themselves having been produced by spin twists of the spin texture as the background spins attempt to accommodate the frustrating magnetic interactions produced by the quenched disorder. A well known extrapolation to higher temperatures then implies that the qualitatively identical spin twists generated by mobile carriers must be part of the mechanism associated with the formation of stripes.
We wish to make clear that our paper does not claim to be the first to propose that magnetic interactions in general, and spin twists in particular, are important in the formation of stripes. Firstly, the work of Salem and one of us investigated the problem of frustrating FM bonds whose locations could be chosen such that the ground-state energy was minimized. It was found that when quantum fluctuations were included, if the magnitude of the frustrating interaction was smaller than that of the background majority spins, periodic stripes of frustrating bonds were the ground state configuration. (So, in these ground states, again there is no kinetic energy but stripe phases are indeed encountered.) Secondly, when the frustrating bonds cannot chose their (static) positions but are fixed by the Sr impurity ions, the numerical simulations (mentioned above) of Ref. suggested the presence of domain walls between the clusters of the cluster spin-glass phase. More recently, work of Stojkovic and coworkers examined a version of the mobile hole problem by implementing a purely magnetic model that included the long-ranged spin twists produced by mobile holes (as well as the frustrating Coulombic energy between the carriers) and found many of the magnetic structures encountered in , including stripe phases. Lastly, White and Scalapino, who find evidence for stripe structures in the $`t`$$`J`$ model for mobile holes , note that the charge and spin distributions of striped structures attempt to accommodate the frustration (read: spin twists) on the magnetic background produced by the mobile holes.
Looking at the totality of the evidence in this and the above-mentioned papers we believe that one can make a strong case that there is a similar mechanism at work in the formation of stripes in all of these situations, and that this mechanism is spin twists.
Our paper is organized as follows. In the next section we present a detailed theoretical analysis of the effects of quenched disorder on the spin texture in systems such as weakly doped LSCO, producing a reliable analytical theory of the spin distortions both near the frustrating bond and in the far field. We use this distortion field to produce an accurate interaction functional between pairs of such bonds, which we then use to demonstrate the stability of such clusters in the cluster spins glass phase. In particular, this leads to a clear identification of the local AFM order parameter of each cluster. Finally, we show the resulting spin texture between two such clusters, and demonstrate how (local) stripe configurations can be stabilized in the cluster spin-glass phase.
## II Core solution and energies of the single bond problem:
### A Hamiltonian and definitions
We consider the familiar model of magnetism in the CuO planes of the high $`T_\text{c}`$ cuprates in which the copper ions and oxygen holes are treated as a lattice of spins governed by a Heisenberg Hamiltonian
$$H=\underset{ij}{}J_{ij}𝐒_i𝐒_j$$
(1)
where $`ij`$ denotes a summation over near neighbour pairs of spins and the $`J_{ij}`$ are the exchange interaction integrals. For an undoped lattice these spins are the Cu spins and the exchange integrals are equal and negative; the ground state corresponds to an AFM ordered state on a square lattice. In what follows we simplify our considerations by using classical spins, implying that only the transverse (viz., moment reorientation) and not the longitudinal (viz., moment magnitude) spin-spin interactions are included .
The simplest model of the effects of doping in weakly doped cuprates at low temperatures was proposed by Emery, and corresponds to localizing the holes on oxygen sites and replacing the AFM Cu-Cu superexchange for this occupied bond with an effective FM exchange. The phase diagram of the multiply doped version of this model was produced by Aharony, et al . Although detailed transport analysis of this part of the LaSrCuO phase diagram has shown that a slightly different model of the localized dopants is required for a direct comparison to experiments, the FM bond model (which we shall refer to as the frustrating bond model) is more amenable to analytical study, for reasons that we shall elaborate on below, and shall be used throughout this paper.
Thus, we consider the Hamiltonian of Eq. (1) wherein the $`i,j`$ label sites of a square lattice (that is, only the Cu spins and the effective interactions between them are considered) and the exchange interaction integral between two adjacent sites $`i`$ and $`j`$ has the form
$$J_{ij}=\{\begin{array}{cc}\lambda J\hfill & \text{with probability }x/2\hfill \\ J\hfill & \text{with probability }1x/2\hfill \end{array}$$
(2)
where $`J`$ and $`\lambda `$ are positive constants and $`\lambda `$ represents the relative strength of the ferromagnetic and antiferromagnetic bonds . The doping level $`x`$ could be, say, either the Sr doping level in $`\mathrm{La}_{2\mathrm{x}}\mathrm{Sr}_\mathrm{x}\mathrm{CuO}_4`$ or the Ca doping level in $`\mathrm{Y}_{1\mathrm{x}}\mathrm{Ca}_\mathrm{x}\mathrm{Ba}_2\mathrm{Cu}_3\mathrm{O}_6`$ (noting that Neidermayer et al. has shown that the phase diagrams in these two systems are identical). If we now choose a coordinate system such that linear combinations of $`x`$\- and $`y`$-directed unit vectors span the lattice, then the Hamiltonian can be written explicitly as
$$H=\frac{1}{2}\underset{i}{}\underset{\widehat{a}}{}J_{i,i+\widehat{a}}𝐒_i𝐒_{i+\widehat{a}}$$
(3)
where $`i`$ is summed over all lattice sites and $`\widehat{a}`$ ranges over $`\pm \widehat{x},\pm \widehat{y}`$. The equilibrium condition corresponds to that of zero torque from the local effective field at each lattice site:
$$\underset{\widehat{a}}{}J_{i,i+\widehat{a}}𝐒_i\times 𝐒_{i+\widehat{a}}=0.$$
(4)
As is well known, the complication of treating a bipartite lattice (labelling the two sublattices as A and B sites) can be avoided by transforming the physical problem of FM bonds in a predominantly AFM background into the mathematically equivalent problem of AFM bonds in a FM background. These two pictures can be converted one to the other under the simple transformation (AFM $``$ FM) given below:
$`J_{ij}`$ $``$ $`J_{ij}`$ (5)
$`𝐒_i`$ $``$ $`\{\begin{array}{cc}+𝐒_i\hfill & \text{for }iA\hfill \\ 𝐒_i\hfill & \text{for }iB\hfill \end{array}.`$ (8)
Lastly, we note that the objects under consideration are classical spins, and we set their length to be one, scaling $`J`$ to be $`JS^2`$. Further, the ground states that we discuss in this paper all correspond to situations in which the spins lie in some plane, and thus from now on we restrict our formalism to describe planar spins. We denote the bulk direction of the spins by $`𝐒_{\mathrm{}}`$, and at any lattice site $`i`$ there exists a spin $`𝐒_i`$ characterized by the angle $`\psi _i`$ between the spin and the $`x`$-axis:
$$𝐒_i=\widehat{x}\mathrm{cos}\psi _i+\widehat{y}\mathrm{sin}\psi _i𝐒_i𝐒_j=\mathrm{cos}(\psi _i\psi _j).$$
(9)
We choose $`\psi _i=\varphi _i+\psi _{\mathrm{}}`$ where $`\psi _{\mathrm{}}`$ is taken to be the average angle of the spins over the bulk of the material. That is $`𝐒_{\mathrm{}}=\widehat{x}\mathrm{cos}\psi _{\mathrm{}}+\widehat{y}\mathrm{sin}\psi _{\mathrm{}}`$ so that the angle $`\varphi _i`$, defined according to $`\mathrm{cos}\varphi _i=𝐒_i𝐒_{\mathrm{}}`$, represents the deviation of the spin at site $`i`$ from the bulk direction. The collection $`\{\varphi _i\}`$ of the spin distortions at each lattice site constitutes the spin texture on the lattice.
The numbering scheme for the lattice sites near the frustrating bond (what from now on we call the bond sites) is shown in Fig. 1.
### B Spin deviations of a frustrating bond
In a FM lattice doped with a single AFM bond, the ground state solution to the spin texture is no longer obvious, and such a situation is called frustrated. We wish to produce an accurate analytical solution to this problem in both the far-field region and close to the frustrating bond. This will allow us to accurately track the energies of both the single and many bond problems.
Consider a purely FM lattice frustrated by the introduction of an $`x`$-directed AFM bond between the $`(0,0)`$ and $`(1,0)`$ lattice sites. We shall denote the spin distortions at these sites by $`\varphi _0`$ and $`\varphi _1`$, respectively. The Hamiltonian is then written as
$$H=J\underset{ij}{}^{}\mathrm{cos}(\varphi _i\varphi _j)+\lambda J\mathrm{cos}(\varphi _0\varphi _1)$$
(10)
where the prime indicates the omission of the AFM bond from the summation. It is no longer clear that the trivial solution $`\varphi _i=0`$ represents the ground state since there may be another solution with $`\pi /2<|\varphi _0\varphi _1|<3\pi /2`$ that the takes the system to a lower energy state.
As is well known, considerable information can be gained from an examination of the dynamical properties of a linearized system of equations of motion about a proposed equilibrium structure. Here, we briefly outline this formalism, since it will be important to our later work. For our purposes, the dynamical behaviour of a lattice of $`n`$ independent spins is modeled sufficiently by
$$\ddot{\varphi _i}=\underset{\widehat{a}}{}\mathrm{sin}(\varphi _i\varphi _{i+\widehat{a}})i=1,2,\mathrm{},n.$$
(11)
Close to the ordered equilibrium state $`\varphi _i=0`$, this behaviour is governed by the linearized system $`\ddot{𝐱}=M𝐱`$ where $`M=Df(\mathrm{𝟎})`$ is the derivative matrix of $`f`$ evaluated at the origin and $`𝐱`$ is related to the spin texture according to the row vector $`𝐱^T=[\varphi _1,\varphi _2,\varphi _3,\mathrm{},\varphi _n]`$.
Solving for the normal modes, and taking note of negative eigenvalues, the instability of the system to a non FM ordered ground state can be identified. This technique, when applied for larger and larger systems, reproduces the known result that an instability is first reached at $`\lambda _c=1`$ and that there is only one stable spin texture for all $`\lambda `$ exceeding $`\lambda _c`$.
Unlike such instability analysis, or the Fourier-based approach of Vanninemus, et al. , here we wish to develop a continuum theory capable of describing an infinite lattice including the spins in the immediate neighbourhood of the frustrating bond. To this end, we proceed as follows.
Let $`\varphi `$ be a function of a continuous variable $`𝐫`$ which ranges over the entire $`xy`$-plane such that $`\varphi _i\varphi (𝐫_i)`$. Then, provided that $`\varphi `$ is a smooth, slowly varying function of position, we may approximate the equilibrium condition for the undoped system (to lowest order ) by
$$^2\varphi =0.$$
(12)
Now consider a FM lattice frustrated by the introduction of a single $`x`$-directed AFM bond. The equilibrium spin distortions away from the core of the frustration are governed by Laplace’s equation. Choose the origin of the coordinate system centred on the bond, and solve Laplace’s equation, in polar coordinates, by separation of variables. The imposition of the appropriate solution symmetries yields
$$\varphi (r,\theta )=\underset{m=1,3,5,\mathrm{}}{\overset{\mathrm{}}{}}r^mA_m\mathrm{cos}m\theta $$
(13)
where we have adopted the convention that summations are over odd indices only.
It is clear that for sufficiently large $`r`$, the lowest order term dominates. That is to say, far from the bond the distortions are dipolar:
$$\varphi (𝐫)=\frac{𝐩𝐫}{r^2}$$
(14)
where $`𝐩=A_1\widehat{x}`$ or $`𝐩=A_1\widehat{y}`$. This agrees with the well known results in the literature (see, e.g., Refs. ). Unfortunately this result is inadequate for our purposes, since we also require the spin distortions near the bond. As we show below, a previous attempt fails, and thus we present new arguments leading to a valid solution close to and far away from the frustrating bond. Other work has been unable to solve analytically this part of the frustrating bond problem (although it is clear that they are aware of the issues that we have finally solved).
We have written down a general solution to the static spin texture on the infinite lattice due to a single AFM bond, and that solution consists of a linear combination of an infinite number of possible solution modes. However, since we have shown that the single bond system has only one stable solution, we expect that any prepared state will decay into the state of lowest energy given by the $`m=1`$ solution in Eq. (13). That is every $`A_m0`$ for $`m1`$. Nonetheless, we run into the difficulty that the field equation to which Eq. (13) is a solution is not strictly valid at the bond sites. Consequently, we cannot expect that these solutions will perform well near the bond itself. Indeed, we find that the purely dipolar solution with $`1/r`$ fall-off fits numerical solutions extremely well $`(0.5\%)`$ up to three or four lattice sites away from the bond, but that in the core of the frustration the deviation becomes quite large. At the bond sites themselves, the error is $`25\%`$.
We may ask, of course, why such a description does not suffice if, for the most part, we are interested in the spin distortions away from the bond. Surely we can tolerate a small error at a handful of lattice sites? The answer is that we cannot. Since the spin distortions are most severe in the immediate vicinity of the bond, the spins in the core of the frustration are a large contributor to the magnitude of the total energy stored in the distortions. Thus a proper calculation of the energy in the system requires that we model the core correctly. Equally important is that, for a given solution mode, the local equilibrium condition at the bond site determines the overall magnitude of the spin distortions. That is, it fixes the magnitude of the dipole moment associated with the distortion field. Kovalev and Bogdan, who suggested a continuum approach for the core region of this problem , fall into precisely this trap and hence obtain the wrong magnitude for the long range behaviour for the spin distortions (see below).
It remains to be answered how we might treat the spin distortions around the bond. Ideally, we would like to treat the AFM character of the interaction in the core as a small perturbation on the field equations, but this is not possible since the continuum formalism is badly behaved at the origin under the symmetries we have imposed. A second possibility would be to derive a separate discrete solution valid in the core and to match it smoothly onto the exterior continuum solution. However, we are inclined to avoid such a patch-work approach. Not only is it somewhat inelegant, but it also defeats the purpose of introducing the continuum formalism, namely, to do away with discrete calculations altogether.
Instead, we make use of the fact that the sets $`\{\mathrm{cos}(m\theta )/r^m\}`$ and $`\{\mathrm{sin}(m\theta )/r^m\}`$ are complete (in the sense that any static spin texture satisfying the symmetries specified by a single $`x`$\- or $`y`$-directed bond can be expanded in one of these bases). Thus, to solve for the spin distortions everywhere on the lattice is essentially to fix the values of the coefficients $`\{A_m\}`$. In the following, we attempt to expand the solution to the spin distortions of an $`x`$-directed AFM bond near the origin in the basis $`\{\mathrm{cos}(m\theta )/r^m\}`$.
To start, we expect that the coefficient $`A_1`$ must dominate the others since $`\mathrm{cos}(\theta )/r`$ is the mode of lowest energy. Further, convergence at the bond sites requires that $`(A_m)0`$ faster than $`2^m`$ as $`m\mathrm{}`$. Thus, it is meaningful to treat the expansion $`\varphi ^{(n)}(r,\theta )`$, consisting of the first $`n`$ terms of Eq. (13) as an approximate solution. We can then apply the local equilibrium condition at $`n`$ sites around the bond to determine the $`n`$ coefficients.
For concreteness, consider the four term expansion
$$\varphi (r,\theta )=\varphi ^{(4)}(r,\theta )=A_1\frac{\mathrm{cos}\theta }{r}+A_3\frac{\mathrm{cos}3\theta }{r^3}+A_5\frac{\mathrm{cos}5\theta }{r^5}+A_7\frac{\mathrm{cos}7\theta }{r^7}.$$
(15)
To solve for its four coefficients we require the $`9\times 4`$ transformation matrix
$$T:=\left[\frac{\varphi _i}{A_j}\right]_{9\times 4}=\left[\begin{array}{cccc}2& 8& 32& 128\\ & & & \\ \frac{2}{5}& \frac{88}{125}& \frac{1312}{3125}& \frac{3712}{78125}\\ & & & \\ \frac{6}{13}& \frac{72}{2197}& \frac{19104}{371293}& \frac{569472}{62748517}\\ & & & \\ \frac{2}{3}& \frac{8}{27}& \frac{32}{243}& \frac{128}{2187}\\ & & & \\ \frac{2}{17}& \frac{376}{4913}& \frac{35872}{1419857}& \frac{2566016}{410338673}\\ & & & \\ \frac{6}{25}& \frac{936}{15625}& \frac{7584}{9765625}& \frac{9784704}{6103515625}\\ & & & \\ \frac{10}{41}& \frac{920}{68921}& \frac{335200}{115856201}& \frac{609920}{194754273881}\\ & & & \\ \frac{10}{29}& \frac{520}{24389}& \frac{47200}{20511149}& \frac{14926720}{17249876309}\\ & & & \\ \frac{2}{5}& \frac{8}{125}& \frac{32}{3125}& \frac{128}{78125}\end{array}\right]$$
(16)
(generated using a symbolic algebra computer package) and the matrix
$$M=\left[\begin{array}{ccccccccc}\hfill (32\lambda )& \hfill 2& \hfill 0& \hfill 1& \hfill 0& \hfill 0& \hfill 0& \hfill 0& \hfill 0\\ \hfill 1& \hfill 5& \hfill 1& \hfill 0& \hfill 1& \hfill 0& \hfill 0& \hfill 0& \hfill 0\\ \hfill 0& \hfill 1& \hfill 4& \hfill 1& \hfill 0& \hfill 1& \hfill 0& \hfill 1& \hfill 0\\ \hfill 1& \hfill 0& \hfill 2& \hfill 4& \hfill 0& \hfill 0& \hfill 0& \hfill 0& \hfill 1\end{array}\right]$$
(17)
of linearized equilibrium conditions at sites 1 through 4. Defining the row vector $`𝐚^T=[A_1,A_3,A_5,A_7]`$ the determination of $`\{A_m\}`$ is equivalent to solving the homogeneous system of equations $`MT𝐚=0`$. The requirement that $`det(MT)=0`$ yields a critical value $`\lambda _\text{c}\dot{=}\mathrm{\hspace{0.17em}1.0113}`$ (very close to the true value $`\lambda _\text{c}=1`$) for which
$$𝐚=[0.5987,0.0742,0.1960,0.0552]\times \varphi _1.$$
(18)
That is to say, the best four term expansion reads
$`\varphi ^{(4)}(r,\theta )`$ $`=`$ $`\varphi (r,\theta )=A_1{\displaystyle \frac{\mathrm{cos}\theta }{r}}+A_3{\displaystyle \frac{\mathrm{cos}3\theta }{r^3}}+A_5{\displaystyle \frac{\mathrm{cos}5\theta }{r^5}}+A_7{\displaystyle \frac{\mathrm{cos}7\theta }{r^7}}`$ (19)
$`=`$ $`{\displaystyle \frac{𝐩𝐫}{r^2}}+{\displaystyle \underset{m=3,5,7}{}}A_m{\displaystyle \frac{\mathrm{cos}m\theta }{r^m}}`$ (20)
with
$$p=|𝐩|=A_1=+0.5987\varphi _1,A_3=+0.0742\varphi _1,A_5=+0.1960\varphi _1,A_7=0.0552\varphi _1$$
(21)
The coefficients of $`\varphi ^{(n)}(r,\theta )`$ for $`n`$ being increased from 1 to 4 are presented in Table I.
What this calculation provides that the other does not is the value of the multiplicative factor $`A_1/\varphi _1=A_1/\varphi _10.6`$ relating the magnitude of the spin distortions at the bond sites to the magnitude of the dipole moment associated with the bond itself.
We have shown that the spin distortions are given everywhere by
$$\varphi (r,\theta )=A_1\frac{\mathrm{cos}\theta }{r}+A_3\frac{\mathrm{cos}3\theta }{r^3}+A_5\frac{\mathrm{cos}5\theta }{r^5}+\mathrm{}$$
(22)
As we have seen, however, this expression is unwieldy in that it requires the application of infinitely many local equilibrium conditions to fully determine the coefficients $`\{A_m\}`$. Even to calculate the coefficients of a finite series expansion of several terms is computationally expensive. Ideally, what we would like to have is a solution dependent on a single parameter whose value is determined by applying a single boundary condition at the bond itself. Here we now outline such a method. We find that a simple assumption on the distribution of modes can provide this very result.
We proceed by assuming that the spectrum of modes can be modeled by
$$A_{2n+1}=(1)^n\frac{2A_1}{(2n+1)2^{2n+1}}$$
(23)
or, in somewhat simplified notation,
$$A_k=(\pm )\frac{2A_1}{k2^k}$$
(24)
where the index $`k`$ is taken to be odd and the sign is taken alternately positive and negative. This form falls off just fast enough to make the series converge — besides this seemingly naive reason, we appeal to its success (described in detail below) to justify its usage.
Under this assumption
$`\varphi (r,\theta )`$ $`=`$ $`A_1{\displaystyle \frac{\mathrm{cos}\theta }{r}}+A_3{\displaystyle \frac{\mathrm{cos}3\theta }{r^3}}+A_5{\displaystyle \frac{\mathrm{cos}5\theta }{r^5}}+\mathrm{}`$ (25)
$`=`$ $`2A_1\left({\displaystyle \frac{\mathrm{cos}\theta }{2r}}{\displaystyle \frac{1}{3}}{\displaystyle \frac{\mathrm{cos}3\theta }{(2r)^3}}+{\displaystyle \frac{1}{5}}{\displaystyle \frac{\mathrm{cos}5\theta }{(2r)^5}}\mathrm{}\right).`$ (26)
Now, the magnitude of the dipole moment in terms of the distortion at the bond site follows immediately from solving $`\varphi (1/2,0)=\varphi _1`$ self-consistently. We find that $`A_1=\frac{2}{\pi }\varphi _1`$ and hence
$$\varphi (r,\theta )=\varphi _1\frac{4}{\pi }\underset{k}{}(\pm )\frac{\mathrm{cos}k\theta }{k(2r)^k}$$
(27)
since
$$\varphi _1=\varphi (1/2,0)=\varphi _1\frac{4}{\pi }\left(1\frac{1}{3}+\frac{1}{5}\mathrm{}\right)$$
(28)
which is identically equal to $`\varphi _1`$. The spin distortions at the remaining sites in the immediate neighbourhood of the bond are as follows:
$$\varphi _2=\varphi (\sqrt{5}/2,\mathrm{arctan}(2))=0.2951672353\varphi _1,\varphi _4=\varphi (3/2,0)=0.4096655294\varphi _1$$
(29)
(In fact, one may prove the identity $`2\varphi _2+\varphi _4\varphi _1`$, which we shall use later on in this paper.)
Notice that as $`r`$ becomes large, we get
$`\varphi (r,\theta )`$ $``$ $`\varphi _1{\displaystyle \frac{4}{\pi }}\mathrm{cos}\theta \mathrm{arctan}\left({\displaystyle \frac{1}{2r}}\right)\varphi _1{\displaystyle \frac{4}{\pi }}{\displaystyle \frac{\mathrm{cos}\theta }{2r}}`$ (30)
$`=`$ $`\varphi _1{\displaystyle \frac{2}{\pi }}{\displaystyle \frac{\mathrm{cos}\theta }{r}}`$ (31)
so that the solution retains its familiar long range behaviour. That is
$$\varphi (𝐫)=\frac{𝐩𝐫}{r^2}$$
(32)
but now with
$$p=|𝐩|=\frac{2}{\pi }\varphi _1.$$
(33)
What remains is to determine the parameter $`\varphi _1`$. As promised, the bond furnishes a single boundary condition in the form of the equilibrium condition applied at either of the bond sites:
$$\lambda \mathrm{sin}(2\varphi _1)+2\mathrm{sin}(\varphi _1\varphi _2)+\mathrm{sin}(\varphi _1\varphi _4)=0.$$
(34)
This is an implicit equation for $`\varphi _1`$, and thus for all of the spin distortions as a function of $`\lambda `$. Its solution is plotted in Fig. 2.
Moreover, the linearized equation gives
$$\lambda _\text{c}2\varphi _1+2(\varphi _1\varphi _2)+(\varphi _1\varphi _4)=0$$
(35)
which can be solved explicitly:
$$\lambda _\text{c}=\frac{3}{2}\frac{2\varphi _2+\varphi _4}{2\varphi _1}1.$$
(36)
That is, our ansatz correctly reproduces the exact critical value of $`\lambda _c`$ !
A comparison of the solution Eq. (27) to numerical simulations is presented in Fig. 3; clearly, the agreement is excellent, providing the most direct support for our ansatz.
### C Energy functional
We now calculate the total energy stored in the spin distortions induced by a single AFM bond. That these distortions are both small and slowly varying in position away from the core allows us to convert the sum of the energy contributions into an integral of the energy density $`(\varphi )^2`$. A complete derivation is provided in Appendix A, and we summarize the results below.
To begin, the Hamiltonian is approximated to second order everywhere except across the AFM bond itself:
$$H\underset{ij}{}J_{ij}+\frac{1}{2}J\underset{ij}{}^{}(\varphi _i\varphi _j)^22\lambda J\mathrm{sin}^2\varphi _1$$
(37)
Of course, the term $`J_{ij}`$ represents the total energy of the system in the absence of spin distortions so that the energy from the distortions alone is given by the latter two terms. They, in turn, can be expanded using the continuum approximation
$`E_{\text{dist}}`$ $`=`$ $`{\displaystyle \frac{1}{4}}J{\displaystyle \underset{i0,1}{}}{\displaystyle \underset{\widehat{a}}{}}(\varphi _i\varphi _{i+\widehat{a}})^2+{\displaystyle \frac{1}{2}}J\left(2(\varphi _1\varphi _4)^2+4(\varphi _1\varphi _2)^2\right)2\lambda J\mathrm{sin}^2\varphi _1`$ (38)
$``$ $`J\left\{{\displaystyle \frac{1}{2}}{\displaystyle _M}(\varphi )^2d^2r+(\varphi _1\varphi _4)^2+2(\varphi _1\varphi _2)^22\lambda \mathrm{sin}^2\varphi _1\right\}`$ (39)
where $`M`$ is the $`xy`$-plane excluding a small region about the bond centre.
An explicit evaluation of the energy using the solution of the previous subsection is presented in Appendix A, wherein the full effect of the core region is accounted for. We find
$$E_{\text{dist}}=2J\left(\varphi _1^2\lambda \mathrm{sin}^2\varphi _1\right).$$
(40)
This result is compared to numerical simulations in Fig. 4, and again, excellent agreement between our numerical solutions and our analytical work is found.
So, now we carry on to the examination of the many-bond problem, having an excellent solution to both the core and far-field distortion patterns of the single-bond problem, as well as an accurate energy functional for an isolated frustrating bond.
## III Interacting Frustrating Bonds:
After dealing with a single bond in isolation, the next step toward treating a non-zero density of bonds is to determine how bonds interact with one another. In this subsection we consider the problem of two bonds.
Suppose that there is an AFM bond, call it $`A`$, between the 0 and 1 sites. Then suppose that another similarly directed bond, call it $`B`$, is placed between the $`s`$ and $`s+1`$ sites and that a vector $`𝐑`$ making an angle $`\mathrm{\Phi }`$ with the $`x`$-axis connects the two bond centres, as in Fig. 5.
We expect that the energy can be parametrized by two variables
$$\alpha =\frac{\varphi _1\varphi _0}{2}\beta =\frac{\varphi _{s+1}\varphi _s}{2}$$
(41)
and that it can be written in the form
$`E_2(\alpha ,\beta )`$ $`=`$ $`E(\alpha )+E(\beta )+Jg\alpha \beta `$ (42)
$`=`$ $`2J(1\lambda )(\alpha ^2+\beta ^2)+\left({\displaystyle \frac{2}{3}}\delta \right)J\lambda (\alpha ^4+\beta ^4)+Jg\alpha \beta `$ (43)
where $`g=g(𝐑)`$ is a real-valued function of the separation and relative orientation of the bonds and $`\delta `$ is a small correction representing the $`4^{\text{th}}`$ order contribution to energy from the spin distortions which were neglected in Eq. (39).
The requirement that the spin texture remain invariant up to a global sign change under interchange of $`\alpha `$ and $`\beta `$ reduces the two parameter energy expression (43) to one of two one parameter expressions $`E_2^+`$ or $`E_2^{}`$ corresponding to the symmetric ($`\alpha =\beta `$) and the antisymmetric ($`\alpha =\beta `$) state.
*Case 1*: $`\varphi =\alpha =\beta `$. In this case, the dipoles associated with the bond are anti-aligned. The total energy is
$$E_2^{}(\varphi )=E_2(\varphi ,\varphi )=4J\left((1\lambda \frac{1}{4}g)\varphi ^2+\frac{1}{2}\left(\frac{2}{3}\delta \right)\lambda \varphi ^4\right)<2E(\varphi ).$$
(44)
Minimization with respect to $`\varphi `$ gives
$$\varphi =\sqrt{(\lambda +\frac{1}{4}g1)/\left(\frac{2}{3}\delta \right)\lambda }.$$
(45)
This implies that the critical value of $`\lambda `$ at which the canted ground state first appears is lower than it is for a single AFM bond $`\lambda _c=1`$.
*Case 2*: $`\varphi =\alpha =\beta `$. This represents a higher energy metastable state characterized by aligned dipoles. The energy for this configuration is
$$E_2^+(\varphi )=E_2(\varphi ,\varphi )=4J\left((1\lambda +\frac{1}{4}g)\varphi ^2+\frac{1}{2}\left(\frac{2}{3}\delta \right)\lambda \varphi ^4\right)>2E(\varphi )$$
(46)
with distortion magnitude
$$\varphi =\sqrt{(\lambda \frac{1}{4}g1)/\left(\frac{2}{3}\delta \right)\lambda }.$$
(47)
Now the critical value of the coupling constant required for a distorted ground state exceeds $`\lambda _c=1`$.
Case 1 is of particular interest since it yields the ground state energy $`E_2^0`$ of the two bond system. Further, by re-expressing that energy in terms of the energy of a single bond (extracted from Eq. (40) in the limit $`\lambda \lambda _c+0^+`$) we can determine the energy of interaction between the two bonds.
$$E_2^0=4J\frac{(\lambda +\frac{1}{4}g1)^2}{\lambda (2/3\delta )}=2E^0J\frac{(\lambda 1)}{\lambda (2/3\delta )}g(𝐑)J\frac{1}{8\lambda (2/3\delta )}g^2(𝐑)$$
(48)
In general, since we expect $`g`$ to be small, we can write
$$E_{\text{int}}=J\frac{(\lambda 1)}{\lambda (2/3\delta )}g(𝐑).$$
(49)
However, as $`\lambda 1`$, we obtain
$$E_{\text{int}}J\frac{1}{8\lambda (2/3\delta )}g^2(𝐑),$$
(50)
a weak, long-range interaction with a higher power law, a consequence of the fact that the dipolar distortions do not pre-exist in the unperturbed medium at $`\lambda =1`$. Such an interaction is analogous to the Van der Waals interaction between thermally fluctuating dipoles.
The function $`g`$ expresses the functional dependence of the interaction energy on the geometrical configuration of the bonds. We have yet to determine its exact form. All we can say now is that
$$\underset{R\mathrm{}}{lim}g(𝐑)=0$$
(51)
which simply formalizes our expectation that two bonds must be non-interacting at infinite separation.
The long range spin distortions arising from each of the bonds $`A`$ and $`B`$ with dipole moments $`𝐩^A`$ and $`𝐩^B`$ is given by
$$\varphi ^A(𝐫)=\frac{𝐩^A𝐫}{r^2}\text{and}\varphi ^B(𝐫)=\frac{𝐩^B𝐫}{r^2}.$$
(52)
The linearity of the Laplacian implies that the field equations admit a superposition principle. Therefore, we take the total spin distortion at each point to be
$$\varphi =\varphi ^A+\varphi ^B$$
(53)
where $`\varphi ^A`$ and $`\varphi ^B`$ are the spin distortions from each bond in the absence of the other.
The energy in the distortions away from the cores goes as
$`{\displaystyle _{M2}}(\varphi )^2d^2r`$ $`=`$ $`{\displaystyle _{M2}}((\varphi ^A+\varphi ^B)^2)d^2r`$ (54)
$`=`$ $`{\displaystyle _{M2}}(\varphi ^A)^2d^2r+{\displaystyle _{M2}}(\varphi ^B)^2d^2r`$ (56)
$`+{\displaystyle _{M2}}\varphi ^A\varphi ^Bd^2r`$
where $`M2\left(=\backslash D_ϵ(\mathrm{𝟎})\backslash D_ϵ(𝐑)\right)`$ is the $`xy`$-plane with disks removed about the bond centres. The last term in this expression vanishes identically, which indicates that the core of each bond interacts with the long-range spin distortions of the other and justifies a rather involved calculation of the interaction energy which we have relegated to Appendix B.
What we find is that $`E_{\text{int}}`$ has the form of a magnetic dipole interaction. Further, although in the preceding discussion we considered only parallel bonds, it is simple to show that these results hold more generally. Thus, for two bonds which are parallel or perpendicular we have
$$E_{\text{int}}=J\frac{2\pi }{R^2}\left\{2(𝐩^A\widehat{𝐑})(𝐩^B\widehat{𝐑})𝐩^A𝐩^B\right\}.$$
(57)
Finally, we can work backwards to find $`g(𝐑)`$. For parallel bonds,
$$E_{\text{int}}=J\frac{2\pi }{R^2}\left\{2(𝐩^A\widehat{𝐑})(𝐩^B\widehat{𝐑})𝐩^A𝐩^B\right\}=\pm J\frac{(\lambda 1)}{\lambda (2/3\delta )}\frac{8}{\pi R^2}\mathrm{cos}2\mathrm{\Phi }.$$
(58)
That is
$$g(𝐑)=\frac{8}{\pi R^2}\mathrm{cos}2\mathrm{\Phi }.$$
(59)
We note that the identical calculation for two perpendicular bonds gives
$$g(𝐑)=\frac{8}{\pi R^2}\mathrm{sin}2\mathrm{\Phi }.$$
(60)
## IV Clusters in the cluster spin glass phase:
We should ask whether the results we have obtained so far for the one and two bond problems can be generalized to allow us to tackle the problem of a lattice frustrated by the presence of any number of arbitrarily placed AFM bonds. It should be clear that, in general, even for relatively few bonds, the induced spin distortions will be very complicated and the energy surface characterized by many closely spaced, low-lying states. In such a case one must resort to sophisticated computer algorithms to numerically generate the ground state spin texture, and the qualitative analysis of the cluster spin-glass phase from such work has been analyzed elsewhere .
In contrast to that work, here we note that there are configurations of suitably high symmetry for which we can confidently treat the spins as planar and even solve analytically for the spin distortions and energies of all the possible states. The smallest such configuration is the square cluster of parallel bonds.
By a cluster we imply a collection of bonds arranged in some local region on the lattice. That collection, call it $`C`$, can be thought of as a set of dipole–position pairs: $`\{(𝐩^\alpha ,𝐫^\alpha )\}_{\alpha C}.`$ Given a high symmetry cluster, for which a spin-planar ground state is justified, the spin distortions away from the cores of the bonds are given by
$$\varphi (𝐫)=\underset{\alpha C}{}\frac{𝐩^\alpha (𝐫𝐫^\alpha )}{|𝐫𝐫^\alpha |^2}$$
(61)
which is the solution (unique for the required symmetry) to the equation
$$^2\varphi (𝐫)=2\pi \underset{\alpha C}{}𝐩^\alpha \delta (𝐫𝐫^\alpha ).$$
(62)
The total interaction energy of the cluster can be written as the sum of all pairwise interactions
$$E_{\text{int}}=\underset{\alpha <\beta C}{}J\frac{2\pi }{|𝐫^\alpha 𝐫^\beta |^2}\left(2\frac{𝐩^\alpha (𝐫^\alpha 𝐫^\beta )𝐩^\beta (𝐫^\alpha 𝐫^\beta )}{|𝐫^\alpha 𝐫^\beta |^2}𝐩^\alpha 𝐩^\beta \right).$$
(63)
Thus, for instance, the $`L\times L`$ square cluster of parallel bonds given by
$$𝐫^1=(0,0),𝐫^2=(L,0),𝐫^3=(0,L),𝐫^4=(L,L),𝐩^\alpha =\pm p\widehat{x}\mathrm{for}\alpha C=\{1,2,3,4\},$$
(64)
has a rather simple interaction energy. There are four possibilities depending on the the orientation of each dipole:
$$E_{\text{int}}=J\frac{8\pi }{L^2}p^2,\mathrm{\hspace{0.17em}0},\mathrm{\hspace{0.17em}0},+J\frac{8\pi }{L^2}p^2$$
(65)
In practice, however, the degeneracy of the middle two states is lifted by higher order terms in the interaction energy. Indeed, the splitting observed in numerical simulations enables us to list the four distinguished states in order of decreasing energy.
$$\left|\begin{array}{cc}& \\ & \end{array}\right|,\left|\begin{array}{cc}& \\ & \end{array}\right|,\left|\begin{array}{cc}& \\ & \end{array}\right|=\text{first excited state},\left|\begin{array}{cc}& \\ & \end{array}\right|=\text{ground state}$$
(66)
The first excited state consists of four similarly directed dipoles. Destructive interference inside the cluster gives zero net distortion, but outside the cluster the dipoles add constructively so that the cluster acts like a single unit with a much stronger moment. This results in very strong long range spin distortions. Far enough from the bond, the spin distortions are given by
$$\varphi (𝐫)=\frac{4𝐩𝐫}{r^2}.$$
(67)
Since the long range spin distortions mediate the interaction between bonds, we expect a cluster of this kind to strongly couple to other bonds in the lattice.
In the ground state, for which the spin distortion pattern is shown in Fig. 8, we have the opposite case: internally, the dipoles add constructively to give large distortions whereas outside they cancel to give very small ones. The absence of long range spin distortions implies that these clusters can only weakly interact with other bonds. Most interesting, though, is that the internal spins are uniformly oriented but differently ordered from those spins outside the cluster.
Numerical solutions of the energies of this square cluster are shown in Fig. 9. The solid curves are our analytical results (apart from a constant times the single-bond energy (that is straightforward to calculate)), and provides strong support for their usage. In the ground state this figure makes clear that the binding energy of the cluster can be quite large, especially for small cluster sizes. Further, square clusters tend to settle into states with strong internal binding and which interact only weakly with other bonds. That is to say, a square cluster is a locally ordered domain whose local order parameter $`\widehat{𝛀}`$ is non-collinear with spins in the rest of the lattice. However, this behaviour is a strong function of cluster size. This construction, and the energy plot of Fig. 9, demonstrates that such clusterings of spins (in regions that are not too large) are stable.
The relation of such clusters to stripes, or in the case of spin modulations, to the physics associated with the appearance of domain walls, can be demonstrated by considering the interface between such clusters, and to this end we have analyzed a six-bond cluster given by
$`𝐫^1=(0,0),𝐫^2`$ $`=`$ $`(L,0),𝐫^3=(2L,0),𝐫^4=(0,L),𝐫^5=(L,L),𝐫^6=(2L,L),`$ (69)
$`𝐩^\alpha =\pm p\widehat{x}\mathrm{for}\alpha C=\{1,2,3,4,5,6\}.`$
Following the above analysis for four bonds, for the six-bond situation there are 64 possible choices of the dipole moments’ orientations, and these states have 15 different energies. The lowest energy configuration corresponds to
$$\left|\begin{array}{ccc}& & \\ & & \end{array}\right|$$
(70)
and has a dipole-pair interaction energy of $`(103\pi J/10)(p/L)^2`$ (which is noticeably lower than the first excited state, which has an energy of $`(38\pi J/10)(p/L)^2`$).
The ground state spin texture for this location of the six frustrating bonds is shown in Fig. 10. From this figure one can see the important result that this arrangement of spins is exactly what would expect if each $`4\times 4`$ cluster within the 6-bond cluster was in its respective ground state. Thus, between these $`4\times 4`$ clusters one obtains a domain wall, of width one lattice spacing, over which the local magnetic order parameter is rotated. Numerical evidence suggestive of this type of domain wall was discussed at length in Ref. for the case of a random distribution (and orientation) of a non-zero density of frustrating spin interactions, and it is clear that the same physics is at work in these two situations: spin twists.
## V Conclusions:
The above formalism has provided a detailed analytical theory of the spin distortions generated by frustrating bonds, and of the interactions mediated by the spin background between them. We have critiqued its validity by comparing to numerical solutions, and have found excellent agreement. Then, by focusing on highly symmetrical distributions of frustrating bonds, reminiscent of local regions of bonds in the multiply doped state, we have used this theory to verify the existence of locally ordered magnetic clusters. Most importantly, our work shows that these clusters are stable. These are the clusters envisioned to exist in the so-called cluster spin-glass phase of LSCO and $`\mathrm{Y}_{1\mathrm{x}}\mathrm{Ca}_\mathrm{x}\mathrm{Ba}_2\mathrm{Cu}_3)_6`$ .
It is to be stressed that it is believed that mobile holes produce the same kinds of spin distortions that are produced by the frustrating bonds discussed in this paper . Thus, we believe that our results support previous suggestions that spin twists and distortions are part of the competing interactions that might lead to rivers of charge appearing as low-energy fluctuations in the doped cuprate systems.
## ACKNOWLEDGMENTS
We wish to thank Marc-Henri Julien, John Tranquada, Kazu Yamada, Noha Salem, and especially Bob Birgeneau and David Johnston, for helpful comments. Also, we thank Frank Marsiglio for a critical reading of the manuscript. The first draft of this paper was written while one of us (RJG) was visiting ICTP, Trieste, and he wishes to thank them for their hospitality and support. This work was supported in part by the NSERC of Canada.
## A Single Bond Energy
The question of the energy stored in the spin distortions can be answered by expanding the Hamiltonian as follows:
$`H`$ $`=`$ $`J{\displaystyle \underset{ij}{}^{}}\mathrm{cos}(\varphi _i\varphi _j)+\lambda J\mathrm{cos}(\varphi _1\varphi _0)`$ (A1)
$``$ $`J{\displaystyle \underset{ij}{}^{}}\left(1{\displaystyle \frac{1}{2}}(\varphi _i\varphi _j)^2\right)+\lambda J\mathrm{cos}(2\varphi _1)`$ (A2)
$`=`$ $`{\displaystyle \underset{ij}{}}J_{ij}+{\displaystyle \frac{1}{2}}J{\displaystyle \underset{ij}{}^{}}(\varphi _i\varphi _j)^22\lambda J\mathrm{sin}^2\varphi _1`$ (A3)
Of course, the term $`J_{ij}`$ represents the energy intrinsic to the lattice. Therefore, the energy from the spin distortions alone is given by
$$\frac{1}{2}J\underset{ij}{}^{}(\varphi _i\varphi _j)^22\lambda J\mathrm{sin}^2\varphi _1.$$
(A4)
This in turn can be expanded using the continuum approximation
$`E_{\text{dist}}`$ $``$ $`J\left\{{\displaystyle \frac{1}{2}}{\displaystyle _M}(\varphi )^2d^2r+(\varphi _1\varphi _4)^2+2(\varphi _1\varphi _2)^22\lambda \mathrm{sin}^2\varphi _1\right\}`$ (A5)
$`=`$ $`J\left\{{\displaystyle \frac{1}{2}}{\displaystyle _M}\varphi \varphi \mathrm{𝐝𝐬}+(\varphi _1\varphi _4)^2+2(\varphi _1\varphi _2)^22\lambda \mathrm{sin}^2\varphi _1\right\}`$ (A6)
where $`M=\backslash D_ϵ(\mathrm{𝟎})`$ is the $`xy`$-plane excluding a small region about the bond centre. The integral term of this expression must be evaluated. Since the core of the bond occupies a unit disk at the origin, the appropriate value for $`ϵ`$ is 1. Thus, one finds
$$E_{\text{dist}}=J\left\{\frac{\pi }{2}\underset{k}{}kA_k^2+(\varphi _1\varphi _4)^2+2(\varphi _1\varphi _2)^22\lambda \mathrm{sin}^2\varphi _1\right\}.$$
(A7)
Using the results for the mode characterization of Eq. (23) gives
$$\underset{k}{}kA_k^2=\underset{k}{}k\left(\frac{2A_1}{k2^k}\right)^2=\frac{8}{\pi ^2}\mathrm{ln}\left(\frac{5}{3}\right)\varphi _1^2$$
(A8)
and the identity $`2\varphi _2+\varphi _4=\varphi _1`$ allows us to write
$$(\varphi _1\varphi _4)^2+2(\varphi _1\varphi _2)^2=\varphi _1^2+2\varphi _2^2+\varphi _4^2.$$
(A9)
Thus the energy in the spin distortions is
$$E_{\text{dist}}=J\left\{\left(\frac{4}{\pi }\mathrm{ln}\left(\frac{5}{3}\right)+1\right)\varphi _1^2+2\varphi _2^2+\varphi _4^22\lambda \mathrm{sin}^2\varphi _1\right\}$$
(A10)
which can be evaluated using Eq. (29) to give
$$E_{\text{dist}}=2J\left(\varphi _1^2\lambda \mathrm{sin}^2\varphi _1\right).$$
(A11)
We stress that this expression includes the energy from the core of the spin distortion field.
## B Interaction Energy
In accordance with Fig. 5, we write the spin distortion at the 0 site from the bond B as
$$\varphi _0^B=\frac{𝐩^B(𝐑+\frac{1}{2}\widehat{x})}{|𝐑+\frac{1}{2}\widehat{x}|^2}𝐩^B(𝐑+\frac{1}{2}\widehat{x})\frac{1}{R^2}\left[1\frac{𝐑\widehat{x}}{R^2}\frac{1}{4R^2}\right].$$
(B1)
A similar expression can be written down for the other bond site. Then, by superposition, the difference between the net distortions at sites 0 and 1 is
$$(\varphi _0\varphi _1)=(\varphi _0^A\varphi _1^A)\frac{2}{R^4}(𝐩^B𝐑)(𝐩^B\widehat{x})+\frac{1}{R^2}𝐩^B\widehat{x}+𝒪\left(\frac{1}{R^2}\right)$$
(B2)
Then we make use of the relationship between the spin distortions at the bond sites and the magnitude of the dipole moment:
$$(\varphi _0^A+\varphi _1^A)2\frac{\pi }{2}p^A.$$
(B3)
Therefore
$$(\varphi _0\varphi _1)^2=(\varphi _0^A+\varphi _1^A)^2\frac{8\alpha }{R^4}(𝐩^A𝐑)(𝐩^B𝐑)+\frac{4\alpha }{R^2}𝐩^A𝐩^B$$
(B4)
so that
$$\frac{1}{2}E=E(𝐩^A)+\frac{1}{2}J\left\{\frac{2\pi }{R^2}𝐩^A𝐩^B\frac{4\pi }{R^4}(𝐩^A𝐑)(𝐩^B𝐑)\right\}$$
(B5)
and thus, by symmetry, the total energy is
$$E=E(𝐩^A)+E(𝐩^B)+J\{\frac{2\pi }{R^2}𝐩^A𝐩^B\frac{4\pi }{R^4}(𝐩^A𝐑)(𝐩^B𝐑)\}.$$
(B6)
We conclude that the interaction energy is given by
$$E_{\text{int}}(𝐩^A,𝐩^B)=J\frac{2\pi }{R^2}\left\{\frac{2}{R^2}(𝐩^A𝐑)(𝐩^B𝐑)𝐩^A𝐩^B\right\}.$$
(B7)
(Repeating the above calculation for two perpendicular bonds produces the identical result.) |
warning/0001/hep-ph0001334.html | ar5iv | text | # INTRODUCTION TO B PHYSICS
## 1 Introduction
The rich phenomenology of weak decays has always been a source of information about the nature of elementary particle interactions. A long time ago, $`\beta `$\- and $`\mu `$-decay experiments revealed the structure of the effective flavor-changing interactions at low momentum transfer. Today, weak decays of hadrons containing heavy quarks are employed for tests of the Standard Model and measurements of its parameters. In particular, they offer the most direct way to determine the weak mixing angles, to test the unitarity of the Cabibbo-Kobayashi-Maskawa (CKM) matrix, and to explore the physics of CP violation. Hopefully, this will provide some hints about New Physics beyond the Standard Model. On the other hand, hadronic weak decays also serve as a probe of that part of strong-interaction phenomenology which is least understood: the confinement of quarks and gluons inside hadrons.
The structure of weak interactions in the Standard Model is rather simple. Flavor-changing decays are mediated by the coupling of the charged current $`J_{\mathrm{CC}}^\mu `$ to the $`W`$-boson field:
$$_{\mathrm{CC}}=\frac{g}{\sqrt{2}}J_{\mathrm{CC}}^\mu W_\mu ^{}+\text{h.c.,}$$
(1)
where
$$J_{\mathrm{CC}}^\mu =(\overline{\nu }_e,\overline{\nu }_\mu ,\overline{\nu }_\tau )\gamma ^\mu \left(\begin{array}{c}e_\mathrm{L}\\ \mu _\mathrm{L}\\ \tau _\mathrm{L}\end{array}\right)+(\overline{u}_\mathrm{L},\overline{c}_\mathrm{L},\overline{t}_\mathrm{L})\gamma ^\mu V_{\mathrm{CKM}}\left(\begin{array}{c}d_\mathrm{L}\\ s_\mathrm{L}\\ b_\mathrm{L}\end{array}\right)$$
(2)
contains the left-handed lepton and quark fields, and
$$V_{\mathrm{CKM}}=\left(\begin{array}{ccc}V_{ud}& V_{us}& V_{ub}\\ V_{cd}& V_{cs}& V_{cb}\\ V_{td}& V_{ts}& V_{tb}\end{array}\right)$$
(3)
is the CKM matrix. At low energies, the charged-current interaction gives rise to local four-fermion couplings of the form
$$_{\mathrm{eff}}=2\sqrt{2}G_FJ_{\mathrm{CC}}^\mu J_{\mathrm{CC},\mu }^{},$$
(4)
where
$$G_F=\frac{g^2}{4\sqrt{2}M_W^2}=1.16639(2)\text{GeV}^2$$
(5)
is the Fermi constant.
According to the structure of the charged-current interaction, weak decays of hadrons can be divided into three classes: leptonic decays, in which the quarks of the decaying hadron annihilate each other and only leptons appear in the final state; semi-leptonic decays, in which both leptons and hadrons appear in the final state; and non-leptonic decays, in which the final state consists of hadrons only. Representative examples of these three types of decays are shown in Fig. 1. The simple quark-line graphs shown in this figure are a gross oversimplification, however. In the real world, quarks are confined inside hadrons, bound by the exchange of soft gluons. The simplicity of the weak interactions is overshadowed by the complexity of the strong interactions. A complicated interplay between the weak and strong forces characterizes the phenomenology of hadronic weak decays. As an example, a more realistic picture of a non-leptonic decay is shown in Fig. 2.
The complexity of strong-interaction effects increases with the number of quarks appearing in the final state. Bound-state effects in leptonic decays can be lumped into a single parameter (a “decay constant”), while those in semi-leptonic decays are described by invariant form factors depending on the momentum transfer $`q^2`$ between the hadrons. Approximate symmetries of the strong interactions help us to constrain the properties of these form factors. Non-leptonic weak decays, on the other hand, are much more complicated to deal with theoretically. Only very recently reliable tools have been developed that allow us to control the complex QCD dynamics in many two-body $`B`$ decays using a heavy-quark expansion.
Over the last decade, a lot of information on heavy-quark decays has been collected in experiments at $`e^+e^{}`$ storage rings operating at the $`\mathrm{{\rm Y}}(4s)`$ resonance, and more recently at high-energy $`e^+e^{}`$ and hadron colliders. This has led to a rather detailed knowledge of the flavor sector of the Standard Model and many of the parameters associated with it. In the years ahead the $`B`$ factories at SLAC, KEK, Cornell, and DESY will continue to provide a wealth of new results, focusing primarily on studies of CP violation and rare decays.
The experimental progress in heavy-flavor physics has been accompanied by a significant progress in theory, which was related to the discovery of heavy-quark symmetry, the development of the heavy-quark effective theory, and more generally the establishment of various kinds of heavy-quark expansions. The excitement about these developments rests upon the fact that they allow model-independent predictions in an area in which “progress” in theory often meant nothing more than the construction of a new model, which could be used to estimate some strong-interaction hadronic matrix elements. In Sec. 2, we review the physical picture behind heavy-quark symmetry and discuss the construction, as well as simple applications, of the heavy-quark effective theory. Section 3 deals with applications of these concepts to exclusive weak decays of $`B`$ mesons. Applications of the heavy-quark expansion to inclusive $`B`$ decays are reviewed in Sec. 4. We then focus on the exciting field of rare hadronic $`B`$ decays, concentrating on the example of the decays $`B\pi K`$. In Sec. 5, we discuss the theoretical description of these decays and explain various strategies for constraining and determining the weak, CP-violating phase $`\gamma =\text{arg}(V_{ub}^{})`$ of the CKM matrix. In Sec. 6, we discuss how rare decays can be used to search for New Physics beyond the Standard Model.
## 2 Heavy-Quark Symmetry
This section provides an introduction to the ideas of heavy-quark symmetry $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$ and the heavy-quark effective theory $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$, which provide the modern theoretical framework for the description of the properties and decays of hadrons containing a heavy quark. For a more detailed description of this subject, the reader is referred to the review articles in Refs. 18–24.
### 2.1 The Physical Picture
There are several reasons why the strong interactions of hadrons containing heavy quarks are easier to understand than those of hadrons containing only light quarks. The first is asymptotic freedom, the fact that the effective coupling constant of QCD becomes weak in processes with a large momentum transfer, corresponding to interactions at short distance scales $`^{\mathrm{?},\mathrm{?}}`$. At large distances, on the other hand, the coupling becomes strong, leading to non-perturbative phenomena such as the confinement of quarks and gluons on a length scale $`R_{\mathrm{had}}1/\mathrm{\Lambda }_{\mathrm{QCD}}1`$ fm, which determines the size of hadrons. Roughly speaking, $`\mathrm{\Lambda }_{\mathrm{QCD}}0.2`$ GeV is the energy scale that separates the regions of large and small coupling constant. When the mass of a quark $`Q`$ is much larger than this scale, $`m_Q\mathrm{\Lambda }_{\mathrm{QCD}}`$, it is called a heavy quark. The quarks of the Standard Model fall naturally into two classes: up, down and strange are light quarks, whereas charm, bottom and top are heavy quarks.<sup>a</sup><sup>a</sup>aIronically, the top quark is of no relevance to our discussion here, since it is too heavy to form hadronic bound states before it decays. For heavy quarks, the effective coupling constant $`\alpha _s(m_Q)`$ is small, implying that on length scales comparable to the Compton wavelength $`\lambda _Q1/m_Q`$ the strong interactions are perturbative and much like the electromagnetic interactions. In fact, the quarkonium systems $`(\overline{Q}Q)`$, whose size is of order $`\lambda _Q/\alpha _s(m_Q)R_{\mathrm{had}}`$, are very much hydrogen-like.
Systems composed of a heavy quark and other light constituents are more complicated. The size of such systems is determined by $`R_{\mathrm{had}}`$, and the typical momenta exchanged between the heavy and light constituents are of order $`\mathrm{\Lambda }_{\mathrm{QCD}}`$. The heavy quark is surrounded by a complicated, strongly interacting cloud of light quarks, antiquarks and gluons. In this case it is the fact that $`\lambda _QR_{\mathrm{had}}`$, i.e. that the Compton wavelength of the heavy quark is much smaller than the size of the hadron, which leads to simplifications. To resolve the quantum numbers of the heavy quark would require a hard probe; the soft gluons exchanged between the heavy quark and the light constituents can only resolve distances much larger than $`\lambda _Q`$. Therefore, the light degrees of freedom are blind to the flavor (mass) and spin orientation of the heavy quark. They experience only its color field, which extends over large distances because of confinement. In the rest frame of the heavy quark, it is in fact only the electric color field that is important; relativistic effects such as color magnetism vanish as $`m_Q\mathrm{}`$. Since the heavy-quark spin participates in interactions only through such relativistic effects, it decouples.
It follows that, in the limit $`m_Q\mathrm{}`$, hadronic systems which differ only in the flavor or spin quantum numbers of the heavy quark have the same configuration of their light degrees of freedom $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. Although this observation still does not allow us to calculate what this configuration is, it provides relations between the properties of such particles as the heavy mesons $`B`$, $`D`$, $`B^{}`$ and $`D^{}`$, or the heavy baryons $`\mathrm{\Lambda }_b`$ and $`\mathrm{\Lambda }_c`$ (to the extent that corrections to the infinite quark-mass limit are small in these systems). These relations result from some approximate symmetries of the effective strong interactions of heavy quarks at low energies. The configuration of light degrees of freedom in a hadron containing a single heavy quark with velocity $`v`$ does not change if this quark is replaced by another heavy quark with different flavor or spin, but with the same velocity. Both heavy quarks lead to the same static color field. For $`N_h`$ heavy-quark flavors, there is thus an SU$`(2N_h)`$ spin-flavor symmetry group, under which the effective strong interactions are invariant. These symmetries are in close correspondence to familiar properties of atoms. The flavor symmetry is analogous to the fact that different isotopes have the same chemistry, since to good approximation the wave function of the electrons is independent of the mass of the nucleus. The electrons only see the total nuclear charge. The spin symmetry is analogous to the fact that the hyperfine levels in atoms are nearly degenerate. The nuclear spin decouples in the limit $`m_e/m_N0`$.
Heavy-quark symmetry is an approximate symmetry, and corrections arise since the quark masses are not infinite. In many respects, it is complementary to chiral symmetry, which arises in the opposite limit of small quark masses. There is an important distinction, however. Whereas chiral symmetry is a symmetry of the QCD Lagrangian in the limit of vanishing quark masses, heavy-quark symmetry is not a symmetry of the Lagrangian (not even an approximate one), but rather a symmetry of an effective theory that is a good approximation to QCD in a certain kinematic region. It is realized only in systems in which a heavy quark interacts predominantly by the exchange of soft gluons. In such systems the heavy quark is almost on-shell; its momentum fluctuates around the mass shell by an amount of order $`\mathrm{\Lambda }_{\mathrm{QCD}}`$. The corresponding fluctuations in the velocity of the heavy quark vanish as $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q0`$. The velocity becomes a conserved quantity and is no longer a dynamical degree of freedom $`^\mathrm{?}`$. Nevertheless, results derived on the basis of heavy-quark symmetry are model-independent consequences of QCD in a well-defined limit. The symmetry-breaking corrections can be studied in a systematic way. To this end, it is however necessary to cast the QCD Lagrangian for a heavy quark,
$$_Q=\overline{Q}(i\text{ /}Dm_Q)Q,$$
(6)
into a form suitable for taking the limit $`m_Q\mathrm{}`$.
### 2.2 Heavy-Quark Effective Theory
The effects of a very heavy particle often become irrelevant at low energies. It is then useful to construct a low-energy effective theory, in which this heavy particle no longer appears. Eventually, this effective theory will be easier to deal with than the full theory. A familiar example is Fermi’s theory of the weak interactions. For the description of the weak decays of hadrons, the weak interactions can be approximated by point-like four-fermion couplings, governed by a dimensionful coupling constant $`G_F`$ \[cf. (4)\]. The effects of the intermediate $`W`$ bosons can only be resolved at energies much larger than the hadron masses.
The process of removing the degrees of freedom of a heavy particle involves the following steps $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$: one first identifies the heavy-particle fields and “integrates them out” in the generating functional of the Green functions of the theory. This is possible since at low energies the heavy particle does not appear as an external state. However, whereas the action of the full theory is usually a local one, what results after this first step is a non-local effective action. The non-locality is related to the fact that in the full theory the heavy particle with mass $`M`$ can appear in virtual processes and propagate over a short but finite distance $`\mathrm{\Delta }x1/M`$. Thus, a second step is required to obtain a local effective Lagrangian: the non-local effective action is rewritten as an infinite series of local terms in an Operator Product Expansion (OPE) $`^{\mathrm{?},\mathrm{?}}`$. Roughly speaking, this corresponds to an expansion in powers of $`1/M`$. It is in this step that the short- and long-distance physics is disentangled. The long-distance physics corresponds to interactions at low energies and is the same in the full and the effective theory. But short-distance effects arising from quantum corrections involving large virtual momenta (of order $`M`$) are not described correctly in the effective theory once the heavy particle has been integrated out. In a third step, they have to be added in a perturbative way using renormalization-group techniques. These short-distance effects lead to a renormalization of the coefficients of the local operators in the effective Lagrangian. An example is the effective Lagrangian for non-leptonic weak decays, in which radiative corrections from hard gluons with virtual momenta in the range between $`m_W`$ and some low renormalization scale $`\mu `$ give rise to Wilson coefficients, which renormalize the local four-fermion interactions $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$.
The heavy-quark effective theory (HQET) is constructed to provide a simplified description of processes where a heavy quark interacts with light degrees of freedom predominantly by the exchange of soft gluons $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. Clearly, $`m_Q`$ is the high-energy scale in this case, and $`\mathrm{\Lambda }_{\mathrm{QCD}}`$ is the scale of the hadronic physics we are interested in. The situation is illustrated in Fig. 3. At short distances, i.e. for energy scales larger than the heavy-quark mass, the physics is perturbative and described by conventional QCD. For mass scales much below the heavy-quark mass, the physics is complicated and non-perturbative because of confinement. Our goal is to obtain a simplified description in this region using an effective field theory. To separate short- and long-distance effects, we introduce a separation scale $`\mu `$ such that $`\mathrm{\Lambda }_{\mathrm{QCD}}\mu m_Q`$. The HQET will be constructed in such a way that it is equivalent to QCD in the long-distance region, i.e. for scales below $`\mu `$. In the short-distance region, the effective theory is incomplete, since some high-momentum modes have been integrated out from the full theory. The fact that the physics must be independent of the arbitrary scale $`\mu `$ allows us to derive renormalization-group equations, which can be employed to deal with the short-distance effects in an efficient way.
Compared with most effective theories, in which the degrees of freedom of a heavy particle are removed completely from the low-energy theory, the HQET is special in that its purpose is to describe the properties and decays of hadrons which do contain a heavy quark. Hence, it is not possible to remove the heavy quark completely from the effective theory. What is possible is to integrate out the “small components” in the full heavy-quark spinor, which describe the fluctuations around the mass shell.
The starting point in the construction of the HQET is the observation that a heavy quark bound inside a hadron moves more or less with the hadron’s velocity $`v`$ and is almost on-shell. Its momentum can be written as
$$p_Q^\mu =m_Qv^\mu +k^\mu ,$$
(7)
where the components of the so-called residual momentum $`k`$ are much smaller than $`m_Q`$. Note that $`v`$ is a four-velocity, so that $`v^2=1`$. Interactions of the heavy quark with light degrees of freedom change the residual momentum by an amount of order $`\mathrm{\Delta }k\mathrm{\Lambda }_{\mathrm{QCD}}`$, but the corresponding changes in the heavy-quark velocity vanish as $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q0`$. In this situation, it is appropriate to introduce large- and small-component fields, $`h_v`$ and $`H_v`$, by
$$h_v(x)=e^{im_Qvx}P_+Q(x),H_v(x)=e^{im_Qvx}P_{}Q(x),$$
(8)
where $`P_+`$ and $`P_{}`$ are projection operators defined as
$$P_\pm =\frac{1\pm \text{/}v}{2}.$$
(9)
It follows that
$$Q(x)=e^{im_Qvx}[h_v(x)+H_v(x)].$$
(10)
Because of the projection operators, the new fields satisfy $`\text{/}vh_v=h_v`$ and $`\text{/}vH_v=H_v`$. In the rest frame, i.e. for $`v^\mu =(1,0,0,0)`$, $`h_v`$ corresponds to the upper two components of $`Q`$, while $`H_v`$ corresponds to the lower ones. Whereas $`h_v`$ annihilates a heavy quark with velocity $`v`$, $`H_v`$ creates a heavy antiquark with velocity $`v`$.
In terms of the new fields, the QCD Lagrangian (6) for a heavy quark takes the form
$$_Q=\overline{h}_vivDh_v\overline{H}_v(ivD+2m_Q)H_v+\overline{h}_vi\text{ /}D_{}H_v+\overline{H}_vi\text{ /}D_{}h_v,$$
(11)
where $`D_{}^\mu =D^\mu v^\mu vD`$ is orthogonal to the heavy-quark velocity: $`vD_{}=0`$. In the rest frame, $`D_{}^\mu =(0,\stackrel{}{D})`$ contains the spatial components of the covariant derivative. From (11), it is apparent that $`h_v`$ describes massless degrees of freedom, whereas $`H_v`$ corresponds to fluctuations with twice the heavy-quark mass. These are the heavy degrees of freedom that will be eliminated in the construction of the effective theory. The fields are mixed by the presence of the third and fourth terms, which describe pair creation or annihilation of heavy quarks and antiquarks. As shown in the first diagram in Fig. 4, in a virtual process, a heavy quark propagating forward in time can turn into an antiquark propagating backward in time, and then turn back into a quark. The energy of the intermediate quantum state $`hh\overline{H}`$ is larger than the energy of the incoming heavy quark by at least $`2m_Q`$. Because of this large energy gap, the virtual quantum fluctuation can only propagate over a short distance $`\mathrm{\Delta }x1/m_Q`$. On hadronic scales set by $`R_{\mathrm{had}}=1/\mathrm{\Lambda }_{\mathrm{QCD}}`$, the process essentially looks like a local interaction of the form
$$\overline{h}_vi\text{ /}D_{}\frac{1}{2m_Q}i\text{ /}D_{}h_v,$$
(12)
where we have simply replaced the propagator for $`H_v`$ by $`1/2m_Q`$. A more correct treatment is to integrate out the small-component field $`H_v`$, thereby deriving a non-local effective action for the large-component field $`h_v`$, which can then be expanded in terms of local operators. Before doing this, let us mention a second type of virtual corrections involving pair creation, namely heavy-quark loops. An example is shown in the second diagram in Fig. 4. Heavy-quark loops cannot be described in terms of the effective fields $`h_v`$ and $`H_v`$, since the quark velocities inside a loop are not conserved and are in no way related to hadron velocities. However, such short-distance processes are proportional to the small coupling constant $`\alpha _s(m_Q)`$ and can be calculated in perturbation theory. They lead to corrections that are added onto the low-energy effective theory in the renormalization procedure.
On a classical level, the heavy degrees of freedom represented by $`H_v`$ can be eliminated using the equation of motion. Taking the variation of the Lagrangian with respect to the field $`\overline{H}_v`$, we obtain
$$(ivD+2m_Q)H_v=i\text{ /}D_{}h_v.$$
(13)
This equation can formally be solved to give
$$H_v=\frac{1}{2m_Q+ivD}i\text{ /}D_{}h_v,$$
(14)
showing that the small-component field $`H_v`$ is indeed of order $`1/m_Q`$. We can now insert this solution into (11) to obtain the “non-local effective Lagrangian”
$$_{\mathrm{eff}}=\overline{h}_vivDh_v+\overline{h}_vi\text{ /}D_{}\frac{1}{2m_Q+ivD}i\text{ /}D_{}h_v.$$
(15)
Clearly, the second term corresponds to the first class of virtual processes shown in Fig. 4.
It is possible to derive this Lagrangian in a more elegant way by manipulating the generating functional for QCD Green functions containing heavy-quark fields $`^\mathrm{?}`$. To this end, one starts from the field redefinition (10) and couples the large-component fields $`h_v`$ to external sources $`\rho _v`$. Green functions with an arbitrary number of $`h_v`$ fields can be constructed by taking derivatives with respect to $`\rho _v`$. No sources are needed for the heavy degrees of freedom represented by $`H_v`$. The functional integral over these fields is Gaussian and can be performed explicitly, leading to the effective action
$$S_{\mathrm{eff}}=\mathrm{d}^4x_{\mathrm{eff}}i\mathrm{ln}\mathrm{\Delta },$$
(16)
with $`_{\mathrm{eff}}`$ as given in (15). The appearance of the logarithm of the determinant
$$\mathrm{\Delta }=\mathrm{exp}\left(\frac{1}{2}\mathrm{Tr}\mathrm{ln}\left[2m_Q+ivDi\eta \right]\right)$$
(17)
is a quantum effect not present in the classical derivation presented above. However, in this case the determinant can be regulated in a gauge-invariant way, and by choosing the gauge $`vA=0`$ one can show that $`\mathrm{ln}\mathrm{\Delta }`$ is just an irrelevant constant $`^{\mathrm{?},\mathrm{?}}`$.
Because of the phase factor in (10), the $`x`$ dependence of the effective heavy-quark field $`h_v`$ is weak. In momentum space, derivatives acting on $`h_v`$ produce powers of the residual momentum $`k`$, which is much smaller than $`m_Q`$. Hence, the non-local effective Lagrangian (15) allows for a derivative expansion:
$$_{\mathrm{eff}}=\overline{h}_vivDh_v+\frac{1}{2m_Q}\underset{n=0}{\overset{\mathrm{}}{}}\overline{h}_vi\text{ /}D_{}\left(\frac{ivD}{2m_Q}\right)^ni\text{ /}D_{}h_v.$$
(18)
Taking into account that $`h_v`$ contains a $`P_+`$ projection operator, and using the identity
$$P_+i\text{ /}D_{}i\text{ /}D_{}P_+=P_+\left[(iD_{})^2+\frac{g_s}{2}\sigma _{\mu \nu }G^{\mu \nu }\right]P_+,$$
(19)
where $`i[D^\mu ,D^\nu ]=g_sG^{\mu \nu }`$ is the gluon field-strength tensor, one finds that $`^{\mathrm{?},\mathrm{?}}`$
$$_{\mathrm{eff}}=\overline{h}_vivDh_v+\frac{1}{2m_Q}\overline{h}_v(iD_{})^2h_v+\frac{g_s}{4m_Q}\overline{h}_v\sigma _{\mu \nu }G^{\mu \nu }h_v+O(1/m_Q^2).$$
(20)
In the limit $`m_Q\mathrm{}`$, only the first term remains:
$$_{\mathrm{}}=\overline{h}_vivDh_v.$$
(21)
This is the effective Lagrangian of the HQET. It gives rise to the Feynman rules shown in Fig. 5.
Let us take a moment to study the symmetries of this Lagrangian $`^\mathrm{?}`$. Since there appear no Dirac matrices, interactions of the heavy quark with gluons leave its spin unchanged. Associated with this is an SU(2) symmetry group, under which $`_{\mathrm{}}`$ is invariant. The action of this symmetry on the heavy-quark fields becomes most transparent in the rest frame, where the generators $`S^i`$ of SU(2) can be chosen as
$$S^i=\frac{1}{2}\left(\begin{array}{cc}\sigma ^i& 0\\ 0& \sigma ^i\end{array}\right);[S^i,S^j]=iϵ^{ijk}S^k.$$
(22)
Here $`\sigma ^i`$ are the Pauli matrices. An infinitesimal SU(2) transformation $`h_v(1+i\stackrel{}{ϵ}\stackrel{}{S})h_v`$ leaves the Lagrangian invariant:
$$\delta _{\mathrm{}}=\overline{h}_v[ivD,i\stackrel{}{ϵ}\stackrel{}{S}]h_v=0.$$
(23)
Another symmetry of the HQET arises since the mass of the heavy quark does not appear in the effective Lagrangian. For $`N_h`$ heavy quarks moving at the same velocity, eq. (21) can be extended by writing
$$_{\mathrm{}}=\underset{i=1}{\overset{N_h}{}}\overline{h}_v^iivDh_v^i.$$
(24)
This is invariant under rotations in flavor space. When combined with the spin symmetry, the symmetry group is promoted to SU$`(2N_h)`$. This is the heavy-quark spin-flavor symmetry $`^{\mathrm{?},\mathrm{?}}`$. Its physical content is that, in the limit $`m_Q\mathrm{}`$, the strong interactions of a heavy quark become independent of its mass and spin.
Consider now the operators appearing at order $`1/m_Q`$ in the effective Lagrangian (20). They are easiest to identify in the rest frame. The first operator,
$$𝒪_{\mathrm{kin}}=\frac{1}{2m_Q}\overline{h}_v(iD_{})^2h_v\frac{1}{2m_Q}\overline{h}_v(i\stackrel{}{D})^2h_v,$$
(25)
is the gauge-covariant extension of the kinetic energy arising from the residual motion of the heavy quark. The second operator is the non-Abelian analogue of the Pauli interaction, which describes the color-magnetic coupling of the heavy-quark spin to the gluon field:
$$𝒪_{\mathrm{mag}}=\frac{g_s}{4m_Q}\overline{h}_v\sigma _{\mu \nu }G^{\mu \nu }h_v\frac{g_s}{m_Q}\overline{h}_v\stackrel{}{S}\stackrel{}{B}_ch_v.$$
(26)
Here $`\stackrel{}{S}`$ is the spin operator defined in (22), and $`B_c^i=\frac{1}{2}ϵ^{ijk}G^{jk}`$ are the components of the color-magnetic field. The chromo-magnetic interaction is a relativistic effect, which scales like $`1/m_Q`$. This is the origin of the heavy-quark spin symmetry.
### 2.3 The Residual Mass Term and the Definition of the Heavy-Quark Mass
The choice of the expansion parameter in the HQET, i.e. the definition of the heavy-quark mass $`m_Q`$, deserves some comments. In the derivation presented earlier in this section, we chose $`m_Q`$ to be the “mass in the Lagrangian”, and using this parameter in the phase redefinition in (10) we obtained the effective Lagrangian (21), in which the heavy-quark mass no longer appears. However, this treatment has its subtleties. The symmetries of the HQET allow a “residual mass” $`\delta m`$ for the heavy quark, provided that $`\delta m`$ is of order $`\mathrm{\Lambda }_{\mathrm{QCD}}`$ and is the same for all heavy-quark flavors. Even if we arrange that such a mass term is not present at the tree level, it will in general be induced by quantum corrections. (This is unavoidable if the theory is regulated with a dimensionful cutoff.) Therefore, instead of (21) we should write the effective Lagrangian in the more general form $`^\mathrm{?}`$
$$_{\mathrm{}}=\overline{h}_vivDh_v\delta m\overline{h}_vh_v.$$
(27)
If we redefine the expansion parameter according to $`m_Qm_Q+\mathrm{\Delta }m`$, the residual mass changes in the opposite way: $`\delta m\delta m\mathrm{\Delta }m`$. This implies that there is a unique choice of the expansion parameter $`m_Q`$ such that $`\delta m=0`$. Requiring $`\delta m=0`$, as it is usually done implicitly in the HQET, defines a heavy-quark mass, which in perturbation theory coincides with the pole mass $`^\mathrm{?}`$. This, in turn, defines for each heavy hadron $`H_Q`$ a parameter $`\overline{\mathrm{\Lambda }}`$ (sometimes called the “binding energy”) through
$$\overline{\mathrm{\Lambda }}=(m_{H_Q}m_Q)|_{m_Q\mathrm{}}.$$
(28)
If one prefers to work with another choice of the expansion parameter, the values of non-perturbative parameters such as $`\overline{\mathrm{\Lambda }}`$ change, but at the same time one has to include the residual mass term in the HQET Lagrangian. It can be shown that the various parameters depending on the definition of $`m_Q`$ enter the predictions for physical quantities in such a way that the results are independent of the particular choice adopted $`^\mathrm{?}`$.
There is one more subtlety hidden in the above discussion. The quantities $`m_Q`$, $`\overline{\mathrm{\Lambda }}`$ and $`\delta m`$ are non-perturbative parameters of the HQET, which have a similar status as the vacuum condensates in QCD phenomenology $`^\mathrm{?}`$. These parameters cannot be defined unambiguously in perturbation theory. The reason lies in the divergent behavior of perturbative expansions in large orders, which is associated with the existence of singularities along the real axis in the Borel plane, the so-called renormalons $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. For instance, the perturbation series which relates the pole mass $`m_Q`$ of a heavy quark to its bare mass,
$$m_Q=m_Q^{\mathrm{bare}}\left\{1+c_1\alpha _s(m_Q)+c_2\alpha _s^2(m_Q)+\mathrm{}+c_n\alpha _s^n(m_Q)+\mathrm{}\right\},$$
(29)
contains numerical coefficients $`c_n`$ that grow as $`n!`$ for large $`n`$, rendering the series divergent and not Borel summable $`^{\mathrm{?},\mathrm{?}}`$. The best one can achieve is to truncate the perturbation series at its minimal term, but this leads to an unavoidable arbitrariness of order $`\mathrm{\Delta }m_Q\mathrm{\Lambda }_{\mathrm{QCD}}`$ (the size of the minimal term) in the value of the pole mass. This observation, which at first sight seems a serious problem for QCD phenomenology, should not come as a surprise. We know that because of confinement quarks do not appear as physical states in nature. Hence, there is no unique way to define their on-shell properties such as a pole mass. Remarkably, QCD perturbation theory “knows” about its incompleteness and indicates, through the appearance of renormalon singularities, the presence of non-perturbative effects. One must first specify a scheme how to truncate the QCD perturbation series before non-perturbative statements such as $`\delta m=0`$ become meaningful, and hence before non-perturbative parameters such as $`m_Q`$ and $`\overline{\mathrm{\Lambda }}`$ become well-defined quantities. The actual values of these parameters will depend on this scheme.
We stress that the “renormalon ambiguities” are not a conceptual problem for the heavy-quark expansion. In fact, it can be shown quite generally that these ambiguities cancel in all predictions for physical observables $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. The way the cancellations occur is intricate, however. The generic structure of the heavy-quark expansion for an observable is of the form:
$$\text{Observable}C[\alpha _s(m_Q)]\left(1+\frac{\mathrm{\Lambda }}{m_Q}+\mathrm{}\right),$$
(30)
where $`C[\alpha _s(m_Q)]`$ represents a perturbative coefficient function, and $`\mathrm{\Lambda }`$ is a dimensionful non-perturbative parameter. The truncation of the perturbation series defining the coefficient function leads to an arbitrariness of order $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q`$, which cancels against a corresponding arbitrariness of order $`\mathrm{\Lambda }_{\mathrm{QCD}}`$ in the definition of the non-perturbative parameter $`\mathrm{\Lambda }`$.
The renormalon problem poses itself when one imagines to apply perturbation theory to very high orders. In practice, the perturbative coefficients are known to finite order in $`\alpha _s`$ (typically to one- or two-loop accuracy), and to be consistent one should use them in connection with the pole mass (and $`\overline{\mathrm{\Lambda }}`$ etc.) defined to the same order.
### 2.4 Spectroscopic Implications
The spin-flavor symmetry leads to many interesting relations between the properties of hadrons containing a heavy quark. The most direct consequences concern the spectroscopy of such states $`^{\mathrm{?},\mathrm{?}}`$. In the limit $`m_Q\mathrm{}`$, the spin of the heavy quark and the total angular momentum $`j`$ of the light degrees of freedom are separately conserved by the strong interactions. Because of heavy-quark symmetry, the dynamics is independent of the spin and mass of the heavy quark. Hadronic states can thus be classified by the quantum numbers (flavor, spin, parity, etc.) of their light degrees of freedom $`^\mathrm{?}`$. The spin symmetry predicts that, for fixed $`j0`$, there is a doublet of degenerate states with total spin $`J=j\pm \frac{1}{2}`$. The flavor symmetry relates the properties of states with different heavy-quark flavor.
In general, the mass of a hadron $`H_Q`$ containing a heavy quark $`Q`$ obeys an expansion of the form
$$m_{H_Q}=m_Q+\overline{\mathrm{\Lambda }}+\frac{\mathrm{\Delta }m^2}{2m_Q}+O(1/m_Q^2).$$
(31)
The parameter $`\overline{\mathrm{\Lambda }}`$ represents contributions arising from terms in the Lagrangian that are independent of the heavy-quark mass $`^\mathrm{?}`$, whereas the quantity $`\mathrm{\Delta }m^2`$ originates from the terms of order $`1/m_Q`$ in the effective Lagrangian of the HQET. For the ground-state pseudoscalar and vector mesons, one can parametrize the contributions from the kinetic energy and the chromo-magnetic interaction in terms of two quantities $`\lambda _1`$ and $`\lambda _2`$, in such a way that $`^\mathrm{?}`$
$$\mathrm{\Delta }m^2=\lambda _1+2\left[J(J+1)\frac{3}{2}\right]\lambda _2.$$
(32)
The hadronic parameters $`\overline{\mathrm{\Lambda }}`$, $`\lambda _1`$ and $`\lambda _2`$ are independent of $`m_Q`$. They characterize the properties of the light constituents.
Consider, as a first example, the SU(3) mass splittings for heavy mesons. The heavy-quark expansion predicts that
$`m_{B_S}m_{B_d}`$ $`=`$ $`\overline{\mathrm{\Lambda }}_s\overline{\mathrm{\Lambda }}_d+O(1/m_b),`$
$`m_{D_S}m_{D_d}`$ $`=`$ $`\overline{\mathrm{\Lambda }}_s\overline{\mathrm{\Lambda }}_d+O(1/m_c),`$ (33)
where we have indicated that the value of the parameter $`\overline{\mathrm{\Lambda }}`$ depends on the flavor of the light quark. Thus, to the extent that the charm and bottom quarks can both be considered sufficiently heavy, the mass splittings should be similar in the two systems. This prediction is confirmed experimentally, since
$`m_{B_S}m_{B_d}`$ $`=`$ $`(90\pm 3)\text{MeV},`$
$`m_{D_S}m_{D_d}`$ $`=`$ $`(99\pm 1)\text{MeV}.`$ (34)
As a second example, consider the spin splittings between the ground-state pseudoscalar ($`J=0`$) and vector ($`J=1`$) mesons, which are the members of the spin-doublet with $`j=\frac{1}{2}`$. From (31) and (32), it follows that
$`m_B^{}^2m_B^2`$ $`=`$ $`4\lambda _2+O(1/m_b),`$
$`m_D^{}^2m_D^2`$ $`=`$ $`4\lambda _2+O(1/m_c).`$ (35)
The data are compatible with this:
$`m_B^{}^2m_B^2`$ $``$ $`0.49\mathrm{GeV}^2,`$
$`m_D^{}^2m_D^2`$ $``$ $`0.55\mathrm{GeV}^2.`$ (36)
Assuming that the $`B`$ system is close to the heavy-quark limit, we obtain the value
$$\lambda _20.12\text{GeV}^2$$
(37)
for one of the hadronic parameters in (32). This quantity plays an important role in the phenomenology of inclusive decays of heavy hadrons.
A third example is provided by the mass splittings between the ground-state mesons and baryons containing a heavy quark. The HQET predicts that
$`m_{\mathrm{\Lambda }_b}m_B`$ $`=`$ $`\overline{\mathrm{\Lambda }}_{\mathrm{baryon}}\overline{\mathrm{\Lambda }}_{\mathrm{meson}}+O(1/m_b),`$
$`m_{\mathrm{\Lambda }_c}m_D`$ $`=`$ $`\overline{\mathrm{\Lambda }}_{\mathrm{baryon}}\overline{\mathrm{\Lambda }}_{\mathrm{meson}}+O(1/m_c).`$ (38)
This is again consistent with the experimental results
$`m_{\mathrm{\Lambda }_b}m_B`$ $`=`$ $`(345\pm 9)\text{MeV},`$
$`m_{\mathrm{\Lambda }_c}m_D`$ $`=`$ $`(416\pm 1)\text{MeV},`$ (39)
although in this case the data indicate sizeable symmetry-breaking corrections. The dominant correction to the relations (2.4) comes from the contribution of the chromo-magnetic interaction to the masses of the heavy mesons,<sup>b</sup><sup>b</sup>bBecause of spin symmetry, there is no such contribution to the masses of $`\mathrm{\Lambda }_Q`$ baryons. which adds a term $`3\lambda _2/2m_Q`$ on the right-hand side. Including this term, we obtain the refined prediction that the two quantities
$`m_{\mathrm{\Lambda }_b}m_B{\displaystyle \frac{3\lambda _2}{2m_B}}`$ $`=`$ $`(311\pm 9)\text{MeV},`$
$`m_{\mathrm{\Lambda }_c}m_D{\displaystyle \frac{3\lambda _2}{2m_D}}`$ $`=`$ $`(320\pm 1)\text{MeV}`$ (40)
should be close to each other. This is clearly satisfied by the data.
The mass formula (31) can also be used to derive information on the heavy-quark masses from the observed hadron masses. Introducing the “spin-averaged” meson masses $`\overline{m}_B=\frac{1}{4}(m_B+3m_B^{})5.31`$ GeV and $`\overline{m}_D=\frac{1}{4}(m_D+3m_D^{})1.97`$ GeV, we find that
$$m_bm_c=(\overline{m}_B\overline{m}_D)\left\{1\frac{\lambda _1}{2\overline{m}_B\overline{m}_D}+O(1/m_Q^3)\right\}.$$
(41)
Using theoretical estimates for the parameter $`\lambda _1`$, which lie in the range $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$
$$\lambda _1=(0.3\pm 0.2)\text{GeV}^2,$$
(42)
this relation leads to
$$m_bm_c=(3.39\pm 0.03\pm 0.03)\text{GeV},$$
(43)
where the first error reflects the uncertainty in the value of $`\lambda _1`$, and the second one takes into account unknown higher-order corrections. The fact that the difference of the pole masses, $`m_bm_c`$, is known rather precisely is important for the analysis of inclusive decays of heavy hadrons.
## 3 Exclusive Semi-Leptonic Decays
Semi-leptonic decays of $`B`$ mesons have received a lot of attention in recent years. The decay channel $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ has the largest branching fraction of all $`B`$-meson decay modes. From a theoretical point of view, semi-leptonic decays are simple enough to allow for a reliable, quantitative description. The analysis of these decays provides much information about the strong forces that bind the quarks and gluons into hadrons. Schematically, a semi-leptonic decay process is shown in Fig. 6. The strength of the $`bc`$ transition vertex is governed by the element $`V_{cb}`$ of the CKM matrix. The parameters of this matrix are fundamental parameters of the Standard Model. A primary goal of the study of semi-leptonic decays of $`B`$ mesons is to extract with high precision the values of $`|V_{cb}|`$ and $`|V_{ub}|`$. We will now discuss the theoretical basis of such analyses.
### 3.1 Weak Decay Form Factors
Heavy-quark symmetry implies relations between the weak decay form factors of heavy mesons, which are of particular interest. These relations have been derived by Isgur and Wise $`^\mathrm{?}`$, generalizing ideas developed by Nussinov and Wetzel $`^\mathrm{?}`$, and by Voloshin and Shifman $`^{\mathrm{?},\mathrm{?}}`$.
Consider the elastic scattering of a $`B`$ meson, $`\overline{B}(v)\overline{B}(v^{})`$, induced by a vector current coupled to the $`b`$ quark. Before the action of the current, the light degrees of freedom inside the $`B`$ meson orbit around the heavy quark, which acts as a static source of color. On average, the $`b`$ quark and the $`B`$ meson have the same velocity $`v`$. The action of the current is to replace instantaneously (at time $`t=t_0`$) the color source by one moving at a velocity $`v^{}`$, as indicated in Fig. 7. If $`v=v^{}`$, nothing happens; the light degrees of freedom do not realize that there was a current acting on the heavy quark. If the velocities are different, however, the light constituents suddenly find themselves interacting with a moving color source. Soft gluons have to be exchanged to rearrange them so as to form a $`B`$ meson moving at velocity $`v^{}`$. This rearrangement leads to a form-factor suppression, reflecting the fact that, as the velocities become more and more different, the probability for an elastic transition decreases. The important observation is that, in the limit $`m_b\mathrm{}`$, the form factor can only depend on the Lorentz boost $`\gamma =vv^{}`$ connecting the rest frames of the initial- and final-state mesons. Thus, in this limit a dimensionless probability function $`\xi (vv^{})`$ describes the transition. It is called the Isgur-Wise function $`^\mathrm{?}`$. In the HQET, which provides the appropriate framework for taking the limit $`m_b\mathrm{}`$, the hadronic matrix element describing the scattering process can thus be written as
$$\frac{1}{m_B}\overline{B}(v^{})|\overline{b}_v^{}\gamma ^\mu b_v|\overline{B}(v)=\xi (vv^{})(v+v^{})^\mu .$$
(44)
Here $`b_v`$ and $`b_v^{}`$ are the velocity-dependent heavy-quark fields of the HQET. It is important that the function $`\xi (vv^{})`$ does not depend on $`m_b`$. The factor $`1/m_B`$ on the left-hand side compensates for a trivial dependence on the heavy-meson mass caused by the relativistic normalization of meson states, which is conventionally taken to be
$$\overline{B}(p^{})|\overline{B}(p)=2m_Bv^0(2\pi )^3\delta ^3(\stackrel{}{p}\stackrel{}{p}^{}).$$
(45)
Note that there is no term proportional to $`(vv^{})^\mu `$ in (44). This can be seen by contracting the matrix element with $`(vv^{})_\mu `$, which must give zero since $`\text{/}vb_v=b_v`$ and $`\overline{b}_v^{}\text{/}v^{}=\overline{b}_v^{}`$.
It is more conventional to write the above matrix element in terms of an elastic form factor $`F_{\mathrm{el}}(q^2)`$ depending on the momentum transfer $`q^2=(pp^{})^2`$:
$$\overline{B}(v^{})|\overline{b}\gamma ^\mu b|\overline{B}(v)=F_{\mathrm{el}}(q^2)(p+p^{})^\mu ,$$
(46)
where $`p^({}_{}{}^{}{}_{}{}^{)}=m_Bv^({}_{}{}^{}^)`$. Comparing this with (44), we find that
$$F_{\mathrm{el}}(q^2)=\xi (vv^{}),q^2=2m_B^2(vv^{}1).$$
(47)
Because of current conservation, the elastic form factor is normalized to unity at $`q^2=0`$. This condition implies the normalization of the Isgur-Wise function at the kinematic point $`vv^{}=1`$, i.e. for $`v=v^{}`$:
$$\xi (1)=1.$$
(48)
It is in accordance with the intuitive argument that the probability for an elastic transition is unity if there is no velocity change. Since for $`v=v^{}`$ the final-state meson is at rest in the rest frame of the initial meson, the point $`vv^{}=1`$ is referred to as the zero-recoil limit.
The heavy-quark flavor symmetry can be used to replace the $`b`$ quark in the final-state meson by a $`c`$ quark, thereby turning the $`B`$ meson into a $`D`$ meson. Then the scattering process turns into a weak decay process. In the infinite-mass limit, the replacement $`b_v^{}c_v^{}`$ is a symmetry transformation, under which the effective Lagrangian is invariant. Hence, the matrix element
$$\frac{1}{\sqrt{m_Bm_D}}D(v^{})|\overline{c}_v^{}\gamma ^\mu b_v|\overline{B}(v)=\xi (vv^{})(v+v^{})^\mu $$
(49)
is still determined by the same function $`\xi (vv^{})`$. This is interesting, since in general the matrix element of a flavor-changing current between two pseudoscalar mesons is described by two form factors:
$$D(v^{})|\overline{c}\gamma ^\mu b|\overline{B}(v)=f_+(q^2)(p+p^{})^\mu f_{}(q^2)(pp^{})^\mu .$$
(50)
Comparing the above two equations, we find that
$`f_\pm (q^2)`$ $`=`$ $`{\displaystyle \frac{m_B\pm m_D}{2\sqrt{m_Bm_D}}}\xi (vv^{}),`$
$`q^2`$ $`=`$ $`m_B^2+m_D^22m_Bm_Dvv^{}.`$ (51)
Thus, the heavy-quark flavor symmetry relates two a priori independent form factors to one and the same function. Moreover, the normalization of the Isgur-Wise function at $`vv^{}=1`$ now implies a non-trivial normalization of the form factors $`f_\pm (q^2)`$ at the point of maximum momentum transfer, $`q_{\mathrm{max}}^2=(m_Bm_D)^2`$:
$$f_\pm (q_{\mathrm{max}}^2)=\frac{m_B\pm m_D}{2\sqrt{m_Bm_D}}.$$
(52)
The heavy-quark spin symmetry leads to additional relations among weak decay form factors. It can be used to relate matrix elements involving vector mesons to those involving pseudoscalar mesons. A vector meson with longitudinal polarization is related to a pseudoscalar meson by a rotation of the heavy-quark spin. Hence, the spin-symmetry transformation $`c_v^{}^{}c_v^{}^{}`$ relates $`\overline{B}D`$ with $`\overline{B}D^{}`$ transitions. The result of this transformation is $`^\mathrm{?}`$
$`{\displaystyle \frac{1}{\sqrt{m_Bm_D^{}}}}D^{}(v^{},\epsilon )|\overline{c}_v^{}\gamma ^\mu b_v|\overline{B}(v)`$ $`=`$ $`iϵ^{\mu \nu \alpha \beta }\epsilon _\nu ^{}v_\alpha ^{}v_\beta \xi (vv^{}),`$
$`{\displaystyle \frac{1}{\sqrt{m_Bm_D^{}}}}D^{}(v^{},\epsilon )|\overline{c}_v^{}\gamma ^\mu \gamma _5b_v|\overline{B}(v)`$ $`=`$ $`\left[\epsilon ^\mu (vv^{}+1)v^\mu \epsilon ^{}v\right]\xi (vv^{}),`$
where $`\epsilon `$ denotes the polarization vector of the $`D^{}`$ meson. Once again, the matrix elements are completely described in terms of the Isgur-Wise function. Now this is even more remarkable, since in general four form factors, $`V(q^2)`$ for the vector current, and $`A_i(q^2)`$, $`i=0,1,2`$, for the axial current, are required to parameterize these matrix elements. In the heavy-quark limit, they obey the relations $`^\mathrm{?}`$
$`{\displaystyle \frac{m_B+m_D^{}}{2\sqrt{m_Bm_D^{}}}}\xi (vv^{})`$ $`=`$ $`V(q^2)=A_0(q^2)=A_1(q^2)`$
$`=`$ $`\left[1{\displaystyle \frac{q^2}{(m_B+m_D)^2}}\right]^1A_1(q^2),`$
$`q^2`$ $`=`$ $`m_B^2+m_D^{}^22m_Bm_D^{}vv^{}.`$ (54)
Equations (3.1) and (3.1) summarize the relations imposed by heavy-quark symmetry on the weak decay form factors describing the semi-leptonic decay processes $`\overline{B}D\mathrm{}\overline{\nu }`$ and $`\overline{B}D^{}\mathrm{}\overline{\nu }`$. These relations are model-independent consequences of QCD in the limit where $`m_b,m_c\mathrm{\Lambda }_{\mathrm{QCD}}`$. They play a crucial role in the determination of the CKM matrix element $`|V_{cb}|`$. In terms of the recoil variable $`w=vv^{}`$, the differential semi-leptonic decay rates in the heavy-quark limit become $`^\mathrm{?}`$
$`{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }(\overline{B}D\mathrm{}\overline{\nu })}{\mathrm{d}w}}`$ $`=`$ $`{\displaystyle \frac{G_F^2}{48\pi ^3}}|V_{cb}|^2(m_B+m_D)^2m_D^3(w^21)^{3/2}\xi ^2(w),`$
$`{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }(\overline{B}D^{}\mathrm{}\overline{\nu })}{\mathrm{d}w}}`$ $`=`$ $`{\displaystyle \frac{G_F^2}{48\pi ^3}}|V_{cb}|^2(m_Bm_D^{})^2m_D^{}^3\sqrt{w^21}(w+1)^2`$ (55)
$`\times \left[1+{\displaystyle \frac{4w}{w+1}}{\displaystyle \frac{m_B^22wm_Bm_D^{}+m_D^{}^2}{(m_Bm_D^{})^2}}\right]\xi ^2(w).`$
These expressions receive symmetry-breaking corrections, since the masses of the heavy quarks are not infinitely large. Perturbative corrections of order $`\alpha _s^n(m_Q)`$ can be calculated order by order in perturbation theory. A more difficult task is to control the non-perturbative power corrections of order $`(\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q)^n`$. The HQET provides a systematic framework for analyzing these corrections. For the case of weak-decay form factors the analysis of the $`1/m_Q`$ corrections was performed by Luke $`^\mathrm{?}`$. Later, Falk and the present author have analyzed the structure of $`1/m_Q^2`$ corrections for both meson and baryon weak decay form factors $`^\mathrm{?}`$. We shall not discuss these rather technical issues in detail, but only mention the most important result of Luke’s analysis. It concerns the zero-recoil limit, where an analogue of the Ademollo-Gatto theorem $`^\mathrm{?}`$ can be proved. This is Luke’s theorem $`^\mathrm{?}`$, which states that the matrix elements describing the leading $`1/m_Q`$ corrections to weak decay amplitudes vanish at zero recoil. This theorem is valid to all orders in perturbation theory $`^{\mathrm{?},\mathrm{?},\mathrm{?}}`$. Most importantly, it protects the $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ decay rate from receiving first-order $`1/m_Q`$ corrections at zero recoil $`^\mathrm{?}`$. \[A similar statement is not true for the decay $`\overline{B}D\mathrm{}\overline{\nu }`$. The reason is simple but somewhat subtle. Luke’s theorem protects only those form factors not multiplied by kinematic factors that vanish for $`v=v^{}`$. By angular momentum conservation, the two pseudoscalar mesons in the decay $`\overline{B}D\mathrm{}\overline{\nu }`$ must be in a relative $`p`$ wave, and hence the amplitude is proportional to the velocity $`|\stackrel{}{v}_D|`$ of the $`D`$ meson in the $`B`$-meson rest frame. This leads to a factor $`(w^21)`$ in the decay rate. In such a situation, kinematically suppressed form factors can contribute $`^\mathrm{?}`$.\]
### 3.2 Short-Distance Corrections
In Sec. 2, we have discussed the first two steps in the construction of the HQET. Integrating out the small components in the heavy-quark fields, a non-local effective action was derived, which was then expanded in a series of local operators. The effective Lagrangian obtained that way correctly reproduces the long-distance physics of the full theory (see Fig. 3). It does not contain the short-distance physics correctly, however. The reason is obvious: a heavy quark participates in strong interactions through its coupling to gluons. These gluons can be soft or hard, i.e. their virtual momenta can be small, of the order of the confinement scale, or large, of the order of the heavy-quark mass. But hard gluons can resolve the spin and flavor quantum numbers of a heavy quark. Their effects lead to a renormalization of the coefficients of the operators in the HQET. A new feature of such short-distance corrections is that through the running coupling constant they induce a logarithmic dependence on the heavy-quark mass $`^\mathrm{?}`$. Since $`\alpha _s(m_Q)`$ is small, these effects can be calculated in perturbation theory.
Consider, as an example, the matrix elements of the vector current $`V=\overline{q}\gamma ^\mu Q`$. In QCD this current is partially conserved and needs no renormalization. Its matrix elements are free of ultraviolet divergences. Still, these matrix elements have a logarithmic dependence on $`m_Q`$ from the exchange of hard gluons with virtual momenta of the order of the heavy-quark mass. If one goes over to the effective theory by taking the limit $`m_Q\mathrm{}`$, these logarithms diverge. Consequently, the vector current in the effective theory does require a renormalization $`^\mathrm{?}`$. Its matrix elements depend on an arbitrary renormalization scale $`\mu `$, which separates the regions of short- and long-distance physics. If $`\mu `$ is chosen such that $`\mathrm{\Lambda }_{\mathrm{QCD}}\mu m_Q`$, the effective coupling constant in the region between $`\mu `$ and $`m_Q`$ is small, and perturbation theory can be used to compute the short-distance corrections. These corrections have to be added to the matrix elements of the effective theory, which contain the long-distance physics below the scale $`\mu `$. Schematically, then, the relation between matrix elements in the full and in the effective theory is
$$V(m_Q)_{\mathrm{QCD}}=C_0(m_Q,\mu )V_0(\mu )_{\mathrm{HQET}}+\frac{C_1(m_Q,\mu )}{m_Q}V_1(\mu )_{\mathrm{HQET}}+\mathrm{},$$
(56)
where we have indicated that matrix elements in the full theory depend on $`m_Q`$, whereas matrix elements in the effective theory are mass-independent, but do depend on the renormalization scale. The Wilson coefficients $`C_i(m_Q,\mu )`$ are defined by this relation. Order by order in perturbation theory, they can be computed from a comparison of the matrix elements in the two theories. Since the effective theory is constructed to reproduce correctly the low-energy behavior of the full theory, this “matching” procedure is independent of any long-distance physics, such as infrared singularities, non-perturbative effects, and the nature of the external states used in the matrix elements.
The calculation of the coefficient functions in perturbation theory uses the powerful methods of the renormalization group. It is in principle straightforward, yet in practice rather tedious. A comprehensive discussion of most of the existing calculations of short-distance corrections in the HQET can be found in Ref. 18.
### 3.3 Model-Independent Determination of $`|V_{cb}|`$
We will now discuss the most important application of the formalism described above in the context of semi-leptonic decays of $`B`$ mesons. A model-independent determination of the CKM matrix element $`|V_{cb}|`$ based on heavy-quark symmetry can be obtained by measuring the recoil spectrum of $`D^{}`$ mesons produced in $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ decays $`^\mathrm{?}`$. In the heavy-quark limit, the differential decay rate for this process has been given in (3.1). In order to allow for corrections to that limit, we write
$`{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }}{\mathrm{d}w}}`$ $`=`$ $`{\displaystyle \frac{G_F^2}{48\pi ^3}}(m_Bm_D^{})^2m_D^{}^3\sqrt{w^21}(w+1)^2`$ (57)
$`\times \left[1+{\displaystyle \frac{4w}{w+1}}{\displaystyle \frac{m_B^22wm_Bm_D^{}+m_D^{}^2}{(m_Bm_D^{})^2}}\right]|V_{cb}|^2^2(w),`$
where the hadronic form factor $`(w)`$ coincides with the Isgur-Wise function up to symmetry-breaking corrections of order $`\alpha _s(m_Q)`$ and $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q`$. The idea is to measure the product $`|V_{cb}|(w)`$ as a function of $`w`$, and to extract $`|V_{cb}|`$ from an extrapolation of the data to the zero-recoil point $`w=1`$, where the $`B`$ and the $`D^{}`$ mesons have a common rest frame. At this kinematic point, heavy-quark symmetry helps us to calculate the normalization $`(1)`$ with small and controlled theoretical errors. Since the range of $`w`$ values accessible in this decay is rather small ($`1<w<1.5`$), the extrapolation can be done using an expansion around $`w=1`$:
$$(w)=(1)\left[1\widehat{\varrho }^2(w1)+\widehat{c}(w1)^2\mathrm{}\right].$$
(58)
The slope $`\widehat{\varrho }^2`$ and the curvature $`\widehat{c}`$, and indeed more generally the complete shape of the form factor, are tightly constrained by analyticity and unitarity requirements $`^{\mathrm{?},\mathrm{?}}`$. In the long run, the statistics of the experimental results close to zero recoil will be such that these theoretical constraints will not be crucial to get a precision measurement of $`|V_{cb}|`$. They will, however, enable strong consistency checks.
Measurements of the recoil spectrum have been performed by several experimental groups. Figure 8 shows, as an example, the data reported some time ago by the CLEO Collaboration. The weighted average of the experimental results is $`^\mathrm{?}`$
$$|V_{cb}|(1)=(35.2\pm 2.6)\times 10^3.$$
(59)
Heavy-quark symmetry implies that the general structure of the symmetry-breaking corrections to the form factor at zero recoil is $`^\mathrm{?}`$
$$(1)=\eta _A\left(1+0\times \frac{\mathrm{\Lambda }_{\mathrm{QCD}}}{m_Q}+\text{const}\times \frac{\mathrm{\Lambda }_{\mathrm{QCD}}^2}{m_Q^2}+\mathrm{}\right)\eta _A(1+\delta _{1/m^2}),$$
(60)
where $`\eta _A`$ is a short-distance correction arising from the finite renormalization of the flavor-changing axial current at zero recoil, and $`\delta _{1/m^2}`$ parameterizes second-order (and higher) power corrections. The absence of first-order power corrections at zero recoil is a consequence of Luke’s theorem $`^\mathrm{?}`$. The one-loop expression for $`\eta _A`$ has been known for a long time $`^{\mathrm{?},\mathrm{?},\mathrm{?}}`$:
$$\eta _A=1+\frac{\alpha _s(M)}{\pi }\left(\frac{m_b+m_c}{m_bm_c}\mathrm{ln}\frac{m_b}{m_c}\frac{8}{3}\right)0.96.$$
(61)
The scale $`M`$ in the running coupling constant can be fixed by adopting the prescription of Brodsky, Lepage and Mackenzie (BLM) $`^\mathrm{?}`$, where it is identified with the average virtuality of the gluon in the one-loop diagrams that contribute to $`\eta _A`$. If $`\alpha _s(M)`$ is defined in the $`\overline{\text{ms}}`$ scheme, the result is $`^\mathrm{?}`$ $`M0.51\sqrt{m_cm_b}`$. Several estimates of higher-order corrections to $`\eta _A`$ have been discussed. A renormalization-group resummation of logarithms of the type $`(\alpha _s\mathrm{ln}m_b/m_c)^n`$, $`\alpha _s(\alpha _s\mathrm{ln}m_b/m_c)^n`$ and $`m_c/m_b(\alpha _s\mathrm{ln}m_b/m_c)^n`$ leads to $`^{\mathrm{?},\mathrm{?}}`$<sup>-</sup>$`^\mathrm{?}`$ $`\eta _A0.985`$. On the other hand, a resummation of “renormalon-chain” contributions of the form $`\beta _0^{n1}\alpha _s^n`$, where $`\beta _0=11\frac{2}{3}n_f`$ is the first coefficient of the QCD $`\beta `$-function, gives $`^\mathrm{?}`$ $`\eta _A0.945`$. Using these partial resummations to estimate the uncertainty gives $`\eta _A=0.965\pm 0.020`$. Recently, Czarnecki has improved this estimate by calculating $`\eta _A`$ at two-loop order $`^\mathrm{?}`$. His result, $`\eta _A=0.960\pm 0.007`$, is in excellent agreement with the BLM-improved one-loop expression (61). Here the error is taken to be the size of the two-loop correction.
The analysis of the power corrections is more difficult, since it cannot rely on perturbation theory. Three approaches have been discussed: in the “exclusive approach”, all $`1/m_Q^2`$ operators in the HQET are classified and their matrix elements estimated, leading to $`^{\mathrm{?},\mathrm{?}}`$ $`\delta _{1/m^2}=(3\pm 2)\%`$; the “inclusive approach” has been used to derive the bound $`\delta _{1/m^2}<3\%`$, and to estimate that $`^\mathrm{?}`$ $`\delta _{1/m^2}=(7\pm 3)\%`$; the “hybrid approach” combines the virtues of the former two to obtain a more restrictive lower bound on $`\delta _{1/m^2}`$. This leads to $`^\mathrm{?}`$ $`\delta _{1/m^2}=0.055\pm 0.025`$.
Combining the above results, adding the theoretical errors linearly to be conservative, gives
$$(1)=0.91\pm 0.03$$
(62)
for the normalization of the hadronic form factor at zero recoil. Thus, the corrections to the heavy-quark limit amount to a moderate decrease of the form factor of about 10%. This can be used to extract from the experimental result (59) the model-independent value
$$|V_{cb}|=(38.7\pm 2.8_{\mathrm{exp}}\pm 1.3_{\mathrm{th}})\times 10^3.$$
(63)
### 3.4 Measurements of $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ and $`\overline{B}D\mathrm{}\overline{\nu }`$ Form Factors and Tests of Heavy-Quark Symmetry
We have discussed earlier in this section that heavy-quark symmetry implies relations between the semi-leptonic form factors of heavy mesons. They receive symmetry-breaking corrections, which can be estimated using the HQET. The extent to which these relations hold can be tested experimentally by comparing the different form factors describing the decays $`\overline{B}D^{()}\mathrm{}\overline{\nu }`$ at the same value of $`w`$.
When the lepton mass is neglected, the differential decay distributions in $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ decays can be parameterized by three helicity amplitudes, or equivalently by three independent combinations of form factors. It has been suggested that a good choice for three such quantities should be inspired by the heavy-quark limit $`^{\mathrm{?},\mathrm{?}}`$. One thus defines a form factor $`h_{A1}(w)`$, which up to symmetry-breaking corrections coincides with the Isgur-Wise function, and two form-factor ratios
$`R_1(w)`$ $`=`$ $`\left[1{\displaystyle \frac{q^2}{(m_B+m_D^{})^2}}\right]{\displaystyle \frac{V(q^2)}{A_1(q^2)}},`$
$`R_2(w)`$ $`=`$ $`\left[1{\displaystyle \frac{q^2}{(m_B+m_D^{})^2}}\right]{\displaystyle \frac{A_2(q^2)}{A_1(q^2)}}.`$ (64)
The relation between $`w`$ and $`q^2`$ has been given in (3.1). This definition is such that in the heavy-quark limit $`R_1(w)=R_2(w)=1`$ independently of $`w`$.
To extract the functions $`h_{A1}(w)`$, $`R_1(w)`$ and $`R_2(w)`$ from experimental data is a complicated task. However, HQET-based calculations suggest that the $`w`$ dependence of the form-factor ratios, which is induced by symmetry-breaking effects, is rather mild $`^\mathrm{?}`$. Moreover, the form factor $`h_{A1}(w)`$ is expected to have a nearly linear shape over the accessible $`w`$ range. This motivates to introduce three parameters $`\varrho _{A1}^2`$, $`R_1`$ and $`R_2`$ by
$`h_{A1}(w)`$ $``$ $`(1)\left[1\varrho _{A1}^2(w1)\right],`$
$`R_1(w)`$ $``$ $`R_1,R_2(w)R_2,`$ (65)
where $`(1)=0.91\pm 0.03`$ from (62). The CLEO Collaboration has extracted these three parameters from an analysis of the angular distributions in $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ decays $`^\mathrm{?}`$. The results are
$$\varrho _{A1}^2=0.91\pm 0.16,R_1=1.18\pm 0.32,R_2=0.71\pm 0.23.$$
(66)
Using the HQET, one obtains an essentially model-independent prediction for the symmetry-breaking corrections to $`R_1`$, whereas the corrections to $`R_2`$ are somewhat model dependent. To good approximation $`^\mathrm{?}`$
$`R_1`$ $``$ $`1+{\displaystyle \frac{4\alpha _s(m_c)}{3\pi }}+{\displaystyle \frac{\overline{\mathrm{\Lambda }}}{2m_c}}1.3\pm 0.1,`$
$`R_2`$ $``$ $`1\kappa {\displaystyle \frac{\overline{\mathrm{\Lambda }}}{2m_c}}0.8\pm 0.2,`$ (67)
with $`\kappa 1`$ from QCD sum rules $`^\mathrm{?}`$. Here $`\overline{\mathrm{\Lambda }}`$ is the “binding energy” as defined in (28). Theoretical calculations $`^{\mathrm{?},\mathrm{?}}`$ as well as phenomenological analyses $`^{\mathrm{?},\mathrm{?}}`$ suggest that $`\overline{\mathrm{\Lambda }}0.45`$–0.65 GeV is the appropriate value to be used in one-loop calculations. A quark-model calculation of $`R_1`$ and $`R_2`$ gives results similar to the HQET predictions $`^\mathrm{?}`$: $`R_11.15`$ and $`R_20.91`$. The experimental data confirm the theoretical prediction that $`R_1>1`$ and $`R_2<1`$, although the errors are still large.
Heavy-quark symmetry has also been tested by comparing the form factor $`(w)`$ in $`\overline{B}D^{}\mathrm{}\overline{\nu }`$ decays with the corresponding form factor $`𝒢(w)`$ governing $`\overline{B}D\mathrm{}\overline{\nu }`$ decays. The theoretical prediction $`^{\mathrm{?},\mathrm{?}}`$
$$\frac{𝒢(1)}{(1)}=1.08\pm 0.06$$
(68)
compares well with the experimental results for this ratio: $`0.99\pm 0.19`$ reported by the CLEO Collaboration $`^\mathrm{?}`$, and $`0.87\pm 0.30`$ reported by the ALEPH Collaboration $`^\mathrm{?}`$. In these analyses, it has also been tested that within experimental errors the shape of the two form factors agrees over the entire range of $`w`$ values.
The results of the analyses described above are very encouraging. Within errors, the experiments confirm the HQET predictions, starting to test them at the level of symmetry-breaking corrections.
## 4 Inclusive Decay Rates
Inclusive decay rates determine the probability of the decay of a particle into the sum of all possible final states with a given set of global quantum numbers. An example is provided by the inclusive semi-leptonic decay rate of the $`B`$ meson, $`\mathrm{\Gamma }(\overline{B}X\mathrm{}\overline{\nu })`$, where the final state consists of a lepton-neutrino pair accompanied by any number of hadrons. Here we shall discuss the theoretical description of inclusive decays of hadrons containing a heavy quark $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. From a theoretical point of view such decays have two advantages: first, bound-state effects related to the initial state, such as the “Fermi motion” of the heavy quark inside the hadron $`^{\mathrm{?},\mathrm{?}}`$, can be accounted for in a systematic way using the heavy-quark expansion; secondly, the fact that the final state consists of a sum over many hadronic channels eliminates bound-state effects related to the properties of individual hadrons. This second feature is based on the hypothesis of quark-hadron duality, which is an important concept in QCD phenomenology. The assumption of duality is that cross sections and decay rates, which are defined in the physical region (i.e. the region of time-like momenta), are calculable in QCD after a “smearing” or “averaging” procedure has been applied $`^\mathrm{?}`$. In semi-leptonic decays, it is the integration over the lepton and neutrino phase space that provides a smearing over the invariant hadronic mass of the final state (so-called global duality). For non-leptonic decays, on the other hand, the total hadronic mass is fixed, and it is only the fact that one sums over many hadronic states that provides an averaging (so-called local duality). Clearly, local duality is a stronger assumption than global duality. It is important to stress that quark-hadron duality cannot yet be derived from first principles; still, it is a necessary assumption for many applications of QCD. The validity of global duality has been tested experimentally using data on hadronic $`\tau `$ decays $`^\mathrm{?}`$.
Using the optical theorem, the inclusive decay width of a hadron $`H_b`$ containing a $`b`$ quark can be written in the form
$$\mathrm{\Gamma }(H_bX)=\frac{1}{m_{H_b}}\text{Im}H_b|𝐓|H_b,$$
(69)
where the transition operator $`𝐓`$ is given by
$$𝐓=i\mathrm{d}^4xT\{_{\mathrm{eff}}(x),_{\mathrm{eff}}(0)\}.$$
(70)
Inserting a complete set of states inside the time-ordered product, we recover the standard expression
$$\mathrm{\Gamma }(H_bX)=\frac{1}{2m_{H_b}}\underset{X}{}(2\pi )^4\delta ^4(p_Hp_X)|X|_{\mathrm{eff}}|H_b|^2$$
(71)
for the decay rate. For the case of semi-leptonic and non-leptonic decays, $`_{\mathrm{eff}}`$ is the effective weak Lagrangian given in (4), which in practice is corrected for short-distance effects $`^{\mathrm{?},\mathrm{?},\mathrm{?}}`$<sup>-</sup>$`^\mathrm{?}`$ arising from the exchange of gluons with virtualities between $`m_W`$ and $`m_b`$. If some quantum numbers of the final states $`X`$ are specified, the sum over intermediate states is to be restricted appropriately. In the case of the inclusive semi-leptonic decay rate, for instance, the sum would include only those states $`X`$ containing a lepton-neutrino pair.
In perturbation theory, some contributions to the transition operator are given by the two-loop diagrams shown on the left-hand side in Fig. 9. Because of the large mass of the $`b`$ quark, the momenta flowing through the internal propagator lines are large. It is thus possible to construct an OPE for the transition operator, in which $`𝐓`$ is represented as a series of local operators containing the heavy-quark fields. The operator with the lowest dimension, $`d=3`$, is $`\overline{b}b`$. It arises by contracting the internal lines of the first diagram. The only gauge-invariant operator with dimension 4 is $`\overline{b}i\text{ /}Db`$; however, the equations of motion imply that between physical states this operator can be replaced by $`m_b\overline{b}b`$. The first operator that is different from $`\overline{b}b`$ has dimension 5 and contains the gluon field. It is given by $`\overline{b}g_s\sigma _{\mu \nu }G^{\mu \nu }b`$. This operator arises from diagrams in which a gluon is emitted from one of the internal lines, such as the second diagram shown in Fig. 9. For dimensional reasons, the matrix elements of such higher-dimensional operators are suppressed by inverse powers of the heavy-quark mass. Thus, any inclusive decay rate of a hadron $`H_b`$ can be written as $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$
$$\mathrm{\Gamma }(H_bX_f)=\frac{G_F^2m_b^5}{192\pi ^3}\left\{c_3^f\overline{b}b_H+c_5^f\frac{\overline{b}g_s\sigma _{\mu \nu }G^{\mu \nu }b_H}{m_b^2}+\mathrm{}\right\},$$
(72)
where the prefactor arises naturally from the loop integrations, $`c_n^f`$ are calculable coefficient functions (which also contain the relevant CKM matrix elements) depending on the quantum numbers $`f`$ of the final state, and $`O_H`$ are the (normalized) forward matrix elements of local operators, for which we use the short-hand notation
$$O_H=\frac{1}{2m_{H_b}}H_b|O|H_b.$$
(73)
In the next step, these matrix elements are systematically expanded in powers of $`1/m_b`$, using the technology of the HQET. The result is $`^{\mathrm{?},\mathrm{?},\mathrm{?}}`$
$`\overline{b}b_H`$ $`=`$ $`1{\displaystyle \frac{\mu _\pi ^2(H_b)\mu _G^2(H_b)}{2m_b^2}}+O(1/m_b^3),`$
$`\overline{b}g_s\sigma _{\mu \nu }G^{\mu \nu }b_H`$ $`=`$ $`2\mu _G^2(H_b)+O(1/m_b),`$ (74)
where we have defined the HQET matrix elements
$`\mu _\pi ^2(H_b)`$ $`=`$ $`{\displaystyle \frac{1}{2m_{H_b}}}H_b(v)|\overline{b}_v(i\stackrel{}{D})^2b_v|H_b(v),`$
$`\mu _G^2(H_b)`$ $`=`$ $`{\displaystyle \frac{1}{2m_{H_b}}}H_b(v)|\overline{b}_v{\displaystyle \frac{g_s}{2}}\sigma _{\mu \nu }G^{\mu \nu }b_v|H_b(v).`$ (75)
Here $`(i\stackrel{}{D})^2=(ivD)^2(iD)^2`$; in the rest frame, this is the square of the operator for the spatial momentum of the heavy quark. Inserting these results into (72) yields
$$\mathrm{\Gamma }(H_bX_f)=\frac{G_F^2m_b^5}{192\pi ^3}\left\{c_3^f\left(1\frac{\mu _\pi ^2(H_b)\mu _G^2(H_b)}{2m_b^2}\right)+2c_5^f\frac{\mu _G^2(H_b)}{m_b^2}+\mathrm{}\right\}.$$
(76)
It is instructive to understand the appearance of the “kinetic energy” contribution $`\mu _\pi ^2`$, which is the gauge-covariant extension of the square of the $`b`$-quark momentum inside the heavy hadron. This contribution is the field-theory analogue of the Lorentz factor $`(1\stackrel{}{v}_b^{\mathrm{\hspace{0.17em}2}})^{1/2}1\stackrel{}{k}^{\mathrm{\hspace{0.17em}2}}/2m_b^2`$, in accordance with the fact that the lifetime, $`\tau =1/\mathrm{\Gamma }`$, for a moving particle increases due to time dilation.
The main result of the heavy-quark expansion for inclusive decay rates is the observation that the free quark decay (i.e. the parton model) provides the first term in a systematic $`1/m_b`$ expansion $`^\mathrm{?}`$. For dimensional reasons, the corresponding rate is proportional to the fifth power of the $`b`$-quark mass. The non-perturbative corrections, which arise from bound-state effects inside the $`B`$ meson, are suppressed by at least two powers of the heavy-quark mass, i.e. they are of relative order $`(\mathrm{\Lambda }_{\mathrm{QCD}}/m_b)^2`$. Note that the absence of first-order power corrections is a consequence of the equations of motion, as there is no independent gauge-invariant operator of dimension 4 that could appear in the OPE. The fact that bound-state effects in inclusive decays are strongly suppressed explains a posteriori the success of the parton model in describing such processes $`^{\mathrm{?},\mathrm{?}}`$.
The hadronic matrix elements appearing in the heavy-quark expansion (76) can be determined to some extent from the known masses of heavy hadron states. For the $`B`$ meson, one finds that
$`\mu _\pi ^2(B)`$ $`=`$ $`\lambda _1=(0.3\pm 0.2)\text{GeV}^2,`$
$`\mu _G^2(B)`$ $`=`$ $`3\lambda _20.36\text{GeV}^2,`$ (77)
where $`\lambda _1`$ and $`\lambda _2`$ are the parameters appearing in the mass formula (32). For the ground-state baryon $`\mathrm{\Lambda }_b`$, in which the light constituents have total spin zero, it follows that
$$\mu _G^2(\mathrm{\Lambda }_b)=0,$$
(78)
while the matrix element $`\mu _\pi ^2(\mathrm{\Lambda }_b)`$ obeys the relation
$$(m_{\mathrm{\Lambda }_b}m_{\mathrm{\Lambda }_c})(\overline{m}_B\overline{m}_D)=\left[\mu _\pi ^2(B)\mu _\pi ^2(\mathrm{\Lambda }_b)\right]\left(\frac{1}{2m_c}\frac{1}{2m_b}\right)+O(1/m_Q^2),$$
(79)
where $`\overline{m}_B`$ and $`\overline{m}_D`$ denote the spin-averaged masses introduced in connection with (41). The above relation implies
$$\mu _\pi ^2(B)\mu _\pi ^2(\mathrm{\Lambda }_b)=(0.01\pm 0.03)\text{GeV}^2.$$
(80)
What remains to be calculated, then, is the coefficient functions $`c_n^f`$ for a given inclusive decay channel.
To illustrate this general formalism, we discuss as an example the determination of $`|V_{cb}|`$ from inclusive semi-leptonic $`B`$ decays. In this case the short-distance coefficients in the general expression (76) are given by $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$
$`c_3^{\mathrm{SL}}`$ $`=`$ $`|V_{cb}|^2\left[18x^2+8x^6x^812x^4\mathrm{ln}x^2+O(\alpha _s)\right],`$
$`c_5^{\mathrm{SL}}`$ $`=`$ $`6|V_{cb}|^2(1x^2)^4.`$ (81)
Here $`x=m_c/m_b`$, and $`m_b`$ and $`m_c`$ are the masses of the $`b`$ and $`c`$ quarks, defined to a given order in perturbation theory $`^\mathrm{?}`$. The $`O(\alpha _s)`$ terms in $`c_3^{\mathrm{SL}}`$ are known exactly $`^\mathrm{?}`$, and reliable estimates exist for the $`O(\alpha _s^2)`$ corrections $`^\mathrm{?}`$. The theoretical uncertainties in this determination of $`|V_{cb}|`$ are quite different from those entering the analysis of exclusive decays. The main sources are the dependence on the heavy-quark masses, higher-order perturbative corrections, and above all the assumption of global quark-hadron duality. A conservative estimate of the total theoretical error on the extracted value of $`|V_{cb}|`$ yields $`^\mathrm{?}`$
$$|V_{cb}|=(0.040\pm 0.003)\left[\frac{\text{B}_{\mathrm{SL}}}{10.5\%}\right]^{1/2}\left[\frac{1.6\text{ps}}{\tau _B}\right]^{1/2}=(40\pm 1_{\mathrm{exp}}\pm 3_{\mathrm{th}})\times 10^3.$$
(82)
The value of $`|V_{cb}|`$ extracted from the inclusive semi-leptonic width is in excellent agreement with the value in (63) obtained from the analysis of the exclusive decay $`\overline{B}D^{}\mathrm{}\overline{\nu }`$. This agreement is gratifying given the differences of the methods used, and it provides an indirect test of global quark-hadron duality. Combining the two measurements gives the final result
$$|V_{cb}|=0.039\pm 0.002.$$
(83)
After $`V_{ud}`$ and $`V_{us}`$, this is the third-best known entry in the CKM matrix.
## 5 Rare $`𝑩`$ Decays and Determination of the Weak Phase $`𝜸`$
The main objectives of the $`B`$ factories are to explore the physics of CP violation, to determine the flavor parameters of the electroweak theory, and to probe for physics beyond the Standard Model. This will test the CKM mechanism, which predicts that all CP violation results from a single complex phase in the quark mixing matrix. Facing the announcement of evidence for a CP asymmetry in the decays $`BJ/\psi K_S`$ by the CDF Collaboration $`^\mathrm{?}`$, the confirmation of direct CP violation in $`K\pi \pi `$ decays by the KTeV and NA48 groups $`^{\mathrm{?},\mathrm{?}}`$, and the successful start of the $`B`$ factories at SLAC and KEK, the year 1999 has been an important step towards achieving this goal.
The determination of the sides and angles of the “unitarity triangle” $`V_{ub}^{}V_{ud}+V_{cb}^{}V_{cd}+V_{tb}^{}V_{td}=0`$ depicted in Fig. 10 plays a central role in the $`B`$ factory program. Adopting the standard phase conventions for the CKM matrix, only the two smallest elements in this relation, $`V_{ub}^{}`$ and $`V_{td}`$, have non-vanishing imaginary parts (to an excellent approximation). In the Standard Model the angle $`\beta =\text{arg}(V_{td})`$ can be determined in a theoretically clean way by measuring the mixing-induced CP asymmetry in the decays $`BJ/\psi K_S`$. The preliminary CDF result implies $`^\mathrm{?}`$ $`\mathrm{sin}2\beta =0.79_{0.44}^{+0.41}`$. The angle $`\gamma =\text{arg}(V_{ub}^{})`$, or equivalently the combination $`\alpha =180^{}\beta \gamma `$, is much harder to determine $`^\mathrm{?}`$. Recently, there has been significant progress in the theoretical understanding of the hadronic decays $`B\pi K`$, and methods have been developed to extract information on $`\gamma `$ from rate measurements for these processes. Here we discuss the charged modes $`B^\pm \pi K`$, which from a theoretical perspective are particularly clean.
In the Standard Model, the main contributions to the decay amplitudes for the rare processes $`B\pi K`$ are due to the penguin-induced flavor-changing neutral current (FCNC) transitions $`\overline{b}\overline{s}q\overline{q}`$, which exceed a small, Cabibbo-suppressed $`\overline{b}\overline{u}u\overline{s}`$ contribution from $`W`$-boson exchange. The weak phase $`\gamma `$ enters through the interference of these two (“penguin” and “tree”) contributions. Because of a fortunate interplay of isospin, Fierz and flavor symmetries, the theoretical description of the charged modes $`B^\pm \pi K`$ is very clean despite the fact that these are exclusive non-leptonic decays $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. Without any dynamical assumption, the hadronic uncertainties in the description of the interference terms relevant to the determination of $`\gamma `$ are of relative magnitude $`O(\lambda ^2)`$ or $`O(ϵ_{\mathrm{SU}(3)}/N_c)`$, where $`\lambda =\mathrm{sin}\theta _C0.22`$ is a measure of Cabibbo suppression, $`ϵ_{\mathrm{SU}(3)}20\%`$ is the typical size of SU(3) breaking, and the factor $`1/N_c`$ indicates that the corresponding terms vanish in the factorization approximation. Factorizable SU(3) breaking can be accounted for in a straightforward way.
Recently, the accuracy of this description has been further improved when it was shown that non-leptonic $`B`$ decays into two light mesons, such as $`B\pi K`$ and $`B\pi \pi `$, admit a systematic heavy-quark expansion $`^\mathrm{?}`$. To leading order in $`1/m_b`$, but to all orders in perturbation theory, the decay amplitudes for these processes can be calculated from first principles without recourse to phenomenological models. The QCD factorization theorem proved in Ref. 122 improves upon the phenomenological approach of “generalized factorization” $`^\mathrm{?}`$, which emerges as the leading term in the heavy-quark limit. With the help of this theorem, the irreducible theoretical uncertainties in the description of the $`B^\pm \pi K`$ decay amplitudes can be reduced by an extra factor of $`O(1/m_b)`$, rendering their analysis essentially model independent. As a consequence of this fact, and because they are dominated by FCNC transitions, the decays $`B^\pm \pi K`$ offer a sensitive probe to physics beyond the Standard Model $`^{\mathrm{?},\mathrm{?}}`$<sup>-</sup>$`^\mathrm{?}`$, much in the same way as the “classical” FCNC processes $`BX_s\gamma `$ or $`BX_s\mathrm{}^+\mathrm{}^{}`$.
### 5.1 Theory of $`B^\pm \pi K`$ Decays
The hadronic decays $`B\pi K`$ are mediated by a low-energy effective weak Hamiltonian $`^\mathrm{?}`$, whose operators allow for three different classes of flavor topologies: QCD penguins, trees, and electroweak penguins. In the Standard Model the weak couplings associated with these topologies are known. From the measured branching ratios one can deduce that QCD penguins dominate the $`B\pi K`$ decay amplitudes $`^\mathrm{?}`$, whereas trees and electroweak penguins are subleading and of a similar strength $`^\mathrm{?}`$. The theoretical description of the two charged modes $`B^\pm \pi ^\pm K^0`$ and $`B^\pm \pi ^0K^\pm `$ exploits the fact that the amplitudes for these processes differ in a pure isospin amplitude, $`A_{3/2}`$, defined as the matrix element of the isovector part of the effective Hamiltonian between a $`B`$ meson and the $`\pi K`$ isospin eigenstate with $`I=\frac{3}{2}`$. In the Standard Model the parameters of this amplitude are determined, up to an overall strong phase $`\varphi `$, in the limit of SU(3) flavor symmetry $`^\mathrm{?}`$. Using the QCD factorization theorem the SU(3)-breaking corrections can be calculated in a model-independent way up to non-factorizable terms that are power-suppressed in $`1/m_b`$ and vanish in the heavy-quark limit.
A convenient parameterization of the non-leptonic decay amplitudes $`𝒜_{+0}𝒜(B^+\pi ^+K^0)`$ and $`𝒜_{0+}\sqrt{2}𝒜(B^+\pi ^0K^+)`$ is $`^\mathrm{?}`$
$`𝒜_{+0}`$ $`=`$ $`P(1\epsilon _ae^{i\gamma }e^{i\eta }),`$
$`𝒜_{0+}`$ $`=`$ $`P\left[1\epsilon _ae^{i\gamma }e^{i\eta }\epsilon _{3/2}e^{i\varphi }(e^{i\gamma }\delta _{\mathrm{EW}})\right],`$ (84)
where $`P`$ is the dominant penguin amplitude defined as the sum of all terms in the $`B^+\pi ^+K^0`$ amplitude not proportional to $`e^{i\gamma }`$, $`\eta `$ and $`\varphi `$ are strong phases, and $`\epsilon _a`$, $`\epsilon _{3/2}`$ and $`\delta _{\mathrm{EW}}`$ are real hadronic parameters. The weak phase $`\gamma `$ changes sign under a CP transformation, whereas all other parameters stay invariant.
Based on a naive quark-diagram analysis one would not expect the $`B^+\pi ^+K^0`$ amplitude to receive a contribution from $`\overline{b}\overline{u}u\overline{s}`$ tree topologies; however, such a contribution can be induced through final-state rescattering or annihilation contributions $`^\mathrm{?}`$<sup>-</sup>$`^\mathrm{?}`$. They are parameterized by $`\epsilon _a=O(\lambda ^2)`$. In the heavy-quark limit this parameter can be calculated and is found to be very small $`^\mathrm{?}`$: $`\epsilon _a2\%`$. In the future, it will be possible to put upper and lower bounds on $`\epsilon _a`$ by comparing the CP-averaged branching ratios for the decays $`^\mathrm{?}`$ $`B^\pm \pi ^\pm K^0`$ and $`B^\pm K^\pm \overline{K}^0`$. Below we assume $`|\epsilon _a|0.1`$; however, our results will be almost insensitive to this assumption.
The terms proportional to $`\epsilon _{3/2}`$ in (5.1) parameterize the isospin amplitude $`A_{3/2}`$. The weak phase $`e^{i\gamma }`$ enters through the tree process $`\overline{b}\overline{u}u\overline{s}`$, whereas the quantity $`\delta _{\mathrm{EW}}`$ describes the effects of electroweak penguins. The parameter $`\epsilon _{3/2}`$ measures the relative strength of tree and QCD penguin contributions. Information about it can be derived by using SU(3) flavor symmetry to relate the tree contribution to the isospin amplitude $`A_{3/2}`$ to the corresponding contribution in the decay $`B^+\pi ^+\pi ^0`$. Since the final state $`\pi ^+\pi ^0`$ has isospin $`I=2`$, the amplitude for this process does not receive any contribution from QCD penguins. Moreover, electroweak penguins in $`\overline{b}\overline{d}q\overline{q}`$ transitions are negligibly small. We define a related parameter $`\overline{\epsilon }_{3/2}`$ by writing
$$\epsilon _{3/2}=\overline{\epsilon }_{3/2}\sqrt{12\epsilon _a\mathrm{cos}\eta \mathrm{cos}\gamma +\epsilon _a^2},$$
(85)
so that the two quantities agree in the limit $`\epsilon _a0`$. In the SU(3) limit this new parameter can be determined experimentally form the relation $`^\mathrm{?}`$
$$\overline{\epsilon }_{3/2}=R_1\left|\frac{V_{us}}{V_{ud}}\right|\left[\frac{2\text{B}(B^\pm \pi ^\pm \pi ^0)}{\text{B}(B^\pm \pi ^\pm K^0)}\right]^{1/2}.$$
(86)
SU(3)-breaking corrections are described by the factor $`R_1=1.22\pm 0.05`$, which can be calculated in a model-independent way using the QCD factorization theorem for non-leptonic decays $`^\mathrm{?}`$. The quoted error is an estimate of the theoretical uncertainty due to corrections of $`O(\frac{1}{N_c}\frac{m_s}{m_b})`$. Using preliminary data reported by the CLEO Collaboration $`^\mathrm{?}`$ to evaluate the ratio of the CP-averaged branching ratios in (86), we obtain
$$\overline{\epsilon }_{3/2}=0.21\pm 0.06_{\mathrm{exp}}\pm 0.01_{\mathrm{th}}.$$
(87)
With a better measurement of the branching ratios the uncertainty in $`\overline{\epsilon }_{3/2}`$ will be reduced significantly.
Finally, the parameter
$`\delta _{\mathrm{EW}}`$ $`=`$ $`R_2\left|{\displaystyle \frac{V_{cb}^{}V_{cs}}{V_{ub}^{}V_{us}}}\right|{\displaystyle \frac{\alpha }{8\pi }}{\displaystyle \frac{x_t}{\mathrm{sin}^2\theta _W}}\left(1+{\displaystyle \frac{3\mathrm{ln}x_t}{x_t1}}\right)`$ (88)
$`=`$ $`(0.64\pm 0.09)\times {\displaystyle \frac{0.085}{|V_{ub}/V_{cb}|}},`$
with $`x_t=(m_t/m_W)^2`$, describes the ratio of electroweak penguin and tree contributions to the isospin amplitude $`A_{3/2}`$. In the SU(3) limit it is calculable in terms of Standard Model parameters $`^{\mathrm{?},\mathrm{?}}`$. SU(3)-breaking as well as small electromagnetic corrections are accounted for by the quantity $`^{\mathrm{?},\mathrm{?}}`$ $`R_2=0.92\pm 0.09`$. The error quoted in (88) includes the uncertainty in the top-quark mass.
Important observables in the study of the weak phase $`\gamma `$ are the ratio of the CP-averaged branching ratios in the two $`B^\pm \pi K`$ decay modes,
$$R_{}=\frac{\text{B}(B^\pm \pi ^\pm K^0)}{2\text{B}(B^\pm \pi ^0K^\pm )}=0.75\pm 0.28,$$
(89)
and a particular combination of the direct CP asymmetries,
$$\stackrel{~}{A}=\frac{A_{\mathrm{CP}}(B^\pm \pi ^0K^\pm )}{R_{}}A_{\mathrm{CP}}(B^\pm \pi ^\pm K^0)=0.52\pm 0.42.$$
(90)
The experimental values of these quantities are derived using preliminary data reported by the CLEO Collaboration $`^\mathrm{?}`$. The theoretical expressions for $`R_{}`$ and $`\stackrel{~}{A}`$ obtained using the parameterization in (5.1) are
$`R_{}^1`$ $`=`$ $`1+2\overline{\epsilon }_{3/2}\mathrm{cos}\varphi (\delta _{\mathrm{EW}}\mathrm{cos}\gamma )`$
$`+\overline{\epsilon }_{3/2}^2(12\delta _{\mathrm{EW}}\mathrm{cos}\gamma +\delta _{\mathrm{EW}}^2)+O(\overline{\epsilon }_{3/2}\epsilon _a),`$
$`\stackrel{~}{A}`$ $`=`$ $`2\overline{\epsilon }_{3/2}\mathrm{sin}\gamma \mathrm{sin}\varphi +O(\overline{\epsilon }_{3/2}\epsilon _a).`$ (91)
Note that the rescattering effects described by $`\epsilon _a`$ are suppressed by a factor of $`\overline{\epsilon }_{3/2}`$ and thus reduced to the percent level. Explicit expressions for these contributions can be found in Ref. 121.
### 5.2 Lower Bound on $`\gamma `$ and Constraint in the $`(\overline{\rho },\overline{\eta })`$ Plane
There are several strategies for exploiting the above relations. From a measurement of the ratio $`R_{}`$ alone a bound on $`\mathrm{cos}\gamma `$ can be derived, implying a non-trivial constraint on the Wolfenstein parameters $`\overline{\rho }`$ and $`\overline{\eta }`$ defining the apex of the unitarity triangle $`^\mathrm{?}`$. Only CP-averaged branching ratios are needed for this purpose. Varying the strong phases $`\varphi `$ and $`\eta `$ independently we first obtain an upper bound on the inverse of $`R_{}`$. Keeping terms of linear order in $`\epsilon _a`$ yields $`^\mathrm{?}`$
$$R_{}^1\left(1+\overline{\epsilon }_{3/2}|\delta _{\mathrm{EW}}\mathrm{cos}\gamma |\right)^2+\overline{\epsilon }_{3/2}^2\mathrm{sin}^2\gamma +2\overline{\epsilon }_{3/2}|\epsilon _a|\mathrm{sin}^2\gamma .$$
(92)
Provided $`R_{}`$ is significantly smaller than 1, this bound implies an exclusion region for $`\mathrm{cos}\gamma `$ which becomes larger the smaller the values of $`R_{}`$ and $`\overline{\epsilon }_{3/2}`$ are. It is convenient to consider instead of $`R_{}`$ the related quantity $`^\mathrm{?}`$
$$X_R=\frac{\sqrt{R_{}^1}1}{\overline{\epsilon }_{3/2}}=0.72\pm 0.98_{\mathrm{exp}}\pm 0.03_{\mathrm{th}}.$$
(93)
Because of the theoretical factor $`R_1`$ entering the definition of $`\overline{\epsilon }_{3/2}`$ in (86) this is, strictly speaking, not an observable. However, the theoretical uncertainty in $`X_R`$ is so much smaller than the present experimental error that it can be ignored for all practical purposes. The advantage of presenting our results in terms of $`X_R`$ rather than $`R_{}`$ is that the leading dependence on $`\overline{\epsilon }_{3/2}`$ cancels out, leading to the simple bound $`|X_R||\delta _{\mathrm{EW}}\mathrm{cos}\gamma |+O(\overline{\epsilon }_{3/2},\epsilon _a)`$.
In Fig. 11 we show the upper bound on $`X_R`$ as a function of $`|\gamma |`$, obtained by varying the input parameters in the intervals $`0.15\overline{\epsilon }_{3/2}0.27`$ and $`0.49\delta _{\mathrm{EW}}0.79`$ (corresponding to using $`|V_{ub}/V_{cb}|=0.085\pm 0.015`$ in (88)). Note that the effect of the rescattering contribution parameterized by $`\epsilon _a`$ is very small. The gray band shows the current value of $`X_R`$, which still has too large an error to provide any useful information on $`\gamma `$. The situation may change, however, once a more precise measurement of $`X_R`$ will become available. For instance, if the current central value $`X_R=0.72`$ were confirmed, it would imply the bound $`|\gamma |>75^{}`$, marking a significant improvement over the indirect limit $`|\gamma |>37^{}`$ inferred from the global analysis of the unitarity triangle including information from $`K`$$`\overline{K}`$ mixing $`^\mathrm{?}`$.
So far, we have used the inequality (92) to derive a lower bound on $`|\gamma |`$. However, a large part of the uncertainty in the value of $`\delta _{\mathrm{EW}}`$, and thus in the resulting bound on $`|\gamma |`$, comes from the present large error on $`|V_{ub}|`$. Since this is not a hadronic uncertainty, it is appropriate to separate it and turn (92) into a constraint on the Wolfenstein parameters $`\overline{\rho }`$ and $`\overline{\eta }`$. To this end, we use that $`\mathrm{cos}\gamma =\overline{\rho }/\sqrt{\overline{\rho }^2+\overline{\eta }^2}`$ by definition, and $`\delta _{\mathrm{EW}}=(0.24\pm 0.03)/\sqrt{\overline{\rho }^2+\overline{\eta }^2}`$ from (88). The solid lines in Fig. 12 show the resulting constraint in the $`(\overline{\rho },\overline{\eta })`$ plane obtained for the representative values $`X_R=0.5`$, 0.75, 1.0, 1.25 (from right to left), which for $`\overline{\epsilon }_{3/2}=0.21`$ would correspond to $`R_{}=0.82`$, 0.75, 0.68, 0.63, respectively. Values to the right of these lines are excluded. For comparison, the dashed circles show the constraint arising from the measurement of the ratio $`|V_{ub}/V_{cb}|=0.085\pm 0.015`$ in semi-leptonic $`B`$ decays, and the dashed-dotted line shows the bound implied by the present experimental limit on the mass difference $`\mathrm{\Delta }m_s`$ in the $`B_s`$ system $`^\mathrm{?}`$. Values to the left of this line are excluded. It is evident from the figure that the bound resulting from a measurement of the ratio $`X_R`$ in $`B^\pm \pi K`$ decays may be very non-trivial and, in particular, may eliminate the possibility that $`\gamma =0`$. The combination of this bound with information from semi-leptonic decays and $`B`$$`\overline{B}`$ mixing alone would then determine the Wolfenstein parameters $`\overline{\rho }`$ and $`\overline{\eta }`$ within narrow ranges,<sup>c</sup><sup>c</sup>cAn observation of CP violation, such as the measurement of $`ϵ_K`$ in $`K`$$`\overline{K}`$ mixing or $`\mathrm{sin}2\beta `$ in $`BJ/\psi K_S`$ decays, is however needed to fix the sign of $`\overline{\eta }`$. and in the context of the CKM model would prove the existence of direct CP violation in $`B`$ decays. If one is more optimistic, one may even hope that in the future the constraint from $`B\pi K`$ decays may become incompatible with the bound from $`B_s`$$`\overline{B}_s`$ mixing, thus indicating New Physics beyond the Standard Model.<sup>d</sup><sup>d</sup>dAt the time of writing, the bound from $`B_s`$$`\overline{B}_s`$ mixing is being pushed further to the right, making such a scenario a tantalizing possibility.
### 5.3 Extraction of $`\gamma `$
Ultimately, the goal is of course not only to derive a bound on $`\gamma `$ but to determine this parameter directly from the data. This requires to fix the strong phase $`\varphi `$ in (5.1), which can be achieved either through the measurement of a CP asymmetry or with the help of theory. A strategy for an experimental determination of $`\gamma `$ from $`B^\pm \pi K`$ decays has been suggested in Ref. 120. It generalizes a method proposed by Gronau, Rosner and London $`^\mathrm{?}`$ to include the effects of electroweak penguins. The approach has later been refined to account for rescattering contributions to the $`B^\pm \pi ^\pm K^0`$ decay amplitudes $`^\mathrm{?}`$. Before discussing this method, we will first illustrate an easier strategy for a theory-guided determination of $`\gamma `$ based on the QCD factorization theorem for non-leptonic decays $`^\mathrm{?}`$. This method does not require any measurement of a CP asymmetry.
Theory-guided determination:
In the previous section the theoretical predictions for the non-leptonic $`B\pi K`$ decay amplitudes obtained using the QCD factorization theorem were used in a minimal way, i.e. only to calculate the size of the SU(3)-breaking effects parameterized by $`R_1`$ and $`R_2`$ in (86) and (88). The resulting bound on $`\gamma `$ and the corresponding constraint in the $`(\overline{\rho },\overline{\eta })`$ plane are therefore theoretically very clean. However, they are only useful if the value of $`X_R`$ is found to be larger than about 0.5 (see Fig. 11), in which case values of $`|\gamma |`$ below $`65^{}`$ are excluded. If it would turn out that $`X_R<0.5`$, then it is possible to satisfy the inequality (92) also for small values of $`\gamma `$, however, at the price of having a very large strong phase, $`\varphi 180^{}`$. But this possibility can be discarded based on the model-independent prediction that $`^\mathrm{?}`$
$$\varphi =O[\alpha _s(m_b),\mathrm{\Lambda }_{\mathrm{QCD}}/m_b].$$
(94)
A direct calculation of this phase to leading power in $`1/m_b`$ yields $`^\mathrm{?}`$ $`\varphi 11^{}`$. Using the fact that $`\varphi `$ is parametrically small, we can exploit a measurement of the ratio $`X_R`$ to obtain a determination of $`|\gamma |`$ – corresponding to an allowed region in the $`(\overline{\rho },\overline{\eta })`$ plane – rather than just a bound. This determination is unique up to a sign. Note that for small values of $`\varphi `$ the impact of the strong phase in the expression for $`R_{}`$ in (5.1) is a second-order effect. As long as $`|\varphi |\sqrt{2\mathrm{\Delta }\overline{\epsilon }_{3/2}/\overline{\epsilon }_{3/2}}`$, the uncertainty in $`\mathrm{cos}\varphi `$ has a much smaller effect than the uncertainty in $`\overline{\epsilon }_{3/2}`$. With the present value of $`\overline{\epsilon }_{3/2}`$ this is the case as long as $`|\varphi |43^{}`$. We believe it is a safe assumption to take $`|\varphi |<25^{}`$ (i.e. more than twice the value obtained to leading order in $`1/m_b`$), so that $`\mathrm{cos}\varphi >0.9`$.
Solving the equation for $`R_{}`$ in (5.1) for $`\mathrm{cos}\gamma `$, and including the corrections of $`O(\epsilon _a)`$, we find
$$\mathrm{cos}\gamma =\delta _{\mathrm{EW}}\frac{X_R+\frac{1}{2}\overline{\epsilon }_{3/2}(X_R^21+\delta _{\mathrm{EW}}^2)}{\mathrm{cos}\varphi +\overline{\epsilon }_{3/2}\delta _{\mathrm{EW}}}+\frac{\epsilon _a\mathrm{cos}\eta \mathrm{sin}^2\gamma }{\mathrm{cos}\varphi +\overline{\epsilon }_{3/2}\delta _{\mathrm{EW}}},$$
(95)
where we have set $`\mathrm{cos}\varphi =1`$ in the numerator of the $`O(\epsilon _a)`$ term. Using the QCD factorization theorem one finds that $`\epsilon _a\mathrm{cos}\eta 0.02`$ in the heavy-quark limit $`^\mathrm{?}`$, and we assign a 100% uncertainty to this estimate. In evaluating the result (95) we scan the parameters in the ranges $`0.15\overline{\epsilon }_{3/2}0.27`$, $`0.55\delta _{\mathrm{EW}}0.73`$, $`25^{}\varphi 25^{}`$, and $`0.04\epsilon _a\mathrm{cos}\eta \mathrm{sin}^2\gamma 0`$. Figure 13 shows the allowed regions in the $`(\overline{\rho },\overline{\eta })`$ plane for the representative values $`X_R=0.25`$, 0.75, and 1.25 (from right to left). We stress that with this method a useful constraint on the Wolfenstein parameters is obtained for any value of $`X_R`$.
Model-independent determination:
It is important that, once more precise data on $`B^\pm \pi K`$ decays will become available, it will be possible to test the prediction of a small strong phase $`\varphi `$ experimentally. To this end, one must determine the CP asymmetry $`\stackrel{~}{A}`$ defined in (90) in addition to the ratio $`R_{}`$. From (5.1) it follows that for fixed values of $`\overline{\epsilon }_{3/2}`$ and $`\delta _{\mathrm{EW}}`$ the quantities $`R_{}`$ and $`\stackrel{~}{A}`$ define contours in the $`(\gamma ,\varphi )`$ plane, whose intersections determine the two phases up to possible discrete ambiguities $`^{\mathrm{?},\mathrm{?}}`$. Figure 14 shows these contours for some representative values, assuming $`\overline{\epsilon }_{3/2}=0.21`$, $`\delta _{\mathrm{EW}}=0.64`$, and $`\epsilon _a=0`$. In practice, including the uncertainties in the values of these parameters changes the contour lines into contour bands. Typically, the spread of the bands induces an error in the determination of $`\gamma `$ of about $`^\mathrm{?}`$ $`10^{}`$. In the most general case there are up to eight discrete solutions for the two phases, four of which are related to the other four by a sign change $`(\gamma ,\varphi )(\gamma ,\varphi )`$. However, for typical values of $`R_{}`$ it turns out that often only four solutions exist, two of which are related to the other two by a sign change. The theoretical prediction that $`\varphi `$ is small implies that solutions should exist where the contours intersect close to the lower portion in the plot. Other solutions with large $`\varphi `$ are strongly disfavored. Note that according to (5.1) the sign of the CP asymmetry $`\stackrel{~}{A}`$ fixes the relative sign between the two phases $`\gamma `$ and $`\varphi `$. If we trust the theoretical prediction that $`\varphi `$ is negative $`^\mathrm{?}`$, it follows that in most cases there remains only a unique solution for $`\gamma `$, i.e. the CP-violating phase $`\gamma `$ can be determined without any discrete ambiguity.
Consider, as an example, the hypothetical case where $`R_{}=0.8`$ and $`\stackrel{~}{A}=15\%`$. Figure 14 then allows the four solutions where $`(\gamma ,\varphi )(\pm 82^{},21^{})`$ or $`(\pm 158^{},78^{})`$. The second pair of solutions is strongly disfavored because of the large values of the strong phase $`\varphi `$. From the first pair of solutions, the one with $`\varphi 21^{}`$ is closest to our theoretical expectation that $`\varphi 11^{}`$, hence leaving $`\gamma 82^{}`$ as the unique solution.
## 6 Sensitivity to New Physics
In the presence of New Physics the theoretical description of $`B^\pm \pi K`$ decays becomes more complicated. In particular, new CP-violating contributions to the decay amplitudes may be induced. A detailed analysis of such effects has been presented in $`^\mathrm{?}`$. A convenient and completely general parameterization of the two amplitudes in (5.1) is obtained by replacing
$$PP^{},\epsilon _ae^{i\gamma }e^{i\eta }i\rho e^{i\varphi _\rho },\delta _{\mathrm{EW}}ae^{i\varphi _a}+ibe^{i\varphi _b},$$
(96)
where $`\rho `$, $`a`$, $`b`$ are real hadronic parameters, and $`\varphi _\rho `$, $`\varphi _a`$, $`\varphi _b`$ are strong phases. The terms $`i\rho `$ and $`ib`$ change sign under a CP transformation. New Physics effects parameterized by $`P^{}`$ and $`\rho `$ are isospin conserving, while those described by $`a`$ and $`b`$ violate isospin symmetry. Note that the parameter $`P^{}`$ cancels in all ratios of branching ratios and thus does not affect the quantities $`R_{}`$ and $`X_R`$ as well as any CP asymmetry. Because the ratio $`R_{}`$ in (89) would be 1 in the limit of isospin symmetry, it is particularly sensitive to isospin-violating New Physics contributions.
New Physics can affect the bound on $`\gamma `$ derived from (92) as well as the extraction of $`\gamma `$ using the strategies discussed above. We will discuss these two possibilities in turn.
### 6.1 Effects on the Bound on $`\gamma `$
The upper bound on $`R_{}^1`$ in (92) and the corresponding bound on $`X_R`$ shown in Fig. 11 are model-independent results valid in the Standard Model. Note that the extremal value of $`R_{}^1`$ is such that $`|X_R|(1+\delta _{\mathrm{EW}})`$ irrespective of $`\gamma `$. A value of $`|X_R|`$ exceeding this bound would be a clear signal for New Physics $`^{\mathrm{?},\mathrm{?},\mathrm{?}}`$.
Consider first the case where New Physics may induce arbitrary CP-violating contributions to the $`B\pi K`$ decay amplitudes, while preserving isospin symmetry. Then the only change with respect to the Standard Model is that the parameter $`\rho `$ may no longer be as small as $`O(\epsilon _a)`$. Varying the strong phases $`\varphi `$ and $`\varphi _\rho `$ independently, and allowing for an arbitrarily large New Physics contribution to $`\rho `$, one can derive the bound $`^\mathrm{?}`$
$$|X_R|\sqrt{12\delta _{\mathrm{EW}}\mathrm{cos}\gamma +\delta _{\mathrm{EW}}^2}1+\delta _{\mathrm{EW}}.$$
(97)
The extremal value is the same as in the Standard Model, i.e. isospin-conserving New Physics effects cannot lead to a value of $`|X_R|`$ exceeding $`(1+\delta _{\mathrm{EW}})`$. For intermediate values of $`\gamma `$ the Standard Model bound on $`X_R`$ is weakened; but even for large $`\rho =O(1)`$, corresponding to a significant New Physics contribution to the decay amplitudes, the effect is small.
If both isospin-violating and isospin-conserving New Physics contributions are present and involve new CP-violating phases, the analysis becomes more complicated. Still, it is possible to derive model-independent bounds on $`X_R`$. Allowing for arbitrary values of $`\rho `$ and all strong phases, one obtains $`^\mathrm{?}`$
$`|X_R|`$ $``$ $`\sqrt{(|a|+|\mathrm{cos}\gamma |)^2+(|b|+|\mathrm{sin}\gamma |)^2}`$ (98)
$``$ $`1+\sqrt{a^2+b^2}{\displaystyle \frac{2}{\overline{\epsilon }_{3/2}}}+X_R,`$
where the last inequality is relevant only in cases where $`\sqrt{a^2+b^2}1`$. The important point to note is that with isospin-violating New Physics contributions the value of $`|X_R|`$ can exceed the upper bound in the Standard Model by a potentially large amount. For instance, if $`\sqrt{a^2+b^2}`$ is twice as large as in the Standard Model, corresponding to a New Physics contribution to the decay amplitudes of only 10–15%, then $`|X_R|`$ could be as large as 2.6 as compared with the maximal value 1.8 allowed (for arbitrary $`\gamma `$) in the Standard Model. Also, in the most general case where $`b`$ and $`\rho `$ are non-zero, the maximal value $`|X_R|`$ can take is no longer restricted to occur at the endpoints $`\gamma =0^{}`$ or $`180^{}`$, which are disfavored by the global analysis of the unitarity triangle $`^\mathrm{?}`$. Rather, $`|X_R|`$ would take its maximal value if $`|\mathrm{tan}\gamma |=|\rho |=|b/a|`$.
The present experimental value of $`X_R`$ in (93) has too large an error to determine whether there is any deviation from the Standard Model. If $`X_R`$ turns out to be larger than 1 (i.e. at least one third of a standard deviation above its current central value), then an interpretation of this result in the Standard Model would require a large value $`|\gamma |>91^{}`$ (see Fig. 11), which would be difficult to accommodate in view of the upper bound implied by the experimental constraint on $`B_s`$$`\overline{B}_s`$ mixing, thus providing evidence for New Physics. If $`X_R>1.3`$, one could go a step further and conclude that the New Physics must necessarily violate isospin $`^\mathrm{?}`$.
### 6.2 Effects on the Determination of $`\gamma `$
A value of the observable $`R_{}`$ violating the bound (92) would be an exciting hint for New Physics. However, even if a future precise measurement will give a value that is consistent with the Standard Model bound, $`B^\pm \pi K`$ decays provide an excellent testing ground for physics beyond the Standard Model. This is so because New Physics may cause a significant shift in the value of $`\gamma `$ extracted using the strategies discussed earlier, leading to inconsistencies when this value is compared with other determinations of $`\gamma `$.
A global fit of the unitarity triangle combining information from semi-leptonic $`B`$ decays, $`B`$$`\overline{B}`$ mixing, CP violation in the kaon system, and mixing-induced CP violation in $`BJ/\psi K_S`$ decays provides information on $`\gamma `$ which in a few years will determine its value within a rather narrow range $`^\mathrm{?}`$. Such an indirect determination could be complemented by direct measurements of $`\gamma `$ using, e.g., $`BDK^{()}`$ decays, or using the triangle relation $`\gamma =180^{}\alpha \beta `$ combined with a measurement of $`\alpha `$. We will assume that a discrepancy of more than $`25^{}`$ between the “true” $`\gamma =\text{arg}(V_{ub}^{})`$ and the value $`\gamma _{\pi K}`$ extracted in $`B^\pm \pi K`$ decays will be observable after a few years of operation at the $`B`$ factories. This sets the benchmark for sensitivity to New Physics effects.
In order to illustrate how big an effect New Physics could have on the extracted value of $`\gamma `$, we consider the simplest case where there are no new CP-violating couplings. Then all New Physics contributions in (96) are parameterized by the single parameter $`a_{\mathrm{NP}}a\delta _{\mathrm{EW}}`$. A more general discussion can be found in Ref. 127. We also assume for simplicity that the strong phase $`\varphi `$ is small, as suggested by (94). In this case the difference between the value $`\gamma _{\pi K}`$ extracted from $`B^\pm \pi K`$ decays and the “true” value of $`\gamma `$ is to a good approximation given by
$$\mathrm{cos}\gamma _{\pi K}\mathrm{cos}\gamma a_{\mathrm{NP}}.$$
(99)
In Fig. 15 we show contours of constant $`X_R`$ versus $`\gamma `$ and $`a`$, assuming without loss of generality that $`\gamma >0`$. Obviously, even a moderate New Physics contribution to the parameter $`a`$ can induce a large shift in $`\gamma `$. Note that the present central value of $`X_R0.7`$ is such that values of $`a`$ less than the Standard Model result $`a0.64`$ are disfavored, since they would require values of $`\gamma `$ exceeding $`100^{}`$, in conflict with the global analysis of the unitarity triangle $`^\mathrm{?}`$.
### 6.3 Survey of New Physics models
In Ref. 127, we have explored how New Physics could affect purely hadronic FCNC transitions of the type $`\overline{b}\overline{s}q\overline{q}`$ focusing, in particular, on isospin violation. Unlike in the Standard Model, where isospin-violating effects in these processes are suppressed by electroweak gauge couplings or small CKM matrix elements, in many New Physics scenarios these effects are not parametrically suppressed relative to isospin-conserving FCNC processes. In the language of effective weak Hamiltonians this implies that the Wilson coefficients of QCD and electroweak penguin operators are of a similar magnitude. For a large class of New Physics models we found that the coefficients of the electroweak penguin operators are, in fact, due to “trojan” penguins, which are neither related to penguin diagrams nor of electroweak origin.
Specifically, we have considered: (a) models with tree-level FCNC couplings of the $`Z`$ boson, extended gauge models with an extra $`Z^{}`$ boson, supersymmetric models with broken R-parity; (b) supersymmetric models with R-parity conservation; (c) two-Higgs-doublet models, and models with anomalous gauge-boson couplings. Some of these models have also been investigated in Refs. 125 and 126. In case (a), the electroweak penguin coefficients can be much larger than in the Standard Model because they are due to tree-level processes. In case (b), these coefficients can compete with the ones of the Standard Model because they arise from strong-interaction box diagrams, which scale relative to the Standard Model like $`(\alpha _s/\alpha )(m_W^2/m_{\mathrm{SUSY}}^2)`$. In models (c), on the other hand, isospin-violating New Physics effects are not parametrically enhanced and are generally smaller than in the Standard Model.
For each New Physics model we have explored which region of parameter space can be probed by the $`B^\pm \pi K`$ observables, and how big a departure from the Standard Model predictions one can expect under realistic circumstances, taking into account all constraints on the model parameters implied by other processes. Table 1 summarizes our estimates of the maximal isospin-violating contributions to the decay amplitudes, as parameterized by $`|a_{\mathrm{NP}}|`$. They are the potentially most important source of New Physics effects in $`B^\pm \pi K`$ decays. For comparison, we recall that in the Standard Model $`a0.64`$. Also shown are the corresponding maximal values of the difference $`|\gamma _{\pi K}\gamma |`$. As noted above, in models with tree-level FCNC couplings New Physics effects can be dramatic, whereas in supersymmetric models with R-parity conservation isospin-violating loop effects can be competitive with the Standard Model. In the case of supersymmetric models with R-parity violation the bound (98) implies interesting limits on certain combinations of the trilinear couplings $`\lambda _{ijk}^{}`$ and $`\lambda _{ijk}^{\prime \prime }`$, as discussed in Ref. 127.
## 7 Concluding Remarks
We have presented an introduction to recent developments in the theory and phenomenology of $`B`$ physics, focusing on heavy-quark symmetry, exclusive and inclusive weak decays of $`B`$ mesons, and rare $`B`$ decays that are sensitive to CP-violating weak phases of the Standard Model. The theoretical tools that allow us to perform quantitative calculations are various forms of heavy-quark expansions, i.e. expansions in logarithms and inverse powers of the large scale provided by the heavy-quark mass, $`m_b\mathrm{\Lambda }_{\mathrm{QCD}}`$.
Heavy-flavor physics is a rich and diverse area of current research, which is characterized by a fruitful interplay between theory and experiments. This has led to many significant discoveries and developments. $`B`$ physics has the potential to determine many important parameters of the electroweak theory and to test the Standard Model at low energies. At the same time, through the study of CP violation it provides a window to physics beyond the Standard Model. Indeed, there is a fair chance that such New Physics will first be seen at the $`B`$ factories, before it can be explored in future collider experiments at the Tevatron and the Large Hadron Collider.
Acknowledgements: It is a great pleasure to thank the Organizers of the Trieste Summer School in Particle Physics for the invitation to present these lectures and for providing a stimulating and relaxing atmosphere, which helped to initiate many physics discussions. In particular, I wish to express my gratitude to Gia Dvali, Antonio Masiero, Goran Senjanovic and Alexei Smirnov for their great hospitality and their many efforts to make my stay in Trieste a memorable one. Last but not least, I wish to thank the students of the school for their lively interest in these lectures.
## References |
warning/0001/nucl-ex0001006.html | ar5iv | text | # Measurement of Tensor Polarization in Elastic Electron-Deuteron Scattering at Large Momentum Transfer
## Abstract
Tensor polarization observables ($`\text{t}_{20}`$, $`\text{t}_{21}`$ and $`\text{t}_{22}`$) have been measured in elastic electron-deuteron scattering for six values of momentum transfer between 0.66 and 1.7 (GeV/c)<sup>2</sup>. The experiment was performed at the Jefferson Laboratory in Hall C using the electron HMS Spectrometer, a specially designed deuteron magnetic channel and the recoil deuteron polarimeter POLDER. The new data determine to much larger $`Q^2`$ the deuteron charge form factors $`G_C`$ and $`G_Q`$. They are in good agreement with relativistic calculations and disagree with pQCD predictions.
PACS numbers: 25.30.Bf,13.40.Gp,21.45.+v,24.70.+s
The development of a quantitative understanding of the structure of the deuteron, the only two-nucleon bound state, has long been considered an important testing ground for models of the nucleon-nucleon potential. Nevertheless, the charge distribution of the deuteron is not well known experimentally, because it is only through the use of both polarization measurements and unpolarized elastic scattering cross sections that it can be unambiguously determined. In the experiment described here, a precise determination of the charge form factor of the deuteron is presented through measurement of the deuteron tensor polarization observables up to a momentum transfer of $`Q^2`$=1.7 (GeV/c)<sup>2</sup>, for the first time well beyond its zero crossing.
Since the deuteron is a spin-1 nucleus, its electromagnetic structure is described by three form factors: the charge monopole $`G_C`$, quadrupole $`G_Q`$ and magnetic dipole $`G_M`$. Thus it is possible to unambiguously separate the three components only through measurement of three observables. In the one-photon exchange approximation, the elastic scattering cross section is typically expressed in terms of structure functions $`A(Q^2)`$ and $`B(Q^2)`$ ($`d\sigma /d\mathrm{\Omega }𝒮`$ with $`𝒮=A(Q^2)+B(Q^2)\mathrm{tan}^2(\theta _e/2)`$, see full expressions e.g. in ) that can be separately determined by variation of the scattered electron angle $`\theta _e`$ for a given momentum transfer $`Q^2`$ to the deuteron.
The third observable can be the cross section dependence on deuteron (tensor or vector) polarization. The tensor analyzing powers can be measured using a polarized deuteron target (with unpolarized beam) . Alternatively, the tensor moments of the outgoing deuterons can be measured using unpolarized beam and target. Both types of experiment result in the same combinations of form factors:
$`\text{t}_{20}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}𝒮}}({\displaystyle \frac{8}{3}}\eta G_CG_Q+{\displaystyle \frac{8}{9}}\eta ^2G_Q^2`$ (2)
$`+{\displaystyle \frac{1}{3}}\eta [1+2(1+\eta )\mathrm{tan}^2{\displaystyle \frac{\theta _e}{2}}]G_M^2)`$
$`\text{t}_{21}`$ $`=`$ $`{\displaystyle \frac{2}{\sqrt{3}𝒮\mathrm{cos}\frac{\theta _e}{2}}}\eta \left[\eta +\eta ^2\mathrm{sin}^2{\displaystyle \frac{\theta _e}{2}}\right]^{\frac{1}{2}}G_MG_Q`$ (3)
$`\text{t}_{22}`$ $`=`$ $`{\displaystyle \frac{1}{2\sqrt{3}𝒮}}\eta G_M^2,\text{with }\eta =Q^2/4M_d^2\text{.}`$ (4)
The tensor moment $`\text{t}_{20}`$ is particularly interesting due to its sensitivity to $`G_C`$. It has been previously measured using either the polarimeter or polarized target technique, up to 0.85 (GeV/c)<sup>2</sup>. In our experiment described below, new measurements of t<sub>2q</sub> were performed between 0.66 and 1.7 (GeV/c)<sup>2</sup>. $`A(Q^2)`$ was measured previously up to 4 (GeV/c)<sup>2</sup>, but with significant discrepancy between data sets in our $`Q^2`$ range. New $`A(Q^2)`$ data , including some from this experiment, resolve many of these discrepancies. $`B(Q^2)`$, which is typically a factor 10 smaller than $`A(Q^2)`$, has been measured up to 2.8 (GeV/c)<sup>2</sup>.
Our experiment was performed at the Thomas Jefferson National Accelerator Facility (JLab) in the experimental Hall C. A continuous electron beam with a typical current between 80 and 120 $`\mu `$A was used together with a 12 cm long liquid deuterium target resulting in an average luminosity of about 3$`\times `$10<sup>38</sup> cm<sup>-2</sup>s<sup>-1</sup>.
The scattered electrons were detected in the High Momentum Spectrometer (HMS), in coincidence with the recoil deuterons. The scattered deuterons were transported by a specially designed magnetic channel composed of warm magnets, three quadrupoles and one dipole, to the POLDER polarimeter. This magnetic channel optimized the acceptance matching between the two arms, which varied from 0.5 to 1 depending on the kinematics, and focussed the elastically scattered deuterons on the target of POLDER. The deuteron magnetic channel was set at a fixed angle of 60.5. The six different $`Q^2`$ values were then obtained by changing both the beam energy (from 1.4 to 4 GeV) and the detection angle of the HMS spectrometer.
The elastic scattering events were selected by setting cuts on the primary vertex position and $`\gamma ^{}d`$ invariant mass, as determined by the HMS, the particle energy loss in two thin plastic scintillators located before the polarimeter target, and the time coincidence measurement between the two arms. The combination of these redundant selection criteria reduced the contribution of remaining background (mainly coming from random coincidences between electrons and protons, and from coherent pion production) to less than 0.2%.
The polarimeter POLDER is based on the charge exchange reaction <sup>1</sup>H($`\stackrel{}{d}`$,2p)n, which provides sizeable angular asymmetries depending on the tensor, but not on the vector, components of the incident deuteron polarization . The direction of deuterons is measured with two multi-wire proportional chambers placed upstream of a 22 cm long liquid hydrogen target. Deuterons that undergo a charge exchange reaction, produce two outgoing protons with small relative angle and momentum in the forward direction. They are detected, and their positions measured, in two hodoscopes, composed of plastic scintillator bars. The polar and azimuthal angle distributions of the center of mass of the two protons are used to determine the deuteron beam polarization. The efficiency of the polarimeter, defined as the fraction of the deuterons undergoing a charge exchange reaction, is of the order of (3–6)$`\times `$10<sup>-3</sup> and must be measured with a precision of 1%. The absolute polarized efficiency $`ϵ_{pol}(\theta ,\phi )`$ of the polarimeter, measured in this experiment, has to be compared to the unpolarized value $`ϵ_0(\theta )`$ through the relation :
$`ϵ_{pol}(\theta ,\phi )=ϵ_0(\theta )[{\displaystyle \frac{}{}}1+\text{t}_{20}\text{T}_{20}(\theta )`$ (5)
$`+`$ $`2\mathrm{cos}(\varphi )\text{t}_{21}\text{T}_{21}(\theta )+2\mathrm{cos}(2\varphi )\text{t}_{22}\text{T}_{22}(\theta ){\displaystyle \frac{}{}}],`$ (6)
where T<sub>kq</sub> are the analyzing powers of the <sup>1</sup>H($`\stackrel{}{d}`$,2p)n reaction, t<sub>kq</sub> the deuteron polarization coefficients to be determined in this experiment, $`\theta `$ is the angle between the incident deuteron and the proton pair momentum and $`\varphi `$ the angle between the normal to the <sup>1</sup>H($`\stackrel{}{d}`$,2p)n reaction plane and the $`e`$-$`d`$ scattering plane.
The analyzing powers and the unpolarized efficiency were measured previously at SATURNE using deuteron beams of known polarization in the range of kinetic energies between 140 and 520 MeV, in 10 to 30 MeV steps . The polarimeter data analysis was identical for the calibration and the JLab measurements. The selection of charge exchange events was achieved by requiring a coincidence between the detection of one incident particle before the target and the detection of two charged particles in the hodoscopes. Events with several incident particles were rejected using cuts on the energy loss measured in the scintillators and on the multiplicity information from the wire chambers. Time of flight was measured between the incident deuteron and the hit bars of hodoscopes. Cuts on this time of flight, together with an algorithm to reconstruct proper proton tracks, led to a clean selection of charge exchange events. Two different tracking algorithms, with different geometrical selection criteria, were used to prove that the background (parasitic reactions in the polarimeter or multiple incident particles not rejected by the front end of the polarimeter) within the charge exchange events was negligible. The angles $`\theta `$ and $`\phi `$ were then calculated using the direction of the deuteron and the proton tracks. The deduced efficiency was then stable within 0.6% under changes of experimental conditions (except for the data at the lowest deuteron energy where variations reached 1.2%).
The distributions of incident deuteron energy on the polarimeter at JLab had a large width, 16 to 51 MeV, and were not centered at any of the energies of the calibration experiment. The observables $`ϵ_0`$ and T<sub>kq</sub> of Eq. 6 were then obtained by weighting with deuteron energies the interpolated SATURNE data. For this procedure, the deuteron energy was calculated for each event from the JLab beam energy and the scattered electron angle, with a correction coming from energy loss (mostly in the LD<sub>2</sub> target).
The tensor polarization observables were obtained from Eq. 6 through a minimization procedure, adjusting the $`\text{t}_{2q}`$ values such that the angular distribution on the right-hand side best reproduced the angular distribution of the polarized efficiency measured in this experiment. In this fit, the resulting value of $`\text{t}_{20}`$ is highly correlated with the fixed value of $`ϵ_0`$, but is uncorrelated with $`\text{t}_{21}`$ and $`\text{t}_{22}`$. A small spin precession correction was then applied, corresponding to a net deviation of 29.7 in the deuteron channel. Our results are given in Table I. The systematic errors include those due to analysis cuts (mostly from geometrical POLDER cuts), the uncertainties in the deuteron energy (from beam energy, electron angle, beam position on target), the uncertainties in calibration results (statistical and systematic errors on analyzing powers, interpolation, absolute stability on unpolarized efficiency) as well as the small instrumental unphysical asymmetries measured in the calibration. The uncertainty coming from the knowledge of the deuteron energy as well as the one due to calibration results were larger at the lowest $`Q^2`$ points because of the energy dependence of $`ϵ_0`$ and the stability of the polarimeter at this deuteron energies. In the case of the point at 1.47 (GeV/c)<sup>2</sup>, the $`\theta `$ distribution of $`ϵ_{pol}`$ did not match exactly the expected behaviour from Eq. 6. This led to the addition of a contribution to the systematic error in t<sub>20</sub> for this point of $`\mathrm{\Delta }t_{20}`$=0.1. These systematic errors were combined quadratically and are mostly uncorrelated for the different data points.
For the sake of comparison with other data and with theoretical models, small corrections (of order $`B/A`$ and $`B\mathrm{tan}^2(\theta _e/2)/A`$, see Eqs. 24) were applied to calculate t<sub>2q</sub> at the conventionally accepted angle of 70. These results obtained for the tensor polarization observables are shown in Fig. 2 and 3, and compared with the existing world data and with several recent theoretical predictions. The error bars include both statistical and systematic errors, combined quadratically.
Where the new data overlap with the earlier Bates data, they agree within the combined uncertainties, although it appears that the Bates $`\text{t}_{20}`$ values are systematically more negative. The indication of t<sub>21</sub> crossing 0 is consistent with the existence of a node of the magnetic form factor $`G_M`$ (see Eq. 3) around 2 (GeV/c)<sup>2</sup>, as first indicated by a measurement of $`B(Q^2)`$ .
A recent non-relativistic impulse approximation prediction (NRIA) calculated using the Argonne $`v_{18}`$ potential for the NN interaction, seems to reproduce the Bates data (the dotted curve in Fig.2 and 3). But to be in a reasonable agreement with our new t<sub>20</sub> data, meson exchange currents (MEC) and relativistic corrections (RC) (solid curve) must be included. The MEC calculation includes pair terms and the $`\rho \pi \gamma `$ mechanism, for which the strength is not well known.
Two relativistic and covariant models, both including MEC, are compared with the data. The dashed curve uses a three-dimensional reduction of the Bethe-Salpeter equation using an equal-time formalism, and includes $`\rho \pi \gamma `$ exchange currents. The long dashed curve is the prediction of a model developed in the framework of the explicitly covariant version of light front dynamics. It uses a full relativistic potential, calculated with the same set of mesons and parameter values used in the construction of the Bonn potential, but does not include the $`\rho \pi \gamma `$ MEC. Both models are in good agreement with our t<sub>20</sub> data, but the prediction based on the light cone formalism agrees better with the last NIKHEF data, at lower $`Q^2`$. However, this model does not reproduce the position of the node of $`G_M`$, which leads to a bad description of t<sub>21</sub>.
Finally, two pQCD calculations, predicting simple relations between the form factors of the deuteron, are shown by dashed-dotted curves in Fig. 2 and 3. One of them uses only the helicity-conserving matrix element of the electromagnetic current, arguing that it should dominate above 1 (GeV/c)<sup>2</sup>. The other one includes the helicity-one-flip matrix element and fixes its contribution using the location of the node of $`B(Q^2`$), taken to be at 2 (GeV/c)<sup>2</sup>. Comparison with t<sub>20</sub> and t<sub>21</sub> measurements clearly shows that both pQCD predictions fail to reproduce our data contrary to the scale in four-momentum transfer given by the authors for the applicability of their calculations.
Deuteron form factors can be expressed in terms of $`A`$ (which have been interpolated using the latest data in our $`Q^2`$ range), $`B`$ and $`\text{t}_{20}`$. These equations are quadratic and admit, in general, two solutions. Ambiguities in the choice of the proper solution remain only for our two highest $`Q^2`$ points, due to the fact that $`\text{t}_{20}`$ is close to its maximum, where the two solutions are nearly degenerate. If we follow the prediction of most theoretical models, according to which the maximum of $`\text{t}_{20}`$ occurs beyond our highest $`Q^2`$ point, one of the two solutions can be selected. This particular issue will be addressed elsewhere in more detail . The errors in $`G_C`$ (see Table I) come predominantly from the errors in the $`\text{t}_{20}`$ measurements.
The results for the charge form factors $`G_C`$ and $`G_Q`$, shown in Fig. 4, lead to the same conclusions made for $`\text{t}_{20}`$ data about the models and the Bates data. The results for the charge form factor $`G_C`$ show a node located at a lower value than inferred from the previous Bates data. This removes the inconsistency, pointed out by Henning , in the location of the minimum for the charge form factor of two- and three-nucleon systems. Our data also suggest for the first time a secondary maximum of $`|G_C|`$. The height of this maximum seems to be inconsistent with that of the corresponding three-nucleon system, within the same non-relativistic models .
In summary, we have measured the tensor polarization observables in electron deuteron elastic scattering between 0.65 and 1.72 (GeV/c)<sup>2</sup>. Our data on t<sub>20</sub>, used in conjunction with data on the structure function $`A(Q^2`$), provide a determination of the charge and quadrupole form factors. We have compared our results with only few recent calculations. Within non-relativistic models, all the observables are in favor of the inclusion of meson exchange currents and relativistic effects in the theoretical calculations. In fact the present data could constitute the best experimental determination of isoscalar meson exchange currents. Recent relativistic models are in remarkable agreement with our data. Finally, the $`Q^2`$ range covered by these data shows that the pQCD predictions are not reliable for these momentum transfers.
Acknowledgements: We acknowledge the outstanding work of the JLab accelerator division and the Hall C engineering staff. We thank the Indiana University Cyclotron Facility for its technical help. This work was supported by the French Centre National de la Recherche Scientifique and Commissariat à l’Energie Atomique, the U.S. Department of Energy and National Science Foundation, the Swiss National Science Foundation, and the K.C. Wong Foundation. |
warning/0001/quant-ph0001107.html | ar5iv | text | # Entanglement and Open Systems in Algebraic Quantum Field Theory
## 1. Introduction
In *PCT, Spin and Statistics, and All That*, Streater and Wightman claim that, as a consequence of the axioms of algebraic quantum field theory (AQFT), “it is difficult to isolate a system described by fields from outside effects” (1989, p. 139). Haag makes a similar claim in *Local Quantum Physics*: “From the previous chapters of this book it is evidently not obvious how to achieve a division of the world into parts to which one can assign individuality…Instead we used a division according to regions in space-time. This leads in general to open systems” (1992, p. 298). By a field system these authors mean that portion of a quantum field within a specified bounded open region $`O`$ of spacetime, with its associated algebra of observables $`𝒜(O)`$ (constructed in the usual way, out of field operators smeared with test functions having support in the region). The environment of a field system, so construed, is naturally taken to be the field in the region $`O^{}`$, the spacelike complement of $`O`$. But then the claims above appear, at first sight, puzzling. After all, it is an axiom of AQFT that the observables in $`𝒜(O^{})`$ commute with those in $`𝒜(O)`$. And this implies — indeed, is *equivalent* to — the assertion that standard von Neumann measurements performed in $`O^{}`$ *cannot* have ‘outside effects’ on the expectations of observables in $`O`$ (Lüders, 1951). What, then, could the above authors possibly mean by saying that the field in $`O`$ must be regarded as an open system?
A similar puzzle is raised by a famous passage in which Einstein (1948) contrasts the picture of physical reality embodied in classical field theories with that which emerges when we try to take quantum theory to be complete:
> “If one asks what is characteristic of the realm of physical ideas independently of the quantum theory, then above all the following attracts our attention: the concepts of physics refer to a real external world, i.e., ideas are posited of things that claim a “real existence” independent of the perceiving subject (bodies, fields, etc.)…it appears to be essential for this arrangement of the things in physics that, at a specific time, these things claim an existence independent of one another, insofar as these things “lie in different parts of space”. Without such an assumption of the mutually independent existence (the “being-thus”) of spatially distant things, an assumption which originates in everyday thought, physical thought in the sense familiar to us would not be possible. Nor does one see how physical laws could be formulated and tested without such clean separation.…For the relative independence of spatially distant things ($`A`$ and $`B`$), this idea is characteristic: an external influence on $`A`$ has no *immediate* effect on $`B`$; this is known as the “principle of local action,” which is applied consistently in field theory. The complete suspension of this basic principle would make impossible the idea of the existence of (quasi-)closed systems and, thereby, the establishment of empirically testable laws in the sense familiar to us” (*ibid*, pp. 321-2; Howard’s (1989) translation).
There is a strong temptation to read Einstein’s ‘assumption of the mutually independent existence of spatially distant things’ and his ‘principle of local action’ as anticipating, respectively, the distinction between separability and locality, or between nonlocal ‘outcome-outcome’ correlation and ‘measurement-outcome’ correlation, which some philosophers argue is crucial to unravelling the conceptual implications of Bell’s theorem (see, e.g., Howard 1989). However, even in nonrelativistic quantum theory, there is no question of any nonlocal *measurement*-outcome correlation between distinct systems or degrees of freedom, whose observables are always represented as commuting. Making the reasonable assumption that Einstein knew this quite well, what is it about taking quantum theory at face value that he saw as a threat to securing the existence of physically closed systems?
What makes quantum systems open for Einstein, as well as for Streater and Wightman, and Haag, is that they can occupy entangled states in which they sustain nonclassical EPR correlations with systems outside their light cones. That is, while it is correct to read Einstein’s discussion of the mutually independent existence of distant systems as an implicit critique of the way in which quantum theory often represents their joint state as entangled, we believe it must be the *outcome-outcome* EPR correlations associated with entangled states that, in Einstein’s view, pose a problem for the legitimate testing of the predictions of quantum theory. One could certainly doubt whether EPR correlations really pose any methodological problem, or whether they truly require the existence of physical (or ‘causal’) influences acting on a quantum system from outside. But the analogy with open systems in thermodynamics that Einstein and the others seem to be invoking is not entirely misplaced.
Consider the simplest toy universe consisting of two nonrelativistic quantum systems, represented by a tensor product of two-dimensional Hilbert spaces $`_A^2_B^2`$, where system $`A`$ is the ‘object’ system, and $`B`$ its ‘environment’. Let $`x`$ be any state vector for the composite system $`A+B`$, and $`D_A(x)`$ be the reduced density operator $`x`$ determines for system $`A`$. Then the von Neumann entropy of $`A`$, $`E_A(x)=\text{Tr}(D_A(x)\mathrm{ln}D_A(x))`$ ($`=E_B(x)`$), varies with the degree to which $`A`$ and $`B`$ are entangled. If $`x`$ is a product vector with no entanglement, $`E_A(x)=0`$, whereas, at the opposite extreme, $`E_A(x)=\mathrm{ln}2`$ when $`x`$ is, say, a singlet or triplet state. The more $`A`$ and $`B`$ are entangled, the more ‘disordered’ $`A`$ becomes, because it will have more than one state available to it and $`A`$’s probabilities of occupying them will approach equality. In fact, exploiting an analogy to Carnot’s heat cycle and the second law of thermodynamics — that it is impossible to construct a *perpetuum mobile* — Popescu and Rohrlich (1997) have shown that the general principle that it is impossible to create entanglement between pairs of systems by local operations on one member of each pair implies that the von Neumann entropy of either member provides the uniquely correct measure of their entanglement when they are in a pure state. Changes in their degree of entanglement, and hence in the entropy of either system, can only come about in the presence of a nontrivial interaction Hamiltonian between them. But the fact remains that there is an intimate connection between a system’s entanglement with its environment and the extent to which the system should be thought of as physically closed.
Returning to AQFT, Streater and Wightman, as well as Haag, all intend to make a far stronger claim about quantum field systems — a point that even applies to spacelike separated regions of a *free* field, and might well have offended Einstein’s physical sensibilities even more. The point is that quantum field systems are *unavoidably* and *intrinsically* open to entanglement. Streater and Wightman’s comment is made in reference to the Reeh-Schlieder (1961) theorem, a consequence of the general axioms of AQFT. We shall show that this theorem entails severe *practical* obstacles to isolating field systems from entanglement with other field systems. Haag’s comment goes deeper, and is related to the fact that the algebras associated with field systems localized in spacetime regions are in all known models of the axioms type III von Neumann algebras. We shall show that this feature of the local algebras imposes a fundamental limitation on isolating field systems from entanglement even *in principle*.
Think again of our toy nonrelativistic universe $`A+B`$, with Alice in possession of system $`A`$, and the state $`x`$ entangled. Although there are no operations that Alice can perform on system $`A`$ which will reduce its entropy, she can still try to destroy its entanglement with $`B`$ by performing a standard von Neumann measurement on $`A`$. If $`P_\pm `$ are the eigenprojections of the observable she measures, and the initial density operator of $`A+B`$ is $`D=P_x`$, where $`P_x`$ is the projection onto the ray $`x`$ generates, then the post-measurement joint state of $`A+B`$ will be given by the new density operator
$$DD^{}=(P_+I)P_x(P_+I)+(P_{}I)P_x(P_{}I).$$
(1)
Since the projections $`P_\pm `$ are one-dimensional, and $`x`$ is entangled, there are nonzero vectors $`a_x^\pm _A^2`$ and $`b_x^\pm _B^2`$ such that $`(P_\pm I)x=a_x^\pm b_x^\pm `$, and a straightforward calculation reveals that $`D^{}`$ may be re-expressed as
$$D^{}=\text{Tr}[(P_+I)P_x]P_+P_{b_x^+}+\text{Tr}[(P_{}I)P_x]P_{}P_{b_x^{}}.$$
(2)
Thus, regardless of the initial state $`x`$, or the degree to which it was entangled, $`D^{}`$ will always be a convex combination of product states, and there will no longer be any entanglement between $`A`$ and $`B`$. One might say that Alice’s operation on $`A`$ has the effect of isolating $`A`$ from any further EPR influences from $`B`$.
Moreover, this result can be generalized. Given any finite or infinite dimension for the Hilbert spaces $`𝖧_A`$ and $`𝖧_B`$, there is always an operation Alice can perform on $`A`$ that will destroy its entanglement with $`B`$ no matter what their initial state $`D`$ was, pure or mixed. In fact, it suffices for Alice to measure any nondegenerate observable $`A`$ with a discrete spectrum (excluding $`0`$). The final state $`D^{}`$ will then be a convex combination of product states, each of which is a product density operator obtained by ‘collapsing’ $`D`$ using some particular eigenprojection of the measured observable. (The fact that disentanglement of a state can always be achieved in this way does not conflict with the recently established result there can be no ‘universal disentangling machine’, i.e., no *unitary* evolution that maps an arbitrary $`A+B`$ state $`D`$ to an unentangled state with the same reduced density operators as $`D`$ (Mor 1998; Mor and Terno 1999). Also bear in mind that we have *not* required that a successful disentangling process leave the states of the entangled subsystems unchanged.)
The upshot is that if entanglement *does* pose a methodological threat, it can at least be brought under control in nonrelativistic quantum theory. Not so when we consider the analogous setup in quantum field theory, with Alice in the vicinity of one region $`A`$, and $`B`$ any other spacelike-seperated field system. We shall see that AQFT puts both practical and theoretical limits on Alice’s ability to destroy entanglement between her field system and $`B`$. Again, while one could doubt whether this poses any real methodological problem for Alice (an issue to which we shall return later), we think it is ironic, considering Einstein’s point of view, that such limits should be forced upon us precisely when we make the transition to a fully *relativistic* formulation of quantum theory.
We begin in Section 2. by reviewing the formalism of AQFT, the concept of entanglement between spacelike separated field systems, and the mathematical representation of an operation performed within a local spacetime region on a field system. In Section 3., we connect the Reeh-Schlieder theorem with the practical difficulties involved in guaranteeing that a field system is disentangled from other field systems. The language of operations also turns out to be indispensible for clearing up some apparently paradoxical physical implications of the Reeh-Schlieder theorem that have been raised in the literature without being properly resolved. In Section 4., we discuss differences between type III von Neumann algebras and the standard type I von Neumann algebras employed in nonrelativistic quantum theory, emphasizing the radical implications type III algebras have for the ignorance interpretation of mixtures and entanglement. We end Section 4. by connecting the type III character of the algebra of a local field system with the inability, in principle, to perform local operations on the system that will destroy its entanglement with other spacelike separated systems. We offer this result as one way to make precise the sense in which AQFT requires a radical change in paradigm — a change that, regrettably, has passed virtually unnoticed by philosophers of quantum theory.
## 2. AQFT, Entanglement, and Local Operations
First, let us recall that an abstract $`C^{}`$-algebra is a Banach -algebra, where the involution and norm are related by $`|A^{}A|=|A|^2`$. Thus the algebra $`(𝖧)`$ of all bounded operators on a Hilbert space $`𝖧`$ is a $`C^{}`$-algebra, with taken to be the adjoint operation, and $`||`$ the standard operator norm. Moreover, any -*sub*algebra of $`(𝖧)`$ that is closed in the operator norm is a $`C^{}`$-algebra, and, conversely, one can show that every abstract $`C^{}`$-algebra has a concrete (faithful) representation as a norm-closed -subalgebra of $`(𝖧)`$, for some appropriate Hilbert space $`𝖧`$ (Kadison and Ringrose (henceforth, KR) 1997, Remark 4.5.7). On the other hand, a von Neumann algebra is always taken to be a concrete collection of operators on some fixed Hilbert space $`𝖧`$. For $`F`$ any set of operators on $`𝖧`$, let $`F^{}`$ denote the commutant of $`F`$, the set of all operators on $`𝖧`$ that commute with *every* operator in $`F`$. Observe that $`FF^{\prime \prime }`$, that $`FG`$ implies $`G^{}F^{}`$, and (hence) that $`A^{}=A^{\prime \prime \prime }`$. $``$ is called a von Neumann algebra exactly when $``$ is a -subalgebra of $`(𝖧)`$ that contains the identity and satisfies $`=^{\prime \prime }`$. This is equivalent, via von Neumann’s famous double commutant theorem (KR 1997, Theorem 5.3.1), to the assertion that $``$ is closed in the strong operator topology, where $`Z_nZ`$ strongly just in case $`|(Z_nZ)x|0`$ for all $`x𝖧`$. If the sequence $`\{Z_n\}`$ converges to $`Z`$ in norm, then since $`|(Z_nZ)x||Z_nZ||x|`$, the convergence is also strong, hence every von Neumann algebra is also a $`C^{}`$-algebra. However, not every $`C^{}`$-algebra of operators is a von Neumann algebra. For example, the $`C^{}`$-algebra $`𝒞`$ of all compact operators on an infinite-dimensional Hilbert space $`𝖧`$ — that is, the norm closure of the -subalgebra of all finite rank operators on $`𝖧`$ — does *not* contain the identity, nor does $`𝒞`$ satisfy $`𝒞=𝒞^{\prime \prime }`$. (Indeed, $`𝒞^{\prime \prime }=(𝖧)`$, because only multiples of the identity commute with all finite-dimensional projections, and of course *every* operator commutes with all multiples of the identity.) Finally, let $`S`$ be any self-adjoint (i.e., -closed) set of operators in $`(𝖧)`$. Then $`S^{}`$ is a -algebra containing the identity, and both $`S^{}`$ ($`=S^{\prime \prime \prime }=(S^{})^{\prime \prime }`$) and $`S^{\prime \prime }`$ ($`=(S^{})^{}=(S^{})^{\prime \prime \prime }=(S^{\prime \prime })^{\prime \prime }`$) are von Neumann algebras. If we suppose there is some other von Neumann algebra $``$ such that $`S`$, then $`^{}S^{}`$, which in turn entails $`S^{\prime \prime }^{\prime \prime }=`$. Thus $`S^{\prime \prime }`$ is actually the smallest von Neumann algebra containing $`S`$, i.e., the von Neumann algebra that $`S`$ generates. For example, the von Neumann algebra generated by all finite rank operators is the whole of $`(𝖧)`$.
The basic mathematical object of AQFT on Minkowski spacetime $`M`$ is an association $`O𝒜(O)`$ between bounded open subsets $`O`$ of $`M`$ and $`C^{}`$-subalgebras $`𝒜(O)`$ of an abstract $`C^{}`$-algebra $`𝒜`$ (Horuzhy 1988, Haag 1992). The motivation for this association is that the self-adjoint elements of $`𝒜(O)`$ represent the physical magnitudes, or observables, of the field intrinsic to the region $`O`$. We shall see below how the elements of $`𝒜(O)`$ can also be used to represent mathematically the physical operations that can be performed within $`O`$, and often it is only this latter interpretation of $`𝒜(O)`$ that is emphasized (Haag 1992, p. 104). One naturally assumes
*Isotony:* If $`O_1O_2`$, then $`𝒜(O_1)𝒜(O_2)`$.
As a consequence, the collection of all local algebras $`𝒜(O)`$ defines a net whose limit points can be used to define algebras associated with unbounded regions, and in particular $`𝒜(M)`$, which is identified with $`𝒜`$ itself.
One of the leading ideas in the algebraic approach to fields is that all of the physics of a particular field theory is encoded in the structure of its net of local algebras. (In particular, while any given field algebra on $`M`$ obtained via smearing will define a unique net, the net underdetermines the field algebra; see Borchers 1960.) But there are some general assumptions about the net $`\{𝒜(O):OM\}`$ that all physically reasonable field theories are held to satisfy. First, one assumes
*Microcausality:* $`𝒜(O^{})𝒜(O)^{}`$.
One also assumes that there is a faithful representation $`𝐱\alpha _𝐱`$ of the spacetime translation group of $`M`$ in the group of automorphisms of $`𝒜`$, satisfying
*Translation Covariance:* $`\alpha _𝐱(𝒜(O))=𝒜(O+𝐱)`$.
*Weak Additivity:* For any $`OM`$, $`𝒜`$ is the smallest $`C^{}`$-algebra containing $`_{𝐱M}𝒜(O+𝐱)`$.
Finally, one assumes that there is some irreducible representation of the net $`\{𝒜(O):OM\}`$ in which these local algebras are identified with von Neumann algebras acting on a (nontrivial) Hilbert space $`𝖧`$, $`𝒜`$ is identified with a strongly dense subset of $`(𝖧)`$, and the following condition holds
> *Spectrum Condition*: The generator of spacetime translations, the energy-momentum of the field, has a spectrum confined to the forward light-cone.
These last three conditions, and their role in the proof of the Reeh-Schlieder theorem (microcausality is not needed), are discussed at length in Halvorson (2000). We wish only to note here that while the spectrum condition itself only makes sense relative to a representation — wherein one can speak, via Stone’s theorem, of a generator of the spacetime translation group of $`M`$ (now concretely represented as a strongly continuous group of unitary operators $`\{U_𝐱\}`$ acting on $`𝖧`$) — the requirement that the abstract net *have* a representation satisfying the spectrum condition does not require that one actually *pass* to such a representation to compute expectation values, cross-sections, etc. Indeed, Haag and Kastler (1964) have argued that there is a precise sense in which all concrete representations of a net are physically equivalent, including representations with and without a translationally invariant vacuum state vector $`\mathrm{\Omega }`$. Since we are not concerned with that argument here, we shall henceforth take the ‘Haag-Araki’ approach of assuming that all the local algebras $`\{𝒜(O):OM\}`$ are von Neumann algebras acting on some $`𝖧`$, with $`𝒜^{\prime \prime }=(𝖧)`$, and there is a translationally invariant vacuum state $`\mathrm{\Omega }𝖧`$.
We turn next to the concept of a state of the field. Generally, a physical state of a quantum system represented by some von Neumann algebra $`(𝖧)`$ is given by a normalized linear expectation functional $`\tau `$ on $``$ that is both positive and countably additive. Positivity is the requirement that $`\tau `$ map any positive operator in $``$ to a nonnegative expectation (a must, given that positive operators have nonnegative spectra), while countable additivity is the requirement that $`\tau `$ be additive over countable sums of mutually orthogonal projections in $``$. (There are also non-countably additive or ‘singular’ states on $``$ (KR 1997, p. 723), but whenever we use the term ‘state’ we shall mean *countably additive* state.) Every state on $``$ extends to a state $`\rho `$ on $`(𝖧)`$ which, in turn, can be represented by a density operator $`D_\rho `$ on $`𝖧`$ via the standard formula $`\rho ()=\text{Tr}(D_\rho )`$ (KR 1997, p. 462). A pure state on $`(𝖧)`$, i.e., one that is not a nontrivial convex combination or mixture of other states of $`(𝖧)`$, is then represented by a vector $`x𝖧`$. We shall always use the notation $`\rho _x`$ for the normalized state functional $`(x,x)/|x|^2`$ ($`=\text{Tr}(P_x)`$). If, furthermore, we consider the restriction $`\rho _x|_{}`$, the induced state on some von Neumann subalgebra $`(𝖧)`$, we cannot in general expect it to be pure on $``$ as well. For example, with $`𝖧=_A^2_B^2`$, $`=(_A^2)I`$, and $`x`$ entangled, we know that the induced state $`\rho _x|_{}`$, represented by $`D_A(x)(_A^2)`$, is *always* mixed. Similarly, one cannot expect that a pure state $`\rho _x`$ of the field algebra $`𝒜^{\prime \prime }=(𝖧)`$ — which supplies a maximal specification of the state of the field *throughout* spacetime — will have a restriction to a local algebra $`\rho _x|_{𝒜(O)}`$ that is itself pure. In fact, we shall see later that the Reeh-Schlieder theorem entails that the vacuum state’s restriction to any local algebra is always highly mixed.
There are two topologies on the state space of a von Neumann algebra $``$ that we shall need to invoke. One is the metric topology induced by the norm on linear functionals. The norm of a state $`\rho `$ on $``$ is defined by $`\rho sup\{|\rho (Z)|:Z=Z^{},|Z|1\}`$. If two states, $`\rho _1`$ and $`\rho _2`$, are close to each other in norm, then they dictate close expectation values uniformly for *all* observables. In particular, if both $`\rho _1`$ and $`\rho _2`$ are vector states, i.e., they are induced by vectors $`x_1,x_2𝖧`$ such that $`\rho _1=\rho _{x_1}|_{}`$ and $`\rho _2=\rho _{x_2}|_{}`$, then $`|x_1x_2|0`$ implies $`\rho _1\rho _20`$. (It is important not to conflate the terms ‘vector state’ and ‘pure state’, unless of course $`=(𝖧)`$ itself.) More generally, whenever the trace norm distance between two density operators goes to zero, the norm distance between the states they induce on $``$ goes to zero. Since every state on $`(𝖧)`$ is given by a density operator, which in turn can be decomposed as an infinite convex combination of one dimensional projections (with the infinite sum understood as trace norm convergence), it follows that every state on $`(𝖧)`$ is the norm limit of convex combinations of vectors states of $``$ (cf. KR 1997, Thm. 7.1.12). The other topology we shall invoke is the weak- topology: a sequence or net of states $`\{\rho _n\}`$ on $``$ weak- converges to a state $`\rho `$ just in case $`\rho _n(Z)\rho (Z)`$ for all $`Z`$. This convergence need not be uniform on all elements of $``$, and is therefore weaker than the notion of approximation embodied by norm convergence. As it happens, any state on $`(𝖧)`$ that is the weak- limit of a set of states is also their norm limit, but this is only true for type I von Neumann algebras (Connes and Størmer, 1978, Cor. 9).
Next, we turn to defining entanglement in a field. Fix a state $`\rho `$ on $`(𝖧)`$, and two mutually commuting subalgebras $`_A,_B(𝖧)`$. To define what it means for $`\rho `$ to be entangled across the algebras, we need only consider the restriction $`\rho |_{_{AB}}`$ to the von Neumann algebra they generate, $`_{AB}=[_A_B]^{\prime \prime }`$, and of course we need a definition that also applies when $`\rho |_{_{AB}}`$ is mixed. A state $`\omega `$ on $`_{AB}`$ is called a product state just in case there are states $`\omega _A`$ of $`_A`$ and $`\omega _B`$ of $`_B`$ such that $`\omega (XY)=\omega _A(X)\omega _B(Y)`$ for all $`X_A`$, $`Y_B`$. Clearly, product states, or convex combinations of product states, possess only classical correlations. Moreover, if one can even just *approximate* a state with convex combinations of product states, its correlations do not significantly depart from those characteristic of a classical statistical theory. Therefore, we define $`\rho `$ to be entangled across $`(_A,_B)`$ just in case $`\rho |_{_{AB}}`$ is *not* a weak- limit of convex combinations of product states of $`_{AB}`$ (Halvorson and Clifton, 2000). Notice that we chose weak- convergence rather than convergence in norm, hence we obtain a strong notion of entanglement. In the case $`𝖧=𝖧_A𝖧_B`$, $`_A=(𝖧_A)I`$, and $`_B=I(𝖧_B)`$, the definition obviously coincides with the usual notion of entanglement for a pure state (convex combinations and approximations being irrelevant in that case), and also coincides with the definition of entanglement (usually called ‘nonseparability’) for a mixed density operator that is standard in quantum information theory (Werner, 1989; Clifton and Halvorson, 2000; Clifton *et al*, 2000). Further evidence that the definition captures an essentially nonclassical feature of correlations is given by the fact that $`_{AB}`$ will possess an entangled state in the sense defined above if and *only if* both $`_A`$ and $`_B`$ are nonabelian (Bacciagaluppi, 1993, Thm. 7; Summers and Werner, 1995, Lemma 2.1). Returning to AQFT, it is therefore reasonable to say that a global state of the field $`\rho `$ on $`𝒜^{\prime \prime }=(𝖧)`$ is entangled across a pair of spacelike-separated regions $`(O_A,O_B)`$ just in case $`\rho |_{𝒜_{AB}}`$, $`\rho `$’s restriction to $`𝒜_{AB}=[𝒜(O_A)𝒜(O_B)]^{\prime \prime }`$, falls outside the weak- closure of the convex hull of $`𝒜_{AB}`$’s product states.
Our next task is to review the mathematical representation of operations, highlight some subtleties in their physical interpretation, and then discuss what is meant by *local* operations on a system. We then end this section by giving the general argument that local operations performed in either of two spacelike separated regions ($`O_A`$,$`O_B`$) cannot create entanglement in a state across the regions.
The most general transformation of the state of a quantum system with Hilbert space $`𝖧`$ is described by an operation on $`(𝖧)`$, defined to be a positive, weak- continuous, linear map $`T:(𝖧)(𝖧)`$ satisfying $`0T(I)I`$ (Haag and Kastler, 1964; Davies, 1976; Kraus, 1983; Busch *et al*, 1995; Werner 1987). (The weak- topology on a von Neumann algebra $``$ is defined in complete analogy to the weak- topology on its state space, i.e., $`\{Z_n\}`$ weak- converges to $`Z`$ just in case $`\rho (Z_n)\rho (Z)`$ for all states $`\rho `$ of $``$.) Any such $`T`$ induces a map $`\rho \rho ^T`$ from the state space of $`(𝖧)`$ into itself or $`0`$, where, for all $`Z(𝖧)`$,
$$\rho ^T(Z)\rho (T(Z))/\rho (T(I))\text{if}\rho (T(I))0;0\text{otherwise}.$$
(3)
The number $`\rho (T(I))`$ is the probability that an ensemble in state $`\rho `$ will respond ‘Yes’ to the question represented by the positive operator $`T(I)`$. An operation $`T`$ is called selective if $`T(I)<I`$, and nonselective if $`T(I)=I`$. The final state after a selective operation on an ensemble of identically prepared systems is obtained by ignoring those members of the ensemble that fail to respond ‘Yes’ to $`T(I)`$. Thus a selective operation involves performing a physical operation on an ensemble followed by a *purely conceptual* operation in which one makes a selection of a subensemble based on the outcome of the physical operation (assigning ‘state’ $`0`$ to the remainder). Nonselective operations, by contrast, always elicit a ‘Yes’ response from any state, hence the final state is not obtained by selection but purely as a result of the physical interaction between object system and the device that effects the operation. (We shall shortly discuss some actual physical examples to make this general description of operations concrete.)
An operation $`T`$, which quantum information theorists call a superoperator (acting, as it does, on operators to produce operators), “can describe any combination of unitary operations, interactions with an ancillary quantum system or with the environment, quantum measurement, classical communication, and subsequent quantum operations conditioned on measurement results” (Bennett *et al*, 1999). Interestingly, a superoperator itself can always be represented in terms of operators, as a consequence of the Kraus representation theorem (1983, p. 42): for any operation $`T:(𝖧)(𝖧)`$, there exists a (not necessarily unique) countable collection of Kraus operators $`\{K_i\}(𝖧)`$ such that
$$T()=\underset{i}{}K_i^{}()K_i,\text{with}0\underset{i}{}K_i^{}K_iI$$
(4)
where both sums, if infinite, are to be understood as weak- convergence. It is not difficult to show that the sum $`_iK_iK_i^{}`$ must also weak- converge, hence we can let $`T^{}`$ denote the operation conjugate to $`T`$ whose Kraus operators are $`\{K_i^{}\}`$. It then follows (using the linearity and cyclicity of the trace) that if a state $`\rho `$ is represented by a density operator $`D`$ on $`𝖧`$, $`\rho ^T`$ will be represented by the density operator $`T^{}(D)`$. If the mapping $`\rho \rho ^T`$, or equivalently, $`DT^{}(D)`$, maps pure states to pure states, then the operation $`T`$ is called a pure operation, and this corresponds to it being representable by a *single* Kraus operator.
More generally, the Kraus representation shows that a general operation is always equivalent to mixing the results of separating an initial ensemble into subensembles to which one applies pure (possibly selective) operations, represented by the individual Kraus operators. To see this, let $`T`$ be an arbitrary operation performed on a state $`\rho `$, where $`\rho ^T0`$, and suppose $`T`$ is represented by Kraus operators $`\{K_i\}`$. Let $`\rho ^{K_i}`$ denote the result of applying to $`\rho `$ the pure operation given by the mapping $`T_i()=K_i^{}()K_i`$, and (for convenience) define $`\lambda _i=\rho (T_i(I))/\rho (T(I))`$. Then, at least when there are finitely many Kraus operators, it is easy to see that $`T`$ itself maps $`\rho `$ to the convex combination $`\rho ^T=_i\lambda _i\rho ^{K_i}`$. In the infinite case, this sum converges not just weak- but *in norm*, and it is a useful exercise in the topologies we have introduced to see why. Letting $`\rho _n^T`$ denote the partial sum $`_{i=1}^n\lambda _i\rho ^{K_i}`$, we need to establish that
$$\underset{n\mathrm{}}{lim}[sup\{|\rho ^T(Z)\rho _n^T(Z)|:Z=Z^{}(𝖧),|Z|1\}]=0.$$
(5)
For *any* $`Z(𝖧)`$, we have
$$|\rho ^T(Z)\rho _n^T(Z)|=\rho (T(I))^1|\underset{i=n+1}{\overset{\mathrm{}}{}}\rho (K_i^{}ZK_i)|.$$
(6)
However, $`\rho (K_i^{}()K_i)`$, being a positive linear functional, has a norm that may be computed by its action on the identity (KR 1997, Thm. 4.3.2). Therefore, $`|\rho (K_i^{}ZK_i)||Z|\rho (K_i^{}K_i)`$, and we obtain
$$|\rho ^T(Z)\rho _n^T(Z)|\rho (T(I))^1|Z|\underset{i=n+1}{\overset{\mathrm{}}{}}\rho (K_i^{}K_i).$$
(7)
However, since $`_iK_i^{}K_i`$ weak- converges, this last summation is just the tail set of a convergent series. Therefore, when $`|Z|1`$, the right-hand side of (7) goes to zero independently of $`Z`$.
To get a concrete idea of how operations work physically, and to highlight two important interpretational pitfalls, let us again consider our toy universe, with $`𝖧=_A^2_B^2`$ and $`x`$ an entangled state. Recall that Alice disentangled $`x`$ by measuring an $`A`$ observable with eigenprojections $`P_\pm `$. Her measurement corresponds to applying the nonselective operation $`T`$ with Kraus operators $`K_1=P_+I`$ and $`K_2=P_{}I`$, resulting in the final state $`T^{}(P_x)=T(P_x)=D^{}`$, as given in (1). If Alice were to further ‘apply’ the pure selective operation $`T^{}`$ represented by the single Kraus operator $`P_+I`$, the final state of her ensemble, as is apparent from (2), would be the product state $`D^{\prime \prime }=P_+P_{b_x^+}`$. But, as we have emphasized, this corresponds to a conceptual operation in which Alice just throws away all members of the original ensemble that yielded measurement outcome $`1`$. On the other hand, it is essential not lose sight of the issue that troubled Einstein. *Whatever* outcome Alice selects for, she will then be in a position to assert that certain $`B`$ observables — those that have either $`b_x^+`$ or $`b_x^{}`$ as an eigenvector, depending on the outcome she favours — have a sharp value in the ensemble she is left with. But prior to Alice performing the first operation $`T`$, such an assertion would have contradicted the orthodox interpretation of the entangled superposition $`x`$. If, contra Bohr, one were to view this change in $`B`$’s state as a *real physical* change brought about by one of the operations Alice performs, surely the innocuous conceptual operation $`T^{}`$ could not be the culprit — it must have been $`T`$ which forced $`B`$ to ‘choose’ between the alternatives $`b_x^\pm `$. Unfortunately, this clear distinction between the physical operation $`T`$ and conceptual operation $`T^{}`$ is not reflected well in the formalism of operations. For we could equally well have represented Alice’s final product state $`D^{\prime \prime }=P_+P_{b_x^+}`$, not as the result of successively applying the operations $`T`$ and $`T^{}`$, but as the outcome of applying the single composite operation $`T^{}T`$, which is just the mapping $`T^{}`$. And *this* $`T^{}`$ now needs to be understood, not purely as a conceptual operation, but as also involving a physical operation, with possibly real nonlocal effects on $`B`$, depending on one’s view of the EPR paradox. (In particular, keep in mind that you are taking the first step on the road to conceding the incompleteness of quantum theory if you attribute the change in the state of $`B`$ brought about by $`T^{}`$ in this case to a mere change in Alice’s *knowledge* about $`B`$’s state.)
There is a second pitfall that concerns interpreting the result of *mixing* subensembles, as opposed to singling out a particular subensemble. Consider an alternative method available to Alice for disentangling a state $`x`$. For concreteness, let us suppose that $`x`$ is the singlet state $`1/\sqrt{2}(a^+b^{}a^{}b^+)`$. Alice applies the nonselective operation with Kraus representation
$$T()=\frac{1}{2}(\sigma _aI)()(\sigma _aI)+\frac{1}{2}(II)()(II),$$
(8)
where $`\sigma _a`$ is the spin observable with eigenstates $`a^\pm `$. Since $`\sigma _aI`$ maps $`x`$ to the triplet state $`1/\sqrt{2}(a^+b^{}+a^{}b^+)`$, $`T^{}`$ ($`=T`$) will map $`P_x`$ to an equal mixture of the singlet and triplet, which admits the following convex decomposition into product states
$$D^{}=\frac{1}{2}P_{a^+b^{}}+\frac{1}{2}P_{a^{}b^+}.$$
(9)
Has Alice truly disentangled $`A`$ from $`B`$? Technically, Yes. Yet all Alice has done, physically, is to separate the initial $`A`$ ensemble into two subensembles in equal proportion, left the second subensemble alone while performing a (pure, nonselective) unitary operation $`\sigma _aI`$ on the first that maps all its $`A+B`$ pairs to the triplet state, and then remixed the ensembles. Thus, notwithstanding the above decomposition of the final density matrix $`D^{}`$, Alice *knows quite well* that she is in possession of an ensemble of $`A`$ systems each of which is entangled either via the singlet or triplet state with the corresponding $`B`$ systems. This will of course be recognized as one aspect of the problem with the ignorance interpretation of mixtures. We have two different ways to decompose $`D^{}`$ — as an equal mixture of the singlet and triplet or of two product states — but which is the correct way to understand how the ensemble is *actually* constituted? The definition of entanglement is just not sensitive to the answer. (It is exactly this insensitivity that is at the heart of the recent dispute over whether NMR quantum computing is correctly understood as implementing genuine *quantum* computing that cannot be simulated classically (Braunstein *et al*, 1999; Laflamme, 1998).) Nevertheless, we are inclined to think the destruction of the singlet’s entanglement that Alice achieves by applying the operation in (8) is an artifact of her mixing process, in which she is represented as simply forgetting about the history of the $`A`$ systems. And this is the view we shall take when we consider similar possibilities for destroying entanglement between field systems in AQFT.
In the two examples considered above, Alice applies operations whose Kraus operators lie in the subalgebra $`(𝖧_A)I`$ associated with system $`A`$. In the case of a nonselective operation, this is clearly sufficient for her operation not to have any effect on the expectations of the observables of system $`B`$. However, it is also necessary. The point is quite general. Let us define a nonselective operation $`T`$ to be (*pace* Einstein!) local to the subsystem represented by a von Neumann subalgebra $`(𝖧)`$ just in case $`\rho ^T|_{^{}}=\rho |_{^{}}`$ for all states $`\rho `$. Thus, we require that $`T`$ leave the expectations of observables outside of $``$, as well as those in its center $`^{}`$, unchanged. Since distinct states of $`^{}`$ cannot agree on all expectation values, this means $`T`$ must act like the identity operation on $`^{}`$. Now fix an arbitrary element $`Y^{}`$, and suppose $`T`$ is represented by Kraus operators $`\{K_i\}`$. A straightforward calculation reveals that
$$\underset{i}{}[Y,K_i]^{}[Y,K_i]=T(Y^2)T(Y)YYT(Y)+YT(I)Y.$$
(10)
Since $`T(I)=I`$, and $`T`$ leaves the elements of $`^{}`$ fixed, the right-hand side of (10) reduces to zero. Thus each of the terms in the sum on the left-hand side, which are positive operators, must individually be zero. Since $`Y`$ was an arbitrary element of $`^{}`$, it follows that $`\{K_i\}(^{})^{}=`$. So we see that nonselective operations local to $``$ *must* be represented by Kraus operators taken from the subalgebra $``$.
As for selective operations, we have already seen that they *can* ‘change’ the global statistics of a state $`\rho `$ outside the subalgebra $``$, particularly when $`\rho `$ is entangled. However, a natural extension of the definition of local operation on $``$ to a cover the case when $`T`$ is selective is to require that $`T(Y)=T(I)Y`$ for all $`Y^{}`$. This implies $`\rho ^T(Y)=\rho (T(I)Y)/\rho (T(I))`$, and so guarantees that $`T`$ will leave the statistics of any observable in $`^{}`$ the same *modulo* whatever correlations that observable might have had in the initial state with the Yes/No question represented by the positive operator $`T(I)`$. Further motivation is provided by the fact this definition is equivalent to requiring that $`T`$ factor across the algebras $`(,^{})`$, in the sense that $`T(XY)=T(X)Y`$ for all $`X`$, $`Y^{}`$ (Werner, 1987, Lemma). If there exist product states across $`(,^{})`$ (an assumption we shall later see does *not* usually hold when $``$ is a local algebra in AQFT), this guarantees that any local selective operation on $``$, when the global state is an entirely uncorrelated product state, will leave the statistics of that state on $`^{}`$ unchanged. Finally, observe that $`T(Y)=T(I)Y`$ for all $`Y^{}`$ implies that the right-hand side of (10) again reduces to zero. Thus it follows (as before) that selective local operations on $``$ must also be represented by Kraus operators taken from the subalgebra $``$.
Applying these considerations to field theory, any local operation on the field system within a region $`O`$, whether or not the operation is selective, is represented by a family of Kraus operators taken from $`𝒜(O)`$. In particular, each individual element of $`𝒜(O)`$ represents a pure operation that can be performed within $`O`$ (cf. Haag and Kastler, 1964, p. 850). We now need to argue that local operations performed by two experimenters in spacelike separated regions cannot create entanglement in a state across the regions where it had none before. This point, well-known by quantum information theorists working in nonrelativistic quantum theory, in fact applies quite generally to any two commuting von Neumann algebras $`_A`$ and $`_B`$.
Suppose that a state $`\rho `$ is not entangled across $`(_A,_B)`$, local operations $`T_A`$ and $`T_B`$ are applied to $`\rho `$, and the result is nonzero (i.e., some members of the initial ensemble are not discarded). Since the Kraus operators of these operations commute, it is easy to check that $`(\rho ^{T_A})^{T_B}=(\rho ^{T_B})^{T_A}`$, so it does not matter in which order we take the operations. It is sufficient to show that $`\rho ^{T_A}`$ will again be unentangled, for then we can just repeat the same argument to obtain that neither can $`(\rho ^{T_A})^{T_B}`$ be entangled. Next, recall that a general operation $`T_A`$ will just produce a mixture over the results of applying a countable collection of pure operations to $`\rho `$; more precisely, the result will be the norm, and hence weak-, limit of finite convex combinations of the results of applying pure operations to $`\rho `$. If the states that result from $`\rho `$ under those pure operations are themselves not entangled, $`\rho ^{T_A}`$ itself could not be either, because the set of unentangled states is by definition convex and weak- closed. Without loss of generality, then, we may assume that the local operation $`T_A`$ is pure and, hence, given by $`T_A()=K^{}()K`$, for some *single* Kraus operator $`K_A`$. As before, we shall denote the resulting state $`\rho ^{T_A}`$ by $`\rho ^K`$ ($`\rho (K^{}K)/\rho (K^{}K)`$).
Next, suppose that $`\omega `$ is any product state on $`_{AB}`$ with restrictions to $`_A`$ and $`_B`$ given by $`\omega _A`$ and $`\omega _B`$, and such that $`\omega ^K0`$. Then, for any $`X_A`$, $`Y_B`$,
$`\omega ^K(XY)`$ $`=`$ $`{\displaystyle \frac{\omega (K^{}(XY)K)}{\omega (K^{}K)}}`$ (11)
$`=`$ $`{\displaystyle \frac{\omega (K^{}XKY)}{\omega (K^{}K)}}`$ (12)
$`=`$ $`{\displaystyle \frac{\omega _A(K^{}XK)}{\omega _A(K^{}K)}}\omega _B(Y)=\omega _A^K(X)\omega _B(Y).`$ (13)
It follows that $`K`$ maps product states of $`_{AB}`$ to product states (or to zero). Suppose, instead, that $`\omega `$ is a convex combination of states on $`_{AB}`$, $`\omega =_{i=1}^n\lambda _i\omega _i`$. Then, setting $`\lambda _i^K=\omega _i(K^{}K)/\omega (K^{}K)`$, it is easy to see that $`\omega ^K=_{i=1}^n\lambda _i^K\omega _i^K`$, hence $`K`$ preserves convex combinations of states on $`_{AB}`$ as well. It is also not difficult to see that the mapping $`\omega \omega ^K`$ is weak- continuous at any point where $`\omega ^K0`$ (cf. Halvorson and Clifton, 2000, Sec. 3). Returning to our original state $`\rho `$, our hypothesis is that it is not entangled. Thus, there is a net of states $`\{\omega _n\}`$ on $`_{AB}`$, each of which is a convex combination of product states, such that $`\omega _n\rho |_{_{AB}}`$ in the weak- topology. It follows from the above considerations that $`\omega _n^K\rho ^K|_{_{AB}}`$, where each of the states $`\{\omega _n\}`$ is again a convex combination of product states. Hence $`\rho ^K|_{_{AB}}`$ is not entangled either.
## 3. The Operational Implications of the <br>Reeh-Schlieder Theorem
Again, let $`(𝖧)`$ be any von Neumann algebra. A vector $`x𝖧`$ is called cyclic for $``$ if the norm closure of the set $`\{Ax:A\}`$ is the *whole* of $`𝖧`$. In AQFT, the Reeh-Schlieder (RS) theorem connects this formal property of cyclicity to the physical property of a field state having bounded energy. (More generally, the connection is between cyclicity and field states that are ‘analytic’ in the energy. This, together with the physical and mathematical origins of the RS theorem, are analyzed in depth in Halvorson (2000).) A pure global state $`x`$ of the field has bounded energy just in case $`E([0,r])x=x`$ for some $`r<\mathrm{}`$, where $`E`$ is the spectral measure for the global Hamiltonian of the field. In other words, the probability in state $`x`$ that the field’s energy is confined to the bounded interval $`[0,r]`$ is unity. In particular, the vacuum $`\mathrm{\Omega }`$ is an eigenstate of the Hamiltonian with eigenvalue $`0`$, and hence trivially has bounded energy. The RS theorem implies that
*If $`x`$ has bounded energy, then x is cyclic for any local algebra $`𝒜(O)`$.*
Our first order of business is to explain Streater and Wightman’s comment that the RS theorem entails “it is difficult to isolate a system described by fields from outside effects” (1989, p. 139).
A vector $`x`$ is called separating for a von Neumann algebra $``$ if $`Ax=0`$ implies $`A=0`$ whenever $`A`$. It is an elementary result of von Neumann algebra theory that $`x`$ will be cyclic for $``$ if and only if $`x`$ is separating for $`^{}`$ (KR 1997, Cor. 5.5.12). To illustrate with a simple example, take $`𝖧=𝖧_A𝖧_B`$. If $`dim𝖧_Adim𝖧_B`$, then it is possible for there to be vectors $`x𝖧`$ that have a Schmidt decomposition $`_ic_ia_ib_i`$ where $`|c_i|^20`$ for *all* $`i=1`$ to $`dim𝖧_B`$. If we act on such an $`x`$ by an operator in the subalgebra $`I(𝖧_B)`$, of form $`IB`$, then the only way $`(IB)x`$ can be the zero vector is if $`B`$ itself maps all the basis vectors $`\{b_i\}`$ to zero, i.e., $`IB=0`$. Thus such vectors are separating for $`I(𝖧_B)`$, and therefore cyclic for $`(𝖧_A)I`$. Conversely, it is easy to convince oneself that $`(𝖧_A)I`$ possesses a cyclic vector — equivalently, $`I(𝖧_B)`$ has a separating vector — *only if* $`dim𝖧_Adim𝖧_B`$. So, to take another example, each of the $`A`$ and $`B`$ subalgebras will possess a cyclic and a separating vector just in case $`𝖧_A`$ and $`𝖧_B`$ have the same dimension (cf. the proof of Clifton *et al* 1998, Thm. 4).
Consider, now, a local algebra $`𝒜(O)`$ with $`O^{}\mathrm{}`$, and a field state $`x`$ with bounded energy. The RS theorem tells us that $`x`$ is cyclic for $`𝒜(O^{})`$, and therefore, separating for $`𝒜(O^{})^{}`$. But by microcausality, $`𝒜(O)𝒜(O^{})^{}`$, hence $`x`$ must be separating for the subalgebra $`𝒜(O)`$ as well. Thus it is an immediate corollary to the RS theorem that
*If $`x`$ has bounded energy, then x is separating for any local algebra $`𝒜(O)`$ with $`O^{}\mathrm{}`$.*
It is this corollary that prompted Streater and Wightman’s remark. But what has it got to do with thinking of the field system $`𝒜(O)`$ as isolated? For a start, we can now show that the local restriction $`\rho _x|_{𝒜(O)}`$ of a state with bounded energy is always a highly ‘noisy’ mixed state. Recall that a state $`\omega `$ on a von Neumann algebra $``$ is a component of another state $`\rho `$ if there is a third state $`\tau `$ such that $`\rho =\lambda \omega +(1\lambda )\tau `$ with $`\lambda (0,1)`$ (Van Fraassen 1991, p. 161). We are going to show that $`\rho _x|_{𝒜(O)}`$ has a *norm* dense set of components in the state space of $`𝒜(O)`$.
Once again, the point is quite general. Let $``$ be any von Neumann algebra, $`x`$ be separating for $``$, and let $`\omega `$ be an arbitrary state of $``$. We must find a sequence $`\{\omega _n\}`$ of states of $``$ such that each $`\omega _n`$ is a component of $`\rho _x|_{}`$ and $`\omega _n\omega 0`$. Since $``$ has a separating vector, it follows that every state of $``$ is a vector state (KR 1997, Thm 7.2.3). (That this should be so is not as surprising as it sounds. Again, if $`𝖧=𝖧_A𝖧_B`$, and $`dim𝖧_Adim𝖧_B`$, then as we have seen, the B subalgebra possesses a separating vector. But it is also easy to see, in this case, that every state on $`I(𝖧_B)`$ is the reduced density operator obtained from a pure state on $`(𝖧)`$ determined by a vector in $`𝖧`$.) In particular, there is a nonzero vector $`y𝖧`$ such that $`\omega =\omega _y`$. Since $`x`$ is separating for $``$, $`x`$ is cyclic for $`^{}`$, therefore we may choose a sequence of operators $`\{A_n\}^{}`$ so that $`A_nxy`$. Since $`|A_nxy|0`$, $`\omega _{A_nx}\omega _y0`$. We claim now that each $`\omega _{A_nx}`$ is a component of $`\rho _x|_{}`$. Indeed, for any positive element $`B^{}B`$, we have:
$`A_nx,B^{}BA_nx`$ $`=`$ $`x,A_n^{}A_nB^{}Bx=Bx,A_n^{}A_nBx`$ (14)
$``$ $`|A_n^{}A_n|Bx,Bx=|A_n|^2x,B^{}Bx.`$ (15)
Thus,
$`\omega _{A_nx}(B^{}B)`$ $`=`$ $`{\displaystyle \frac{A_nx,B^{}BA_nx}{|A_nx|^2}}{\displaystyle \frac{|A_n|^2}{|A_nx|^2}}\rho _x(B^{}B).`$ (16)
If we now take $`\lambda =|A_nx|^2/|A_n|^2(0,1)`$, and consider the linear functional $`\tau `$ on $``$ given by $`\tau =(1\lambda )^1(\rho _x|_{}\lambda \omega _{A_nx})`$, then (16) implies that $`\tau `$ is a state (in particular, positive), and we see that $`\rho _x|_{}=\lambda \omega _{A_nx}+(1\lambda )\tau `$ as required. (This result also holds more generally for states $`\rho `$ of $``$ that are faithful, i.e., $`\rho (Z)=0`$ entails $`Z=0`$ for any positive $`Z`$; see the first part of the proof of Summers and Werner, 1988, Thm. 2.1.)
So bounded energy states are, locally, highly mixed. And such states are far from special — they lie norm dense in the pure state space of $`(𝖧)`$. To see this, just recall that it is part of the spectral theorem for the global Hamiltonian that $`E([0,n])`$ converges strongly to the identity as $`n\mathrm{}`$. Thus we may approximate any vector $`y𝖧`$ by the sequence of bounded energy states $`\{E([0,n])y/|E([0,n])y|\}_{n=0}^{\mathrm{}}`$. Since there are so many bounded energy states of the field, that are locally so ‘noisy’, Streater and Wightman’s comment is entirely warranted. But somewhat more can be said. As we saw with our toy example in Section 1, when a local subsystem of a global system in a pure state is itself in a mixed state, this is a sign of that subsystem’s entanglement with its environment. And there is entanglement lurking in bounded energy states too. But, first, we need to take a closer look at the operational implications of local cyclicity.
If a vector $`x`$ is cyclic for $``$, then for any $`y`$, there is a sequence $`A_n`$ such that $`A_nxy`$. Thus for any $`ϵ>0`$ there is an $`A`$ such that $`\rho _{Ax}\rho _y<ϵ`$. However, $`\rho _{Ax}`$ is just the state one gets by applying the pure operation given by the Kraus operator $`K=A/|A|`$ to $`\rho _x`$. It follows that if $`x`$ is cyclic for $``$, one can get arbitrarily close in norm to any other pure state of $`(H)`$ by applying an appropriate pure local operation in $``$ to $`\rho _x`$. In particular, pure operations on the vacuum $`\mathrm{\Omega }`$ within a local region $`O`$, no matter how small, can prepare essentially any global state of the field. As Haag emphasizes, to do this the operation must “judiciously exploit the small but nonvanishing long distance correlations which exist in the vacuum” (1992, p. 102). This, as Redhead (1995) has argued by analogy to the singlet state, is made possible by the fact that the vacuum is highly entangled (cf. Clifton *et al* 1998). But the first puzzle we need to sort out is that it looks as though entirely *physical* operations in $`O`$ can change the global state, in particular the vacuum $`\mathrm{\Omega }`$, to any desired state! (For example, Segal and Goodman (1965) have called this “bizarre” and “physically quite surprising”, sentiments echoed recently by Fleming who calls it “amazing!” (1999).)
Redhead’s analysis of the cyclicity of the singlet state $`1/\sqrt{2}(a^+b^{}a^{}b^+)`$ for the subalgebra $`(_A^2)I`$ is designed to remove this puzzle (*ibid*, p. 128). (Note that in this simple $`2\times 2`$-dimensional case, he could equally well have chosen *any* entangled state, since they are all separating for $`I(_B^2)`$.) Redhead writes:
> “…we want to distinguish clearly two senses of the term “operation”. Firstly there are physical operations such as making measurements, selecting subensembles according to the outcome of measurements, and mixing ensembles with probabilistic weights, and secondly there are the mathematical operations of producing superpositions of states by taking linear combinations of pure states produced by appropriate selective measurement procedures. These superpositions are of course quite different from the mixed states whose preparation we have listed as a physical operation” (1995, pp. 128-9).
Note that, in stark contrast to our discussion in the previous section, Redhead counts selecting subensembles and mixing as physical operations; it is only the operation of superposition that warrants the adjective ‘mathematical’. When he explains why it is possible that $`x`$ can be cyclic, Redhead first notes (*ibid*, p. 129) that the four basis states
$$a^+b^{},a^{}b^{},a^{}b^+,a^+b^{},$$
(17)
are easily obtained by the physical operations of applying projections and unitary transformations to the singlet state, and exploiting the fact that the singlet strictly correlates $`\sigma _a`$ with $`\sigma _b`$. He goes on:
> “But *any* state for the joint system is some linear combination of these four states, so by the *mathematical* operation of linear combination, we can see how to generate an arbitrary state in $`𝖧_1𝖧_2`$ from physical operations performed on particle one. But all the operations we have described can be represented in the algebra of operators on $`𝖧_1`$ (extended to $`𝖧_1𝖧_2`$)” (*ibid*, p. 129).
Now, while Redhead’s explanation of why it is mathematically possible for $`x`$ to be cyclic is perfectly correct, he actually misses the mark when it comes to the physical interpretation of cyclicity. The point is that superposition *of states* is a red-herring. Certainly a superposition of the states in (17) could not be prepared by physical operations confined to the $`A`$ system. But, as Redhead himself notes in the final sentence above, one can get the same *effect* as superposing those states by acting on $`x`$ with an operator of form $`AI`$ in the subalgebra $`(_A^2)I`$ — an operator that is itself a ‘superposition’ of other operators in that algebra. What Redhead fails to point out is that the action of this operator on $`x`$ *does have a local physical interpretation*: as we have seen, it is a Kraus operator that represents the outcome of a generalized positive operator valued measurement on the $`A`$ system. The key to the puzzle is, rather, that this positive operator valued measurement will generally have to be *selective*. For one certainly could never, with nonselective operations on $`A`$ alone, get as close as one likes to any state vector in $`_A^2_B^2`$ (otherwise all state vectors would induce the same state on $`I(_B^2)`$!). We conclude that the correct way to view the physical content of cyclicity is that changes in the global state are partly due to an experimenter’s ability to perform a generalized measurement on $`A`$, and partly due (*pace* Redhead) to the purely conceptual operation of selecting a subensemble based on the outcome of the experimenter’s measurement together with the consequent ‘change’ in the state of $`B`$ via the EPR correlations between $`A`$ and $`B`$.
One encounters the same interpretational pitfall concerning the cyclicity of the vacuum in relation to localized states in AQFT. A global state of the field is said to be localized in $`O`$ if its expectations on the algebra $`𝒜(O^{})`$ agree with vacuum expectation values (Haag, 1992, p. 102). Thus localized states are ‘excitations’ of the vacuum confined to $`O`$. In particular, $`U\mathrm{\Omega }`$ will be a localized state whenever $`U`$ is a unitary operator taken from $`𝒜(O)`$ (since unitary operations are nonselective). But every element of a $`C^{}`$-algebra is a finite linear combination of unitary operators (KR 1997, Thm. 4.1.7). Since $`\mathrm{\Omega }`$ is cyclic for $`𝒜(O)`$, this means we must be able to approximate any global state by linear superpositions of vectors describing states localized in $`O`$ — even approximate states that are localized in regions spacelike separated with $`O`$! Haag, rightly cautious, calls this a “(superficial) paradox” (1992, p. 254), but he fails to put his finger on its resolution: while unitary operations are nonselective, a local operation in $`𝒜(O)`$ given by a Kraus operator that is a linear combination of local unitary operators will generally be *selective*. (Haag *does* make the interesting point out that only a proper subset of the state space of a field can be approximated if we restrict ourselves to local operations that involve a physically reasonable expenditure of energy. But we do not share the view of Schroer (1999) that this point by itself reconciles the RS theorem with ‘common sense’.)
The (common) point of the previous two paragraphs is perhaps best summarized as follows. Both Redhead and Haag would agree that unitary Kraus operators in $`𝒜(O)`$ give rise to purely physical operations in the local region $`O`$. But there are many Kraus operators in $`𝒜(O)`$ that do not represent purely physical operations in $`O`$ insofar as they are selective. Since every Kraus operator is a linear superposition of unitary operators, it follows that “superposition of local operations” does not preserve (pure) physicality. Redhead is right that the key to diffusing the paradox is in noting that superpositions are involved — but it is essential to understand these superpositions as occurring locally in $`𝒜(O)`$, not in the Hilbert space.
Our next order of business is to supply the rigorous argument behind Redhead’s intuition about the connection between cyclicity and entanglement. The point, again, is quite general: for any two commuting nonabelian von Neumann algebras $`_A`$ and $`_B`$, and any state vector $`x`$ cyclic for $`_A`$ (or $`_B`$), $`\rho _x`$ will be entangled across the algebras (Halvorson and Clifton, 2000, Prop. 2). For suppose, in order to extract a contradiction, that $`\rho _x`$ is *not* entangled. Then as we have seen, operations on $`\rho _x`$ that are local to $`_A`$ cannot turn that state into an entangled state across $`(_A,_B)`$. Yet, by the cyclicity of $`x`$, we know that we can apply pure operations to $`\rho _x`$, that are local to $`_A`$ (or $`_B`$), and approximate in norm (and hence weak- approximate) any other vector state of $`_{AB}`$. It follows that no vector state of $`_{AB}`$ could be entangled across $`(_A,_B)`$, and the same goes for all its mixed states, which lie in the the norm closed convex hull of the vector states. But this means that $`_{AB}`$ would possess *no* entangled states at all — in flat contradiction with the fact that neither $`_A`$ nor $`_B`$ is abelian.
Returning to the context of AQFT, if we now consider *any* two spacelike separated field systems, $`𝒜(O_A)`$ and $`𝒜(O_B)`$, then the argument we just gave establishes that the dense set of field states bounded in the energy will *all* be entangled across the regions $`(O_A,O_B)`$. (Note that the fact that $`𝒜(O_A)`$ and $`𝒜(O_B)`$ are nonabelian is *itself* a consequence of the RS theorem. For if, say, $`𝒜(O_A)`$ were abelian, then since by the RS theorem that algebra possesses a cyclic vector, it must be a maximal abelian subalgebra of $`(𝖧)`$ (KR 1997, Cor. 7.2.16). The same conclusion would have to follow for any subregion $`\stackrel{~}{O}_AO_A`$ whose closure is a proper subset of $`O_A`$. And this, by isotony, would lead to the absurd conclusion that $`𝒜(\stackrel{~}{O}_A)=𝒜(O_A)`$, which is easily shown to be inconsistent with the axioms of AQFT (Horuzy, 1988, Lemma 1.3.10).) However, by itself this result does not imply that Alice cannot destroy a bounded energy state $`x`$’s entanglement across $`(O_A,O_B)`$ by performing local operations in $`O_A`$. In fact, Borchers (1965, Cor. 7) has shown that any vector state of form $`Ax`$ for any nontrivial $`A𝒜(O_A)`$ *never* has bounded energy (nor is ‘analytic’ in the energy). So it seems that all Alice needs to do is perform any pure operation within $`O_A`$ and the resulting state, because it is no longer subject to the RS theorem, need no longer be entangled across $`(O_A,O_B)`$.
However, the RS theorem gives only a sufficient, *not* a necessary, condition for a state $`x`$ of the field to be cyclic for $`𝒜(O_A)`$. And notwithstanding that no pure operation Alice performs can preserve boundedness in the energy, *almost all* the pure operations she could perform *will* preserve the state’s cyclicity! The reason is, once again, quite general. Again let $`_A`$ and $`_B`$ be two commuting nonabelian von Neumann algebras, suppose $`x`$ is cyclic for $`_A`$, and consider the state induced by the vector $`Ax`$ where $`A_A`$. Now every element in a von Neumann algebra is the strong limit of invertible elements in the algebra (Dixmier and Maréchal, 1971, Prop. 1). Therefore, there is a sequence of invertible operators $`\{\stackrel{~}{A}_n\}_A`$ such that $`\stackrel{~}{A}_nxAx`$, i.e., $`\rho _{\stackrel{~}{A}_nx}\rho _{Ax}0`$. Notice, however, that since each $`\stackrel{~}{A}_n`$ is invertible, each vector $`\stackrel{~}{A}_nx`$ is again cyclic for $`_A`$, because we can ‘cycle back’ to $`x`$ by applying to $`\stackrel{~}{A}_nx`$ the inverse operator $`\stackrel{~}{A}_n^1_A`$, and from there we know, by hypothesis, that we can cycle with elements of $`_A`$ arbitrarily close to any other vector in $`𝖧`$. It follows that, even though Alice may *think* she has applied the pure operation given by some Kraus operator $`A/|A|`$ to $`x`$, she could well have *actually* applied an invertible Kraus operation given by one of the operators $`\stackrel{~}{A}_n/|\stackrel{~}{A}_n|`$ in a strong neighborhood of $`A/|A|`$. And if she actually did this, then she certainly would *not* disentangle $`x`$, because she would not have succeeded in destroying the *cyclicity* of the field state for her local algebra. We could, of course, give Alice the freedom to employ more general mixing operations in $`O_A`$. But as we saw in the last section, it is far from clear whether a mixing operation should count as a successful disentanglement when all the states that are mixed by her operation are themselves entangled — or at least not *known* by Alice to be disentangled (given her practical inability to specify exactly which Kraus operations go into the pure operations of her mixing process).
Besides this, there is a more fundamental practical limitation facing Alice, even if we allow her any local operation she chooses. If, as we have seen, we can approximate the result of acting on $`x`$ with any given operator in von Neumann algebra $``$ by acting on $`x`$ with an invertible operator that preserves $`x`$’s cyclicity, then the set of all such ‘invertible actions’ on $`x`$ must itself produce a dense set of vector states, given that $`\{Ax:A\}`$ is dense. It follows that if a von Neumann algebra possesses even just one cyclic vector, it must possess a dense set of them (Dixmier and Maréchal, 1971, Lemma 4; cf. Clifton *et al* 1998). Now consider, again, the general situation of two commuting nonabelian algebras $`_A`$ and $`_B`$, where either algebra possesses a cyclic vector, and hence a dense set of such. If, in addition, the algebra $`_{AB}`$ possesses a separating vector, then *all* states of that algebra will be vector states, a *norm* dense set of which must therefore be entangled across $`(_A,_B)`$. And since the entangled states of $`_{AB}`$ are open in the weak- topology, they must be open in the (stronger) norm topology too — so we are dealing with a truly generic set of states. It follows, quite independently of the RS theorem, that
*Generic Result: If $`_A`$ and $`_B`$ are commuting nonabelian von Neumann algebras either of which possesses a cyclic vector, and $`_{AB}`$ possesses a separating vector, then the generic state of $`_{AB}`$ will be entangled across $`(_A,_B)`$.*
The role that the RS theorem plays is to guarantee that the antecedent conditions of this Generic Result are satisfied whenever we consider spacelike separated regions (and corresponding algebras) satisfying $`(O_AO_B)^{}\mathrm{}`$. This is a very weak requirement, which is satisfied, for example, when we assume both regions are bounded in spacetime. In that case, in order to be *certain* that her local operation in $`O_A`$ (pure or mixed) produced a disentangled state, Alice would need the extraordinary ability to distinguish the state of $`𝒜_{AB}`$ which results from her operation from the generic set states of $`𝒜_{AB}`$ that are entangled!
Finally, while we noted in our introduction the irony that limitations on disentanglement arise precisely when one considers *relativistic* quantum theory, the practical limitations we have just identified — as opposed to the *intrinsic* limits on disentanglement which are the subject of the next section — are not characteristic of AQFT alone. In particular, the existence of locally cyclic states does not depend on field theory. As we have seen, both the $`A`$ and $`B`$ subalgebras of $`(𝖧_A𝖧_B)`$ possess a cyclic vector just in case $`dim𝖧_A=dim𝖧_B`$. Indeed, operator algebraists so often find themselves dealing with von Neumann algebras that, together with their commutants, possess a cyclic vector, that such algebras are said by them to be in ‘standard form’. So we should not think that local cyclicity is somehow peculiar to the states of local quantum fields.
Neither is it the case that our Generic Result above finds its only application in quantum *field* theory. For example, consider the infinite-by-infinite state space $`𝖧_A𝖧_B`$ of any two nonrelativistic particles, ignoring their spin degrees of freedom. Take the tensor product with a third auxiliary infinite-dimensional Hilbert space $`𝖧_A𝖧_B𝖧_C`$. Then obviously $`\mathrm{}=dim𝖧_Cdim(𝖧_A𝖧_B)=\mathrm{}`$, whence the $`C`$ subalgebra possesses a cyclic vector, which is therefore separating for the $`A+B`$ algebra. On the same dimensional grounds, both the $`A`$ and $`B`$ subalgebras possess cyclic vectors of their own. So our Generic Result applies immediately yielding the conclusion that a typical state of $`A+B`$ will be entangled (cf. Clifton and Halvorson, 2000).
Nor should we think of local cyclicity or the applicability of our Generic Result as peculiar to standard *local* quantum field theory. After noting that the local cyclicity of the vacuum in AQFT was a “great, counterintuitive, surprise” (p. 4) when it was first proved, Fleming (1999) proposes, instead, to build up local algebras associated with bounded open spatial sets within hyperplanes from raising and lowering operators associated with nonlocal Newton-Wigner position eigenstates — a proposal that goes back at least as far as Segal (1964). Fleming then observes, as did Segal (1964, p. 143), that the resulting vacuum state will *not* be entangled nor cyclic for any such local algebra. Nevertheless, as Segal points out, each Segal-Fleming local algebra will be isomorphic to the algebra $`(𝖧)`$ of all bounded operators on an *infinite*-dimensional Hilbert space $`𝖧`$, and algebras associated with spacelike-separated regions in the same hyperplane commute. It follows that if we take any two spacelike separated bounded open regions $`O_A`$ and $`O_B`$ lying in the same hyperplane, $`[𝒜(O_A)𝒜(O_B)]^{\prime \prime }`$ will be naturally isomorphic to $`(𝖧_A)(𝖧_B)`$ (Horuzhy 1988, Lemma 1.3.28), and the result of the previous paragraph applies. So Fleming’s ‘victory’ over the RS theorem of standard local quantum field theory rings hollow. Even though the Newton-Wigner vacuum is not itself entangled or locally cyclic across the regions $`(O_A,O_B)`$, it will be indistinguishable from globally pure states of the Newton-Wigner field that are! (For further discussion of the Segal-Fleming approach to quantum fields, see Halvorson (2000).)
On the other hand, generic entanglement is certainly not to be expected in every quantum-theoretic context. For example, if we ignore external degrees of freedom, and just consider the spins of two particles with joint state space $`𝖧_A𝖧_B`$, where both spaces are nontrivial and *finite*-dimensional, then the Generic Result no longer applies. Taking the product with a third auxiliary Hilbert space $`H_C`$ does not work, because in order for the $`A+B`$ subalgebra to have a separating vector we would need $`dimH_CdimH_AdimH_B`$, but for either the $`A`$ or $`B`$ subalgebras to possess a cyclic vector we would *also* need that either $`dimH_AdimH_BdimH_C`$ or $`dimH_BdimH_AdimH_C`$ — both of which contradict the fact $`𝖧_A`$ and $`𝖧_B`$ are nontrivial and finite-dimensional. (In fact, it can be shown that the spins of any pair of particles are *not* generically entangled, unless of course we ignore their mixed spin states; see Clifton and Halvorson, 2000 for further discussion.) The point is that while the conditions for generic entanglement may or may not obtain in *any* quantum-theoretical context — depending on the observables and dimensions of the state spaces involved — the beauty of the RS theorem is that it allows us to deduce that generic entanglement between bounded open spacetime regions *must* obtain just by making some very general and natural assumptions about what should count as a physically reasonable relativistic quantum field theory.
## 4. Type III von Neumann Algebras and <br>Intrinsic Entanglement
Though it is not known to follow from the general axioms of AQFT (cf. Kadison, 1963), all known concrete models of the axioms are such that the local algebras associated with bounded open regions in $`M`$ are type III factors (Horuzy, 1988, pgs. 29, 35; Haag, 1992, Sec. V.6). We start by reviewing what precisely is meant by the designation ‘type III factor’.
A von Neumann algebra $``$ is a factor just in case its center $`^{}`$ consists only of multiples of the identity. It is easy to verify that this is equivalent to $`(^{})^{\prime \prime }=(𝖧)`$, thus $``$ induces a ‘factorization’ of the total Hilbert space algebra $`(𝖧)`$ into two subalgebras which together generate that algebra.
To understand what ‘type III’ means, a few further definitions need to be absorbed. A partial isometry $`V`$ is an operator on a Hilbert space $`𝖧`$ that maps some particular closed subspace $`C𝖧`$ isometrically onto another closed subspace $`C^{}𝖧`$, and maps $`C^{}`$ to zero. (Think of $`V`$ as a ‘hybrid’ unitary/projection operator.) Given the set of projections in a von Neumann algebra $``$, we can define the following equivalence relation on this set: $`PQ`$ just in case there is a partial isometry $`V`$ that maps the range of $`P`$ onto the range of $`Q`$. (It is important to notice that this definition of equivalence is relative to the particular von Neumann algebra $``$ that the projections are considered to be members of.) For example, any two infinite-dimensional projections in $`(𝖧)`$ are equivalent (when $`𝖧`$ is separable), including projections one of whose range is properly contained in the other (cf. KR 1997, Cor. 6.3.5). A nonzero projection $`P`$ is called abelian if the von Neumann algebra $`PP`$ acting on the subspace $`P𝖧`$ (with identity $`P`$) is abelian. One can show that the abelian projections in a factor $``$ are exactly the atoms in its projection lattice (KR 1997, Prop. 6.4.2). For example, the atoms of the projection lattice of $`(𝖧)`$ are all its one-dimensional projections, and they are all (trivially) abelian, whereas it is clear that higher-dimensional projections are not. Finally, a projection $`P`$ is called infinite (relative to $``$!) when it is equivalent to another projection $`Q`$ such that $`Q<P`$, i.e., Q projects onto a proper subspace of the range of $`P`$. One can also show that any abelian projection in a von Neumann algebra must be *finite*, i.e., not infinite (KR 1997, Prop. 6.4.2).
A type I von Neumann factor is now defined as one that possesses an abelian projection. For example, $`(𝖧)`$ for any Hilbert space $`𝖧`$ is always type I, and, indeed, every type I factor arises as the algebra of all bounded operators on some Hilbert space (KR 1997, Thm 6.6.1). On the other hand, a factor is type III if all its nonzero projections are infinite and equivalent. In particular, this entails that the algebra itself is not abelian, nor could it even possess an abelian projection — which would have to be finite. And since a type III factor contains no abelian projections, its projection lattice cannot have any atoms. Another fact about type III algebras is that they *always* possess a vector that is both cyclic and separating (Sakai, 1971, Cor. 2.9.28). Therefore we know that type III algebras will always possess a dense set of cyclic vectors, and that all their states will be vector states. *Notwithstanding this*, type III algebras possess *no* pure states, as a consequence of the fact that they lack atoms.
To get some feeling for why this is the case — and for the general connection between the failure of the projection lattice of an algebra to possess atoms and its failure to possess pure states — let $``$ be any non-atomic von Neumann algebra possessing a separating vector (so all of its states are vector states), and let $`\rho _x`$ be any state of $``$. We shall need two further definitions. The support projection, $`S_x`$, of $`\rho _x`$ in $``$ is defined to be the meet of all projections $`P`$ such that $`\rho _x(P)=1`$. (So $`S_x`$ is the smallest projection in $``$ that $`\rho _x`$ ‘makes true’.) The left-ideal, $`I_x`$, of $`\rho _x`$ in $``$ is defined to be the set of all $`A`$ such that $`\rho _x(A^{}A)=0`$. Now since $`S_x`$ is not an atom, there is some nonzero $`P`$ such that $`P<S_x`$. Choose any vector $`y`$ in the range of $`P`$ (noting it follows that $`S_yP`$). We shall first show that $`I_x`$ is a proper subset of $`I_y`$. So let $`AI_x`$. Clearly this is equivalent to saying that $`Ax=0`$, or that $`x`$ lies in the range of $`N(A)`$, the projection onto the null-space of $`A`$. $`N(A)`$ itself lies $``$ (KR 1997, Lemma 5.1.5 and Prop. 2.5.13), thus, $`\rho _x(N(A))=1`$, and accordingly $`S_xN(A)`$. But since $`S_yP<S_x`$, we also have $`\rho _y(N(A))=1`$. Thus, $`y`$ too lies in the range of $`N(A)`$, i.e., $`Ay=0`$, and therefore $`AI_y`$. To see that the inclusion $`I_xI_y`$ is proper, note that since $`(y,S_yy)=1`$, $`(y,[IS_y]^2y)=0`$, and thus $`IS_yI_y`$. However, certainly $`IS_yI_x`$, for the contrary would entail that $`(x,S_yx)=1`$, in other words, $`S_xS_yP<S_x`$ — a contradiction. We can now see, finally, that $`\rho _x`$ cannot be pure. For, quite generally, the pure states of a von Neumann algebra $``$ determine *maximal* left-ideals in $``$ (KR 1997, Thm. 10.2.10), yet we have just shown, under the assumption that $``$ is non-atomic, that $`I_xI_y`$.
The fact that every state of a type III algebra $``$ is mixed throws an entirely new wrench into the works of the ignorance interpretation of mixtures. (To our knowledge, Van Aken (1985) is the only philosopher of quantum theory to have noticed this.) Not only is there no preferred way to pick out components of a mixture, but the components of states of $``$ will always *themselves* be mixtures. Thus, it is impossible to understand the preparation of such a mixture in terms of mixing pure states — the states of $``$ are always irreducibly or *intrinsically* mixed. Note, however, that while the states of type III factors fit this description, so do the states of certain *abelian* von Neumann algebras. For example, the ‘multiplication’ algebra $`(L_2())`$ of all bounded functions of the position operator for a single particle lacks atomic projections because position has no eigenvectors. Moreover, all the states of $``$ are vector states, because any state vector that corresponds to a wavefunction whose support is the whole of $``$ is separating for $``$. Thus the previous paragraph’s argument applies equally well to $``$.
Of course no properly *quantum* system has an abelian algebra of observables, and, as we have already noted, systems with abelian algebras are never entangled with other systems. This makes the failure of a type III factor $``$ to have pure states importantly different from that failure in the case of an abelian algebra. Because $``$ is *non*abelian, and taking the commutant preserves type (KR 1997, Thm. 9.1.3) so that $`^{}`$ will also be nonabelian, one suspects that any pure state of $`(^{})^{\prime \prime }=(𝖧)`$ — which must restrict to an intrinsically mixed state on both subalgebras $``$ and $`^{}`$ — has to be *intrinsically entangled* across $`(,^{})`$. And that intuition is exactly right; indeed, one can show that there are not even any *product* states across $`(,^{})`$ (Summers 1990, p. 213). And, of course, if there are no unentangled states across $`(,^{})`$, then the infamous distinction, some have argued is important to preserve, between so-called ‘improper’ mixtures that arise by restricting an entangled state to a subsystem, and ‘proper’ mixtures that do not, becomes *irrelevant*.
Even more interesting is the fact that in all known models of AQFT, the local algebras are ‘type III<sub>1</sub>’. It would take us too far afield to explain the standard sub-classification of factors presupposed by the subscript ‘$`1`$’. We wish only to draw attention to an equivalent characterization of type III<sub>1</sub> algebras established by Connes and Størmer (1978, Cor. 6): A factor $``$ acting standardly on a (separable) Hilbert space is type III<sub>1</sub> just in case for *any two* states $`\rho ,\omega `$ of $`(𝖧)`$, and any $`ϵ>0`$, there are unitary operators $`U`$, $`U^{}^{}`$ such that $`\rho \omega ^{UU^{}}<ϵ`$. Notice that this result immediately implies that there can be no unentangled states across $`(,^{})`$; for, if some $`\omega `$ were not entangled, it would be impossible to act on this state with local unitary operations in $``$ and $`^{}`$ and get arbitrarily close to the states that *are* entangled across $`(,^{})`$. Furthermore — and this is the interesting fact — the Connes-Størmer characterization immediately implies the impossibility of distinguishing in any reasonable way between the different degrees of entanglement that states might have across $`(,^{})`$. For it is a standard assumption in quantum information theory that all reasonable measures of entanglement must be *invariant* under unitary operations on the separate entangled systems (cf. Vedral *et al*, 1997), and presumably such a measure should assign close degrees of entanglement to states that are close to each other in norm. In light of the Connes-Størmer characterization, imposition of both these requirements forces triviality on any proposed measure of entanglement across $`(,^{})`$. Of course, the standard von Neumann entropy measure we discussed in Section 1. is norm continuous, and, because of the unitary invariance of the trace, this measure is invariant under unitary operations on the component systems. But in the case of a type III factor $``$, that measure, as we should expect, is *not* available. Indeed, the state of a system described by $``$ cannot be represented by any density operator *in* $``$ because $``$ cannot contain compact operators, like density operators, whose spectral projections are all finite!
The above considerations have particularly strong physical implications when we consider local algebras associated with diamond regions in $`M`$, i.e., regions given by the intersection of the timelike future of a given spacetime point $`p`$ with the timelike past of another point in $`p`$’s future. When $`\mathrm{}M`$ is a diamond, it can be shown in many models of AQFT, including for *non*interacting fields, that $`𝒜(\mathrm{}^{})=𝒜(\mathrm{})^{}`$ (Haag 1992, Sec. III.4.2). Thus every global state of the field will be intrinsically entangled across $`(𝒜(\mathrm{}),𝒜(\mathrm{}^{}))`$, and it is never possible to think of the field system in a diamond region $`\mathrm{}`$ as disentangled from that of its spacelike complement. Though he does not use the language of entanglement, this is precisely the reason for Haag’s remark that field systems are always open. In particular, Alice would have *no hope whatsoever* of using local operations in $`\mathrm{}`$ to disentangle that region’s state from that of the rest of the world.
Suppose, however, that Alice has only the more limited goal of disentangling a state of the field across some isolated pair of *strictly* spacelike-separated regions $`(O_A,O_B)`$, i.e., regions which remain spacelike separated when either is displaced by an arbitrarily small amount. It is also known that in many models of AQFT the local algebras possess the split property: for any bounded open $`OM`$, and any larger region $`\stackrel{~}{O}`$ whose interior contains the closure of $`O`$, there is a type I factor $`𝒩`$ such that $`𝒜(O)𝒩𝒜(\stackrel{~}{O})`$ (Bucholz 1974, Werner 1987). This implies that the von Neumann algebra generated by a pair of algebras for strictly spacelike separated regions is isomorphic to their tensor product and, as a consequence, that there *are* product states across $`(𝒜(O_A),𝒜(O_B))`$ (cf. Summers 1990, pgs. 239-40). Since, therefore, not every state of $`𝒜_{AB}`$ is entangled, we might hope that whatever the global field state is, Alice could *at least in principle* perform an operation in $`O_A`$ on that state that disentangles it across $`(O_A,O_B)`$. However, we are now going use the fact that $`𝒜(O_A)`$ lacks abelian projections to show that a norm dense set of entangled states of $`𝒜_{AB}`$ cannot be disentangled by any pure local operation performed in $`𝒜(O_A)`$.
Let $`\rho _x`$ be any one of the norm dense set of entangled states of $`𝒜_{AB}`$ induced by a vector $`x𝖧`$ cyclic for $`𝒜(O_B)`$, and let $`K𝒜(O_A)`$ be an arbitrary Kraus operator. (Observe that $`\rho _x^K0`$ because $`x`$ is separating for $`𝒜(O_B)^{}`$ — which includes $`𝒜(O_A)`$ — and $`K^{}K𝒜(O_A)`$ is positive.) Suppose, for the purposes of extracting a contradiction, that $`\omega _x^K`$ is not also entangled. Let $`Ky`$, with $`y𝖧`$, be any nonzero vector in the range of $`K`$. Then, since $`x`$ is cyclic for $`𝒜(O_B)`$, we have, for some sequence $`\{B_i\}𝒜(O_B)`$, $`Ky=K(limB_ix)=lim(B_iKx)`$, which entails $`(\omega _x^K)^{B_i/|B_i|}\omega _{Ky}0`$. Since $`\omega _x^K`$ is not entangled across $`(𝒜(O_A),𝒜(O_B))`$, and the local pure operations on $`𝒜(O_B)`$ given by the Kraus operators $`B_i/|B_i|`$ cannot create entanglement, we see that $`\omega _{Ky}`$ is the norm (hence weak-) limit of a sequence of unentangled states and, as such, is not itself entangled either. Since $`y`$ was arbitrary, it follows that every nonzero vector in the range of $`K`$ induces an unentangled state across $`(𝒜(O_A),𝒜(O_B))`$. Obviously, the same conclusion follows for any nonzero vector in the range of $`R(K)`$ — the range projection of $`K`$ — since the range of the latter lies dense in that of the former.
Next, consider the von Neumann algebra
$$𝒞_{AB}[R(K)𝒜(O_A)R(K)R(K)𝒜(O_B)R(K)]^{\prime \prime }$$
(18)
acting on the Hilbert space $`R(K)𝖧`$. Since $`K𝒜(O_A)`$, $`R(K)𝒜(O_A)`$ (KR 1997, p. 309), and thus the subalgebra $`R(K)𝒜(O_A)R(K)`$ cannot be abelian — on pain of contradicting the fact that $`𝒜(O_A)`$ has no abelian projections. And neither is $`R(K)𝒜(O_B)R(K)`$ abelian. For since $`𝒜(O_B)`$ itself is nonabelian, there are $`Y_1,Y_2𝒜(O_B)`$ such that $`[Y_1,Y_2]0`$. And because our regions $`(O_A,O_B)`$ are strictly spacelike separated, they have the Schlieder property: $`0A𝒜(O_A),0B𝒜(O_B)`$ implies $`AB0`$ (Summers 1990, Thm. 6.7). Therefore,
$$[R(K)Y_1R(K),R(K)Y_2R(K)]=[Y_1,Y_2]R(K)0.$$
(19)
So we see that neither algebra occurring in $`𝒞_{AB}`$ is abelian; yet they commute, and so there must be at least one entangled state across those algebras. But this conflicts with the conclusion of the preceding paragraph! For the vector states of $`𝒞_{AB}`$ are precisely those induced by the vectors in the range of $`R(K)`$, and we deduced that these all induce unentangled states across $`(𝒜(O_A),𝒜(O_B))`$. Therefore, by restriction, they all induce unentangled states across the algebra $`𝒞_{AB}`$. But if none of $`𝒞_{AB}`$’s vector states are entangled, it can possess *no* entangled states at all.
The above argument still goes through under the weaker assumption that Alice applies any mixed *projective* operation, i.e., any operation $`T`$ corresponding to a standard von Neumann measurement associated with a mutually orthogonal set $`\{P_i\}𝒜(O_A)`$ of projection operators. For if we suppose, again for reductio, that $`\rho _x^T=_i\lambda _i\rho _x^{P_i}`$ is not entangled across the regions, then since entanglement cannot be created by a further application to $`\rho _x^T`$ of the local projective operation given by (say) $`T_1()=P_1()P_1`$, it follows that $`(\rho _x^T)^{T_1}=(\rho _x^{T_1T})=\rho _x^{P_1}`$ must again be unentangled, and the above reasoning to a contradiction goes through *mutatis mutandis* with $`K=P_1`$. This is to be contrasted to the nonrelativistic case we considered in Section 1, where Alice *was* able to disentangle an arbitrary state of $`(H_AH_B)`$ by a nonselective projective operation on $`A`$. And a moment’s reflection will reveal that that was possible precisely because of the availability of abelian projections in the algebra of her subsystem $`A`$. We have not, of course, shown that the above argument covers *arbitrary* mixing operations Alice might perform in $`O_A`$; in particular, positive-operator valued mixings, where the Kraus operators $`\{K_i\}`$ of a local operation $`T`$ in $`O_A`$ do not have mutually orthogonal ranges. However, although it would be interesting to know how far the result could be pushed, we have already expressed our reservations about whether arbitrary mixing operations should count as disentangling when none of the pure operations of which they are composed could possibly produce disentanglement on their own.
In summary:
*There are many regions of spacetime within which no local operations can be performed that will disentangle that region’s state from that of its spacelike complement, and within which no pure or projective operation on any one of a norm dense set of states can yield disentanglement from the state of any other strictly spacelike-separated region.*
Clearly the advantage of the formalism of AQFT is that it allows us to see clearly just how much more deeply entrenched entanglement is in *relativistic* quantum theory. At the very least, this should serve as a strong note of caution to those who would quickly assert that quantum nonlocality cannot peacefully exist with relativity.
As far as what becomes of Einsteinian worries about the possibility of doing science in such a deeply entangled world, the split property of local algebras comes to the rescue. For let us suppose Alice knows nothing more than that she wants to prepare some state $`\rho `$ on $`𝒜(O_A)`$ for subsequent testing. (The following argument is simply an amplification of the reasoning in Werner 1987 and Summers 1990, Thm. 3.13.) Since there is a type I factor $`𝒩`$ satisfying $`𝒜(O_A)𝒩𝒜(\stackrel{~}{O}_A)`$ for any super-region $`\stackrel{~}{O}_A`$, and $`\rho `$ is a vector state (when we assume $`(O_A)^{}\mathrm{}`$), its vector representative defines a state on $`𝒩`$ that extends $`\rho `$ and is, therefore, represented by some density operator $`D_\rho `$ in the type I algebra $`𝒩`$. Now $`D_\rho `$ is an infinite convex combination $`_i\lambda _iP_i`$ of mutually orthogonal atomic projections in $`𝒩`$ satisfying $`_iP_i=I`$ with $`_i\lambda _i=1`$. But each such projection is equivalent, *in the type* III *algebra* $`𝒜(\stackrel{~}{O}_A)`$, to the identity operator. Thus, for each $`i`$, there is a partial isometry $`V_i𝒜(\stackrel{~}{O}_A)`$ satisfying $`V_iV_i^{}=P_i`$ and $`V_i^{}V_i=I`$. Next, consider the nonselective operation $`T`$ on $`𝒜(\stackrel{~}{O}_A)`$ given by Kraus operators $`K_i=\sqrt{\lambda _i}V_i`$, and fix an arbitrary $`X𝒜(O_A)`$. We claim that $`T(X)=\rho (X)I`$. Indeed, because each $`P_i`$ is abelian in $`𝒩𝒜(O_A)`$, the operator $`P_iXP_i`$ acting on $`P_i𝖧`$ can only be some multiple, $`c_i`$, of the identity operator $`P_i`$ on $`P_i𝖧`$, and taking the trace of both sides of the equation
$$P_iXP_i=c_iP_i$$
(20)
immediately reveals that $`c_i=\text{Tr}(P_iX)`$. Moreover, acting on the left of (20) with $`V_i^{}`$ and on the right with $`V_i`$, we obtain $`V_i^{}XV_i=\text{Tr}(P_iX)I`$, which yields the desired conclusion when multiplied by $`\lambda _i`$ and summed over $`i`$. Finally, since $`T(X)=\rho (X)I`$ for all $`X𝒜(O_A)`$, obviously $`\omega ^T=\rho `$ for all initial states $`\omega `$ of $`𝒜(O_A)`$. Thus, once we allow Alice to perform an operation like $`T`$ that is *approximately* local to $`𝒜(O_A)`$ (choosing $`\stackrel{~}{O}_A`$ to approximate $`O_A`$ as close as we like), she has the freedom to prepare any state of $`𝒜(O_A)`$ that she pleases.
Notice that, ironically, testing the theory is actually *easier* here than in nonrelativistic quantum theory! For we were able to exploit above the type III character of $`𝒜(\stackrel{~}{O}_A)`$ to show that Alice can always prepare her desired state on $`𝒜(O_A)`$ *nonselectively*, i.e., without ever having to sacrifice any members of her ensemble! Also observe that the result of her preparing operation $`T`$, because it is local to $`𝒜(\stackrel{~}{O}_A)`$, will always produce a product state across $`(O_A,O_B)`$ when $`O_B(\stackrel{~}{O}_A)^{}`$. That is, for any initial state $`\omega `$ across the regions, and all $`X𝒜(O_A)`$ and $`Y𝒜(O_B)`$, we have
$$\omega ^T(XY)=\omega (T(X)Y)=\omega (\rho (X)Y)=\rho (X)\omega (Y).$$
(21)
So as soon as we allow Alice to perform *approximately* local operations on her field system, she *can* isolate it from entanglement with other strictly spacelike-separated field systems, while simultaneously preparing its state as she likes and with relative ease. God is subtle, but not malicious.
*Acknowledgments*—The authors are grateful to Paul Busch for helpful discussions, Jeremy Butterfield for helping us to clarify our critique of Redhead’s discussion of the operational implications of cyclicity, and Reinhard Werner for filling in for us the argument of Eqn. (10). R. K. C. wishes to thank All Souls College, Oxford for support under a Visiting Fellowship.
References
Bacciagaluppi, G. (1993), ‘Separation Theorems and Bell Inequalities in Algebraic Quantum Mechanics’, in P. Busch, P. Lahti, and P. Mittelstaedt (eds), *Symposium on the Foundations of Modern Physics 1993* (Singapore: World Scientific) pp. 29–37.
Bennett, C. H., DiVincenzo, D. P., Fuchs, C. A., Mor, T., Rains, E., Schor, P. W., Smolin, J. A., and Wooters, W. K. (1999), ‘Quantum Nonlocality without Entanglement’, *Physical Review A* 59, 1070–1091.
Borchers, H. J. (1965), ‘On the Vacuum State in Quantum Field Theory. II.’, *Communications in Mathematical Physics* 1, 57–79.
Braunstein, S. L., Caves, C. M., Jozsa, R., Linden, N., Popescu, S., and Schack, R. (1999), ‘Separability of Very Noisy Mixed States and Implications for NMR Quantum Computing’, *Physical Review Letters* 83, 1054–1057.
Bucholz, D. (1974), ‘Product States for Local Algebras’, *Communications in Mathematical Physics* 36, 287–304.
Busch, P., Grabowski, M., and Lahti, P. J. (1995), *Operational Quantum Physics* (Berlin: Springer).
Clifton, R. and Halvorson, H. (2000), ‘Bipartite mixed states of infinite-dimensional systems are generically nonseparable’, *Physical Review A* 61, 012108.
Clifton, R., Feldman, D. V., Halvorson, H., Redhead, M. L. G., and Wilce, A. (1998), ‘Superentangled States’, *Physical Review A* 58, 135–145.
Clifton, R., Halvorson, H., and Kent, A. (2000), ‘Non-local correlations are generic in infinite-dimensional bipartite systems’, forthcoming in *Physical Review A*, April issue.
Connes, A. and Størmer, E. (1978), ‘Homogeneity of the State Space of Factors of Type III<sub>1</sub>’, *Journal of Functional Analysis* 28, 187-196.
Dixmier, J. and Maréchal, O. (1971), ‘Vecteurs totalisateurs d’une algèbre de von Neumann’, *Communications in Mathematical Physics* 22, 44–50.
Davies, E. B. (1976), *Quantum Theory of Open Systems* (London: Academic Press).
Einstein, A. (1948), ‘Quantenmechanik und Wirklichkeit’, *Dialectica* 2, 320–324.
Fleming, G. (1999), ‘Reeh-Schlieder Meets Newton-Wigner’, forthcoming in *PSA 1998 Vol. II* (East Lansing, MI: Philosophy of Science Association).
Haag, R. (1992), *Local Quantum Physics*, 2nd edition (New York: Springer).
Haag, R. and Kastler, D. (1964), ‘An Algebraic Approach to Quantum Field Theory’, *Journal of Mathematical Physics* 5, 848–861.
Halvorson, H. (2000), ‘Does Relativity imply Nonlocality? Unpacking the Reeh-Schlieder Theorem’, forthcoming.
Halvorson, H. and Clifton. R. (2000), ‘Generic Bell Correlation between Arbitrary Local Algebras in Quantum Field Theory’, forthcoming in *Journal of Mathematical Physics* April issue.
Horuzhy, S. S. (1988), *Introduction to Algebraic Quantum Field Theory* (Dordrecht: Kluwer).
Howard, D. (1989), ‘Holism, Separability, and the Metaphysical Implications of the Bell Experiments’, in J. T. Cushing and E. McMullin (eds.), *The Philosophical Consequences of Bell’s Theorem* (Notre Dame: Notre Dame University Press) pp. 224–253.
Kadison, R. (1963), ‘Remarks on the type of von Neumann algebras of local observables in quantum field theory’, *Journal of Mathematical Physics* 4, 1511–1516.
Kadison, R. and Ringrose, J. (1997), *Fundamentals of the Theory of Operator Algebras* (Providence, R. I.: American Mathematical Society).
Kraus, K. (1983), *States, Effects, and Operations* (Berlin: Springer).
Laflamme, R. (1998), Review of ‘Separability of Very Noisy Mixed States and Implications for NMR Quantum Computing’ by Braunstein *et al* (1998), in *Quick Reviews in Quantum Computation and Information*, http://quantum-computing.lanl.gov/qcreviews/qc/.
Lüders, G. (1951), ‘Über die Zustandsänderung durch den Messprozess’, *Annalen der Physik* 8, 322–328.
Mor, T. (1998), ‘On the Disentanglement of States’, quant-ph/9812020.
Mor, T. and Terno, D. R. (1999), ‘Sufficient Conditions for Disentanglement’, quant-ph/9907036.
Popescu, S. and Rohrlich, D. (1997), ‘Thermodynamics and the Measure of Entanglement’, *Physical Review A* 56, R3319–R3321.
Redhead, M. L. G. (1995), ‘More Ado about Nothing’, *Foundations of Physics* 25, 123–137.
Reeh, H. and Schlieder, S. (1961), ‘Bemerkungen zur unitaraquivalenz von Lorentzinvarianten Feldern’, *Nuovo Cimento* 22, 1051–1068.
Sakai, S. (1971), *$`C^{}`$-Algebras and $`W^{}`$-algebras* (Berlin: Springer-Verlag).
Schroer, B. (1998), *A Course on: ‘Modular Localization and Nonperturbative Local Quantum Physics’*, hep-th/9805093.
Schroer, B. (1999), ‘Basic Quantum Theory and Measurement from the Viewpoint of Local Quantum Physics’, quant-ph/9904072.
Segal, I. E. (1964), ‘Quantum Fields and Analysis in the Solution Manifolds of Differential Equations’, in W. T. Martin and I. E. Segal (eds), *Proceedings of a Conference on the Theory and Applications of Analysis in Function Space* (Boston: MIT Press) Ch. 8.
Segal, I. E. and Goodman, R. W. (1965), ‘Anti-Locality of Certain Lorentz-Invariant Operators’, *Journal of Mathematics and Mechanics* 14, 629–638.
Streater, R. F. and Wightman, A. S. (1989), *PCT, Spin and Statistics, and All That* (New York: Addison-Wesley).
Summers, S. J. (1990), ‘On the independence of local algebras in quantum field theory’, *Reviews of Mathematical Physics* 2, 201–247.
Summers, S. J. and Werner, R. F. (1988), ‘Maximal violation of Bell’s inequalities for algebras of observables in tangent spacetime regions’, Annales Institut Henri Poincaré 49, 215–243.
Summers, S. J., and Werner, R. F. (1995), ‘On Bell’s inequalities and algebraic invariants’, *Letters in Mathematical Physics* 33, 321–334.
Van Aken, J. (1985), ‘Analysis of Quantum Probability Theory’, *Journal of Philosophical Logic* 14, 267–296.
Van Fraassen, B. C. (1991), *Quantum Mechanics: An Empiricist View* (Oxford: Clarendon Press).
Vedral, V., Plenio, M. B., Rippin, M. A., and Knight, P. L. (1997), ‘Quantifying Entanglement’, *Physical Review Letters* 78, 2275–2279.
Werner, R. F. (1987), ‘Local Preparability of States and the Split Property’, *Letters in Mathematical Physics* 13, 325–329.
Werner, R. F. (1989), ‘Quantum States with Einstein-Podolsky-Rosen correlations admitting a hidden-variable model’, *Physical Review A* 40, 4277–4281. |
warning/0001/nlin0001011.html | ar5iv | text | # Lyapunov Instability for a hard-disk fluid in equilibrium and nonequilibrium thermostated by deterministic scattering
## I Introduction
Transport of energy and momentum is a central problem in nonequilibrium statistical mechanics, but so far most of our knowledge is confined to the macroscopic level. There is still a long way to go when it comes to understanding this phenomena in microscopic terms although significant progress has been made during the last years, with help from dynamical systems theory and computer simulations. In general, external forces are needed to drive a system out of equilibrium, but in order to prepare a nonequilibrium steady state the redundant energy has to be removed thus preventing the system from heating up indefinitely. One way out is the introduction of thermostating mechanisms . Both stochastic and deterministic/time-reversible thermostats are in use, the latter having been introduced to remain close to Hamiltonian dynamics.
Recently, an alternative thermostating mechanism acting via deterministic time-reversible boundary-scattering has been applied on a hard disk fluid to model heat and shear flow nonequilibrium steady states . The calculated transport coefficients have been found to be in agreement with the theoretical values obtained from kinetic theory, but only for special cases and in the thermodynamic limit the conjectured identity between exponential phase-space contraction and entropy production rate holds. In the present paper we investigate further the dynamical properties of this system by computing the full Lyapunov spectra and related quantities like the Kaplan-Yorke dimension or the Kolmogorov-Sinai entropy. In Sec. II we briefly recapitulate the model and its thermostating mechanism and Sec. III serves to outline the method used for computing the Lyapunov exponents . The results are presented in Sec. IV and conclusions are drawn in Sec. V.
## II Model
Consider a two-dimensional system of hard disks confined in a square box of length $`L`$ with periodic boundary conditions along the x-axis, i.e., the left and right sides at $`x=\pm L/2`$ are identified. The $`N`$ disks interact among themselves via elastic hard collisions, thus the bulk dynamics is purely conservative. In the following and in all the numerical computations we use reduced units by setting the particle mass $`m`$, the disk diameter $`\sigma `$ and the Boltzmann constant $`k_B`$ equal to one. Now, denote with $`p_x^i`$, $`p_y^i`$ and with $`p_x^f`$, $`p_y^f`$ the tangential and normal momentum of a disk before and after a collision with the wall. Then the scattering prescription is given as
$$(p_x^f,p_y^f)=\{\begin{array}{cc}𝓣\text{ }^1𝓣\text{ }(p_x^i,p_y^i),\hfill & p_x^i0\hfill \\ \mathrm{\Pi }𝓣\text{ }^1^1𝓣\text{ }(p_x^i,p_y^i),\hfill & p_x^i<0,\hfill \end{array}$$
(1)
where $`𝓣\text{ }:[0,\mathrm{})\times [0,\mathrm{})[0,1]\times [0,1]`$ is the invertible map
$$(\zeta ,\xi )=𝓣\text{ }(p_x,p_y)=(\text{erf}\left(|p_x|/\sqrt{2T}\right),\mathrm{exp}\left(p_y^2/2T\right))$$
(2)
and $`:[0,1]\times [0,1][0,1]\times [0,1]`$ is a two-dimensional, invertible, phase-space conserving chaotic map to be specified later. $`\mathrm{\Pi }(p_x,p_y)=(p_x,p_y)`$ only serves to produce the right sign for the backward scattering. The parameter $`T`$ plays the role of a temperature . Note that the colliding disk retains its tangential direction and that the scattering is reversible by construction.
So far, the model has only been defined in equilibrium. In order to drive the system into a nonequilibrium steady state (NSS) the collision rule only has to be modified appropriately, which will be done in Sec. IV B (see also ).
## III Method
Like all hard disk systems our model is chaotic in the sense that two nearby phase space trajectories diverge exponentially with time. This is mainly due to the dispersing action of the hard disk collisions in the bulk, but especially for small particle numbers we also expect a contribution to this divergence due the chaotic nature of our scattering mechanism. The average logarithmic divergence rate in phase space are described by the so-called Lyapunov exponents $`\lambda _l`$. Denote by $`𝚪=\{𝐪_1,𝐪_2,\mathrm{},𝐪_N,𝐩_1,𝐩_2,\mathrm{},𝐩_N\}`$ the 4N dimensional phase space vector for $`N`$ disks. Then, the time evolution
$$𝚪(0)=\mathrm{\Phi }^t[𝚪(0)]$$
(3)
of an initial state $`𝚪(0)`$ consists of a smooth streaming which is interrupted by particle-particle and particle-wall collisions. Next, consider a satellite trajectory $`𝚪_s(t)`$ initially displaced from the reference trajectory by an infinitesimal vector $`\delta 𝚪(0)`$. In a chaotic system $`|\delta 𝚪(0)|`$ is growing on average exponentially, thus rendering the system unpredictable for long times. Then there exists a complete set of linear-independent initial vectors $`\{\delta 𝚪_l(0):l=1,\mathrm{},4N\}`$ and Lyapunov exponents defined as
$$\lambda _l=\underset{t\mathrm{}}{lim}\frac{1}{t}\mathrm{ln}\frac{|\delta 𝚪_l(t)|}{|\delta 𝚪_l(0)|}.$$
(4)
The $`\lambda _l`$, which we order according to $`\lambda _1\lambda _2\mathrm{}\lambda _{4N}`$, are independent of the coordinate system and the metric. The whole set of Lyapunov exponents is referred to as the Lyapunov spectrum.
In Hamiltonian systems the Lyapunov exponents appear in pairs summing up to zero, $`\lambda _i+\lambda _{4Ni+1}=0`$ for $`i=1,\mathrm{},2N`$, due to the symplectic nature of the equations of motion. In a continuous dynamical system one Lyapunov exponent associated with the direction of the phase flow vanishes. Moreover, each conserved quantity leads to an additional vanishing Lyapunov exponent. The symmetry found in symplectic dynamical systems is lost when the system is driven to a nonequilibrium stationary state. However, for homogeneous driving the symmetry is replaced by the so-called conjugate pairing rule saying that after excluding the vanishing exponents associated with the flow direction and the conservation of energy the remaining pairs, i.e. $`\{\lambda _1,\lambda _{4N}\}`$, $`\{\lambda _2,\lambda _{4N1}\}`$, and so on, each sum up to the same negative value $`C`$. For inhomogeneously driven systems such as ours or the Chernov-Lebowitz shear flow model , however, the symmetry is lost and no pairing rules exist.
As can be seen from Eqs.(1,2) the phase space volume is in general changed during each disk-wall collision. In equilibrium this averages up to zero, whereas in NSS the average phase space contraction rate is negative and is given by the sum of all Lyapunov exponents. Consequently, the phase volume shrinks continuously in NSS and and the phase-space distribution collapses onto a multifractal strange attractor. The fractal dimension of this strange attractor can be estimated with the conjecture of Kaplan-Yorke ,
$$D_{KY}=n+\frac{_{l=1}^n\lambda _l}{|\lambda _{n+1}|}$$
(5)
where $`n`$ is the largest integer for which $`_{l=1}^n\lambda _l0`$. $`D_{KY}`$ is the dimension of a phase space object which neither shrinks nor grows and for which the natural measure is conserved by the flow.
For the calculation of the full Lyapunov spectrum we use a method worked out by Dellago et al. which is actually a generalization of the algorithm of Benettin et al. for smooth dynamical systems. The latter follows the time evolution of a reference trajectory and of a complement set of tangent vectors by solving the original and the linearized equations of motion, respectively. Periodic reorthonormalization prevents the tangent vectors from collapsing all into the direction of fastest growth. Averaging the logarithmic expansion and contraction rates of the tangent vectors then yields the Lyapunov exponents. For a hard disk system the free streaming is interrupted by impulsive collisions, either with another particle or the boundary. This certainly affects both the trajectory and the tangent space and has to be included in the calculation. The free streaming and the particle-particle collisions in the bulk have been treated in Section III-D of reference , where the same notation was used as in the present work. We refer to Eqs. (39)-(42) and Eqs. (68)-(73) of this article for explicit expressions of the particle-particle collision rules in phase space and tangent space, respectively. It remains to consider the particle-wall collisions and the following lines are formulated in parallel to the respective treatment of Dellago and Posch for the Chernov-Lebowitz model . In fact, the only difference lies in the ’scattering matrix’ and its derivatives.
If particle $`k`$ collides with the walls its position remains unchanged whereas its momentum is changed according to the scattering rules Eq. (1). The collision map $`𝚪^f=𝐌(𝚪^i)`$ in phase space becomes
$`𝐪_j^f`$ $`=`$ $`𝐪_j^i\text{for}j=1,\mathrm{},N`$ (6)
$`𝐩_j^f`$ $`=`$ $`𝐩_j^i\text{for}jk`$ (7)
$`𝐩_k^f`$ $`=`$ $`𝓒\text{ }^+(𝐩_k^i),\text{for}p_x^i0`$ (8)
$`𝐩_k^f`$ $`=`$ $`𝓒\text{ }^{}(𝐩_k^i),\text{for}p_x^i<0`$ (9)
using the abbreviations $`𝓒^\mathbf{+}=𝓣\text{ }^1𝓣\text{ }`$ for the forward scattering and $`𝓒^{\mathbf{}}=\mathrm{\Pi }𝓣\text{ }^1^1𝓣\text{ }`$for the backward scattering (Eq. (1)).
In order to obtain the corresponding transformation for the tangent space vector $`\delta 𝚪`$ at a particle-wall collision we assume that the collision takes place at phase point $`𝚪`$ at time $`\tau _c`$. Then the satellite trajectory, displaced by the infinitesimal vector $`\delta 𝚪`$, collides at a different phase point $`𝚪+\delta 𝚪_c`$ at a different time $`\tau _c+\delta \tau _c`$. A linear approximation in phase space and time yields
$$\delta 𝚪^f=\frac{𝐌}{𝚪}\delta 𝚪^i+\left[\frac{𝐌}{𝚪}𝐅(𝚪^i)𝐅(𝐌(𝚪^i))\right]\delta \tau _c$$
(10)
where $`𝐅`$ is the right hand side of the equation of motion during the free streaming , and $`𝐌/𝚪`$ is the matrix of the derivatives of the full collision map with respect to the phase-space coordinates. Obviously, the delay time $`\delta \tau _c`$ is a function of the phase point $`𝚪^i`$ and of the tangent vector $`\delta 𝚪^i`$. For a disk-wall collision of the $`k`$th particle the delay time $`\delta \tau _c`$ is given by
$$\delta \tau _c=\frac{(\delta 𝐪_k𝐧)}{(𝐩_k/m𝐧)}.$$
(11)
Here, $`𝐧`$ is the normal vector of the wall pointing into the simulation box. Since the scattering rules Eq. (1) for the momentum components is independent of the position of the particle, the matrix $`𝐌/𝚪`$ has the form
$$\frac{𝐌}{𝚪}=\left(\begin{array}{cc}\mathrm{𝟏}& \mathrm{𝟎}\\ \mathrm{𝟎}& \frac{𝓒\text{ }^\pm (𝐩^i)}{(𝐩^i)}\end{array}\right)$$
(12)
where $`\mathrm{𝟏}`$ and $`\mathrm{𝟎}`$ are the $`2N\times 2N`$ unit and zero matrices, respectively. $`𝓒\text{ }^\pm (𝐩^i)/(𝐩^i)`$ is the matrix of the derivatives of the outgoing momenta with respect to the incoming momenta and only the components of the colliding particle $`k`$ are different from zero. From Eq. (10) the following transformation rules for the tangent vectors can be deduced:
$`\delta 𝐪_j^f`$ $`=`$ $`\delta 𝐪_j^i\text{for}jk`$ (13)
$`\delta 𝐩_j^f`$ $`=`$ $`\delta 𝐩_j^i\text{for}jk`$ (14)
$`\delta 𝐪_k^f`$ $`=`$ $`\delta 𝐪_k^i(𝐩_k^f𝐩_k^i)\delta \tau _c`$ (15)
$`\delta 𝐩_k^f`$ $`=`$ $`{\displaystyle \frac{𝓒\text{ }^\pm (𝐩_k^i)}{(𝐩_k^i)}}\delta 𝐩_k^i.`$ (16)
Omitting for notational convenience the index $`k`$ indicating the colliding particle we obtain from Eq. (1) the following expressions for the $`2\times 2`$ matrix $`(𝓒\text{ }^\pm (𝐩_k^i)/(𝐩_k^i))_{\alpha \beta }=p_\alpha ^f/p_\beta ^i`$, $`\alpha ,\beta \{x,y\}`$:
$`{\displaystyle \frac{p_x^f}{p_x^i}}=\left(𝐃\right)_{11}\mathrm{exp}\left[((p_x^f)^2(p_x^i)^2)/(2T)\right],`$ $`{\displaystyle \frac{p_x^f}{p_y^i}}=\left(𝐃\right)_{12}{\displaystyle \frac{\pi p_y^i}{\sqrt{2T}}}\mathrm{exp}\left[((p_x^f)^2(p_y^i)^2)/(2T)\right]`$ (17)
$`{\displaystyle \frac{p_y^f}{p_x^i}}=\left(𝐃\right)_{21}{\displaystyle \frac{\sqrt{2T}}{\pi p_y^f}}\mathrm{exp}\left[((p_y^f)^2(p_x^i)^2)/(2T)\right],`$ $`{\displaystyle \frac{p_y^f}{p_y^i}}=\left(𝐃\right)_{22}{\displaystyle \frac{p_y^i}{p_y^f}}\mathrm{exp}\left[((p_y^f)^2(p_y^i)^2)/(2T)\right],`$ (18)
$`\text{for}p_x^i0.`$ (19)
Here, $`𝐃`$ denotes the matrix of the derivatives of the chaotic map $``$. Eqs. (17,18) are stated for positive tangential velocities, for negative tangential velocities $``$ has only to be replaced by $`^1`$, see Eq. (1).
Combining the free streaming with the transformation for the disk-disk and the disk-wall collisions, one is now able to follow the exact time evolution of the trajectory and of the tangent-space vector.
## IV Results
Using the algorithm outlined in the previous section we are now able to calculate the full Lyapunov spectrum for our hard disk model with deterministic scattering at the boundary. As already mentioned we use reduced units by setting the particle mass $`m`$, the disk diameter $`\sigma `$ and the Boltzmann constant $`k_B`$ equal to unity. We define the number density by $`\overline{n}=N/L^2`$. For simulation we use a collision-to-collision approach and neighbor lists . For an initial configuration the centers of the disks are positioned on a triangular lattice and the momenta are chosen from a Gaussian with zero mean. The total momentum is then set to zero and the momenta are rescaled to obtain the total kinetic energy $`E_{kin}=N(T_u+T_d)/2`$, $`T_u`$ and $`T_d`$ being the imposed ’parametrical’ temperatures of the upper and the lower wall, respectively.
### A Equilibrium
We now set both wall temperatures $`T_u`$, $`T_d`$ equal to one and compute the full Lyapunov spectra for a four-particle system at number density $`\overline{n}=0.2`$ using three different chaotic maps:
$`_B(\zeta ,\xi )`$ $`=`$ $`(k\zeta ,\xi /k)\text{modulo }1,\text{(baker map)}`$ (20)
$`_C(\zeta ,\xi )`$ $`=`$ $`((k+1)\zeta +\xi ,k\zeta +\xi )\text{modulo }1,\text{(cat map)}`$ (21)
and
$$_S:\{\begin{array}{c}\xi ^{}=\xi \frac{k}{2\pi }\mathrm{sin}(2\pi \zeta ),\hfill \\ \zeta ^{}=\zeta +\xi ^{},\hfill \end{array}\text{modulo }1,\text{(standard map)}$$
(22)
with $`0\zeta ,\xi 1`$. $`k2\text{}`$ is a parameter controlling the chaoticity of the map, i.e. the magnitude of the Lyapunov exponents.
The resulting spectra are shown in Fig. 1 where we have also plotted the Lyapunov spectrum for elastic reflection as reference. To emphasize the conjugate pairs, the Lyapunov exponents are ordered as $`\{\lambda _{2Ni+1},\lambda _{2N+i}\}`$, with $`i=1,\mathrm{},2N`$. Errors are estimated as in from the convergence of the exponents as a function of simulation time such that the time-dependent exponents did not deviate more than $`\pm \mathrm{\Delta }\lambda `$ from their mean values during the second half of the simulation run. For high accuracy more than $`10^7`$ disk-disk collisions and more than $`510^6`$ disk-wall were simulated yielding errors less than $`\pm 0.001`$ for the exponents and less than $`\pm 0.002`$ for the pair sums. In the case of elastic reflection three Lyapunov exponents vanish. One exponent vanishes due to the neutral expansion behavior in the direction of the flow, a second due to the conservation of kinetic energy. The third exponents is zero due to the translational invariance of the system in the $`x`$-direction . The fourth vanishing exponent then verifies the pairing rule. For a hard disk system thermostated by deterministic scattering only two Lyapunov exponents vanish. The kinetic energy is now allowed to fluctuate around a mean value, so only the neutral expansion and the translational invariance remain. As we expect the maximum Lyapunov exponent increases with increasing chaoticity of the map, i.e. when going from a baker map ($`k=2`$) to a cat map ($`k=2`$) to a standard map ($`k=100`$) (Results not plotted here show a similar behavior when $`k`$ is increased for a given map.). The pairing rule for these models is satisfied with an error of $`\pm 0.002`$ in the pair sums. At this point we add a remark which might seem at first purely technical. For all simulations we used a symmetrical configuration, i.e. Eq. (1) is used for the upper wall whereas $``$ and $`^1`$ are interchanged in Eq. (1) for the lower wall. Using the same scattering rules for both walls results in an asymmetry and eventually in an asymmetric Lyapunov spectrum even in equilibrium violating the pairing rule.
### B NSS
We move on to the nonequilibrium stationary state and turn first to the case of an imposed temperature gradient by the walls. Since in the thermodynamic limit the Lyapunov spectrum is mainly determined by the bulk behavior we use in the following only a cat map with $`k=2`$ as chaotic map $``$. In order to determine the macroscopic state of the system the velocity and density profiles of the bulk are measured as well as the temperatures and velocities of the walls. Wall velocities are defined as the mean tangential velocity of the incoming and outgoing particles. Wall temperatures are defined as mean temperature of the incoming and outgoing fluxes, $`\overline{T}=\left(\overline{T}_i+\overline{T}_o\right)/2`$, with $`\overline{T}_{i/o}=\left((v_xv_x_x)^2_x+\left[v_y\right]_y/\left[v_y^1\right]_y\right)/2`$ where $`_x`$ and $`[]_y`$ represent an average over the density $`\rho (v_x)`$ and the flux $`\mathrm{\Phi }`$ to and from the wall, respectively (see also ).
#### 1 Heat flow
Again, we first investigate a small system with four particles at $`\overline{n}=0.2`$ with high accuracy. In order to impose a temperature difference on the system we only have to choose two different parametrical temperatures $`T_u`$, $`T_d`$ (see Eq.(2) and ). Note that this also affects the derivatives of the collision matrix (Eqs. (17, 18)). Figure 2 shows the spectra for this system under a temperature gradient, the numbers denoting the parametrical temperatures $`T_{u/d}`$ of the upper and the lower wall. In NSS the sum of the Lyapunov exponents is negative and exactly equal to the phase space contraction rate. $`T_uT_d=13`$ results in $`\lambda _l=1.029`$ and $`T_uT_d=15`$ in $`\lambda _l=2.703`$. We find again two vanishing Lyapunov exponents but with increasing temperature difference all nonzero exponents also increase in magnitude, the negative ones certainly stronger to yield an overall negative sum. The pair sums are also shown and, as we expect, the driving shifts the sums towards negative values thus destroying the symmetry. The deviations from the pairing rule are particularly strong for pairs with large $`i`$. Figure 3 shows the results for a $`36`$-particle system under the same setting. At least $`210^6`$ disk-disk collisions and $`210^5`$ disk-wall collisions have been simulated in each run. The spectra are plotted as connected lines only for graphical reasons, it is understood that the exponents are defined for integer $`i`$ only. The change in the Lyapunov spectrum under thermal driving are similar to the four-particle system. Increasing the density from $`\overline{n}=0.2`$ to $`\overline{n}=0.6`$ results in a larger magnitude of all nonzero exponents due to the higher collision rate. The Kaplan-Yorke dimension$`D_{KY}`$ \[Fig. 4\] is decreasing for increasing temperature gradient, with a larger dimensionality loss $`\mathrm{\Delta }D_{KY}`$ for higher densities than for lower densities at given $`\mathrm{\Delta }\overline{T}`$. As can immediately be guessed from the positive branch of the spectra thermal driving also results in an increasing Kolmogorov-Sinai entropy $`h_{KS}`$ \[Fig. 4\], defined as the sum over all positive exponents,
$$h_{KS}=\underset{\{\lambda _l>0\}}{}\lambda _l,$$
(23)
with increasing temperature gradient. So, as we expect from thermodynamics, thermal driving reduces the ordering of the system. Higher collision rates at higher densities lead to more viscous heating in the bulk and eventually to an increasing disorder ($`h_{KS}`$) of the system. The second, lower data point at $`\mathrm{\Delta }T=0`$ shows $`h_{KS}/N`$ for elastic reflection as reference.
#### 2 Shear flow
One way to model moving walls is to add some tangential momentum $`d`$ to $`p_x`$ before and after the collision of a particle with the boundary (model I in ),
$$(p_x^f,p_y^f)=\{\begin{array}{cc}𝒮_d𝓒^\mathbf{+}𝒮_d(p_x^i,p_y^i),\hfill & p_x^id\hfill \\ 𝒮_d𝓒^{\mathbf{}}𝒮_d(p_x^i,p_y^i),\hfill & p_x^i<d,\hfill \end{array}$$
(24)
with
$$𝒮_d(p_x^i,p_y^i)=(p_x^i+d,p_y^i).$$
(25)
In order to impose shear the shift $`d`$ has only to be chosen with different signs for the upper and the lower wall. Note that these scattering rules are time reversible. Certainly, the drift also affects the derivatives of the collision matrix, Eqs. (17,18), where $`p_x^f`$ goes to $`p_x^fd`$ and $`p_x^i`$ to $`p_x^i+d`$. Figure 5 shows the full Lyapunov spectra and the pair sums for a 36-particle system at $`\overline{n}=0.6`$ under shear while keeping $`T_u=T_d=1`$ fixed. The negative exponents increase in magnitude whereas the positive branch changes very little. Before we take a closer look at the Kaplan-Yorke dimension and the Kolmogorov-Sinai entropy let us investigate another scattering rule (model III in ) which also models moving walls:
$$(p_x^{},p_y^{})=𝓣\text{ }_{}^1𝓣\text{ }_{}(p_x,p_y)$$
(26)
with
$$𝓣\text{ }_{}(p_x,p_y)=(\frac{\text{erf}\left[(p_xd)/\sqrt{2T}\right]+1}{2},\mathrm{exp}(p_y^2/2T))$$
(27)
Model III is still deterministic but no longer time reversible and only using this shear model a (numerical) equality between phase space contraction rate and entropy production was found in . Note that $`p_x^f`$ changes now to $`p_x^fd`$ and $`p_x^i`$ to $`p_x^id`$ in Eqs. (17,18). Figure 6 shows the corresponding Lyapunov spectra for a 36-particle system at $`\overline{n}=0.6`$ under shear, but in contrast to model I the Lyapunov exponents of both the positive and the negative branch now increase in magnitude with increasing shear rate $`\gamma `$. For comparison the Lyapunov spectrum for the Chernov-Lebowitz model is also plotted, with $`E_{kin}/N`$ and $`\gamma `$ equal to values obtained with model III at $`d=1.5`$. $`D_{KY}`$ for both models is compared in Fig. 7(a) and we see that the dimensionality loss with increasing shear rate $`\gamma `$ is stronger for model I than for model III. The graph of $`h_{KS}`$, plotted for both models in Fig. 7(b), asks for more explanation. Firstly, the overall behavior of $`h_{KS}`$ is increasing with larger $`\gamma `$, which seems to be the opposite of the observation made for the Chernov-Lebowitz shear model in . But there the total kinetic energy is kept constant for all shear rates whereas here both models try to fix the wall temperature. Increasing shear results in an increasing viscous heat production in the bulk which is reflected by a larger mean kinetic energy per particle \[Fig. 8(a)\]. Hence, the loss in $`h_{KS}`$ due to the ordering introduced by the shear is more than compensated by an increase of disorder due to a higher temperature in the bulk. Secondly, the only minor changes in the positive branch of the Lyapunov spectra for model I under shear yield an initially almost constant or even decreasing Kolmogorov-Sinai entropy. This, and the even more puzzling behavior of the wall temperature \[Fig. 8(b)\] can be explained by the fact that model I does not produce a Gaussian outgoing flux after the scattering. We found in that model I leads to an outgoing distribution with strong discontinuities in NSS whereas model III yields proper outgoing Gaussians. For comparison we have also computed $`D_{KY}`$ and $`h_{KS}`$ for the Chernov-Lebowitz model when it is approximately in the same macroscopic state as model III, i.e. we set the Chernov-Lebowitz system on the same kinetic energy shell and tried to find the appropriate shear parameter which results in the same shear rate. The Kaplan-Yorke dimension seems to be almost identical with that of model I, Fig. 7(a), and furthermore, the Kolmogorov-Sinai entropy now also increases with larger shear rate, only differing by a constant with the one obtained from model III, Fig. 7(b). This offset, depending on the special type of chaotic map chosen, originates from the fluctuating character of the model and should vanish in the thermodynamic limit.
## V Conclusion
We have calculated the full Lyapunov spectrum for a hard-disk fluid in equilibrium and nonequilibrium steady states thermostated by deterministic scattering. Since the model allows for fluctuations around a mean total energy only two vanishing Lyapunov exponents are found in both equilibrium and nonequilibrium states. In nonequilibrium the system is dissipative with a mean phase-space contraction rate smaller than zero. The magnitude of the Lyapunov exponents increases with increasing temperature gradient or shear rate, with a stronger increase for the negative branch. Thus both heat and shear flow situations result in a decreasing Kaplan-Yorke dimension and an increasing Kolmogorov-Sinai entropy with stronger nonequilibrium. Due to the inhomogeneous driving at the boundary the pairing rule does not hold. We did not verify the relation between Lyapunov exponents and transport coefficients (as e.g. in ) since in absence of a pairing rule this would be equivalent to checking the relation between phase space contraction and entropy production rates, where the latter has been done in . The main difference of our shear flow model III and the Chernov-Lebowitz shear flow model is the fact that in the latter the total kinetic energy is fixed whereas our models tries to fix the wall temperature. A direct comparison of both models reveals that they yield identical Kaplan-Yorke dimensions and only differ by a constant in the Kolmogorov-Sinai entropy, which vanishes in the thermodynamic limit.
## Acknowledgments
Special thanks go to Ch. Dellago for providing and explaining the original code used to compute the Lyapunov spectra and for helping to interpret the results. Furthermore, the author likes to thank O. Agullo for his technical support and R. Klages for pointing out relevant literature and for giving valuables advice. This work is supported, in part, by the Interuniversity Attraction Pole program of the Belgian Federal Office of Scientific, Technical and Cultural Affairs and by the Training and Mobility Program of the European Commission. |
warning/0001/nucl-ex0001010.html | ar5iv | text | # Spectroscopy of 13,14B via the one-neutron knockout reaction.
## I Introduction
The structure and reactions of “neutron-halo” nuclei, weakly bound systems that display a very diffuse surface of nearly pure neutron matter at densities far below that of normal nuclear matter, have recently been subjects of intense study . In a highly simplified picture, the longitudinal momentum distribution of fragments from the breakup of a loosely bound projectile directly reflects the internal momentum distribution of the valence nucleon and hence the square of the Fourier transform of its wave function. Thus, halo formation in such loosely bound nuclei can be investigated by measuring the momentum distribution of the fragment from a breakup reaction. The wide spatial dispersion characteristic of a halo neutron translates into a narrow momentum distribution. In this model, one treats the core of the nucleus as an inert spectator to the breakup process. While early experiments were based on the assumption that the measured momentum distribution represented a single reaction channel, it has recently become possible to go beyond this naive approach by measuring $`\gamma `$-radiation from the decay of excited states of the core in coincidence with the breakup fragments . This technique provides a further benefit in that partial cross sections for the excitation of different final states of the core can be measured, together with the longitudinal momentum distributions associated with inert and ‘active’ cores. This allows for spectroscopic investigation of both the core nucleus and the valence nucleon(s). The method, originally developed in a search for proton halo nuclei , has also been applied to the neutron case .
Among the candidates for neutron halo formation, the <sup>14</sup>B nucleus is of particular interest as the lowest-mass bound system amongst the N=9 isotones. The well-known halo nuclei <sup>11</sup>Be, <sup>11</sup>Li, <sup>14</sup>Be, <sup>17</sup>B, and <sup>19</sup>C occupy a similar position for N=7, N=8, N=10, N=12, and N=13, respectively (see Refs. and references therein for a discussion of <sup>19</sup>C structure). Moreover, <sup>14</sup>B is an odd-odd nucleus and thus different from any other neutron-halo system observed to date. In the first attempt to probe the structure of <sup>14</sup>B, Bazin, et al. reported a broadening of the longitudinal momentum distribution when compared with the shape expected from theoretical predictions, and suggested that this difference might be due to a strong contribution from core excitation. As a result, they were unable to draw any firm conclusion about halo formation in this nucleus. In order to investigate the structure of <sup>14</sup>B, and to determine the role that the core plays in the dissociation reaction, the momentum distributions of <sup>13</sup>B fragments corresponding to removal of neutrons from different orbitals in <sup>14</sup>B were measured using the technique described in Ref.. The contribution from core excitation was isolated by measuring the momentum distribution of excited <sup>13</sup>B fragments in coincidence with $`\gamma `$-rays from their decay.
This paper is divided into the following sections: the experimental setup and the procedure are described in Section II, while the experimental results and the analysis of the longitudinal momentum distributions is presented in Section III. Finally, a summary is given in Section IV.
## II Experiment
The experiment was performed at the National Superconducting Cyclotron Laboratory (NSCL) at Michigan State University. The <sup>14</sup>B radioactive secondary beam was produced by fragmentation of an 80 MeV per nucleon primary <sup>18</sup>O beam on a 790 mg/cm<sup>2</sup> Be target. The secondary <sup>14</sup>B beam, having an energy of 59.2 $`\pm `$ 0.3 MeV per nucleon, was then selected by the A1200 fragment separator and transmitted to a target chamber where a neutron was removed during an interaction with a Be or Au target, leaving the <sup>13</sup>B core in its ground state or in an excited state. The $`\gamma `$-rays from the decay of <sup>13</sup>B excited states were detected by an array of 38 cylindrical NaI (Tl) detectors which completely surrounded the target. Recoiling <sup>13</sup>B core fragments were momentum-analyzed using the S800 spectrograph . The momentum acceptance of the spectrograph was 6$`\%`$ and the angular acceptances were $`\pm 3.5^\mathrm{o}`$ and $`\pm 5^\mathrm{o}`$ in the dispersive and non-dispersive directions, respectively. These acceptances resulted in an estimated 99$`\%`$ efficiency for detecting events corresponding to the ground-state momentum distribution, and 95$`\%`$ for the momentum distributions of the excited states, as determined from the observed shapes. This point is discussed further below. The spectrograph was operated in dispersion-matched mode, in which the intrinsic dispersion of the secondary beam (0.5$`\%`$) is compensated by the last section of the spectrometer. The targets of Be and Au were 228 mg/cm<sup>2</sup> and 256 mg/cm<sup>2</sup> thick, respectively, chosen to produce an energy straggling of about the same magnitude as the resolution of the spectrograph. The overall resolution (including target thickness effects) was measured to be 9 MeV/c full width at half maximum (FWHM) in a direct-beam run with the target in place. This run also gave the normalization for the beam flux, determined to be $`5\times 10^3`$ particles/s.
Time-of-flight information (over a distance of 70 m) combined with the energy loss and total energy signal obtained with a segmented ion chamber and a 5 cm thick plastic scintillator, respectively, were used to identify and measure the yields of the fragments in the focal plane of the spectrograph. Two $`x/y`$ position-sensitive cathode-readout drift chambers in the focal plane were used to determine the momentum and angle information of the fragments.
## III Experimental Results
### A Cross sections and spin assignments for <sup>13</sup>B
An energy spectrum of $`\gamma `$-rays in coincidence with <sup>13</sup>B fragments is shown in Fig. 1. This spectrum has been transformed into the projectile rest frame using the position of the incident $`\gamma `$-ray measured in the NaI array, which allows for Doppler correction on an event-by-event basis. The solid curve is a fit to the data obtained via Monte-Carlo simulation using the code GEANT . The simulation took into account Doppler broadening, the distortion of the shapes caused by the back transformation to the projectile rest frame, and the calculated $`\gamma `$-ray detection efficiencies. Experimental efficiencies measured with calibrated radioactive sources agreed with the simulation to within 5$`\%`$. The background, which presumably results from a combination of breakup reactions leaving the target in an excited state and secondary interactions of neutrons with the detector and surrounding materials, was parameterized by a simple exponential dependence. Taking into account the 3.48 MeV, 3.68 MeV and 4.13 MeV transitions in <sup>13</sup>B, the fit to the experimental spectrum is excellent (the $`\chi ^2`$ per degree of freedom N is 1.08). The choice of these particular $`\gamma `$-ray transitions is discussed below. Eliminating the 4.13 MeV $`\gamma `$-ray from the fit causes $`\chi ^2`$/N to increase to 1.10 for 170 degrees of freedom, which is marginally significant. This indicates that all the important transitions were accounted for and that there are no other significant $`\gamma `$-rays between 1.5 and 3.5 MeV or above 4.2 MeV.
The level scheme of <sup>13</sup>B is presented in Fig. 2. The ground state has J$`{}_{}{}^{\pi }=\frac{3}{2}^{}`$, resulting from a \[$`1p_{3/2}`$\]<sup>-1</sup> proton configuration. Below 4.5 MeV, three negative parity and two positive parity excited states have been identified. Although the energies and parities of these states are well established from particle-transfer reaction studies, very little is known about their spins. Based on the $`\gamma `$-ray spectrum of Fig. 1, we are interested in the group of states between 3.48 and 4.13 MeV.
In the simplest model, the J$`{}_{}{}^{\pi }=2^{}`$ ground state of <sup>14</sup>B results from the coupling of a $`2s1d`$-shell neutron to the <sup>13</sup>B ground state. Then, the neutron removal from <sup>14</sup>B could proceed by the removal of the $`sd`$-shell neutron to leave <sup>13</sup>B in its ground state, by removal of a neutron from the $`1p_{1/2}`$ orbital leaving <sup>13</sup>B in an excited state with J$`{}_{}{}^{\pi }=3/2^+,5/2^+`$, or by removal of a neutron from the $`1p_{3/2}`$ orbital leaving <sup>13</sup>B in a more highly excited state with J$`{}_{}{}^{\pi }=1/2^+,3/2^+,5/2^+`$, or $`7/2^+`$.
This simple model is confirmed by shell-model calculations in the $`1p2s1d`$ model space with the WBT and WBP residual interactions . Calculations were carried out for the <sup>14</sup>B ground state with a $`(1p)^3(sd)^1`$ configuration (relative to a closed shell for for <sup>16</sup>O), and for the <sup>13</sup>B spectrum with $`(1p)^3(2s1d)^0`$, $`(1p)^4(2s1d)^1`$ and $`(1p)^5(2s1d)^2`$ configurations. (The total number of eigenstates are 5, 299 and 2973, respectively.) The low-lying energy levels obtained for these configurations, labeled by (0), (1) and (2), respectively, are shown in Fig. 2. The spectroscopic factors for the WBT interaction are given in Table 1 (the spectroscopic factors obtained with WBP are similar to those of WBT). The higher levels, which are reached by predominantly $`1p_{3/2}`$ removal, lie above the neutron decay threshold of 4.9 MeV. The spectroscopic factors leading to the (2) configurations are zero with our configuration assumptions. All of the (0), (1) and (2) configurations are allowed in the <sup>11</sup>B(t,p)<sup>13</sup>B reaction which has been previously used to identify the <sup>13</sup>B states . In principle, the (0) and (2) configurations can be mixed, which could give some spectroscopic strength leading to negative parity states around 4 MeV in excitation. Our estimates within the mixed (0)+(2) model space indicate that the effect of this mixing leads to spectroscopic factors of about 0.02 or less, and the effect of this mixing will be ignored here.
Partial cross sections obtained from the absolute $`\gamma `$-ray intensities corrected by the computed efficiencies and the spectrometer acceptance are presented in Table I. The overall error in these cross sections are of the order of 18$`\%`$, due mainly to systematic errors in the fit to the experimental gamma-ray spectrum, and to the uncertainties in the number of particles in the beam (12$`\%`$) and the target thickness (5$`\%`$).
The calculated cross section for each final state of the core fragment in a one-nucleon knockout reaction is given by:
$$\sigma (n)=\underset{j}{}C^2S(j,n)\sigma _{sp}(j,B_n),$$
(1)
where $`C^2S(j,n)`$ is the computed spectroscopic factor for the removed nucleon with respect to a given core state and $`\sigma _{sp}(j,B_n)`$ is the reaction cross section for the removal of a nucleon from a single-particle state with total angular momentum $`j`$. $`B_n`$ is the sum of the separation energy and excitation energy of the state $`n`$. The single-particle cross sections were calculated by Tostevin in a eikonal model assuming that there are two reaction mechanisms involved in the knock-out reaction: (i) nucleon stripping in which the halo nucleon interacts strongly with the target and leaves the beam, and (ii) diffraction dissociation in which the nucleon moves forward with essentially the beam velocity but away from the core.
The calculated cross sections for the states below 4.5 MeV are given in Table I. The best agreement with experiment is achieved with the spin assignments shown there. It should be noted that the evidence for the existence of the 4.13 MeV transition in the experimental $`\gamma `$-ray spectrum is marginal. Furthermore, the calculated strength for this transition (Table I) does not take into account the configuration mixing discussed above. Overall, it appears that the combination of the theoretical spectroscopic factors and computed single-particle cross sections does a remarkable job of reproducing the experimental yields.
### B Momentum distributions
The momentum vector of the fragment at the target position after the reaction was reconstructed using the known magnetic fields and the positions and angles at the focal plane, and the ion-optics code COSY . The momentum components transverse to the beam direction are sensitive to the reaction mechanism, while the longitudinal momentum component is relatively independent of the details of the reaction mechanism as discussed by Orr, et al. . The longitudinal momentum distribution is obtained as the projection of the total momentum in the direction parallel to the beam. From the coincident $`\gamma `$-ray data, it is possible to generate the distribution for the case when <sup>13</sup>B is left in its ground state, by subtracting the contribution from the excited states from the singles spectrum after correcting for the $`\gamma `$-ray detection efficiencies. The result of this subtraction process is presented in Fig. 3(a) and corresponds to the removal of a single neutron from the $`sd`$ shell in <sup>14</sup>B due to a reaction with a <sup>9</sup>Be target. Fig. 3(b) shows the longitudinal momentum distribution for the core-excited states, which is considerably broader.
A more precise comparison can be made through the use of the reaction theory. Calculated longitudinal momentum distributions, also shown in Fig. 3, were determined within the framework of the eikonal model following the procedure of Ref. . Since the dissociation products are formed at an impact parameter greater than $`b_{min}=R_C+R_n`$ (where $`R_C`$ and $`R_n`$ are the energy-dependent radii for the core and valence nucleon, usually chosen to reproduce the measured interaction cross section) the black disc approximation can be used. The cross section is then expressed in terms of the impact parameter as the one-dimensional Wigner transform of the wave function after the reaction, where the wave function profile is unity outside of a cutoff radius and zero inside. The momentum distribution corresponding to the <sup>13</sup>B core in its ground state, shown in Fig. 3(a), agrees reasonably well with the l = 0 curve of the eikonal model calculation, which represents the removal of an $`s`$-wave valence neutron from the $`sd`$ shell in <sup>14</sup>B. Similar results have been obtained for <sup>11</sup>Be and <sup>15</sup>C . The width obtained for the distribution is $`55\pm 2`$ MeV/c FWHM, after correcting for the 9 MeV/c experimental momentum resolution. This width results from a Lorentzian fit to the experimental data and is in excellent agreement with the value of 56 MeV/c FWHM displayed by the theoretical l = 0 curve in the lab. system. The disagreement in the tails of the distribution suggests that there may be a contribution from the l = 2 distribution and possibly other effects due to the acceptance of the spectrometer, approximations in the reaction model used to analyze the data, or unidentified contributions to the line shape .
Spectroscopic factors of 0.306 and 0.662, respectively, are obtained from the shell model calculation for the removal of a valence neutron from the $`1d_{5/2}`$ and $`2s_{1/2}`$ states in <sup>14</sup>B, which indicates that a non-negligible l = 2 admixture is expected. Individual cross sections for both configurations have been derived by fitting the experimental distribution with a linear combination of l = 0 and l = 2 shapes. The best fit, shown in Fig. 4, implies that the removal of a valence neutron from the $`2s_{1/2}`$ ($`1d_{5/2}`$) orbital in the <sup>14</sup>B ground state accounts for 89$`\pm 3\%`$ (11$`\pm 3\%`$) of the total yield. The agreement with cross sections derived from the eikonal model and the theoretical spectroscopic factors is very good (Table I), as is the quality of the fit to the experimental data shown in Fig. 4.
The core-excited longitudinal momentum distribution, shown in Fig. 3(b), was obtained from the $`\gamma `$-ray-coincident <sup>13</sup>B yield after subtracting a background distribution derived by gating on the exponential part of the $`\gamma `$-ray spectrum at high energy and normalizing to the extrapolated background under the peaks. The background computed in this way constituted 26$`\%`$ of the coincident data in the region of the three identified $`\gamma `$-ray transitions. The core-excited distribution has a shape that agrees well with that expected for l = 1, confirming the expectation of the removal of a $`1p_{1/2}`$ neutron. The experimental width, obtained from a Lorentzian fit to the central part of the distribution, is $`135\pm 15`$ MeV/c FWHM. The calculated width in the lab. frame of the theoretical l = 1 curve is 144 MeV/c. The distributions computed for $`p`$ and $`d`$ wave breakup are broader than the actual momentum acceptance of the spectrometer. A (small) correction to the integrated yield due to this limited acceptance has been applied to the experimental cross sections given in Table I.
### C Coulomb breakup
The longitudinal momentum distribution obtained in the one-neutron breakup reaction of <sup>14</sup>B on a <sup>197</sup>Au target is shown in Fig. 5. The $`\gamma `$-ray coincident yield has been subtracted using the same procedure as described in the previous section. In this case, the $`\gamma `$-ray spectrum was essentially equal to the exponential background distribution, and no core-excited transitions were observed. The width obtained from a gaussian fit (dashed curve in Fig. 5) to the experimental longitudinal momentum distribution is $`59\pm 3`$ MeV/c (after correcting for the experimental resolution). This width is larger than the value of $`48\pm 3`$ MeV/c measured by Bazin, et al. for Coulomb breakup on a tantalum target at 86 MeV/A. However, it agrees within the errors with that extracted from the <sup>13</sup>B ground-state longitudinal momentum distribution for the Be target in the present work. The integrated cross section derived for Coulomb breakup of <sup>14</sup>B on <sup>197</sup>Au (leaving the <sup>13</sup>B core fragment in the ground-state) is $`638\pm 45`$ mb, much larger than the yield quoted in Ref.. It was suggested there that the smaller cross section reflected a quenching of the soft dipole strength that implied “normal” nuclear structure for <sup>14</sup>B. However, a possible experimental problem due to the restricted angular acceptance of the A1200 spectrometer was acknowledged. The present observations of a broader momentum distribution coupled with a larger cross section suggest that this was indeed the case.
The Coulomb breakup cross section was calculated using a Yukawa potential with finite-size corrections . The result, 543 mb, is less than our measurement. However, the yield calculated in Ref. using a Woods-Saxon potential was 50$`\%`$ larger than that from the Yukawa form. Our data suggest that the actual situation is intermediate between these two approximations. The Coulomb contribution to our Be-target data was also computed and found to be negligibly small (2.3 mb). The solid curve in Fig. 5 shows the longitudinal momentum distribution for the Au target predicted in the model of Ref. , normalized to the experimental data. The width of this Lorentzian function compares well with experiment, but the agreement in the tails of the distribution is unsatisfactory.
### D Neutron halo in <sup>14</sup>B
Zhongzhou, et al. have predicted an inversion of the $`1d_{5/2}`$ and $`2s_{1/2}`$ orbitals for <sup>14</sup>B within the framework of a non-linear relativistic mean-field calculation. This is in agreement with the shell-model calculation cited above and with our observation of a dominant l = 0 component in the ground state of this nucleus. Their calculation shows that, under these conditions, <sup>14</sup>B is a neutron halo nucleus. On the other hand, measurements of the interaction cross sections at relativistic energies for particle-stable B isotopes ranging from mass 8-15, by Tanihata, et al. , suggested that their effective RMS radii are practically constant, which does not support the halo hypothesis. Measurements at intermediate energy, analyzed by Liatard, et al. , gave larger RMS radii in all cases than those of Ref. , together with a significant mass dependence and an anomalously large radius for <sup>14</sup>B. The radius difference is not by any means as large as in the case of <sup>11</sup>Li, for example, but this is not unexpected given the fact that the valence neutron is more tightly bound in <sup>14</sup>B. Of course, the RMS radius extracted from an interaction cross section is not the only signature for halo structure , and the narrow longitudinal momentum distribution and large cross section for dissociation of the ground state of <sup>14</sup>B observed in the present experiment strongly argue for halo structure in this system.
## IV Summary
In this experiment, we have measured the longitudinal momentum of the <sup>13</sup>B core fragment in one-neutron knockout from <sup>14</sup>B, on both <sup>9</sup>Be and <sup>197</sup>Au targets. The contribution from core excitation was isolated by measuring the excited <sup>13</sup>B fragments in coincidence with $`\gamma `$-rays from their decay. Comparison of the observed intensities of core-excited $`\gamma `$-ray transitions with a shell-model calculation, using an eikonal reaction theory, allowed us to make spin assignments for two positive-parity excited states in <sup>13</sup>B. The longitudinal momentum distributions in coincidence with the positive-parity core-excited transitions are consistent with knockout from a $`1p_{3/2}`$ state, as expected. An excellent fit to the distribution obtained when the <sup>13</sup>B core remains in its ground state is also provided by the shell-model calculation, which predicts an admixture of $`2s_{1/2}`$ and $`1d_{5/2}`$ neutron knockout. The experimental data imply an 89$`\pm 3\%`$ l = 0 component, in agreement with the shell-model result. The Coulomb breakup cross section, measured with the <sup>197</sup>Au target, is consistent with expectations for a weakly-bound $`2s_{1/2}`$ neutron. The width of the longitudinal momentum distribution obtained with the <sup>197</sup>Au target ($`59\pm 3`$ MeV/c) agrees with that obtained using the <sup>9</sup>Be target after the core-excited component is subtracted.
The results of the present experiment lend very strong support to the idea that <sup>14</sup>B is a “neutron-halo” system, the first odd-odd nucleus to display this structure. It also appears that, with the sole exception of <sup>17</sup>C at N=11, all of the lowest-mass, particle-stable isotones from N=7-13 are halo nuclei. It therefore seems that re-investigation of the structure of <sup>17</sup>C is in order.
## V acknowledgments
The first author (V.G.) was financially supported by FAPESP (Fundação de Amparo a Pesquisa do Estado de São Paulo - Brazil) while on leave from the UNIP (Universidade Paulista). This work was supported by the National Science Foundation under Grants No. PHY94-02761, PHY99-01133, PHY95-14157, and PHY96-05207. |
warning/0001/nucl-th0001002.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The interaction of the nucleon with the $`\mathrm{\Lambda }`$ hyperon is of great theoretical interest. Since the one-pion-exchange interaction is absent (except under broken isospin ), one is able to focus on the shorter range contributions to the interaction. The singlet and triplet scattering lengths are predicted to be equal by SU(6) symmetry.
Unfortunately, experiments are difficult to perform for this system because of the short lifetime of the $`\mathrm{\Lambda }`$. Nevertheless, some experimental results do exist and have been analyzed. Rijken et al. have recently generated a set of potentials which fit these data. The range of scattering lengths obtained from this analysis is rather large, which is perhaps a natural consequence considering that the data do not extend to low energies.
Use of the radiative capture of a $`\pi ^{}`$ from an s-wave atomic state by deuterium to measure the neutron-neutron (nn) scattering length, through the influence of the final state interaction, was investigated in an experiment by Phillips and Crowe and soon after studied theoretically by McVoy and Bander. The basic technique consists of comparing the shape of the of the photon spectrum with calculations. Because of the final-state interaction of the two neutrons, a peak results near the high-energy end of the spectrum corresponding to a maximum in the s-wave phase shifts at low neutron-neutron energy. Since this phase shift rises very quickly to a maximum, as a function of energy in the neutron-neutron frame, the peak in the photon spectrum is very near the maximum energy. It is also possible to use the neutron spectrum at low energies for the studies provided that the angle between the neutron and photon is known. Experiments have been performed in both geometries.
These early exploratory efforts were followed by an experiment detecting the neutrons which gave useful results. Following a study of limits due to the theoretical uncertainties, experiments were carried out at the Paul Scherrer Institute (PSI) detecting only the photon, as well as experiments in which the photon and neutron were detected in coincidence. These experiments resulted in values for the nn scattering length (and effective range) which were considered to be the best obtained to date. Recently another experiment was performed using this reaction at the Clinton P. Anderson Meson Physics Facility. Results from this experiment appear to confirm the PSI results. The uncertainty in the scattering length is of the order of $`\pm 0.5`$ fm for a value of -18.58 fm, about 2-3%.
A determination of the $`\mathrm{\Lambda }n`$ scattering lengths from the photon spectrum of the reaction K<sup>-</sup>d$``$n$`\mathrm{\Lambda }\gamma `$, where the capture takes place from an atomic state, was suggested by Gibson et al.. A study of this process performed by Workman and Fearing concluded that the different hadronic routes considered by Akhiezer et al. had a negligible impact on a possible measurement (however, see Refs. for comments on their representation of the amplitudes). They also concluded that the dominant operator for the conversion of a proton to a $`\mathrm{\Lambda }`$ through radiative capture of a kaon was of the Kroll-Ruderman form
$$𝒪=𝝈\mathit{ϵ}$$
(1)
where $`\mathit{ϵ}`$ is the polarization of the outgoing photon and $`𝝈`$ is the spin operator for the proton on which the capture takes place.
At the same time as this theoretical effort, a feasibility study demonstrated that the spectrum for this reaction could be separated from the background. Workman and Fearing compared their analysis with these data to show that reasonable values of the scattering lengths were consistent with the data.
Two important issues were left unresolved by the work of Workman and Fearing. They used only the asymptotic $`\mathrm{\Lambda }`$-neutron wave functions. In the work of Ref. the lack of knowledge of the short range final-state wave function was found to be the dominant theoretical uncertainty in the analysis of the scattering length, of the order of $`\pm 0.3`$ fm. If the uncertainty is of the same order in the case of $`\mathrm{\Lambda }`$-neutron scattering the consequences are more serious since the scattering length itself is thought to be an order of magnitude smaller.
A second problem, which may be more serious, is that, while in the case of the neutron-neutron final state the spin state of the two neutrons is restricted to be singlet by the Pauli principle, the $`\mathrm{\Lambda }`$n final state consists of a mixture of a singlet and triplet states. One expects that the scattering in these two states may be different, and indeed the dependence of the scattering length on the spin might provide some very important clues to the structure of the interaction. While it has been suspected that polarization information may provide a tool to obtain a singlet-triplet separation (see Balewski et al. for example), we provide here the formalism and expressions necessary to obtain such a separation for this reaction.
We first give a brief heuristic description of the basic physics underlying the development to follow. Consider a deuteron target prepared in a configuration with only magnetic spin projection zero along a z-axis defined by the direction of the photon. Assume also that the final state of the $`\mathrm{\Lambda }`$-n system is in a singlet state. Since the $`K^{}`$ is assumed to be captured from an atomic s-state, there will be a total spin projection of zero in the initial state and zero angular momentum projection in the final state, other than that due to the spin of the $`\gamma `$. Since the “transversality” relation,
$$\mathit{ϵ}𝐤=0,$$
(2)
requires that the spin projection of the photon must lie along its direction of travel (i.e. there is no zero spin projection) this assumed transition to the singlet final state must have zero amplitude due to the conservation of the z-projection of angular momentum. Thus, for example, a spectrum taken under these conditions could be analyzed to obtain the triplet scattering parameters alone.
The previous paragraph gives only an example of how deuteron spin information can affect the analysis. A measurement of the photon circular polarization can also help to separate the singlet and triplet scattering states. In the subsequent sections we develop the expressions necessary to calculate the photon spectrum and provide specific results.
## 2 Expressions for the Rate
For comparison with the measurement we need the theoretical prediction for the shape of the spectrum. The absolute rate is not usually measured in this type of experiment.
### 2.1 Kinematics and Phase Space
The magnitude of the momentum, $`𝐩`$, of each baryon in this pair’s center of mass is given by
$$p^2=\frac{(s+M_n^2M_\mathrm{\Lambda }^2)^2}{4s}M_n^2$$
(3)
with $`s`$ the center of mass energy of the $`\mathrm{\Lambda }`$-neutron pair given by
$$s=E_0^22kE_0,$$
(4)
where $`E_0`$ is the total energy of the initial system and $`k`$ is the final photon momentum. $`M_\mathrm{\Lambda }`$ and $`M_n`$ are the masses of the $`\mathrm{\Lambda }`$ and neutron respectively.
Since we shall assume that the final-state baryons are not observed, we must integrate over their directions, giving for the differential rate of the observed photon
$$\frac{d\mathrm{\Gamma }}{d\mathrm{\Omega }_𝐤dk}\frac{pk\omega _\mathrm{\Lambda }\omega _n}{(\omega _\mathrm{\Lambda }+\omega _n)}|M|^2𝑑\mathrm{\Omega }_𝐩$$
(5)
The total energies, $`\omega _n`$ and $`\omega _\mathrm{\Lambda }`$ are very nearly constant over the energy range of interest.
It should be noted that a single factor of the photon momentum $`k`$ appears. It is linear only with relativistic phase space; it would appear to the second power in non-relativistic phase space. When this reaction was first proposed, the non-relativistic version was usually used (in fact Williams used this reaction as an example of the use of non-relativistic phase space). Of course, either type of phase space can be used provided that the proper corresponding operator is used. The difference between a first and second power of the photon momentum is relativistic in origin and the effect is very small in many cases. In the analysis of Ref. only the very end of the spectrum was used and the photon momentum is sufficiently constant over this range that such a factor is immaterial. For the analysis of Ref. the larger range of the spectrum used leads to a small sensitivity to the power of $`k`$ used in this expression. For the present reaction, since the peak lies farther from the end point of the spectrum than for the $`\pi ^{}dnn\gamma `$ reaction, a larger range of photon momenta may be included in the analysis and the correct factor is more important.
### 2.2 Matrix Element
The mathematical development of the expressions for the matrix element used here follows a different procedure than that of Refs. and . The previous methods have written the matrix element for the transition first as a plane wave transition then corrected this expression by adding and subtracting the s-wave contribution. The s-wave contribution with a final state interaction is then substituted in the added element. For a matrix element with two particles detected in the final state this is the appropriate method. However, for the case in which only the photon is detected, there is an average over the unobserved baryon momenta which must be made as outlined in the previous section. This integral was carried out numerically in the previous developments.
In the method used in the present analysis the partial-wave quantities are calculated in terms of the relative $`\mathrm{\Lambda }`$-neutron momentum. This technique requires several partial waves to be computed for a handful of magnetic quantum numbers. While it may seem a priori less efficient, in the end it has several advantages. The first is that the final averaging over the direction of the unobserved momenta can be done analytically. Since the contributions from the higher partial waves depend only on the initial state wave functions (we assume no interaction in partial waves higher than $`\mathrm{}=0`$), if one wishes to fit the data by varying the scattering length and effective range for the singlet and triplet states, these higher partial waves need be calculated only once.
A second advantage is that the degree of coherence or incoherence of the singlet and triplet states is manifest. For example, we will see that if all magnetic quantum numbers of the initial deuteron are equally populated and the polarization of the final photon is not observed, the transitions to the singlet and triplet final states are incoherent.
The matrix element for the reaction which proceeds from an initial deuteron with spin-projection $`S_z`$ to a final state of the $`\mathrm{\Lambda }`$-neutron system in a total spin state $`(S^{},S_z^{})`$ is
$$M^{S^{},S_z^{},S_z}=𝑑𝐫_1𝑑𝐫_2<S^{}S_z^{}|\psi _S^{}^{}(𝐩,𝐫)e^{i𝐊(𝐫_1+𝐫_2)}|𝒪e^{i𝐤𝐫_1}|\mathrm{\Psi }_D^{S_z}(𝐫)>$$
(6)
where $`\psi _S^{}^{}(𝐩,𝐫)`$ is the final state wave function of the neutron-$`\mathrm{\Lambda }`$ system and the variable $`S^{}`$ is 0 or 1 for the singlet or triplet case. $`𝐊`$ is the center-of-mass momentum of the baryon pair. The quantity $`|\mathrm{\Psi }_D^{S_z}(𝐫)>`$ is the deuteron wave function with an initial spin projection $`S_z`$. We obtain the radial deuteron wave function from the solution of the coupled Schrödinger equations for a pure one-pion-exchange potential. It has been shown that this interaction reproduces all of the low-energy properties of the deuteron (likely to be the most important in this calculation). The operator $`𝒪=𝝈_1\mathit{ϵ}`$.
Transforming to relative and center-of-mass coordinates in the spatial variables we have
$$M^{S^{},S_z^{},S_z}=(2\pi )^3\delta (𝐊𝐤)𝑑𝐫<S^{}S_z^{}|\psi _S^{}^{}(𝐩,𝐫)|𝒪e^{i\frac{𝐤𝐫}{2}}|\mathrm{\Psi }_D^{S_z}(𝐫)>.$$
(7)
We write the deuteron wave function as
$$|\mathrm{\Psi }_D^{S_z}(𝐫)>=S(r)Y_0^0|1S_z>+D(r)Y_2^m^{}(𝐫)|1\sigma >C_{1,2,1}^{\sigma ,m^{},S_z},$$
(8)
and the $`\mathrm{\Lambda }`$-n relative motion wave function as
$$\psi _S^{}(𝐩,𝐫)=4\pi i^{\mathrm{}}Y_{\mathrm{}}^m(𝐫)Y_{\mathrm{}}^m(𝐩)\varphi _{\mathrm{}}^S^{}(r).$$
(9)
With these expansions we can write (with an arbitrary overall normalization, omitting the delta function)
$$M^{S^{},S_z^{},S_z}=4\pi <S^{}S_z^{}|𝒪|𝑑𝐫\underset{\mathrm{},m,L,M}{}i^L\mathrm{}Y_{\mathrm{}}^m(𝐫)Y_L^M(𝐫)Y_{\mathrm{}}^m(𝐩)Y_L^M(𝐤)\varphi _{\mathrm{}}^S^{}(r)j_L\left(\frac{kr}{2}\right)$$
$$\times \left[S(r)Y_0^0|1S_z>+D(r)Y_2^m^{}(𝐫)|1\sigma >C_{1,2,1}^{\sigma ,m^{},S_z}\right].$$
(10)
With the photon direction, $`𝐤`$, along the z axis, one has
$$Y_L^M(𝐤)=\delta _{M,0}\sqrt{\frac{2L+1}{4\pi }}.$$
(11)
Performing the integral over the angles of $`𝐫`$, we obtain
$$M^{S^{},S_z^{},S_z}=\underset{\mathrm{},m}{}Y_{\mathrm{}}^m(𝐩)[<S^{}S_z^{}|𝒪|1S_z>\delta _{m0}\sqrt{2\mathrm{}+1}_0^{\mathrm{}}r^2dr\varphi _{\mathrm{}}^S^{}(r)j_{\mathrm{}}(\frac{1}{2}kr)S(r)$$
$$+\underset{L,\sigma }{}i^L\mathrm{}\sqrt{\frac{5(2L+1)}{(2\mathrm{}+1)}}C_{L,2,\mathrm{}}^{0,m,m}C_{L,2,\mathrm{}}^{0,0,0}\sqrt{2L+1}C_{1,2,1}^{\sigma ,m,S_z}<S^{}S_z^{}|𝒪|1\sigma >$$
$$\times _0^{\mathrm{}}r^2drD(r)\varphi _{\mathrm{}}^S^{}(r)j_L\left(\frac{kr}{2}\right)]$$
(12)
or,
$$M^{S^{},S_z^{},S_z}=\underset{\mathrm{},m}{}Y_{\mathrm{}}^m(𝐩)[S_{\mathrm{}}(S^{})\delta _{m,0}<S^{}S_z^{}|𝒪|1S_z>$$
$$+\sqrt{5}\underset{\sigma }{}C_{1,2,1}^{\sigma ,m,S_z}<S^{}S_z^{}|𝒪|1\sigma >\underset{L}{}i^L\mathrm{}C_{\mathrm{}\mathrm{\hspace{0.17em}2},L}^{m,m,0}C_{\mathrm{},2,L}^{0,0,0}D_{\mathrm{},L}(S^{})]$$
(13)
with the definitions
$$S_{\mathrm{}}(S^{})=\sqrt{2\mathrm{}+1}_0^{\mathrm{}}r^2𝑑r\varphi _{\mathrm{}}^S^{}(r)j_{\mathrm{}}(\frac{kr}{2})S(r)$$
(14)
$$D_{\mathrm{},L}^S^{}=\sqrt{2\mathrm{}+1}_0^{\mathrm{}}r^2𝑑rD(r)\varphi _{\mathrm{}}^S^{}(r)j_L\left(\frac{kr}{2}\right).$$
(15)
If we define
$$B_{\mathrm{}}^m(S^{})\underset{L}{}i^L\mathrm{}C_{\mathrm{},2,L}^{m,m,0}C_{\mathrm{},2,L}^{0,0,0}D_{\mathrm{},L}(S^{}),$$
(16)
we can write
$$M^{S^{},S_z^{},S_z}=\underset{\mathrm{},m}{}Y_{\mathrm{}}^m(𝐩)\left[S_{\mathrm{}}\delta _{m,0}<S^{}S_z^{}|𝒪|1S_z>+\sqrt{5}\underset{\sigma }{}C_{1,2,1}^{\sigma ,m,S_z}<S^{}S_z^{}|𝒪|1\sigma >B_{\mathrm{}}^m\right],$$
(17)
where the explicit dependence on $`S^{}`$ of $`S_{\mathrm{}}`$ and $`B_{\mathrm{}}^m`$ has been suppressed.
We define the spherical components of the photon polarization vector as
$$ϵ^{+1}=\frac{1}{\sqrt{2}}(ϵ_x+iϵ_y);ϵ^1=\frac{1}{\sqrt{2}}(ϵ_xiϵ_y);ϵ^0=ϵ_z.$$
(18)
The spin structure of the radiative capture operator is assumed to be of the form
$$𝒪=𝝈_1\mathit{ϵ}$$
(19)
where particle 1 is the proton which is transformed into a $`\mathrm{\Lambda }`$. The expectation values of this operator for the baryon spin states are
$$\begin{array}{cc}& \\ \mathrm{Triplet}\hfill & \\ & \\ <1,0|𝒪|1,1>\hfill & \hfill ϵ^+\\ <1,0|𝒪|1,1>\hfill & \hfill ϵ^{}\\ <1,0|𝒪|1,0>\hfill & \hfill 0\\ <1,1|𝒪|1,1>\hfill & \hfill ϵ^0\\ <1,1|𝒪|1,1>\hfill & \hfill ϵ^0\\ <1,1|𝒪|1,0>\hfill & \hfill ϵ^{}\\ <1,1|𝒪|1,0>\hfill & \hfill ϵ^+\\ <1,1|𝒪|1,1>\hfill & \hfill 0\\ <1,1|𝒪|1,1>\hfill & \hfill 0\\ & \\ \mathrm{Singlet}\hfill & \\ & \\ <0,0|𝒪|1,1>\hfill & \hfill ϵ^+\\ <0,0|𝒪|1,1>\hfill & \hfill ϵ^{}\\ <0,0|𝒪|1,0>\hfill & \hfill ϵ^0\end{array}$$
The transversality condition gives $`ϵ^0=0`$.
Because of the averaging over the direction of the unobserved momentum $`𝐩`$, each term in $`\mathrm{}`$ and $`m`$ will contribute incoherently. We now consider each of the cases $`m=0`$, $`m=\pm 1`$ and $`m=\pm 2`$ separately. In the square-averaging over the photon polarizations we omit the uniform factor of 1/2.
#### 2.2.1 m=0
For $`m=0`$
$$M_{m=0}^{S^{},S_z^{},S_z}=<S^{}S_z^{}|𝒪|1S_z>\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)(S_{\mathrm{}}+\sqrt{5}C_{1,2,1}^{S_z,0,S_z}B_{\mathrm{}}^0)$$
(20)
This term will contribute to the transitions $`(S_z=\pm 1S_z^{}=0)`$ and $`(S_z=0S_z^{}=\pm 1)`$
$$M_{m=0}^{1,0,1}=ϵ^{}\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)(S_{\mathrm{}}+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0)=M_{m=0}^{0,0,1}$$
(21)
$$M_{m=0}^{1,0,1}=ϵ^+\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)(S_{\mathrm{}}+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0)=M_{m=0}^{0,0,1}$$
(22)
$$M_{m=0}^{1,1,0}=ϵ^{}\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)(S_{\mathrm{}}\sqrt{2}B_{\mathrm{}}^0);M_{m=0}^{1,1,0}=ϵ^+\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)(S_{\mathrm{}}\sqrt{2}B_{\mathrm{}}^0)$$
(23)
giving a contribution to the total capture rate of
$$\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}(S_{\mathrm{}}+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0)^2$$
(24)
from each of the initial magnetic states $`\pm 1`$ for either the singlet or triplet final state and
$$2\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}(S_{\mathrm{}}\sqrt{2}B_{\mathrm{}}^0)^2$$
(25)
from the initial magnetic state $`0`$ for the triplet final state.
Since the singlet and triplet states add constructively for $`S_z=1`$ and destructively for $`S_z=1`$, they are incoherent if and only if the populations of the these two initial states are equal.
#### 2.2.2 $`|m|=1`$
For $`|m|=1`$
$$M_{m=1}^{S^{},S_z^{},S_z}=<S^{}S_z^{}|𝒪|1S_z1>\sqrt{5}C_{1,2,1}^{S_z1,1,S_z}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1$$
(26)
contributes to $`(S_z=0S_z^{}=0)`$;
$$M_{m=1}^{1,0,0}=ϵ^{}\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1=M_{m=1}^{0,0,0}$$
(27)
and $`(S_z=1S_z^{}=\pm 1)`$
$$M_{m=1}^{1,1,1}=ϵ^{}\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1;M_{m=1}^{1,1,1}=ϵ^+\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1$$
(28)
while
$$M_{m=1}^{S^{},S_z^{},S_z}=<S^{}S_z^{}|𝒪|1S_z+1>\sqrt{5}C_{1,2,1}^{S_z+1,1,S_z}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1$$
(29)
contributes to $`(S_z=0S_z^{}=0)`$;
$$M_{m=1}^{1,0,0}=ϵ^+\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1=M_{m=1}^{0,0,0}$$
(30)
and $`(S_z=1S_z^{}=\pm 1)`$
$$M_{m=1}^{1,1,1}=ϵ^{}\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1;M_{m=1}^{1,1,1}=ϵ^+\sqrt{\frac{3}{2}}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1$$
(31)
giving a contribution from $`|m|=1`$ to the capture rate of
$$3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left(B_{\mathrm{}}^1\right)^2;\mathrm{for}\mathrm{each}\mathrm{of}S_z=0,\pm 1.$$
(32)
Only the $`S_z=0`$ state contributes for the singlet final state. Since the relative sign of the singlet and triplet is different for $`m=1`$ and $`m=1`$, the singlet and triplet states are always incoherent for $`|m|=1`$.
#### 2.2.3 $`|m|=2`$
For $`|m|=2`$
$$M_{m=2}^{S^{},S_z^{},S_z}=<S^{}S_z^{}|𝒪|1S_z2>\sqrt{5}C_{1,2,1}^{S_z2,2,S_z}\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^2(𝐩)B_{\mathrm{}}^2$$
(33)
contributes to $`(S_z=1S_z^{}=0)`$
$$M_{m=2}^{1,0,1}=ϵ^{}\sqrt{3}\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^2(𝐩)B_{\mathrm{}}^2=M_{m=2}^{0,0,1}$$
(34)
while
$$M_{m=2}^{S^{},S_z^{},S_z}=<S^{}S_z^{}|𝒪|1S_z+2>\sqrt{5}C_{1,2,1}^{S_z+2,2,S_z}\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^2(𝐩)B_{\mathrm{}}^2$$
(35)
contributes to $`(S_z=1S_z^{}=0)`$.
$$M_{m=2}^{1,0,1}=ϵ^+\sqrt{3}\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^2(𝐩)B_{\mathrm{}}^2=M_{m=2}^{0,0,1}$$
(36)
giving a contribution to the capture rate for either initial spin projection $`\pm 1`$ of
$$3\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left(B_{\mathrm{}}^2\right)^2$$
(37)
Since the singlet and triplet states add constructively for $`S_z=1`$ and destructively for $`S_z=1`$, they are incoherent only if the populations of the these two initial states are equal.
### 2.3 Summary
Combining the expressions in the previous section we obtain the following results for the capture rate for the various initial and final states. The notation for quantities calculated with the singlet or triplet final state on the $`\mathrm{\Lambda }`$-n system is now shown explicitly. For the amplitudes in the form $`M_m^{S_z^{},S_z}`$
$`m=0,ϵ^{}`$
$$M_0^{0,1}=\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)\left[S_{\mathrm{}}(1)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(1)+S_{\mathrm{}}(0)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(0)\right]$$
(38)
$$M_0^{1,0}=\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)\left(S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right)$$
(39)
$`m=0,ϵ^+`$
$$M_0^{0,1}=\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)\left[S_{\mathrm{}}(1)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(1)S_{\mathrm{}}(0)\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(0)\right]$$
(40)
$$M_0^{1,0}=\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^0(𝐩)\left(S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right)$$
(41)
$`m=1,ϵ^{}`$
$$M_1^{0,0}=\sqrt{\frac{3}{2}}Y_{\mathrm{}}^1(𝐩)\left[B_{\mathrm{}}^1(1)+B_{\mathrm{}}^1(0)\right]$$
(42)
$$M_1^{1,1}=\sqrt{\frac{3}{2}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1(1)$$
(43)
$`m=1,ϵ^+`$
$$M_1^{1,1}=\sqrt{\frac{3}{2}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1(1)$$
(44)
$`m=1,ϵ^{}`$
$$M_1^{1,1}=\sqrt{\frac{3}{2}}\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^0(1)$$
(45)
$`m=1,ϵ^+`$
$$M_1^{0,0}=\sqrt{\frac{3}{2}}Y_{\mathrm{}}^11(𝐩)\left[B_{\mathrm{}}^1(1)B_{\mathrm{}}^1(0)\right]$$
(46)
$$M_1^{1,1}=\sqrt{\frac{3}{2}}Y_{\mathrm{}}^1(𝐩)B_{\mathrm{}}^1(1)$$
(47)
$`m=2,ϵ^{}`$
$$M_2^{0,1}=\sqrt{3}Y_{\mathrm{}}^2(𝐩)\left[B_{\mathrm{}}^1(1)+B_{\mathrm{}}^1(0)\right]$$
(48)
$`m=2,ϵ^+`$
$$M_2^{0,1}=\sqrt{3}Y_{\mathrm{}}^2(𝐩)\left[B_{\mathrm{}}^1(1)B_{\mathrm{}}^1(0)\right].$$
(49)
For the capture rates we have
$$S_z=1;\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(1)S_{\mathrm{}}(0)\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(0)\right]^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1(1)\right]^2$$
$$+3\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^2(1)+B_{\mathrm{}}^2(0)\right]^2$$
(50)
$$S_z=0;2\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right]^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1(1)\right]^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1(0)\right]^2$$
(51)
$$S_z=1;\underset{\mathrm{}=0}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(1)+S_{\mathrm{}}(0)+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0(0)\right]^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1(1)\right]^2$$
$$+3\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^2(1)B_{\mathrm{}}^2(0)\right]^2.$$
(52)
Using plane waves for $`\mathrm{}1`$
$$S_z=1;\left\{S_0(1)S_0(0)+\frac{1}{\sqrt{2}}\left[B_0^0(1)B_0^0(0)\right]\right\}^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2+12\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^2\right]^2$$
(53)
$$S_z=0;2\left[S_0(1)\sqrt{2}B_0^0(1)\right]^2+2\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right]^2+6\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2$$
(54)
$$S_z=1;\left\{S_0(1)+S_0(0)+\frac{1}{\sqrt{2}}\left[B_0^0(1)+B_0^0(0)\right]\right\}^2+4\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0\right]^2+3\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2$$
(55)
Adding, and assuming all magnetic projections of the deuteron to be equally populated, we find
$$2\left[S_0(1)+\frac{1}{\sqrt{2}}B_0^0(1)\right]^2+2\left[S_0(0)+\frac{1}{\sqrt{2}}B_0^0(0)\right]^2+2\left[S_0(1)\sqrt{2}B_0^0(1)\right]^2$$
$$+6\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left\{\left[S_{\mathrm{}}\right]^2+\left[B_{\mathrm{}}^0\right]^2\right\}+12\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2+12\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^2\right]^2.$$
(56)
Also useful are the expressions for a given photon polarization.
$`ϵ^{}`$
$$S_z=1;\frac{3}{2}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2+12\underset{\mathrm{}=2}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^2\right]^2$$
(57)
$$S_z=0;\left|S_0(1)\sqrt{2}B_0^0(1)\right|^2+\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right]^2+6\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2$$
(58)
$$S_z=1;\left|S_0(1)+S_0(0)+\frac{1}{\sqrt{2}}\left[B_0^0(1)+B_0^0(0)\right]\right|^2+4\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}+\frac{1}{\sqrt{2}}B_{\mathrm{}}^0\right]^2+\frac{3}{2}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2$$
(59)
$`ϵ^+`$
$$S_z=1;\left|S_0(1)S_0(0)+\frac{1}{\sqrt{2}}\left[B_0^0(1)B_0^0(0)\right]\right|^2+\frac{3}{2}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2$$
(60)
$$S_z=0;\left|S_0(1)\sqrt{2}B_0^0(1)\right|^2+\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[S_{\mathrm{}}(1)\sqrt{2}B_{\mathrm{}}^0(1)\right]^2$$
(61)
$$S_z=1;\frac{3}{2}\underset{\mathrm{}=1}{\overset{\mathrm{}}{}}\left[B_{\mathrm{}}^1\right]^2.$$
(62)
### 2.4 Discussion
The ratio of the triplet to singlet rates, in the limit of s-wave contributions and an s-wave deuteron only, and assuming equal singlet and triplet scattering in the s-wave, is two. The introduction of the higher partial waves, and a difference in the scattering lengths, modifies this ratio but it is still roughly two. Hence there is a greater sensitivity to the triplet scattering length than to the singlet.
The S and D states of the deuteron interfere coherently, with different coefficients depending on whether the initial state is $`S_z=0`$ or $`S_z=\pm 1`$. For this reason the shape of the spectrum is different for the transition from the $`S_z=0`$ sub-state than from the $`S_z=\pm 1`$ states due to the D-state contribution. Figure 1 shows a typical result for the pure $`\mathrm{}=0`$ contribution calculated with an asymptotic $`\mathrm{\Lambda }`$-n wave function.
## 3 Solutions with Exponential Potentials
Rijken et al. have recently fit potential models to the $`\mathrm{\Lambda }`$-nucleon scattering data. We use the phase shifts for the s wave determined by this group to define our potentials so that the asymptotic form which is singular at the origin can be replaced with a more realistic wave function.
We will represent the effective interaction as a sum of exponential potentials. To this end we write the true potential as a Laplace transform
$$V(r)=_0^{\mathrm{}}\lambda (\mu )e^{\mu r}𝑑\mu .$$
(63)
Consider a calculation of this integral using a Gauss-Laguerre integration scheme with a 5 point approximation. The integral has maximum precision for only a single value of $`r`$, and the value that is chosen as typical for $`r`$ represents the scale at which the integration is made. We choose to work, primarily, at the scale of 1 fm, in which case the points in the integration scheme (normally dimensionless) represent inverse ranges in units of fm<sup>-1</sup>. The smallest of these inverse ranges has a value of 0.26356 fm<sup>-1</sup>. Since no exchange with such a small mass is believed to take place in this reaction, one expects to find zero for the coefficient of this term. Indeed, fits to the phase shifts produce very small numbers. We set these values to zero and consider fits with only 4 parameters. Thus, the potential is parameterized with the form
$$V(r)=\underset{i=1}{\overset{4}{}}\lambda _ie^{\mu _ir}.$$
(64)
In this case the longest range entering into the problem is that of the second Gauss-Laguerre point, which has an inverse range of 1.41340 fm<sup>-1</sup> and corresponds very well with two pion masses. The other ranges are 3.59642 fm<sup>-1</sup>, 7.08581 fm<sup>-1</sup> and 12.64080 fm<sup>-1</sup>. We have also considered the scale of one-half fm (which means that the inverse ranges are doubled in value) and 2 fm (which means that the inverse ranges have been multiplied by 1/2). While we work primarily with the fits on the scale of 1 fm, the alternate fits provide estimates of the model dependence. It is the 2 fm range fit which is more useful for reasons discussed in section 4.
### 3.1 Jost Solutions
Consider the solution of the Schrödinger equation for a sum of exponential potentials:
$$V(r)=\underset{j=1}{\overset{N}{}}\lambda _je^{\mu _jr}.$$
(65)
Jost writes the solution for the s-wave, $`f(p,r)`$, as
$$f(p,r)=e^{ipr}\underset{\gamma }{}C_\gamma (p)e^{m_\gamma r}$$
(66)
where the subscript $`\gamma `$ is a compound index representing a set of N integers. For example, for a three-term potential
$$\gamma [j,k,l],j,k,l=0,1,2\mathrm{}.$$
(67)
The coefficients $`C_\gamma (p)`$ are given by the recursion relation
$$C_{[j,k,l]}(p)=\frac{\lambda _1C_{[j1,k,l]}(p)+\lambda _2C_{[j,k1,l]}(p)+\lambda _3C_{[j,k,l1]}(p)}{m_\gamma (m_\gamma +2ip)},$$
(68)
where
$$m_\gamma m_{[j,k,l]}j\mu _1+k\mu _2+l\mu _3.$$
(69)
The recursion is started with
$$C_{[0,0,0]}=1,C_{[1,k,l]}=C_{[j,1,l]}=C_{[j,k,1]}=0$$
(70)
and is built up by first computing all coefficients with the sum of indices equal to one, then two, etc. with no negative index.
The solution with the proper boundary condition at the origin for an incoming spherical wave with unit amplitude at infinity is
$$\psi (p,r)=\frac{f(p,r)S(p)f(p,r)}{2ipr},$$
(71)
where
$$S(p)=\frac{f(p,0)}{f(p,0)}.$$
(72)
These expressions can be used to calculate the values of the S-matrix for any value of $`p`$.
In order to calculate the overlap with the deuteron wave function, we require the wave function for real (positive) values of $`p`$. In this case we can write (for real $`\lambda _j`$)
$$f(p,r)=f^{}(p,r)\mathrm{with}S(p)=e^{2i\delta (p)}.$$
(73)
We can now write the wave function (Eq. 71) as
$$\psi (p,r)=\frac{e^{i\delta (p)}}{2ipr}\left[e^{i\delta (p)}e^{ikr}\underset{\gamma }{}C_\gamma (p)e^{m_\gamma r}e^{i\delta (p)}e^{ikr}\underset{\gamma }{}C_\gamma (p)e^{m_\gamma r}\right].$$
(74)
Note that the lowest order term is given by
$$\frac{e^{i\delta (p)}\mathrm{sin}(pr+\delta (p))}{pr},$$
(75)
which is identical to the asymptotic wave function. We shall use these results to represent the $`\mathrm{\Lambda }`$-n wave functions.
### 3.2 Fits to the Phase Shifts
Potentials were fit to the phase shifts for each of the six cases of Ref. , for both the singlet and triplet states and for each of the scales mentioned above. Even though the interaction of interest is for $`\mathrm{\Lambda }`$-neutron scattering, the fits were made to the $`\mathrm{\Lambda }`$-proton phase shifts since they were more closely related to data. The results are shown in Fig. 2 for the triplet case. The potentials themselves are shown in Fig. 3. Table 1 lists the strengths from the fit for the scale of 1 fm. It is this fit which is used in the remainder of the paper unless otherwise noted.
Table 2 summarizes the scattering lengths and effective ranges corresponding to these fits. Also given are the scattering lengths found by Rijken et al.with the original potential.
In Fig. 4 are plotted the full wave function and the asymptotic wave function for the singlet “a” case of Ref. .
## 4 Results
We now turn to problems in the analysis and some possible solutions. The first issue is the use of the effective range expansion (ERE). In the $`\pi ^{}dnn\gamma `$ measurement the ERE is adequate to describe the s-wave phase shift over the full range of data being analyzed. As is shown in Figure 5, this may not be the case for the n$`\mathrm{\Lambda }`$ measurement. Certainly it is not adequate over the full range, and it may be somewhat questionable even if only the upper 10 MeV of the spectrum is used. Of course the ERE is only useful if the asymptotic wave functions are used or if the short-range corrections are made with a technique similar to that employed in Ref. .
To find an alternative to the ERE we examined the parameters which came from the fitting of the exponential potentials to the phase shifts of Rijken et al. It was found that an acceptable fit could be obtained with the longest range parameter held fixed. Plots of the other three parameters (see Fig. 6 for the triplet case) show a linear behavior with case number. We therefore defined continuous variables, which reproduce the phase shifts (to a good approximation) at the integers 1 through 6 (corresponding to cases a through f), and create an interpolation and extrapolation procedure.
We define strengths
$$\lambda _i=c_i+xd_i$$
(76)
for i=2, 3, 4. Here $`x=x_s`$ for the singlet case and $`x=x_t`$ for the triplet case. The values of $`c_i`$, $`d_i`$ and $`\lambda _1`$ are given in Table 3. While this method allows a range of scattering lengths to be produced with an appropriate short-range wave function, there are limitations. Since we would like to be able to study scattering lengths outside the range given in Ref. , we wish to consider values of $`x`$ outside the range (1-6). Indeed this does give an extension of the range of scattering lengths, but for the 1 fm scale the range is limited. Outside this range the scattering lengths return to previous values. This constraint dictates the limits of the analysis shown in the figures to follow. It was found for the fits on the scale of 1/2 fm that the problem was greater, so that the analysis would be restricted to an even smaller area. The 2 fm range fits were used for the model dependence estimates because, in this case, a larger range of scattering lengths was feasible. However, the scattering lengths do not reproduce those of Ref. as well as the 1 fm scale, as may be seen in Table 2.
In performing the analysis one may use the asymptotic wave functions or the full wave functions. The use of the full wave functions should be preferable, but one can obtain an idea of the sensitivity to the details of the full wave function by comparing results with those for the asymptotic wave functions. Figure 7 shows the sum of the $`S_z=0`$ and $`S_z=\pm 1`$ s-wave contributions to the spectra as shown in Fig. 1 for the asymptotic wave functions (solid lines) and the full wave functions (dashed lines) for the parameters of the singlet “a” and “f” cases.
Perhaps the largest problem in the measurement is the separation of the singlet and triplet scattering lengths in the final state. If no spin degrees of freedom are measured then the singlet and triplet states contribute incoherently to the rate, and it might seem impossible to separate them. However, because the interference between the S and D states of the deuteron differs, the shapes of the singlet and triplet spectra are different, as illustrated in Fig. 8. Because the triplet gives a larger contribution than the singlet, a greater sensitivity to the triplet scattering is seen in the fits of the data and pseudo-data in the figures which follow.
Figure 9 shows an analysis of the data of Gall et al.. Since this experiment was only a feasibility study, one can not expect to obtain much information about the scattering lengths. It is interesting to see, however, that the values are in the range expected. The open circle in the upper left hand corner shows the point of minimum $`\chi ^2`$.
Figure 10 shows a similar analysis of the Gall et al. data with the 2 fm scale fits. In this case a larger range of scattering lengths can be studied so that an error estimate can be given. One-standard-deviations limits can be read from the graph (the inner contour) with the triplet length lying between –1.3 and –2.6 fm and the singlet length between –0.2 and –6.3 fm.
Figure 11 provides a comparison of the experimental and theoretical spectra for the scattering lengths corresponding to the minimum $`\chi ^2`$ (=37.8 for 37 data points) at $`a_s=2.96`$ and $`a_t=1.72`$ for the 1 fm scale.
Figure 12 presents a similar analysis of a pseudo data spectrum. The pseudo data were generated from a selected spectrum by including errors chosen from a Gaussian distribution such that they have a value of 3% of the rate at the maximum and are proportional to the square root of the rate at other points, as would be the case with errors dominated by counting statistics. The analyses shown were made over the full range of the Gall et al. data from 255 to 293 MeV unless otherwise specified. The case chosen for presentation in Fig. 12 is a favorable one in the sense that the scattering lengths used for generation of the pseudo-data (solid circle) fall near the center of the inner ellipse. As is necessary statistically, most of the results do fall nearer the edge of this ellipse with about 1/3 falling outside. One can observe from this figure that the experimental uncertainty in such an experiment would be $`\pm 0.3`$ fm for the triplet scattering length and $`\pm 0.8`$ fm for the singlet.
It is possible to use such an analysis of pseudo data to estimate errors due to model assumptions. For the following we use the same pseudo data that were described above. Figure 13a shows the results of an analysis using asymptotic wave functions with phase shifts determined from the Jost solutions. Figure 13b shows an analysis with the additional approximation that the phase shifts are taken from the ERE. As can be seen a significant error results.
Use of the asymptotic wave functions is an extreme approximation. Although we do not know exactly which more realistic wave function should be used, we can choose one among the Jost models we have been considering. To this end we analyze the spectrum created with the 1 fm scale using the 2 fm scale spectra. The result can be seen in Fig. 14. The positions of the singlet and triplet scattering lengths at the minimum are –1.613 and –1.884 fm to be compared with the values found with the (1 fm scale) analysis above of –1.896 and –2.031 fm (the input values are –1.968 and –1.827 fm). Thus we find model errors of $`\pm 0.28`$ fm for the singlet and $`\pm 0.15`$ fm for the triplet case. We will take these as estimates of the model (theoretical) errors.
Figure 15 shows the same type of analysis but including only the upper 10 MeV of the spectrum in the comparison. It is seen that the results are very similar, due perhaps to the fact that the main difference in shapes between singlet and triplet occurs in this region (see Fig. 8). The analysis over this range may offer experimental advantages in addition, since the background due to neutral pion decay is less severe.
We have seen that it is difficult to determine the scattering lengths individually. It is possible to separate the singlet and triplet states with the use of spin information from the deuteron initial state or the photon final state. Figure 16 shows the spectra for the six possible combinations of these spin projections for four selected parameter sets. While the overall normalization of these curves is arbitrary, the relative normalization among them is correct. The upper left hand curve (for $`S_z=+1`$ and right circularly polarized photon) shows the maximum sensitivity since it corresponds to the case where the singlet and triplet final states are coherent in a destructive manner. The disadvantage is that this same cancellation leads to a small rate if the singlet and triplet scatterings are not too dissimilar, as can be expected from SU(6) symmetry. However, this small quantity is a direct measure of the difference between the singlet and triplet lengths. Note that the spectra from an initial state $`S_z=0`$ are independent of the singlet parameters.
We can express the results in terms of more conventional variables. Figure 17 illustrates the spectrum expected from a pure vector polarized deuteron target with the asymmetry defined as
$$A_y\frac{\mathrm{\Gamma }(S_z=+1)\mathrm{\Gamma }(S_z=1)}{\mathrm{\Gamma }(S_z=+1)+\mathrm{\Gamma }(S_z=1)}.$$
(77)
We see that a measurement with precision of the order of 2% is needed near the peak of the spectrum or 10% at the upper part of the spectrum. What is being searched for is a deviation of the asymmetry from –1 or, in other words, a non-null value of the rate from the initial state $`S_z=+1`$. Such an observation would prove that the two scattering lengths are unequal. The dashed curve corresponds to a difference of 0.1 fm, the dash-dot curve to 0.66 fm, the solid curve to 0.91 fm and the dotted curve to 1.46 fm. This result is almost completely free from the model dependent effects discussed previously. The observation of a difference from –1 in the asymmetry would be a direct indication of a difference in the scattering lengths.
Figure 18 shows the spectra expected from measurements in which the circular polarization of the final photon is measured. The asymmetry is defined as
$$A_ϵ\frac{\mathrm{\Gamma }(ϵ=+1)\mathrm{\Gamma }(ϵ=1)}{\mathrm{\Gamma }(ϵ=+1)+\mathrm{\Gamma }(ϵ=1)}.$$
(78)
Here again the same order of accuracy is needed to separate the dotted and dashed curves.
## 5 Conclusions
A formalism has been presented with attention to the spin degrees of freedom. Due to the difference in shape of the spectrum for the reaction proceeding from the zero magnetic quantum number projection of the deuteron from that with magnetic quantum numbers of $`\pm 1`$, the singlet-triplet separation can be made, although the difference is not large and the triplet state tends to dominate.
An analysis of the present data which exist for this reaction has been made. Even though the uncertainties in the data from this feasibility experiment are large, some information can be obtained; in particular, separate numbers can be extracted for the singlet and triplet scattering lengths (with large overlapping errors). We find for the singlet scattering length in the range $`0.15`$ to $`5.0`$ and the triplet value from $`1.3`$ to $`2.65`$ (from the 2 fm range fit).
While this precision may seem rather modest, it is useful to compare with current values in the literature. Tan obtained a value (believed to be mostly the triplet scattering length) of $`2.0\pm 0.5`$ fm from an analysis of data on the reaction $`K^{}d\pi ^{}p\mathrm{\Lambda }`$. Recent data on the reaction $`pppK^+\mathrm{\Lambda }`$, in conjunction with previous elastic scattering data, lead to a value for the spin-averaged scattering length of $`2.0\pm 0.2`$ fm. The importance of the influence of the third strongly interacting particle in the final state is difficult to estimate and the errors are experimental only.
In a three-body calculation for the hyper-triton bound state Miyagawa concludes that the best values of the singlet scattering length lie between $`2.7`$ fm and $`2.4`$ fm and the triplet between $`1.6`$ fm and $`1.3`$ fm, conforming to the “f” solution of Rijken et al..
An analysis of pseudo data for a $`K^{}d\mathrm{\Lambda }n\gamma `$ experiment with a cross-comparison of models leads to an estimate of the model dependence of the order of $`\pm 0.2`$ fm. The same type of study gives an estimate of the error in such an experiment of $`\pm 0.3`$ fm for the triplet scattering length and $`\pm 0.8`$ fm for the singlet. These estimates are for measurements without spin information and assume an uncertainty of 3% for the maximum rate in the spectrum.
It was shown that spin information in either the initial or final state would be valuable in separating the scattering lengths. If it were possible to perform the capture experiment on a deuteron target of purely magnetic quantum number zero the triplet scattering length alone would be measured.
The use of a polarized deuteron target would also allow the separation since the singlet and triplet states interfere destructively or constructively according to the relative alignment of the deuteron spin along or against the direction of the photon.
Since the previous feasibility study showed that the measurement was possible but very difficult, one might ask if the increased difficulty due to the requirement of polarizing the deuteron might not make it impossible. Certainly it adds another constraint, especially if it is necessary to add high-Z material which would preferentially capture the Kaons. In this regard we mention a possible polarized deuteron target which is made from a hydrogen-deuteron molecular system. It has a high density and contains no heavy materials.
For a measurement of the circular polarization of the photon, the spectra show a significant sensitivity of a similar type.
This work was supported by the U. S. Department of Energy and the National Science Foundation. HKH thanks the Korean Scientific and Engineering Foundation for their support. The research of BFG was supported by the Department of Energy (LA-UR-99-6261). |
warning/0001/cond-mat0001161.html | ar5iv | text | # Multiple scaling regimes in simple aging models
## Abstract
We investigate aging in glassy systems based on a simple model, where a point in configuration space performs thermally activated jumps between the minima of a random energy landscape. The model allows us to show explicitly a subaging behavior and multiple scaling regimes for the correlation function. Both the exponents characterizing the scaling of the different relaxation times with the waiting time and those characterizing the asymptotic decay of the scaling functions are obtained analytically by invoking a ‘partial equilibrium’ concept.
PACS numbers: 02.50.-r, 75.10.Nr, 05.20.-y
The dynamics of glassy materials can be strongly dependent on the history of glass formation . Generally speaking, one finds that the relaxation dynamics becomes increasingly slower with the “age” of the system, that means with the time $`t_w`$ expired since the material was brought into the glassy state. Such aging phenomena have been identified in many systems and various dynamical probes (for a recent review, see e.g. ). Prominent examples are shear stress relaxations in structural glasses , thermoremanent magnetizations in spin glasses , and electric field relaxations in dipolar glasses . A convenient way to quantify aging in such experiments is to disturb the probe at time $`t_w`$ by a sudden change of external field, and to measure the response $`R(t_w+t,t_w)`$ at a later time $`t_w+t`$. Often, the characteristic relaxation time grows proportionally to the age $`t_w`$. When $`t`$ is larger than all microscopic times associated with fast time-translational invariant relaxations, one then expects $`R(t_w+t,t_w)`$ to depend on the ratio $`t/t_w`$ only, i.e. $`R(t_w+t,t_w)=F(t/t_w)`$.
In principle, however, one cannot rule out other scaling forms, as e.g. $`R(t_w+t,t_w)=F(t/t_w^\mu )`$ with $`\mu >0`$ being different from one. In particular the case $`\mu <1`$, which has been called ‘subaging’ because the effective relaxation time grows more slowly than the age of the system, seems to be of experimental relevance . It is moreover possible that there exist, for given waiting time $`t_w`$, various scaling regimes in time $`t`$, which are governed by different relaxation times $`t_w^{\mu _s}`$, $`s=1,2,\mathrm{}`$. More precisely, depending on how $`t`$ is scaled with $`t_w`$, one can obtain different asymptotic scaling functions in the limit $`t_w\mathrm{}`$. For example, for $`\mu _1>\mu _2>0`$, one may find $`R(t_w+\mathrm{\Lambda }_1t_w^{\mu _1},t_w)F_1(\mathrm{\Lambda }_1)`$ and $`R(t_w+\mathrm{\Lambda }_2t_w^{\mu _2},t_w)F_2(\mathrm{\Lambda }_2)`$ when $`t_w\mathrm{}`$. In fact, the occurrence of different scaling functions being associated with various time regimes has recently been conjectured on the basis of analytical results for the Langevin dynamics of mean-field spin glass models . So far, however, it was not possible to validate these conjectures, or to exemplify them in some reasonable phenomenological models.
In this Letter we will discuss a model that allows us to demonstrate for the first time explicitly the possible occurrence of subaging behavior and multiple scaling regimes. This model, which has a strong resemblance to the previously studied “trap model” , is motivated by the simple and widespread view that glassy dynamics may be described by a thermally activated motion of a point (“particle”) that jumps among the deep (free) energy minima $`E_i`$ of a complex configuration space. According to extreme value statistics one may expect the distribution $`\rho (E)`$ of these deep minima to be exponential, and indeed, mean-field theories of spin glasses and recent results from molecular dynamics simulations suggest this to be the case.
To be specific, let us consider a $`d`$-dimensional cubic lattice and assign to each lattice site $`i`$ an energy $`E_i`$, $`\mathrm{}<E_i0`$, drawn from the distribution $`\rho (E)=T_\mathrm{g}^1\mathrm{exp}(E/T_\mathrm{g})`$. The particle jumps among nearest neighbor sites only, and the jump rate from site $`i`$ to a neighboring site $`j`$ is
$$w_{i,j}=\nu \mathrm{exp}\left(\beta [\alpha E_j(1\alpha )E_i]\right),$$
(1)
where the “attempt frequency” $`\nu 1`$ sets our time unit, $`\beta ^1T`$ is the temperature (or thermal energy), and the parameter $`\alpha `$ specifies how the energies of the initial and target site are weighted. In order for the $`w_{i,j}`$ to obey detailed balance, $`\alpha `$ can assume any real value, but on physical grounds it is reasonable to restrict $`\alpha `$ to the range $`0\alpha 1`$. Independent of $`\alpha `$, the system undergoes a “dynamical phase transition” at $`T=T_\mathrm{g}`$: In the high-temperature phase, where $`T>T_\mathrm{g}`$, the probability $`\phi (E)\rho (E)\mathrm{exp}(\beta E)`$ for finding the system in a state with energy $`E`$ is normalizable and thermal equilibrium will be approached with a characteristic equilibration time that diverges for $`TT_\mathrm{g}`$. By contrast, in the glassy phase, where $`T<T_\mathrm{g}`$ and $`\phi (E)`$ is not normalizable, the dynamics never becomes stationary but ‘ages’.
It is important at this point to stress the differences between the above model and the earlier studied trap model . In the latter, the jump rates depend only on the energy of the initial site corresponding to $`\alpha =0`$ in eq. (1), and this allows a straightforward mapping onto a continuous time-random walk with a waiting time distribution decaying as a power law. Much more important,
the trap model was so far considered only on a mean-field level corresponding to an “annealed situation”, where the site energies are drawn anew after each jump.
In order to study aging effects in the glassy phase, we focus, as in the trap model, on the (disorder averaged) probability $`\mathrm{\Pi }(t_w+t,t_w)`$ that the system does not change its state between $`t_w`$ and $`t_w+t`$. For particles hopping on a lattice, this can be interpreted as a dynamical structure factor . Initially, the particle is located on any of the sites and then it starts to explore the configuration space at some $`T<T_\mathrm{g}`$. Physically, this means that we are considering an instantaneous quench from $`T=\mathrm{}`$.
Our idea to explore the scaling properties of $`\mathrm{\Pi }(t_w+t,t_w)`$ for both $`t`$ and $`t_w`$ becoming large is based on the following “partial equilibrium” concept: From the time after the quench up to $`t_w`$ the particle has followed a Brownian path in configuration space that on average consists of $`S(t_w)`$ distinct and mutually connected sites. On a typical path with $`SS(t_w)`$ sites then, it is reasonable to think that the particles should effectively equilibrate, i.e. the probability to be on a particular site $`j`$ of the path may be approximated by $`\tau _j/_{k=1}^S\tau _k`$, where $`\tau _j\mathrm{exp}(\beta E_j)`$ ($`1\tau _j<\mathrm{}`$). Conditioned on being at the site $`j`$, the system has a probability $`\mathrm{exp}(t_{n_j}w_{j,n_j})=\mathrm{exp}(t\tau _j^{\alpha 1}_{n_j}\tau _{n_j}^\alpha )`$ not to change state within time $`t`$, where the sum over $`n_j`$ runs over all nearest neighbor sites of $`j`$. Exactly two of these neighboring sites are considered to belong to the Brownian path in view of its one-dimensional topology. The remaining $`2(d1)`$ neighboring sites are assumed to have not been visited before. Hence, we may deduce the scaling properties of $`\mathrm{\Pi }(t_w+t,t_w)`$ from
$$\stackrel{~}{\mathrm{\Pi }}(t_w+t,t_w)\frac{_{j=1}^{S(t_w)}\tau _j\mathrm{exp}\left(t\tau _j^{\alpha 1}_{n_j}\tau _{n_j}^\alpha \right)}{_{k=1}^{S(t_w)}\tau _k},$$
(2)
where $`\mathrm{}`$ denotes an average over $`(2d1)S(t_w)`$ uncorrelated random numbers $`\tau _j`$ that are distributed according to the power law $`\varphi (\tau )=\theta \tau ^{1\theta }`$ with $`\theta T/T_\mathrm{g}`$.
Clearly, the partial equilibrium concept is an approximation that needs to be tested. To this end we have determined $`\mathrm{\Pi }(t_w+t,t_w)`$ and $`S(t_w)`$ in $`d=1,3`$ for various values of $`\theta `$ and $`\alpha `$ by Monte Carlo simulations. Then we took $`S(t_w)`$ from these simulations to calculate $`\stackrel{~}{\mathrm{\Pi }}(t_w+t,t_w)`$ from eq. (2). The disorder average in this simple numerical evaluation was performed over a set of realizations independent of the ones taken in the simulations.
Figure 1 shows the results for one representative parameter pair $`(\theta ,\alpha )=(1/4,3/8)`$. The data have been collected as functions of $`t`$ for a broad range of fixed waiting times $`t_w`$ and are plotted already in scaled form as functions of $`\mathrm{\Lambda }_1=t/t_w^{\mu _1}`$ with exponents $`\mu _1=\mu _1(\theta ,\alpha )`$ being specified below. As can be seen from the figure, the data for both $`\mathrm{\Pi }(t_w+t,t_w)`$ and $`\stackrel{~}{\mathrm{\Pi }}(t_w+t,t_w)`$ collapse onto single curves $`F_1(\mathrm{\Lambda }_1)`$ and $`\stackrel{~}{F}_1(\mathrm{\Lambda }_1)`$, respectively. Although the two scaling functions are different, their asymptotic behavior for large $`\mathrm{\Lambda }_1`$ is the same, $`F_1(\mathrm{\Lambda }_1)\stackrel{~}{F}_1(\mathrm{\Lambda }_1)\mathrm{\Lambda }_1^\delta `$ (see the solid lines in Fig. 1). The values of the exponent $`\delta =\delta (\theta ,\alpha )`$ are given below.
It is important to inspect also more closely the ‘short’ time regime, where the complementary probability $`1\mathrm{\Pi }(t_w+t,t_w)`$ for the system to change state between $`t_w`$ and $`t_w+t`$ is small. This complementary probability can be as relevant as $`\mathrm{\Pi }(t_w+t,t_w)`$ dependent on the physical quantity being measured. Scaling plots of $`1\mathrm{\Pi }(t_w+t,t_w)`$ and $`1\stackrel{~}{\mathrm{\Pi }}(t_w+t,t_w)`$ as functions of $`\mathrm{\Lambda }_1=t/t_w^{\mu _1}`$ in $`d=1,3`$ are also shown in Fig. 1. Again there is a good data collapse and for $`\mathrm{\Lambda }_10`$ we find $`1F_1(\mathrm{\Lambda }_1)1\stackrel{~}{F}_1(\mathrm{\Lambda }_1)\mathrm{\Lambda }_1^ϵ`$ with exponents $`ϵ=ϵ(\theta ,\alpha )`$ given below. An analogous overall behavior as displayed in Fig. 1 was found also for other pairs $`(\theta ,\alpha )`$ (with $`0<\theta <1`$, $`0\alpha 1`$). We thus conclude that the partial equilibrium concept not only yields the correct scaling behavior (same $`\mu _1`$ exponents) but also the correct asymptotics of the scaling functions . However, a study of the ‘participation ratios’ (see for their definition) shows that the partial equilibrium concept is not exact, even for large times .
Now we turn to the analytical study of eq. (2). Since $`\stackrel{~}{\mathrm{\Pi }}(t_w+t,t_w)`$ depends on $`t_w`$ only via $`S(t_w)`$ let us first discuss the scaling of $`S(t_w)`$ with time $`t_w`$. For $`\alpha =0`$ this problem has been addressed some time ago (see e.g. ) and for $`t_w\mathrm{}`$ one finds
$$S(t_w)t_w^\gamma ,\gamma =\{\begin{array}{cc}\frac{d\theta }{d+(2d)\theta },& 1d<2\hfill \\ \theta ,& d>2\hfill \end{array}$$
(3)
(In $`d=2`$ there are logarithmic corrections, $`S(t_w)[t_w/\mathrm{log}t_w]^\theta `$.) For $`0<\alpha 1`$ one expects (3) not to change, since $`\alpha `$ only affects the nearest-neighbor hopping rates but not the transport properties on large length scales. In fact, in our simulations we always found eq. (3) to hold true. Let us note also that in $`d=1`$ one can give a simple finite-size scaling argument to show that (3) remains valid for $`0<\alpha 1`$.
Next we derive the scaling properties of $`\stackrel{~}{\mathrm{\Pi }}`$ as a function of $`t`$ and $`S=S(t_w)`$, and then use eq. (3) to obtain the corresponding scaling properties of $`\stackrel{~}{\mathrm{\Pi }}`$ as a function of $`t`$ and $`t_w`$. When replacing the denominator in eq. (2) by $`_0^{\mathrm{}}𝑑\lambda \mathrm{exp}(\lambda _{k=1}^S\tau _k)`$, the average over the $`\tau _j`$ can to a large extent be factorized, and one obtains the following asymptotic formula valid in the limit of large $`S`$
$`\stackrel{~}{\mathrm{\Pi }}`$ $``$ $`{\displaystyle \frac{\theta ^3\stackrel{~}{S}}{\kappa }}{\displaystyle _0^{\mathrm{}}}𝑑\lambda e^{\lambda ^\theta \stackrel{~}{S}}{\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau _1}{\tau _1^{1+\theta }}}e^{\lambda \tau _1}{\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau _2}{\tau _2^{1+\theta }}}e^{\lambda \tau _2}`$ (6)
$`\times {\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau }{\tau ^\theta }}e^{\lambda \tau }\left[f\left({\displaystyle \frac{t}{\tau ^{1\alpha }}}\right)\right]^{2(d1)}\mathrm{exp}(t{\displaystyle \frac{\tau _1^\alpha +\tau _2^\alpha }{\tau ^{1\alpha }}}),`$
$`f(x)\theta {\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau }{\tau ^{1+\theta }}}\mathrm{exp}(x\tau ^\alpha ).`$
Here $`\kappa \mathrm{\Gamma }(1\theta )`$, where $`\mathrm{\Gamma }(.)`$ is the Gamma function, and $`\stackrel{~}{S}\kappa S`$. Note that for $`t=0`$ and $`\stackrel{~}{S}\mathrm{}`$, eq. (6) yields the correct normalization $`\stackrel{~}{\mathrm{\Pi }}1`$.
After two transformations $`\lambda \stackrel{~}{S}\lambda ^\theta `$ and $`\tau \stackrel{~}{S}^{1/\theta }\tau `$ we can identify $`\mathrm{\Lambda }_1=t/\stackrel{~}{S}^{(1\alpha )/\theta }`$ as a scaling variable corresponding to a first characteristic time $`t_1\stackrel{~}{S}(t_w)^{(1\alpha )/\theta }t_w^{\gamma (1\alpha )/\theta }t_w^{\mu _1}`$. We thus obtain
$$\mu _1=\frac{\gamma (1\alpha )}{\theta }$$
(7)
with $`\gamma `$ from eq. (3). This exponent $`\mu _1`$ has been used to collapse the data in Fig. 1, i.e. we took $`\mu _1=(1\alpha )/(1+\theta )`$ in $`d=1`$ and $`\mu _1=(1\alpha )`$ in $`d=3`$. Based on the equilibrium concept it is is easy to show that $`t_1`$ has a simple physical interpretation: It scales as the typical maximum trapping time $`t_{\mathrm{max}}(t_w)`$ encountered after $`t_w`$. In $`d=1`$ one thus finds a subaging behavior even for $`\alpha =0`$ since the deepest trap is visited a large number of times $`N(t_w)t_w^{\theta /(1+\theta )}`$, so that $`t_1t_{\mathrm{max}}(t_w)t_w/N(t_w)t_w`$. Similarly, for $`\alpha 0`$, the deepest trap is revisited a large number of times in all dimensions $`d`$ due to a strong backward jump correlation when the particle leaves a site with low energy.
The scaling function in the first time domain $`t=\mathrm{\Lambda }_1t_w^{\mu _1}`$ reads
$`\stackrel{~}{F}_1(\mathrm{\Lambda }_1)`$ $`=`$ $`{\displaystyle \frac{\theta ^3}{\kappa }}{\displaystyle _0^{\mathrm{}}}𝑑\lambda e^{\lambda ^\theta }{\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau _1}{\tau _1^{1+\theta }}}{\displaystyle _1^{\mathrm{}}}{\displaystyle \frac{d\tau _2}{\tau _2^{1+\theta }}}`$ (9)
$`\times {\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{d\tau }{\tau ^\theta }}e^{\lambda \tau }\left[f\left({\displaystyle \frac{\mathrm{\Lambda }_1}{\tau ^{1\alpha }}}\right)\right]^{2(d1)}\mathrm{exp}(\mathrm{\Lambda }_1{\displaystyle \frac{\tau _1^\alpha +\tau _2^\alpha }{\tau ^{1\alpha }}}),`$
and has the limiting behavior
$$\stackrel{~}{F}_1(\mathrm{\Lambda }_1)\{\begin{array}{ccc}1c_>\mathrm{\Lambda }_1^{(1\theta )/(1\alpha )},\hfill & \theta >\alpha \hfill & \\ & & \mathrm{\Lambda }_10\hfill \\ 1c_<\mathrm{\Lambda }_1^{\theta /\alpha },\hfill & \theta <\alpha \hfill & \\ c_{\mathrm{}}\mathrm{\Lambda }_1^{\theta /(1\alpha )}\hfill & \mathrm{\Lambda }_1\mathrm{}\hfill & \end{array}$$
(10)
The constants $`c_>`$, $`c_<`$, and $`c_{\mathrm{}}`$ can be expressed in terms of $`\alpha `$, $`\theta `$ and $`d`$ but are not of interest here. Equation (10) yields the exponents $`\delta =\theta /(1\alpha )`$ and $`ϵ=(1\theta )/(1\alpha )`$ taken in Fig. 1 to characterize the decay of $`\mathrm{\Pi }`$ and rise of $`1\mathrm{\Pi }`$, respectively. For $`\alpha =0`$, one recovers the results of the annealed model , namely $`\delta =\theta `$ and $`ϵ=1\theta `$.
For $`\theta >\alpha `$ the function $`\stackrel{~}{F}_1(\mathrm{\Lambda }_1)`$ describes the scaling properties of $`\stackrel{~}{\mathrm{\Pi }}`$ completely. However, based on eq. (6) one finds that for $`\theta <\alpha <1/2`$, there exists a second scaling variable $`\mathrm{\Lambda }_2=t/\stackrel{~}{S}^{(12\alpha )/\theta }t/t_w^{\gamma (12\alpha )/\theta }t/t_w^{\mu _2}`$ yielding
$$\mu _2=\frac{\gamma (12\alpha )}{\theta }.$$
(11)
Therefore, a second characteristic time scale $`t_2S^{(12\alpha )/\theta }`$, diverging when $`t_w\mathrm{}`$, governs the behavior when $`\stackrel{~}{\mathrm{\Pi }}`$ is close to one. (Note that for fixed $`\mathrm{\Lambda }_2`$, $`\mathrm{\Lambda }_1=\stackrel{~}{S}^{\alpha /\theta }\mathrm{\Lambda }_20`$ for $`\stackrel{~}{S}\mathrm{}`$). In the new time domain one finds the following generalized scaling form,
$$\stackrel{~}{S}(1\stackrel{~}{\mathrm{\Pi }})=\stackrel{~}{F}_2(\mathrm{\Lambda }_2),\theta <\alpha <1/2,$$
(12)
where $`F_2(.)`$ is given by
$`\stackrel{~}{F}_2(\mathrm{\Lambda }_2)`$ $`=`$ $`{\displaystyle \frac{\psi _{\mathrm{}}\theta ^2\mathrm{\Lambda }_2}{(1\alpha )\kappa }}{\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{d\lambda e^{\lambda ^\theta }}{\lambda ^{\alpha \theta }}}{\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{d\tau e^{\lambda \tau }}{\tau ^{1\alpha +x}}}g(\mathrm{\Lambda }_2\lambda ^{1\alpha }\tau ^\alpha ),`$ (14)
$`g(x)x^{\frac{\alpha \theta }{1\alpha }}{\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{du}{u^{1+\frac{1\theta }{1\alpha }}}}(1e^u)\mathrm{exp}[(x/u)^{\frac{1}{(1\alpha )}}].`$
Here $`\psi _{\mathrm{}}lim_{\overline{\tau }\mathrm{}}\psi (\overline{\tau })/(\theta \overline{\tau }^{1\theta })`$, and $`\psi (\overline{\tau })`$ denotes the probability distribution for the variable $`\overline{\tau }(\tau _1^\alpha +\tau _2^\alpha )^{1/\alpha }`$, i.e. $`\psi (\overline{\tau })\delta (\overline{\tau }[\tau _1^\alpha +\tau _2^\alpha ]^{1/\alpha })`$. From (14) follows
$$\stackrel{~}{F}_2(\mathrm{\Lambda }_2)\{\begin{array}{cc}c_0\mathrm{\Lambda }_2,\hfill & \mathrm{\Lambda }_20\hfill \\ c_<\mathrm{\Lambda }_2^{\theta /\alpha },\hfill & \mathrm{\Lambda }_2\mathrm{}\hfill \end{array}$$
(15)
where $`c_0`$ is a constant dependent on $`\theta `$ and $`\alpha `$. We note that $`1\stackrel{~}{\mathrm{\Pi }}`$ matches continuously as one leaves the short time scaling regime ($`tt_w^{\mu _2}`$) described by $`F_2`$ to enter the regime described by the scaling function $`F_1`$ (where $`tt_w^{\mu _1}`$). Figure 2 shows the short time behavior of $`\stackrel{~}{\mathrm{\Pi }}`$ and $`\mathrm{\Pi }`$, rescaled as in (12) for the same parameters as in Fig. 1. As can be seen from the figure, the data approach the two power laws predicted by eq. (15) for large $`t_w`$.
For clarity, we illustrate the overall behavior of $`\mathrm{\Pi }(t_w+t,t_w)`$ as a function of $`t`$ in Fig. 3. For $`(\theta ,\alpha )`$ values in the two-time-scaling region $`0<\theta <\alpha <1/2`$ (shaded area of the $`\alpha `$-$`\theta `$-diagram shown in the inset), there exist three different $`t`$ regimes: (I ) $`\mathrm{\Pi }(t_w+t,t_w)1\mathrm{const}.t_w^\gamma [t/t_w^{\mu _2}]`$ for $`tt_w^{\mu _2}`$, (II ) $`\mathrm{\Pi }(t_w+t,t_w)1\mathrm{const}.^{}t_w^\gamma [t/t_w^{\mu _2}]^{\theta /\alpha }=1\mathrm{const}.^{}[t/t_w^{\mu _1}]^{\theta /\alpha }`$ for $`t_w^{\mu _2}tt_w^{\mu _1}`$, and (III ) $`\mathrm{\Pi }(t_w+t,t_w)[t/t_w^{\mu _1}]^\delta `$ for $`t_w^{\mu _1}t`$. When $`(\theta ,\alpha )`$ lies in the unshaded area of the $`\alpha `$-$`\theta `$-diagram, the first regime $`tt_w^{\mu _2}`$ does not exist (or, more precisely, it then becomes irrelevant in the limit of large $`t_w`$). Note that for $`t_w\mathrm{}`$ and $`\mathrm{\Lambda }_2=t/t_w^{\mu _2}`$ fixed, the long-time regime in Fig. 3 “moves toward infinity” and the behavior is fully described by the second scaling function $`F_2(\mathrm{\Lambda }_2)`$, while for $`t_w\mathrm{}`$ and $`\mathrm{\Lambda }_1=t/t_w^{\mu _1}`$ fixed the short-time regime in Fig. 3 “moves toward zero”, and the behavior is fully described by the first scaling function $`F_1(\mathrm{\Lambda }_1)`$.
In summary we have shown (i) that generalized trap models can exhibit subaging behavior, induced by multiple visits to the same trap, and (ii) the possible existence
of several distinct scaling regimes in the two-time plane. Such a possibility is of crucial importance for the interpretation of experiments, since the waiting time can typically be varied between one minute and a few days only (with some notable exceptions ). The occurrence of several time regimes then may get masked by an apparent rescaling of the relaxation curves by a single effective value of $`\mu `$. From a theoretical perspective, it would be interesting to study the existence of a generalized Fluctuation-Dissipation Theorem in the aging regime, as it is predicted by mean-field spin-glass models . This problem is intimately related to the validity of the partial equilibrium concept introduced above.
We should like to thank W. Dieterich and E. Pitard for stimulating discussions. P.M. gratefully acknowledges financial support from the Deutsche Forschunggsgemeinschaft (Ma 1636/2-1). |
warning/0001/physics0001004.html | ar5iv | text | # Untitled Document
SINGLE–PASS LASER POLARIZATION OF
ULTRARELATIVISTIC POSITRONS.
Alexander Potylitsyn
Tomsk Polytechnic University,
pr. Lenina 2A, Tomsk, 634050, Russia
e-mail: pap@phtd.tpu.edu.ru
## Abstract
The new method for producing of the polarized relativistic positrons is suggested. A beam of unpolarized positrons accelerated up to a few GeV can be polarized during a head-on collision with an intense circularly polarized lazer wave. After a multiple Compton backscattering process the initial positrons may lose a substantial part of its energy and, as consequence, may acquire the significant longitudinal polarization. The simple formulas for the final positron energy and polarization degree depended on the laser flash parameters have been obtained. The comparison of efficiences for the suggested technique and known ones is carried out. Some advantages of the new method were shown.
The experiments with polarized electron-positron beams in future linear colliders will furnish a means for studying a number of intriguing physical problems . While the problem of generation and acceleration of longitudinally polarized electron beams seems to be solved , the approach for production of polarized positron beams with the required parameters has not been finally defined yet. In \[3-7\] methods were proposed for the generation of longitudinally polarized positrons during $`\mathrm{e}^+\mathrm{e}^{}`$– pair production by circularly polarized photons with the energy $`\omega 10^1`$ MeV, which are, in their turn, generated by either passing electrons with the energy $`10^2`$ GeV through a helical undulator , or through Compton backscattering of circularly polarized laser photons on a beam of electrons with the energy $``$5 GeV , or through bremsstrahlung of longitudinally polarized $``$50 MeV electrons . To achieve the needed intensity of a positron source ($`\mathrm{N}_{\mathrm{e}^+,\mathrm{pol}}10^{10}`$ particles/bunch) it is suggested to use an undulator of the length $`\mathrm{L}>`$ 100 m , or to increase the laser power , or to use a high-current accelerator of polarized electrons .
The present paper considers an alternative way to approach this problem.
A beam of unpolarized positrons from a conventional source being cooled in a damping ring and preliminary accelerated to an energy $`\mathrm{E}_0`$ can be polarized during a head-on-collision with a high-intensity circularly polarized laser wave.
It is well known that during Compton backscattering of circularly polarized laser photons on unpolarized positrons (electrons) with the energy E$`{}_{0}{}^{}`$100 GeV the scattered photon takes up to 90$`\%`$ of the initial positron energy while the recoil positron aquires $`100\%`$ longitudinal polarization . At $`\mathrm{E}_0`$10 GeV, however, the positron loses too little of its energy during single Compton backscattering (a few percent), and the longitudinal polarization of the recoil positron is, therefore, of the same order of magnitude. Current advances of laser physics make it possible to obtain parameters of laser flash such that the positron successively interacts with $`\mathrm{N}`$ 1 identical circularly polarized photons. It is apparent that in this case the positron can lose a substantial fraction of its energy (comparable with $`\mathrm{E}_0`$). To evaluate the resulting polarization of the recoil positron, let us consider multiple Compton backscattering in greater detail.
Let us carry out calculations in a positron rest frame (PRF) and in a laboratory frame (LF). Following , let us write the Compton scattering cross section in PRF where spin correlations of three particles will be viewed– initial photon, and initial and recoil positrons (upon summation over the scattered photon polarizations):
$$\frac{\mathrm{d}\sigma }{\mathrm{d}\mathrm{\Omega }}=2\mathrm{r}_0^2\left(\frac{\mathrm{k}}{\mathrm{k}_0}\right)^2\left\{\mathrm{\Phi }_0+\mathrm{\Phi }_2(\mathrm{P}_\mathrm{c},\stackrel{}{\xi }_0)+\mathrm{\Phi }_2(\mathrm{P}_\mathrm{c},\stackrel{}{\xi })+\mathrm{\Phi }_2(\stackrel{}{\xi }_0,\stackrel{}{\xi })+\mathrm{\Phi }_3(\mathrm{P}_\mathrm{c},\stackrel{}{\xi }_0,\stackrel{}{\xi })\right\}$$
$`(1)`$
Here $`\mathrm{r}_0`$ is the electron classical radius; $`\mathrm{k}_0`$$`\mathrm{k}`$ are the initial and scattered photon energy; $`\mathrm{P}_\mathrm{c}=\pm `$ 1 is the circular polarization of the initial photon; and $`\stackrel{}{\xi _0},\stackrel{}{\xi }`$ are the spin vectors of the initial and final positrons. Functions $`\mathrm{\Phi }_0,\mathrm{\Phi }_2,\mathrm{\Phi }_3`$ were obtained in paper .
In (1) and further in the paper use is made of the system of units $`\mathrm{}=\mathrm{m}_\mathrm{e}=\mathrm{c}=1`$, unless otherwise indicated.
Since the scattered photons are not detected, the cross section (1) has to be integrated over the photon outgoing angles. Due to azymuthal symmetry, it will depend on the average longitudinal polarization components $`\xi _{0\mathrm{l}},\xi _\mathrm{l}`$ solely. On this basis we will keep only these components which remain the same in LF.
For positrons with $`\gamma _010^4(\gamma _0`$ is the Lorentz-factor of the initial positron), the laser photon energy in PRF ( $`\omega _0`$1 eV in LF ) will satisfy the relation
$$\mathrm{k}_0=2\gamma _0\omega _01$$
$`(2)`$
Using (2) let us write the expression for the scattered photon energy in PRF:
$$\mathrm{k}=\frac{\mathrm{k}_0}{1+\mathrm{k}_0(1\mathrm{cos}\theta )}\mathrm{k}_0\left[1\mathrm{k}_0(1\mathrm{cos}\theta )\right]$$
$`(3)`$
Here $`\theta `$ is the polar angle of the scattered photon in PRF.
Leaving the terms not higher than $`\mathrm{k}_0^2`$ , let us write in explicit form the $`\mathrm{\Phi }_\mathrm{i}`$ functions derived in for electrons :
$$\mathrm{\Phi }_0=\frac{1}{8}\left[1+\mathrm{cos}^2\theta +\mathrm{k}_0^2(1\mathrm{cos}\theta ^2)\right],$$
$$\mathrm{\Phi }_2(\mathrm{P}_\mathrm{c},\xi _{0\mathrm{l}})=\frac{1}{8}\mathrm{P}_\mathrm{c}\xi _{0\mathrm{l}}\mathrm{k}_0\mathrm{cos}\theta ,$$
$`(4)`$
$$\mathrm{\Phi }_2(\mathrm{P}_\mathrm{c},\xi _\mathrm{l})=\frac{1}{8}\mathrm{P}_\mathrm{c}\xi _\mathrm{l}\left(1\mathrm{cos}\theta \right)\left[2\mathrm{k}_0\mathrm{cos}\theta \mathrm{k}_0^2(\mathrm{cos}\theta \mathrm{cos}^2\theta +\mathrm{sin}^2\theta )\right],$$
$$\mathrm{\Phi }_2(\xi _{0\mathrm{l}},\xi _\mathrm{l})=\frac{1}{8}\xi _{0\mathrm{l}}\xi _\mathrm{l}\left[1+\mathrm{cos}^2\theta \mathrm{k}_0^2\mathrm{cos}\theta \mathrm{sin}^2\theta \right],$$
$$\mathrm{\Phi }_3(\mathrm{P}_\mathrm{c},\xi _{0\mathrm{l}},\xi _\mathrm{l})=0.$$
Upon routine integration we obtain:
$$\sigma =\frac{\pi \mathrm{r}_0^2}{2}\left\{\frac{8}{3}(12\mathrm{k}_0)+\frac{4}{3}\mathrm{P}_\mathrm{c}\xi _{0\mathrm{l}}\mathrm{k}_0(12\mathrm{k}_0)+\xi _\mathrm{l}\left[\frac{8}{3}\xi _{0\mathrm{l}}(12\mathrm{k}_0)+\frac{4}{3}\mathrm{P}_\mathrm{c}\mathrm{k}_0\right]\right\}$$
$`(5)`$
It is obvious that in averaging with respect to the initial particles spin and taking the summation with respect to two spin states of the recoil positron, instead of (5) we get Klein-Nishina’s cross section for $`\mathrm{k}_0`$ 1 :
$$\sigma =\frac{8}{3}\pi \mathrm{r}_0^2\left(12\mathrm{k}_0\right)$$
$`(6)`$
From (5) follows that longitudinal polarization of a recoil positron (electron) is determined by both its initial polarization and the circular polarization of a photon (later the longitudinal polarization indices $`l`$ will be omitted):
$$\xi =\frac{\xi _0{\displaystyle \frac{\mathrm{k}_0}{2}}\mathrm{P}_\mathrm{c}}{1{\displaystyle \frac{\mathrm{k}_0}{2}}\mathrm{P}_\mathrm{c}\xi _0}$$
$`(7)`$
The upper (lower) sing refers to a positron (electron).
If the initial positron is unpolarized ($`\xi _0`$ = 0), then upon a single interaction with a laser photon the recoil positron becames polarized :
$$|\xi _{(1)}|=|\frac{\mathrm{k}_0}{2}\mathrm{P}_\mathrm{c}|1.$$
$`(8)`$
In order to consider the next scattering act, let us calculate the average longitudinal momentum $`<\mathrm{k}_{}>`$ along the initial photon direction and the average energy $`<\mathrm{k}>`$ of the scattered photon in PRF using the same approximation as before:
$$<\mathrm{k}_{}>=\frac{\mathrm{k}\mathrm{cos}\theta \left({\displaystyle \frac{\mathrm{k}}{\mathrm{k}_0}}\right)^2\mathrm{\Phi }_0d\mathrm{\Omega }}{\left({\displaystyle \frac{\mathrm{k}}{\mathrm{k}_0}}\right)^2\mathrm{\Phi }_0d\mathrm{\Omega }}=\frac{6}{5}\mathrm{k}_0^2,$$
$`(9)`$
$$<\mathrm{k}>=\frac{\mathrm{k}\left(\frac{\mathrm{k}}{\mathrm{k}_0}\right)^2\mathrm{\Phi }_0d\mathrm{\Omega }}{\left({\displaystyle \frac{\mathrm{k}}{\mathrm{k}_0}}\right)^2\mathrm{\Phi }_0d\mathrm{\Omega }}=\mathrm{k}_0(1\mathrm{k}_0).$$
Thus, upon the first event of interaction, the photon in LF aquires, on average, the energy
$$<\omega _{\mathrm{sc}}>=\gamma _0(<\mathrm{k}>\beta _0<\mathrm{k}_{}>)\gamma _0<\mathrm{k}>=\gamma _0\mathrm{k}_0.$$
$`(10)`$
In (10) $`\beta _0=1{\displaystyle \frac{\gamma _0^2}{2}}`$ is the velocity of PRF with respect to LF.
It is apparent that the recoil positron loses its energy (10) and hence
$$\gamma _{(1)}=\gamma _0<\omega _{\mathrm{sc}}>=\gamma _0(1\mathrm{k}_0)=\gamma _0(12\gamma _0\omega _0).$$
$`(11)`$
In PRF before the second interaction the initial photon, in view of (11), will have a lower energy
$$\mathrm{k}_{(1)}=2\gamma _{(1)}\omega _0=2\gamma _0\omega _0(12\gamma _0\omega _0)=\mathrm{k}_0(1\mathrm{k}_0),$$
$`(12)`$
and the recoil positron will have a polarization:
$$\xi _{(2)}=\frac{\xi _{(1)}{\displaystyle \frac{\mathrm{k}_{(1)}}{2}}\mathrm{P}_\mathrm{c}}{1{\displaystyle \frac{\mathrm{k}_{(1)}}{2}}\mathrm{P}_\mathrm{c}\xi _{(1)}}.$$
$`(13)`$
Substituting its value from (8) for $`\xi _{(1)}`$, we obtain:
$$\xi _{(2)}=\mathrm{P}_\mathrm{c}\frac{{\displaystyle \frac{\mathrm{k}_0}{2}}+{\displaystyle \frac{\mathrm{k}_{(1)}}{2}}}{1\mathrm{P}_\mathrm{c}{\displaystyle \frac{\mathrm{k}_0}{2}}{\displaystyle \frac{\mathrm{k}_{(1)}}{2}}},|\xi _{(2)}|>|\xi _{(1)}|$$
$`(14)`$
It follows from (14) that as a result of multiple Compton backscattering the longitudinal polarization of positrons builds up, while their energy decreases in LF (so-called laser cooling, see ).
Let us write expressions relating the polarization and energy for two subsequent acts of scattering:
$$\gamma _{(\mathrm{i}+1)}=\gamma _{(\mathrm{i})}(12\omega _0\gamma _{(\mathrm{i})}),$$
$`(15)`$
$$\xi _{(\mathrm{i}+1)}=\frac{\xi _{(\mathrm{i})}\gamma _{(\mathrm{i})}\omega _0\mathrm{P}_\mathrm{c}}{1\gamma _{(\mathrm{i})}\omega _0\mathrm{P}_\mathrm{C}\xi _{(\mathrm{i})}}.$$
$`(16)`$
From these we can obtain the equations for the finite differences:
$$\mathrm{\Delta }\gamma _{(\mathrm{i})}=\gamma _{(\mathrm{i}+1)}\gamma _{(\mathrm{i})}=2\omega _0\gamma _{(\mathrm{i})}^2,$$
$`(17)`$
$$\mathrm{\Delta }\xi _{(\mathrm{i})}=\xi _{(\mathrm{i}+1)}\xi _{(\mathrm{i})}\omega _0\mathrm{P}_\mathrm{c}\gamma _{(\mathrm{i})}(1\xi _{(\mathrm{i})}^2).$$
$`(18)`$
When $`\mathrm{N}`$1, instead of (17) and (18) we can arrive at differential equations, whose solution with proper initial conditions will yield
$$\gamma _{\left(\mathrm{N}\right)}=\frac{\gamma _0}{1+2\gamma _0\omega _0\mathrm{N}}$$
$`(19)`$
$$\xi _{(\mathrm{N})}=\frac{\gamma _0\omega _0\mathrm{N}}{1+\gamma _0\omega _0\mathrm{N}}.$$
$`(20)`$
When deriving the above relation, there was taken the left circular polarization $`\mathrm{P}_\mathrm{c}=1`$ for the sake of simplicity.
Equations (19) and (20) describe the positron characteristics after $`N`$ collisions with circularly polarized laser photons. The number of collisions $`\mathrm{N}`$ is controlled by the luminosity of the process $`\mathrm{L}`$:
$$\mathrm{N}=\frac{\mathrm{N}_{\mathrm{scat}}}{\mathrm{N}_\mathrm{e}^+}=\mathrm{N}_0\mathrm{L}=\mathrm{N}_0\frac{{\displaystyle \frac{8}{3}}\pi \mathrm{r}_0^2}{2\pi (\sigma _{\mathrm{e}^+}^2+\sigma _{\mathrm{ph}}^2)}.$$
$`(21)`$
In (21) $`\mathrm{N}_0=\mathrm{A}/\omega _0`$ is the number of photons per laser flash, $`\mathrm{A}`$ is the laser energy, and $`\sigma _{\mathrm{ph}},\sigma _{\mathrm{e}^+}`$, are the laser focus and positron bunch radii. We can expect that after cooling in the damping ring $`\sigma _{\mathrm{e}^+}\sigma _{\mathrm{ph}}`$. In this case, substituting (21) into (19) and (20), we obtain the following simple formulas for positron’s characteristics:
$$\gamma _{\left(\mathrm{N}\right)}=\frac{\gamma _0}{1+2\mu },$$
$`(22)`$
$$\xi _{(\mathrm{N})}=\frac{\mu }{1+\mu },$$
$`(23)`$
which depend on the dimensionless parameter $`\mu `$ solely
$$\mu =\gamma _0\omega _0\mathrm{N}=\frac{4}{3}\frac{\mathrm{A}}{\mathrm{mc}^2}\gamma _0\left(\frac{\mathrm{r}_0}{\sigma _{\mathrm{ph}}}\right)^2.$$
$`(24)`$
It follows from (24) that the $`\mu `$ parameter depends linearly on the laser flash energy and the initial positron energy, but it is inversely proportional to the laser focus area and does not depend on the interaction time (duration flash). Having written (22) as:
$$\frac{\gamma _0}{\gamma _{\left(\mathrm{N}\right)}}=1+2\mu ,$$
$`(25)`$
we will compare the result with the estimate by V. Telnov obtained in a classical approximation. Substituting into (24) the estimate used in $`\sigma _{\mathrm{ph}}^2={\displaystyle \frac{\lambda _0\mathrm{l}_\mathrm{e}}{8\pi }}`$ ( $`\lambda _0`$ is the laser photon wavelength and $`\mathrm{l}_\mathrm{e}`$ is the positron bunch length), we get :
$$\frac{\gamma _0}{\gamma _{\left(\mathrm{N}\right)}}=1+\frac{64}{3}\frac{\mathrm{A}}{\mathrm{mc}^2}\gamma _0\frac{\pi \mathrm{r}_0^2}{\lambda _0\mathrm{l}_\mathrm{e}}.$$
$`(26)`$
The resulting expression is closed to a similar one in but the second term in (26) is by a factor of $`\pi `$ smaller. This dicrepancy is connected with rough calculation of the luminosity (constant area of the laser focus) used in (21).
By way of illustration let us consider an example (see ):
$$\gamma _0=10^4,\mathrm{A}=5\mathrm{J},\lambda _0=500\mathrm{nm},\mathrm{l}_\mathrm{e}=0.2\mathrm{mm},\sigma _{\mathrm{ph}}^2=\lambda _0\mathrm{l}_\mathrm{e}/8\pi .$$
$`(27)`$
In this case $`\mu `$= 1.6 and, therefore, $`\gamma _{\left(\mathrm{N}\right)}0.3\gamma _0;\xi _\mathrm{l}60\%`$.
Thus, when a positron bunch interacts with a laser flash of the given parameters, all the positrons acquire longitudinal polarization of about 60$`\%`$. The change in the polarization sign is obtained by inverting the sign of the circular polarization of laser radiation.
It should be noted that with a proper selection of the sign of circular polarization, the process of laser cooling would give rise to a longitudinal polarization increase of the electrons rather than to depolarization of electrons beam (as in the case of unpolarized laser radiation considered in ).
Note that, generally speaking, the laser parameters (27) correspond to the so-called ”strong” field, when the contribution from non-linear Compton scattering would be considerably high.
Non-linear processes, i.e., simultaneous scattering of a few laser photons on the moving positron, are characterized by an increase in the effective positron mass in PRF, which, in its turn, leads to a decrease in the Lorentz-factor and the energy transferred to the positron through scattering. It is to be expected that the $`\mu `$ parameter (24) for a fixed value of the laser flash energy A will be sufficiently lower for a non-linear case as compared to the linear one, and hence a lower attainable polarization (23).
In order to reach a linear mode of the Compton scattering process, one has to stretch the laser flash (the length of its interaction with the positron bunch) (see, for instance, ).
In conclusion, let us estimate the energy $`\mathrm{A}_{+,\mathrm{pol}}`$ necessary to obtain one polarized positron with the energy $`\mathrm{E}_+>10^1`$ MeV and the longitudinal polarization $`\xi _\mathrm{l}>`$ 0,5 i.e., the parameters acceptable for consequent acceleration.
i) According to the estimates an electron with the energy $`\mathrm{E}_{}`$ 200 GeV passing through a helical undulator of the length $`\mathrm{L}150\mathrm{m}`$ can generate a number of circularly polarized photons needed to obtain one polarized positron to be later accelerated (conversion efficiency $`\eta =\mathrm{N}_{\mathrm{e}^+,\mathrm{pol}}/\mathrm{N}_\mathrm{e}^{}1`$). Hence, $`\mathrm{A}_{+,\mathrm{pol}}\mathrm{E}_{}/\eta =200\mathrm{GeV}`$.
ii) The author of the paper considered a scheme for production of $`\mathrm{N}_{\mathrm{e}^+}=10^9`$ polarized positrons when the laser radiation of the total energy $`\mathrm{A}_\mathrm{\Sigma }20\mathrm{J}`$ is scattered on an electron bunch with $`\mathrm{E}_{}`$ = 5 GeV and $`\mathrm{N}_\mathrm{e}^{}=10^{10}\mathrm{e}^{}`$ / bunch. Thus
$$\mathrm{A}_{+,\mathrm{pol}}\frac{\mathrm{N}_\mathrm{e}^{}\mathrm{E}_{}+\mathrm{A}_\mathrm{\Sigma }}{\mathrm{N}_{\mathrm{e}^+}}170\mathrm{GeV}.$$
iii) In the author estimated the conversion efficiency for longitudinally polarized electrons with the energy $`\mathrm{E}_{}`$ = 50 MeV:
$$\eta 10^3.$$
Therefore $`\mathrm{A}_{+,\mathrm{pol}}\mathrm{E}_{}/\eta =`$ 50 GeV.
iv) For the method suggested in the present paper, evaluation of $`\mathrm{A}_{+,\mathrm{pol}}`$ can be made for parameters of the unpolarized positron source used in SLAC .
The conversion efficiency for the electron energy $`\mathrm{E}_{}`$ = 33 GeV equals:
$$\eta 1.$$
Therefore, for a bunch with $`\mathrm{N}_{\mathrm{e}^+}=10^{10}`$ and the positron energy $`\mathrm{E}_0`$ = 5 GeV interacting with the laser flash ($`\mathrm{A}=5\mathrm{J}`$) we have:
$$\mathrm{A}_{+,\mathrm{pol}}=\frac{\mathrm{E}_{}}{\eta }+\mathrm{E}_0+\frac{\mathrm{A}}{\mathrm{N}_{\mathrm{e}^+}}=33\mathrm{GeV}+5\mathrm{GeV}+3\mathrm{GeV}40\mathrm{GeV}.$$
Thus, the scheme proposed here seems to be most energy effective.
The author is grateful to V. Telnov and J. Clendenin for stimulating discussions.
References
1. P.M. Zerwas. Preprint DESY 94-001, 1994.
2. J.E. Clendenin, R. Alley, J. Frish, T. Kotseroglou, G. Mulhollan, D. Schultz, H. Tang, J. Turner and A.D. Yeremian. The SLAC Polarized Electron Source. AIP Conf. Proceedings, N. 421, pp. 250-259, 1997.
3. V.E. Balakin, A.A. Mikhailichenko. Preprint INP 79-85, Novosibirsk, 1979.
4. Yung Su Tsai. Phys. Rev. D, v.48 (1993), pp.96-115.
5. T. Okugi, Y. Kurihara, M. Chiba, A. Endo, R. Hamatsu, T. Hirose, T. Kumita, T. Omori, Y. Takeuchi, M. Yoshioka. Jap.J.Appl.Phys. v. 35(1996), pp. 3677-3680.
6. A.P. Potylitsyn. Nucl. Instr. and Meth. v. A398 (1997), pp.395-398.
7. E.G. Bessonov, A.A. Mikhailichenko. Proc. of V European Particle Accelerator Conference, 1996, pp. 1516-1518.
8. K. Flottmann. Preprint DESY 93-161, 1993.
9. T. Omori. KEK- Proceedings 99-12, 1999, pp. 161-179.
10. T. Kotseroglou, V. Bharadwaj, J.E. Clendenin, S. Ecklund, J. Frisch, P. Krejcik, A.V. Kulikov, J. Liu, T. Maruyama, K.K. Millage, G. Mulhollan, W.R. Nelson, D.C. Schultz, J.C. Sheppard, J. Turner, K. Van Bibber, K. Flottmann, Y. Namito. Particle Accelerator Conference (PAC’99). Proceedings (to be published).
11. H. Tolhoek. Rev. of Mod. Phys. v. 28 (1956), pp. 277-298.
12. G.I. Kotkin, S.I. Polityko, V.G. Serbo. Physics of Atomic Nuclei, v. 59 (1996), pp. 2229-2234.
13. F.W. Lipps, H.A. Tolhoek. Physika, v. 20 (1954), pp. 395-405.
14. P. Sprangle, E. Esarey. Phys. Fluids. v.B4 (1992), pp. 2241-2248.
15. V. Telnov. Phys. Rev. Lett. v.78 (1997), pp.4757-4760.
16. I.V. Pogorelsky, I. Ben-Zvi, T. Hirose. BNL Report No.65907, October, 1998.
17. S. Ecklund. SLAC-R-502 (1997), pp. 63-98. |
warning/0001/physics0001023.html | ar5iv | text | # Measurement of mechanical vibrations excited in aluminium resonators by 0.6 GeV electrons.
## 1 Introduction
A key issue for a Resonant Mass Gravitational Wave Detector of improved sensitivity with respect to the existing detectors, is the background due to impinging cosmic ray particles . The energy deposited in the detector’s mass along a particle’s track may excite the very vibrational modes that are to signal the passing of a gravitational wave. Computer simulations of such effects are based on the thermo-acoustic conversion model and earlier measurements of resonant effects in Beron et al. and Grassi Strini et al. . According to the model, the energy deposited by a traversing particle heats the material locally around the particle track, which leads to mechanical tension and thereby excites acoustic vibrational modes . At a strain sensitivity of the order of $`10^{21}`$ envisaged for a next generation gravitational-wave detector, computer simulations show that operation of the instrument at the surface of the earth would be prohibited by the effect of the cosmic ray background. Since the applicability of the thermo-acoustic conversion model would thus yield an important constraint on the operating conditions of resonant mass gravitational wave detectors, Grassi-Strini, Strini and Tagliaferri measured the mechanical vibrations in a bar resonator bombarded by 0.02 GeV protons and 5\*10<sup>-4</sup> GeV electrons. We extended that experiment by measuring the excitation patterns in more detail for a bar and a sphere excited by 0.6 GeV electrons. Even though we cannot think of a reason why the model, if applicable to the bar, would not hold for a sphere, we did turn to measuring with a sphere also.
We exposed two aluminium 50ST alloy cylindrical bars and an aluminium alloy sphere, each equipped with piezoelectric ceramic sensors, to a beam of $`0.6`$ GeV electrons used in single bunch mode with a pulse width of up to $`2\mu `$s, and adjustable intensity of $`10^9`$ to $`10^{10}`$ electrons. We recorded the signals from the piezo sensors, and Fourier analysed their time series. Before and after the beam run we calibrated the sensor response of one of the bars for its first longitudinal vibrational mode at $``$ 13 kHz to calibrated accelerometers.
## 2 Experiment setup and method
In the experiment we used three different setups in various runs, as summarised in table I: two bars and a sphere. With the un-calibrated bar BU we explored the feasibility of the measurement. Also, bar BU proved useful to indirectly determine the relative excitation amplitudes of higher longitudinal vibrational modes, see sec. 4.1. With bar BC calibrated at its first longitudinal vibrational mode, we measured directly its excitation amplitude in the beam. Finally, with the sphere we further explored the applicability of the model.
Table I. Characteristics of our setup. Setup code name: BC BU SU Resonator type: bar bar sphere Diameter: 0.035 m 0.035 m 0.150 m Length: 0.2 m 0.2 m - Suspension: plastic string plastic string brass rod 0.15 m\*0.002 m Piezo sensors: 1 2 2 Piezo hammer: 0 1 1 Capacitor driver 1 0 0 Direct calibration yes no no Beam energy 0.76 GeV 0.62 GeV 0.35 GeV Beam peak current 3 mA 18 mA 19 mA Electrons per burst $`10^9`$ $`510^{10}`$ $`510^{10}`$ Mean absorbed energy per electron 0.02 GeV 0.02 GeV 0.1 GeV Typical absorbed energy per burst 0.01 J 0.6 J 3.0 J
### 2.1 Electron beam
We used the Amsterdam linear electron accelerator MEA delivering an electron beam with a pulse-width of up to 2 $`\mu `$s in its hand-triggered, single bunch mode. The amount of charge per beam pulse was varied, recorded by a calibrated digital oscilloscope, photographed and analysed off line to determine the number of impinging electrons per burst.
### 2.2 Suspension and positioning
In both setups BC and BU, see fig. 1, the cylindrical aluminium bar was horizontally suspended in the middle, as indicated in the figure, with a plastic string.
The bar’s cylinder axis was positioned at 90<sup>0</sup> to the beam direction. The bar’s suspension string was connected to a horizontally movable gliding construction, enabling us to handle the resonator by remote control, and let the impinging electron beam hit it at different horizontal positions. The aluminium sphere SU, see fig. 2, was suspended from its centre by a brass rod. Either the bar’s gliding construction, or the sphere’s suspension bar, was attached to an aluminium tripod mounted inside a vacuum chamber , which was evacuated to about $`10^5`$mbar. By remote control, we rotated the tripod and moved it vertically to either let the beam pass the resonator completely, or let it traverse the resonator. We let the beam traverse the sphere at different heights and different incident angles with respect to the piezo sensors positions on the sphere.
We mark the beam heights as E (Equator) and A (Africa) at 0.022 m below the equator. The E beam passed horizontally through the sphere’s origin, remaining in the same vertical plane for the A beam.
### 2.3 Sensors and signal processing
In setup BC we used a single piezo sensor of $`1531`$mm<sup>3</sup> and glued it over it full length at 0.01 m off the centre on the top of the bar. Bar BC was equipped with a capacitor plate of 0.03 m diameter at a distance of $`0.004`$ m from one of its end faces. In setup BU one piezo sensor of $`360.3`$mm<sup>3</sup> was fixed on one end face of the bar.
A similar sensor of about the same dimensions was fixed in the same manner, oriented parallel to the cylinder’s long axis at a position 35 mm away from the end-face, In the third setup, SU, see fig. 2, two piezo sensors of $`360.3`$mm<sup>3</sup> were glued to the sphere’s surface. One was situated at the equator, with respect to the vertical rotation axis, the other one at a relative displacement of $`45^0`$ west longitude, and at $`45^0`$ north latitude.
For the setup in use, each sensor was connected to a charge amplifier of $`210^{10}`$V/C gain. The signals were sent through a Krohn-Hite 3202R low-pass 100 kHz pre-filter, to a R9211C Advantest spectrum analyser with internal 2 MHz pre-sampling and 125 kHz digital low-pass filtering. The oscillation signals were recorded for 64 ms periods at a 4 $`\mu `$s sample rate. The beam pulse could be used as a delayed trigger to the Advantest. Using the memory option of the Advantest, the piezo signals were recorded from 0.3 ms onward before the arrival of the trigger. The data were stored on disk and were Fourier analysed off line.
### 2.4 Checks and stability
The data were taken at an ambient temperature of $`23^0`$C. By exciting the resonator with the piezo-hammer we checked roughly its overall performance. As to be discussed in section 3, setup BC was calibrated before and after the beam run. The instrument’s stability was checked several times during the run by an electric driving signal on its capacitor endplate.
## 3 Calibration of bar BC’s piezo ceramic sensor
A standard accelerometer mounted on the bar damped the vibrations too strongly to confidently measure their excitations in the electron beam. Therefore the response of the piezoelectric ceramic together with its amplifier was first calibrated against two 2.4 gramme Bruel&Kjaer 4375 accelerometers glued, one at a time, to bar BC’s end face and connected to a 2635 charge amplifier. The resonator was excited through air by a nearby positioned loud-speaker driven from the Advantest digitally tunable sine-wave generator. The output signals from both the piezoelectric ceramic amplifier and the accelerometer amplifier were fed into the Advantest. Stored time series were read out by an Apple Mac 8100 AV, running Lab-View for on-line Fourier analysis, peak selection, amplitude and decay time determination.
We took nine calibration runs varying the charge amplifier’s sensitivity setting, and dismounting and remounting either of the two accelerometers to the bar. For the lowest longitudinal vibrational mode we calculated the ratio of the Fourier peak signal amplitudes, $`R`$, from the piezoelectric ceramic and accelerometer.
With the calibrated bar BC positioned in the electron beam line we checked the stability of the piezoelectric ceramic’s response intermittently with the beam runs by exciting the bar through its capacitor plate at one end face, electrically driving it at and around half the bar’s resonance frequency. We found the response to remain stable within a few percent.
After the beam runs we took additional calibration values in air with a newly acquired Bruel&Kjaer 0.5 g 4374S subminiature accelerometer and a Nexus 2692 AOS4 charge amplifier.
In fig. 3, typical frequency responses are shown when driving the bar by a loud-speaker signal. The upper part gives the Fourier peak amplitude of the bar’s 13 kHz resonance as measured with the accelerometer. The lower part gives the corresponding amplitude for the signal from the piezoelectric ceramic. The right hand side of the picture shows the amplitudes to be smaller, as expected when driving the bar slightly off resonance. We calculate the decay time, $`\tau `$, of the k-th mode amplitude $`A_k\left(t\right)=A_k\left(0\right)e^{t/\tau }`$ to be $`\tau =0.4`$ s for this setup, equipped with the relatively light accelerometer.
Figure 4 shows the corresponding two signals when driving the bar by the capacitor plate at 6.5 kHz, that is at half the bar’s resonance frequency. Here, the direct electric response of the piezoelectric ceramic’s signal to the driving sine-wave is present, clearly without a mechanic signal, as would have shown up in the accelerometer. The direct signal at 6.5 kHz remains constant. On the other hand, the bar’s mechanical signals on and off its resonance frequency around 13 kHz show the expected amplitude change again, thereby demonstrating that around the bar’s resonance, the piezoelectric ceramic does only respond to the mechanical signal, not to the electric driving signal. See also the caption of fig. 4.
We calculated the average value of $`R_0=V_{piezo}^{Fourier}/V_{accel.}^{Fourier}`$ and the error over all 29 measurements, finding for the calibration factor at f=13 kHz,
$$\beta =R_0S\left(2\pi f\right)^2=\left(2.2\pm 0.3\right)\text{V/nm}$$
(1)
where $`S=0.1`$V/ms<sup>-2</sup> is the amplifier setting of the accelerometer.
## 4 Beam experiments
Sensor signals way above the noise level were observed for every beam pulse hitting the sphere or the bar. We ascertained that: a) the signals arose from mechanical vibrations in the resonator, and b) they were directly initiated by the effect of the beam on the resonator, and not arising from an indirect effect of the beam on the piezo sensors. Our assertion is based on a combination of test results observed for both the bars and the sphere, as now to be discussed.
First, when the beam passed underneath the resonator without hitting it, we observed no sensor signal above the noise. Second, as shown in fig. 5, the sensors’ delayed responses after the impact of the beam agreed with the sound velocity. Here the beam was hitting the sphere at a position 5 mm above the sphere’s south pole. The middle trace shows the beam pulse of $`2\mu `$s duration. The two other traces show both piezo sensors to respond with a transient signal right from the start time of the beam’s arrival and to begin oscillating after some delay, depending on their distance from the beam. The distance of the equatorial sensor to the beam hitting the sphere at the south pole was 0.11 m, corresponding to $`22\mu `$s travel time for a sound velocity of $`510^3`$ m/s. The signal is indeed seen in the lowest trace starting to oscillate at that delay time. The upper trace shows the signal from the second sensor situated on the northern hemisphere at 0.14 m from the traversing beam, correspondingly starting to oscillate with a delay of $`28\mu `$s after the impact of the beam. Third, after dismounting the piezo-hammer from the resonator, we observed that the sensor signals did not change, which showed that the activation is not caused by the beam inducing a triggering of the piezo-hammer. Fourth, to simulate the electric effect of the beam pulse on the sensors, we coupled a direct current of 60 mA and 2.5 $`\mu `$s duration from a wave packet generator to the bar. Apart from the direct response of the piezo-sensor during the input driving wave, no oscillatory signal was detected above the noise level. Finally, we measured the dependence of the amplitudes in several vibrational modes on the hit position of the beam, as will be described in the following sections. We found the amplitudes to follow the patterns as calculated with the thermo-acoustic conversion model.
### 4.1 Results for the bar
In fig. 6 a typical Fourier spectrum of bar BC is shown up to 55 kHz. The arrows point to identified vibrational modes . From a fit of $`K`$ and $`f_0`$ of the longitudinal frequencies $`f_L=Lf_0\left(1L^2K\right)`$ of the modes for $`L`$=1,..,4, we find $`f_0=12933,K=0.0022`$, where $`f_0`$ is related to the sound velocity by $`v_s=2lf_0=5173`$ m/s for our bar length of $`l`$=0.2 m. For the Poisson-ratio $`\sigma =2l\sqrt{K}/\left(\pi r\right)`$, $`r`$ being the cylinder radius of the bar, from our fit we get $`\sigma =0.338`$. The values agree well with the handbook quoting $`\sigma `$=0.33 and $`v_s`$=5000 m/s for aluminium. The root mean square error of the fit is 35 Hz, in correspondence with the 30 Hz frequency resolution used in the Fourier analysis. Other peaks correspond to torsional and transverse modes .
The Fourier amplitudes $`A_k`$ of $`f\left(t\right)=A_ke^{i\omega _kt}`$ of the modes depend linearly (as shown for the 13 kHz, $`L`$=1 mode in fig. 7) on the integrated charge in the beam pulse for a fixed beam position, and therefore also linearly on the energy deposited by the beam, which ranged in these runs from 0.06-0.8 J. The spread in the ratios of the amplitudes to the beam charge, shows the Fourier amplitudes to reproduce within $`\pm 10\%`$.
The agreement of the model to within 10% with the measured data is shown in fig. 8. The figure shows the measured Fourier amplitudes of bar BC at the piezo sensor and the calculations following Grassi Strini et al. as a function of the hit position along the cylinder axis for the first four longitudinal modes. For each mode the average model value was scaled to the average measured value. The best fit was found with a shift of the hit positions along the bar, by an overall offset of $`x_0`$=-0.0075 m, which corresponds to the crude way we aligned the bar with the beam line.
#### 4.1.1 Lowest bar mode excitation amplitude
For the 13 kHz, $`L`$=1 mode we determine the absolute amplitude for a comparison with the model calculation of ref. . Firstly, we use the amplitude function B<sub>0</sub>(x), see eq. 9 of ref. , by rewriting it in the form:
$`B_0\left(x\right)=2\kappa _0\mathrm{\Delta }E/\pi \times cos\left(\pi x/l\right)sin\left(\pi \eta /\left(2l\right)\right)/\pi \eta /\left(2l\right)`$ (2)
with
$`\kappa _0=\alpha l/\left(c_vM\right)=\alpha /\left(c_v\rho O\right).`$ (3)
In this expressions $`x`$ is the hit position along the the cylinder axis, $`l`$ the bar length, $`\eta `$ the beam diameter, $`\alpha `$ the thermal linear expansion coefficient, $`\rho `$ the density, c<sub>v</sub> the specific heat, $`O`$ the cylindrical surface area of the bar, and $`\mathrm{\Delta }E`$ the energy absorbed by the bar. From $`B_0\left(x\right)`$ we derive the functional form for the measured values of $`W_{sens}`$ as
$`W_{sens}\left(x\right)={\displaystyle \frac{B_0\left(x\right)}{\mathrm{\Delta }E}}\beta D{\displaystyle \frac{dE}{dQ}}`$ (4)
where $`dE/dQ`$ is the beam energy absorbed by the bar per unit of impinging beam charge, $`\beta `$ the calibration factor as discussed in section 3, and D the decay factor $`e^{t/\tau }`$, since eq. 2 applies at excitation time and we have to correct the amplitude at measuring time for the mode’s decay, corresponding to its Q-factor. Therefore,
$$W_{sens}\left(x\right)=\kappa _{exp}2/\pi \times cos\left(\pi x/l\right)sin\left(\pi \eta /\left(2l\right)\right)/\pi \eta /\left(2l\right),$$
(5)
with
$$\kappa _{exp}=\beta D\kappa _0\frac{dE}{dQ}$$
(6)
From fitting eq. 5 to the measured values $`W_{sens}\left(x\right)`$ given in table II with $`\kappa _{exp}`$ as the free variable, we find our presently measured value for $`\kappa _0^{exp}=\kappa _{exp}/(dE/dQD\beta )`$ which we compare to the model value in eq. 3. Secondly, the decay time was measured by recording the sensor signals after a trigger delayed by up to 1.6 s at a fixed beam hit position. An exponential fit $`A\left(t\right)=A_0e^{t/\tau }`$ to the mode amplitude gives $`\tau =\left(0.36\pm 0.01\right)`$s for the $`L`$=1 mode. This corresponds to a Q-value of $``$ 15000, a value consistent with the room temperature measurement of aluminium as in ref. , and indicating a negligible influence of the suspension and piezoelectric ceramic sensor for this mode. From the measured value of $`\tau `$ and a mean delay time from the start of the beam pulse of 0.016 s, we calculate the decay factor to be $`D=0.95`$.
Table II. Excitation values W<sub>sens</sub>, equalling the ratio of the measured Fourier amplitude and the measured beam pulse charge at each of the indicated hit positions on the bar for the 13 kHz, $`L`$=1 mode.
| hit position $`x`$ cm | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| W<sub>sens</sub> V/nC | 0.185 | 0.216 | 0.167 | 0.180 | 0.225 | 0.152 | 0.152 | 0.157 | 0.112 | 0.089 | 0.057 |
Thirdly, as indicated, we use the data for W<sub>sens</sub> in the second row of table II to fit the variable $`\kappa _{exp}`$ in eq. 5, where now x is the hit position as given in row 1, $`l`$=0.2 m , and $`\eta =0.002`$ m. The value found in the fit is $`\kappa _{exp}=\left(0.300\pm 0.025\right)`$ V/nC. Fourthly, from a Monte Carlo simulation at the beam energy of 570 MeV used for these runs, we calculate the mean absorbed energy and the mean energy spread which results from the fluctuating energy losses of the passing electrons and the energies of the secondaries escaping from the bar, as $`\mathrm{\Delta }`$E<sub>e</sub>=(19$`\pm `$ 2) MeV. The electron beam pulse thus deposits $`dE/dQ=\left(0.019\pm 0.002\right)`$ J/nC in the bar. Using the measured calibration value at f=12986 Hz as given in eq. 1, $`\beta `$=(2.2 $`\pm `$ 0.3) V/nm, we arrive at
$`\kappa _0^{exp}=\left(7.4\pm 1.4\right)\text{nm/J}.`$ (7)
Finally, we calculate the model value of $`\kappa _0`$ from the material constants as being $`\kappa _0`$=10 nm/J, neglecting the much smaller error as arising from some uncertainty in the parameters. We conclude that $`\kappa _0^{exp}/\kappa _0=\left(0.74\pm 0.14\right)`$, a result that is consistent with the validity of the model of ref. .
The measured maximum excitation amplitude at beam position $`x`$=0, see fig. 8 for the 13 kHz, $`L`$=1 longitudinal mode thus corresponds to (0.13 $`\pm `$ 0.02) nm.
#### 4.1.2 Higher bar mode excitation amplitudes
Having determined the correspondence between the model calculation and the experiment’s result for the first longitudinal vibrational mode amplitude, we return to some of the higher vibrational modes. To compare the modes we need to take the sensor position on the bar into account. We rewrite the displacement amplitude of eq. 5 from ref. as a function of hit position $`x_h`$ and sensor position $`x_s`$ as:
$$\mathrm{\Phi }_{oddL}=\left(2\kappa /L\pi \right)sin\left(L\pi x_s/l\right)cos\left(L\pi x_h/l\right),$$
$$\mathrm{\Phi }_{evenL}=\left(2\kappa /L\pi \right)cos\left(L\pi x_s/l\right)sin\left(L\pi x_h/l\right),$$
(8)
where $`l`$ is the bar length. We dropped the beam width correction term which would lead to a less than 0.1% correction even for $`L`$=4. We approximate the sensor response by the local strain along bar BC’s cylinder axis, that is to the d$`\mathrm{\Phi }`$/dx<sub>s</sub> of eq. 8, arriving at a sensor response, S<sub>L</sub>:
$`S_{oddL}=B_Lcos\left(L\pi x_h/l\right),B_{oddL}=\left(2ϵ\kappa /l\right)cos\left(L\pi x_s/l\right),`$
$`S_{evenL}=B_Lsin\left(L\pi x_h/l\right),B_{evenL}=\left(2ϵ\kappa /l\right)sin\left(L\pi x_s/l\right),`$ (9)
where $`ϵ`$ is a sensor response parameter. The $`x_s`$ dependent term did not enter into the calculation of $`\kappa _0^{exp}`$ in the previous section, since the calibration was done at the same sensor position as the beam measurement. However, for a comparison between the modes, the dependence on the sensor position $`x_s`$ has to be taken into account. Since the variables are strongly correlated, we, first, fitted for each mode the term $`B_L`$ in the $`x_h`$ dependent part of eq. 9 to the measured value of $`W_{sens}`$ for the mode, shifting the origin of $`x_h`$ by 0.0075 m, as mentioned before. The results are given in the first row of table III.
Table III. Bar BC modes comparison. The piezoelectric ceramic sensor responds to the bar’s strain.
| description | symbol | 13 kHz,$`L`$=1 | 25.6 kHz,$`L`$=2 | 38 kHz,$`L`$=3 | 50 kHz,$`L`$=4 |
| --- | --- | --- | --- | --- | --- |
| amplitude | $`B_L^{meas}`$ | 0.12$`\pm 0.01`$ | 0.021$`\pm 0.002`$ | 0.033$`\pm 0.03`$ | 0.052$`\pm 0.005`$ |
| decay correction | D | 1.04$`\pm `$0.001 | 1.17$`\pm 0.02`$ | 1.49 $`\pm `$0.06 | 1.14$`\pm `$0.01 |
| $`B_L^{meas}\times D`$ | $`B_L^{exp}`$ | 0.12$`\pm 0.01`$ | 0.025$`\pm 0.003`$ | 0.049$`\pm 0.005`$ | 0.059$`\pm 0.006`$ |
| sensor position factor | $`P_L`$ | 1.02 | 2.61 | 1.20 | 1.41 |
| $`ϵ\kappa 2/l=B_L^{exp}\times P_L(arb.u.)`$ | $`\kappa ^{}`$ | 0.12$`\pm 0.01`$ | 0.065$`\pm 0.007`$ | 0.059$`\pm 0.006`$ | 0.083$`\pm 0.008`$ |
Second, we corrected the amplitudes $`B_L^{meas}`$ for the mode decay with a factor $`D`$, given in row 2, and corresponding to the times $`\tau _1=0.36s,\tau _2=0.10s,\tau _3=0.04s,\tau _4=0.12s`$, which leads to the values of $`B_L^{exp}`$ in row 3. Finally, we multiplied with the factor $`P_{oddL}=1/cos\left(L\pi x_s/l\right)`$, $`P_{evenL}=1/sin\left(L\pi x_s/l\right)`$, where the bar length is $`l`$=0.2 m. Since the sensor extends from 0.005 through 0.020 m from the center of the bar, we use the mean sensor position $`x_s=0.0125`$ m. The resulting values of $`\kappa ^{}=2ϵ\kappa /l`$, shown in the last row, should be independent of L. For $`L`$=2,3,4 they are rather closely scattered around a mean value of $`\kappa ^{}=0.07`$ which is, however, at about half the $`L`$=1 value. This discrepancy might have originated from some resonances of the sensor itself, and we suspect the strong peak at 23 kHz, shown in figure 6, to be an indication of such resonances playing a role.
Since the amplitudes of the higher modes for bar BC do not comply with our expectations we turn, as a further check, to our un-calibrated measurements with bar BU. It had been equipped with a piezoelectric sensor at one end face where the longitudinal modes have maximum amplitude. The sensor had been mounted flatly with about half of its surface glued to the bar, and responding to the bar’s surface acceleration, not its strain as at bar BC. We extract the $`\kappa _L`$ values from our measurement analogously as for bar BC, following again the model calculations of Grassi Strini et al. , using the $`L`$=1 mode as the reference. The results are given in table IV.
Table IV. Bar BU modes comparison. The piezoelectric ceramic sensor responds to the bar’s acceleration. The value of $`\kappa _1^{meas}`$ for the $`L`$=1, 13 kHz mode is used as the reference for the higher modes.
| description | symbol | 13 kHz,$`L`$=1 | 25.6 kHz,$`L`$=2 | 38 kHz,$`L`$=3 | 50 kHz,$`L`$=4 |
| --- | --- | --- | --- | --- | --- |
| relative amplitude | $`B^{meas}`$ | 1 | 1.15 | 11.5 | 3.6 |
| relative decay correction | D | 1 | 3$`\pm `$1 | 0.7$`\pm `$0.2 | 1.3$`\pm `$0.9 |
| $`\left(\omega _{L=1}/\omega _L\right)^2`$ | $`\mathrm{\Omega }`$ | 1 | 0.26 | 0.12 | 0.07 |
| $`B^{meas}\times D\times \mathrm{\Omega }`$ | $`\kappa _L^{meas}/\kappa _1^{meas}`$ | $`\mathrm{𝟏}`$ | 0.8$`\pm `$0.3 | 0.9$`\pm `$0.4 | 0.3$`\pm `$0.3 |
After applying the decay correction factor $`D`$ and the frequency normalisation factor $`\mathrm{\Omega }`$, the results should be independent of L. The $`L`$=4 value is significantly low, which, again, might be due to some interfering resonance. The $`L`$=2 and $`L`$=3 values, however, do not significantly deviate from the $`L`$=1 value, thus confirming the model calculations for these higher modes too.
### 4.2 Results for the sphere
Our measurements on the sphere consisted of a) hitting the sphere with the beam at one of two heights in the vertically oriented plane through its suspension: at the equator (E) and at 0.022 m southward (A); b) rotating the sphere with its two fixed sensors over $`180^0`$ around the suspension axis at each beam height, and measuring several times back and forth by steps of $`30^0`$ to diminish the influence of temperature and beam fluctuations, ending up on a $`10^0`$ angular lattice.
The Fourier amplitude spectrum of sensor-1, averaged over the angular positions, is shown in fig. 9. The lowest spheroidal mode is most relevant for a spherical resonant mass gravitational wave detector, and we therefore focus on a few spheroidal modes. As expected, the lowest spheroidal $`L`$=2 mode is seen at 17.6 kHz, the lowest spheroidal $`L`$=1 mode at 24 kHz, and the lowest spheroidal $`L`$=0 mode at 37 kHz. Some other peaks are also indicated in the figure, though not the toroidal modes, which we neglect completely. It should be noted that while the $`L`$0 amplitudes oscillate over the angles, the $`L`$=0 amplitude does not, leading to a relative enhancement of the latter in the angle-averaged fig. 9.
The Fourier amplitudes, again, showed a linear dependence on the deposited energy.
To determine the decay times at $`f`$=17.6, 24 kHz and 37 kHz, see fig. 10, we took data with up to 4 s delay in the spectrum analyser, and found $`\tau 1`$s, 0.4 s and 0.1 s respectively.
The angular distributions for the amplitude of the 37 kHz, $`L`$=0 mode at the two vertical beam positions E and A are shown in figures 11. The $`L`$=0 amplitude is independent of the angle, and since the amplitude is constant to within 20% we infer from the Fourier-modulus’ deviation from flatness a 20% variation of the beam intensity from one shot to another.
During our measurement with the sphere we were unable to use the beam pulse as a trigger, implying that the start time of data acquisition with respect to the beam pulse is unknown. In our further analysis we will therefore use only the Fourier modulus, and not analyse the phases.
The absolute scale of the 37 kHz Fourier amplitude turned out to be $`5`$ times larger than the model value for sensor-2 and $`50`$ times for sensor-1. We assume this discrepancy to be based on some interference effects, possibly with a sensor resonance and a suspension bar mode, and we do not further analyse the $`L`$=0 mode. To disentangle the angular distributions in general, we felt, would squeeze the results of our simple measurement too much, for a couple of reasons. First, the Fourier amplitude $`A_L`$ for any multi-pole order $`L`$ in the sphere’s case is actually a sum of $`M`$-submodes. Though they would be degenerate for an ideal sphere, in practice some $`M`$-modes might or might not turn out to be split beyond the frequency-resolution of $`\mathrm{\Delta }f`$=30 Hz. Second, both sensors $`s_1,s_2`$ should be taken to have unknown sensitivities, e$`_{s_j}`$, in three orthogonal directions, with phase factors +1 or -1 for their orientation. Third, though each mode would start to be excited within the same sub-nanosecond time interval of the beam crossing, the building up of each mode’s resonance vibration may lead to a specific phase $`t_{M_L,b_k}^0`$ depending on the mode’s spatial relation to the beam path.
We now show that the calculated angular distributions have the signature of the $`L`$-character of the measurement. Therefore, we write the Fourier modulus at different impinging beam positions $`b_k`$ as a function of the angle $`\varphi `$ as:
$$A_{L,s_j,b_k}\left(\varphi \right)=F_{L,s_j,b_k}\mathit{ϵ}_{𝒔_𝒋}\underset{M_L}{\overset{+M_L}{}}s_{L,M_L,b_k}𝐮_{L,M_L,s_j,b_k}\left(\varphi \right)e^{\omega _Lt_{M_L,b_k}^0},$$
(10)
where $`F_{L,s_j,b_k}`$ is a frequency response function for each sensor, which may depend also on the beam position. This normalisation factor is expected to be of order 1, and is kept fixed at 1 for the $`L`$=2 distributions. It is used as a free parameter for the $`L`$=1 distributions to compensate for the rather inaccurate knowledge of a) the sensor positions on the sphere’s surface, b) the beam track location and c) the electrons and photons shower development along the track, since the exact excitation strengths of the modes are quite sensitive to such data. As the first step in the fitting procedure we separately calculated the $`s_{L,M_L,b_k}𝐮_{L,M_L,s_j,b_k}`$, where $`s_{L,M_L,b_k}`$ is the mode’s strength from the beam excitation, as detailed in the appendix. We inserted the calculated $`s_{L,M_L,b_k}𝐮_{L,M_L,s_j,b_k}`$ in a hierarchical fitting model, to simultaneously fit the relevant parameters of eq. 10 to the 17.6 kHz, $`L`$=2 Fourier modulus $`A_{L,s_j,b_k}`$ for both sensors $`s_1`$ and $`s_2`$ at both beam positions E and A. This fit led to a reduced $`\chi ^2`$=1.3 at 59 degrees of freedom. Next, with fixed values for the sensor efficiencies $`\mathit{ϵ}_{𝒔_𝒋}`$ so established, we fitted the relevant parameters for the 24 kHz, $`L`$=1 Fourier peaks, including the $`L`$=1 sensor response factors $`F`$. At all stages the $`t_{M_L,b_k}^0`$ of the phases were kept within the bounds of the period of mode-$`L`$. With an uncertainty in the beam charge and in the Fourier peak amplitudes of $`20\%`$ each, the error amounts to $``$30%, and we took a minimum absolute error of 2$`\times 10^5`$ for sensor $`s_1`$ and 1$`\times 10^5`$ for sensor $`s_2`$. In total we have 152 data points, while the total number of fitted parameters is 27, including a relative normalising factor for the mean beam current at beam position A with respect to the mean current at beam position E. We found for the total fit a reduced $`\chi ^2`$=1.6 at 125 degrees of freedom. The $`L`$=1 response factors remain within 1.1 and 0.2.
The results of the fits to the 17.6 kHz and 24 kHz are given in figures 12 and 13 and table V. Note the different vertical scales used for sensor-1 and sensor-2 in both pictures. Some of the parameters given in the table are strongly correlated.
Table V. Results of a hierarchical fit to the data of the sphere at the 19 measuring angles of both sensors $`s_1`$ and $`s_2`$ of the Fourier modulus at 17.6 kHz and 24 kHz at beam positions E and A.
| sensor efficiency | fit result | error |
| --- | --- | --- |
| | $`10^4`$ V/ms<sup>-2</sup> | $`10^4`$ V/ms<sup>-2</sup> |
| $`ϵ_{s_1}r`$ | +0.10 | 0.02 |
| $`ϵ_{s_1}\theta `$ | -0.5 | 0.2 |
| $`ϵ_{s_1}\varphi `$ | +0.4 | 0.1 |
| $`ϵ_{s_2}r`$ | -0.5 | 0.2 |
| $`ϵ_{s_2}\theta `$ | -0.5 | 0.2 |
| $`ϵ_{s_2}\varphi `$ | +4.0 | 0.3 |
| $`intensity\left(\frac{beamA}{beamE}\right)`$ | 0.7 | 0.1 |
| response factor | | |
| $`F_{L=1,s_1,beamA}`$ | 0.3 | 0.1 |
| $`F_{L=1,s_2,beamA}`$ | 1.1 | 0.1 |
| $`F_{L=1,s_1,beamE}`$ | 0.2 | 0.1 |
| $`F_{L=1,s_2,beamE}`$ | 1.1 | 0.2 |
We conclude that the measured Fourier amplitude angular distributions are consistent with the model value for the $`L`$=2 and $`L`$=1 mode signature.
#### 4.2.1 The sphere’s absolute displacement
Finally, to estimate the order of magnitude of the sphere’s absolute displacement, we have to take an intermediate step by first normalising bar BU to the calibrated results for bar BC and then use bar BU as a calibration for the sphere. The sensors used on bar BU consisting of the same sensor material and having been cut roughly to the same size, we assume to be identical to the ones used on sphere SU. The amplifiers used are identical. The bar BU sensors, however, differ strongly from those of bar BC.
We arrive at an indirectly calibrated value for $`ϵ_{BU,2}=\left(3.410^4\pm 30\%\right)`$ V/ms<sup>-2</sup> of sensor-2 on bar BU. The value of $`ϵ_{BU,1}`$ for sensor-1 is about ten times smaller.
Then, for the sphere, the fitted value of $`\mathit{ϵ}_{𝒔_𝒋}`$ given in table V, shows the largest value of sensor-2, $`ϵ_{\varphi SU}=\left(410^4\pm 10\%\right)`$V/ms<sup>-2</sup>, to lead to a ratio $`\left(1.2\pm 0.4\right)`$ with $`ϵ_{BU}`$. Again, the values for sensor-1 are about ten times smaller. The error of $`33\%`$ is the propagated statistical error only. The result seems reasonable. So, the model calculation and our sphere measurement results are of the same order of magnitude on an absolute scale too.
From the maximum Fourier modulus, $`V^{max}=\left(0.003\pm 0.001\right)`$ V, of the 17.6 kHz, $`L`$=2 sphere mode measured in sensor sensor-2 as given in fig. 13, and the absorbed energy of 3.1 J, we find the maximum sphere’s displacement to correspond to $`\left(0.2\pm 0.1\right)`$nm/J.
## 5 Discussion
Having confirmed the thermo-acoustic conversion model in the present experiment, we discuss some points about extrapolating these results to the actual operation of a resonant gravitational wave detector. Firstly, in our experiment many incident particles deposited their energy in the resonator, in contrast to a single muon hitting an actual detector. However, from this difference it seems unlikely to reach different conclusions, especially since in the process of depositing energy along its track, the muon will generate lots of secondary particles too. Also, we measured at room temperature while actual detectors would have to operate in the millikelvin range. An aluminium resonator, for instance, at such a temperature, would be superconducting, and it is as yet unclear how the decoupling of the electron gas from the lattice would affect the process of acoustic excitation.
Therefore, we hold it of particular importance for the prospected shielding of a next generation resonant mass gravitational wave detector that an existing millikelvin detector like the Nautilus , would succeed in measuring the impinging cosmic rays in correlation with the resonator mode. Such a result, as a test for the further applicability of the thermo-acoustic conversion model at operating temperature, would come closest to the real situation envisaged for the new detectors.
Apart from such temperature effects, the applicability of the thermal acoustic conversion model is confirmed by the data and therefore cosmic rays should be expected to seriously disrupt, as calculated by the model, the possibility of detecting gravitational waves. It is beyond the scope of this article to go into any detail . We want to point, however, to earlier calculations which, having used the model, clearly show, firstly, that a next generation spherical resonant mass gravitational wave detector of ultra high sensitivity will be significantly excited by cosmic rays. Secondly, the high impact rate of cosmic rays will prohibit gravitational wave detection at the earth’s surface, with the required sensitivity.
Finally, shielding the instrument by an appreciable layer of rock as available in, for instance, the Gran Sasso laboratory, would suppress the cosmic ray background by a factor of $`10^6`$. Even then a vetoing system will be necessary and, at the radical reduction of the background rate so established, it may indeed work effectively.
## Acknowledgements
We thank A. Henneman for the computer code to calculate a sphere’s vibrational modes, R. Rumphorst for his knowledgeable estimate of sensor sensitivity, J. Boersma for digging out the formal orthogonality proof of a sphere’s eigenmodes, and the members of the former GRAIL team for expressing their interest in this study, especially P.W. van Amersfoort, J. Flokstra, G. Frossati, H. Rogalla, A.T.M. de Waele. This work is part of the research programme of the National Institute for Nuclear Physics and High-Energy Physics (NIKHEF) which is financially supported through the Foundation for Fundamental Research on Matter (FOM), by the Dutch Organisation for Science Research (NWO).
## Appendix: Sphere excitation model calculation
Our calculation of the ($`L,M`$)-mode excitation strengths is based on the source term of eq. 5.10/11 of ref. , $`s=\mathrm{\Sigma }/\left(\rho V\right)𝑑z_{}𝐮`$, with $`\mathrm{\Sigma }=\gamma dE/dz`$. Here, $`\gamma `$ is the Grueneisen constant, $`\rho V`$ de sphere’s mass, and $`dE/dz`$ the absorbed energy per unit track length. The Fourier amplitudes, measuring the second time derivative of the mode amplitudes, are directly proportional to $`s`$, and the mode amplitudes follow from $`s/\omega ^2`$, as in eq. 5.18 of . However, the amount of energy absorbed per unit length in our case depends on the particle’s position along the track. We therefore re-included the $`\mathrm{\Sigma }`$-term under the source term’s integral by letting $`dE\left(z\right)/dz`$ represent the electromagnetic cascade development of ref. as an approximation to the amount of energy absorbed per unit track length by the sphere at position z along the beam track,
$$s_{L,M_L}=\kappa _L_{}𝐮_{L,M_L}\left(z\right)\frac{dE\left(z\right)}{dz}𝑑z,$$
(11)
where $`z`$ is being measured from the beam’s entrance point into the sphere. With $`E_{abs}`$ being the total amount of energy absorbed by the sphere from the electron bunch, we write $`dE\left(z\right)/dz=E_{abs}d\left(E\left(z\right)/E_{abs}\right)/dz`$ and use the polynomial expansion $`d\left(E\left(z\right)/E_{abs}\right)/dz=_{i=0}^3c_iz^i`$. Then $`_{z_{min}}^{z_{max}}𝑑zd\left(E\left(z\right)/E_{abs}\right)/𝑑z`$=1. For the polynomial, measuring $`z`$ in meter, we acquired the values $`c_0`$=0.8332 m<sup>-1</sup>, $`c_1`$=226 m<sup>-2</sup>, $`c_2`$=-1832 m<sup>-3</sup>, $`c_3`$=4909 m<sup>-4</sup> from a fit to the form given in ref. , with less than a percent deviation for our case of $`0z0.15`$ m. The value for the energy absorbed by the sphere from a single electron, $`E_{abs}^e`$=123 MeV, we got from both our Monte Carlo simulation using GEANT , and from EGS4 . At the measured 25 nC beam pulse charge this corresponds to a total $`E_{abs}=3.1`$J absorbed by the sphere. Then the value of $`\kappa E_{abs}=\gamma E_{abs}/M`$=1.00 m<sup>2</sup>/s<sup>2</sup>, for our case of $`M=\rho V`$=4.95 kg and $`\gamma `$=1.6. Our sphere has a suspension hole which leads to a slight shift in the frequencies and the spatial distribution of the modes, with respect to those of a sphere without a hole . We approximated, however, our sphere’s modes by the ideal hole-free sphere’s eigenmode solutions u($`z`$) , using the available computer code as established in ref. , and renormalising to $`𝐮𝐮𝑑V=V`$, as used in ref. from eq. 5.6 onward. The source term $`s_{L,M_L}`$ was calculated for each mode ($`L,M`$) by numerically integrating eq. 11. We checked the surface term in the numerical procedure to be negligible, as assumed in the partial integration leading to the form of $`s`$ used by ref. . Each u($`\varphi `$) in eq. 10 is the eigenmode solution, calculated for each sensor on the $`\varphi `$-grid of the measured data, and each term $`s_{L,M_L,b_k}`$ is the excitation factor $`s_{L,M_L}`$ at the specific beam position $`b_k`$. |
warning/0001/math0001069.html | ar5iv | text | # Maslov class and minimality in Calabi-Yau manifolds
## 1 Introduction
The Maslov class $`[\mu _\mathrm{\Lambda }]`$ of a Lagrangian embedding $`j:\mathrm{\Lambda }V`$ in the standard Euclidean symplectic vector space $`V`$ has been constructed by Maslov in the study of global patching problem for asymptotic solutions of some PDEs (see for further details on this point of view). Subsequently, this cohomological class has found applications in the analysis of several quantization procedure, starting from up to recent aspects on its relations with asymptotic, semiclassical and geometric quantization, for which we refer to . In spite of this, there are several problems in the very definition of the Maslov class for Lagrangian submanifolds of generic symplectic manifolds.
In it has been proved that, for a Lagrangian embedding $`j:\mathrm{\Lambda }V`$ in a Euclidean symplectic vector space $`(V,\omega )`$, the Maslov form $`\mu _\mathrm{\Lambda }`$ can be represented by $`\mu _\mathrm{\Lambda }=i_H\omega `$, that is by the contraction of the symplectic form with the mean curvature vector field $`H`$ of the embedding $`j`$. Unfortunately, the very definition of Maslov form (and related class) as exposed in , and , depends on the fact that the Lagrangian submanifold $`\mathrm{\Lambda }`$ is embedded in a symplectic vector space, in which we have chosen a projection $`\pi :V\mathrm{\Lambda }_0`$ over a fixed Lagrangian subspace $`\mathrm{\Lambda }_0`$; then the Maslov class $`[\mu _\mathrm{\Lambda }]H^1(\mathrm{\Lambda },R)`$ can be defined as the Poincare’ dual to the singular locus $`Z(\mathrm{\Lambda })\mathrm{\Lambda }`$, where $`Z(\mathrm{\Lambda }):=\{\lambda \mathrm{\Lambda }|\mathrm{rk}(\pi _{}(\lambda ))<\mathrm{max}\}H_{n1}(\mathrm{\Lambda },Z)`$. In the classical literature it is proved that if one changes projection $`\pi `$, that is if one changes the reference Lagrangian subspace $`\mathrm{\Lambda }_0`$, then the Maslov class $`\mu _\mathrm{\Lambda }`$ does not change, while its representative changes. This is achieved using the so called universal Maslov class construction on the Lagrangian Grassmannian $`GrL(V)`$, (the homogeneous space which parametrizes Lagrangian subspaces of $`(V,\omega )`$, see , and ). These formulations depend heavily on the linear structure of the ambient manifold $`V`$; in particular it is assumed that $`V`$ is endowed with the trivial connection. Therefore, it seems difficult even to define the Maslov class for Lagrangian submanifolds of symplectic manifolds, which are not vector spaces. For instance, it is possible to define the Maslov class of a Lagrangian embedding via the so called generating functions, or their generalization (Morse families), for which we refer to , and particularly . In this way, one obtains a notion of Maslov class for Lagrangian submanifolds embedded in any cotangent bundle $`T^{}M`$ over a Riemannian manifold $`M`$, constructing a $`Z`$-valued $`\stackrel{ˇ}{\mathrm{C}}`$ech cocycle, starting from the signature of the Hessian of a Morse family; however this construction depends strongly on the choice of a base manifold ($`M`$ in the case of the cotangent bundle) and does not seem to be generalizable to Lagrangian embedding in any symplectic manifold. (See for more details on this kind of construction).
Recently (see ), Fukaya has shown how to define a Maslov index for closed loops on Lagrangian submanifolds of a quite general class of symplectic manifolds, the so called pseudo-Einstein symplectic manifolds. The construction is developed using non trivial assumptions on the structure of the ambient manifold and is carried on only for a particular subclass of Lagrangian submanifolds; moreover, there is no explicit reference to the corresponding Maslov class.
In this paper we show that, whenever the ambient manifold is Calabi-Yau, it is possible to give a consistent definition of Maslov class for its Lagrangian submanifolds, generalizing the approach of Arnol’d with the so called universal Maslov class. In this framework, we show that it is possible to generalize the result of Morvan and then we comment on various consequences of our construction, in particular on the possible definition of Maslov class for Lagrangian embedding in any symplectic manifold.
## 2 The Maslov class for Lagrangian embedding in Calabi-Yau
Let us briefly recall the standard construction of the Maslov class $`\mu _\mathrm{\Lambda }`$, for a Lagrangian submanifold $`\mathrm{\Lambda }`$, embedded in a symplectic vector space $`(V,\omega )`$, of real dimension $`2n`$: first of all, one considers the tangent spaces to $`\mathrm{\Lambda }`$ as (affine) subspaces of $`V`$. Then, using the trivial parallel displacement one transports every tangent plane in a fixed point $`P`$ of $`V`$, (for example the origin). Now, one has to consider the Lagrangian Grassmannian $`GrL(T_PV)`$, which by definition parametrizes all Lagrangian subspaces of $`T_PV`$. Using the trivial connection, we have thus obtained a map:
$$G:\mathrm{\Lambda }GrL(T_PV).$$
It is easy to see (, ),that $`GrL(T_PV)`$ has the natural structure of the homogeneous space $`{\displaystyle \frac{U(n)}{O(n)}}`$; then by the standard tool of the exact homotopy sequence for a fibration (see), it is proved that $`\pi _1(GrL(T_PV))Z`$. In fact, having fixed a Lagrangian plane $`\mathrm{\Lambda }_0`$ in $`T_PV`$, all other Lagrangian planes are obtained via a unitary automorphism $`AU(n)`$. Obviously, we have a fibration:
$$SU(n)U(n)\stackrel{det}{}S^1,$$
but this does not descend to $`GrL(T_PS)`$, since we have to quotient out the possible orthogonal automorphisms. However, since the square of the determinant of an orthogonal automorphism is always 1, we have a well defined map:
$$det^2:GrL(T_PS)S^1,$$
which sits in the following commutative diagram of fibrations:
$$\begin{array}{cccccc}SO(n)& & O(n)& \stackrel{det}{}& S^0& \\ & & & & & \\ SU(n)& & U(n)& \stackrel{det}{}& S^1& \\ & & & & & z^2\\ GrSL(C^n)& & GrL(C^n)& \stackrel{det^2}{}& S^1& \end{array}$$
In this diagram the space $`GrSL(C^n)`$ denotes the Grassmannian of special Lagrangian planes in $`C^n`$, that is the Grassmannian of Lagrangian planes which are calibrated by the top holomorphic form of $`C^n`$; the corresponding Lagrangian submanifolds are called special Lagrangian (see for more details). Notice that this space is always simply connected.
Finally, using Hurewicz isomorphism and taking a generator belonging to $`H^1(GrL(T_PV),Z)`$, which is thought as the pull-back via $`det^2`$ of the generator $`[\alpha ]H^1(S^1,Z)`$, one defines the Maslov class $`[\mu _\mathrm{\Lambda }]:=G^{}(det^2)^{}[\alpha ]`$. Obviously, this construction is indipendent on the choice of the point $`P`$, since if another point is chosen it is possible to construct a homotopy in such a way to prove the invariance of $`[\mu _\mathrm{\Lambda }]`$. It is clear that, in this framework, the existence of the trivial connection is an (almost!) essential requirement for the construction to work. In fact, we will see in this section that, to have a consistent definition of Maslov class it is not necessary that the ambient manifold is endowed with the trivial connection, but is sufficient that the global holonomy of the symplectic manifold is small in a suitable sense.
From now on we restrict our attention to Lagrangian submanifolds of Calabi-Yau manifolds. Recall that Calabi-Yau manifolds can be defined as compact Kaehler manifolds with vanishing first Chern class; recall also that a celebrated theorem by Yau (proving a previous conjecture by Calabi) implies that for every choice of the Kaehler class on a Calabi-Yau, there exists a unique Ricci-flat Kaehler metric. Moreover, while the holonomy of a Kaehler manifold is contained in U(n), if $`g`$ is the Ricci-flat metric of an n-dimensional Calabi-Yau, then the corresponding holonomy group is contained in SU(n). Finally, let us recall that, on every Kaehler manifold $`(X,g,J)`$ (where $`g`$ is a Kaehler metric and $`J`$ the integrable almost complex structure) the corresponding symplect or Kaehler form $`\omega `$ is related to $`g`$ via:
$$\omega (X,Y):=g(X,JY)X,Y\mathrm{\Gamma }(TX),$$
(1)
and that the almost complex structure tensor $`J`$ is covariantly constant with respect to the Levi-Civita connection induced by $`g`$. Considering a Kaehler metric $`g`$ on a Calabi-Yau, we will always mean the Ricci-flat metric. Typical examples of Calabi-Yau are given by the zero locus of a homogeneous polynomial of degree $`n+1`$ in $`P^n(C)`$ (whenever this locus is smooth); however it is by no means true that all Calabi-Yau are algebraic. For further details on this class of manifolds see for example and .
The construction of Fukaya for defining the Maslov index of closed loops goes as follow (see for details and motivations). He considers symplectic manifolds $`(X,\omega )`$ which are pseudo-Einstein in the sense that there exists an integer $`N`$ such that $`N\omega =c_1(X)`$. By this relation, the line bundle $`det(TX)`$ is flat when restricted to every Lagrangian submanifold $`\mathrm{\Lambda }`$ of $`X`$, but Fukaya restricts further the class of Lagrangian submanifolds considering only the so called Bohr-Sommerfeld orbit $`\mathrm{\Lambda }`$ (BS-orbit for short), which are defined as the Lagrangian submanifolds for which the restriction of $`det(TX)`$ is not only flat, but even trivial. This implies that if we consider a closed loop $`h:S^1\mathrm{\Lambda }`$ ($`\mathrm{\Lambda }`$ is a BS-orbit), then the monodromy $`M`$ of the tangent bundle $`TX`$ along $`h(S^1)`$ is contained in $`SU(n)`$. Then the idea is to take a path in $`SU(n)`$ joining $`M`$ with the identity, in order to get an induced trivialization of $`h^{}(TX_{|h(S^1)})S^1\times C^n`$. In this trivial bundle there is a family of Lagrangian vector subspaces $`T_{h(t)}\mathrm{\Lambda }`$ and in this way we get a loop in $`GrL(C^n)`$ and hence a well-defined integer (the Maslov index) $`m(h)`$. Obviously $`m(h)`$ is independent of the choice of the path in $`SU(n)`$ which joins $`M`$ to the unit, since $`\pi _1(SU(n))1`$.
Now we come to our construction. Consider embedded Lagrangian submanifolds $`\mathrm{\Lambda }`$ of a Calabi-Yau $`(X,\omega ,g,J)`$, where $`\omega ,g,J`$ are related by (1). Define the Lagrangian Grassmannization $`GrL(X)`$ of $`TX`$ as the fibre bundle over $`X`$ obtained substituting $`T_xX`$ with $`GrL(T_xX)`$, thus:
$$GrL(X):=\underset{xX}{}GrL(T_xX)$$
and in particular:
$$GrL(X)_\mathrm{\Lambda }:=\underset{x\mathrm{\Lambda }}{}GrL(T_xX).$$
Let $`G(j)`$ be the Gauss map, which takes $`x\mathrm{\Lambda }`$ in $`T_x\mathrm{\Lambda }`$ thought as a Lagrangian subspace of $`T_xX`$. Via $`G(j)`$, the embedding $`j:\mathrm{\Lambda }X`$ lifts to a section $`G(j):\mathrm{\Lambda }GrL(X)_\mathrm{\Lambda }`$. We would like to define the Maslov class of $`\mathrm{\Lambda }`$ via a map $`:\mathrm{\Lambda }S^1`$ in the following way: to every point $`x\mathrm{\Lambda }`$, we consider $`G(j)(x)`$ and then through the isomorphism $`GrL(T_xX){\displaystyle \frac{U(n)}{O(n)}}`$, taking the map $`det^2`$ we get a point in $`S^1`$. However, as we have seen, to establish an isomorphism to every space $`GrL(T_xX)`$ ($`x\mathrm{\Lambda }`$) with $`{\displaystyle \frac{U(n)}{O(n)}}`$ we need a reference Lagrangian plane in $`GrL(T_xX)`$ $`x\mathrm{\Lambda }`$, that is we need another section of $`GrL(X)_\mathrm{\Lambda }`$, besides $`G(j)(\mathrm{\Lambda })`$.
To this aim, fix a point $`p\mathrm{\Lambda }`$, consider $`T_p\mathrm{\Lambda }`$ and use the parallel displacement, induced by the Levi Civita connection of $`g`$, along a system $`\gamma `$ of paths on $`\mathrm{\Lambda }`$ starting from $`p`$, to construct a reference distribution of Lagrangian planes $`𝒟_\gamma `$ over $`\mathrm{\Lambda }`$, that is another section of $`GrL(X)_\mathrm{\Lambda }`$. This is indeed possible, since the holonomy is contained in $`U(n)`$, the parallel displacement is an isometry for $`g`$ and $`J`$ is covariantly constant: these facts, combined with the relation 1 imply that parallel transport sends Lagrangian planes in Lagrangian planes. Obviously this distribution $`𝒟_\gamma `$ is not uniquely determined, since it depends on the choice of the system of paths $`\gamma `$ starting from $`p`$. In spite of this, due to the fact that the holonomy of a Calabi-Yau metric is very constrained, this dependence does not prevent us to reach our goal. Indeed, consider $`q\mathrm{\Lambda }`$ and compare the two Lagrangian planes $`(𝒟_\gamma )_q`$ and $`(𝒟_\delta )_q`$ obtained by parallel transport of $`T_p\mathrm{\Lambda }`$ along two different paths $`\gamma `$ and $`\delta `$. By the holonomy property of a Calabi-Yau metric we have:
$$(𝒟_\gamma )_q=M(𝒟_\delta )_qMSU(n).$$
Thus, if $`AU(n)`$ is such that $`T_q\mathrm{\Lambda }=A(𝒟_\gamma )_q`$, then $`T_q\mathrm{\Lambda }=AM(𝒟_\delta )_q`$; so to every $`q\mathrm{\Lambda }`$ we can associate $`A_q`$ such that $`G(j)(q)=T_q\mathrm{\Lambda }=A_q(𝒟_\gamma )_q`$, where $`A_q`$ is determined up to multiplication by a matrix $`MSU(n)`$. At this point the key observation is that $`det^2(A_q)S^1`$ is a well defined point, which is not affected by the ambiguity of $`A_q`$. In this way we have a well-defined map, the Maslov map:
$$\begin{array}{cccc}:& \mathrm{\Lambda }& & S^1\\ & q& & det^2(A_q)\end{array}$$
Take the generator $`[\alpha ]`$ of $`H^1(S^1,Z)`$ represented by the form $`\alpha :=\frac{1}{2\pi }d\theta `$. Observe that the target space of the Maslov map, is not only topologically a circle, but even a Lie group, the group $`U(1)`$: this implies that the choice of the form $`\frac{1}{2\pi }d\theta `$ is compulsory, since it is the unique normalized invariant 1-form. Now we can give the following:
Definition: Using the previous notations, we define the Maslov form of the Lagrangian embedding $`j:\mathrm{\Lambda }X`$ as $`\mu _\mathrm{\Lambda }:=^{}\alpha `$ and the corresponding Maslov class as $`[\mu _\mathrm{\Lambda }]=^{}[\alpha ]H^1(\mathrm{\Lambda },Z)`$.
Remark 1 : The Maslov map $``$ has been built up fixing a reference point $`p`$, from which we constructed $`𝒟_\gamma `$; in this way the map $``$ associates to $`p`$ $`1S^1`$. It is clear that if one takes a different reference point $`p^{}`$, then the map $``$ changes (this time $`p^{}`$ goes to 1), but the Maslov class and the Maslov form do not change, as it is immediate to see. In particular, the invariance of the Maslov form is due to the invariance of $`\alpha `$ under the action of the Lie group $`U(1)`$.
Remark 2 : In , Trofimov costructed a generalized Maslov class, as a cohomological class defined on the space of paths $`[X,\mathrm{\Lambda }]`$; these paths start from a fix point $`x_0`$ in a symplectic manifold $`X`$ and end to a fixed Lagrangian submanifold $`\mathrm{\Lambda }`$ of $`X`$. We argue that the the Maslov class we have just defined can be obtained as a finite dimensional reduction of the class built up in , when one uses the Levi-Civita connection induced by the Calabi-Yau metric. In fact, Trofimov did not use metric connections, but instead affine torsion free connections, preserving the symplectic structure, which are generally not induced by a metric.
## 3 Representation of the Maslov class via the mean curvature vector field
In this section, generalizing what has been proved by Morvan in for Lagrangian embeddings in Euclidean symplectic vector space, we prove the following:
Theorem: Let $`j:\mathrm{\Lambda }X`$ be a Lagrangian embedding in a Calabi-Yau X and let $`H\mathrm{\Gamma }(N\mathrm{\Lambda })`$ be the mean curvature vector field of the embedding $`j`$ (with repect to the Calabi-Yau metric), then:
$$\mu _\mathrm{\Lambda }=\frac{1}{\pi }i_H\omega ,$$
where $`\omega `$ is the Kaehler form constructed from the Calabi-Yau metric $`g`$, and $`\mu _\mathrm{\Lambda }`$ is the Maslov form previously defined.
Before proving the theorem we need various preliminary results, which we are going to state and prove, and we need also to decompose into simpler pieces the action of $`^{}`$ on $`[\alpha ]`$.
Recall that given an embedding $`j`$, the associated second fundamental form $`\sigma :T\mathrm{\Lambda }\times T\mathrm{\Lambda }N\mathrm{\Lambda }`$ is a symmetric tensor defined by:
$$\sigma (X,Y):=_X^gY_X^{j^{}g}Y,X,Y\mathrm{\Gamma }(T\mathrm{\Lambda }),$$
where $`^g`$ is the Levi-Civita connection in the ambient manifold, while $`^{j^{}g}`$ is the connection induced on $`\mathrm{\Lambda }`$ via the pulled-back metric. If $`\sigma `$ is identically vanishing, then the submanifold is called totally geodesic. Taking the trace of $`\sigma `$ we get a field of normal vectors, that is the mean curvature vector field $`H`$ of the embedding $`j`$. Those embeddings for which $`H`$ is identically vanishing are called minimal.
First of all we need to understand the local structure of $`TGrL(T_xX)`$. Fix a point $`q\mathrm{\Lambda }`$ and set $`V:=T_qX`$ for short. We can prove the following:
Lemma 1: The space $`T_\pi GrL(V)`$ over a Lagrangian n-plane $`\pi `$ of $`V`$ can be identified with the subspace of linear maps $`\psi :\pi \pi ^{}`$ ($`\pi ^{}`$ denotes the orthogonal subspace in $`V`$ with respect to the metric $`g`$ in $`q`$) such that:
$$g(\psi (X),JY)=g(\psi (Y),JX),X,Y\pi .$$
Proof: First of all, we have $`T_\pi GrL(V)S(\pi )`$, where $`S(\pi )`$ is the space of all symmetric bilinear forms on $`\pi `$. In fact every $`vT_\pi GrL(V)`$ can be represented as $`{\displaystyle \frac{d}{dt}}B(t)\pi _{|t=0}`$, where $`B(t)`$ is a path of linear symplectic transformation of $`V`$, with the condition $`B(0)=id_V`$. To $`vT_\pi GrL(V)`$ we can associate a form $`S_v`$ given by:
$$S_v(X,Y):=\omega (\frac{d}{dt}B(t)X_{|t=0},Y).$$
This form is clearly bilinear and is symmetric:
$$S_v(X,Y)=\omega (\frac{d}{dt}B(t)X_{|t=0},B(t)Y_{|t=0})=$$
$$=\frac{d}{dt}\omega (B(t)X,B(t)Y)_{|t=0}\omega (B(t)X_{|t=0},\frac{d}{dt}B(t)Y_{|t=0})=0\omega (X,\frac{d}{dt}B(t)Y_{|t=0})=$$
$$=\omega (\frac{d}{dt}B(t)Y_{|t=0},X)=S_v(Y,X),$$
by the fact that $`B(t)`$ is a symplectic linear transformation of $`V`$ and by skewsymmetry of $`\omega `$. It is easy to verify that the corresponding map $`T_\pi GrL(V)S(\pi )`$ is an isomorphism. Moreover we have:
$$S_v(X,Y)=\omega (\frac{d}{dt}B(t)X_{|t=0},Y)\stackrel{(\text{1})}{=}g(\frac{d}{dt}B(t)X_{|t=0},JY)$$
and thus, identifying $`\psi :\pi \pi ^{}`$ with $`{\displaystyle \frac{d}{dt}}B(t)\pi _{|t=0}`$ we get the result.
By Lemma 1 it is clear that $`J`$ itself, restricted to $`q`$, can be considered not only as an element of $`T_\pi GrL(V)`$ but even as an invariant vector field on $`GrL(V)`$, that is $`J_q\mathrm{\Gamma }(TGrL(V))`$. Let $`e_1,\mathrm{},e_n`$ be an orthonormal basis of $`\pi `$ and $`f^1,\mathrm{},f^n`$ the corresponding dual basis, in such a way that $`Je_1,\mathrm{},Je_n`$ is a basis of $`\pi ^{}`$ and $`Jf^1,\mathrm{},Jf^n`$ the associated dual basis. Then $`J`$ as a vector belonging to $`T_\pi GrL(V)`$, can be represented as a section of $`\pi ^{}\pi ^{}`$, that is $`J=f^iJe_i`$ (Einstein summation convention is intended). From $`J`$ in this representation one can construct a 1-form $`\stackrel{~}{J}\mathrm{\Omega }^1(GrL(V))`$ using the paring induced by the metric, that is $`\stackrel{~}{J}=e_iJf^i`$. This 1-form has a quite outstanding role:
Lemma 2: Fix an arbitrary Lagrangian plane in $`V`$ in order to have a map $`det^2:GrL(V)S^1`$. Then:
$$(det^2)^{}(\alpha )=\frac{1}{\pi }\stackrel{~}{J},$$
so that $`\stackrel{~}{J}`$ defines a closed form on $`GrL(V)`$.
Proof: It is sufficient to prove that for every $`XT_\pi GrL(V)`$ one has $`(det^2)^{}(\alpha )(X)=\frac{1}{\pi }\stackrel{~}{J}(X)`$. Indeed:
$$(det^2)^{}(\alpha )(X)=(\alpha )(det_{}^2(X)),$$
so we are led to compute the tangent map to $`det^2`$. Assume for simplicity that $`\pi `$ is the reference Lagrangian plane in the isomorphism $`GrL(V){\displaystyle \frac{U(n)}{O(n)}}`$, so that it is represented by the identity matrix. Then, since $`T_\pi GrL(V){\displaystyle \frac{u(n)}{o(n)}}`$, consider a path $`\gamma :(ϵ,ϵ)u(n)`$, such that $`\gamma (0)=O`$ and such that its image in $`u(n)`$ has empty intersection with $`o(n)`$ (except for the zero matrix). The exponential mapping determines in this way a path in $`GrL(V)`$ through $`\pi `$. Now, we have:
$$\frac{d}{dt}det^2(e^{\gamma (t)})_{|t=0}=\frac{d}{dt}det(e^{2\gamma (t)})_{|t=0}=\frac{d}{dt}(e^{2Tr(\gamma (t))})_{|t=0}=$$
$$2Tr(\dot{\gamma }(0))=2Tr(X),$$
where $`\dot{\gamma }(0)`$ is identified with the tangent vector $`X`$ in $`T_\pi GrL(V)`$. Hence one gets:
$$(det^2)^{}(\alpha )(X)=(\alpha )(det_{}^2(X))=(\alpha )(2Tr(X))=\frac{1}{\pi }Tr(X).$$
On the other hand, $`X\mathrm{\Gamma }(\pi ^{}\pi ^{})`$, so that it can be represented as $`X=X_k^lf^kJe_l`$; thus one gets:
$$\stackrel{~}{J}(X)=(e_iJf^i)(X_k^lf^kJe_l)=X_i^i=Tr(X).$$
Till now we have worked only locally, having fixed a point $`q\mathrm{\Lambda }`$. To proceed we need to globalize the properties stated in lemma 1 and 2. Let us define the vertical tangent bundle $`VT(GrL(X)_\mathrm{\Lambda })`$ ($`VT(GrL)`$ for short) over $`GrL(X)_\mathrm{\Lambda }`$ as:
$$VT(GrL(X)_\mathrm{\Lambda }):=\underset{x\mathrm{\Lambda }}{}TGrL(T_xX);$$
notice that this is not the tangent bundle of $`GrL(X)_\mathrm{\Lambda }`$, since it is obtained taking the tangent bundle of the fibre only (thus the name vertical). Analogously, one can define the vertical cotangent bundle over $`GrL(X)_\mathrm{\Lambda }`$ as:
$$VT^{}(GrL(X)_\mathrm{\Lambda }):=\underset{x\mathrm{\Lambda }}{}T^{}GrL(T_xX),$$
(from now on denoted as $`VT^{}(GrL)`$ for short).
Now, by the previous reasoning and since $`J`$ is covariantly constant on a Kaehler manifold $`X`$, we have that $`J`$ defines a section of $`VT(GrL)`$ and analogously $`\stackrel{~}{J}`$ induces a section of $`VT^{}(GrL)`$. In order to globalize the result of lemma 2, observe that the section $`𝒟_\gamma `$ of $`GrL(X)_\mathrm{\Lambda }`$ over $`\mathrm{\Lambda }`$, defined in the previous section, enables one to give a well-defined map $`Det^2:GrL(X)_\mathrm{\Lambda }S^1`$ (one takes as a reference Lagrangian plane in $`GrL(T_xX)`$ the subspace $`(𝒟_\gamma )_x`$). It is clear that one gets immediately the following:
Corollary 1: Under the previous notations and considering the fibration $`Det^2:GrL(X)_\mathrm{\Lambda }S^1`$ induced by the reference distribution $`𝒟_\gamma `$ one has:
$$(Det^2)^{}(\alpha )=\frac{1}{\pi }\stackrel{~}{J}$$
where $`\stackrel{~}{J}`$ is viewed as a section of $`VT^{}(GrL)`$.
Via the Gauss map we can pull-back $`VT(GrL)`$ to $`\mathrm{\Lambda }`$:
$$\begin{array}{ccc}G(j)^{}(VT(GrL))& & VT(GrL)\\ & & pr_{VT}\\ \mathrm{\Lambda }& & GrL(X)_\mathrm{\Lambda }\end{array}$$
Lemma 3: The bundle $`G(j)^{}VT(GrL)`$ can be identified with the subspace of $`T^{}\mathrm{\Lambda }N\mathrm{\Lambda }`$ consisting of those sections $`\psi \mathrm{\Gamma }(T^{}\mathrm{\Lambda }N\mathrm{\Lambda })`$ (that is $`N\mathrm{\Lambda }`$-valued 1-forms on $`\mathrm{\Lambda }`$) such that:
$$g(\psi (X),JY)=g(\psi (Y),JX),X,Y\mathrm{\Gamma }(T\mathrm{\Lambda }).$$
Proof: By the very definition of pulled-back bundle, we have that:
$$G(j)^{}VT(GrL)\{(x;x^{},\pi ,X)\mathrm{\Lambda }\times VT(GrL):(x,T_x\mathrm{\Lambda })=G(j)(x)=$$
$$=pr_{VT}(x^{},\pi ,X)=(x^{},\pi )\},$$
which clearly implies the constraint $`x=x^{}`$ and $`T_x\mathrm{\Lambda }=\pi `$ so that:
$$G(j)^{}VT(GrL)\underset{x\mathrm{\Lambda }}{}T_{\pi =T_x\mathrm{\Lambda }}GrL(T_xX).$$
On the other hand, by lemma 1:
$$T_{\pi =T_x\mathrm{\Lambda }}GrL(T_xX)\{\psi \mathrm{\Gamma }(T_x^{}\mathrm{\Lambda }N_x\mathrm{\Lambda })\mathrm{such}\mathrm{that}:$$
$$g(\psi (X),JY)=g(\psi (Y),JX),X,YT_x\mathrm{\Lambda }\},$$
so one gets immediately the thesis.
The tangent application to the Gauss map is related to the second fundamental form as shown in the following:
Lemma 4: The tangent map to $`G(j)`$ in a point $`x\mathrm{\Lambda }`$ can be identified with the second fundamental form $`\sigma `$, thought of as an application with values in $`T^{}\mathrm{\Lambda }N\mathrm{\Lambda }`$; more exactly $`\sigma `$ takes values in the subspace $`G(j)^{}(VT(GrL))`$ of $`T^{}\mathrm{\Lambda }N\mathrm{\Lambda }`$, in the sense that it satisfies $`g(\sigma (X,Y),JZ)=g(\sigma (X,Z),JY)`$.
Proof: First of all, the identity $`g(\sigma (X,Y),JZ)=g(\sigma (X,Z),JY)`$ is a consequence of the fact that Lagrangian submanifolds of Kähler manifolds are always anti-invariant (also called totally real) submanifolds of top dimension (see page 35). Hence, always by result of , page 43, we have the desired relation. Finally, the fact that the tangent map to the Gauss map can be identified with the second fundamental form, via the action of the almost complex structure $`J`$ and the metric $`g`$, is a classically known result which can be found, for example in , page 196.
Observe that by lemma 3 and 4, the second fundamental form $`\sigma (X,.)`$, considered as a map taking values in $`T^{}\mathrm{\Lambda }N\mathrm{\Lambda }`$ is an element of $`G(j)^{}(VT(GrL))`$. Let us summarize the situation in the following diagram:
$$\begin{array}{ccccc}T\mathrm{\Lambda }& \stackrel{G(j)_{}}{}& G(j)^{}(VT(GrL))& T^{}\mathrm{\Lambda }N\mathrm{\Lambda }& VT(GrL)\\ & & & & \\ \mathrm{\Lambda }& \stackrel{G(j)}{}& G(j)(\mathrm{\Lambda })& & GrL(X)_\mathrm{\Lambda }\end{array}$$
Denote again with $`\stackrel{~}{J}`$ the restriction of $`\stackrel{~}{J}`$ to the bundle $`G(j)^{}(VT^{}(GrL))`$. By the previous diagram we can pull-back $`\stackrel{~}{J}`$ to a closed 1-form on $`\mathrm{\Lambda }`$ via $`G(j)^{}`$:
$$(G(j)^{}(\stackrel{~}{J}))(X)=\stackrel{~}{J}(G(j)_{}(X))=\stackrel{~}{J}(\sigma (X,.))X\mathrm{\Gamma }(T\mathrm{\Lambda }),$$
(2)
where the last equality in equation (2) is due to lemma 4 and the pairing between $`\stackrel{~}{J}`$ and $`\sigma (X,.)`$ is induced by the natural pairing between $`G(j)^{}(VT^{}(GrL))`$ and $`G(j)^{}(VT(GrL))`$, respectively.
Proof of the theorem: First of all, notice that the Maslov map $`:\mathrm{\Lambda }S^1`$ can be decomposed as $`=Det^2G(j)`$, as is immediate to see. Then $`\mu _\mathrm{\Lambda }:=^{}(\alpha )=G(j)^{}(Det^2)^{}(\alpha )`$ and so $`\mu _\mathrm{\Lambda }=\frac{1}{\pi }G(j)^{}(\stackrel{~}{J})`$, by lemma 2. Now $`\stackrel{~}{J}=e_lJf^l`$ and $`\sigma (X,.)`$ can be represented as $`\mathrm{\Gamma }(T^{}\mathrm{\Lambda }N\mathrm{\Lambda })\sigma (X,.)=\sigma _i^k(X)f^iJe_k`$. In this way we have that for all $`X\mathrm{\Gamma }(T\mathrm{\Lambda })`$:
$$(G(j)^{}(\stackrel{~}{J}))(X)=(e_lJf^l)(\sigma _i^k(X)f^iJe_k)=\sigma _i^i(X)=$$
$$=\underset{i}{}g(\sigma (X,e_i),Je_i)=(\mathrm{by}\mathrm{lemma}4)=\underset{i}{}g(\sigma (e_i,e_i),JX)=$$
$$=g(H,JX)=\omega (H,X)=i_H\omega (X).$$
Hence, one gets the result:
$$\mu _\mathrm{\Lambda }=G(j)^{}(\frac{1}{\pi }\stackrel{~}{J})=\frac{1}{\pi }i_H\omega H^1(\mathrm{\Lambda },Z).$$
(3)
By the result of the theorem, one can give the following:
Definition: Let $`\mathrm{\Lambda }X`$ a Lagrangian embedding in a Calabi-Yau $`X`$; then the Maslov index $`m`$ of a closed loop $`\gamma `$ on $`\mathrm{\Lambda }`$ is given by:
$$m(\gamma ):=\frac{1}{\pi }_\gamma i_H\omega Z.$$
## 4 Conclusions
Calabi-Yau manifolds have received great attention as target spaces for superstring compactifications. Moreover their Lagrangian and special Lagrangian submanifolds are now considered as the cornerstones for understanding the mirror symmetry phenomenon between pairs of Calabi-Yau spaces, both from a categorical point of view (), and from a physical-geometrical standpoint (). Let us recall that special Lagrangian submanifolds $`\mathrm{\Lambda }`$ of a Calabi-Yau $`X`$ are exactly what are called BPS states or supersymmetric cycles in the physical literature; on the other hand, it is known that special Lagrangian submanifolds are nothing else that minimal Lagrangian submanifolds (compare page 96, where this is proved for special Lagrangian submanifolds of $`C^n`$). From our result it turns out that the Maslov class of special Lagrangian submanifolds is identically vanishing; on the other hand, this can be seen just by considering the Grassmannian of special Lagrangian planes, which turns out to be diffeomorphic to $`{\displaystyle \frac{SU(n)}{SO(n)}}`$, hence simply connected (notice that the Grassmannian of special Lagrangian planes is isomorphic to the fibre in the fibration $`det^2:GrL(C^n)S^1`$). It is then clear that the Maslov index is identically vanishing for all special Lagrangian submanifolds $`\mathrm{\Lambda }`$ of a Calabi-Yau $`X`$. We believe that this simple observation can enhance our understanding of the structure of the $`A^{\mathrm{}}`$-Fukaya category, whenever its objects are restricted to minimal Lagrangian submanifolds (see for a definition of $`A^{\mathrm{}}`$ category, and for its application in the study of mirror symmetry). Indeed, this is a key point for the proof of homological mirror symmetry for K3 surfaces, for which we refer to .
The Maslov class so far constructed does not depend on the choice of a canonical projection, from which one could determine the singular locus (as usually happens when one considers Lagrangian embedding in cotangent bundles over an arbitrary Riemannian manifold). However, it is still possible to determine, rather then the singular locus, the homology class $`[Z]H_{n1}(\mathrm{\Lambda },Z)`$ of a singular locus , just considering the Poincare’ dual to $`[\mu _\mathrm{\Lambda }]`$, and setting $`[Z]:=Pd([\mu _\mathrm{\Lambda }])`$ ($`Pd`$ stands for Poncare’ duality). We have said a singular locus , because $`Z`$ is not determined at all uniquely, but only up to its homology class; in spite of this one could take as singular locus any representative of $`[Z]`$. So it makes sense to speak of a singular locus, even if there is no projection to which to refer it.
It is clear that it is not possible to extend our definition of Maslov class for Lagrangian embedding in arbitrary symplectic manifolds; even the construction of Fukaya (which is specifically designed for Maslov index of closed loops only on BS orbits) needs several assumption such that the ambient manifold admits a prequantum bundle and so on. We are thus tempted to suggest the following alternative description: we would like to define the Maslov class for a Lagrangian embedding in any symplectic manifold $`(X,\omega )`$, via the mean curvature representation $`i_H\omega `$. Two problems arise following this approach. First of all, to define the mean curvature vector field $`H`$ it is necessary to fix a Riemannian metric on $`X`$; as it is well known, on any symplectic manifold one has lots of Riemannian metrics $`g_J(X,Y):=\omega (X,JY)`$, constructed using the given symplectic form $`\omega `$ and choosing an $`\omega `$-compatible almost complex structure $`J`$; (recall that the set of $`\omega `$-compatible almost complex structures on a given symplectic manifold is always non empty and contractible, see ). What is the right choice for $`g_J`$?
Once we have fixed the right metric, the second problem is related to the closure of the 1-form $`i_H\omega `$, considered as a form on $`\mathrm{\Lambda }`$; indeed there is no reason, a priori, for which $`i_H\omega `$ has to be closed. We are thus led to the following:
Conjecture: Having fixed the Lagrangian embedding $`j:\mathrm{\Lambda }X`$, on any symplectic manifold $`(X,\omega )`$ there exists at least one Riemannian metric $`g_J`$ built up from an $`\omega `$-compatible almost complex structure $`J`$, such that the 1-form $`i_H\omega `$, considered as a form on $`\mathrm{\Lambda }`$ is closed. Multiplying the corresponding cohomological class $`[i_H\omega ]`$ for a suitable constant in such a way that it is integer valued, we call this class the Maslov-Morvan class of the Lagrangian submanifold $`\mathrm{\Lambda }`$.
It does not seem possible to give an interpretation of this conjectured Maslov-Morvan class via the universal Maslov class, as we have done for Calabi-Yau manifolds, since, in general, we have no control on the holonomy of $`g_J`$.
Clearly, the study of the relations between the conjectured class $`[i_H\omega ]`$ and the ordinary Maslov class for a Lagrangian embedding in cotangent bundles (via Morse families) deserves further effort and is left for future investigations.
Acknowledgements: It is a pleasure to thank E. Aldrovandi, U.Bruzzo, F. Cardin and M. Spera for encouragement and useful discussions. |
warning/0001/cond-mat0001074.html | ar5iv | text | # Condensation of a Hard-Core Bose Gas
## I Introduction
Interacting bosonic quantum systems are of special interest because of the effect of Bose condensation. The classical example is the Bose-Einstein condensation in an ideal Bose gas . The analogous phenomenon in a real (i.e. interacting) Bose gas is the transition from a normal to a superfluid state (e.g., in <sup>4</sup>He). Another related phenomenon is the condensation of a cold gas of bosonic atoms in a magnetic trap . There is a number of recent theoretical investigations of the effect of the interaction on the critical properties of the normal-superfluid transition based on Monte Carlo simulations and analytic calculations .
A fundamental model for the description of interacting bosons is the Ginzburg-Landau theory, usually called the Gross-Pitaevskii approach to Bose condensates . However, there are dense systems of bosons (like the classical <sup>4</sup>He superfluid), for which the Ginzburg-Landau theory is valid only very close to the phase transition from the normal fluid to the superfluid, where the order parameter is small. Away from the phase transition the interaction of the order parameter field is more complicated than described by this theory. Moreover, the Ginzburg-Landau theory is only a description of the order parameter (which is related to the density of the superfluid) but does not take into account the interaction with the non-superfluid part of the system. Consequently, it does not provide reliable information, e.g., for the value of the critical temperature. The question of the effect of the interaction on the latter has been discussed in the recent literature in great detail and with controversal results . It seems that the critical temperature can be shifted by a variation of the density of bosons. In particular, at low density it was found that the shift $`\mathrm{\Delta }T_c/T_0(na^3)^\alpha `$, where $`T_0`$ is the critical temperature of the ideal (non-interacting) Bose gas, $`n`$ the total density of bosons and $`a`$ the scattering radius of the hard-core interaction. Depending on the calculational method and approximations, the exponent $`\alpha `$ varies between $`1/3`$ and $`1/2`$ . Recent Monte Carlo simulations support $`\alpha =1/3`$. This value was also obtained by a self-consistent calculation of the quasiparticle spectrum and in a $`1/N`$ expansion . At a high density the critical temperature reaches a maximum and decreases with even higher densities. The latter is a consequence of the depletion of the condensate due to interaction.
In order to give a complete overview of the properties of an interacting Bose gas we need a model which takes fully into account all parts of the system of bosons. Close to the critical point, however, it should lead to the Ginzburg-Landau theory. Such a model was given by hard-core bosons, based on a slave-boson representation . Here we will briefly discuss this model and evaluate its critical temperature for different densities.
The paper is organized as follows: In Sect. 2 the slave-boson respresentation of hard-core bosons on a lattice is introduced. Then in Sect. 3 two collective fields are defined. One represents the superfluid condensate the other the total density of bosons, as discussed in Sect. 4. In Sect. 5 the total density, the condensate density and the critical temperature are calculated in mean-field approximation. Finally, concluding remarks and a discussion are given in Sect. 6.
## II The Model: Slave-Boson Representation
A continuous system of hard-core bosons with scattering length $`a`$ is approximated by a lattice Bose gas with lattice constant $`a`$. Although this approximation is limited because it restricts configurations of bosons to be commensurate with the lattice structure, it is more suitable for the investigation of a dense system of bosons than the usually considered $`|\mathrm{\Phi }|^4`$ (Ginzburg-Landau) theory. The representation of the model uses the slave-boson approach to hard-core bosons . (Originally the slave-boson approach was invented for the (fermionic) Hubbard model .) The latter relies on a picture in which a particle trades its position with an empty site on the lattice. Both, the particle as well as the empty site, are described by corresponding creation and annihilation operators. In a functional integral representation of a grand canonical ensemble of bosons with chemical potential $`\mu `$ at temperature $`T`$ this can be formulated in terms of a complex field $`b_x`$ of bosons and a complex field $`e_x`$ of empty sites with the action
$$S_{s.b.}=\frac{1}{T}\underset{x,x^{}}{}b_x^{}e_xt_{x,x^{}}b_x^{}e_x^{}^{}\frac{1}{T}\underset{x}{}\mu |b_x|^2,$$
where the first term describes the exchange of bosons and empty sites in a hopping process at sites $`x`$ and $`x^{}`$ with rate $`t_{x,x^{}}`$. A local constraint $`|e_x|^2+|b_x|^2=1`$ takes care of the complementary character of the bosons and empty sites. In $`S_{s.b.}`$ we consider only the thermal fluctuations (i.e. a vanishing Matsubara frequency) because non-zero Matsubara frequencies are separated by a gap if the temperature $`T`$ is non-zero. Here we assume that the temperature enters through the action $`S_{s.b.}`$ but not through the constraint. Physical quantities can be calculated from the partition function
$$Z=e^{S_{s.b.}}\underset{x}{}\delta (|e_x|^2+|b_x|^21)db_xdb_x^{}de_xde_x^{}.$$
For instance, the total density of bosons $`n`$ is given as the response to a change of the chemical potential $`\mu `$
$$n=\frac{T}{N}\frac{\mathrm{log}Z}{\mu }=\frac{T}{N}\frac{1}{Z}\frac{Z}{\mu },$$
(1)
where $`N`$ is the number of lattice sites. To evaluate the density of the condensate $`\rho _s`$ we must add an external vector potential to the action $`S_{s.b.}`$ (i.e. a Peierls factor to the hopping matrix $`t`$) and measure the response to this potential . We shall return to the corresponding expression subsequently.
## III Collective-Field Representation
Since the fields $`b_x`$ and $`e_x`$ are subject to the local constraint, we can not treat them in a conventional way as order parameter fields. It is necessary to eliminate the constraint in the partition function which can be achieved by integration over the fields. For this purpose we introduce a complex collective field $`\mathrm{\Phi }_x`$ and a real field $`\phi _x`$ which break up the biquadratic term in the action $`S_{s.b.}`$. This can be written as the (Hubbard-Stratonovich) transformation
$$S_{s.b.}S^{}=T\underset{x,x^{}}{}\mathrm{\Phi }_x(1t)_{x,x^{}}^1\mathrm{\Phi }_x^{}^{}+T\underset{x}{}\phi _x^2+\underset{x}{}\left(\begin{array}{c}e_x\\ b_x\end{array}\right)\left(\begin{array}{cc}2\phi _x+T^1& \mathrm{\Phi }_x\\ \mathrm{\Phi }_x^{}& \mu T^1\end{array}\right)\left(\begin{array}{c}e_x^{}\\ b_x^{}\end{array}\right).$$
(2)
The new action $`S^{}`$ gives the same partition function which can be seen by integrating over the collective fields. The field $`\phi `$ is necessary in order to invert the hopping term which is now the positive matrix $`1t`$. Then we can perform the integration of the slave-boson fields. This becomes simple if the $`2\times 2`$ Hermitean matrix in Eq. (2) is diagonalized by a unitary transformation. The latter leaves the constraint $`|b_x|^2+|e_x|^2=1`$ invariant and gives the eigenvalues
$$\lambda _{x,\pm }=\phi _x+1/2T\mu /2T\pm \sqrt{(\phi _x+1/2T+\mu /2T)^2+|\mathrm{\Phi }_x|^2}.$$
Now the partition function reads
$$Z=Z_0e^{S_b}\underset{x}{}\delta (|b_x|^2+|e_x|^21)e^{T\phi _x^2\lambda _{x,+}|e_x|^2\lambda _{x,}|b_x|^2}$$
$$\times db_xdb_x^{}de_xde_x^{}d\phi _xd\mathrm{\Phi }_xd\mathrm{\Phi }_x^{},$$
where the only non-local term is
$$S_b=T\underset{x,x^{}}{}\mathrm{\Phi }_x(1t)_{x,x^{}}^1\mathrm{\Phi }_x^{}^{}.$$
The $`e_x`$ and $`b_x`$ integration can be carried out (s. Appendix A) which yields
$$Z=Z_0e^{S_b}\underset{x}{}e^{T\phi _x^2}\frac{e^{\lambda _{x,+}}e^{\lambda _{x,}}}{\lambda _{x,+}\lambda _{x,}}d\phi _xd\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}$$
$$=Z_0e^{S_b}\underset{x}{}e^{1/4T+\mu /2T}e^{T\phi _x^2}\frac{sinh(\sqrt{(\phi _x+\mu /2T)^2+|\mathrm{\Phi }_x|^2})}{\sqrt{(\phi _x+\mu /2)^2+|\mathrm{\Phi }_x|^2}}𝑑\phi _x𝑑\mathrm{\Phi }_x𝑑\mathrm{\Phi }_x^{}.$$
The constant $`Z_0`$ is the normalization factor of the Hubbard-Stratonovich transformation of Eq. (2). It is convenient to separate the field-independent factor $`_xe^{1/4T+\mu /2T}`$ to define the partition function
$$\overline{Z}=e^{S_b}\underset{x}{}e^{T\phi _x^2}\frac{sinh(\sqrt{(\phi _x+\mu /2T)^2+|\mathrm{\Phi }_x|^2})}{\sqrt{(\phi _x+\mu /2T)^2+|\mathrm{\Phi }_x|^2}}𝑑\phi _x𝑑\mathrm{\Phi }_x𝑑\mathrm{\Phi }_x^{}.$$
This partition function has the effective action for the collective field $`\mathrm{\Phi }_x`$
$$S=S_b+S_0,$$
where $`S_b`$ is its non-local (i.e. hopping) part and
$$S_0=\underset{x}{}\mathrm{log}(Z_1(|\mathrm{\Phi }_x|^2)$$
is its local (i.e. potential) part with
$$Z_1(|\mathrm{\Phi }|^2)=e^{T\phi ^2}\frac{sinh(\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }|^2})}{\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }|^2}}𝑑\phi .$$
## IV Interpretation of the Fields
The introduction of the collective field has completely separated the hopping part $`S_b`$ from the potential part $`S_0`$. The latter does not depend on the phase of the collective field. The hopping part alone describes free bosons (ideal Bose gas) with the usual complex field. This can be seen by writing the partition function of the ideal Bose gas as
$$Z_{IBG}=\underset{k}{}\left[\underset{n_k0}{}e^{n_k(ϵ_k\mu _0)/T}\right]=\underset{k}{}[1e^{(ϵ_k\mu _0)/T}]^1,$$
where $`k`$ is a quantum number which characterizes the system of bosons. This can also be expressed in terms of an integral over a complex field as
$$Z_{IBG}=\mathrm{exp}\left(\underset{k}{}[1e^{(ϵ_k\mu _0)/T}]|\mathrm{\Phi }_k|^2\right)\underset{k}{}d\mathrm{\Phi }_kd\mathrm{\Phi }_k^{}/\pi .$$
(Notice that a rescaling of the field yields only a factor to $`Z_{IBG}`$.) For the translational-invariant Bose gas $`k`$ is the wavevector and we have the energy
$$ϵ_k=\frac{\mathrm{}^2k^2}{2m}$$
with the boson mass $`m`$.
Using the Fourier components $`(1\stackrel{~}{t}_k)^1`$ in the non-local term $`S_b`$ we can compare the latter with the corresponding expression of the ideal Bose gas
$$\underset{k}{}[1e^{(ϵ_k\mu _0)/T}]|\mathrm{\Phi }_k|^2.$$
(3)
By setting the hopping term $`(1\stackrel{~}{t}_k)^1`$ equal to $`1e^{(1/T)(ϵ_k\mu _0)}`$ we obtain
$$\stackrel{~}{t}_k=(1e^{(ϵ_k\mu _0)/T})^1.$$
The chemical potential of the ideal Bose gas is restricted to $`\mu _00`$ whereas it can have any real value in the interacting Bose gas. In particular, we can choose $`\mu _00`$ such that $`\stackrel{~}{t}_k>0`$. Moreover, in the Bose gas we can now apply the continuum approximation
$$S_b/T=\underset{x,x^{}}{}\mathrm{\Phi }_x(1t)_{x,x^{}}^1\mathrm{\Phi }_x^{}^{}\left[\frac{\mathrm{}^2}{2m}\mathrm{\Phi }_x(^2\mathrm{\Phi }^{})_x+\alpha |\mathrm{\Phi }_x|^2\right]d^3x,$$
(4)
where $`\alpha =_x(1t)_{x,x^{}}^1`$. It is important to notice that $`\alpha =(1\stackrel{~}{t}_0)^1`$ is positive because $`1t`$ was defined as a positive matrix. This implies that $`S_b`$ is a positive quadratic form. $`S_b`$ can be compared with the expression of the ideal Bose gas if we replace $`\alpha `$ by $`\mu _0`$. Then the expression in (4) is an approximation of Eq. (3) for small $`ϵ_k\mu _0`$, which applies, e.g., to high temperatures. This approximation is very common in the literature and can also be used in the case of hard-core bosons.
Physical quantities can be expressed as expectation values of the new fields $`\phi `$ and $`\mathrm{\Phi }`$. For instance, we obtain from Eq. (1) for the total density of bosons the expression (see Appendix B)
$$n=\frac{1}{2}+T\phi _x,$$
where
$$\mathrm{}=\frac{1}{\overline{Z}}\mathrm{}e^{S_b}\underset{x}{}e^{T\phi _x^2}\frac{sinh(\sqrt{(\phi _x+\mu /2T)^2+|\mathrm{\Phi }_x|^2})}{\sqrt{(\phi _x+\mu /2T)^2+|\mathrm{\Phi }_x|^2}}d\phi _x\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}.$$
(5)
Thus $`\phi _x`$, which is conjugate to $`|e_x|^2`$ according to Eq. (2), is related to the total density of bosons. Conversely, $`\mathrm{\Phi }_x`$, which is conjugate to $`e_xb_x^{}`$ according to Eq. (2), corresponds to the density of the condensate and can be expressed as the expectation value
$$\rho _s=T\alpha |\mathrm{\Phi }_x|^2=\frac{T\alpha }{\overline{Z}}|\mathrm{\Phi }_x|^2e^{S_bS_0}\underset{x}{}d\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}.$$
We notice that the potential part of the partition function $`S_0`$ is symmetric with respect to $`\mu \mu `$. This implies that the expectation value $`\phi _x`$ is an odd function of $`\mu `$. Therefore, the density varies monotoneously with $`\mu `$, as it should. The symmetry of $`S_0`$ with respect to $`\mu \mu `$ implies that $`\rho _s`$ is an even function of $`\mu `$. This is a characteristic feature of our lattice hard-core bosons in which bosons and empty sites are dual to each other.
### A Near the Critical Point
The $`\phi `$-integration in $`Z_1`$ can be performed numerically in order to obtain an effective potential
$$\underset{x}{}\mathrm{log}(Z_1(|\mathrm{\Phi }_x|^2).$$
It is interesting to notice that the expression $`\mathrm{log}(Z_1(|\mathrm{\Phi }_x|^2)`$ is linear for large values of $`|\mathrm{\Phi }_x|`$ (cf. Fig. 1). Therefore, the quadratic term in $`S_b`$ suppresses the large fluctuations. This means that the $`|\mathrm{\Phi }_x|^4`$ approximation of the Ginzburg-Landau theory provides a stronger suppression of large fluctuations than the complete theory.
Near the critical point we can expand the free energy around $`|\mathrm{\Phi }_x|=0`$ because the order parameter field $`\mathrm{\Phi }`$ is small. The result of this expansion is the Ginzburg-Landau functional for the collective field with
$$T\underset{x,x^{}}{}\mathrm{\Phi }_x(1t)_{x,x^{}}^1\mathrm{\Phi }_x^{}^{}\underset{x}{}\left(b_1|\mathrm{\Phi }_x|^2+b_2|\mathrm{\Phi }_x|^4\right)$$
$$\left[\frac{T\mathrm{}^2}{2m}\mathrm{\Phi }_x(^2\mathrm{\Phi }^{})_x+(T\alpha b_1)|\mathrm{\Phi }_x|^2b_2|\mathrm{\Phi }_x|^4\right]d^3x,$$
(6)
where
$$b_1=\frac{Z_1^{}(0)}{Z_1(0)},b_2=\frac{1}{2}\left[\frac{Z_1^{\prime \prime }(0)}{Z_1(0)}\left(\frac{Z_1^{}(0)}{Z_1(0)}\right)^2\right].$$
Expression (6) is also known as the Gross-Pitaevskii functional which is often used as a model of a dilute Bose system . In our derivation we have $`\mu `$\- and $`T`$-dependent coefficients. As one can see in Fig. 1 the term with $`|\mathrm{\Phi }|^2`$ can change its sign with a changing $`\mu `$. On the other hand, away from $`\mathrm{\Phi }=0`$ the behavior is controlled by the confining part $`|\mathrm{\Phi }|^4`$. This is the typical Ginzburg-Landau picture of a second-order phase transition.
The effect of the $`|\mathrm{\Phi }|^4`$ interaction on $`T_c`$ in the model of Eq. (6) was recently investigated in detail in a self-consistent calculation and in a $`1/N`$ expansion . In this model it is assumed that the coefficient of the $`|\mathrm{\Phi }|^4`$ term depends on the the scattering length $`a`$ and the total density by a factor $`a^3n`$. This is valid at low density. The calculation shows that the critical temperature of the interacting system $`T_c`$, normalized with the critical temperature of the ideal Bose gas, is shifted as
$$1T_0/T_cc_0an^{1/3}.$$
(7)
Numerical investigations show that the critical temperature must descrease at higher densities. This depletion effect, which cannot be seen in the Gross-Pitaevskii approach, will be discussed in the slave-boson approach subsequently.
## V Mean-Field Approximation
Since $`\phi _x`$ is a field which appears only in local terms, it can be integrated out at each site independently for a given value of the condensate field $`\mathrm{\Phi }_x`$. The treatment of $`\mathrm{\Phi }_x`$ is more difficult since it appears in non-local term $`S_b`$. It can be studied in terms of the classical field equation
$`\left[{\displaystyle \frac{\mathrm{}^2T}{2m}}^2+T\alpha {\displaystyle \frac{Z_1^{}(|\mathrm{\Phi }_x|^2)}{Z_1(|\mathrm{\Phi }_x|^2)}}\right]\mathrm{\Phi }_x=0,`$
which is the extremum of the action $`S_b+S_0`$. A further simplification is the additional assumption that the condensate field varies only weakly in space: $`^2\mathrm{\Phi }0`$. This gives the mean-field or Thomas-Fermi approximation. Both densities, $`n`$ and $`\rho _s`$, can be evaluated in mean-field approximation. The mean-field free energy reads
$$F_{MF}=\frac{1}{N}\mathrm{log}Z=\frac{1}{4}\frac{\mu }{2T}+T\alpha |\mathrm{\Phi }|^2\mathrm{log}(Z_1(|\mathrm{\Phi }|^2)),$$
(8)
where $`|\mathrm{\Phi }|`$ must be at the minimum of $`F_{MF}`$. For a given chemical potential $`\mu `$ there is a critical value $`T_c`$ which separates two regimes: one regime for $`T<T_c`$ in which the minimum of the mean-field free energy is $`|\mathrm{\Phi }|^2>0`$ and another regime with $`|\mathrm{\Phi }|^2=0`$ for $`TT_c`$. This can be seen by plotting $`T\alpha |\mathrm{\Phi }|^2\mathrm{log}(Z_1(|\mathrm{\Phi }|^2))`$ (Fig. 1) which has, depending on $`\mu `$, either a minimum at $`|\mathrm{\Phi }|=0`$ or another one at $`|\mathrm{\Phi }|0`$. The minimal value $`|\mathrm{\Phi }|`$ varies continuously as one goes through $`T_c\alpha `$. The behavior of the densities as functions of $`\mu `$ at a fixed temperature is shown in Fig. 2.
### A Near the Critical Point
For a very dilute system the mean-field approximation is insufficient and the calculations of Refs. should be applied. However, at higher densities (more than $`n0.2`$) the mean-field approximation should be reliable. Then the critical temperature $`T_c`$ of the mean-field calculation is
$$T_c=b_1/\alpha Z_1^{}(0)/(\alpha Z_1(0)).$$
The decreasing behavior of $`T_c/T_0`$ ($`T_0n^{2/3}`$ is the condensation temperature of the ideal Bose gas) is shown in Fig. 3. Our mean-field result for $`T_c/T_0`$ agrees qualitatively with the Monte Carlo result of Ref. . However, we expect that fluctuations might reduce the critical temperature substantially close to $`n=1`$.
We can expand the mean-field free energy (8) in powers of $`|\mathrm{\Phi }|^2`$ up to $`|\mathrm{\Phi }|^4`$. Then the minimum of $`F`$ must satisfy the mean-field equation
$$|\mathrm{\Phi }|^2\frac{T\alpha Z_1(0)Z_1^{}(0)}{T\alpha Z_1^{}(0)Z_1^{\prime \prime }(0)}=\frac{T_cT}{(TT_c)b_12b_2/\alpha }\frac{T_cT}{2b_2/\alpha }(TT_c)$$
(9)
if the right-hand side is non-negative and $`|\mathrm{\Phi }|^2=0`$ otherwise. Since $`b_20`$ in a typical situation (cf. Fig. 1), there is a non-zero solution. The coefficient on the right-hand side of (9) can be evaluated numerically for a given value of $`\mu `$.
## VI Discussion
The slave-boson representation is given by two fields $`b_x`$, $`e_x`$ which are subject to a constraint: one represents empty the other singly occupied sites. These two fields are replaced by two collective fields $`\phi _x`$ and $`\mathrm{\Phi }_x`$ which have a direct physical interpretation. The former couples to $`|e_x|^2`$ and the latter couples to the product $`e_xb_x^{}`$. The introduction of the collective field $`\mathrm{\Phi }_x`$ enables us to integrate the slave-boson fields and to eliminate the constraint of the these fields. The effective field theory provides a “two-liquid” theory which is represented by the fields $`\phi _x`$ and $`\mathrm{\Phi }_x`$. These fields correspond with the total density $`n`$ and and the superfluid density $`\rho _s`$, respectively. The latter is the order parameter of the condensation whereas the total density $`n`$ does not indicate a critical behavior at the condensation point (cf. Fig. 2). Therefore, $`n`$ can be fixed near the critical point and the theory can be expanded in terms of a small order parameter field $`\mathrm{\Phi }_x`$. This yields the well-known Ginzburg-Landau (or Gross-Pitaevskii) theory with density-dependent parameters. In other words, there is a phase boundary in the $`\mu `$-$`T`$ phase diagram which separates the normal from the superfluid phase. In the vicinity of this phase boundary we can apply the Gross-Pitaevskii approach.
The condensate field can be treated in mean-field approximation, assuming a homogeneous order parameter. This reveals the phenomenon of depletion of the condensate due to a strong interaction among the bosons which has been observed in superfluid <sup>4</sup>He. In the slave-boson representation it is a consequence of the duality of empty and singly occupied lattice sites which is reflected by the constraint $`|e_x|^2+|b_x|^2=1`$. Thus physical quantities are symmetric with respect to a half-filled system (i.e. $`n=1/2`$ or $`\mu =0`$).
The lattice theory is not very accurate at densities $`n0.5`$ because it reduces the motion of bosons significantly in comparison with a continuous system. Moreover, the mean-field approximation neglects vortices and flucuations of the order parameter. From this point of view we can only expect a qualitative agreement of our results with those, e.g., obtained from experiments. Nevertheless, the mean-field approximation should be reasonable if the density is not too low. The critical exponent of the order parameter should be renormalized due to fluctuations, using the renormalization group for the three-dimensional $`|\mathrm{\Phi }|^4`$ model.
The main result is that the phase diagram of the slave-boson theory of the strongly interacting bose system has two normal phases (i.e. a dense and a dilute one) and a superfluid phase for intermediate densities. Near the transition points a Gross-Pitaevskii approach can be used to describe the physics of small order parameter field. The superfluid density is low in our hard-core system. This might be a consequence of the strong interaction which suppresses the superfluid density.
In conclusion, we established an effective field theory which enables us to evaluate the properties of a strongly interacting system of bosons. It takes into account the order parameter of the condensate and the total density by interacting fields. It describes the phase transition between a normal phase and a condensed phase. The phase transition were studied in mean-field approximation. We evaluated the density-dependent critical temperature at densities $`n>0.2`$ in which the mean-field approximation of the order parameter is reliable.
Acknowledgement:
I am grateful to S. Girvin for bringing Ref. to my attention.
Appendix A
The integration of the $`e`$\- and the $`b`$-field can be performed in $`Z`$ for each point $`x`$ independently
$$\delta (|b_x|^2+|e_x|^21)e^{\lambda _{x,+}|e_x|^2\lambda _{x,}|b_x|^2}𝑑b_x𝑑b_x^{}𝑑e_x𝑑e_x^{}.$$
$`(A.1)`$
The integrand does not depend on the phases of the field. Therefore, the phase integration contributes a factor $`4\pi ^2`$. Moreover, we set $`s:=|b_x|^2`$ and $`t:=|e_x|^2`$ which yields for (A.1)
$$\pi ^2_0^{\mathrm{}}_0^{\mathrm{}}\delta (s+t1)e^{\lambda _{x,+}t\lambda _{x,}s}𝑑s𝑑t=\pi ^2_0^1e^{\lambda _{x,+}t\lambda _{x,}(1t)}𝑑t$$
$$=\pi ^2\frac{e^{\lambda _{x,+}}e^{\lambda _{x,}}}{\lambda _{x,+}\lambda _{x,}}$$
Appendix B
To write the total density of bosons we have to evaluate
$$\frac{T}{N}\frac{\mathrm{log}Z}{\mu }=\frac{1}{2}+\frac{T}{N}\frac{1}{\overline{Z}}\frac{\overline{Z}}{\mu }.$$
Differentation yields
$$\frac{\overline{Z}}{\mu }=\underset{\overline{x}}{}\frac{Z_1(|\mathrm{\Phi }_{\overline{x}}|^2)}{\mu }\left[\underset{x\overline{x}}{}Z_1(|\mathrm{\Phi }_x|^2)\right]\underset{x}{}d\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}.$$
Since
$$\frac{Z_1(|\mathrm{\Phi }_{\overline{x}}|^2)}{\mu }=\phi e^{T\phi ^2}\frac{sinh(\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2})}{\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2}}𝑑\phi $$
we can write for the previous expression
$$\frac{\overline{Z}}{\mu }=\underset{\overline{x}}{}\phi _{\overline{x}}e^{S_b}\underset{x}{}e^{T\phi ^2}\frac{sinh(\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2})}{\sqrt{(\phi +\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2}}d\phi _xd\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}.$$
This implies
$$\frac{1}{\overline{Z}}\frac{\overline{Z}}{\mu }=\frac{1}{\overline{Z}}\underset{\overline{x}}{}\phi _{\overline{x}}e^{S_b}e^{T\phi _{\overline{x}}^2}\frac{sinh(\sqrt{(\phi _{\overline{x}}+\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2})}{\sqrt{(\phi _{\overline{x}}+\mu /2T)^2+|\mathrm{\Phi }_{\overline{x}}|^2}}𝑑\phi _{\overline{x}}\left[\underset{x\overline{x}}{}Z_1(|\mathrm{\Phi }_x|^2)\right]\underset{x}{}d\mathrm{\Phi }_xd\mathrm{\Phi }_x^{}$$
$$=\underset{\overline{x}}{}\phi _{\overline{x}}.$$ |
warning/0001/hep-ph0001236.html | ar5iv | text | # R-parity Violation and Neutrino Masses
## Abstract
R-parity violation in the supersymmetric standard model could be the origin of neutrino masses and mixing accounting for the atmospheric and solar neutrino oscillations. More interestingly, this hypothesis may be tested in future colliders by detecting lepton number violating decays of the lightest supersymmetric particle. Here, we present a comprehensive analysis for the determination of sneutrino vacuum expectation values from the one-loop effective scalar potential, and also for the one-loop renormalization of neutrino masses and mixing. Applying our results to theories with gauge mediated supersymmetry breaking, we discuss the effects of the one-loop corrections and how the realistic neutrino mass matrices arise.
The minimal supersymmetric standard model (MSSM) may allow for explicit lepton number and thus R-parity violation through which neutrinos get nonzero masses . As it is an attractive possibility to explain the neutrino mass matrix consistent with the current data coming from, in particular, the atmospheric and solar neutrino experiments, many works have been devoted to investigating the properties of neutrino masses and mixing arising from R-parity violation \[see references in \]. The lepton number and R-parity violating terms in the MSSM superpotential are
$$W=\mu _iL_iH_2+\lambda _{ijk}L_iL_jE_k^c+\lambda _{ijk}^{}L_iQ_jD_k^c.$$
(1)
We first recall that there are two contributions to neutrino masses from R-parity violation. One is the loop mass arising from one-loop diagrams with the exchange of squarks, sleptons or gauginos. The other is the tree mass arising from the misalignment of the bilinear couplings in the superpotential and sneutrino vacuum expectation values (VEVs) determined by the minimization of the scalar potential,
$$V=[(m_{L_iH_1}^2+\mu \mu _i)L_iH_1^{}+B_iL_iH_2+\mathrm{h}.\mathrm{c}.]+m_{L_i}^2|L_i|^2+V_D+\mathrm{}.$$
(2)
Here $`m_{L_iH_1}^2`$, $`B_i`$ and $`m_{L_i}^2`$ are the soft terms, $`\mu `$ is the supersymmetric Higgs mass parameter, and $`V_D`$ denotes the SU(2)xU(1) D-terms.
The first attempt to investigate whether R-parity violation can provide the solution to the atmospheric and solar neutrino problems has been made in , where it was found that the minimal supergravity model with bilinear R-parity violating terms naturally yields the desired neutrino masses and mixing angles. According to the scatter plot study of minimal supergravity parameter space , the matter conversion (vacuum oscillation) solution to the solar neutrino problem was found to be realized in a few % (20 %) of the selected parameter space. After the observation of muon neutrino oscillation in Super-Kamiokande , the similar attempt has been made in the context of minimal supergravity models with generic (trilinear) R-parity violating couplings to find out the preferred ranges of the soft parameters depending on $`\mathrm{tan}\beta `$, and to obtain the correlation between the neutrino properties (the atmospheric mixing angle and the ratio between two heaviest masses) and the soft parameters.
A remarkable feature of R-parity violation as the origin of neutrino masses and mixing is that this idea can be tested in future collider experiments despite the small R-parity violating couplings. For instance, the large mixing angle between the muon and tau neutrino implies the observation of comparable numbers of muons and taus produced in the decays of the lightest supersymmetric particle (LSP) . Moreover, the factorization of the R-parity even and odd quantities in the neutrino-neutralino mixing matrix enables us to probe the $`\nu _\mu `$$`\nu _\tau `$ and $`\nu _\mu `$$`\nu _e`$ oscillation amplitudes measured in Super-Kamiokande and the CHOOZ experiment directly in colliders through the measurement of the electron, muon and tau branching ratios of the neutralino LSP . In a detailed analysis, it was found that this testability holds in most of $`\mathrm{tan}\beta `$ and neutralino mass parameter space, in particular, for the case of the bilinear R-parity violating models .
More recently, further development has been made to include the one-loop effect in the determination of the sneutrino VEVs . For this, one adds the one-loop correction,
$$V_1=\frac{1}{64\pi ^2}\mathrm{Str}^4\left(\mathrm{ln}\frac{^2}{Q^2}\frac{3}{2}\right),$$
(3)
to the scalar potential (2). Then, it is straightforward from the one-loop effective scalar potential to calculate the sneutrino VEVs,
$$\frac{L_i^0}{H_1^0}=\frac{B_i\mathrm{tan}\beta +(m_{L_iH_1}^2+\mu \mu _i)+\mathrm{\Sigma }_{L_i}^{(1)}}{m_{L_i}^2+\frac{1}{2}M_Z^2c_{2\beta }+\mathrm{\Sigma }_{L_i}^{(2)}},$$
(4)
where the one-loop correction terms $`\mathrm{\Sigma }_{L_i}^{(1,2)}`$ are given by
$$\mathrm{\Sigma }_{L_i}^{(1)}=\frac{V_1}{H_1^0L_i^0}|_{L_i^0=0},\mathrm{\Sigma }_{L_i}^{(2)}=\frac{V_1}{L_1^0L_i^0}|_{L_i^0=0},$$
(5)
under the condition that the R-parity violating parameters are small, $`\mu _i/\mu ,\lambda ,\lambda ^{}1`$, which is the case with small neutrino masses, $`m_\nu M_Z`$. The essential step in determining $`\mathrm{\Sigma }^{(1,2)}`$’s is the diagonalization of the mass matrices of neutralinos/neutrinos, charginos/charged leptons and Higgses/sleptons which get mixed due to R-parity violation. This procedure can be done analytically when the R-parity violating parameters are small. The complete analytic formulae for the $`\mathrm{\Sigma }`$’s are then obtained in including the contributions of all the particles in the MSSM. Having determined the sneutrino VEVs, one obtains the tree mass given by
$$M_{ij}^\nu =\frac{M_Z^2}{F_N}\xi _i\xi _jc_\beta ^2\text{with}F_N=\frac{M_1M_2}{M_1c_W^2+M_2s_W^2}\frac{M_Z^2}{\mu }s_{2\beta }$$
(6)
where $`\xi _iL_i^0/H_1^0\mu _i/\mu `$, and $`M_{1,2}`$ are the masses of the neutral gauginos, bino and wino, respectively. Recall that the matrix (6) makes only one neutrino massive.
To obtain the complete neutrino mass eigenvalues and mixing angles, one needs to perform one-loop renormalization of the neutrino/neutralino mass matrix through the general formula,
$$M^{pole}(p^2)=M(Q)+\mathrm{\Pi }(p^2)\frac{1}{2}\left(M(Q)\mathrm{\Sigma }(p^2)+\mathrm{\Sigma }(p^2)M(Q)\right)$$
(7)
where $`Q`$ is the renormalization scale, and $`M(Q)`$ is the the $`\overline{DR}`$ renormalized tree-level mass matrix, $`\mathrm{\Pi }`$ and $`\mathrm{\Sigma }`$ are the contributions from one-loop self-energy diagrams. In our case, $`M`$ is the 7x7 neutrino/neutralino mass matrix consisting of the 3x4 neutrino-neutralino mixing mass matrix $`M_D`$ and the 4x4 neutralino mass matrix $`M_N`$. To find out the 3x3 neutrino mass matrix, usually used is the on-shell renormalization scheme in which one works in the tree-level mass basis and rediagonalizes the one-loop corrected 7x7 mass matrix (7). But, as far as only the neutrino sector is concerned, it is more useful to work in the weak basis and obtain the effective neutrino mass matrix by one-step diagonalization . Following this procedure, one finds the neutrino mass matrix;
$`M^\nu (p^2)`$ $`=`$ $`M_DM_N^1M_D^T(Q)+\mathrm{\Pi }_n(p^2)`$ (8)
$`+`$ $`M_DM_N^1\mathrm{\Pi }_D^T(p^2)+\mathrm{\Pi }_D(p^2)M_N^1M_D^T`$
neglecting the subleading contributions. Here, $`\mathrm{\Pi }_{n,D}`$ are the neutrino-neutrino and neutrino-neutralino one-loop masses. Note that the first term on the right-hand side of Eq. (8) is the tree mass given in Eq. (6). As can be seen from (8), there is no need to calculate the other one-loop terms such as $`\mathrm{\Pi }_N`$ for neutralino self-energies and $`\mathrm{\Sigma }`$’s. Furthermore, we can simply take $`p^2=m_\nu ^2=0`$ to calculate the physical neutrino masses and mixing. This should be contrast with the on-shell renormalization scheme where the neutralino masses are also involved in calculating the one-loop renormalized neutrino masses.
Applying the above results to the theories with gauge mediated supersymmetry breaking, one can find various interesting properties . First of all, the one-loop corrections $`\mathrm{\Sigma }_{L_i}^{(1,2)}`$ in Eq. (4) can lead to a drastic change in the sneutrino VEVs. This effect is magnified in the parameter ranges where a partial cancellation between the two quantities, $`(m_{L_iH_1}^2+\mu \mu _i)`$ and $`B_i\mathrm{tan}\beta `$, occurs. In certain cases, it can even change the order of magnitudes of the neutrino mass eigenvalues. The most interesting question is again what kind of realistic neutrino mass matrices accounting for the atmospheric and solar neutrino oscillations can be obtained. In the bilinear models, it turns out that only realizable is the small mixing MSW solution to the solar neutrino problem with $`\mathrm{tan}\beta 20`$. In the trilinear models, one has more freedom and thus more solutions. For instance, the small mixing MSW solution can be obtained even for small $`\mathrm{tan}\beta `$. An interesting aspect is that the bi-maximal mixing solution to the atmospheric and solar neutrino problems can also be realized with small $`\mathrm{tan}\beta `$.
Finally, we would like to emphasize that the desirable neutrino mass matrices with R-parity violation arise without fine-tuning of parameters given the overall smallness of R-parity violating parameters, $`\mu _i/\mu ,\lambda `$ and $`\lambda ^{}`$. This smallness may be a consequence of a certain flavor symmetry responsible for the quark and lepton mass hierarchies, such as horizontal U(1) symmetry . |
warning/0001/astro-ph0001445.html | ar5iv | text | # The FIR-Radio Correlation of Wolf-Rayet Galaxies and the Role of Star Formation in LINERs
## 1 Introduction
Low ionization nuclear emission regions (LINERs, Heckamn 1980) are galaxies with strong forbidden lines from low ionization states, compared with those from higher ionization states, typically \[O ii\] 3727 $`>`$ \[O iii\] 5007. LINER phenomena are the most common activities of galaxies known in the local universe (Ho, Felippenko, & Sargent 1997). If most LINERs are truely low luminosity active galactic nuclei (LLAGN) (Ho 1998), it will be important to understand the nature of AGN by investigating the role of LINERs. However, the origin of LINER phenomena is debated (Maoz et al 1998; Lawrence 1998; Ho 1998; and the references therein). Maoz et al (1998) claimed that the young stellar population may provide enough ionizing photons for the observed spectra of a significant fraction of LINERs, and it is also possible that LINERs may be a heterogeneous class. A good example is M81, one of the nearest bright LINERs. Ho (1998) stressed the incontrovertible, nonstellar nature of M81, though the stellar contribution cannot be ruled out (Maoz et al. 1998). Recently, Hameed & Devereux (1999) discussed the extended nuclear emission-line regions having a LINER spectrum in M81, and argued that M81 is likely a composite object. They also proposed that shock heating or UV-photons from post asymptotic giant branch stars are probably responsible for the extended LINER emission. Further evidence for the “composite” nature for M81 is provided by the ROSAT HRI data (Colbert & Mushotzky 1999), in which the intensity ratio of the point-like component to the extended one is about 3/2 (Colbert, private communication).
It has been suggested (Condon et al. 1982) that the AGN activities can be distinguished from starbursts by using the FIR-radio correlation, which is more significant for starbursts than for AGN. The weak FIR-radio correlation of Seyferts may imply the starburst (SB)-dominated bolometric luminosity for Seyferts (Forbes & Norris 1998; also see, Norris et al. 1988). The hypothesis of a close connection between starbursts and Seyferts has been proposed by Terlevich and his collaborators (Terlevich & Melnick 1985; Terlevich et al 1992). They claim that the radio-quiet AGN are powered by massive nuclear starburst in a metal-rich environment. Recently, Heckman et al. (1997) discovered a powerful nuclear starburst (in the Wolf-Rayet (WR) phase) in Seyfert galaxy Mrk 477. The bolometric luminosity of Mrk 477 in the central region might be dominated by the nuclear starburst. It was remarked by Maiolino et al. (1998) that Heckman et al’s study would provide observational evidence partly supporting Terlevich’s hypothesis.
Following these leads, we have explored the energetics of LINERs by using the FIR-radio correlation for Wolf-Rayet (WR) galaxies, suggesting different energy budgets of LINERs.
## 2 FIR-radio correlation of WR galaxies
WR galaxies are extragalactic sources that exhibit broad emission lines characteristic of WR stars in their spectra (Conti 1991). Their typical burst ages are $`2`$-$`8`$ Myr. The 50 WR galaxies that show a good FIR-radio correlation, as shown in Fig 1a, have detected flux at 1.4 GHz in NVSS Catalog (Condon et al. 1998) and 60$`\mu m`$ in IRAS (Moshir et al. 1992) among 139 known sources (Schaerer et al. 1999). Such a correlation of WR galaxies is apparently nonlinear, with a regression coefficient of about 1.20, as obtained before for other samples of galaxies (e.g. Fitt et al. 1988; Cox et al. 1988).
This correlation can be understood in the framework of starburst phenomenon (Moorwood 1996; Lisenfeld et al. 1996). For simplicity, we take the stars/dust geometry to be close to a star-free shell of dust surrounding a central dust-free sphere of stars (Mas-Hesse & Kunth 1999). In this scenario, the radiation from the nuclear starburst (the optical, ionizing and non-ionizing photons) heats the dust grains, and the UV photons emitted from the nuclear massive stars photoionize the gas.
The radio flux at 1.4 GHz consists of thermal bremsstrahlung emission from photoionized gas and synchrotron radiation from supernova remnants. The luminosities of the two radiation mechanisms are respectively given by Rubin (1968)
$$L_\mathrm{T}^{\mathrm{f}\mathrm{f}}(\nu )=1.59\times 10^{32}\left(\frac{\nu }{\mathrm{GHz}}\right)^{0.1}T_4^{0.45}N_{\mathrm{UV}}\mathrm{W}\mathrm{Hz}^1$$
(1)
where $`T_4`$ is electron temperature in units of 10<sup>4</sup>K and $`N_{\mathrm{UV}}`$ ionizing photons per second, and by Colina & Pérez-Olea (1992)
$$L_{\mathrm{NT}}^{\mathrm{SNR}}(\nu )=4.45\times 10^{22}\left(\frac{\nu }{\mathrm{GHz}}\right)^{0.7}T_{\mathrm{SNII}}\mathrm{W}\mathrm{Hz}^1$$
(2)
where $`T_{\mathrm{SNII}}`$ is the Type ii supernova rate.
The FIR radiation at 60$`\mu `$m is assumed to be composed of two parts: the warm dust component caused by the same starburst event, and the cool dust component outside the starburst region heated by the general interstellar radiation. Xu et al. (1994) modelled the contribution of cool and warm components in FIR - radio correlation for late-type galaxies Several authors have virtually tried correcting or linearizing the FIR-radio correlation (Condon 1992; also see Fitt et al. 1988; Devereux & Eales 1989). The luminosity of the FIR radiation is described by
$`L_{\mathrm{IR}}(\nu )`$ $`=`$ $`4\pi B_\nu (T_{\mathrm{d},\mathrm{w}})Q_{\mathrm{abs}}(\nu )\pi a^2N_{\mathrm{d},\mathrm{w}}`$ (3)
$`+4\pi B_\nu (T_{\mathrm{d},\mathrm{c}})Q_{\mathrm{abs}}(\nu )\pi a^2N_{\mathrm{d},\mathrm{c}}`$
where $`T_{\mathrm{d},\mathrm{w}}`$ and $`T_{\mathrm{d},\mathrm{c}}`$ are the warm and cool dust temperatures, $`N_{\mathrm{d},\mathrm{w}}`$ and $`N_{\mathrm{d},\mathrm{c}}`$ the total number of warm and cool dust grains, $`a`$ is the average radius of dust grains, $`Q_{\mathrm{abs}}(\nu )`$ the absorption efficiency of dust grains, and $`B_\nu (T_\mathrm{d})`$ the Planck function. We adopt the “astronomical silicate” dust model (Draine & Lee 1984), which is most likely suitable to starburst galaxies (Mas Hesse & Kunth 1999).
Assuming a “steady-state” case for the dust grains, the dust temperatures can be derived from the equilibrium between dust absorption and dust emission,
$$c_0^{\mathrm{}}U_\lambda Q_{\mathrm{abs}}(\lambda )𝑑\lambda =4\pi _0^{\mathrm{}}B_\lambda (T_\mathrm{d})Q_{\mathrm{abs}}(\lambda )𝑑\lambda $$
(4)
where $`U_\lambda `$ is the energy density of a diluted radiation field that heats the dust, which is satisfied with
$$U_\lambda =\frac{4\pi }{c}B_\lambda (T_{\mathrm{eff}})W$$
(5)
where $`W`$ is the dilution factor, and $`T_{\mathrm{eff}}`$ the equivalent effective temperature for the radiation field generated by starburst activities. Using the $`\lambda ^1`$ dependence for $`Q_{\mathrm{abs}}(\lambda )`$, one can yield the dust temperature from equa. (4): $`T_\mathrm{d}T_{\mathrm{eff}}W^{\frac{1}{5}}`$ (Spitzer 1978). The FIR luminosity at 60$`\mu `$m can be obtained from equa. (3), scaling the value of $`Q_{\mathrm{abs}}(\lambda )`$ to fit the Draine & Lee (1984) model at $`60\mu m`$: \[$`\lambda Q_{\mathrm{abs}}(\lambda )/a]_{60\mu m}`$2.5 .
At any given burst age, the evolutionary synthesis model, GISSEL95 (Bruzual & Charlot 1996), is used to provide the relevant quantities such as $`N_{\mathrm{UV}}`$, $`T_{\mathrm{SNII}}`$, and the bolometric corrections for deriving the effective temperatures. Considering the discussion by Mas Hesse & Kunth (1999), we have assumed 50% of $`N_{\mathrm{UV}}`$ are absorbed by dust.
We have estimated the possible values of the dilution factor in various ways and adopt their average, 10<sup>-14</sup>, which is compatible with the usual interstellar value (Spitzer 1978). The radiation transfer is not taken into account. A dust-to-gas mass ratio $`1/100`$ is assumed. We also assume that the gas mass is comparable to the star mass in the starburst region ($`M_{\mathrm{SB}}`$) (namely the gas-to-star mass ratio is roughly unity). The total grain number is interpreted as $`M_{\mathrm{SB}}/\rho _da^3`$ where the density of the ’astronomical silicate’ is adopted as $`\rho _d3\mathrm{g}\mathrm{cm}^3`$ (Draine & Lee 1984).
For calculating the cool dust temperature, we assume that the cool dust component may be heated by the general interstellar radiation field that arises from a past starburst event with a typical burst age $``$ 1 Gyr. The mass of the cool component is a free parameter, and we try fixing its value, 10$`{}_{}{}^{6}M_{}^{}`$ or $`5\times 10^4M_{}`$, for any $`M_{\mathrm{SB}}`$. It means that the contribution from the cool component is relatively significant for small burst strength (small $`M_{\mathrm{SB}}`$), and relatively unimportant for ultraluminous infrared galaxies (ULIGs; large $`M_{\mathrm{SB}}`$). Generally, we have $`T_{\mathrm{d},\mathrm{c}}20`$ K, similar to the assuming cool dust temperature by Fitt et al. (1988).
To perform the calculations, we take $`M_{\mathrm{SB}}`$ as an independent variable, which is in the range of $`10^{4.5}`$-$`10^{10}M_{}`$. The upper end of $`M_{\mathrm{SB}}`$ corresponds to the case of ULIGs (e.g. Genzel et al. 1998). The stronger the starburst (i.e., the larger $`M_{\mathrm{SB}}`$), the higher the FIR and radio luminosities. With various adopted parameters (burst ages, dust-to-gas ratio, etc.), we obtain linear FIR-radio correlations if taking only the warm dust component into consideration, or a nonlinear correlations if both the warm and cool dust components. The model curves are plotted in Fig 1a, the model parameters are listed in Table 1.
The solid line I in Fig 1a represents the linear part of our model prediction at the burst age of 6 Myr, in which the contribution of cool dust emission is neglected and thermal (bremsstrahlung) emission is dominant at 1.4 GHz. The dashed lines in Fig 1a illustrate the lower-right envelopes for models, in which the contribution of cool dust emission is taken into account. For lines IIa and IIb, the cool dust mass is taken as $`5\times 10^6M_{}`$ and $`10^6M_{}`$ at the age of 3 Myr, respectively, and for line IIc, the cool dust mass is $`5\times 10^4M_{}`$ at the age of 6 Myr. As expected, counting the cool dust emission reproduces the nonlinear trend in the correlation lines. It is quite reasonable to see in Fig 1a that the cool dust component makes significant contribution in the case of small burst strength, while it is negligible compared with warm component for large burst strength. Satisfactorily, the majority of WR galaxies are located in a “passage” escorted by the upper and lower envelopes in the diagram. Reasonably, this passage may be considered to be typical of the positions of the SB-dominated galaxies.
In Fig 1b, we have added two prototypical starburst galaxies, M 82 and NGC 253, corresponding to a burst age of 10<sup>7</sup>-10<sup>8</sup> yr. The dotted line Ia in Fig 1b indicates our model prediction (with a dust-to-gas ratio of 1/100) at the burst age of 2$`\times `$10<sup>7</sup> yr, which represents the upper end of age for the supernova, set by GISSEL95. The non-thermal (synchrotron) radiation dominates at 1.4 GHz in this case. Considering the enrichment of the dust grains by supernova explosions at this age, the dust-to-gas ratio can increase by several times, up to an order of magnitude (Hirashita 1999), so it would be reasonable to replace the dust-to-gas ratio of 1/100 with 1/20. As a result, the model curve will shift to a position indicated by the dashed line Ib in Fig 1b.
In order to fit the galaxies that exhibit ongoing star formation, such as a transition object NGC 5194 (Heckman 1980; Larkin et al. 1998), we tried to add a shock wave (that may be related with the supernova explosions and/or outflowing winds from starburst) as additional mechanism for heating the dust, following Dwek (1986) and Contini et al. (1998). The model curve containing a shock-heating phase is represented by the dot-dashed line Is in Fig 1b, which is below the lower border of the passage mentioned above. Here, we have adopted a shock velocity $`v_\mathrm{s}=200\mathrm{k}\mathrm{m}\mathrm{s}^1`$, a shock covering fraction 1/10, and a dust-to-gas ratio 1/20 at the age of $`2\times 10^7`$ yr. It is worth noting that a strong near-infrared \[Fe ii\] line has been observed in NGC 5194, and the shock excitation in supernova remnants is probably the mechanism responsible for this line (Larkin et al 1998). This excitation mechanism may be consistent with our consideration of shock-heating of dust in this galaxy, with $`v_\mathrm{s}`$ in order $`100\mathrm{km}\mathrm{s}^1`$.
## 3 Energetics of LINERs
### 3.1 Classification of LINERs
Recently, more than a dozen of LINERs have been studied with space facilities or large ground-based telescopes. Table 2 lists these sources with claims for their energetics, except for M81 because of the debate mentioned in Sect. 1. Fig 2a indicates the positions where these LINERs are located in the FIR-radio diagram of WR galaxies. It is very instructive to see that the LINERs with SB-supported claims or with the existence of nuclear starburst, in notation of SB in Table 2, have a similar distribution to WR galaxies, while the AGN-supported LINERs or those with the existence of AGN are distinctively located in the upper-left part of the FIR-radio diagram. The latter category of objects is found above the upper border (solid line in Fig 2a) of our models for starburst events.
Furthermore, we investigate two cases of LINERs that are considered to have a composite nature, NGC6240 and M81. For NGC6240, Schulz et al. (1999) claimed, based on ROSAT data, that both AGN and starburst contribute in roughly equal proportion to the energetics of this galaxy. In our Fig. 2a, both NGC6240 and M81 are approximately located on the border of our model prediction. Due to the enrichment of dust and the influence of cool dust component as discussed in Sect. 2, the actual position of our model prediction in Fig 2a might somewhat move to the right and curve up at small burst strength.
Of particular importance to the AGN nature in LINERs is the detection of broad (FWHM $``$ a few thousand km $`\mathrm{s}^1`$) permitted lines in these sources, which may arise from the broad-line regions (BLR). On the analogy of the nomenclature for Seyferts, Ho (1998) has designated those sources having visible BLR as LINER 1, and others as LINER 2.
It is enlightening to see that the most part of LINER 2’s in Terashima (1999) have SB notations in our Table 2, while the majority of LINERs with AGN notations in our Table 2 are listed as LINER 1’s in Terashima (1999). Now we have seen a fact that these LINER 1’s are basically segregated from LINER 2’s in the FIR-radio diagram. This segregation confirms the early suggestion by Condon et al (1982) of distinguishing AGN from starburst by use of the FIR-radio correlation.
The majority of sources studied in this paper are at distances of 10 Mpc or beyond, indicating that the FIR-radio diagrams shown in our figures are basically referred to the global properties of galaxies, due to the large resolution/apertures used in the NVSS and the IRAS. Nevertheless, what we have seen in the segregation is the pairing of the observed SB-supported LINERs to modeled SB-dominated passage, not in a cross-pair of the AGN- to SB-dominated passage. It would be hard to understand that the segregation could be just caused by chance.
The studies on LINERs in the IRAS 1-Jy (f(60$`\mu \mathrm{m})>`$ 1 Jy) sample of ultraluminous infrared galaxies (LINER ULIGs) by Veilleux et al. (1999) may shed light on the above fact. They conclude that “there is no convincing optical or infrared evidence for an AGN in LINER ULIGs”, and “the main source of energy in these LINERs is a starburst rather than an AGN.” We have put these LINER ULIGs in the FIR-radio diagram in Fig 2a by symbols of crosses. Their positions are in the SB-dominated passage. Recent studies (see, Veilleux et al 1999 and the references therein) strongly suggest that the overwhelming part of the bolometric luminosity of ULIGs stems from the inner kpc regions, indicating that the large resolution/apertures used in the NVSS and the IRAS would not make any obvious change in the FIR-radio correlation for LINER ULIGs. In fact, these studies are consistent with the early work by Kennicutt & Kent (1983), which demonstrated that in the case of EW(H$`\alpha `$) $``$ 10Å, suitable to LINER ULIGs, the H$`\alpha `$ emission observed with a slit is comparable to that obtained using large apertures.
Similar analyses of early-type galaxies (Walsh et al 1989) suggest no obvious variance in the FIR-radio correlation with the sizes of galaxies. Radio observations of a sample of ellipticals and S0s (Fabbiano et al 1987) have not found evidence for the extended disk emission, confirming the suggestion that the radiation is from nuclear “starburst” instead of extended disk sources (Dressel 1988). These investigations provide a sound explanation for the statistical study by Walsh et al. (1989), implying that the location of the AGN-supported LINERs (the majority of which are E/S0 sources) in the upper-left part of FIR-radio diagram (Fig 2a) may not be significantly influenced by the size of apertures used.
On the other hand, NGC 6500, a spiral far above the border of the starburst events in Fig 2a, has EW(H$`\alpha `$) = 27Å (Ho et al 1997), indicating little effect of the aperture sizes on the FIR-radio correlation, according to Kennicutt & Kent (1983). The same argument of EW(H$`\alpha `$) $``$ 10Å holds for NGC 404, one nearby galaxy, and four other LINERs selected from the Pico dos Dias Survey (PDS) (Coziol et al. 1998). Three of the PDS LINERs are classified as transition sources by Coziol et al. (1998), including NGC 3310, designated as starburst in Véron Catalog, that is classified as a transition source, SB/LINER.
The remaining sources, NGC 4736, NGC 5055, and NGC 7217, are three LINERs with EW(H$`\alpha `$) $`<`$ 10Å. For example, the EW(H$`\alpha `$) of NGC 7217 obtained with a slit and a large aperture are about 3Å and 6Å (Ho et al 1997; Kennicutt & Kent 1983), respectively. After correcting the aperture effect over this source, the position of NGC 7217 in Fig 2a may move a bit down- and leftward in the FIR-radio diagram, remaining in the SB-dominated passage. The same argument would be applicable to other sources with EW(H$`\alpha `$) $`<`$ 10Å.
From the above discussion, one can see that the aperture effects may not significantly change the situation of segregation for different types of LINERs shown in Fig 2a. Therefore, as suggested by Condon et al (1982), the FIR-radio correlation may provide a preliminary classification of LINERs according to their locations in the diagram as we described above. In other words, one may classify LINERs in terms of their FIR-to-radio ratio: $`QL(1.4\mathrm{GHz})/L(60\mu \mathrm{m})`$: one has $`Q>0.01`$ for the AGN-supported LINERs and $`Q<0.01`$ for the SB-supported ones.
In Table 3 we list part of LINERs extracted from Véron Catalog (Véron-Cetty & Véron 1996) that have detected fluxes at 1.4 GHz and 60$`\mu `$m, and the preliminary classification of their energetics is given in column 6. Their distributions in the FIR-radio diagram are shown in Fig 2b.
As further evidence our classification the types of LINERs, we mention the new results of Alonso-Herrero et al. (1999). The different types of LINERs identified by these authors are consistent with our predictions for those common ones, as listed in our Table 3. For example, the claimed SB-dominated LINERs in their paper, NGC 3504, NGC 3367, NGC 4569, NGC 4826, and NGC 7743 are all located in our SB-dominated passage, and the AGN-dominated LINER NGC 2639 (Alonso-Herrero et al 1999) is designated to be an AGN-supported LINER in Table 3. Further observations of these sources are certainly needed, especially for the LINERs located near the boundary of the starburst events shown with solid line in Fig 2, which might be composite like NGC 6240.
### 3.2 LINERs with inner rings
The effects of rings or bars on Seyfert activities have been studied since the early work in the 1980’s (e.g. Simkin, Su, & Schwarz 1980; Arsenault 1989). Recent studies show, however, that the frequency of barred systems is the same in Seyferts and in normal spirals (McLeod & Rieke 1995; Ho, Filippenko, & Sargent 1997; Mulchaey & Regan 1997). A latest study on the morphology of the 12$`\mu `$m Seyfert Sample (Hunt et al 1999) indicates that LINERs have higher rates of inner rings than normal galaxies.
One striking feature in Fig 2a and 2b is that almost all the LINERs having inner rings are located in the SB-dominated passage. This strong morphological tendency in the LINER sample should have important implication in starburst-AGN connection. Our preliminary analysis shows that the AGN activities of these LINERs are lower than the AGN-supported LINERs located in the upper-left part of the FIR-radio diagram. The reason for the reduced activities might be caused by the reduced fueling gas to the central black holes. We will discuss this topic in a separate paper (Lei et al. 1999), along with the morphological study of Seyferts.
###### Acknowledgements.
We wish to thank Dan Maoz for valuable comments that helped improve our manuscript. The anonymous referee is thanked for his/her critical comments and kind help to improve our English presentation. We are very grateful to Wei Zheng for his careful reading of the manuscript. This work is supported by grants from the NSF of China, and grants from the Ascent Project of the State Scientific Commission of China. |
warning/0001/math0001043.html | ar5iv | text | # Braid group actions on derived categories of coherent sheaves
## 1. Introduction
### 1a. Derived categories of coherent sheaves
Let $`X`$ be a smooth complex projective variety and $`D^b(X)`$ the bounded derived category of coherent sheaves. It is an interesting question how much information about $`X`$ is contained in $`D^b(X)`$.
Certain invariants of $`X`$ can be shown to depend only on $`D^b(X)`$. This is obviously true for $`K(X)`$, the Grothendieck group of both the abelian category $`Coh(X)`$ of coherent sheaves and of $`D^b(X)`$. A deep result of Orlov implies that the topological $`K`$-theory $`K_{\mathrm{top}}^{}(X)`$ is also an invariant of $`D^b(X)`$; hence, so are the sums of its even and odd Betti numbers. Because of the uniqueness of Serre functors , the dimension of $`X`$ and whether it is Calabi-Yau ($`\omega _X𝒪_X`$) or not, can be read off from $`D^b(X)`$. Using Orlov’s theorem quoted above, one can prove that the Hochschild cohomology of $`X`$, $`HH^{}(X)=\mathrm{Ext}_{X\times X}^{}(𝒪_\mathrm{\Delta },𝒪_\mathrm{\Delta })`$, depends only on $`D^b(X)`$. As pointed out by Kontsevich \[29, p. 131\], it is implicit in a paper of Gerstenhaber and Schack that
$$HH^r(X)\underset{p+q=r}{}H^p(X,\mathrm{\Lambda }^qTX).$$
Thus for Calabi-Yau varieties $`dimHH^r(X)=_{p+q=r}h^{p,nq}(X)`$; in mirror symmetry these are the Betti numbers of the mirror manifold. Finally, a theorem of Bondal and Orlov says that if the canonical sheaf $`\omega _X`$ or its inverse is ample, $`X`$ can be entirely reconstructed from $`D^b(X)`$. Contrary to what this list of results might suggest, there are in fact non-isomorphic varieties with equivalent derived categories. The first examples are due to Mukai: abelian varieties and $`K3`$ surfaces . Examples with nontrivial $`\omega _X`$ have been found by Bondal and Orlov .
This paper is concerned with a closely related object, the self-equivalence group $`\mathrm{Auteq}(D^b(X))`$. Recall that an exact functor between two triangulated categories $`,𝔇`$ is a pair $`(F,\nu _F)`$ consisting of a functor $`F:𝔇`$ and a natural isomorphism $`\nu _F:F[1]_{}[1]_𝔇F`$ (here $`[1]_{},[1]_𝔇`$ are the translation functors) with the property that exact triangles in $``$ are mapped to exact triangles in $`𝔇`$. The appropriate equivalence relation between such functors is ‘graded natural isomorphism’ which means natural isomorphism compatible with the maps $`\nu _F`$ \[4, section 1\]. Ignoring set-theoretic difficulties, which are irrelevant for $`=D^b(X)`$, the equivalence classes of exact functors from $``$ to itself form a monoid. $`\mathrm{Auteq}()`$ is defined as the group of invertible elements in this monoid. Known results about $`\mathrm{Auteq}(D^b(X))`$ parallel those for $`D^b(X)`$ itself. It always contains a subgroup $`A(X)(\mathrm{Aut}(X)\mathrm{Pic}(X))\times `$ generated by the automorphisms of $`X`$, the functors of tensoring with an invertible sheaf, and the translation. Bondal and Orlov have shown that if $`\omega _X`$ or $`\omega _X^1`$ is ample then $`\mathrm{Auteq}(D^b(X))=A(X)`$. Mukai’s arguments imply that $`\mathrm{Auteq}(D^b(X))`$ is bigger than $`A(X)`$ for all abelian varieties (recent work of Orlov describes $`\mathrm{Auteq}(D^b(X))`$ completely in this case).
Our own interest in self-equivalence groups comes from Kontsevich’s homological mirror conjecture . One consequence of this conjecture is that for Calabi-Yau varieties to which mirror symmetry applies, the group $`\mathrm{Auteq}(D^b(X))`$ should be related to the symplectic automorphisms of the mirror manifold. This conjectural relationship is rather abstract, and difficult to spell out in concrete examples. Nevertheless, as a first and rather naive check, one can look at some special symplectic automorphisms of the mirror and try to guess the corresponding self-equivalences of $`D^b(X)`$. Having made this guess in a sufficiently plausible way (which means that the two objects show similar behaviour), the next step might be to take some unsolved questions about symplectic automorphisms and translate it into one about $`\mathrm{Auteq}(D^b(X))`$. Using the smoother machinery of sheaf theory one stands a good chance of solving this analogue, and this in turn provides a conjectural answer, or ‘mirror symmetry prediction’, for the original problem. The present paper is an experiment in this mode of thinking. We now state the main results independently of their motivation; the discussion of mirror symmetry will be taken up again in the next section.
Let $`X,Y`$ be two (as before, smooth complex projective) varieties. The Fourier-Mukai transform (FMT) by an object $`𝒫D^b(X\times Y)`$ is the exact functor
$$\mathrm{\Phi }_𝒫:D^b(X)D^b(Y),\mathrm{\Phi }_𝒫(𝒢)=𝐑\pi _2(\pi _1^{}𝒢\stackrel{𝐋}{}𝒫),$$
where $`\pi _1:X\times YX`$, $`\pi _2:X\times YY`$ are the projections. This is a very practical way of defining functors. Orlov has proved that any equivalence $`D^b(X)D^b(Y)`$ can be written as a FMT. Earlier work of Maciocia shows that if $`\mathrm{\Phi }_𝒫`$ is an equivalence, then $`𝒫`$ must satisfy a partial Calabi-Yau condition: $`𝒫\pi _1^{}\omega _X\pi _2^{}\omega _Y^1𝒫`$. Bridgeland provides a partial converse to this.
Now take an object $`D^b(X)`$ which is a complex of locally free sheaves. We define the twist functor $`T_{}:D^b(X)D^b(X)`$ as the FMT with
(1.1)
$$𝒫=\mathrm{Cone}(\eta :^{}𝒪_\mathrm{\Delta }),$$
where $`^{}`$ is the dual complex, $``$ the exterior tensor product, $`\mathrm{\Delta }X\times X`$ is the diagonal, and $`\eta `$ the canonical pairing. Since quasi-isomorphic $``$ give rise to isomorphic functors $`T_{}`$, one can use locally free resolutions to extend the definition to arbitrary objects of $`D^b(X)`$.
###### Definition 1.1.
* $`D^b(X)`$ is called spherical if $`\mathrm{Hom}_{D^b(X)}^r(,)`$ is equal to $``$ for $`r=0,dimX`$ and zero in all other degrees, and if in addition $`\omega _X`$.
* An $`(A_m)`$-configuration, $`m1`$, in $`D^b(X)`$ is a collection of $`m`$ spherical objects $`_1,\mathrm{},_m`$ such that
$$dim_{}\mathrm{Hom}_{D^b(X)}^{}(_i,_j)=\{\begin{array}{cc}1\hfill & |ij|=1,\hfill \\ 0\hfill & |ij|2.\hfill \end{array}$$
Here, as elsewhere in the paper, $`\mathrm{Hom}^r(,)`$ stands for $`\mathrm{Hom}(,[r])`$, and $`\mathrm{Hom}^{}(,)`$ is the total space $`_r\mathrm{Hom}^r(,)`$.
###### Theorem 1.2.
The twist $`T_{}`$ along any spherical object $``$ is an exact self-equivalence of $`D^b(X)`$. Moreover, if $`_1,\mathrm{},_m`$ is an $`(A_m)`$-configuration, the twists $`T__i`$ satisfy the braid relations up to graded natural isomorphism:
$`T__iT_{_{i+1}}T__iT_{_{i+1}}T__iT_{_{i+1}}`$ $`\text{for }i=1,\mathrm{},m1,`$
$`T__iT__jT__jT__i`$ $`\text{for }|ij|2.`$
We should point out that the first part, the invertibility of $`T_{}`$, was also known to Kontsevich, Bridgeland and Maciocia. Let $`\rho `$ be the homomorphism from the braid group $`B_{m+1}`$ to $`\mathrm{Auteq}(D^b(X))`$ defined by sending the standard generators $`g_1,\mathrm{},g_mB_{m+1}`$ to $`T__1,\mathrm{},T__m`$. We call this a weak braid group action on $`D^b(X)`$ (there is a better notion of a group action on a category which requires the presence of certain additional natural transformations ; we have not checked whether these exist in our case). $`\rho `$ induces a representation $`\rho _{}`$ of $`B_{m+1}`$ on $`K(X)`$. Concretely, the twist along an arbitrary $`D^b(X)`$ acts on $`K(X)`$ by
(1.2)
$$(T_{})_{}(y)=y[],y[],$$
where $`[],[𝒢]=_i(1)^idim\mathrm{Hom}^i(,𝒢)`$ is the Mukai pairing or ‘Euler form’. If $`dimX`$ is even then $`\rho _{}`$ factors through the symmetric group $`S_{m+1}`$. The odd-dimensional case is slightly more complicated, but still $`\rho _{}`$ is far from being faithful, at least if $`m`$ is large.
For $`\rho `$ itself we have the following contrasting result:
###### Theorem 1.3.
Assume that $`dimX2`$. Then the homomorphism $`\rho `$ generated by the twists in any $`(A_m)`$-configuration is injective.
The assumption $`dimX1`$ cannot be removed; indeed, there is a $`B_4`$-action on the derived category of an elliptic curve which is not faithful (see section 3d).
### 1b. Homological mirror symmetry and self-equivalences
We begin by recalling Kontsevich’s homological mirror conjecture . On one hand, one takes Calabi-Yau varieties $`X`$ and their derived categories $`D^b(X)`$. On the other hand, using entirely different techniques, it is thought that one can attach to any compact symplectic manifold $`(M,\beta )`$ with zero first Chern class a triangulated category, the derived Fukaya category $`D^bFuk(M,\beta )`$ (despite the notation, this is not constructed as the derived category of an abelian category). Kontsevich’s conjecture is that whenever $`X`$ and $`(M,\beta )`$ form a mirror pair, there is a (non-canonical) exact equivalence
(1.3)
$$D^b(X)D^bFuk(M,\beta ).$$
A more prudent formulation would be to say that (1.3) should hold for the generally accepted constructions of mirror manifolds. Before discussing this conjecture further, we need to explain what $`D^bFuk(M,\beta )`$ looks like. This is necessarily a tentative description, since a rigorous definition does not exist yet. Moreover, for simplicity we have omitted some of the more technical aspects.
Let $`(M,\beta )`$ be as before, of real dimension $`2n`$. To simplify things we assume that $`\pi _1(M)`$ is trivial; this excludes the case of the two-torus, so that $`n2`$. Recall that a submanifold $`L^nM`$ is called Lagrangian if $`\beta |L\mathrm{\Omega }^2(L)`$ is zero. Following Kontsevich \[29, p. 134\] one considers objects, denoted by $`\stackrel{~}{L}`$, which are Lagrangian submanifolds with some extra structure. We will call such objects ‘graded Lagrangian submanifolds’ and the extra structure the ‘grading’. This grading amounts approximately to an integer choice. In fact there is a free $``$-action, denoted by $`\stackrel{~}{L}\stackrel{~}{L}[j]`$ for $`j`$, on the set of graded Lagrangian submanifolds; and if $`L`$ is a connected Lagrangian submanifold, all its possible gradings (assuming that there are any) form a single orbit of this action. For details we refer to . For any pair $`(\stackrel{~}{L}_1,\stackrel{~}{L}_2)`$ of graded Lagrangian submanifolds one expects to have a Floer cohomology group $`HF^{}(\stackrel{~}{L}_1,\stackrel{~}{L}_2)`$, which is a finite-dimensional graded $``$-vector space satisfying $`HF^{}(\stackrel{~}{L}_1,\stackrel{~}{L}_2[j])=HF^{}(\stackrel{~}{L}_1[j],\stackrel{~}{L}_2)=HF^{+j}(\stackrel{~}{L}_1,\stackrel{~}{L}_2)`$. Defining this is a difficult problem; a fairly general solution has been announced recently by Fukaya, Kontsevich, Oh, Ohta and Ono.
The most essential property of $`D^bFuk(M,\beta )`$ is that any graded Lagrangian submanifold $`\stackrel{~}{L}`$ defines an object in this category. The translation functor (which is part of the structure of $`D^bFuk(M,\beta )`$ as a triangulated category) acts on such objects by $`\stackrel{~}{L}\stackrel{~}{L}[1]`$. The morphisms between two objects of this kind are given by the degree zero Floer cohomology with complex coefficients:
$$\mathrm{Hom}_{D^bFuk(M,\beta )}(\stackrel{~}{L}_1,\stackrel{~}{L}_2)=HF^0(\stackrel{~}{L}_1,\stackrel{~}{L}_2)_{}$$
(Floer groups in other degrees can be recovered by changing $`\stackrel{~}{L}_2`$ to $`\stackrel{~}{L}_2[j]`$). Composition of such morphisms is given by certain products on Floer cohomology, which were first introduced by Donaldson. There is also a slight generalisation of this: any pair $`(\stackrel{~}{L},E)`$ consisting of a graded Lagrangian submanifold together with a flat unitary vector bundle $`E`$ on the underlying Lagrangian submanifold, defines an object of $`D^bFuk(M,\beta )`$. The morphisms between such objects are a twisted version of Floer cohomology. It is important to keep in mind that $`D^bFuk(M,\beta )`$ contains many objects other than those which we have described. This is necessarily so because it is triangulated: there must be enough objects to complete each morphism to an exact triangle, and these objects will not usually have a direct geometric meaning. However, it is expected that the objects of the form $`(\stackrel{~}{L},E)`$ generate the category $`D^bFuk(M,\beta )`$ in some sense.
###### Remark 1.4.
In the traditional picture of mirror symmetry, $`M`$ carries a $``$-valued closed two-form $`\beta _{}`$ with real part $`\beta `$. What we have said concerns the Fukaya category for $`\mathrm{im}(\beta _{})=0`$. Apparently, the natural generalisation to $`\mathrm{im}(\beta _{})0`$ would be to take objects $`(\stackrel{~}{L},E,A)`$ consisting of a graded Lagrangian submanifold $`\stackrel{~}{L}`$, a complex vector bundle $`E`$ on the underlying Lagrangian submanifold $`L`$, and a unitary connection $`A`$ on $`E`$ with curvature $`F_A=\beta _{}|L\mathrm{id}_E`$. The point is that to any map $`w:(D^2,D^2)(M,L)`$ one can associate a complex number
$$\frac{\mathrm{trace}(\text{monodromy of }A\text{ around }w|D^2)}{\mathrm{rank}(E)}\mathrm{exp}(_{D^2}w^{}\beta _{}),$$
which is invariant under deformations of $`w`$. These numbers, as well as certain variations of them, would be used as weights in the counting procedure which underlies the definition of Floer cohomology. For simplicity, we will stick to the case $`\mathrm{im}(\beta _{})=0`$ in our discussion.
In parallel with graded Lagrangian submanifolds, there is also a notion of graded symplectic automorphisms; in fact these are just a special kind of graded Lagrangian submanifolds on $`(M,\beta )\times (M,\beta )`$. The graded symplectic automorphisms form a topological group $`\mathrm{Symp}^{\mathrm{gr}}(M,\beta )`$ which is a central extension of the usual symplectic automorphism group $`\mathrm{Symp}(M,\beta )`$ by $``$. $`\mathrm{Symp}^{\mathrm{gr}}(M,\beta )`$ acts naturally on the set of graded Lagrangian submanifolds. Moreover, the central subgroup $``$ is generated by a graded symplectic automorphism denoted by $`[1]`$, which maps each graded Lagrangian submanifold $`\stackrel{~}{L}`$ to $`\stackrel{~}{L}[1]`$; we refer again to for details. Because $`D^bFuk(M,\beta )`$ is defined in what are essentially symplectic terms, every graded symplectic automorphism of $`M`$ induces an exact self-equivalence of it. Moreover, an isotopy of graded symplectic automorphisms will give rise to an equivalence between the induced functors. Thus one has a canonical map
$$\pi _0(\mathrm{Symp}^{\mathrm{gr}}(M,\beta ))\mathrm{Auteq}(D^bFuk(M,\beta )).$$
Now we return to Kontsevich’s conjecture. Assume that $`(M,\beta )`$ has a mirror partner $`X`$ such that (1.3) holds. Then there is an isomorphism between $`\mathrm{Auteq}(D^bFuk(M,\beta ))`$ and $`\mathrm{Auteq}(D^b(X))`$. Combining this with the canonical map above yields a homomorphism
(1.4)
$$\mu :\pi _0(\mathrm{Symp}^{\mathrm{gr}}(M,\beta ))\mathrm{Auteq}(D^b(X)).$$
Somewhat oversimplified, and ignoring the conjectural nature of the whole discussion, one can say that symplectic automorphisms of $`M`$ induce self-equivalences of the derived category of coherent sheaves on its mirror partner. Note that the map $`\mu `$ depends on the choice of equivalence (1.3) and hence is not canonical.
###### Remark 1.5.
One can see rather easily that the central element $`[1]\mathrm{Symp}^{\mathrm{gr}}(M,\beta )`$ induces the translation functor on $`D^bFuk(M,\beta )`$ and hence on $`D^b(X)`$. Passing to the quotient yields a map
$$\overline{\mu }:\pi _0(\mathrm{Symp}(M,\beta ))\mathrm{Auteq}(D^b(X))/(\mathrm{translations}).$$
This simplified version may be more convenient for those readers who are unfamiliar with the ‘graded symplectic’ machinery.
### 1c. Dehn twists and mirror symmetry
A Lagrangian sphere in $`(M,\beta )`$ is a Lagrangian submanifold $`SM`$ which is diffeomorphic to $`S^n`$. One can associate to any Lagrangian sphere a symplectic automorphism $`\tau _S`$ called the generalized Dehn twist along $`S`$, which is defined by a local construction in a neighbourhood of $`S`$ (see or for details; strictly speaking, $`\tau _S`$ depends on various choices, but since the induced functor on $`D^bFuk(M,\beta )`$ is expected to be independent of these choices, we will ignore them in our discussion). These maps are symplectic versions of the classical Picard-Lefschetz transformations. In particular, their action on $`H_{}(M)`$ is given by
(1.5)
$$(\tau _S)_{}(x)=\{\begin{array}{cc}x([S]x)[S]\hfill & \text{if }dim(x)=n\text{,}\hfill \\ x\hfill & \text{otherwise.}\hfill \end{array}$$
where $``$ is the intersection pairing twisted by a dimension-dependent sign. As explained in \[48, section 5b\] $`\tau _S`$ has a preferred lift $`\stackrel{~}{\tau }_S\mathrm{Symp}^{\mathrm{gr}}(M,\beta )`$ to the graded symplectic automorphism group. Suppose that $`(M,\beta )`$ has a mirror partner $`X`$ such that Kontsevich’s conjecture (1.3) holds. Choose some lift $`\stackrel{~}{S}`$ of $`S`$ to a graded Lagrangian submanifold, and let $`D^b(X)`$ be the object which corresponds to $`\stackrel{~}{S}`$. Then
(1.6)
$$\mathrm{Hom}_{D^b(X)}^{}(,)\mathrm{Hom}_{D^bFuk(M,\beta )}^{}(\stackrel{~}{S},\stackrel{~}{S})=HF^{}(\stackrel{~}{S},\stackrel{~}{S})_{}.$$
The Floer cohomology $`HF^{}(\stackrel{~}{S},\stackrel{~}{S})`$ is isomorphic to the ordinary cohomology $`H^{}(S;)`$; this is not true for general Lagrangian submanifolds, but it holds for spheres. Therefore $``$ must be spherical object (this motivated our use of the word spherical). A natural conjecture about the homomorphism $`\mu `$ introduced in the previous section is that
(1.7)
$$\mu ([\stackrel{~}{\tau }_S])=[T_{}],$$
where $`T_{}`$ is the twist functor as defined in section 1a. Roughly speaking, the idea is twist functors and generalized Dehn twists correspond to each other under mirror symmetry. At present this is merely a guess, which can be motivated e.g. by comparing (1.2) with (1.5). But supposing that one wanted to actually prove this claim, how should one go about it? The first step would be to observe that for any $`D^b(X)`$ there is an exact triangle
Here $`\mathrm{Hom}^{}(,)`$ is the graded group of homs in the derived category, $`\mathrm{Hom}^{}(,)_{}`$ is the corresponding direct sum of shifted copies of $``$, and the first arrow is the evaluation map. This exact triangle determines $`T_{}()`$ up to isomorphism; moreover, it does so in purely abstract terms, which involve only the triangulated structure of the category $`D^b(X)`$. Hence if there was an analogous abstract description of the action of $`\stackrel{~}{\tau }_S`$ on $`D^bFuk(M,\beta )`$ one could indeed prove (1.7) (this is slightly imprecise, since it ignores a technical problem about non-functoriality of cones in triangulated categories). The first step towards such a description will be provided in . Note that here, for the first time in our discussion of mirror symmetry, we have made essential use of the triangulated structure of the categories.
Now define an $`(A_m)`$-configuration of Lagrangian spheres in $`(M,\beta )`$ to be a collection of $`m1`$ pairwise transverse Lagrangian spheres $`S_1,\mathrm{},S_mM`$ such that
(1.8)
$$|S_iS_j|=\{\begin{array}{cc}1\hfill & |ij|=1,\hfill \\ 0\hfill & |ij|2.\hfill \end{array}$$
Such configurations occur in Kähler manifolds that can be degenerated into a manifold with a singular point of type $`(A_m)`$ (see or ). The generalized Dehn twists $`\stackrel{~}{\tau }_{S_1},\mathrm{},\stackrel{~}{\tau }_{S_m}`$ along such spheres satisfy the braid relations up to isotopy inside $`\mathrm{Symp}^{\mathrm{gr}}(M,\beta )`$. For $`n=2`$, and ignoring the issue of gradings, this was proved in \[49, Appendix\]; the argument given there can be adapted to yield the slightly sharper and more general statement which we are using here. Thus, by mapping the standard generators of the braid group to the classes $`[\stackrel{~}{\tau }_{S_i}]`$ one obtains a homomorphism from $`B_{m+1}`$ to $`\pi _0(\mathrm{Symp}^{\mathrm{gr}}(M,\beta ))`$. It is a difficult open question in symplectic geometry whether this homomorphism, which we denote by $`\rho ^{}`$, is injective; see for a partial result. We will now see what mirror symmetry has to say about this.
Assume as before that Kontsevich’s conjecture holds, and let $`_1,\mathrm{},_mD^b(X)`$ be the objects corresponding to some choice of gradings $`\stackrel{~}{S}_1,\mathrm{},\stackrel{~}{S}_m`$ for the $`S_j`$. We already know that each $`_i`$ is a spherical object. An argument similar to (1.6) but based on (1.8) shows that $`_1,\mathrm{},_m`$ is an $`(A_m)`$-configuration in $`D^b(X)`$ in the sense of Definition 1.1. Hence the twist functors $`T__i`$ satisfy the braid relations (Theorem 1.2) and generate a homomorphism $`\rho `$ from $`B_{m+1}`$ to $`\mathrm{Auteq}(D^b(X))`$. Assuming that our claim (1.7) is true, one would have a commutative diagram
Since $`dim_{}X=n2`$, we have Theorem 1.3 which says that $`\rho `$ is injective. In the diagram above this would clearly imply that $`\rho ^{}`$ is injective. Thus we are led to a conjectural answer ‘based on mirror symmetry’ to a question of symplectic geometry:
###### Conjecture 1.6.
Let $`(M,\beta )`$ be a compact symplectic manifold with $`\pi _1(M)`$ trivial and $`c_1(M,\beta )=0`$, and $`(S_1,\mathrm{},S_m)`$ an $`(A_m)`$-configuration of Lagrangian spheres in $`M`$ for some $`m1`$. Then the map $`\rho ^{}:B_{m+1}\pi _0(\mathrm{Symp}^{\mathrm{gr}}(M,\beta ))`$ generated by the generalized Dehn twists $`\stackrel{~}{\tau }_{S_1},\mathrm{},\stackrel{~}{\tau }_{S_m}`$ is injective.
### 1d. A survey of the paper
Section 2 introduces spherical objects and twists functors for derived categories of fairly general abelian categories. The main result is the construction of braid group actions, Theorem 2.17.
Section 3a explains how the abstract framework specializes in the case of coherent sheaves; this recovers the definitions presented in section 1a, and in particular Theorem 1.2. More generally, in section 3b, we consider singular and quasi-projective varieties, as well as equivariant sheaves on varieties with a finite group action; the latter give rise to what are probably the simplest examples of our theory. In section 3c we present a more systematic way of producing spherical objects, which exploits their relations with the (much studied) exceptional objects on Fano varieties. Elliptic curves provide the only example where both sides of the homological mirror conjecture are completely understood; in section 3d the group of symplectic automorphisms and the group of autoequivalences of the derived category are compared in an explicit way. Section 3e gives more explicit examples on $`K3`$ surfaces, then finally section 3f puts our results in the framework of mirror symmetry for singularities; this was the underlying motivation for much of this work.
Section 4 contains the proof of the faithfulness result, Theorem 2.18. For the benefit of the reader, we provide here an outline of the argument, in the more concrete situation stated as Theorem 1.3 above; the general case does not differ greatly from this. Let $`_1,\mathrm{},_m`$ be a collection of spherical objects in $`D^b(X)`$, and set $`=_1\mathrm{}_m`$. For a fixed $`m`$ and dimension $`n`$ of the variety, the endomorphism algebra
$$\mathrm{End}^{}()=\underset{i,j}{}\mathrm{Hom}^{}(_i,_j)$$
is essentially the same for all $`(_1,\mathrm{},_m)`$. More precisely, after possibly shifting each $`_i`$ by some amount, one can achieve that $`\mathrm{End}^{}()`$ is equal to a specific graded algebra $`A_{m,n}`$ depending only on $`m,n`$. Moreover, one can define a functor $`\mathrm{\Psi }^{\text{naive}}:D^b(X)A_{m,n}\text{-}mod`$ into the category of graded modules over $`A_{m,n}`$ by mapping $``$ to $`\mathrm{Hom}^{}(,)`$. By a result of the derived category $`D^b(A_{m,n}\text{-}mod)`$ carries a weak action of $`B_{m+1}`$, and one might hope that $`\mathrm{\Psi }^{\text{naive}}`$ should be compatible with these two actions. A little thought shows that this cannot possibly be true: $`A_{m,n}\text{-}mod`$ can be embedded into $`D^b(A_{m,n}\text{-}mod)`$ as the subcategory of complexes of length one, but the braid group action on $`D^b(A_{m,n}\text{-}mod)`$ does not preserve this subcategory. Nevertheless, the basic idea can be saved, at the cost of introducing some more homological algebra.
Take resolutions $`_i^{}`$ of $`_i`$ by bounded below complexes of injective quasi-coherent sheaves. Then one can define a differential graded algebra $`end(^{})`$ whose cohomology is $`\mathrm{End}^{}()`$. The quasi-isomorphism type of $`end(^{})`$ is independent of the choice of resolutions, so it is an invariant of the $`(A_m)`$-configuration $`_1,\mathrm{},_m`$. As before there is an exact functor $`hom(^{},):D^b(X)D(end(^{}))`$ to the derived category of differential graded modules over $`end(^{})`$. Now assume that $`end(^{})`$ is formal, that is to say, quasi-isomorphic to the differential graded algebra $`𝒜_{m,n}=(A_{m,n},0)`$ with zero differential. Quasi-isomorphic differential graded algebras have equivalent derived categories, so what one obtains is an exact functor
$$\mathrm{\Psi }:D^b(X)D(𝒜_{m,n}),$$
which replaces the earlier $`\mathrm{\Psi }^{\text{naive}}`$. A slight modification of the arguments of shows that there is a weak braid group on $`D(𝒜_{m,n})`$; moreover, in contrast with the situation above, the functor $`\mathrm{\Psi }`$ now relates the two braid group actions. Still borrowing from , one can interpret the braid group action on $`D(𝒜_{m,n})`$ in terms of low-dimensional topology, and more precisely geometric intersection numbers of curves on a punctured disc. This leads to a strong faithfulness result for it, which through the functor $`\mathrm{\Psi }`$ implies the faithfulness of the original braid group action on $`D^b(X)`$.
This argument by reduction to the known case of $`D(𝒜_{m,n})`$ hinges on the formality of $`end(^{})`$. We will prove that this assumption is always satisfied when $`n2`$. This has nothing to do with the geometric origin of $`end(^{})`$; in fact, what we will show is that $`A_{m,n}`$ is intrinsically formal for $`n2`$, which means that all differential graded algebras with this cohomology are formal. There is a general theory of intrinsically formal algebras, which goes back to the work of Halperin and Stasheff in the commutative case; the Hochschild cohomology computation necessary to apply this theory to $`A_{m,n}`$ is the final step in the proof of Theorem 2.18.
Acknowledgments. Although he does not figure as an author, the paper was originally conceived jointly with Mikhail Khovanov, and several of the basic ideas are his. At an early stage of this work, we had a stimulating conversation with Maxim Kontsevich. We would also like to thank Mark Gross for discussions about mirror symmetry and singularities, and Brian Conrad, Umar Salam, and Balazs Szendroi for helpful comment. As mentioned earlier, Kontsevich, Bridgeland and Maciocia also knew about the invertibility of the twist functors. Financial support came from Max Planck Institute (Bonn) and Hertford College (Oxford).
Addendum. The results here were first announced at the Harvard Winter School on Mirror Symmetry in January of 1999 (published in ). In the meantime, a preprint by Horja has appeared which is inspired by similar mirror symmetry considerations. While there is little actual overlap ( does not operate in the derived category) Horja uses monodromy calculations to predict corresponding conjectural mirror Fourier-Mukai transforms that ought to be connected to our work, linking it to the toric construction of mirror manifolds.
## 2. Braid group actions
### 2a. Generalities
Fix a field $`k`$; all categories are assumed to be $`k`$-linear. If $`𝔖`$ is an abelian category, $`Ch(𝔖)`$ is the category of cochain complexes in $`𝔖`$ and cochain maps, $`K(𝔖)`$ the corresponding homotopy category (morphisms are homotopy classes of cochain maps), and $`D(𝔖)`$ the derived category. The variants involving bounded (below, above, or on both sides) complexes are denoted by $`Ch^+(𝔖)`$, $`Ch^{}(𝔖)`$, $`Ch^b(𝔖)`$ and so on. Let $`(C_j,\delta _j)_j`$ be a cochain complex of objects and morphisms in $`Ch(𝔖)`$, that is to say $`C_jCh(𝔖)`$ and $`\delta _j\mathrm{Hom}_{Ch(𝔖)}(C_j,C_{j+1})`$ satisfying $`\delta _{j+1}\delta _j=0`$. Such a complex is exactly the same as a bicomplex in $`𝔖`$. In this case we will write $`\{\mathrm{}C_1C_0C_1\mathrm{}\}`$ for the associated total complex, obtained by collapsing the bigrading; this is a single object in $`Ch(𝔖)`$. The same notation will be applied to bicomplexes of objects of $`Ch(𝔖)`$ (which are triple complexes in $`𝔖`$).
For $`C,DCh(𝔖)`$, let $`hom(C,D)`$ be the standard cochain complex of $`k`$-vector spaces whose cohomology is $`H^ihom(C,D)=\mathrm{Hom}_{K(𝔖)}^i(C,D)`$, that is, $`hom^i(C,D)=_j\mathrm{Hom}_𝔖(C^j,D^{j+i})`$ with $`d_{hom(C,D)}^i(\varphi )=d_D\varphi (1)^i\varphi d_C`$. Now suppose that $`𝔖`$ contains infinite direct sums and products. Given an object $`CCh(𝔖)`$ and a cochain complex $`b`$ of $`k`$-vector spaces, one can form the tensor product $`bC`$ and the complex of linear maps $`lin(b,C)`$, both of which are again objects of $`Ch(𝔖)`$. They are defined by choosing a basis of $`b`$ and taking a corresponding direct sum (for $`bC`$) or product (for $`lin(b,C)`$) of shifted copies of $`C`$, with a differential which combines $`d_b`$ and $`d_C`$. The outcome is independent of the chosen basis up to canonical isomorphism. The definition of $`bC`$ is clear, but for $`lin(b,C)`$ there are two possible choices of signs. Ours is fixed to fit in with an evaluation map $`blin(b,C)C`$. To clarify the issue we will now spell out the definition. Take a homogeneous basis $`(x_i)_{iI}`$ of the total space $`b`$, and write $`d_b(x_i)=_jz_{ji}x_j`$. Then $`lin^q(b,C)=_{iI}C_i^q`$, where $`C_i`$ is a copy of $`C`$ shifted by $`\mathrm{deg}(x_i)`$. The differential $`d^q:lin^q(b,C)lin^{q+1}(b,C)`$ has components $`d_{ji}^q:C_i^qC_j^{q+1}`$ which are given by
$$d_{ji}^q=\{\begin{array}{cc}(1)^{\mathrm{deg}(x_i)}d_C\hfill & i=j,\hfill \\ (1)^{\mathrm{deg}(x_i)}z_{ij}\mathrm{id}_C\hfill & \mathrm{deg}(x_i)=\mathrm{deg}(x_j)+1,\hfill \\ 0\hfill & \text{otherwise.}\hfill \end{array}$$
One can verify that the map $`blin(b,C)C`$, $`x_j(c_i)_{iI}c_j`$, is indeed a morphism in $`Ch(𝔖)`$. Moreover, there are canonical monomorphic cochain maps
(2.1) $`bhom(D,C)`$ $`hom(D,bC),`$
$`hom(D,C)b`$ $`hom(lin(b,D),C),`$
$`hom(B,lin(b,C))D`$ $`lin(b,hom(B,C)D),`$
where $`b`$ is as before and $`B,C,DCh(𝔖)`$. These maps are isomorphisms if $`b`$ is finite-dimensional, and quasi-isomorphisms if $`b`$ has finite-dimensional cohomology.
From now on $`𝔖`$ will be an abelian category and $`𝔖^{}𝔖`$ a full subcategory, such that the following conditions hold:
* $`𝔖^{}`$ is a Serre subcategory of $`𝔖`$ (this means that any subobject and quotient object of an object in $`𝔖^{}`$ lies again in $`𝔖^{}`$, and that $`𝔖^{}`$ is closed under extension);
* $`𝔖`$ contains infinite direct sums and products;
* $`𝔖`$ has enough injectives, and any direct sum of injectives is again injective (this is not a trivial consequence of the definition of an injective object);
* for any epimorphism $`f:AA^{}`$ with $`A𝔖`$ and $`A^{}𝔖^{}`$, there is a $`B^{}𝔖^{}`$ and a $`g:B^{}A`$ such that $`fg`$ is an epimorphism (because $`𝔖^{}`$ is a Serre subcategory, $`g`$ may be taken to be mono):
###### Lemma 2.1.
Let $`X`$ be a noetherian scheme over $`k`$ and $`𝔖=Qco(X)`$, $`𝔖^{}=Coh(X)`$ the categories of quasi-coherent resp. coherent sheaves. Then properties (C1)–(C4) are satisfied.
###### Proof.
(C1) and (C2) are obvious. $`𝔖`$ has enough injectives by \[19, II 7.18\]. Moreover, it is locally noetherian, which implies that direct sums of injectives are again injective; see \[19, p. 121\] and the references quoted there. This proves (C3). Finally, we need to verify that a diagram as in (C4) with $`A`$ quasi-coherent and $`A^{}`$ coherent, can be completed with a coherent sheaf $`B^{}`$. Such a $`B^{}`$ certainly exists locally, and replacing it by its image in $`A`$ (which is also coherent) we may extend it to be a coherent subsheaf on all of $`X`$ (see EGA I 9.4.7). Since $`X`$ is quasi-compact, repeating this a finite number of times and taking the union yields a $`B^{}`$ whose map to $`A^{}`$ is globally onto. ∎
As indicated by this example, our main interest is in $`D^b(𝔖^{})`$. However we find it convenient to replace all complexes by injective resolutions. These resolutions may exist only in $`𝔖`$, and they are not necessarily bounded. The precise category we want to work with is this:
###### Definition 2.2.
$`𝔎K^+(𝔖)`$ is the full subcategory whose objects are those bounded below cochain complexes $`C`$ of $`𝔖`$-injectives which satisfy $`H^i(C)𝔖^{}`$ for all $`i`$, and $`H^i(C)=0`$ for $`i0`$.
We will now prove, in several steps, that $`𝔎`$ is equivalent to $`D^b(𝔖^{})`$. First of all, let $`𝔇D^+(𝔖)`$ be the full subcategory of objects whose cohomology has the same properties as in Definition 2.2. The assumption that $`𝔖`$ has enough injectives implies that the obvious functor $`𝔎𝔇`$ is an equivalence. Now let $`Ch_𝔖^{}^b(𝔖)`$ be the category of bounded cochain complexes in $`𝔖`$ whose cohomology objects lie in $`𝔖^{}`$, and $`D_𝔖^{}^b(𝔖)`$ the corresponding full subcategory of $`D^b(𝔖)`$. It is a standard result (proved by truncating cochain complexes) that the obvious functor $`D_𝔖^{}^b(𝔖)𝔇`$ is an equivalence. The final step (and the only nontrivial one) is to relate $`D_𝔖^{}^b(𝔖)`$ and $`D^b(𝔖^{})`$.
###### Lemma 2.3.
For any $`CCh_𝔖^{}^b(𝔖)`$ there is an $`ECh^b(𝔖^{})`$ and a monomorphic cochain map $`\iota :EC`$ which is a quasi-isomorphism.
###### Proof.
Recall that, as an abelian category, $`𝔖`$ has fibre products. The fibre product of two maps $`f_1:A_1A`$, $`f_2:A_2A`$ is the kernel of $`f_100f_2:A_1A_2A`$. If $`f_1`$ is mono (thought of as an inclusion) we write $`f_2^1(A_1)`$ for the fibre product, and if both $`f_1`$ and $`f_2`$ are mono we write $`A_1A_2`$. In the latter case one can also define the sum $`A_1+A_2`$ as the image (kernel of the map to the cokernel) of $`f_100f_2`$.
Let $`N`$ be the largest integer such that $`C^N0`$. Set $`E^n=0`$ for all $`n>N`$. For $`nN`$ define $`E^nC^n`$ (for brevity, we write the monomorphisms as inclusions) inductively as follows. By invoking (C4) one finds subobjects $`F^n,G^nC^n`$ which lie in $`𝔖^{}`$ and complete the diagrams
Set $`E^n=F^n+G^n`$ (this is again in $`𝔖^{}`$) and define $`d_E^n=d_C^n|E^n`$. Since $`E^n`$ is a subobject of $`C^n`$ for any $`n`$, $`E`$ is a bounded complex. Consider the obvious map $`j^n:\mathrm{ker}d_E^n=E^n\mathrm{ker}d_C^nH^n(C)`$ . The definition of $`G^n`$ implies that $`j^n`$ is an epimorphism, and the definition of $`F^{n1}`$ yields $`\mathrm{ker}j^n=E^n\mathrm{im}d_C^{n1}=\mathrm{im}d_E^{n1}`$. It follows that the inclusion induces an isomorphism $`H^{}(E)H^{}(C)`$. ∎
From this Lemma it now follows by standard homological algebra \[14, Proposition III.2.10\] that the obvious functor $`D^b(𝔖^{})D_𝔖^{}^b(𝔖)`$ is an equivalence of categories. Combining this with the remarks made above, one gets
###### Proposition 2.4.
There is an exact equivalence (canonical up to natural isomorphism) $`D^b(𝔖^{})𝔎`$. ∎
### 2b. Twist functors and spherical objects
###### Definition 2.5.
Let $`E𝔎`$ be an object satisfying the following finiteness conditions:
* $`E`$ is a bounded complex,
* for any $`F𝔎`$, both $`\mathrm{Hom}_𝔎^{}(E,F)`$ and $`\mathrm{Hom}_𝔎^{}(F,E)`$ have finite (total) dimension over $`k`$.
Then we define the twist functor $`T_E:𝔎𝔎`$ by
(2.2)
$$T_E(F)=\{hom(E,F)E\stackrel{ev}{}F\}.$$
This expression requires some explanation. $`ev`$ is the obvious evaluation map. The grading is such that if one ignores the differential, $`T_E(F)=F(hom(E,F)E)[1]`$. In other words $`T_E(F)`$ is the cone of $`ev`$. Since $`E`$ is bounded and $`F`$ is bounded below, $`hom(E,F)`$ is again bounded below. Hence $`hom(E,F)E`$ is a bounded below complex of injectives in $`𝔖`$ (this uses property (C3) of $`𝔖`$). Its cohomology $`H^{}(hom(E,F)E)`$ is isomorphic to $`\mathrm{Hom}_𝔎^{}(E,F)H^{}(E)`$ (for instance because $`hom(E,F)`$ is quasi-isomorphic to $`Hom_𝔎^{}(E,F)`$, which is finite dimensional), and so is bounded, and the finiteness conditions imply that each cohomology group lies in $`𝔖^{}`$. Therefore $`hom(E,F)E`$ lies in $`𝔎`$, and the same holds for $`T_E(F)`$. The functoriality of $`T_E`$ is obvious, and one sees easily that it is an exact functor. Actually, for any $`F,G𝔎`$ there is a canonical map of complexes $`(T_E)_{}:hom(F,G)hom(T_E(F),T_E(G))`$. In fancy language, this means that $`T_E`$ is functorial on the differential graded category which underlies $`𝔎`$.
###### Proposition 2.6.
If two objects $`E_1,E_2𝔎`$ satisfying (K1), (K2) are isomorphic, the corresponding functors $`T_{E_1},T_{E_2}`$ are isomorphic.
###### Proof.
Take cones of the rows of the following commutative diagram,
Here the vertical arrows are induced by a quasi-isomorpism of complexes $`E_1E_2`$. ∎
Note also that $`T_{E[j]}`$ is isomorphic to $`T_E`$ for any $`j`$.
###### Definition 2.7.
For an object $`E`$ as in Definition 2.5 we define the dual twist functor $`T_E^{}:𝔎𝔎`$ by $`T_E^{}(F)=\{ev^{}:Flin(hom(F,E),E)\}`$.
Here the grading is such that $`F`$ lies in degree zero. $`ev^{}`$ is again some kind of evaluation map. To write it down explicitly, choose a homogeneous basis $`(\psi _i)`$ of $`hom(F,E)`$. Then $`lin^q(hom(F,E),E)=_iE_i^q`$, where $`E_i`$ is a copy of $`E[\mathrm{deg}(\psi _i)]`$, and the $`i`$-th component of $`ev^{}`$ is simply $`\psi _i`$ itself. $`T_E^{}`$ is again an exact functor from $`𝔎`$ to itself.
###### Lemma 2.8.
$`T_E^{}`$ is left adjoint to $`T_E`$.
###### Proof.
Using the maps from (2.1) and condition (K2) one constructs a chain of natural (in $`F,G𝔎`$) quasi-isomorphisms
$`hom(F,T_E(G))=`$ $`\{hom(F,hom(E,G)E)hom(F,G)\}`$
$``$ $`\{hom(E,G)hom(F,E)hom(F,G)\}`$
$``$ $`\{hom(lin(hom(F,E),E),G)hom(F,G)\}`$
$`=`$ $`hom(T_E^{}(F),G).`$
Here the chain map $`hom(E,G)hom(F,E)hom(F,G)`$ is just composition. The reader may easily check that the required diagrams commute. Taking $`H^0`$ on both sides yields a natural isomorphism $`\mathrm{Hom}_𝔎(F,T_E(G))\mathrm{Hom}_𝔎(T_E^{}(F),G)`$. ∎
###### Definition 2.9.
An object $`E𝔎`$ is called $`n`$-spherical for some $`n>0`$ if it satisfies (K1), (K2) above and in addition,
* $`\mathrm{Hom}_𝔎^i(E,E)`$ is equal to $`k`$ for $`i=0,n`$ and zero in all other degrees;
* The composition $`\mathrm{Hom}_𝔎^j(F,E)\times \mathrm{Hom}_𝔎^{nj}(E,F)\mathrm{Hom}_𝔎^n(E,E)k`$ is a nondegenerate pairing for all $`F𝔎`$, $`j`$.
One can also define $`0`$-spherical objects: these are objects $`E`$ for which $`\mathrm{Hom}_𝔎^{}(E,E)`$ is two-dimensional and concentrated in degree zero, and such that the pairings $`\mathrm{Hom}_𝔎^j(E,F)\times \mathrm{Hom}_𝔎^j(F,E)\mathrm{Hom}_𝔎^0(E,E)/k\mathrm{id}_E`$ are nondegenerate (this means in particular that $`\mathrm{Hom}_𝔎^0(E,E)`$ is isomorphic to $`k[t]/t^2`$ as a $`k`$-algebra). We will not pursue this further; the interested reader can easily verify that the proof of the next Proposition extends to this case.
###### Proposition 2.10.
If $`E`$ is $`n`$-spherical for some $`n>0`$, both $`T_E^{}T_E`$ and $`T_ET_E^{}`$ are naturally isomorphic to the identity functor $`\mathrm{Id}_𝔎`$. In particular, $`T_E`$ is an exact self-equivalence of $`𝔎`$.
###### Proof.
<sup>1</sup><sup>1</sup>1We thank one of the referees for simplifying our original proof of this result.
$`T_ET_E^{}(F)`$ is a total complex
(2.3)
$$\left\{\begin{array}{ccc}hom(E,F)E& \stackrel{\delta }{}& hom(E,lin(hom(F,E),E))E\\ \alpha & & \gamma & & \\ F& \stackrel{\beta }{}& lin(hom(F,E),E)\end{array}\right\}$$
Here $`\alpha =ev`$, $`\beta =ev^{}`$, $`\gamma `$ is a map induced by $`ev`$, and $`\delta `$ a map induced by $`ev^{}`$. We shall need to know a little more about $`\delta `$. By the very definition of $`ev^{}`$ by duality, $`\delta `$’s induced map on cohomology
(2.4)
$$\mathrm{Hom}_𝔎^{}(E,F)H^{}(E)\mathrm{Hom}_𝔎^{}(F,E)^{}\mathrm{Hom}_𝔎^{}(E,E)H^{}(E)$$
is dual to the the composition $`\mathrm{Hom}_𝔎^{}(F,E)\mathrm{Hom}_𝔎^{}(E,F)\mathrm{Hom}_𝔎^{}(E,E)`$, tensored with the identity map on $`H^{}(E)`$. This second pairing is, by the conditions (K3) and (K4) on $`E`$, perfect when we divide $`\mathrm{Hom}_𝔎^{}(E,E)`$ by its degree zero piece $`(k\mathrm{id}_E)`$. Thus the following modification of the map (2.4),
(2.5)
$$\mathrm{Hom}_𝔎^{}(E,F)H^{}(E)\mathrm{Hom}_𝔎^{}(F,E)^{}\frac{\mathrm{Hom}_𝔎^{}(E,E)}{k\mathrm{id}_E}H^{}(E),$$
is an isomorphism.
We now enlarge slightly the object in the top right hand corner of (2.3) to produce a new, quasi-isomorphic, complex $`Q_E(F)`$. The last equation in (2.1) gives a map $`hom(E,lin(hom(F,E),E))Elin(hom(F,E),hom(E,E)E)`$. Since $`hom(F,E)`$ has finite-dimensional cohomology, this is a quasi-isomorphism. $`\gamma `$ extends naturally to $`\overline{\gamma }:lin(hom(F,E),hom(E,E)E)lin(hom(F,E),E)`$; it is just the map induced by $`ev:hom(E,E)EE`$.
In fact $`\overline{\gamma }`$ splits canonically: define the map $`\varphi :lin(hom(F,E),E)lin(hom(F,E),hom(E,E)E`$ induced by $`khom(E,E)`$, $`1\mathrm{id}_E`$. From the definition of $`\overline{\gamma }`$ it follows that $`\overline{\gamma }\varphi =\mathrm{id}`$. This splitting gives a way of embedding an acyclic complex $`\{\mathrm{id}:lin(hom(F,E),E)lin(hom(F,E),E)\}`$ into our enlarged complex $`Q_E(F)`$; the cokernel is
$$\{hom(E,F)E\stackrel{\delta \alpha }{}lin(hom(F,E),\frac{hom(E,E)}{k\mathrm{id}_E}E)F\}.$$
There is an obvious map of $`F`$ to this, and everything we have done is functorial in $`F`$; thus to prove that $`T_ET_E^{}\mathrm{Id}_𝔎`$ we are left with showing that the cokernel
(2.6)
$$\{hom(E,F)E\stackrel{𝛿}{}lin(hom(F,E),\frac{hom(E,E)}{k\mathrm{id}_E}E)\}$$
is acyclic, i.e. the arrow induces an isomorphism on cohomology. But passing to cohomology yields (2.5), which we already noted was an isomorphism.
The proof that $`T_E^{}T_E\mathrm{Id}_𝔎`$ is similar; one passes from $`T_E^{}T_E(F)`$ to a quasi-isomorphic but slightly smaller object, which then has a natural map to $`F`$. The details are almost the same as before, and we leave them to the reader. ∎
### 2c. The braid relations
###### Lemma 2.11.
Let $`E_1,E_2𝔎`$ be two objects such that $`E_1`$ satisfies the conditions (K1), (K2) of Definition 2.5, and $`E_2`$ is $`n`$-spherical for some $`n>0`$. Then $`T_{E_2}(E_1)`$ also satisfies (K1), (K2) and $`T_{E_2}T_{E_1}`$ is naturally isomorphic to $`T_{T_{E_2}(E_1)}T_{E_2}`$.
###### Proof.
Since $`E_1`$ and $`E_2`$ are bounded complexes, so are $`hom(E_1,E_2)`$ and $`T_{E_2}(E_1)`$. Lemma 2.8 says that $`\mathrm{Hom}_𝔎^{}(F,T_{E_2}(E_1))\mathrm{Hom}_𝔎^{}(T_{E_2}^{}(F),E_1)`$. By assumption on $`E_1`$, this implies that $`\mathrm{Hom}_𝔎^{}(F,T_{E_2}(E_1))`$ is always finite-dimensional. Similarly, the finite-dimensionality of $`\mathrm{Hom}_𝔎^{}(T_{E_2}(E_1),F)`$ follows from Proposition 2.10 since $`\mathrm{Hom}_𝔎^{}(T_{E_2}(E_1),F)\mathrm{Hom}_𝔎^{}(E_1,T_{E_2}^{}(F))`$. We have now proved that $`T_{E_2}(E_1)`$ satisfies (K1), (K2). $`T_{E_2}T_{E_1}(F)`$ is a total complex
$$\left\{\begin{array}{ccc}hom(E_2,hom(E_1,F)E_1)E_2& & hom(E_2,F)E_2\\ & & & & \\ hom(E_1,F)E_1& & F\end{array}\right\}$$
where all arrows are evaluation maps or induced by them. We will argue as in the proof of Proposition 2.10. Using (2.1) one sees that the object in the top left hand corner can be replaced by the smaller quasi-isomorphic one $`hom(E_1,F)hom(E_2,E_1)E_2`$. More precisely, this modification defines another functor $`R_{E_1,E_2}`$ on $`𝔎`$ which is naturally isomorphic to $`T_{E_2}T_{E_1}`$. One can rewrite the definition of this functor as
(2.7)
$$R_{E_1,E_2}(F)=\{hom(E_1,F)T_{E_2}(E_1)T_{E_2}(F)\}.$$
The arrow in (2.7) is obtained by composing
$$hom(E_1,F)T_{E_2}(E_1)\stackrel{(T_{E_2})_{}\mathrm{id}}{}hom(T_{E_2}(E_1),T_{E_2}(F))T_{E_2}(E_1)$$
with the evaluation map $`ev:hom(T_{E_2}(E_1),T_{E_2}(F))T_{E_2}(E_1)T_{E_2}(F)`$. This means that one has a natural map from $`R_{E_1,E_2}(F)`$ to $`T_{T_{E_2}(E_1)}T_{E_2}(F)`$, given by $`(T_{E_2})_{}\mathrm{id}`$ on the first component and by the identity on the second one. Since $`(T_{E_2})_{}`$ is a quasi-isomorphism by Proposition 2.10, this natural transformation is an isomorphism. ∎
###### Proposition 2.12.
Let $`E_1,E_2`$ be as before, and assume in addition that $`\mathrm{Hom}_𝔎^i(E_2,E_1)=0`$ for all $`i`$. Then $`T_{E_1}T_{E_2}T_{E_2}T_{E_1}`$.
###### Proof.
The assumption implies that $`T_{E_2}(E_1)`$ is isomorphic to $`E_1`$. Hence the result follows directly from Lemma 2.11 and Proposition 2.6 (one can also prove this by a direct computation, without using Lemma 2.11). ∎
###### Proposition 2.13.
Let $`E_1,E_2𝔎`$ be two $`n`$-spherical objects for some $`n>0`$. Assume that the total dimension of $`\mathrm{Hom}_𝔎^{}(E_2,E_1)`$ is one. Then $`T_{E_1}T_{E_2}T_{E_1}T_{E_2}T_{E_1}T_{E_2}`$.
###### Proof.
Since the twists are not affected by shifting, we may assume that $`\mathrm{Hom}_𝔎^i(E_2,E_1)`$ is one-dimensional for $`i=0`$ and zero in all other dimensions. A simple computation shows that
$$T_{E_2}(E_1)\{E_2\stackrel{g}{}E_1\},T_{E_1}^{}(E_2)\{E_2\stackrel{h}{}E_1\}$$
where $`g`$ and $`h`$ are nonzero maps. As $`\mathrm{Hom}_𝔎(E_2,E_1)`$ is one-dimensional it follows that $`T_{E_2}(E_1)`$ and $`T_{E_1}^{}(E_2)`$ are isomorphic up to the shift $`[1]`$. By applying Lemma 2.11 and Proposition 2.6 one finds that
$$T_{E_1}T_{E_2}T_{E_1}T_{E_1}T_{T_{E_2}(E_1)}T_{E_2}T_{E_1}T_{T_{E_1}^{}(E_2)}T_{E_2}.$$
On the other hand, applying Lemma 2.11 to $`T_{E_1}^{}(E_2)`$ and $`E_1`$, and using Proposition 2.10, shows that $`T_{E_1}T_{T_{E_1}^{}(E_2)}T_{E_2}T_{E_2}T_{E_1}T_{E_2}`$. ∎
We will now carry over the results obtained so far to the derived category $`D^b(𝔖^{})`$. During the rest of this section, $`\mathrm{Hom}`$ always means $`\mathrm{Hom}_{D^b(𝔖^{})}`$.
###### Definition 2.14.
An object $`ED^b(𝔖^{})`$ is called $`n`$-spherical for some $`n>0`$ if it has the following properties:
* $`E`$ has a finite resolution by injective objects in $`𝔖`$;
* $`\mathrm{Hom}^{}(E,F)`$, $`\mathrm{Hom}^{}(F,E)`$ are finite-dimensional for any $`FD^b(𝔖^{})`$.
* $`\mathrm{Hom}^i(E,E)`$ is equal to $`k`$ for $`i=0,n`$ and zero in all other dimensions;
* The composition map $`\mathrm{Hom}^i(F,E)\times \mathrm{Hom}^{ni}(E,F)\mathrm{Hom}^n(E,E)k`$ is a nondegenerate pairing for all $`F𝔎`$ and $`i`$.
Clearly, if $`E`$ is such an object, any finite resolution by $`𝔖`$-injectives is an $`n`$-spherical object of $`𝔎`$ in the sense of Definition 2.9. Using such a resolution, and the equivalence of categories from Proposition 2.4, one can associate to $`E`$ a twist functor $`T_E`$ which, by Proposition 2.10, is an exact self-equivalence of $`D^b(𝔖^{})`$. This will be independent of the choice of resolution up to isomorphism, thanks to Proposition 2.6.
###### Lemma 2.15.
In the presence of (S2) and (S3), condition (S4) is equivalent to the following apparently weaker one:
* There is an isomorphism $`\mathrm{Hom}(E,F)\mathrm{Hom}^n(F,E)^{}`$ which is natural in $`FD^b(𝔖^{})`$.
###### Proof.
The proof is by a ‘general nonsense’ argument. Take any natural isomorphism as in (S4’) and let $`q_F:\mathrm{Hom}(E,F)\times \mathrm{Hom}^n(F,E)k`$ be the family of nondegenerate pairings induced by it. Because of the naturality, these pairings satisfy $`q_F(\varphi ,\psi )=q_F(\varphi \mathrm{id}_E,\psi )=q_E(\mathrm{id}_E,\varphi \psi )`$. Since the pairings are all nondegenerate, $`q_E(\mathrm{id}_E,):\mathrm{Hom}^n(E,E)k`$ is nonzero, hence by (S3) an isomorphism. We have therefore shown that
$$\mathrm{Hom}(E,F)\times \mathrm{Hom}^n(F,E)\stackrel{\text{composition}}{}\mathrm{Hom}^n(E,E)k$$
is a nondegenerate pairing for any $`F`$, which is the special case $`i=0`$ of (S4). The other cases follow by replacing $`F`$ by $`F[i]`$. ∎
###### Lemma 2.16.
Let $`X`$ be a noetherian scheme over $`k`$ and $`𝔖=Qco(X)`$, $`𝔖^{}=Coh(X)`$. Then condition (S4) or (S4’) for an object of $`D^b(𝔖^{})`$ implies condition (S1).
###### Proof.
Let $``$ be an object of $`D^b(𝔖^{})`$ and $`𝔖^{}`$ a coherent sheaf. Since $``$ is bounded, and $``$ has a bounded below resolution by $`𝔖`$-injectives, one has $`\mathrm{Hom}^i(,)=0`$ for $`i0`$. Using (S4) or (S4’) it follows that $`\mathrm{Hom}^i(,)=0`$ for $`i0`$, and \[19, Proposition II.7.20\] completes the proof. ∎
Now define an $`(A_m)`$-configuration $`(m>0)`$ of $`n`$-spherical objects in $`D^b(𝔖^{})`$ to be a collection $`(E_1,\mathrm{},E_m)`$ of such objects, satisfying
(2.8)
$$dim_k\mathrm{Hom}_{D^b(𝔖^{})}^{}(E_i,E_j)=\{\begin{array}{cc}1\hfill & |ij|=1,\hfill \\ 0\hfill & |ij|2.\hfill \end{array}$$
###### Theorem 2.17.
Let $`(E_1,\mathrm{},E_m)`$ be an $`(A_m)`$-configuration of $`n`$-spherical objects in $`D^b(𝔖^{})`$. Then the twists $`T_{E_1},\mathrm{},T_{E_m}`$ satisfy the relations of the braid group $`B_{m+1}`$ up to graded natural isomorphism. That is to say, they generate a homomorphism $`\rho :B_{m+1}\mathrm{Auteq}(D^b(𝔖^{}))`$.
This follows immediately from the corresponding results for $`𝔎`$ (Propositions 2.12 and 2.13). One minor point remains to be cleared up: the Theorem states that the braid relations hold up to graded natural isomorphism, whereas before we have only talked about ordinary natural isomorphism. But one can easily see all the natural isomorphisms which we have constructed are graded ones, essentially because everything commutes with the translation functors. We can now state the main result of this paper:
###### Theorem 2.18.
Suppose that $`n2`$. Then the homomorphism $`\rho `$ defined in Theorem 2.17 is injective, and in fact the following stronger statement holds: if $`gB_{m+1}`$ is not the identity element, then $`\rho (g)(E_i)\cong ̸E_i`$ for some $`i\{1,\mathrm{},m\}`$.
## 3. Applications
### 3a. Smooth projective varieties
We now return to the concrete situation of derived categories of coherent sheaves. The main theme will be the use of suitable duality theorems to simplify condition (S4’) in the definition of spherical objects. Throughout, all varieties will be over an algebraically closed field $`k`$.
For the moment we consider only smooth projective varieties $`X`$, of dimension $`n`$. Let us recall some facts about duality on such varieties. Serre duality says that for any $`𝒢D^b(X)`$ the composition
(3.1)
$$\mathrm{Hom}^n(𝒢,\omega _X)\mathrm{Hom}^{}(𝒪,𝒢)\mathrm{Hom}^n(𝒪,\omega _X)=H^n(\omega _X)k$$
is a nondegenerate pairing (the classical form is for a single sheaf $`𝒢`$; the general case can be derived from this by induction on the length, using the Five-Lemma). Now let $``$ be a bounded complex of locally free coherent sheaves on $`X`$. For all $`𝒢_1,𝒢_2D^+(X)`$ there is a natural isomorphism
(3.2)
$$\mathrm{Hom}^{}(𝒢_1,𝒢_2)\mathrm{Hom}^{}(𝒢_1,𝒢_2^{}).$$
This is proved using a resolution $`𝒢_2^{}`$ of $`𝒢_2`$ by injective quasi-coherent sheaves; the point is that $`𝒢_2^{}^{}`$ is an injective resolution of $`𝒢_2^{}`$ \[19, Proposition 7.17\]. Setting $`𝒢=^{}`$ in (3.1) for some $`D^b(X)`$ and using (3.2) shows that there is an isomorphism, natural in $``$,
(3.3)
$$\mathrm{Hom}^{}(,)\mathrm{Hom}^n(,\omega _X)^{}.$$
Again by (3.2) and the standard finiteness theorems, $`\mathrm{Hom}^{}(,)^{}(^{})`$ is of finite total dimension; hence so is $`\mathrm{Hom}^{}(,)`$ by (3.3). Finally, because of the existence of finite locally free resolutions, everything we have said holds for an arbitrary $`D^b(X)`$.
###### Lemma 3.1.
An object $`D^b(X)`$ is spherical, in the sense of Definition 2.14, iff it satisfies the following two conditions: $`\mathrm{Hom}^j(,)`$ is one-dimensional for $`j=0,n`$ and zero for all other $`j`$; and $`\omega _X`$.
###### Proof.
It follows from (3.3) and the previous discussion that the conditions are sufficient. Conversely, assume that $``$ is a spherical object. Then property (S4) and (3.3) imply that the functors $`\mathrm{Hom}(,\omega _X)`$ and $`\mathrm{Hom}(,)`$ are isomorphic. By a general nonsense argument $``$ must be isomorphic to $`\omega _X`$. ∎
This shows that the abstract definition of spherical objects specializes to the one in section 1a. We will now prove the corresponding statement for twist functors.
###### Lemma 3.2.
Let $`D^b(X)`$ be a bounded complex of locally free sheaves, which is a spherical object. Then the twist functor $`T_{}`$ as defined in section 2c is isomorphic to the FMT by $`𝒫=\mathrm{Cone}(\eta :^{}𝒪_\mathrm{\Delta })`$.
###### Proof.
Let $`^{}𝔎`$ be a bounded resolution of $``$ by injective quasi-coherent sheaves. Let $`T:𝔎D^+(X)`$ be the functor which sends $``$ to $`\mathrm{Cone}(ev:hom(,))`$. We will show that the diagram
(3.4)
where the unlabeled arrows are the equivalence $`𝔎D^b(X)`$ and its inclusion into $`D^+(X)`$, commutes up to isomorphism. Since $`T_{}`$ is defined using the twist functor $`T_{^{}}`$ on $`𝔎`$ and $`𝔎D^b(X)`$, the commutativity of (3.4) implies that $`\mathrm{\Phi }_𝒫T_{}`$. Take an object $`D^b(X)`$ and a resolution $`^{}𝔎`$. Then
$`\mathrm{\Phi }_𝒫()`$ $`=𝐑\pi _2\left\{\pi _1^{}\pi _1^{}^{}\pi _2^{}𝒪_\mathrm{\Delta }\pi _1^{}\right\}`$
$`𝐑\pi _2\left\{\pi _1^{}^{}\pi _1^{}^{}\pi _2^{}𝒪_\mathrm{\Delta }\pi _1^{}^{}\right\}`$
$`\pi _{2,}\left\{\pi _1^{}\text{om}(,^{})\pi _2^{}𝒪_\mathrm{\Delta }\pi _1^{}^{}\right\}`$
$`\left\{hom(,^{})^{}\right\}=T(^{}),`$
where the arrow in the last line is evaluation. This provides a natural isomorphism which makes the left lower triangle in (3.4) commute. To deal with the other triangle, set up a diagram as in the proof of Proposition 2.6. ∎
###### Example 3.3.
Let $`X`$ be a variety which is Calabi-Yau in the strict sense, that is to say $`\omega _X𝒪`$ and $`H^i(X,𝒪)=0`$ for $`0<i<n`$. Then any invertible sheaf on $`X`$ is spherical. For the trivial sheaf, the twist $`T_𝒪`$ is the FMT given by the object on $`X\times X`$ which is the ideal sheaf of the diagonal shifted by $`[1]`$. This is what Mukai calls the ‘reflection functor’.
###### Lemma 3.4.
Let $`YX`$ be a connected subscheme which is a local complete intersection, with (locally free) normal sheaf $`\nu =(𝒥_Y/𝒥_Y^2)^{}`$. Assume that $`\omega _X|Y`$ is trivial, and that $`H^i(Y,\mathrm{\Lambda }^j\nu )=0`$ for all $`0<i+j<n`$. Then $`𝒪_YD^b(X)`$ is a spherical object.
###### Proof.
Denote by $`\iota `$ the embedding of $`Y`$ into $`X`$. The local Koszul resolution of $`\iota _{}𝒪_Y`$ gives the well-known formula for the sheaf Exts, $`\text{xt}^j(\iota _{}𝒪_Y,\iota _{}𝒪_Y)\iota _{}(\mathrm{\Lambda }^j\nu )`$. The assumptions and the spectral sequence $`H^i(\text{xt}^j)\mathrm{Ext}^{i+j}`$ (i.e. the hypercohomology spectral sequence of $`(𝐑\text{om})=\mathrm{Ext}`$) give $`\mathrm{Ext}^r(\iota _{}𝒪_Y,\iota _{}𝒪_Y)=0`$ for $`0<r<n`$. We have $`\mathrm{Hom}(\iota _{}𝒪_Y,\iota _{}𝒪_Y)k`$, hence $`\mathrm{Ext}^n(\iota _{}𝒪_Y,\iota _{}𝒪_Y)k`$ by duality. ∎
###### Example 3.5.
Let $`X`$ be a surface. Then any smooth rational curve $`CX`$ with $`CC=2`$ satisfies the conditions of Lemma 3.4. Now take a chain $`C_1,\mathrm{},C_m`$ of such curves such that $`C_iC_j=\mathrm{}`$ for $`|ij|2`$, and $`C_iC_{i+1}=1`$ for $`i=1,\mathrm{},m1`$. Then $`(𝒪_{C_1},\mathrm{},𝒪_{C_m})`$ is an $`(A_m)`$-configuration of spherical objects.
###### Remark 3.6.
As far as Lemma 3.1 is concerned, one could remove the assumption of smoothness and work with arbitrary projective varieties $`X`$. Serre duality must then be replaced by the general duality theorem \[19, Theorem III.11.1\] applied to the projection $`\pi :X\mathrm{Spec}k`$. This yields a natural isomorphism, for $`𝒢D^{}(X)`$,
$$\mathrm{Ext}^n(𝒢,\omega _X)\mathrm{Ext}^{}(𝒪_X,𝒢)^{},$$
where now $`\omega _X=\pi ^!(𝒪_{\mathrm{Spec}k})D^+(X)`$ is the dualizing complex. With this replacing (3.1) one can essentially repeat the same discussion as in the smooth case, leading to an analogue of Lemma 3.1. The only difference is that the condition that $``$ has a finite locally free resolution must be included as an assumption. We do not pursue this further, for lack of a really relevant application.
### 3b. Two generalisations
We will now look at smooth quasi-projective varieties. Rather than aiming at a comprehensive characterisation of spherical objects, we will just carry over Lemma 3.4 which provides one important source of examples.
Let $`X`$ be a smooth quasi-projective variety of dimension $`n`$, and $`YX`$ a complete subscheme, of codimension $`c`$. $`\iota `$ denotes the embedding $`YX`$. Complete $`X`$ to a projective variety $`\overline{X}`$. Then $`Y\overline{X}`$ is closed, and $`X`$ is smooth, so $`\iota _{}𝒪_Y`$ has a finite locally free resolution; thus we may use Serre duality \[19, Theorem III.11.1\] on $`\overline{X}`$, and the methods of (3.3), to conclude that
$$\mathrm{Hom}(\iota _{}𝒪_Y,)\mathrm{Hom}^n(,\iota _{}𝒪_Y\omega _X)^{},$$
on $`X`$. By continuing as in the projective case, and using the same spectral sequence as in Lemma 3.4, one obtains the following result<sup>2</sup><sup>2</sup>2We thank one of the referees for simplifying our original version of the above proof.:
###### Lemma 3.7.
Assume that $`H^i(Y,\mathrm{\Lambda }^j\nu )=0`$ for all $`0<i+j<n`$, and that $`\iota ^{}\omega _X`$ is trivial. Then $`\iota _{}𝒪_Y`$ is a spherical object in $`D^b(X)`$. ∎
One can now e.g. extend Example 3.5 to quasi-projective surfaces. For subschemes of codimension one, we will later on provide a stronger result, Proposition 3.15, which can be used to construct more interesting spherical objects.
The other generalisation which we want to look at is technically much simpler. Let $`X`$ be a smooth $`n`$-dimensional projective variety over $`k`$ with an action of a finite group $`G`$. We will assume that $`\mathrm{char}(k)=0`$; this implies the complete reducibility of $`G`$-representations, which will be used in an essential way. Let $`Qco_G(X)`$ be the category whose objects are $`G`$-equivariant quasi-coherent sheaves, and whose morphisms are the $`G`$-equivariant sheaf homomorphisms. One can write
(3.5)
$$\mathrm{Hom}_{Qco_G(X)}(_1,_2)=\mathrm{Hom}_{Qco(X)}(_1,_2)^G$$
with respect to the obvious $`G`$-action on $`\mathrm{Hom}_{Qco(X)}(_1,_2)`$. Because taking the invariant part of a $`G`$-vector space is an exact functor, it follows that a $`G`$-sheaf is injective in $`Qco_G(X)`$ iff it is injective in $`Qco(X)`$. This can be used to show that $`Qco_G(X)`$ has enough injectives, and also that $`𝔖=Qco_G(X)`$ and its Serre subcategory $`𝔖^{}=Coh_G(X)`$ of coherent $`G`$-sheaves satisfy the conditions (C1)–(C4) from section 2a. As a further application one derives a formula similar to (3.5) for the derived category:
(3.6)
$$\mathrm{Hom}_{D^+(Qco_G(X))}(_1,_2)=\mathrm{Hom}_{D^+(Qco(X))}(_1,_2)^G$$
for all $`_1,_2D^+(Qco_G(X))`$. This allows one to carry over the usual finiteness results for coherent sheaf cohomology, as well as Serre duality, to the equivariant context. The same argument as in the non-equivariant case now leads to
###### Lemma 3.8.
An object $``$ in the derived category $`D_G^b(X)=D^b(Coh_G(X))`$ of coherent equivariant sheaves is spherical iff the following two conditions are satisfied: $`\mathrm{Hom}_{D_G^b(X)}^j(,)`$ is one-dimensional for $`j=0,n`$ and zero in other degrees; and $`\omega _X`$ is equivariantly isomorphic to $``$. ∎
Finally, one can combine the two generalisations and obtain an equivariant version of Lemma 3.7. This is useful in examples which arise in connection with the McKay correspondence. We will concentrate on the simplest of these examples, which also happens to be particularly relevant for our purpose.
Consider the diagonal subgroup $`G/(m+1)`$ of $`SL_2(k)`$. Write $`R`$ for its regular representation and $`V_1,\mathrm{},V_m`$ for its (nontrivial) irreducible representations. Let $`X`$ be a smooth quasiprojective surface with a complex symplectic form, carrying an effective symplectic action of $`G`$. Choose a fixed point $`xX`$; the tangent space $`T_xX`$ must necessarily be isomorphic to $`R`$ as a $`G`$-vector space. For $`i=1,\mathrm{},m`$ set $`_i=𝒪_xV_iCoh_G(X)`$. The Koszul resolution of $`𝒪_x`$ together with (3.6) shows that
$$\mathrm{Hom}_{D_G^b(X)}^r(_i,_j)(\mathrm{\Lambda }^rRV_i^{}V_j)^G.$$
This implies that each $`_i`$ is a spherical object, and that these objects form an $`(A_m)`$-configuration, so that we obtain a braid group action on $`D_G^b(X)`$.
###### Example 3.9.
In particular, we have a braid group action on the equivariant derived category of coherent sheaves over $`𝔸^2`$, with respect to the obvious linear action of $`G`$ (this is probably the simplest example of a braid group action on a category in the present paper).
Let $`\pi :ZX/G`$ be the minimal resolution. This is again a quasiprojective surface with a symplectic form; it can be constructed as Hilbert scheme of $`G`$-clusters on $`X`$. The irreducible components of $`\pi ^1(x)`$ are smooth rational curves $`C_1,\mathrm{},C_m`$ which are arranged as in Example 3.5, so that their structure sheaves generate a braid group action on $`D^b(Z)`$. A theorem of Kapranov and Vasserot provides an equivalence of categories
(3.7)
$$D_G^b(X)D^b(Z),$$
which takes $`_j`$ to $`𝒪_{C_j}`$ up to tensoring by a line bundle \[23, p. 7\]. This means that the braid group actions on the two categories essentially correspond to each other. Adding the trivial one-dimensional representation $`V_0`$, and the corresponding equivariant sheaf $`_0=𝒪_x=𝒪_xV_0`$, extends the action on $`D_G^b(X)`$ to an action of the affine braid group, except for $`m=1`$. Interestingly, the cyclic symmetry between $`V_0,V_1,\mathrm{},V_m`$ is not immediately visible on $`D^b(Z)`$; the equivalence (3.7) takes $`_0`$ to the structure sheaf of the whole exceptional divisor $`\pi ^1(x)`$. Finally, everything we have said carries over to the other finite subgroups of $`SL(2,k)`$ with the obvious modifications: the Dynkin diagram of type $`(A_m)`$ which occurs implicitly several times in our discussion must be replaced by those of type D/E, and one obtains actions of the corresponding (affine) generalized braid groups.
A recent deep theorem of Bridgeland, King and Reid extends the equivalence (3.7) to certain higher-dimensional quotient singularities. We consider only one very concrete case.
###### Example 3.10.
Let $`X`$ be the Fermat quintic in $`^4`$ with the diagonal action of $`G=(/5)^3`$ familiar from mirror symmetry. The fixed point set $`X^H`$ of the subgroup $`H=(/5)^2\times 1`$ consists of a single $`G`$-orbit $`\mathrm{\Sigma }`$, whose structure sheaf is a spherical object in $`D_G^b(X)`$. By considering other subgroups of the same kind one finds a total of ten spherical objects, with no Homs between any two of them. Now let $`\pi :ZX/G`$ be the crepant resolution given by the Hilbert scheme of $`G`$-clusters. Then $`D_G^b(X)D^b(Z)`$ by so that one gets corresponding spherical objects on $`Z`$. Because of the nature of the equivalence, the object corresponding to $`𝒪_\mathrm{\Sigma }`$ must be supported on the exceptional set $`p^1(\mathrm{\Sigma })`$ of the resolution. We have not determined its precise nature, but this is clearly related to Proposition 3.15 and Examples 3.20 of the next sections.
### 3c. Spherical and exceptional objects
The reader familiar with the theory of exceptional sheaves , or with certain aspects of tilting theory in representation theory, will have noticed a similarity between our twist functors and mutations of exceptional objects. (See also , and note their ‘elliptical exceptional’ objects are examples of $`1`$-spherical objects.) The braid group also occurs in the mutation context, but there it acts on collections of exceptional objects in a triangulated category instead of on the category itself. The relation of the two kinds of braid group actions is not at all clear. We will here content ourselves with two observations, the first of which is motivated by examples in .
###### Definition 3.11.
Let $`X,Y`$ be smooth projective varieties, with $`\omega _X`$ trivial. A morphism $`f:XY`$ (of codimension $`c=dimXdimY`$) is called *simple* if there is an exact triangle
$$𝒪_Y𝐑f_{}𝒪_X\omega _Y[c].$$
In most applications $`Y`$ would be Fano, because one could then use the wealth of known results about exceptional sheaves on such varieties. However, the general theory does not require this assumption on $`Y`$.
###### Lemma 3.12.
Suppose that $`c>0`$ and
$$R^if_{}𝒪_X\{\begin{array}{cc}𝒪_Y\hfill & \text{for }i=0,\hfill \\ 0\hfill & \text{for }0<i<c,\hfill \\ \omega _Y\hfill & \text{for }i=c.\hfill \end{array}$$
Then $`f`$ is simple.
###### Proof.
$`𝐑f_{}𝒪_X`$ is a complex of sheaves whose cohomology is nonzero only in two degrees; a general argument, valid in any derived category, shows that there is an exact triangle $`R^0f_{}𝒪_X𝐑f_{}𝒪_X(R^cf_{}𝒪_X)[c]`$. ∎
###### Proposition 3.13.
Suppose that $`f`$ is simple, and $`D^b(Y)`$ is an exceptional object, in the sense that $`\mathrm{Hom}(,)k`$ and $`\mathrm{Hom}^i(,)=0`$ for all $`i0`$. Then $`𝐋f^{}D^b(X)`$ is a spherical object.
###### Proof.
One can easily show, using e.g. a finite locally free resolution of $``$ and a finite injective quasi-coherent resolution of $`𝒪_X`$, that $`𝐑f_{}𝐋f^{}^𝐋(𝐑f_{}𝒪_X)`$. Hence, by tensoring the triangle in Definition 3.11 with $``$, one obtains another exact triangle $`𝐑f_{}𝐋f^{}\omega _Y[c]`$. This yields a long exact sequence
$$\mathrm{}\mathrm{Hom}^{}(,)\mathrm{Hom}^{}(,𝐑f_{}𝐋f^{})\mathrm{Hom}^c(,\omega _Y)\mathrm{}$$
The second and third group are $`\mathrm{Hom}^{}(𝐋f^{},𝐋f^{})`$ and $`\mathrm{Hom}^{dimX}(,)`$ by, respectively, adjointness and Serre duality. From the assumption that $``$ is exceptional, one now immediately obtains the desired result. ∎
###### Examples 3.14.
* (This assumes $`\text{char}(k)=0`$.) Consider a Calabi-Yau $`X`$ with a fibration $`f:XY`$ over a variety $`Y`$ such that the generic fibres are elliptic curves or $`K3`$ surfaces. Clearly $`f_{}𝒪_X𝒪_Y`$; relative Serre duality shows that $`R^cf_{}𝒪_X\omega _Y`$; and in the $`K3`$ fibred case one has also $`R^1f_{}𝒪_X=0`$. Hence $`f`$ is simple.
* (This assumes $`\text{char}(k)2`$.) Let $`f:XY`$ be a twofold covering branched over a double anticanonical divisor. One can use the $`/2`$-action on $`X`$ to split $`f_{}𝒪_X`$ into two direct summands, which are isomorphic to $`𝒪_Y`$ and $`\omega _Y`$ respectively; this implies that $`f`$ is simple. An example, already considered in , is a $`K3`$ double covering of $`^2`$ branched over a sextic. Another example, which is slightly degenerate but still works, is the unbranched covering map from a $`K3`$ surface to an Enriques surface.
* Examples with $`c=1`$ come from taking $`X`$ to be a smooth anticanonical divisor in $`Y`$, and $`f`$ the embedding. Then $`𝐑f_{}𝒪_X=f_{}𝒪_X\{\omega _Y𝒪_Y\}`$ with the map given by the section of $`\omega _Y^1`$ defining $`X`$. Quartic surfaces in $`^3`$ are an example considered in .
We will now describe a second connection between spherical and exceptional objects, this time using pushforwards instead of pullbacks. The result applies to quasi-projective varieties as well, but it is limited to embeddings of divisors. Let $`X^N`$ be a smooth quasi-projective variety and $`\iota :YX`$ an embedding of a complete connected hypersurface $`Y`$. As in the parallel argument in the previous section, we work on the projective completion $`\overline{X}`$ of $`X`$, in which $`Y`$ is closed. By the smoothness of $`X`$, given $`D^b(Y)`$, $`\iota _{}`$ has a finite locally free resolution, and Serre duality on $`\overline{X}`$ \[19, Theorem III.11.1\] yields
$$\mathrm{Hom}(\iota _{},𝒢)\mathrm{Hom}^{dimX}(𝒢,\iota _{}\omega _X)^{},$$
on $`X`$.
###### Proposition 3.15.
Assume that $`\iota ^{}\omega _X`$ is trivial. If $`D^b(Y)`$ is an exceptional object *with a finite locally free resolution*, then $`\iota _{}`$ is spherical in $`D^b(X)`$.
###### Proof.
In view of the previous discussion, what remains to be done is to compute $`\mathrm{Hom}^i(\iota _{},\iota _{})`$, which by \[19, Theorem III.11.1\] applied to $`\iota _{}`$, is isomorphic to $`\mathrm{Hom}^{i1}(,𝐋\iota ^{}\iota _{}\omega _Y)`$. We will need the following result (which, perhaps surprisingly, need not be true without the $`\iota _{}`$ s).
###### Lemma 3.16.
$`\iota _{}𝐋\iota ^{}(\iota _{})\iota _{}(\omega _Y^1)[1]\iota _{}`$.
###### Proof.
Replacing $`\iota _{}`$ by a quasi-isomorphic complex $`^{}`$ of locally free sheaves, the left hand side of the above equation is $`\iota _{}(^{}|_Y)=^{}𝒪_Y`$ which is quasi-isomorphic to
$$^{}\{𝒪(Y)𝒪\}\iota _{}\{𝒪(Y)𝒪\},$$
where the arrow is multiplication by the canonical section of $`𝒪(Y)`$. Since this vanishes on $`Y`$, which contains the support of $`\iota _{}`$, we obtain $`\iota _{}(𝒪(Y)|_Y)[1]\iota _{}`$ as required. ∎
By hypothesis we may assume that $``$ is a finite complex of locally frees on $`Y`$, so that $`\text{om}(F,F)^{}`$. Thus, computing $`\mathrm{Hom}^i(\iota _{},\iota _{})`$ as the $`(i1)`$th (derived/hyper) sheaf cohomology of the complex of $`𝒪_Y`$-module sheaves $`𝐋\iota ^{}\iota _{}^{}\omega _Y`$, we may push forward to $`X`$ and there use the Lemma above. That is, pick an injective resolution $`𝒪_YI`$ on $`Y`$, so that $`\mathrm{Hom}^i(\iota _{},\iota _{})`$ is the $`(i1)`$th cohomology of
$$\mathrm{\Gamma }_Y(𝐋\iota ^{}\iota _{}^{}\omega _YI),$$
where $`\mathrm{\Gamma }`$ is the global section functor. Pushing forward to $`X`$, this is
$$\mathrm{\Gamma }_X(\iota _{}(𝐋\iota ^{}\iota _{})\iota _{}(()^{}\omega _YI)),$$
which by Lemma 3.16 is
$$\mathrm{\Gamma }_X(\iota _{}(^{}I))[1]\mathrm{\Gamma }_X(\iota _{}(^{}\omega _YI)).$$
This may be brought back onto $`Y`$ to give the $`(i1)`$th cohomology of $`\mathrm{\Gamma }_Y(^{}I)[1]\mathrm{\Gamma }_Y(^{}\omega _YI)`$. This is $`\mathrm{Hom}^i(,)\mathrm{Hom}^{ni}(,)^{}`$, where for the second term we have used Serre duality on $`Y`$. Since $``$ is exceptional this completes the proof. ∎
### 3d. Elliptic curves
The homological mirror conjecture for elliptic curves has been studied extensively by Polishchuk and Zaslow (unfortunately, their formulation of the conjecture differs somewhat from that in section 1b, so that their results cannot be applied directly here). Polishchuk and Orlov , following earlier work of Mukai , have completely determined the automorphism group of the derived category of coherent sheaves. These are difficult results, to which we have little to add. Still, it is maybe instructive to see how things work out in a well-understood case.
We begin with the symplectic side of the story. Let $`(M,\beta )`$ be the torus $`M=/\times /`$ with its standard volume form $`\beta =ds_1ds_2`$. Matters are slightly more complicated than in section 1b, because the fundamental group is nontrivial. In particular, the $`C^{\mathrm{}}`$-topology on $`\mathrm{Symp}(M,\beta )`$ is no longer the correct one; this is due to the fact that Floer cohomology is not invariant under arbitrary isotopies, but only under Hamiltonian ones. There is a bi-invariant foliation $`𝔉`$ of codimension two on $`\mathrm{Symp}(M,\beta )`$, and the Hamiltonian isotopies are precisely those which are tangent to the leaves. To capture this idea one introduces a new topology, the Hamiltonian topology, on $`\mathrm{Symp}(M,\beta )`$. This is the topology generated by the leaves of $`𝔉|U`$, where $`U\mathrm{Symp}(M,\beta )`$ runs over all $`C^{\mathrm{}}`$-open subsets. To avoid confusion, we will write $`\mathrm{Symp}^h(M,\beta )`$ whenever we have the Hamiltonian topology in mind, and call this the Hamiltonian automorphism group; this differs from the terminology in most of the literature where the name is reserved for what, in our terms, is the connected component of the identity in $`\mathrm{Symp}^h(M,\beta )`$. The difference between the two topologies becomes clear if one considers the group $`\mathrm{Aff}(M)=MSL(2,)`$ of oriented affine diffeomorphisms of $`M`$. As a subgroup of $`\mathrm{Symp}(M,\beta )`$ this has its Lie group topology, in which the translation subgroup $`M`$ is connected. In contrast, as a subgroup of $`\mathrm{Symp}^h(M,\beta )`$ it has the discrete topology.
###### Lemma 3.17.
The embedding of $`\mathrm{Aff}(M)`$ into $`\mathrm{Symp}^h(M,\beta )`$ as a discrete subgroup is a homotopy equivalence.
The proof consists of combining the known topology of $`\mathrm{Diff}^+(M)`$, Moser’s theorem that $`\mathrm{Symp}(M,\beta )\mathrm{Diff}^+(M)`$ is a homotopy equivalence, and the flux homomorphism which describes the global structure of the foliation $`𝔉`$. We omit the details.
Let $`\pi :\mathrm{P}^1`$, $`s[\mathrm{cos}(\pi s):\mathrm{sin}(\pi s)]`$ be the universal covering of $`\mathrm{P}^1`$. Consider the subgroup $`\stackrel{~}{SL}(2,)SL(2,)\times \mathrm{Diff}()`$ of pairs $`(g,\stackrel{~}{g})`$ such that $`\stackrel{~}{g}`$ is a lift of the action of $`g`$ on $`\mathrm{P}^1`$. $`\stackrel{~}{SL}(2,)`$ is a central extension of $`SL(2,)`$ by $``$ (topologically, it consists of two copies of the universal cover). We define a graded symplectic automorphism of $`(M,\beta )`$ to be a pair
$$(\varphi ,\stackrel{~}{\varphi })\mathrm{Symp}^h(M,\beta )\times C^{\mathrm{}}(M,\stackrel{~}{SL}(2,))$$
such that $`\stackrel{~}{g}`$ is a lift of $`Dg:MSL(2,)`$; here we have used the standard trivialisation of $`TM`$. The graded symplectic automorphisms form a group under the composition $`(\varphi ,\stackrel{~}{\varphi })(\psi ,\stackrel{~}{\psi })=(\varphi \psi ,(\stackrel{~}{\varphi }\psi )\stackrel{~}{\psi })`$. We denote this group by $`\mathrm{Symp}^{h,gr}(M,\beta )`$, and equip it with the topology induced from $`\mathrm{Symp}^h(M,\beta )\times C^{\mathrm{}}(M,\stackrel{~}{SL}(2,))`$. It is a central extension of $`\mathrm{Symp}^h(M,\beta )`$ by $``$. One can easily verify that the definition is equivalent to that in , which in turn goes back to ideas of Kontsevich .
Even in this simplest example, the construction of the derived Fukaya category $`D^bFuk(M,\beta )`$ has not yet been carried out in detail, so we will proceed on the basis of guesswork in the style of section 1b. The basic objects of $`D^bFuk(M,\beta )`$ are pairs $`(L,E)`$ consisting of a Lagrangian submanifold and a flat unitary bundle on it. Thus, in addition to symplectic automorphisms, the category should admit another group of self-equivalences, which act on all objects $`(L,E)`$ by tensoring $`E`$ with some fixed flat unitary line bundle $`\xi M`$. The two kinds of self-equivalence should give a homomorphism
(3.8)
$$\gamma :G\stackrel{\mathrm{def}}{=}M^{}\pi _0(\mathrm{Symp}^{h,gr}(M,\beta ))\mathrm{Auteq}(D^bFuk(M,\beta )),$$
where $`M^{}=H^1(M;/)`$ is the Jacobian, or dual torus. In order to make the picture more concrete, we will now write down the group $`G`$ explicitly. Take the standard presentation of $`SL(2,)`$ by generators $`g_1=\left(\begin{array}{cc}1& 1\\ 0& 1\end{array}\right)`$, $`g_2=\left(\begin{array}{cc}1& 0\\ 1& 1\end{array}\right)`$ and relations $`g_1g_2g_1=g_2g_1g_2`$, $`(g_1g_2)^6=1`$. Let $`\stackrel{~}{SL}(2,)\stackrel{~}{SL}(2,)`$ be the preimage of $`SL(2,)`$. One can lift $`g_1,g_2`$ to elements $`a_1=(g_1,\stackrel{~}{g}_1)`$ and $`a_2=(g_2,\stackrel{~}{g}_2)`$ in $`\stackrel{~}{SL}(2,)`$ which satisfy $`\stackrel{~}{g}_1(1/2)=1/4`$ and $`\stackrel{~}{g}_2(1/4)=0`$. Together with the central element $`t=(\mathrm{id},ss1)`$ these generate $`\stackrel{~}{SL}(2,)`$, and one can easily work out what the relations are:
$$\stackrel{~}{SL}(2,)=a_1,a_2,t|a_1a_2a_1=a_2a_1a_2,(a_1a_2)^6=t^2,[a_1,t]=[a_2,t]=1.$$
Any element of $`(g,\stackrel{~}{g})\stackrel{~}{SL}(2,)`$ defines a graded symplectic automorphism of $`(M,\beta )`$: one simply takes $`\varphi =g`$ and $`\stackrel{~}{\varphi }`$ to be the constant map with value $`\stackrel{~}{g}`$. Moreover, any translation of $`M`$ has a canonical lift to a graded symplectic automorphism, by taking $`\stackrel{~}{\varphi }`$ to be the constant map with value $`1\stackrel{~}{SL}(2,)`$. These two observations together give a subgroup $`\stackrel{~}{\mathrm{Aff}}(M)=M\stackrel{~}{SL}(2,)`$ of $`\mathrm{Symp}^{h,gr}(M,\beta )`$, which fits into a commutative diagram
Hence, in view of Lemma 3.17, $`\pi _0(\mathrm{Symp}^{h,gr}(M,\beta ))\stackrel{~}{\mathrm{Aff}}(M)`$. After spelling out everything one finds that $`G`$ is the semidirect product $`(/)^4\stackrel{~}{SL}(2,)`$, with respect to the action of $`\stackrel{~}{SL}(2,)`$ on $`^4`$ given by
(3.9)
$$a_1\left(\begin{array}{cccc}1& 1& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 1& 1\end{array}\right),a_2\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& 1& 0& 0\\ 0& 0& 1& 1\\ 0& 0& 0& 1\end{array}\right),t\mathrm{id}.$$
We now pass to the mirror dual side. Let $`X`$ be a smooth elliptic curve over $``$. We choose a point $`x_0X`$ which will be the identity for the group law on $`x`$. The derived category $`D^b(X)`$ has self-equivalences
$$T_𝒪,S\text{and}R_x,L_x,T_{𝒪_x}(xX)$$
defined as follows: $`T_𝒪`$ is the twist by $`𝒪`$, which is spherical for obvious reasons. $`S`$ is the original example of a Fourier-Mukai transform, $`S=\mathrm{\Phi }_{}`$ with $`=𝒪(\mathrm{\Delta }\{x_0\}\times XX\times \{x_0\})`$ the Poincaré line bundle. It maps the structure sheaves of points $`𝒪_x`$ to the line bundles $`𝒪(xx_0)`$, and was shown to be an equivalence by Mukai . $`R_x`$ is the self-equivalence induced by the translation $`yy+x`$; $`L_x`$ is the functor of tensoring with the degree zero line bundle $`𝒪(xx_0)`$; and $`T_{𝒪_x}`$ is the twist along $`𝒪_x`$ which is spherical by Lemma 3.4. These functors have the following properties:
(3.10) $`[L_x,R_y]\mathrm{id}\text{for all }x,y,`$
(3.11) $`T_{𝒪_x}\text{ is isomorphic to }𝒪(x),`$
(3.12) $`S^4[2],`$
(3.13) $`T_{𝒪_{x_0}}T_𝒪T_{𝒪_{x_0}}T_𝒪T_{𝒪_{x_0}}T_𝒪S^1,`$
(3.14) $`T_{𝒪_{x_0}}R_xT_{𝒪_{x_0}}^1R_xL_x^1,`$
(3.15) $`T_{𝒪_{x_0}}L_xT_{𝒪_{x_0}}^1L_x,`$
(3.16) $`T_𝒪R_xT_𝒪^1R_x,`$
(3.17) $`T_𝒪L_xT_𝒪^1R_xL_x.`$
Most of these isomorphisms are easy to prove; those which present any difficulties are (3.11), (3.12), and (3.13). The first and third of these are proved below, and the second one is a consequence of \[36, Theorem 3.13(1)\].
###### Proof of (3.11).
(This argument is valid for the structure sheaf of a point on any algebraic curve.) A simple computation shows that the dual in the derived sense is $`𝒪_x^{}𝒪_x[1]`$. The formula for inverses of FMTs, for which see e.g. \[7, Lemma 4.5\], shows that $`T_{𝒪_x}^1\mathrm{\Phi }_𝒬`$ for some object $`𝒬`$ fitting into an exact triangle
$$𝒬𝒪_\mathrm{\Delta }\stackrel{f}{}𝒪_{(x,x)}.$$
When following through the computation it is not easy to keep track of the map $`f`$, but that is not really necessary. All we need to know is that $`f0`$, which is true because the converse would violate the fact that $`\mathrm{\Phi }_𝒬`$ is an equivalence. Then, since any morphism $`𝒪_\mathrm{\Delta }𝒪_{(x,x)}`$ in the derived category is represented by a genuine map of sheaves, $`f`$ must be some nonzero multiple of the obvious restriction map. It follows $`𝒬`$ is isomorphic to the kernel of $`f`$, which is $`𝒪_\mathrm{\Delta }\pi _1^{}𝒪(x)`$. This means that $`T_{𝒪_x}^1`$ is the functor of tensoring with $`𝒪(x)`$. Passing to inverses yields the desired result. ∎
###### Proof of (3.13).
The equality between the first two terms follows from Theorem 2.17, because $`𝒪_{x_0},𝒪`$ form an $`(A_2)`$-configuration of spherical objects. By the standard formula for the adjoints of a FM transform, the inverse of $`S`$ is the FMT with $`^{}[1]`$. By definition $`T_𝒪`$ is the FMT with $`𝒪(\mathrm{\Delta })[1]`$. Using (3.11) it follows that $`T_{𝒪_{x_0}}T_𝒪T_{𝒪_{x_0}}`$ is the FMT with $`\pi _1^{}𝒪(x_0)𝒪(\mathrm{\Delta })[1]\pi _2^{}𝒪(x_0)^{}[1]`$. ∎
Equations (3.12) and (3.13) show that $`(T_𝒪T_{𝒪_{x_0}})^6[2]`$. Therefore one can define a homomorphism
$$\stackrel{~}{SL}(2,)\mathrm{Auteq}(D^b(X))$$
by mapping the generators $`a_1,a_2,t`$ to $`T_𝒪,T_{𝒪_{x_0}}`$ and the translation $`[1]`$; this already occurs in Mukai’s paper , slightly disguised by the fact that he uses a different presentation of $`SL(2,)`$. The functors $`L_x,R_x`$ yield another homomorphism $`X\times X\mathrm{Auteq}(D^b(X))`$; and one can combine the two constructions into a map
(3.18)
$$\gamma ^{}:G^{}\stackrel{\text{def}}{=}(X\times X)\stackrel{~}{SL}(2,)\mathrm{Auteq}(D^b(X)).$$
Here the semidirect product is taken with respect to the $`\stackrel{~}{SL}(2,)`$-action on $`X\times X`$ indicated by (3.14)–(3.17); explicitly, it is given by the matrices
(3.19)
$$a_1\left(\begin{array}{cc}1& 1\\ 0& 1\end{array}\right),a_2\left(\begin{array}{cc}1& 0\\ 1& 1\end{array}\right),t\mathrm{id}.$$
###### Lemma 3.18.
The group $`G`$ in (3.8) is isomorphic to the group $`G^{}`$ in (3.18).
###### Proof.
Introduce complex coordinates $`z_1=r_1+ir_4`$, $`z_2=r_2ir_3`$ on $`^4/^4`$. Then the action of $`\stackrel{~}{SL}(2,)`$ described in (3.9) becomes $``$-linear, and is given by the same matrices as in (3.19). This is sufficient to identify the two semidirect products which define $`G`$ and $`G^{}`$. We should point out that although the argument is straightforward, the change of coordinates is by no means obvious from the geometric point of view: a look back at the definition of $`G`$ shows that $`z_1,z_2`$ mix genuine symplectic automorphisms with the extra symmetries of $`D^bFuk(M,\beta )`$ which come from tensoring with flat line bundles. ∎
The way in which this fits into the general philosophy is that one expects to have a commutative diagram, with the right vertical arrow given by Kontsevich’s conjecture,
(3.20)
To be accurate, one should adjust the modular parameter of $`X`$ and the volume of $`(M,\beta )`$, eventually introducing a complex part $`\beta _{}`$ as in Remark 1.4, so that they are indeed mirror dual. This has not played any role up to now, since the groups $`G`$ and $`G^{}`$ are independent of the parameters, but it would become important in further study. A theorem of Orlov says that $`\gamma ^{}`$ is always injective, and is an isomorphism iff $`X`$ has no complex multiplication. Only the easy part of the theorem is important for us here: if $`X`$ has complex multiplication then its symmetries induce additional automorphisms of $`D^b(X)`$, which are not contained in the image of $`\gamma ^{}`$. Therefore, if the picture (3.20) is correct, the derived Fukaya category for the corresponding values of $`\beta _{}`$ must admit exotic automorphisms which do not come from symplectic geometry or from flat line bundles. It would be interesting to check this claim, especially because similar phenomena may be expected to occur in higher dimensions.
We will now apply the intuition provided by the general discussion to the specific topic of braid group actions. To a simple closed curve $`S`$ on $`(M,\beta )`$ one can associate a Dehn twist $`\tau _S\mathrm{Symp}^h(M,\beta )`$ which is unique up to Hamiltonian isotopy. This is defined by taking a symplectic embedding $`\iota `$ of $`(U,\theta )=([ϵ;ϵ]\times /,ds_1ds_2)`$ into $`M`$ for some $`ϵ>0`$, with $`\iota (\{0\}\times /)=S`$, and using a local model
$$\tau :UU,\tau (s_1,s_2)=(s_1,s_2h(s_1))$$
where $`hC^{\mathrm{}}(,)`$ is some function with $`h(s)=0`$ for $`sϵ/2`$, $`h(s)=1`$ for $`sϵ/2`$, and $`h(s)+h(s)=1`$ for all $`s`$. The interesting fact is that the Dehn twists along two parallel geodesic lines are not Hamiltonian isotopic: they differ by a translation which depends on the area lying between the two lines. Now take
$$S_1=/\times \{0\},S_2=\{0\}\times /,S_3=/\times \{1/2\}.$$
This is an $`(A_3)`$-configuration of circles. Hence the Dehn twists $`\tau _{S_1},\tau _{S_2},\tau _{S_3}`$ define a homomorphism from the braid group $`B_4`$ to $`\pi _0(\mathrm{Symp}^h(M,\beta ))`$. However, this is not injective: $`\tau _{S_3}^1\tau _{S_1}`$ is Hamiltonian isotopic to a translation which has order two, so that the nontrivial braid $`(g_3^1g_1)^2B_4`$ gets mapped to the identity element. The natural lift of this homomorphism to $`\mathrm{Symp}^{h,gr}(M,\beta )`$ has the same non-injectivity property. Guided by mirror symmetry, one translates this example into algebraic geometry as follows: $`_1=𝒪_{x_0},_2=𝒪,_3=𝒪_xD^b(X)`$, where $`xx_0`$ is a point of order two on $`X`$, form an $`(A_3)`$-configuration of spherical objects. Hence their twist functors generate a weak action of $`B_4`$ on $`D^b(X)`$. By (3.11) $`T__3^1T__1`$ is the functor of tensoring with $`𝒪(xx_0)`$. Since the square of this is the identity functor, we have the same relation as in the symplectic case, so that the action is not faithful.
### 3e. $`K3`$ surfaces
Let $`X`$ be a smooth complex $`K3`$ surface. Consider, as in Example 3.5, a chain of embeddings $`\iota _1,\mathrm{},\iota _m:^1X`$ whose images $`C_i`$ satisfy $`C_iC_j=1`$ for $`|ij|=1`$, and $`C_iC_j=0`$ for $`|ij|2`$. One can then use the structure sheaves $`𝒪_{C_i}`$ to define a braid group action on $`D^b(X)`$. However, this is not the only way:
###### Proposition 3.19.
For each $`i=1,\mathrm{},m`$ choose $`_i`$ to be either $`𝒪(C_i)`$ or else $`𝒪_{C_i}(1):=(\iota _i)_{}𝒪__1(1)`$. Then the $`_i`$ form an $`(A_m)`$-configuration of spherical objects in $`D^b(X)`$, and hence generate a weak braid group action on that category. ∎
The choice can be made for each $`_i`$ independently. These multiple possibilities are relevant from the mirror symmetry point of view. This is explained in , so we will only summarize the discussion here.
Suppose that $`X`$ is elliptically fibred with a section $`S`$. Its mirror should be the symplectic four-manifold $`(M,\beta )`$ with $`M=X`$ and where $`\beta `$ is the real part of some holomorphic two-form on $`X`$ (hyperkähler rotation). The smooth holomorphic curves in $`M`$ are precisely the Lagrangian submanifolds in $`(M,\beta )`$ which are special (with respect to the calibration given by the Kähler form of a Ricci-flat metric on $`X`$). In particular, the curves $`C_i`$ turn into an $`(A_m)`$-configuration of Lagrangian two-spheres; hence the generalized Dehn twists along them generate a homomorphism $`B_{m+1}\pi _0(\mathrm{Symp}^{\mathrm{gr}}(M,\beta ))`$. One can wonder what the corresponding braid group action on $`D^b(X)`$ should be. This question is not really meaningful without a distinguished equivalence between the derived Fukaya category of $`(M,\beta )`$ and that of coherent sheaves on $`X`$, which is not what is predicted by Kontsevich’s conjecture. But if we adopt the Strominger-Yau-Zaslow picture of mirror symmetry, then conjecturally there should be a distinguished full and faithful embedding of triangulated categories
$$D^bFuk(M,\beta )D^b(X)$$
induced by the particular special Lagrangian torus fibration of $`M`$ that comes from the elliptic fibration of $`X`$ (this fibration may, of course, not be distinguished). That this is an embedding, and not an equivalence, is a feature of even-dimensional mirror symmetry. This embedding should be an extension of the Fourier-Mukai transform which takes special Lagrangian submanifolds of $`M`$ (algebraic curves in $`X`$) to coherent sheaves on $`X`$ using the relative Poincaré sheaf on $`X\times _^1X`$ that comes from considering the elliptic fibres to be self-dual using the section; see e.g. .
Assuming this, it now makes sense to ask what spherical objects of $`D^b(X)`$ correspond to the special Lagrangian spheres $`C_1,\mathrm{},C_m`$. The Fourier-Mukai transform takes any special Lagrangian submanifold $`C`$ which is a section of the elliptic fibration to the invertible sheaf $`𝒪(SC)`$; and if $`C`$ lies in a fibre of the fibration, it goes to the structure sheaf $`𝒪_C`$. If we assume that all curves $`C_i`$ fall into one of these two categories, and that $`S`$ intersects all those of them which lie in one fibre, then the Fourier-Mukai transform takes the special Lagrangian submanifolds $`C_1,\mathrm{},C_m`$ in $`(M,\beta )`$ to sheaves $`_1,\mathrm{},_m`$ as in Proposition 3.19, tensored with $`𝒪(S)`$. Then, up to the minor difference of tensoring by $`𝒪(S)`$, one of the braid group actions mentioned in that Proposition would be the correct mirror dual of the symplectic one.
As mentioned in section 3b, such configurations of curves $`C_i`$ are the exceptional loci in the resolution of any algebraic surface with an $`(A_m)`$-singularity. Now, $`(A_m)`$-configurations of Lagrangian two-spheres occur as vanishing cycles in the smoothing of the same singularity. Thus, in a sense, mirror symmetry interchanges smoothings and resolutions. A more striking, though maybe less well understood, instance of this phenomenon is Arnold’s strange duality (see e.g. ), which has been interpreted as a manifestation of mirror symmetry by a number of people (Aspinwall and Morrison, Kobayashi, Dolgachev, Ebeling, etc.). Each of the 14 singular affine surfaces $`S(c_1,c_2,c_3)`$ on Arnold’s list has a natural compactification $`\overline{S}(c_1,c_2,c_3)`$ which has four singular points. One of these points is the original singularity at the origin; the other three are quotient singularities lying on the divisor at infinity, which is a $`^1`$. One can smooth the singular point at the origin; the intersection form of the vanishing cycles obtained in this way is $`T(c_1,c_2,c_3)H`$, where $`T(c_1,c_2,c_3)`$ is the matrix associated to the Dynkin-type diagram
and $`H=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)`$. On the other hand, one can resolve the three singular points at infinity. Inside the resolution, this yields a configuration of smooth rational curves of the form $`T(b_1,b_2,b_3)`$ for certain other numbers $`(b_1,b_2,b_3)`$. One can also do the two things together: this removes all singularities, yielding a smooth $`K3`$ surface $`X(c_1,c_2,c_3)`$ with a splitting of its intersection form as
$$T(c_1,c_2,c_3)HT(b_1,b_2,b_3).$$
Strange duality is the observation that the numbers $`(b_1,b_2,b_3)`$ associated to one singularity on the list occur as $`(c_1,c_2,c_3)`$ for another singularity, and vice versa. Kobayashi (extended by Ebeling to more general singularities) explains this by showing that the $`K3`$s $`X(c_1,c_2,c_3)`$ and $`X(b_1,b_2,b_3)`$ belong to mirror dual families. The associated map on homology interchanges the $`T(c_1,c_2,c_3)`$ and $`T(b_1,b_2,b_3)`$ summands of the intersection form (the extra hyperbolic of vanishing cycles goes to the $`H^0H^4`$ of the other $`K3`$ surface). Thus, the smoothing of the singular point at the origin in $`\overline{S}(c_1,c_2,c_3)`$ corresponds, in a slightly vague sense, to the resolution of the divisor at infinity of $`\overline{S}(b_1,b_2,b_3)`$. From our point of view, since the rational curves at infinity in $`X(b_1,b_2,b_3)`$ can be used to define a braid group action on its derived category, one would like to have a similar configuration of Lagrangian two-spheres (vanishing cycles) in the finite part of $`X(c_1,c_2,c_3)`$. On the level of homology, such a configuration exists of course, but it is apparently unknown whether it can be realized geometrically (recall that Lagrangian submanifolds can have many more non-removable intersection points than their intersection number suggests).
### 3f. Singularities of threefolds
Throughout the following discussion, all varieties will be smooth projective threefolds which are Calabi-Yau in the strict sense (some singular threefolds will also occur, but they will be specifically designated as such). Let $`X`$ be such a variety.
###### Examples 3.20.
Any invertible sheaf on $`X`$ is a spherical object in $`D^b(X)`$. If $`S`$ is a smooth connected surface in $`X`$ with $`H^1(S,𝒪_S)=H^2(S,𝒪_S)=0`$ (e.g. a rational surface or Enriques surface), the structure sheaf $`𝒪_S`$ is a spherical object, by Lemma 3.4. Similarly, for $`C`$ a smooth rational curve in $`X`$ with normal bundle $`\nu _C𝒪_^1(1)𝒪_^1(1)`$ (usually referred to as a $`(1,1)`$-curve), $`𝒪_C`$ is spherical. The ideal sheaf $`𝒥_C`$ of such a curve is also a spherical object; this follows from $`𝒥_C[1]T_𝒪(𝒪_C)`$.
Supposing the ground field to be $`k=`$, we will now return to the conjectural duality between smoothings and resolutions that already played a role in the previous section, and which has been considered by many physicists. (Of course our interest in this is in trying to mirror Dehn twists on smoothings, which arise as monodromy transformations around a degeneration of the smoothing collapsing the appropriate spherical vanishing cycle, by twists on the derived categories of the resolutions.) To explain the approach of physicists (as described in , for instance), it is better to adopt the traditional point of view in which mirror symmetry relates the combined complex and (complexified) Kähler moduli spaces of two varieties, rather than Kontsevich’s conjecture which considers a fixed value of the moduli variables. Then the idea can be phrased like this: moving towards the discriminant locus in the complex moduli space of a variety $`X`$, which means degenerating it to a singular variety $`Y`$, should be mirror dual to going to a ‘boundary wall’ of the complexified Kähler cone of the mirror $`\widehat{X}`$ (the annihilator of a face of the Mori cone), thus inducing an extremal contraction $`\widehat{X}\widehat{Y}`$. A second application of the same idea, with the roles of the mirrors reversed, shows that an arbitrary crepant resolution $`ZY`$ should be mirror dual to a smoothing $`\widehat{Z}`$ of $`\widehat{Y}`$. A case that is reasonably well-understood is that of the ordinary double point (ODP: $`x^2+y^2+z^2+t^2=0`$ in local analytic coordinates) singularity. ODPs should be self-dual, in the sense that if $`Y`$ has $`d`$ distinct ODPs then so does $`\widehat{Y}`$ (this can be checked for Calabi-Yau hypersurfaces in toric varieties, for instance). We now review briefly Clemens’ work on the homology of smoothings and resolutions of such singularities.
A degeneration of $`X`$ to a variety $`Y`$ with $`d`$ ODPs determines $`d`$ vanishing cycles in $`X`$, and hence a map $`v:^dH_3(X)`$. Let $`v^{}:H_3(X)^d`$ be the Poincaré dual of $`v`$. Suppose that $`Y`$ has a crepant resolution $`Z`$ which, in local analytic coordinates near each ODP, looks like the standard small resolution. This means that the exceptional set in $`Z`$ consists of $`d`$ disjoint $`(1,1)`$-curves. By one has
$$H_3(Z)\mathrm{ker}(v^{})/\mathrm{im}(v)$$
and exact sequences
$`H_3(X)\stackrel{v^{}}{}^dH_2(Z)H_2(X)0,`$
$`0H_4(X)H_4(Z)^d\stackrel{v}{}H_3(X).`$
Thus, if there are $`r`$ relations between the vanishing cycles (the image of $`v`$ is of rank $`dr`$), the Betti numbers are
(3.21)
$$\begin{array}{c}\hfill b_2(Z)=b_2(X)+r,b_3(Z)=b_3(X)2(dr),\\ \hfill b_4(Z)=b_4(X)+r.\end{array}$$
Topologically, $`Z`$ arises from $`X`$ through codimension three surgery along the vanishing cycles, and the statements above can be proved e.g. by considering the standard cobordism between them. More intuitively, one can explain matters as follows. Going from $`X`$ to $`Y`$ shrinks the vanishing cycles to points; at the same time, the relations between vanishing cycles are given by four-dimensional chains which become cycles in the limit $`Y`$, because their boundaries shrink to points. This means that we lose $`dr`$ generators of $`H_3`$ and get $`r`$ new generators of $`H_4`$. In $`Z`$, there are $`dr`$ relations between the homology classes of the exceptional $`^1`$s; these relations are pullbacks of closed three-dimensional cycles on $`Y`$ which do not lift to cycles on $`Z`$, so that going from $`Y`$ to $`Z`$ adds $`r`$ new generators to $`H_2`$ while removing another $`dr`$ generators from $`H_3`$. Finally $`H_4(Z)=H_4(Y)`$ for codimension reasons.
Mirror symmetry exchanges odd and even-dimensional homology, so if $`X`$ and $`Z`$ have mirrors $`\widehat{X}`$ and $`\widehat{Z}`$ then
$$\begin{array}{c}\hfill b_2(\widehat{Z})=b_2(\widehat{X})(dr),b_3(\widehat{Z})=b_2(\widehat{X})+2r,\\ \hfill b_4(\widehat{Z})=b_4(\widehat{X})(dr).\end{array}$$
The suggested explanation, in the general framework explained above, is that $`\widehat{Z}`$ should contain $`d`$ vanishing cycles with $`dr`$ relations between them, obtained from a degeneration to a variety $`\widehat{Y}`$ with $`d`$ ODP, and that $`\widehat{X}`$ should be a crepant resolution of $`\widehat{Y}`$. Thus, mirror symmetry exchanges ODPs with the opposite number of relations between their vanishing cycles. Moreover, to the $`d`$ vanishing cycles in the original variety $`X`$ correspond $`d`$ rational $`(1,1)`$-curves in its mirror $`\widehat{X}`$. It seems plausible to think that the structure sheaves or ideal sheaves of these curves (possibly twisted by some line bundle) should be mirror dual to the Lagrangian spheres representing the vanishing cycles in $`X`$; however, as in the $`K3`$ case, such a statement is not really meaningful unless one has chosen some specific equivalence $`D^b(X)D^bFuk(\widehat{X})`$.
###### Remark 3.21.
When $`r=0`$, $`H_2(Y)H_2(X)`$ so that the exceptional cycles in $`Y`$ are homologous to zero. This is not possible if the resolution is algebraic, so we exclude this case, and also the case $`d=r`$ to avoid the same problem on the mirror.
Going a bit beyond this, we will now propose a possible mirror dual to the $`(A_{2d1})`$-singularity. Let $`X`$ be a variety which can be degenerated to some $`Y`$ with an $`(A_{2d1})`$-singularity, and let $`v_1,\mathrm{},v_{2d1}H_n(X)`$ be the corresponding vanishing cycles. The signs will be fixed in such a way that $`v_iv_{i+1}=1`$ for all $`i`$. We impose two additional conditions. One is that $`Y`$ should have a partial smoothing $`Y^{}`$ (equivalently, $`X`$ a partial degeneration) having $`d`$ ODPs, built according to the local model
$$x^2+y^2+z^2+\underset{i=1}{\overset{d}{}}(tϵ_i)^2=0$$
with the $`ϵ_i`$s distinct and small. This means that in the $`(A_{2d1})`$-configuration of vanishing cycles in $`X`$, one can degenerate the 1st, 3rd, …, (2d-1)st to ODPs. The second additional condition is that $`Y^{}`$ should admit a resolution $`Z^{}`$ of the standard kind considered above. Then, according to Remark 3.21, there is at least one relation between $`v_1,v_3,\mathrm{},v_{2d1}`$. In fact, since the intersection matrix of all $`v_i`$ has only a one-dimensional nullspace, there must be precisely one relation.
###### Remark 3.22.
This relation is in fact $`v_1+v_3+\mathrm{}+v_{2d1}=0`$. The corresponding situation on $`Z^{}`$ is that all the exceptional $`^1`$s are homologous. This should not be too surprising: they can be moved back together to give the $`d`$-times thickened $`^1`$ in the resolution of the original $`A_{2d1}`$-singularity that one gets by taking the $`d`$-fold branch cover $`tt^d`$ of the resolution of the ODP $`x^2+y^2+z^2+t^2=0`$. We note in passing that out of the $`2^d`$ possible ways of resolving the ODPs in $`Y^{}`$ (differing by flops) at most two can lead to an algebraic manifold, since an exceptional $`^1`$ cannot be homologous to minus another one.
In view of our previous discussion, we expect that the mirror $`\widehat{X}`$ of $`X`$ admits a contraction $`\widehat{X}\widehat{Y}^{}`$ to a variety with $`d`$ ODPs; any smoothing $`\widehat{Z}^{}`$ of $`\widehat{Y}^{}`$ should contain $`d`$ vanishing cycles with $`(d1)`$ relations between them. By (3.21) these give rise to a $`(d1)`$-dimensional subspace of $`H_4(\widehat{X};)H^{1,1}(\widehat{X})`$. There is a natural basis for the relations between the exceptional $`^1`$s in $`Z^{}`$, which comes from the even-numbered vanishing cycles $`v_{2i}`$. The corresponding basis of the subspace of $`H^{1,1}(\widehat{X})`$ can be represented by divisors $`S_2,S_4,\mathrm{},S_{2d2}`$ such that $`S_{2i}`$ intersects only the $`(i1)`$-th and $`i`$-th exceptional $`^1`$. Based on these considerations and others described below, we make a concrete guess as to what $`\widehat{X}`$ looks like:
###### Definition 3.23.
An $`(\widehat{A}_{2d1})`$-configuration of subvarieties inside a smooth threefold consists of embedded smooth surfaces $`S_2`$, $`S_4`$,…, $`S_{2d2}`$ and curves $`C_1`$, $`C_3`$,…, $`C_{2d1}`$ such that
1. the canonical sheaf of the threefold is trivial along each $`S_{2i}`$;
2. each $`S_{2i}`$ is isomorphic to $`^2`$ with two points blown up;
3. $`S_{2i}S_{2j}=\mathrm{}`$ for $`|ij|2`$;
4. $`S_{2i2},S_{2i}`$ are transverse and intersect in $`C_{2i1}`$, which is a rational curve and exceptional (i.e. has self-intersection $`1`$) both in $`S_{2i2}`$ and $`S_{2i}`$.
Note that the last condition implies that $`C_{2i1}`$ is a $`(1,1)`$-curve in the threefold. What we postulate is that the mirror $`\widehat{X}`$ contains such a configuration of subvarieties, with the $`C_{2i1}`$ being the exceptional set of the contraction $`\widehat{X}\widehat{Y}^{}`$. Apart from being compatible with the informal discussion above, there are some more feasibility arguments in favour of this proposal. Firstly, such configurations exist as exceptional loci in crepant resolutions of singularities: Figure 2 represents a toric 3-fold with trivial canonical bundle containing such a configuration. The thick lines represent the $`C_{2i1}`$s joining consecutive surfaces $`S_{2i2},S_{2i}`$, which are themselves represented by the nodes. Removing these nodes and lines gives the singularity of which it is a resolution by collapsing the whole chain of surfaces and lines; this singularity we think of as the dual of the $`(A_{2d1})`$-singularity.
We could have deformed the $`(A_{2d1})`$-singularity in $`X`$ differently, for instance by degenerating an even-numbered vanishing cycle $`v_{2i}`$ to an ODP. This should correspond to contracting a $`^1`$ in $`\widehat{X}`$. Assuming that our guess is right, so that $`\widehat{X}`$ contains a $`(\widehat{A}_{2d1})`$-configuration, this should be the other exceptional curve in the $`S_{2i}`$ besides $`C_{2i1}`$ and $`C_{2i+1}`$ (i.e. the line which we will call $`C_{2i}^1`$ joining $`C_{2i1}`$ and $`C_{2i+1}`$; in Figure 2 these are represented by the vertical lines). Contracting these curves while not contracting $`C_{2i1},C_{2i+1}`$ turns $`S_{2i}`$ into a $`^1\times ^1`$. The whole 4-cycle $`S_{2i}`$ contracts to a lower dimensional cycle only when we contract another of the $`^1`$ s in it, leaving the final $`^1`$ (over which the surface fibres) still uncontracted (on $`X`$, this corresponds to degenerating two consecutive vanishing cycles while leaving the others finite). Thus, there are contractions of $`\widehat{X}`$ mirroring various possible partial degenerations of $`X`$.
A final argument in favour of our proposal, and much of the motivation for it, is that it leads to braid group actions on derived categories of coherent sheaves. These are of interest in themselves, independently of whether or not they can be considered to be mirror dual to the braid groups of Dehn twist symplectomorphisms on smoothings of $`(A_{2d1})`$-singularities.
###### Proposition 3.24.
Let $`X`$ be a smooth quasiprojective threefold, and $`S_2`$, $`S_4`$,…,$`S_{2d2}`$, $`C_1`$, $`C_3`$,…,$`C_{2d1}`$ an $`(\widehat{A}_{2d1})`$-configuration of subvarieties in $`X`$. Then taking $`_i=𝒪_{C_i}`$ if $`i`$ is odd, or $`𝒪_{S_i}`$ if $`i`$ is even, gives an $`(A_{2d1})`$-configuration $`(_1,_2,\mathrm{},_{2d1})`$ of spherical objects in $`D^b(X)`$. ∎
The assumption that the $`S_i`$ are $`^2`$s with two points blown up can be weakened considerably for this result to hold; any other rational surface will do. Proposition 3.24 is a three-dimensional analogue of Example 3.5 and hence, as a comparison with our discussion of $`K3`$ surfaces shows, possibly too naive from the mirror symmetry point of view. There is an alternative way of constructing spherical objects, closer to Proposition 3.19.
###### Proposition 3.25.
Let $`X`$ be a smooth projective threefold which is Calabi-Yau in the strict sense, containing an $`(\widehat{A}_{2d1})`$-configuration as in the previous Proposition. Take rational curves $`L_{2i}`$ in $`S_{2i}`$ such that $`L_{2i}C_{2j+1}=\mathrm{}`$ for all $`i,j`$ (the inverse image of the generic line in $`^2`$ is such a rational curve in the blow-up of $`^2`$). Choose
$$_i=\{\begin{array}{cc}𝒪_{C_i}(1)\text{ or }𝒥_{C_i}\hfill & \text{if }i\text{ is odd,}\hfill \\ 𝒪_{S_i}(L_i)\text{ or }𝒪_X(S_i)\hfill & \text{if }i\text{ is even.}\hfill \end{array}$$
Then the $`_i`$, $`i=1,2,\mathrm{},2d1`$, form an $`(A_{2d1})`$-configuration of spherical objects in $`D^b(X)`$. ∎
Here $`𝒪_{C_i}(1)`$ is shorthand for $`\iota _{}(𝒪_^1(1))`$ where $`\iota :^1X`$ is some embedding with image $`C_i`$, and $`𝒪_{S_i}(L_i)`$ should be interpreted in the same way. As in Lemma 3.19, the choice of $`_i`$ can be made independently for each $`i`$.
There are many other interesting configurations of spherical objects which arise in connection with threefold singularities. Their mirror symmetry interpretations are mostly unclear. For instance, a slight variation of the situation above yields braid group actions built only from structure sheaves of surfaces:
###### Proposition 3.26.
Let $`X`$ be a smooth quasiprojective threefold, and $`S_1`$, $`S_2`$, …, $`S_m`$ a chain of smooth embedded rational surfaces in $`X`$ with the following properties: $`S_iS_{i+1}`$ is transverse and consists of one rational curve, whose self-intersection in $`S_i`$ and $`S_{i+1}`$ is either zero or $`2`$; $`S_iS_j=\mathrm{}`$ for $`|ij|2`$; and $`\omega _X|S_i`$ is trivial. Then $`_i=𝒪_{S_i}`$ is an $`(A_m)`$-configuration of spherical objects in $`D^b(X)`$. ∎
The conditions actually imply that every intersection $`S_iS_{i+1}`$ is a rational curve with normal bundle $`𝒪_^1𝒪_^1(2)`$ in $`X`$. Note also that the presence of rational curves with self-intersection zero forces at least every second of the surfaces $`S_i`$ to be fibred over $`^1`$. Configurations of this kind are the exceptional loci of crepant resolutions of suitable toric singularities.
In a different direction, Nakamura’s resolutions of abelian quotient singularities using Hilbert schemes of clusters, with their toric representations as tessellations of hexagons , lead to situations similar to Proposition 3.24. The nodes of the hexagons in Figure 3 represent surfaces that are the blow-ups of $`^1\times ^1`$ in two distinct points; the six lines emanating from a node represent the six exceptional $`^1`$s in the surface, in which it intersects the other surfaces represented by the other nodes that the lines join.
The structure sheaves of these curves and surfaces give rise to twists on the derived category satisfying braid relations according to the Dynkin-type diagram obtained by adding a vertex in the middle of each edge (these added vertices represent the structure sheaves of the curves – see Figure 3). The McKay correspondence (see section 3b) translates this into a group of twists on the equivariant derived category of the threefold on which the finite group acted.
## 4. Faithfulness
### 4a. Differential graded algebras and modules
The notions discussed in this section are, for the most part, familiar ones; we collect them here to set up the terminology, and also for the reader’s convenience. A detailed exposition of the general theory of differential graded modules can be found in \[1, section 10\].
Fix a field $`k`$ and an integer $`m1`$. Take the semisimple $`k`$-algebra $`R=k^m`$ with generators $`e_1,\mathrm{},e_m`$ and relations $`e_i^2=e_i`$ for all $`i`$, $`e_ie_j=0`$ for $`ij`$ (so $`1=e_1+\mathrm{}+e_m`$ is the unit element). $`R`$ will play the role of ground ring in the following considerations. In particular, by a graded algebra we will always mean a $``$-graded unital associative $`k`$-algebra $`A`$, together with a homomorphism (of algebras, and unital) $`\iota _A:RA^0`$. This equips $`A`$ with the structure of a graded $`R`$-bimodule, and the multiplication becomes a bimodule map. For the sake of brevity, we will denote the bimodule structure by $`e_ia`$ and $`ae_i`$ ($`aA`$) instead of $`\iota _A(e_i)a`$ resp. $`a\iota _A(e_i)`$. All homomorphisms $`AB`$ between graded algebras will be required to commute with the maps $`\iota _A,\iota _B`$. A differential graded algebra (dga) $`𝒜=(A,d_A)`$ is a graded algebra $`A`$ together with a derivation $`d_A`$ of degree one, which satisfies $`d_A^2=0`$ and $`d_A\iota _A=0`$. The cohomology $`H(𝒜)`$ of a dga is a graded algebra. A homomorphism of dgas is called a quasi-isomorphism if it induces an isomorphism on cohomology. Two dgas $`𝒜,`$ are called quasi-isomorphic if there is a chain of dgas and quasi-isomorphisms $`𝒜𝒞_1\mathrm{}𝒞_k`$ connecting them (in fact it is sufficient to allow $`k=1`$, since the category of dgas admits a calculus of fractions \[25, Lemma 3.2\]). A dga $`𝒜`$ is called formal if it is quasi-isomorphic to its own cohomology algebra $`H(𝒜)`$, thought of as a dga with zero differential.
By a graded module over a graded algebra $`A`$ we will always mean a graded right $`A`$-module. Through the map $`\iota _A`$, any such module $`M`$ becomes a right $`R`$-module: again, we will write $`xe_i`$ ($`xM`$) instead of $`x\iota _A(e_i)`$. A differential graded module (dgm) over a dga $`𝒜=(A,d_A)`$ is a pair $`=(M,d_M)`$ consisting of a graded $`A`$-module $`M`$ and a $`k`$-linear map $`d_M:MM`$ of degree one, such that $`d_M^2=0`$ and $`d_M(xa)=(d_Mx)a+(1)^{\mathrm{deg}(x)}x(d_Aa)`$ for $`aA`$. The cohomology $`H()`$ is a graded module over $`H(𝒜)`$. For instance, $`𝒜`$ is a dgm over itself, and as such it splits into a direct sum of dgms
(4.1)
$$𝒫_i=(e_iA,d_A|_{e_iA}),1im.$$
By definition, a dgm homomorphism $`MN`$ is a homomorphism of graded modules which is at the same time a homomorphism of chain complexes. Dgms over $`𝒜`$ and their homomorphisms form an abelian category $`Dgm(𝒜)`$. One can also define chain homotopies between dgm homomorphisms. The category $`K(𝒜)`$ with the same objects as $`Dgm(𝒜)`$ and with the homotopy classes of dgm homomorphisms as morphisms, is triangulated. The translation functor in it takes $`=(M,d_M)`$ to $`[1]=(M[1],d_M)`$, with no change of sign in the module structure. Exact triangles are all those isomorphic to one of the standard triangles involving a dgm homomorphism and its cone.
Having mentioned cones, we use the opportunity to introduce a slight generalisation, which will be used later on. Assume that one has a chain complex in $`Dgm(𝒜)`$, namely dgms $`𝒞_i`$, $`i`$, and dgm homomorphisms $`\delta _i:𝒞_i𝒞_{i+1}`$ such that $`\delta _{i+1}\delta _i=0`$. Then one can form a new dgm $`𝒞`$ by setting $`C=_iC_i[i]`$ and
$$d_C=\left(\begin{array}{cccc}\mathrm{}& & & \\ \delta _{i1}& (1)^id_{C_i}& & \\ & \delta _i& (1)^{i+1}d_{C_{i+1}}& \\ & & \delta _{i+1}& \mathrm{}\end{array}\right).$$
We refer to this as collapsing the chain complex (it can also be viewed as a special case of a twisted complex, see e.g. ), and write $`𝒞=\{\mathrm{}𝒞_i𝒞_{i+1}\mathrm{}\}`$; for complexes of length two, it specializes to the cone of a dgm homomorphism.
Inverting the dgm quasi-isomorphisms in $`K(𝒜)`$ yields another triangulated category $`D(𝒜)`$, in which any short exact sequence of dgms can be completed to an exact triangle. As usual, $`D(𝒜)`$ can also be defined by inverting the quasi-isomorphisms directly in $`Dgm(𝒜)`$, but then the triangulated structure is more difficult to see. We call $`D(𝒜)`$ the derived category of dgms over $`𝒜`$.
###### Warning.
Even though we use the same notation as in ordinary homological algebra, the expressions $`K(𝒜)`$ and $`D(𝒜)`$ have a different meaning here. In particular $`D(𝒜)`$ is not the derived category, in the usual sense, of $`Dgm(𝒜)`$.
For any dga homomorphism $`f:𝒜`$ there is a ‘restriction of scalars’ functor $`Dgm()Dgm(𝒜)`$. This preserves homotopy classes of homomorphisms, takes cones to cones, and commutes with the shift functors. Hence it descends to an exact functor $`K()K(𝒜)`$. Moreover, it obviously preserves quasi-isomorphisms, so that it also descends to an exact functor $`D()D(𝒜)`$; we will denote any of these functors by $`f^{}`$. The next result, taken from \[1, Theorem 10.12.5.1\], shows that two quasi-isomorphic dgas have equivalent derived categories.
###### Theorem 4.1.
If $`f`$ is a quasi-isomorphism, $`f^{}:D()D(𝒜)`$ is an exact equivalence.
Let $`𝒜`$ be a dga. The standard twist functors $`t_1,\mathrm{},t_m`$ from $`Dgm(𝒜)`$ to itself are defined by
$$t_i()=\{e_i_k𝒫_i\}.$$
The tensor product of $`e_i=(Me_i,d_M|_{Me_i})`$ with the dgm $`𝒫_i`$ of (4.1) is one of complexes of $`k`$-vector spaces; it becomes a dgm with the module structure inherited from $`𝒫_i`$. The arrow is the multiplication map $`Me_i_ke_iAM`$, which is a homomorphism of dgms, and we are taking its cone. $`t_i`$ descends to exact functors $`K(𝒜)K(𝒜)`$ and $`D(𝒜)D(𝒜)`$, for which we will use the same notation. This is straightforward for $`K(𝒜)`$. As for $`D(𝒜)`$, one needs to show that $`t_i`$ preserves quasi-isomorphisms; this follows from looking at the long exact sequence
$$\mathrm{}H()e_i_ke_iH(𝒜)H()H(t_i())\mathrm{}$$
###### Lemma 4.2.
Let $`f:𝒜`$ be a quasi-isomorphism of dgas. Then the following diagram commutes up to isomorphism, for each $`1im`$:
###### Proof.
Let $`=(M,d_M)`$ be a dgm over $``$. Consider the commutative diagram of dgms over $`𝒜`$
The upper horizontal arrow is $`mamf(a)`$, and the lower one is multiplication. The cone of the upper row is $`t_i(f^{}())`$, while that of the lower one is $`f^{}(t_i())`$. The two vertical arrows combine to give a quasi-isomorphism between these cones. ∎
Now let $`𝔖^{}𝔖`$ be as in section 2a, and $`𝔎`$ the category from Definition 2.2. Let $`E_1,\mathrm{},E_m`$ be objects of $`𝔎`$, and $`E`$ their direct sum. The chain complex of endomorphisms
$$end(E):=hom(E,E)=\underset{1i,jm}{}hom(E_i,E_j)$$
has a natural structure of a dga. Multiplication is given by composition of homomorphisms. $`\iota _{end(E)}`$ maps $`e_iR`$ to $`\mathrm{id}_{E_i}hom(E_i,E_i)`$, so that left multiplication with $`e_i`$ is the projection to $`hom(E,E_i)`$ while right multiplication is the projection to $`hom(E_i,E)`$. In the same way, for any $`F𝔎`$, the complex $`hom(E,F)`$ is a dgm over $`end(E)`$. The functor $`hom(E,):𝔎K(end(E))`$ defined in this way is exact, because it carries cones to cones. The objects $`E_i`$ get mapped to the dgms $`hom(E,E_i)=e_iend(E)`$, which are precisely the $`𝒫_i`$ from (4.1). We define a functor $`\mathrm{\Psi }_E`$ to be the composition
$$𝔎\stackrel{hom(E,)}{}K(end(E))\stackrel{\text{quotient functor}}{}D(end(E)).$$
###### Lemma 4.3.
Assume that $`E_1,\mathrm{},E_m`$ satisfy the conditions from Definition 2.5, so that the twist functors $`T_{E_i}`$ are defined. Then the following diagram is commutative up to isomorphism, for each $`1im`$:
###### Proof.
For $`F𝔎`$, consider the commutative diagram of dgms over $`end(E)`$
with the following maps: the horizontal arrow in the first row is the composition, that in the second row is induced by the evaluation map $`hom(E_i,F)_kE_iF`$. The left hand vertical arrow is the first of the canonical maps from (2.1), which is a quasi-isomorphism since $`hom(E_i,F)`$ has finite-dimensional cohomology. The cone of the first row is $`t_i(\mathrm{\Psi }_E(F))`$, while that of the second row is $`\mathrm{\Psi }_E(T_{E_i}(F))`$. The vertical arrows combine to give a natural quasi-isomorphism between these cones. ∎
Later on, in our application, the $`E_i`$ occur as resolutions of objects in $`D^b(𝔖^{})`$. The next two Lemmas address the question of how the choice of resolutions affects the construction. This is not strictly necessary for our purpose, but it rounds off the picture.
###### Lemma 4.4.
Let $`E_i,E_i^{}`$ ($`1im`$) be objects in $`𝔎`$ such that $`E_iE_i^{}`$ for all $`i`$. Then the dgas $`end(E)`$ and $`end(E^{})`$ are quasi-isomorphic.
###### Proof.
Choose for each $`i`$ a map $`g_i:E_iE_i^{}`$ which is a chain homotopy equivalence. Set $`C_i=\mathrm{Cone}(g_i)`$, and let $`C`$ be the direct sum of these cones; this is the same as the cone of $`g=g_1\mathrm{}g_m`$. Let $`end(C)`$ be the endomorphism dga of $`C_1,\mathrm{},C_m`$. An element of $`end(C)`$ of degree $`r`$ is a matrix
$$\varphi =\left(\begin{array}{cc}\varphi _{11}& \varphi _{12}\\ \varphi _{21}& \varphi _{22}\end{array}\right)$$
with $`\varphi _{11}hom^r(E,E)`$, $`\varphi _{21}hom^{r1}(E,E^{})`$, $`\varphi _{12}hom^{r+1}(E^{},E)`$, $`\varphi _{22}hom^r(E^{},E^{})`$. The differential in $`end(C)`$ maps $`\varphi `$ to
$$\left(\begin{array}{cc}d_E\varphi _{11}+(1)^r\varphi _{11}d_E(1)^r\varphi _{12}g& d_E\varphi _{12}(1)^r\varphi _{12}d_E\\ g\varphi _{11}(1)^r\varphi _{22}g+d_E^{}\varphi _{21}+(1)^r\varphi _{21}d_E& g\varphi _{12}+d_E^{}\varphi _{22}(1)^r\varphi _{22}d_E^{}\end{array}\right)$$
Let $`𝒞end(C)`$ be the subalgebra of matrices which are lower-triangular ($`\varphi _{12}=0`$). The formula above shows that this is closed under the differential, and hence again a dga. The projection $`\pi _2:𝒞end(E^{})`$, $`\pi _2(\varphi )=\varphi _{22}`$, is a homomorphism of dgas. Its kernel is isomorphic (as a complex of $`k`$-vector spaces, and up to a shift) to the cone of the composition with $`g`$ map
$$g:hom(E,E)hom(E,E^{}).$$
Since $`g`$ is a homotopy equivalence this cone is acyclic, so that $`\pi _2`$ is a quasi-isomorphism of dgas. A similar argument shows that the projection $`\pi _1:𝒞end(E)`$, $`\pi _1(\varphi )=(1)^{\mathrm{deg}(\varphi )}\varphi _{11}`$, is a quasi-isomorphism of dgas. The two maps together prove that $`end(E)`$ and $`end(E^{})`$ are quasi-isomorphic. ∎
As a consequence of this and Theorem 4.1, the categories $`D(end(E))`$ and $`D(end(E^{}))`$ are equivalent. Actually, we have shown a more precise statement: any choice of $`g_i:E_iE_i^{}`$ yields, up to isomorphism of functors, an exact equivalence $`(\pi _2^{})^1\pi _1^{}:D(end(E))D(end(E^{}))`$. We will now see that this equivalence is compatible with the functors $`\mathrm{\Psi }_E,\mathrm{\Psi }_E^{}`$.
###### Lemma 4.5.
In the situation of Lemma 4.4, $`(\pi _2^{})^1\pi _1^{}\mathrm{\Psi }_E\mathrm{\Psi }_E^{}`$.
###### Proof.
The obvious short exact sequence $`0E^{}CE[1]0`$ induces, for any $`F𝔎`$, a short exact sequence of dgms over $`𝒞`$
$$0\pi _1^{}hom(E,F)[1]hom(C,F)\pi _2^{}hom(E^{},F)0.$$
In the derived category $`D(𝒞)`$, this short exact sequence can be completed to an exact triangle by a morphism
(4.2)
$$\pi _2^{}hom(E^{},F)\pi _1^{}hom(E,F).$$
One can define such a morphism explicitly by replacing the given sequence with a (canonically constructed) quasi-isomorphic one, for which the corresponding morphism can be realized by an actual homomorphism of dgms; compare \[14, Proposition III.3.5\]. The advantage of this explicit construction is that (4.2) is now natural in $`F`$. Since $`C`$ is a contractible complex, $`hom(C,F)`$ is acyclic, which implies that (4.2) is an isomorphism in $`D(𝒞)`$ for any $`F`$. This shows that the diagram
commutes up to isomorphism, as desired. ∎
### 4b. Intrinsic formality
Applications of dg methods to homological algebra often hinge on constructing a chain of quasi-isomorphisms connecting two given dgas. For instance, in the situation explained in the previous section, one can try to use the dga $`end(E)`$ to study the twists $`T_{E_i}`$ via the functor $`\mathrm{\Psi }_E`$. What really matters for this purpose is only the quasi-isomorphism type of $`end(E)`$. In general, quasi-isomorphism type is a rather subtle invariant. However, there are some cases where the cohomology already determines the quasi-isomorphism type.
###### Definition 4.6.
A graded algebra $`A`$ is called intrinsically formal if any two dgas with cohomology $`A`$ are quasi-isomorphic; or equivalently, if any dga $``$ with $`H()A`$ is formal.
For instance, one can show easily that any graded algebra $`A`$ concentrated in degree zero is intrinsically formal (this particular example can be viewed as the starting point for Rickard’s theory of derived Morita equivalences , as recast in dga language by Keller ). However, our intended application is to algebras of a rather different kind.
An augmented graded algebra is a graded algebra $`A`$ together with a graded algebra homomorphism $`ϵ_A:AR`$ which satisfies $`ϵ_A\iota _A=\mathrm{id}_R`$. Its kernel is a two-sided ideal, called the augmentation ideal; we write it as $`A^+A`$. A special case is when $`A`$ is connected, which means $`A^i=0`$ for $`i<0`$ and $`\iota _A:RA^0`$ is an isomorphism; then there is of course a unique augmentation map, and $`A^+`$ is the subspace of elements of positive degree.
###### Theorem 4.7.
Let $`A`$ be an augmented graded algebra. If $`HH^q(A,A[2q])=0`$ for all $`q>2`$, then $`A`$ is intrinsically formal.
We remind the reader that the Hochschild cohomology $`HH^{}(A,M)`$ of a graded $`A`$-bimodule $`M`$ is the cohomology of the cochain complex
$`C^q(A,M)=\mathrm{Hom}_{RR}(\stackrel{q}{\stackrel{}{A^+_R\mathrm{}_RA^+}},M),`$
$`(^q\varphi )(a_1,\mathrm{},a_{q+1})=(1)^ϵa_1\varphi (a_2,\mathrm{},a_{q+1})+`$
$`+_{i=1}^q(1)^{ϵ_i}\varphi (a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_{q+1})(1)^{ϵ_q}\varphi (a_1,\mathrm{},a_q)a_{q+1},`$
where $`\mathrm{Hom}_{RR}`$ denotes homomorphisms of graded $`R`$-bimodules (by definition, these are homomorphisms of degree zero). The signs are $`ϵ=q\mathrm{deg}(a_1)`$, $`ϵ_i=\mathrm{deg}(a_1)+\mathrm{}+\mathrm{deg}(a_i)i`$. The bimodules relevant for our application are $`M=A[s]`$ with the left multiplication twisted by a sign: $`axa^{}=(1)^{s\mathrm{deg}(a)}axa^{}`$ for $`a,a^{}A`$ and $`xM`$. Note that the chain complex $`C^{}(A,A[s])`$ depends on $`s`$, so that the cohomology groups which occur in the Theorem above belong to different complexes.
We will give a proof of Theorem 4.7 for lack of an accessible reference, and also because our framework (in which dgas may be nonzero in positive and negative degrees) differs slightly from the usual one. However, the result is by no means new. Originally, the phenomenon of intrinsic formality was discovered by Halperin and Stasheff in the framework of commutative dgas. They constructed a series of obstruction groups, whose vanishing implies intrinsic formality. Later Tanré identified these obstruction groups as Harrison cohomology groups. To the best of our knowledge, the non-commutative version, in which Hochschild cohomology replaces Harrison cohomology, is due to Kadeishvili , who also realized the importance of $`A_{\mathrm{}}`$-algebras in this context. A general survey of $`A_{\mathrm{}}`$-algebras and applications is . It is difficult to find a concrete counterexample, but apparently Theorem 4.7 is not true without the augmentedness assumption. This is related to a fundamental problem, which is that the notion of $`A_{\mathrm{}}`$-algebra with unit is not homotopy invariant (there is no ‘homological perturbation Lemma’ for it).
Let $`A`$ be an augmented graded algebra and $`=(B,d_B)`$ a dga. An $`A_{\mathrm{}}`$-morphism $`\gamma :A`$ is a sequence of maps of graded $`R`$-bimodules $`\gamma _q\mathrm{Hom}_{RR}((A^+)^{_Rq},B[1q])`$, $`q1`$, satisfying the equations
($`E_q`$)
$$\begin{array}{c}\hfill d_B\gamma _q(a_1,\mathrm{},a_q)=_{i=1}^{q1}(1)^{ϵ_i}(\gamma _{q1}(a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_q)\\ \hfill \gamma _i(a_1,\mathrm{},a_i)\gamma _{qi}(a_{i+1},\mathrm{},a_q)).\end{array}$$
The $`ϵ_i`$ are as in the definition of $`HH^{}(A,M)`$ above. The first two of these equations are
($`E_1`$) $`d_B\gamma _1(a_1)=0,`$
($`E_2`$) $`d_B\gamma _2(a_1,a_2)=(1)^{\mathrm{deg}(a_1)1}(\gamma _1(a_1a_2)\gamma _1(a_1)\gamma _1(a_2)).`$
This means that $`\gamma _1`$, which needs not be a homomorphism of algebras, nevertheless induces a multiplicative map $`(\gamma _1)_{}:A^+H()`$. In a sense, the non-multiplicativity of $`\gamma _1`$ is corrected by the higher order maps $`\gamma _q`$, so that $`A_{\mathrm{}}`$-morphisms are ‘approximately multiplicative maps’.
From a more classical point of view, one can see $`A_{\mathrm{}}`$-morphisms simply as a convenient way of encoding dga homomorphisms from a certain large dga canonically associated to $`A`$, a kind of ‘thickening of $`A`$’. Consider $`V=A^+[1]`$ as a graded $`R`$-bimodule, and let $`T^+V=_{q1}V^{_Rq}`$ be its tensor algebra, without unit. We will write $`a_1,\mathrm{},a_qT^+V`$ instead of $`a_1\mathrm{}a_q`$. Now consider $`W=T^+V[1]`$ as a graded $`R`$-bimodule in its own right, and form its tensor algebra with unit $`TW=R_{r1}V^{_Rr}`$. The elements of $`TW`$ (apart from $`RTW`$) are linear combinations of expressions of the form
$$x=a_{11},a_{12},\mathrm{},a_{1,q_1}\mathrm{}a_{r1},\mathrm{},a_{r,q_r}$$
with $`r>0`$, $`q_1,\mathrm{},q_r>0`$, and $`a_{ij}A^+`$. The degree of such an expression is $`\mathrm{deg}_{TW}(x)=_{ij}\mathrm{deg}_A(a_{ij})_iq_i+r`$. One defines a dga $`𝒳=(X,d_X)`$ by taking $`X=TW`$ with the tensor multiplication, and $`d_X`$ to be the derivation which acts on elements of $`W`$ as follows:
$$\begin{array}{c}d_Xa_1,\mathrm{},a_q=_{i=1}^{q1}(1)^{ϵ_i}(a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_q\hfill \\ \hfill a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_q).\end{array}$$
The passage from $`A`$ to $`𝒳`$ is usually written as composition of the bar and cobar functors, which go from augmented dg algebras to dg coalgebras and back, see e.g. . We can now make the above-mentioned connection with $`A_{\mathrm{}}`$-morphisms.
###### Lemma 4.8.
For any $`A_{\mathrm{}}`$-morphism $`\gamma :A`$ one can define a dga homomorphism $`\mathrm{\Gamma }:𝒳`$ by setting $`\mathrm{\Gamma }|R`$ to be the unit map $`\iota _B`$, and $`\mathrm{\Gamma }(a_1,\mathrm{},a_q)=\gamma _q(a_1,\mathrm{},a_q)`$. $`\mathrm{\Gamma }`$ is a quasi-isomorphism iff $`\iota _B\gamma _1`$ induces an isomorphism between $`RA^+A`$ and $`H()`$.
###### Proof.
The first part follows immediately from comparing the equations $`(E_q)`$ with the definition of the differential $`d_X`$. As for the second part, a classical computation due to Moore \[33, Théorème 6.2\] shows that the inclusion $`RA^+\mathrm{ker}d_X`$ induces an isomorphism $`RA^+H(𝒳)`$. This implies the desired result. ∎
As a trivial example, let $`𝒜=(A,0)`$ be the dga given by $`A`$ with zero differential, and take the $`A_{\mathrm{}}`$-morphism $`\gamma :A𝒜`$ given by $`\gamma _1=\mathrm{id}:A^+A`$, $`\gamma _q=0`$ for all $`q2`$. Then Lemma 4.8 shows that the corresponding map $`\mathrm{\Gamma }:𝒳𝒜`$ is a quasi-isomorphism of dgas.
The next Lemma is an instance of ‘homological perturbation theory’, see e.g. . Let $`A`$ be an augmented graded algebra, $``$ be a dga, and $`\varphi :AH()`$ a homomorphism of graded algebras. This makes the cohomology $`H()`$ into a graded $`A`$-bimodule.
###### Lemma 4.9.
Assume that $`HH^q(A,H()[2q])=0`$ for all $`q>2`$. Then there is an $`A_{\mathrm{}}`$-morphism $`\gamma :A`$ such that the induced map $`(\gamma _1)_{}:A^+H()`$ is equal to $`\varphi |A^+`$.
###### Proof.
Choose a map of graded $`R`$-bimodules $`\gamma _1:A^+\mathrm{ker}d_BB`$ which induces $`\varphi |A^+`$. Since $`\gamma _1`$ is multiplicative on cohomology, we can find a map $`\gamma _2`$ such that $`(E_2)`$ is satisfied. From here onwards the construction is inductive. Suppose that $`\gamma _1,\mathrm{},\gamma _{q1}`$, for some $`q3`$, are maps such that $`(E_1),\mathrm{},(E_{q1})`$ hold. Denote the right hand side of equation $`(E_q)`$ for these maps by $`\psi :(A^+)^{_Rq}B[2q]`$. One can compute directly that
(4.3)
$$d_B\psi (a_1,\mathrm{},a_q)=0$$
for all $`a_1,\mathrm{},a_qA^+`$, and that
(4.4)
$$\begin{array}{cc}& \begin{array}{c}\gamma _1(a_1)\psi (a_2,\mathrm{},a_{q+1})+_{i=1}^q(1)^{ϵ_i}\psi (a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_{q+1})\\ (1)^{ϵ_q}\psi (a_1,\mathrm{},a_q)\gamma _1(a_{q+1})=\end{array}\hfill \\ & =d_B\left(_{i=1}^q(1)^{ϵ_i}\gamma _i(a_1,\mathrm{},a_i)\gamma _{q+1i}(a_{i+1},\mathrm{},a_{q+1})\right).\hfill \end{array}$$
By (4.3) $`\psi `$ induces a map $`\overline{\psi }:(A^+)^{_Rq}H()[2q]`$, which is just an element of the Hochschild chain group $`C^q(A,H()[2q])`$. Equation (4.4) says that $`\overline{\psi }`$ is a Hochschild cocycle. By assumption there is an $`\overline{\eta }C^{q1}(A,H()[2q])`$ such that $`^{q1}\overline{\eta }=\overline{\psi }`$. Choose any map of graded $`R`$-bimodules $`\eta :(A^+)^{_Rq1}(\mathrm{ker}d_B)[1q]`$ which induces $`\overline{\eta }`$, and set $`\gamma _{q1}^{\mathrm{new}}=\gamma _{q1}\eta `$. The equations $`(E_1),\mathrm{},(E_{q1})`$ will continue to hold if one replaces $`\gamma _{q1}`$ by $`\gamma _{q1}^{\mathrm{new}}`$. Moreover, if $`\psi ^{\mathrm{new}}`$ denotes the r.h.s. of $`(E_q)`$ after this replacement, one computes that
(4.5)
$$\begin{array}{c}(\psi \psi ^{\mathrm{new}})(a_1,\mathrm{},a_q)=(1)^{\mathrm{deg}(a_1)}\gamma _1(a_1)\eta (a_2,\mathrm{},a_{q1})+\\ +_{i=1}^{q1}(1)^{ϵ_i}\eta (a_1,\mathrm{},a_ia_{i+1},\mathrm{},a_q)(1)^{ϵ_q}\eta (a_1,\mathrm{},a_{q1})\gamma _1(a_q).\end{array}$$
This means that $`\overline{\psi }^{\mathrm{new}}=\overline{\psi }^{q1}\overline{\eta }=0`$. Clearly, the vanishing of $`\overline{\psi }^{\mathrm{new}}`$ ensures that one can extend the sequence $`\gamma _1,\mathrm{},\gamma _{q2},\gamma _{q1}^{\mathrm{new}}`$ by a map $`\gamma _q`$ such that $`(E_q)`$ holds. This completes the induction step.
Note that in the $`q`$-th step, only the $`(q1)`$-st of the given maps $`\gamma _i`$ is changed. Therefore the sequence which we construct does indeed converge to an $`A_{\mathrm{}}`$-morphism $`\gamma `$. ∎
###### Proof of Theorem 4.7.
Let $``$ be a dga whose cohomology algebra is isomorphic to $`A`$. Choose an isomorphism $`\varphi :AH()`$. By Lemma 4.9 there is an $`A_{\mathrm{}}`$-morphism $`\gamma :A`$ such that $`\gamma _1`$ induces $`\varphi |A^+`$. This obviously means that $`(\iota _B\gamma _1)_{}:RA^+H()`$ is an isomorphism. Hence, by Lemma 4.8 the induced map $`\mathrm{\Gamma }:𝒳`$ is a quasi-isomorphism of dgas. We have already seen that there is a quasi-isomorphism $`𝒳𝒜=(A,0)`$. This shows that $``$ is quasi-isomorphic to $`𝒜`$, hence formal. ∎
### 4c. The graded algebras $`A_{m,n}`$
We assume from now on that $`m2`$; this assumption will be retained throughout this section and the following one. In addition, choose an $`n1`$.
Let $`\mathrm{\Gamma }`$ be a quiver (an oriented graph) with vertices numbered $`1,\mathrm{},m`$, and with a ‘degree’ (an integer label) attached to each edge. One can associate to it a graded algebra $`k[\mathrm{\Gamma }]`$, the path algebra, as follows. As a $`k`$-vector space $`k[\mathrm{\Gamma }]`$ is freely generated by the set of all paths (not necessarily closed, of arbitrary length $`0`$) in $`\mathrm{\Gamma }`$. The degree of a path is the sum of all ‘degrees’ of the edges along which it runs. The product of two paths is their composition if the endpoint of the first one coincides with the starting point of the second one, and zero otherwise. The map $`\iota _{k[\mathrm{\Gamma }]}:R(k[\mathrm{\Gamma }])^0`$ maps $`e_i`$ to the path of length zero at the $`i`$-th vertex.
The example we are interested in is the quiver $`\mathrm{\Gamma }_{m,n}`$ shown in Figure 4. Paths of length $`l0`$ in this quiver correspond to $`(l+1)`$-tuples $`(i_0|\mathrm{}|i_l)`$ with $`i_\nu \{1,\mathrm{},m\}`$ and $`|i_{\nu +1}i_\nu |=1`$. The product of two paths in $`k[\mathrm{\Gamma }_{m,n}]`$ is given by $`(i_0|\mathrm{}|i_l)(i_0^{}|\mathrm{}|i_l^{}^{})=(i_0|\mathrm{}|i_l|i_1^{}|\mathrm{}|i_l^{}^{})`$ if $`i_l=i_0^{}`$, or zero otherwise. The grading is $`\mathrm{deg}(i)=0`$, $`\mathrm{deg}(i|i+1)=d_i`$, $`\mathrm{deg}(i+1|i)=nd_i`$, where we set
(4.6)
$$d_i=\{\begin{array}{cc}\frac{1}{2}n\hfill & \text{if }n\text{ is even,}\hfill \\ \frac{1}{2}(n+(1)^i)\hfill & \text{if }n\text{ is odd.}\hfill \end{array}$$
We introduce a two-sided homogeneous ideal $`J_{m,n}k[\mathrm{\Gamma }_{m,n}]`$ as follows. If $`m3`$ then $`J_{m,n}`$ is generated by $`(i|i1|i)(i|i+1|i)`$, $`(i1|i|i+1)`$ and $`(i+1|i|i1)`$ for all $`i=2,\mathrm{},m1`$; in the remaining case $`m=2`$, $`J_{m,n}`$ is generated by $`(1|2|1|2)`$ and $`(2|1|2|1)`$. Now define $`A_{m,n}=k[\mathrm{\Gamma }_{m,n}]/J_{m,n}`$. This is again a graded algebra. It is finite-dimensional over $`k`$; an explicit basis is given by the $`(4m2)`$ elements
(4.7)
$$\{\begin{array}{cc}& (1),\mathrm{},(m),\hfill \\ & (1|2),\mathrm{},(m1|m),\hfill \\ & (2|1),\mathrm{},(m|m1),\hfill \\ & (1|2|1),(2|3|2)=(2|1|2),\mathrm{},(m1|m|m1)=\hfill \\ & =(m1|m2|m1),(m|m1|m).\hfill \end{array}$$
Here we have used the same notation for elements of $`k[\mathrm{\Gamma }_{m,n}]`$ and their images in $`A_{m,n}`$. We will continue to do so in the future, in particular $`(i|i\pm 1|i)`$ will be used to denote the image of both $`(i|i+1|i)`$ and $`(i|i1|i)`$ in $`A_{m,n}`$.
We will now explain why these algebras are relevant to our problem. Let $`𝔎`$ be a category as in Definition 2.2 and $`E_1,\mathrm{},E_m𝔎`$ an $`(A_m)`$-configuration of $`n`$-spherical objects.
###### Lemma 4.10.
Suppose that for each $`i=1,\mathrm{},m1`$ the one-dimensional space $`\mathrm{Hom}^{}(E_{i+1},E_i)`$ is concentrated in degree $`d_i`$. Then the cohomology algebra of the dga $`end(E)`$ is isomorphic to $`A_{m,n}`$.
We should say that the assumption on $`\mathrm{Hom}^{}(E_{i+1},E_i)`$ is not really restrictive since, given an arbitrary $`(A_m)`$-configuration, it can always be achieved by shifting each $`E_i`$ suitably.
###### Proof.
Since each $`E_i`$ is $`n`$-spherical, the pairings
(4.8) $`\mathrm{Hom}^{}(E_{i+1},E_i)\mathrm{Hom}^{}(E_i,E_{i+1})`$ $`\mathrm{Hom}^n(E_i,E_i)k,`$
$`\mathrm{Hom}^{}(E_i,E_{i+1})\mathrm{Hom}^{}(E_{i+1},E_i)`$ $`\mathrm{Hom}^n(E_{i+1},E_{i+1})k`$
are nondegenerate for $`i=1,\mathrm{},m1`$. Hence $`\mathrm{Hom}^{}(E_i,E_{i+1})k`$ is concentrated in degree $`nd_i`$. Choose nonzero elements $`\alpha _i\mathrm{Hom}^{}(E_{i+1},E_i)`$ and $`\beta _i\mathrm{Hom}^{}(E_i,E_{i+1})`$. Then, again because of the nondegeneracy of (4.8), one has
(4.9)
$$\alpha _i\beta _i=c_i(\beta _{i1}\alpha _{i1})$$
in $`\mathrm{Hom}^{}(E,E)`$ for some nonzero constants $`c_2,\mathrm{},c_{m1}k`$. Without changing notation, we multiply each $`\beta _i`$ with $`c_2c_3\mathrm{}c_i`$; then the same equations (4.9) hold with all $`c_i`$ equal to $`1`$. Since $`\mathrm{Hom}^{}(E_i,E_j)=0`$ for all $`|ij|2`$, we also have $`\beta _i\beta _{i1}=0`$, $`\alpha _{i1}\alpha _i=0`$ for all $`i=2,\mathrm{},m1`$. If $`m3`$ then this shows that there is a homomorphism of graded algebras $`A_{m,n}\mathrm{Hom}^{}(E,E)`$ which maps $`(i)`$ to $`\mathrm{id}_{E_i}`$, $`(i|i+1)`$ to $`\alpha _i`$, and $`(i+1|i)`$ to $`\beta _i`$. One sees easily that this is an isomorphism. In the remaining case $`m=2`$ one has to consider
(4.10)
$$\beta _1\alpha _1\beta _1\mathrm{Hom}^{2nd_1}(E_1,E_2),\alpha _1\beta _1\alpha _1\mathrm{Hom}^{n+d_1}(E_2,E_1).$$
By assumption $`\mathrm{Hom}^{}(E_1,E_2)`$ is concentrated in degree $`nd_1<2nd_1`$, and $`\mathrm{Hom}^{}(E_2,E_1)`$ is concentrated in degree $`d_1<n+d_1`$. Hence both elements in (4.10) are zero, which allows one to define $`A_{m,n}\mathrm{Hom}^{}(E,E)`$ as before. The proof that this is an isomorphism is again straightforward. ∎
An inspection of the preceding proof shows that the result remains true for any other choice of numbers $`d_i`$ in the definition of $`A_{m,n}`$. Our particular choice (4.6) makes the algebra as ‘highly connected’ as possible: $`A_{m,n}/R1`$ is concentrated in degrees $`[n/2]`$. This will be useful in the Hochschild cohomology computations of section 4e.
Let $`𝒜_{m,n}`$ be the dga given by $`A_{m,n}`$ with zero differential. We will now consider the properties of the functors $`t_i`$ on the category $`D(𝒜_{m,n})`$.
###### Lemma 4.11.
The functors $`t_i:D(𝒜_{m,n})D(𝒜_{m,n})`$, $`1im`$, are exact equivalences.
###### Proof.
This is closely related to the parallel statements in and in our section 2b. The strategy, as in Proposition 2.10, is to introduce a left adjoint $`t_i^{}`$ of $`t_i`$, and then to prove that the canonical natural transformations $`\mathrm{Id}t_it_i^{}`$, $`t_i^{}t_i\mathrm{Id}`$ are isomorphisms.
Set $`𝒜=𝒜_{m,n}`$ and $`𝒬_i=𝒫_i[n]Dgm(𝒜)`$. Define functors $`t_i^{}`$ ($`1im`$) from $`Dgm(𝒜)`$ to itself by
$$t_i^{}()=\{\stackrel{\eta _i}{}e_i_k𝒬_i\}$$
where $``$ is placed in degree zero, and $`\eta _i(x)=x(i|i\pm 1|i)(i)+x(i+1|i)(i|i+1)+x(i1|i)(i|i1)+x(i)(i|i\pm 1|i)`$ (in this formula, the second term should be omitted for $`i=m`$ and the third term for $`i=1`$; the same convention will be used again later on). To understand why $`\eta _i`$ is a module homomorphism, it is sufficient to notice that the element
(4.11)
$$\begin{array}{c}\hfill (i|i\pm 1|i)(i)+(i+1|i)(i|i+1)+(i1|i)(i|i1)+\\ \hfill +(i)(i|i\pm 1|i)Ae_ie_iA\end{array}$$
is central, in the sense that left and right multiplication (with respect to the obvious $`A`$-bimodule structure of $`Ae_ie_iA`$) with any $`aA`$ have the same effect on it. The same argument as for $`t_i`$ shows that $`t_i^{}`$ descends to exact functors on $`K(𝒜)`$ and $`D(𝒜)`$. For any $`Dgm(𝒜)`$ consider the complex of dgms
$$𝒞_1=e_i𝒫_i\stackrel{\delta _1}{}𝒞_0=(e_ie_iAe_i𝒬_i)\stackrel{\delta _0}{}𝒞_1=e_i𝒬_i,$$
where $`\delta _1(xa)=(xa,xa(i|i\pm 1|i)(i)+xa(i+1|i)(i|i+1)+xa(i1|i)(i|i1)+xa(i)(i|i\pm 1|i))=(xa,x(i)(i|i\pm 1|i)a+x(i|i\pm 1|i)a)`$ and $`\delta _0(x,yab)=(\eta _i(x)yab)`$. The reason why the second expression for $`\delta _1`$ is equal to the first one is again that the element (4.11) is central. A straightforward computation (including some tedious sign checking) shows that the dgm $`𝒞`$ obtained by collapsing this complex is equal to $`t_i^{}t_i()`$.
$`e_iAe_i=k(i)k(i|i\pm 1|i)`$ is simply a two-dimensional graded $`k`$-vector space, nontrivial in degrees zero and $`n`$. Take the homomorphism of dgms
(4.12) $`𝒞_0=(e_ie_iAe_i𝒬_i)`$ $`,`$
$`(x,y_1(i)b_1+y_2(i|i\pm 1|i)b_2)`$ $`xy_2b_2.`$
Extending this by zero to $`𝒞_1,𝒞_1`$ yields a dgm homomorphism $`\psi _{}:𝒞=t_i^{}t_i()`$, because (4.12) vanishes on the image of $`\delta _1`$. This homomorphism is surjective for any $``$, and a computation similar to that in Proposition 2.10 shows that the kernel is always an acyclic dgm. Since $`\psi _{}`$ is natural in $``$, we have indeed provided an isomorphism $`t_i^{}t_i\mathrm{Id}_{D(𝒜)}`$. The proof that $`t_it_i^{}\mathrm{Id}_{D(𝒜)}`$ is parallel. ∎
###### Lemma 4.12.
The functors $`t_i`$ on $`D(𝒜_{m,n})`$ satisfy the braid relations (up to graded natural isomorphism):
$`t_it_{i+1}t_i`$ $`t_{i+1}t_it_{i+1}`$ $`\text{for }i=1,\mathrm{},m1,`$
$`t_it_j`$ $`t_jt_i`$ $`\text{for }|ij|2.`$
###### Proof.
The second relation is easy (it follows immediately from the fact that $`e_iA_{m,n}e_j=0`$ for $`|ij|`$), and we will therefore concentrate on the first one. Moreover, we will only explain the salient points of the argument (a different version of it is described in with full details). Note that the approach taken in Proposition 2.13 cannot be adapted directly to the present case, since we have not developed a general theory of twist functors on derived categories of dgms.
Set $`𝒜=𝒜_{m,n}`$ and $`_i=𝒫_i[n]`$. For any $`Dgm(𝒜)`$ consider the complex of dgms
(4.13)
$$𝒞_3\stackrel{\delta _3}{}𝒞_2\stackrel{\delta _2}{}𝒞_1\stackrel{\delta _1}{}𝒞_0,$$
where
$`𝒞_3`$ $`=e_i_i,`$
$`𝒞_2`$ $`=(e_ie_iAe_i𝒫_i)(e_ie_iAe_{i+1}𝒫_{i+1})`$
$`(e_{i+1}e_{i+1}Ae_i𝒫_i),`$
$`𝒞_1`$ $`=(e_i𝒫_i)(e_{i+1}𝒫_{i+1})(e_i𝒫_i),`$
$`𝒞_0`$ $`=`$
and
$`\delta _3`$ $`:(xa)\left(\begin{array}{c}x(i|i+1|i)a\\ x(i|i+1)(i+1|i)a\\ x(i|i+1)(i+1|i)a\end{array}\right),`$
$`\delta _2`$ $`:\left(\begin{array}{c}x_1a_1b_1\\ x_2(i|i+1)b_2\\ x_3(i+1|i)b_3\end{array}\right)\left(\begin{array}{c}x_1a_1b_1+x_2(i|i+1)b_2\\ x_2(i|i+1)b_2+x_3(i+1|i)b_3\\ x_1ab_1x_3(i+1|i)b_3\end{array}\right),`$
$`\delta _1`$ $`:\left(\begin{array}{c}x_1a_1\\ x_2a_2\\ x_3x_3\end{array}\right)x_1a_1+x_2a_2+x_3a_3.`$
As in the proof of the previous Lemma, one can contract this complex to a single dgm, which is in fact canonically isomorphic to to $`t_it_{i+1}t_i()`$. Now, one can map the whole complex (4.13) surjectively to an acyclic complex (concentrated in degrees $`3`$ and $`2`$)
$$e_i_i\stackrel{\mathrm{id}}{}e_i_i.$$
This is done by taking the identity map on $`𝒞_3`$ together with the homomorphism $`𝒞_2e_ie_iAe_i𝒫_ie_i_i`$, $`m_1(i)b_1+m_2(i|i+1|i)b_2m_2b_2`$, and extending this by zero to the other summands of $`𝒞_2`$ and to $`𝒞_1`$, $`𝒞_0`$. The kernel of the dgm homomorphism defined in this way is a certain subcomplex of (4.13). When writing this down explicitly (which we will not do here) one notices that it contains an acyclic subcomplex isomorphic to
$$e_i𝒫_i\stackrel{\mathrm{id}}{}e_i𝒫_i,$$
located in degrees $`2`$ and $`1`$. If one divides out this acyclic subcomplex, what remains is the complex
(4.14)
$$\begin{array}{c}\hfill (e_ie_iAe_{i+1}𝒫_{i+1})(e_{i+1}e_{i+1}Ae_i𝒫_i)\stackrel{\delta _1^{}}{}\\ \hfill (e_i𝒫_i)(e_{i+1}𝒫_{i+1})\stackrel{\delta _0^{}}{}\end{array}$$
with $`\delta _1^{}(x_1(i|i+1)b_1,x_2(i+1|i)b_2)=(x_1(i|i+1)b_1x_2(i+1|i)b_2,x_1(i|i+1)b_1+x_2(i+1|i)b_2)`$, $`\delta _0^{}(x_1a_1,x_2a_2)=x_1a_1+x_2a_2`$. The remarkable fact about (4.14) is that it is symmetric with respect to exchanging $`i`$ and $`i+1`$. Indeed, one can arrive at the same complex by starting with $`t_{i+1}t_it_{i+1}()`$ and removing acyclic parts. This shows that $`t_{i+1}t_it_{i+1}()`$ and $`t_it_{i+1}t_i()`$ are quasi-isomorphic for all $``$. We leave it to the reader to verify that the argument provides a chain of exact functors and graded natural isomorphisms between them, with $`t_it_{i+1}t_i`$ and $`t_{i+1}t_it_{i+1}`$ at the two ends of the chain. ∎
### 4d. Geometric intersection numbers
Consider the weak braid group action $`\rho _{m,n}:B_{m+1}\mathrm{Auteq}(D(𝒜_{m,n}))`$ generated by $`t_1,\mathrm{},t_m`$. The aim of this section is prove a strong form of faithfulness for it:
###### Theorem 4.13.
Let $`R_{m,n}^g`$ be a functor representing $`\rho _{m,n}(g)`$ for some $`gB_{m+1}`$. If $`R_{m,n}^g(𝒫_j)𝒫_j`$ for all $`j`$, then $`g`$ must be the identity element.
We begin by looking at the center of $`B_{m+1}`$. It is infinite cyclic and generated by an element which, in terms of the standard generators $`g_1,\mathrm{},g_m`$, can be written as $`(g_1g_2\mathrm{}g_m)^{m+1}`$.
###### Lemma 4.14.
For any $`1jm`$, $`(t_1t_2\mathrm{}t_m)^{m+1}(𝒫_j)`$ is isomorphic to $`𝒫_j[2m(m+1)n]`$ in $`D(𝒜_{m,n})`$.
###### Proof.
For each $`1jm`$ there is a short exact sequence of dgms
$$0𝒫_j[n]\stackrel{\alpha }{}e_jA_{m,n}e_j_k𝒫_j\stackrel{\text{multiplication}}{}𝒫_j0,$$
where $`\alpha (x)=(j|j\pm 1|j)x(j)(j|j\pm 1|j)x`$. This implies that the cone of the multiplication map, which is $`t_j(𝒫_j)`$, is isomorphic to $`𝒫_j[1n]`$ in $`D(𝒜_{m,n})`$. Note also that $`t_i(𝒫_j)𝒫_j`$ whenever $`|ij|2`$.
Consider the $`m+1`$ differential graded modules
$`_0=\{𝒫_1[n1]𝒫_2[2n1d_1]\mathrm{}𝒫_m[mn1d_1\mathrm{}d_{m1}]\},`$
$`_1=𝒫_1,`$
$`_2=𝒫_2[1d_1],`$
$`_3=𝒫_3[2d_1d_2],`$
$`\mathrm{}`$
$`_m=𝒫_m[m1d_1\mathrm{}d_{m1}].`$
The definition of $`_0`$ is by collapsing the complex of dgms in which $`𝒫_1[n1]`$ is placed in degree zero, and where the maps are given by left multiplication with $`(i+1|i)`$. We will prove that
(4.15)
$$\{\begin{array}{cc}(t_1t_2\mathrm{}t_m)(_0)_1,\hfill & \\ (t_1t_2\mathrm{}t_m)(_i)_{i+1}\text{ for }1i<m\text{,}\hfill & \\ (t_1t_2\mathrm{}t_m)(_m)_0[2m(m+1)n],\hfill & \end{array}$$
which clearly implies the desired result. By the definitions of $`t_i`$ and $`t_i^{}`$, the second of which is given in the proof of Lemma 4.11, one has
$$t_{i+1}(𝒫_i)=\{𝒫_{i+1}[d_i]𝒫_i\}t_i^{}(𝒫_{i+1})[1d_i]$$
where $`𝒫_i`$ is placed in degree zero and the arrow is left multiplication with $`(i|i+1)`$. This shows that $`t_it_{i+1}(𝒫_i)𝒫_{i+1}[1d_i]`$, and since $`t_i(𝒫_j)𝒫_j`$ whenever $`|ij|2`$, it proves the second equation in (4.15). To verify the other two equations one computes
$`(t_1t_2\mathrm{}t_m)(𝒫_m[n1])`$
$`(t_1t_2\mathrm{}t_{m1})(𝒫_m)`$
$`(t_1t_2\mathrm{}t_{m2})(\{𝒫_{m1}[n+d_{m1}]𝒫_m\})`$
$`(t_1t_2\mathrm{}t_{m3})(\{𝒫_{m2}[2n+d_{m2}+d_{m1}]𝒫_{m1}[n+d_{m1}]𝒫_m\})`$
$`\mathrm{}_0[m(1n)+d_1+\mathrm{}+d_{m1}]`$
and
$`(t_m^{}\mathrm{}t_2^{}t_1^{})(𝒫_1)`$
$`(t_m^{}\mathrm{}t_2^{})(𝒫_1[n1])`$
$`(t_m^{}\mathrm{}t_3^{})(\{𝒫_1[n1]𝒫_2[2n1d_1]\})`$
$`(t_m^{}\mathrm{}t_4^{})(\{𝒫_1[n1]𝒫_2[2n1d_1]𝒫_3[3n1d_1d_2]\})`$
$`\mathrm{}_0.\mathit{}`$
It seems likely that $`(t_1t_2\mathrm{}t_m)^{m+1}`$ is in fact isomorphic to the translation functor $`[2m(m+1)n]`$, but we have not checked this.
Before proceeding further, we need to recall some basic notions from the topology of curves on surfaces. Let $`D`$ be a closed disc, and $`\mathrm{\Delta }DD`$ a set of $`m+1`$ marked points. $`\mathrm{Diff}(D,D;\mathrm{\Delta })`$ denotes the group of diffeomorphisms $`f:DD`$ which satisfy $`f|D=\mathrm{id}`$ and $`f(\mathrm{\Delta })=\mathrm{\Delta }`$. We write $`f_0f_1`$ for isotopy within this group. By a curve in $`(D,\mathrm{\Delta })`$ we mean a subset $`cDD`$ which can be represented as the image of a smooth embedding $`\gamma :[0;1]D`$ such that $`\gamma ^1(\mathrm{\Delta })=\{0;1\}`$. In other words, $`c`$ is an unoriented embedded path in $`DD`$ whose endpoints lie in $`\mathrm{\Delta }`$, and which does not meet $`\mathrm{\Delta }`$ anywhere else. There is an obvious notion of isotopy for curves, denoted again by $`c_0c_1`$. For any two curves $`c_0,c_1`$ there is a geometric intersection number $`I(c_0,c_1)0`$, which is defined by $`I(c_0,c_1)=|(c_0^{}c_1)\mathrm{\Delta }|+\frac{1}{2}|(c_0^{}c_1)\mathrm{\Delta }|`$ for some $`c_0^{}c_0`$ which has minimal intersection with $`c_1`$ (this means, roughly speaking, that $`c_0^{}`$ is obtained from $`c_0`$ by removing all unnecessary intersection points with $`c_1`$). We refer to \[27, section 2a\] for the proof that this is well-defined. Once one has shown this, the following properties are fairly obvious:
1. $`I(c_0,c_1)`$ depends only on the isotopy classes of $`c_0`$ and $`c_1`$;
2. $`I(c_0,c_1)=I(f(c_0),f(c_1))`$ for all $`f\mathrm{Diff}(D,D;\mathrm{\Delta })`$;
3. $`I(c_0,c_1)=I(c_1,c_0)`$.
Note that in general $`I(c_0,c_1)`$ is only a half-integer, because of the weight $`1/2`$ with which the common endpoints of $`c_0`$ and $`c_1`$ contribute. The next Lemma, whose proof we omit, is a modified version of \[13, Proposition III.16\].
###### Lemma 4.15.
Let $`c_0,c_1`$ be two curves in $`(D,\mathrm{\Delta })`$ such that $`I(d,c_0)=I(d,c_1)`$ for all $`d`$. Then $`c_0c_1`$. ∎
From now on, fix a collection of curves $`b_1,\mathrm{},b_m`$ as in Figure 5, as well as an orientation of $`D`$. Then one can identify $`\pi _0(\mathrm{Diff}(D,D;\mathrm{\Delta }))`$ with the braid group by mapping the standard generators $`g_1,\mathrm{},g_mB_{m+1}`$ to positive half-twists along $`b_1,\mathrm{},b_m`$.
###### Lemma 4.16.
Let $`f\mathrm{Diff}(D,D;\mathrm{\Delta })`$ be a diffeomorphism which satisfies $`f(b_j)b_j`$ for all $`1jm`$. The the corresponding element $`gB_{m+1}`$ must be of the form $`g=(g_1g_2\mathrm{}g_m)^{\nu (m+1)}`$ for some $`\nu `$.
###### Proof.
Since $`f(b_j)b_j`$, $`f`$ commutes up to isotopy with the half-twist along $`b_j`$, and hence with any element of $`\mathrm{Diff}(D,D;\mathrm{\Delta })`$. This implies that $`g`$ is central. ∎
The next Lemma, which is far more substantial than the previous ones, establishes a relationship between the topology of curves in $`(D,\mathrm{\Delta })`$ and the algebraically defined braid group action $`\rho _{m,n}`$.
###### Lemma 4.17.
For $`gB_{m+1}`$, let $`f\mathrm{Diff}(D,D;\mathrm{\Delta })`$ be a diffeomorphism in the isotopy class corresponding to $`g`$, and $`R_{m,n}^g`$ a functor which represents $`\rho _{m,n}(g)`$. Then
$$\underset{r}{}dim_k\mathrm{Hom}_{D(𝒜_{m,n})}(𝒫_i,R_{m,n}^g(𝒫_j)[r])=2I(b_i,f(b_j))$$
for all $`1i,jm`$.
A statement of the same kind, concerning a category and braid group action slightly different from ours, has been proved in \[27, Theorem 1.1\]. In principle, the proof given there can be adapted to our situation, but verifying all the details is a rather tedious business. For this reason we take a slightly different approach, which is to derive the result as stated here from its counterpart in . To do this, we first need to recall the situation considered in that paper. In order to avoid confusion, objects which belong to the setup of will be denoted by overlined symbols.
Consider the quiver $`\overline{\mathrm{\Gamma }}_m`$ in Figure 6 with vertices numbered $`0,\mathrm{},m`$ and whose edges are labelled with ‘degrees’ zero or one. Paths of length $`l`$ in $`\overline{\mathrm{\Gamma }}_m`$ are described by $`(l+1)`$-tuples of numbers $`i_0,\mathrm{},i_l\{0,\mathrm{},m\}`$; we will use the notation $`(\overline{i_0|\mathrm{}|i_l})`$ for them. The path algebra $`k[\overline{\mathrm{\Gamma }}_m]`$ is a graded algebra, whose ground ring is $`\overline{R}=k^{m+1}`$. Let $`\overline{J}_m`$ be the homogeneous two-sided ideal in it generated by the elements $`(\overline{i1|i|i+1})`$, $`(\overline{i+1|i|i1})`$, $`(\overline{i|i+1|i})(\overline{i|i1|i})`$ ($`1im1`$), and $`(\overline{0|1|0})`$. The quotient $`\overline{A}_m=k[\overline{\mathrm{\Gamma }}_m]/\overline{J}_m`$ is a finite-dimensional graded algebra; a concrete basis is given by the $`4m+1`$ elements
(4.16)
$$\{\begin{array}{cc}& (\overline{0}),\mathrm{},(\overline{m}),(\overline{0|1}),\mathrm{},(\overline{m1|m})\text{of degree zero, and}\hfill \\ & (\overline{1|0}),\mathrm{},(\overline{m|m1}),(\overline{1|2|1})=(\overline{1|0|1}),\mathrm{},(\overline{m1|m2|m1})=\hfill \\ & =(\overline{m1|m|m1}),(\overline{m|m1|m})\text{of degree one.}\hfill \end{array}$$
$`\overline{A}_m`$ is evidently a close cousin of our algebras $`A_{m,n}`$. We will now make the relationship precise on the level of categories. Let $`\overline{A}_m\text{-}mod`$ be the abelian category of finitely-generated graded right modules over $`\overline{A}_m`$, and $`D^b(\overline{A}_m\text{-}mod)`$ its bounded derived category (in contrast to the situation in section 4a, this is the derived category in the ordinary sense, not in the differential graded one). There is an automorphism $`\{1\}`$ which shifts the grading of a module up by one. This descends to an automorphism of $`D^b(\overline{A}_m\text{-}mod)`$, which is not the same as the translation functor. In particular, for any $`X,YD^b(\overline{A}_m\text{-}mod)`$ there is a bigraded vector space
$$\underset{r_1,r_2}{}\mathrm{Hom}_{D^b(\overline{A}_m\text{-}mod)}(X,Y\{r_1\}[r_2]).$$
We denote by $`\overline{P}_i\overline{A}_m\text{-}mod`$ the projective modules $`(\overline{i})\overline{A}_m`$, for $`0im`$. Let $`𝔓\overline{A}_m\text{-}mod`$ be the full subcategory whose objects are direct sums of $`\overline{P}_i\{r\}`$ for $`i=1,\mathrm{},m`$ and $`r`$; the important thing is that $`\overline{P}_0`$ is not allowed. We write $`K^b(𝔓)`$ for the full subcategory of $`K^b(\overline{A}_m\text{-}mod)`$ whose objects are finite complexes in $`𝔓`$. This is an abuse of notation since $`𝔓`$ is not an abelian category; however, $`K^b(𝔓)`$ is still a triangulated category, because it contains the cone of any homomorphism.
###### Lemma 4.18.
There is an exact functor $`\mathrm{\Pi }:K^b(𝔓)D(𝒜_{m,n})`$ with the following properties:
1. $`\mathrm{\Pi }(\overline{P}_i)`$ is isomorphic to $`𝒫_i`$ up to some shift;
2. There is a canonical isomorphism of functors $`\mathrm{\Pi }\{1\}[n]\mathrm{\Pi }`$;
3. The natural map, which exists in view of property (2),
$$\underset{r_2=nr_1}{}\mathrm{Hom}_{K^b(𝔓)}(X,Y\{r_1\}[r_2])\mathrm{Hom}_{D(𝒜_{m,n})}(\mathrm{\Pi }(X),\mathrm{\Pi }(Y)),$$
is an isomorphism for all $`X,YK^b(𝔓)`$.
###### Proof.
As a first step, consider the functor $`\mathrm{\Pi }^{}:𝔓Dgm(𝒜_{m,n})`$ defined as follows. The object $`\overline{P}_i\{r\}`$ goes to the dgm $`𝒫_i[\sigma _inr]`$, where $`\sigma _i=d_1d_2\mathrm{}d_{i1}`$, and this is extended to direct sums in the obvious way. Let $`\overline{A}_m^d`$ be the space of elements of degree $`d`$ in $`\overline{A}_m`$. Homomorphisms of graded modules $`\overline{P}_i\{r\}\overline{P}_j\{s\}`$ correspond in a natural way to elements of $`(\overline{j})\overline{A}_m^{rs}(\overline{i})`$. On the other hand, dgm homomorphisms between $`𝒫_i[\sigma _inr]`$ and $`𝒫_j[\sigma _jns]`$ correspond to elements of degree $`\sigma _j\sigma _in(sr)`$ in $`(j)A_{m,n}(i)`$. There is an obvious isomorphism, for any $`1i,jm`$ and $`d`$,
(4.17)
$$(\overline{j})\overline{A}_m^d(\overline{i})(j)A_{m,n}^{\sigma _j\sigma _i+nd}(i)$$
which sends any basis element in (4.16) of the form $`(\overline{i_0|\mathrm{}|i_\nu })`$ to the corresponding element $`(i_0|\mathrm{}|i_\nu )A_{m,n}`$; one needs to check, case by case, that the degrees turn out right. We use (4.17) to define $`\mathrm{\Pi }^{}`$ on morphisms; this is obviously compatible with composition, so that the outcome is indeed a functor. Note that $`\mathrm{\Pi }^{}\{1\}[n]\mathrm{\Pi }^{}`$.
Now take a finite chain complex in $`𝔓`$. Applying $`\mathrm{\Pi }^{}`$ to each object in the complex yields a chain complex in $`Dgm(𝒜_{m,n})`$, which one can then collapse into a single dgm. This procedure yields a functor $`K^b(𝔓)K(𝒜_{m,n})`$, which is exact since it carries cones to cones. We define $`\mathrm{\Pi }`$ to be the composition of this with the quotient functor $`K(𝒜_{m,n})D(𝒜_{m,n})`$. Properties (1) and (2) are now obvious from the definition of $`\mathrm{\Pi }^{}`$. The remaining property (3) can be reduced, by repeated use of the Five Lemma, to the case when $`X=\overline{P}_i\{r\}`$, $`Y=\overline{P}_i\{s\}`$; and then it comes down to the fact that (4.17) is an isomorphism. ∎
Define exact functors $`\overline{t}_1,\mathrm{},\overline{t}_m`$ from $`D^b(\overline{A}_m\text{-}mod)`$ to itself by
(4.18)
$$\overline{t}_i(X)=\{X(\overline{i})_k\overline{P}_iX\}.$$
Here $`X(\overline{i})`$ is considered as a complex of graded $`k`$-vector spaces; tensoring with $`\overline{P}_i`$ over $`k`$ makes this into a complex of graded $`\overline{A}_m`$-modules; and the arrow is the multiplication map. We can now state the results of .
###### Lemma 4.19.
$`\overline{t}_1,\mathrm{},\overline{t}_m`$ are exact equivalences and generate a weak braid group action $`\overline{\rho }_m:B_{m+1}\mathrm{Auteq}(D^b(\overline{A}_m\text{-}mod))`$.
###### Lemma 4.20.
For $`gB_{m+1}`$, let $`f\mathrm{Diff}(D,D;\mathrm{\Delta })`$ be a diffeomorphism in the isotopy class corresponding to $`g`$, and $`\overline{R}_m^g`$ a functor which represents $`\overline{\rho }_m(g)`$. Then
$$\underset{r_1,r_2}{}dim_k\mathrm{Hom}_{D^b(\overline{A}_m\text{-}mod)}(\overline{P}_i,\overline{R}_m^g(\overline{P}_j)\{r_1\}[r_2])=2I(b_i,f(b_j))$$
for all $`1i,jm`$.
Lemma 4.19 essentially summarizes the contents of \[27, section 3\], and Lemma 4.20 is \[27, Theorem 1.1\]. The notation here is slightly different (our $`\overline{A}_m`$, $`\overline{P}_i`$ and $`\overline{t}_i`$ are the $`A_m`$, $`P_i`$ and $`_i`$ of that paper). We have also modified the definitions very slightly, namely, we use right modules instead of left modules as in , and the coefficients are $`k`$ instead of $``$. These changes do not affect the results at all (a very conscientious reader might want to check that inversion of paths defines an isomorphism between $`\overline{A}_m`$ and its opposite, and that a result similar to Lemma 4.18 can be proved for an algebra $`\overline{A}_m`$ defined over $``$).
###### Proof of Lemma 4.17.
Since the modules $`\overline{P}_i`$ are projective, the obvious exact functor $`K^b(𝔓)D^b(\overline{A}_m\text{-}mod)`$ is full and faithful. To save notation, we will consider $`K^b(𝔓)`$ simply as a subcategory of $`D^b(\overline{A}_m\text{-}mod)`$. An inspection of (4.18) shows that the $`\overline{t}_i`$ preserve this subcategory, and the same is true of their inverses, defined in . In other words, the weak braid group action $`\overline{\rho }_m`$ restricts to one on $`K^b(𝔓)`$. It follows from the definition of $`\mathrm{\Pi }`$ that $`\mathrm{\Pi }\overline{t}_i|K^b(𝔓)t_i\mathrm{\Pi }`$. Hence, if $`\overline{R}_m^g`$ and $`R_{m,n}^g`$ are functors representing $`\overline{\rho }_m(g)`$ respectively $`\rho _{m,n}(g)`$, the diagram
commutes up to isomorphism. Using this, Lemma 4.18(3) and Lemma 4.20, one sees that
$`_rdim_k\mathrm{Hom}_{D(𝒜_{m,n})}(𝒫_i,R_{m,n}^g(𝒫_j)[r])`$
$`=_rdim_k\mathrm{Hom}_{D(𝒜_{m,n})}(\mathrm{\Pi }(\overline{P}_i),\mathrm{\Pi }\overline{R}_m^g(\overline{P}_j)[r])`$
$`=_{r_1,r_2}dim_k\mathrm{Hom}_{D^b(\overline{A}_m\text{-}mod)}(\overline{P}_i,\overline{R}_m^g(\overline{P}_j)\{r_1\}[r_2])`$
$`=2I(b_i,f(b_j)).\mathit{}`$
###### Proof of Theorem 4.13.
For $`gB_{m+1}`$, choose $`f`$ and $`R_{m,n}^g`$ as in Lemma 4.17. Take also another element $`g^{}B_{m+1}`$ and correspondingly $`f^{}`$, $`R_{m,n}^g^{}`$. Applying Lemma 4.17 to $`(g^{})^1g`$ shows that
$`I(f^{}(b_i),f(b_j))=I(b_i,(f^{})^1f(b_j))`$
$`=\frac{1}{2}_rdim_k\mathrm{Hom}(𝒫_i,(R_{m,n}^g^{})^1R_{m,n}^g(𝒫_j))`$
and assuming that $`R_{m,n}^g(𝒫_j)𝒫_j`$ for all $`j`$,
$`=\frac{1}{2}_rdim_k\mathrm{Hom}(𝒫_i,(R_{m,n}^g^{})^1(𝒫_j))`$
$`=I(b_i,(f^{})^1(b_j))=I(f^{}(b_i),b_j).`$
Since $`i`$ and $`f^{}`$ can be chosen arbitrary, it follows from Lemma 4.15 that $`f(b_j)b_j`$ for all $`j`$. Hence, by Lemma 4.16, $`g=(g_1g_2\mathrm{}g_m)^{\nu (m+1)}`$ for some $`\nu `$. But then $`R_{m,n}^g(𝒫_j)𝒫_j[\nu (2m(m+1)n)]`$ by Lemma 4.14. In view of the assumption that $`R_{m,n}^g(𝒫_j)𝒫_j`$, this implies that $`\nu =0`$, hence that $`g=1`$. ∎
### 4e. Conclusion
The graded algebras $`A_{m,n}`$ are always augmented. For $`n2`$ they are even connected, so that there is only one choice of augmentation map. This makes it possible to apply Theorem 4.7.
###### Lemma 4.21.
$`A_{m,n}`$ is intrinsically formal for all $`m,n2`$.
The proof is by a straight computation of Hochschild cohomology (it would be nice to have a more conceptual explanation of the result). Its difficulty depends strongly on the parameter $`n`$. The easy case is when $`n>2`$, since then already the relevant Hochschild cochain groups are zero; this is no longer true for $`n=2`$. At first sight the computation may appear to rely on our specific choice (4.6) of degrees $`d_i`$, but in fact this only serves to simplify the bookkeeping: the Hochschild cohomology remains the same for any other choice. Throughout, we will write $`\mathrm{\Gamma },A`$ instead of $`\mathrm{\Gamma }_{m,n},A_{m,n}`$.
###### Proof for $`n>2`$.
Note that the ‘degree’ label on any edge of $`\mathrm{\Gamma }`$ is $`[n/2]`$. Moreover, the labels on any two consecutive edges add up to $`n`$. These two facts imply that the degree of any nonzero path $`(i_0|\mathrm{}|i_l)`$ of length $`l`$ in $`k[\mathrm{\Gamma }]`$ is $`[(nl)/2]`$. Now, any element of $`(A^+)^{_Rq}`$ can be written as a sum of expressions of the form
$$c=(i_{1,0}|\mathrm{}|i_{1,l_1})(i_{2,0}|\mathrm{}|i_{2,l_2})\mathrm{}(i_{q,0}|\mathrm{}|i_{q,l_q}),$$
with all $`l_q>0`$. Because the tensor product is over $`R`$, such a $`c`$ can be nonzero only if the paths $`(i_{\nu ,0}|\mathrm{}|i_{\nu ,l_\nu })`$ match up, in the sense that $`i_{\nu ,l_\nu }=i_{\nu +1,0}`$. Then, using the observation made above, one finds that
$$\mathrm{deg}(c)=\mathrm{deg}(i_{1,0}|\mathrm{}|i_{1,l_1}|i_{2,1}|\mathrm{}|i_{2,l_2}|i_{3,1}|\mathrm{}|i_{q,l_q})[n(l_1+\mathrm{}+l_q)/2].$$
Hence $`(A^+)^q`$ is concentrated in degrees $`[(nq)/2]`$. On the other hand, $`A[2q]`$ is concentrated in degrees $`n+q2`$, which implies that
$$C^q(A,A[2q])=\mathrm{Hom}_{RR}((A^+)^{_Rq},A[2q])=0\text{if }n4\text{ or }q4\text{.}$$
We will now focus on the remaining case $`(n,q)=(3,3)`$. Then $`(A^+)^{_R3}`$ is concentrated in degrees $`4`$ while $`A[1]`$ is concentrated in degrees $`4`$. The degree four part of $`(A^+)^{_R3}`$ is spanned by elements $`c=(i_0|i_1)(i_1|i_2)(i_2|i_3)`$, which obviously satisfy $`i_3i_0`$. It follows that as an $`R`$-bimodule, the degree four part satisfies $`e_i((A^+)^{_R3})^4e_i=0`$. On the other hand, the degree four part of $`A[1]`$ is spanned by the elements $`(i|i\pm 1|i)`$, so it satisfies $`e_iA[1]^4e_j=0`$ for all $`ij`$. This implies that there can be no nonzero $`R`$-bimodule maps between $`(A^+)^{_R3}`$ and $`A[1]`$, and hence that $`C^3(A,A[1])`$ is after all trivial.
###### Proof for $`n=2`$.
Consider the relevant piece of the Hochschild complex,
$$C^{q1}(A,A[2q])\stackrel{^{q1}}{}C^q(A,A[2q])\stackrel{^q}{}C^{q+1}(A,A[2q]).$$
$`C^{q+1}(A,A[2q])`$ is zero for degree reasons. In fact, since all edges in $`\mathrm{\Gamma }`$ have ‘degree’ labels one, paths are now graded by their length, so that $`(A^+)^{_Rq+1}`$ is concentrated in degrees $`q+1`$, while $`A[2q]`$ is concentrated in degrees $`q`$. In contrast $`C^q(A,A[2q])`$ is nonzero for all even $`q`$. To give a more precise description of this group we will use the basis of $`A`$ from (4.7), and the basis of $`(A^+)^{_Rq}`$ derived from that. Let $`(i_0|\mathrm{}|i_q)`$, $`i_q=i_0`$, be a closed path of length $`q`$ in $`\mathrm{\Gamma }`$. Define $`\varphi _{i_0,\mathrm{},i_m}C^q(A,A[2q])`$ by setting
$$\varphi _{i_0,\mathrm{},i_q}(c)=\{\begin{array}{cc}(i_0|i_0\pm 1|i_0)\hfill & \text{if }c=(i_0|i_1)\mathrm{}(i_{q1}|i_q),\hfill \\ 0\hfill & \text{on all other basis elements }c\text{.}\hfill \end{array}$$
We claim that the elements defined in this way, with $`(i_0|\mathrm{}|i_q)`$ ranging over all closed paths, form a basis of $`C^q(A,A[2q])`$. To prove this, note that there is only one degree, which is $`q`$, where both $`(A^+)^q`$ and $`A[2q]`$ are nonzero. The degree $`q`$ part of $`(A^+)^q`$ is spanned by expressions $`c=(i_0|i_1)\mathrm{}(i_{q1}|i_q)`$, with $`i_q`$ not necessarily equal to $`i_0`$. The degree $`q`$ part of $`A[2q]`$ is spanned by elements $`(i|i\pm 1|i)`$. Hence, an argument using the $`R`$-bimodule structure shows that if $`i_qi_0`$, then $`\varphi (c)=0`$ for all $`\varphi C^q(A,A[2q])`$. This essentially implies what we have claimed.
We now turn to $`C^{q1}(A,A[2q])`$; for this group we will not need a complete description, but only some sample elements. Given a closed path $`(i_0|\mathrm{}|i_q)`$ as before in $`\mathrm{\Gamma }`$, we define $`\varphi ^{}C^{q1}(A,A[2q])`$ by setting $`\varphi ^{}(c)=(i_0|i_{q1})`$ if $`c=(i_0|i_1)\mathrm{}(i_{q2}|i_{q1})`$, and zero on all other basis elements $`c`$. A simple computation shows that $`\delta ^{q1}(\varphi ^{})=\varphi _{i_0,\mathrm{},i_q}\varphi _{i_{q1},i_0,i_1,\mathrm{},i_{q1}}`$. Also, for any closed path $`(i_0|\mathrm{}|i_q)`$ with $`i_2=i_0`$ and $`i_1=i_0+1`$, define $`\varphi ^{\prime \prime }C^{q1}(A,A[2q])`$ by setting $`\varphi ^{\prime \prime }(c)=(i_0|i_0\pm 1|i_0)`$ for $`c=(i_0|i_1|i_2)(i_2|i_3)\mathrm{}(i_{q1}|i_q)`$, and again zero for all other basis elements $`c`$. Then $`\delta ^{q1}(\varphi ^{\prime \prime })`$ is equal to $`\varphi _{i_0,i_1,\mathrm{},i_q}\varphi _{i_0,i_12,i_2,\mathrm{},i_q}`$ for $`i_0>1`$, and to $`\varphi _{i_0,i_1,\mathrm{},i_q}`$ for $`i_0=1`$.
To summarize, we have now established that the following relations hold in $`HH^q(A,A[2q])`$:
1. $`[\varphi _{i_0,\mathrm{},i_q}]=[\varphi _{i_{q1},i_q,i_1,\mathrm{},i_{q1}}]`$ for all closed paths $`(i_0|\mathrm{}|i_q)`$ in the quiver $`\mathrm{\Gamma }`$.
2. $`[\varphi _{i_0,\mathrm{},i_q}]=[\varphi _{i_0,i_12,i_2,\mathrm{},i_q}]`$ whenever $`i_0=i_22`$ and $`i_1=i_0+1`$.
3. $`[\varphi _{i_0,\mathrm{},i_q}]=0`$ whenever $`i_0=i_2=1`$ and $`i_1=2`$.
Take an arbitrary element $`\varphi _{i_0,\mathrm{},i_q}`$. By applying (1) repeatedly, one can find another element $`\varphi _{i_0^{},\mathrm{},i_q^{}}`$ which represents the same Hochschild cohomology class, up to a sign, and such that $`i_1^{}`$ is maximal among all $`i_\nu ^{}`$. This implies that $`i_0^{}=i_2^{}=i_1^{}1`$. If $`i_1^{}=2`$ then we can apply (3) to show that our Hochschild cohomology class is zero. Otherwise pass to $`\varphi _{i_0^{},i_1^{}2,\mathrm{},i_q^{}}`$, which represents the same Hochschild cohomology class up to sign due to (2), and repeat the argument. The iteration terminates after finitely many moves, because the sum of the $`i_\nu `$ decreases by two in each step. Hence $`HH^q(A,A[2q])`$ is zero for all $`q1`$. ∎
###### Proof of Theorem 2.18.
We first need to dispose of the trivial case $`m=1`$. In that case, choose a resolution $`F_1𝔎`$ of $`E_1`$. Pick a nonzero morphism $`\varphi :F_1F_1[n]`$. This, together with $`\mathrm{id}_{F_1}`$, determines an isomorphism of graded vector spaces $`\mathrm{Hom}^{}(F_1,F_1)kk[n]`$, and hence an isomorphism in $`𝔎`$ between $`F_1F_1[n]`$ and $`\mathrm{Hom}^{}(F_1,F_1)F_1`$. Consider the commutative diagram
The upper row is a piece of the exact triangle which comes from the definition of $`T_{F_1}`$ as a cone, and the lower row is obviously also a piece of an exact triangle. By the axioms of a triangulated category, the diagram can be filled in with an isomorphism between $`F_1[n]`$ and $`T_{F_1}(F_1)[1]`$. Transporting the result to $`D^b(𝔖^{})`$ yields $`T_{E_1}(E_1)E_1[1n]`$. Since $`n2`$ by assumption, it follows that $`T_{E_1}^r(E_1)\cong ̸E_1`$ unless $`r=0`$.
From now on suppose that $`m2`$. After shifting each $`E_i`$ by some amount, we may assume that $`\mathrm{Hom}^{}(E_{i+1},E_i)`$ is concentrated in degree $`d_i`$ for $`i=1,\mathrm{},m1`$ (shifting will not affect the statement because $`T_{E_i[j]}`$ is isomorphic to $`T_{E_i}`$ for any $`j`$). Choose resolutions $`E_1^{},\mathrm{},E_m^{}𝔎`$ for $`E_1,\mathrm{},E_m`$. Lemma 4.10 shows that the endomorphism dga $`end(E^{})`$ has $`H(end(E^{}))A_{m,n}`$. By Lemma 4.21, $`end(E^{})`$ must be quasi-isomorphic to $`𝒜_{m,n}`$. Define an exact functor $`\mathrm{\Psi }`$ to be the composition
$$D^b(𝔖^{})\stackrel{}{}𝔎\stackrel{\mathrm{\Psi }_E^{}}{}D(end(E^{}))\stackrel{}{}D(𝒜_{m,n}).$$
The first arrow is the standard equivalence, and the last one is the equivalence induced by some sequence of dgas and quasi-isomorphisms. By construction $`\mathrm{\Psi }(E_i)𝒫_i`$ for $`i=1,\mathrm{},m`$. In the diagram
the first square commutes because that is the definition of $`T_{E_i}`$, the second square by Lemma 4.3, and the third one by Lemma 4.2. Now let $`g`$ be an element of $`B_{m+1}`$, $`R^g:D^b(𝔖^{})D^b(𝔖^{})`$ a functor which represents $`\rho (g)`$, and $`R_{m,n}^g:D(𝒜_{m,n})D(𝒜_{m,n})`$ a functor which represents $`\rho _{m,n}(g)`$. By applying the previous diagram several times one sees that
$$R_{m,n}^g\mathrm{\Psi }\mathrm{\Psi }R^g.$$
Assume that $`R_g(E_i)E_i`$ for all $`i`$; then also $`R_{m,n}^g(𝒫_i)=R_{m,n}^g\mathrm{\Psi }(E_i)\mathrm{\Psi }R^g(E_i)\mathrm{\Psi }(E_i)𝒫_i`$. By Theorem 4.13 it follows that $`g`$ must be the identity. ∎
We have not tried to compute the Hochschild cohomology of $`A_{m,n}`$ for $`n=1`$. However, an indirect argument using the non-faithful $`B_4`$-action of section 3d shows that $`A_{3,1}`$ cannot be intrinsically formal. More explicitly, if one takes the sheaves $`𝒪_x,𝒪,𝒪_y`$ used in that example, and chooses injective resolutions by quasi-coherent sheaves for them, then the resulting dga $`end(E^{})`$ is not formal. One can give a more direct proof of the same fact by using essentially the same Massey product computation as Polishchuk in \[43, p. 3\]. |
warning/0001/cond-mat0001330.html | ar5iv | text | # Non-Ohmic Coulomb drag in the ballistic electron transport regime
## Abstract
We work out a theory of the Coulomb drag current created under the ballistic transport regime in a one-dimensional nanowire by a ballistic non-Ohmic current in a nearby parallel nanowire. As in the Ohmic case, we predict sharp oscillation of the drag current as a function of gate voltage or the chemical potential of electrons. We also study dependence of the drag current on the voltage $`V`$ across the driving wire. For relatively large values of $`V`$ the drag current is proportional to $`V^2`$.
The purpose of the present paper is to study the Coulomb drag current in the course of ballistic (collisionless) electron transport in a nanowire due to a ballistic driving non-Ohmic current in an adjacent parallel nanowire. The possibility of the Coulomb drag effect in the ballistic regime in quantum wires has been demonstrated by Gurevich, Pevzner and Fenton <sup>*</sup><sup>*</sup>*A number of references to early papers on the Coulomb drag is given in . and has been experimentally observed by Debray et al. . If two wires, 1 and 2, are near one another and are parallel, the drag force due to the ballistic current in wire 2 acts as a sort of permanent acceleration on the electrons of wire 1 via the Coulomb interaction.
We assume that the largest dimension of the structure is smaller than the electron mean free path in the problem (typically a few $`\mu `$m). Such nanoscale systems are characterized by low electron densities, which may be varied by means of the gate voltage.
Following we assume that the drag current in wire 1 is much smaller than the driving ballistic current in wire 2 and calculate the drag current by iterating the Boltzmann equation for wire 1. We have
$$v\frac{F^{(1)}}{z}=I^{(12)}\{F^{(1)},F^{(2)}\},$$
(1)
where $`F^{(1,2)}`$ are the electron distribution functions in wires 1 and 2 respectively, $`v=p/m`$ is the electron velocity, $`p`$ is the $`x`$-component of the electron quasimomentum, the $`x`$ axis is parallel to the wires. The collision integral takes into account only the interwire electron-electron scattering, so that otherwise the electron motion in both wires is considered as ballistic. Now,
$$I^{(12)}\{F^{(1)},F^{(2)}\}=\underset{n^{}p^{}q}{}W_{1pn,2p^{}n^{}}^{1p+qn,2p^{}qn^{}}𝒫$$
(2)
where
$$𝒫=\left[F_{np}^{(1)}F_{n^{}p^{}}^{(2)}\left(1F_{np+q}^{(1)}\right)\left(1F_{n^{}p^{}q}^{(2)}\right)F_{np+q}^{(1)}F_{n^{}p^{}q}^{(2)}\left(1F_{np}^{(1)}\right)\left(1F_{n^{}p^{}}^{(2)}\right)\right].$$
(3)
As in Ref. , we assume the wires to be different though having the same lengths $`L`$ and consider the interaction processes when electrons in the two nanowires after scattering remain within the initial subbands $`\epsilon _{np}^{(1)}=\epsilon _n^{(1)}(0)+p^2/2m`$ and $`\epsilon _{n^{}p}^{(2)}=\epsilon _n^{}^{(2)}(0)+p^2/2m`$, $`n`$ being the bands’ number.
The first iteration of Eq.(1) gives for the nonequilibrium part of the distribution function $`\mathrm{\Delta }F_{np}^{(1)}`$
$$\mathrm{\Delta }F_{np}^{(1)}=\left(z\pm \frac{L}{2}\right)\frac{1}{v_{np}}I^{(12)}\{F^{(1)},F^{(2)}\},$$
(4)
for $`p>\mathrm{\hspace{0.17em}0}`$ ($`p<\mathrm{\hspace{0.17em}0}`$) respectively. Using the particle conserving property of the scattering integral
$$\underset{n}{}𝑑pI^{(12)}\{F^{(1)},F^{(2)}\}=0$$
(5)
we get for the drag current
$$J=2eL\underset{n}{}_0^{\mathrm{}}\frac{dp}{2\pi \mathrm{}}I^{(12)}\{F^{(1)},F^{(2)}\}.$$
(6)
The scattering probability $`W_{1pn,2p^{}n^{}}^{1p+qn,2p^{}qn^{}}`$ in Eq.(2) includes a delta-function describing the energy conservation for the electrons belonging to two different wires
$`W_{1pn,2p^{}n^{}}^{1p+qn,2p^{}qn^{}}={\displaystyle \frac{2\pi }{\mathrm{}}}\left|V_{1pn,2p^{}n^{}}^{1p+qn,2p^{}qn^{}}\right|^2\delta (\epsilon _{np}^{(1)}+\epsilon _{n^{}p^{}}^{(2)}\epsilon _{np+q}^{(1)}\epsilon _{n^{}p^{}q}^{(2)})`$ (7)
which following Ref. can be brought into the form
$$\delta (\epsilon _{np}^{(1)}+\epsilon _{n^{}p^{}}^{(2)}\epsilon _{np+q}^{(1)}\epsilon _{n^{}p^{}q}^{(2)})=\frac{m}{|pp^{}|}\delta (q+pp^{}).$$
(8)
This means that here we have backscattering processes and the electrons swap their quasimomenta as a result of collision.
To calculate the drag current we do the first iteration of the Boltzmann equation in the collision term. One can insert the equilibrium distribution functions into the collision term, e.g. $`F_{np}^{(1)}=f(\epsilon _{np}^{(1)}\mu )`$ for the first wire. Here $`f(\epsilon \mu )`$ is the Fermi function.
We assume, in the spirit of the Landauer-Buttiker-Imry approach, the driving quantum wire to be connected to reservoirs which we call ’left‘ $`(l)`$ and ’right‘ $`(r)`$, each of them being in independent equilibrium described by shifted chemical potentials $`\mu ^{(l)}=\mu eV/2`$ and $`\mu ^{(r)}=\mu +eV/2`$. Here $`\mu `$ is the average chemical potential while $`\mathrm{\Delta }\mu /e=V`$ is the voltage across wire 2 (we will assume that $`eV>0`$) and $`e<0`$ is the electron charge. Therefore, the electrons entering the wire from the ’left‘ (’right‘) and having quasimomenta $`p^{}>0`$ ($`p^{}<0`$) are described by
$$\genfrac{}{}{0pt}{}{F_{n^{}p^{}}^{(2)}=f(\epsilon _{n^{}p^{}}^{(2)}\mu ^{(l)})p^{}>0,}{F_{n^{}p^{}}^{(2)}=f\left(\epsilon _{n^{}p^{}}^{(2)}\mu ^{(r)}\right)p^{}<0}$$
(9)
and we see that the collision integral Eqs. (2), (3) is identically zero if the initial quasimomentum $`p^{}`$ and the final quasimomentum $`p^{}q`$ of electron are of the same sign. In other words we have here backscattering processes; otherwise the equilibrium distribution functions on the right-hand side of Eqs. (2), (3) would depend on the same chemical potential and the collision term would vanish.
Due to Eq.(8) we will be interested only in the values $`p^{}<\mathrm{\hspace{0.17em}0}`$ \[because of the restriction $`p^{}q=p>0`$ which follows from Eq.(6)\] and get the following product of distribution functions in the collision term Eqs. (2) and (3)
$$𝒫=F_{np}^{(1)}F_{n^{}p^{}}^{(2r)}\left(1F_{np^{}}^{(1)}\right)\left(1F_{n^{}p}^{(2l)}\right)F_{np^{}}^{(1)}F_{n^{}p}^{(2l)}\left(1F_{np}^{(1)}\right)\left(1F_{n^{}p^{}}^{(2r)}\right),$$
(10)
or
$`𝒫=f(\epsilon _{np}^{(1)}\mu )f(\epsilon _{n^{}p^{}}^{(2)}\mu ^{(r)})[1f(\epsilon _{np^{}}^{(1)}\mu )][1f(\epsilon _{n^{}p}^{(2)}\mu ^{(l)})]`$ (11)
$`f(\epsilon _{np^{}}^{(1)}\mu )f(\epsilon _{n^{}p}^{(2)}\mu ^{(l)})[1f(\epsilon _{np}^{(1)}\mu )][1f(\epsilon _{n^{}p^{}}^{(2)}\mu ^{(r)})].`$ (12)
This expression can be recast into the form
$`𝒫=2\mathrm{sinh}\left(eV/2k_\mathrm{B}T\right)\mathrm{exp}\{(\epsilon _{np}^{(1)}\mu )/k_\mathrm{B}T\}\mathrm{exp}\{(\epsilon _{np}^{(2)}\mu )/k_\mathrm{B}T\}`$ (13)
$`\times f(\epsilon _{np}^{(1)}\mu )f(\epsilon _{n^{}p^{}}^{(2)}\mu eV/2)f(\epsilon _{np^{}}^{(1)}\mu )f(\epsilon _{n^{}p}^{(2)}\mu +eV/2)`$ (14)
and for the drag current we have
$$J=2e\mathrm{sinh}\left(\frac{eV}{2k_\mathrm{B}T}\right)\frac{2\pi }{\mathrm{}}\frac{mL}{2\pi \mathrm{}}\left(\frac{2L}{2\pi \mathrm{}}\right)^2\left(\frac{2e^2}{\kappa L}\right)^2\underset{nn^{}}{}_0^{\mathrm{}}𝑑p_0^{\mathrm{}}𝑑p^{}\frac{g_{nn^{}}(p+p^{})}{p+p^{}}𝒬$$
(15)
where
$$𝒬=\mathrm{exp}\frac{\epsilon _{np}^{(1)}\mu }{k_\mathrm{B}T}\mathrm{exp}\frac{\epsilon _{np}^{(2)}\mu }{k_\mathrm{B}T}f(\epsilon _{np}^{(1)}\mu )f(\epsilon _{n^{}p^{}}^{(2)}\mu \frac{eV}{2})f(\epsilon _{np^{}}^{(1)}\mu )f(\epsilon _{n^{}p}^{(2)}\mu +\frac{eV}{2}),$$
(16)
$`\kappa `$ is the dielectric susceptibility of the sample and
$$g_{nn^{}}(q)=\left[𝑑𝐫_{}𝑑𝐫_{}^{}|\varphi _n(𝐫_{})|^2K_0\left(|q||𝐫_{}𝐫_{}^{}|/\mathrm{}\right)|\varphi _n^{}(𝐫_{}^{})|^2\right]^2.$$
(17)
Some estimates of this function are given in Ref. .
According to the reasoning given in Ref. all the terms of the sum (15) where the differences $`|\epsilon ^{(1)}(0)\epsilon ^{(2)}(0)|`$ are much bigger than both $`k_\mathrm{B}T`$ and $`eV`$ do not contribute to the current $`J`$. Therefore we are left with the terms of the sum where $`|\epsilon ^{(1)}(0)\epsilon ^{(2)}(0)|`$ is smaller than or of the order of $`k_\mathrm{B}T`$ or $`eV`$. We will assume for simplicity that there is only one such difference (otherwise we would have gotten a sum of several terms of the same structure).
As $`𝒬`$ is a sharp function of $`p`$ and $`p^{}`$ one can take out of the integral all the slowly varying functions and get
$$J=J_0\frac{1}{2}\mathrm{sinh}\left(\frac{eV}{2k_\mathrm{B}T}\right)\frac{{\displaystyle \frac{eV}{4k_\mathrm{B}T}}{\displaystyle \frac{\epsilon _{nn^{}}}{2k_\mathrm{B}T}}}{\mathrm{sinh}\left({\displaystyle \frac{eV}{4k_\mathrm{B}T}}{\displaystyle \frac{\epsilon _{nn^{}}}{2k_\mathrm{B}T}}\right)}\frac{{\displaystyle \frac{eV}{4k_\mathrm{B}T}}+{\displaystyle \frac{\epsilon _{nn^{}}}{2k_\mathrm{B}T}}}{\mathrm{sinh}\left({\displaystyle \frac{eV}{4k_\mathrm{B}T}}+{\displaystyle \frac{\epsilon _{nn^{}}}{2k_\mathrm{B}T}}\right)}$$
(18)
where
$$J_0=\frac{8e^5m^3L(k_\mathrm{B}T)^2}{\kappa ^2\pi ^2\mathrm{}^4}.\frac{g_{nn^{}}(2p_n)}{p_n^3}$$
(19)
Here we have introduced notation
$$\epsilon _{nn^{}}=\epsilon _n^{(1)}(0)\epsilon _n^{}^{(2)}(0),mv_n=p_n=\sqrt{2m[\mu \epsilon _n^{(1)}(0)]}.$$
Let us give an order-of-magnitude estimate of the drag current $`J`$ for a realistic situation. We assume $`T=`$1K, $`\mu `$=14 meV, the widths of the wires are 25 nm, the distance between thr central lines of the wires is 50 nm, $`\kappa `$=13, $`m=6.710^{29}`$. Then
$$J_010^{10}\mathrm{A}.$$
(20)
For $`eVk_\mathrm{B}T`$ one gets from Eq. (18) the result of Ref. . Let us consider in detail the opposite case $`eVk_\mathrm{B}T`$. In this case one gets a nonvanishing result for Eq.(18) only if $`|\epsilon _{nn}|<eV/2`$ and one obtains the following equation for the drag current
$$J=\left[\left(\frac{eV}{2}\right)^2\left(\epsilon _{nn^{}}\right)^2\right],=\frac{2e^5m^3L}{\kappa ^2\pi ^2\mathrm{}^4}\frac{g_{nn^{}}(2p_n)}{p_n^3}.$$
(21)
The situation is illustrated in Fig. 1. The straight lines correspond to the positions of the chemical potentials $`\mu ^{(r)}`$ and $`\mu ^{(l)}`$ while the dashed line corresponds to the average value $`\mu `$. Parabolas (1) and (2) represent the dispersion law of electrons in wires (1) and (2) respectively. The full circles correspond to the initial states of colliding electrons.
Before the collision states 1a and 2a are occupied. The circle representing state 1a is below the dashed line, i.e. below the Fermi level $`\mu `$. The circle 2a represents a state with $`p>0`$ which is also occupied as the corresponding energy is below $`\mu ^{(r)}`$.
After the collision state 1b is occupied. It is represented by a circle above the dashed line which means that it has been free before the collision. In wire 2 state 2b with $`p<0`$ is also occupied. It is above $`\mu ^{(l)}`$, i.e. it has been free before the transition.
The width of the stripe between the two straight lines is $`eV`$. The drag current should be proportional to the number of the occupied initial states as well as to the number of free final states. As a result we have for $`eV|\epsilon _{nn^{}}|`$, $`JV^2`$ — see Fig. 2.
In summary, we have developed a theory of Coulomb drag current in a quantum wire brought about by a non-Ohmic current in a nearby parallel nanowire. A ballistic transport in both nanowires is assumed. The drag current $`J`$ as a function of the gate voltage comprises a system of spikes; the position of each spike is determined by a coincidence of a pair of levels of transverse quantization, $`\epsilon _n(0)`$ and $`\epsilon _n^{}(0)`$ in both wires. For $`eVk_\mathrm{B}T`$, $`J`$ is a parabolic function of the driving voltage $`V`$. The effect may play an important role in the investigation of the interwire Coulomb scattering as well as 1D band structure of the wires.
The authors are grateful to P. Debray for sending them a preprint of paper prior to publication. The authors are pleased to acknowledge the support for this work by the Russian National Fund of Fundamental Research (Grant No 97-02-18286-a).
FIGURE CAPTIONS
Fig. 1. Schematic representation of simultaneous transitions due to the interaction between electrons of the two wires for $`eVk_\mathrm{B}T`$. Circles $``$ and $``$ represent the initially unoccupied and occupied states respectively.
Fig. 2. Dependence of the drag current on the driving voltage for $`T=`$1 K. The values of $`|\epsilon _{nn^{}}|`$ go up from left to right. |
warning/0001/astro-ph0001514.html | ar5iv | text | # DYNAMICS OF WARM ABSORBING GAS IN SEYFERT GALAXIES: NGC 5548
## 1 INTRODUCTION
Highly-ionized, tenuous, absorbing gas has been detected in the soft X-ray spectra of many active galactic nuclei (AGNs) at low and moderate ($`\mathrm{\Delta }E/E=0.020.1`$) spectral resolutions. At least half of the Seyfert 1 galaxies observed by the ASCA satellite indicate the presence of this “warm absorber” gas (hereafter WA) in their X-ray spectra (e.g., review by Mushotzky 1997). Other AGNs, such as radio-quiet and radio-loud quasars, host WA as well (e.g., Green and Mathur 1996, Mathur and Elvis 1995, Siebert et al. 1996, Ulrich et al. 1999), though no statistics are currently available. In addition, broad absorption-line quasars (hereafter BAL QSOs), which constitute at least 10% of all quasars, are conspicuously weak in the soft X-rays — a possible signature of a WA gas (Mathur, Elvis, & Singh 1995; Green & Mathur 1996; Gallagher et al. 1999).
The X-ray WA gas is inferred from the softening of the continuum at the low X-ray energies, between $`0.6`$ keV and a few keV, and is due to many unresolved absorption edges from highly ionized species of elements, such as O, Ne, etc. The strongest edge depths correspond to O vii and O viii at 0.72 and 0.85 keV, respectively. It is generally found that the total hydrogen column density of absorbing material lies between $`10^{21}\mathrm{cm}^2`$ and $`10^{23}\mathrm{cm}^2`$, and may even reach $`10^{24}\mathrm{cm}^2`$ (Reynolds et al. 1997, hereafter R97; George et al. 1998, hereafter G98). The energy resolution of the ASCA satellite data does not allow for a study of the kinematics of the WA gas based on the X-ray absorption profiles. The thermodynamic state of the gas, however, including its temperature and density, can be inferred with a moderate amount of modeling and invoking observed X-ray variability (Otani et al. 1995, Reynolds & Fabian 1995; R97).
Intrinsic blueshifted UV absorption was detected in AGNs by the IUE (Bromage et al. 1985; Koratkar et al. 1996; Leech et al. 1991; Walter et al. 1990; Voit et al. 1987; Verón et al. 1985) and HUT (e.g., Kriss et al. 1992). High and sometimes low-ionization lines are observed, e.g., C iv, N v, O vi, Si iv, Mg ii, and others. The higher S/N and spectral resolution data taken by the HST found many more cases of blueshifted intrinsic absorption in the UV resonance lines. Preliminary statistics indicate that intrinsic UV absorption is present in at least half of Seyfert 1 galaxy spectra (Crenshaw et al. 1999 \[C99\]). In addition to the BAL QSOs, there are some quasars that exhibit intrinsic narrow absorption features (see, for example, Mathur, Wilkes, & Elvis 1999).
In all known cases, the UV resonance line absorption is present in sources with known soft X-ray absorption, hinting of a connection between the two types of absorption (C99; Mathur et al. 1999; Brandt, Laor, & Wills 1999), though the nature of this connection is not yet clear. The correspondence between the UV and X-ray absorbing gas was advanced by Mathur et al. (1995), on the basis of single slab modeling of UVX absorbing gas in NGC 5548, using HST/FOS, ROSAT, and ASCA data in their analysis. Similar analysis was performed for other objects (e.g., NGC 3516, Mathur et al. 1997; NGC 3783, Shields & Hamann 1997). The mean FOS spectrum of NGC 5548 represented the best UV data at the time, and its low spectral resolution ($`200\mathrm{km}\mathrm{s}^1`$) justified the single slab approximation. Much higher spectral resolution HST GHRS and STIS/Echelle have since revealed the presence of a number of distinct kinematic components within the outflowing material, with FWHM of $`50\mathrm{km}\mathrm{s}^1160\mathrm{km}\mathrm{s}^1`$ (Crenshaw & Kraemer 1999). These high resolution spectra of NGC 5548 have uncovered at least five separate C iv absorption components (C99; Crenshaw & Kraemer 1999; Mathur, Elvis, & Wilkes 1999 \[MEW99\]).
The presence of a structure within the intrinsic absorption adds another degree of complexity to the kinematical models of AGNs. It is important, therefore, to understand the relationship between the outflowing highly-ionized gas, the broad emission-line (BEL) and absorption-line regions in these objects. Numerous models have been proposed to address this problem (e.g., Blandford 1990; Arav, Shlosman & Weymann 1997; Königl & Kartje 1994). In this paper we extend the hydromagnetic wind model of Emmering, Blandford and Shlosman (1992, hereafter EBS), which provides a dynamical explanation for the the properties of the broad emission-line regions (BLRs) in AGNs (Bottorff et al. 1997, hereafter Paper 1), to the UVX absorption phenomena. In the EBS model, gas in a dusty molecular accretion disk is loaded on magnetic field lines and centrifugally launched forming a magnetized wind. The basic difference between the EBS model and other outflow models is that it provides for a 3D wind dynamics, with helical motion around the axis being superposed onto the radial motion. The wind is clumpy and stratified both in density and ionization. Paper 1 addressed the radial wind structure at smaller latitudes above the optically-thick disk. First, it was found that the BLR is geometrically thick and has substantial optical depth. As a result, the inner part of the wind becomes progressively ionized with increasing latitude, while the outer part of the wind remains cold and molecular as long as it is shielded by the BLR. As the outer material continues to rise it becomes ionized and eventually passes the observer’s line-of-sight, where the wind is detected via absorption in both UV and X-ray energy bands. The line-of-sight velocity of the wind depends on its injection radius on the disk, so a range of velocities is expected. The hydromagnetic wind model, therefore, naturally explains both the UV emission and absorption systems seen in C iv and other lines, as well as X-ray absorption. While our approach is general, we emphasize the results within the context of observations of NGC 5548.
In Section 2, we summarize the important observations of the WA gas, including the column densities of O vii, O viii and total H from the X-ray data, and the column densities of C iv, N v and H i from the UV data. In Section 3, we consider specifically the case of NGC 5548 and its discrete UV absorption components. The velocity field of the EBS model adopted in Paper 1 is then used to determine the location and line-of-sight width of the UV absorption components. The observed column densities of N v and C iv in conjunction with the location and line-of-sight width of each component allow us to determine the density, temperature, volume filling fraction and column densities of O vii, O viii, H i and total H via photoionization modeling. The calculated ionic column densities are compared with observations, to determine whether or not the X-ray absorption in NGC 5548 is compatible with the model and in order to determine the association between the UV and X-ray warm absorbers.
Section 4 describes a generic WA model. The multicomponent absorption, as seen in the C iv UV absorber of NGC 5548, is smoothed into an equivalent single absorbing column, to study the effects of a distributed WA in Seyfert galaxies without specific knowledge of kinematic absorption components. The generic model is investigated for arbitrary aspect angles to the observers. We estimate the fraction of AGNs having detectable WA gas, and larger O viii than O vii optical depth. Finally, in the Appendix, we discuss the general thermal stability of magnetized UVX absorbing gas.
## 2 OBSERVATIONAL SUMMARY
The spectral resolution of ASCA is insufficient to clearly resolve and measure oxygen and other element absorption edges without additional modeling. This is achieved by producing simplified photoionization models of WA gas using observed AGN continua. Free parameters in the model are subsequently tuned to obtain the best fit for the calculated transmitted spectrum to the observed one.
A common approach taken by R97 and G98 makes a number of essential assumptions, namely, the WA is considered (1) to be a thin slab; (2) to have a constant total hydrogen number density; (3) to be in thermal equilibrium; (4) to be exposed to a power-law soft X-ray continuum of photon index $`\mathrm{\Gamma }2`$, and (5) to possess solar abundances. The total hydrogen column density (or equivalently slab thickness) $`N_H`$, the X-ray ionization parameter $`U_X`$ and $`\mathrm{\Gamma }`$ are then varied until an optimal fit to the data is obtained. Additional free parameters include Galactic absorption, a fraction of unattenuated central continuum, a covering factor of the emitting gas, etc.
The main inferred properties of WA gas from X-ray observations are the optical depths of O vii and O viii, and the column density of total H along the line-of-sight. Ionization stratification in WA gas due to differences in density and location is indicated in at least one object, namely MCG 6-30-15. The latter exhibits variability in the O viii edge depth which is anticorrelated with the continuum, and little or no variability in O vii (Fabian et al. 1994, 1995; Otani et al. 1995). The difference in response is frequently attributed in literature to the recombination time scale, $`t(X_i)`$, of ion $`X_i`$ given by
$$t(X_i)=1/\alpha (X_i,T_e)n_e,$$
(1)
where $`n_e`$ is the electron density, and $`\alpha (X_i,T_e)`$ is the recombination coefficient, which is a function of the ion species $`X_i`$ and a slowly varying function of electron temperature $`T_e`$ (but see section 3.3.5). In such a case, over a wide range in temperature $`\alpha (O\mathrm{vii},T_e)`$ and $`\alpha (O\mathrm{viii},T_e)`$ are of the same order of magnitude, so recombination time differences between the two ions are mainly due to differences in density. The response of O viii and the lack of response of O vii in MCG 6-30-15 during the observing time is interpreted as evidence that O vii absorption occurs in a lower density gas than the O viii absorption. Photoionization modeling and absorption depth variability in MCG 6-30-15 also indicate that the O vii and O viii absorption occurs at different locations (Otani et al. 1995). An additional indication of ionization stratification in WA AGN gas is that objects with significant optical reddening exhibit largest O vii absorption — a hint that X-ray absorption may occur in regions where dust can survive (Komossa & Fink 1997).
If the UV and X-ray absorbers are related, then some of the assumptions used in previous modeling efforts must be modified in view of the detection of multiple kinematic components in Seyfert galaxies, in particular in NGC 5548, which exhibits 5 separate, blueshifted UV absorption components. Thus, we abandon the single zone model in favor of a multiple zone constrained by the EBS model. The constant density approximation is relaxed by estimating density for each of the kinematic components separately, by placing them at different distances from the central continuum source, based on kinematical constraints of the EBS model. We assume thermal and ionization equilibrium in the UVX gas (but see Krolik & Kriss 1995, and Nicastro et al. 1999). Target values for the total hydrogen column density, $`N_H`$, and the optical depths $`\tau (O\mathrm{vii})`$ and $`\tau (O\mathrm{viii})`$, which we convert to column densities, are adopted from the X-ray observations discussed in R97 and are listed in Table 1a.
In addition to the observations at X-ray energies, we also consider the HST UV GHRS observations of the 5 discrete, blueshifted absorption components of C iv, N v and H i detected in NGC 5548 on 1996 Aug. 24 and on 1996 Feb. 17 (C99) (see Table 1b). We adopt the ionic column densities measured from the GHRS observations over the STIS observations (Crenshaw & Kraemer 1999) because the former are closer in time to the ASCA X-ray observations of NGC 5548 on 1993 July 27 (R97). The single exception is for the column density of $`H\mathrm{i}`$ which is lacking in component 1 in the GHRS spectra so we adopt the value from the STIS data. Note that the UV observations are separated by seven months and the X-ray observations are separated from the UV observations by about 2 years. Extensive multiwavelength monitoring campaigns of this object (Clavel et al. 1991; Korista et al. 1995) show variations of UV emission lines which correlate with the continuum over times of days to weeks. C iv emission in NGC 5548 has, for example, varied in strength by a factor of $`2`$ in a month’s time and C iv absorption equivalent width may be similarly variable (Shull & Sachs 1993). Caution must therefore be taken when using noncontemporaneous data to determine the ionization state of the absorbing gas. We also point out that reported values of the ionic column density depend upon the procedure used to fit absorption line profiles. For example, given the same set of HST/GHRS spectra of the C iv absorption systems, MEW99 and C99 derived values of the C iv column densities of components $`13`$ which differed by $`0.20.5`$ dex. Next, modeling that goes into determining observationally-derived ionic column densities does not explicitly account for the possible effects of continuum scattering into the absorption troughs, although these effects do appear grossly in the estimate of the incomplete line-of-sight continuum coverage. This modeling must also make assumptions concerning the underlying emission from the narrow and BEL regions. Finally, the absorption lines are often observed to be saturated, yet non-black. The line-of-sight covering fraction of the ions along the various line-of-sight velocities must be estimated from the degree of saturation present. This is usually done by comparing the depths of the resonance line doublet pairs. Differences in line-of-sight effective continuum coverage amongst the various ions can be expected and are often found. The H i column densities reported by C99 and Crenshaw & Kraemer (1999) suffer from a lack of constraints in this respect (Ly$`\alpha `$ is a singlet, and no higher order Lyman line data are available), and should therefore be considered as lower bounds (Crenshaw, private communication). Thus the uncertainties in the ionic column densities reported in the literature are internal uncertainties only, the total uncertainties must be significantly larger.
Our adopted ionic column density set is listed in Tables 1a and 1b. Given the above considerations, we have disregarded their reported uncertainties, and instead assigned a factor of 2 uncertainty above and below their reported values. An exception is component 1 of C iv, which at the limit of detection is treated as a lower bound (Crenshaw, private communication). These assigned ranges of uncertainty about the reported ionic column density values, which we will refer to as observational constraints, will be used to find a plausible set of solutions to the UVX absorber in NGC 5548.
## 3 DYNAMICS AND EMISSION MODELING FOR AGNs
### 3.1 Hydromagnetic Wind Model
The magnetohydrodynamic (MHD) solution for a stationary, axisymmetric flow in cylindrical coordinates is given by Blandford & Payne (1982) and describes a self-similar, cold, nonrelativistic MHD flow from an idealized Keplerian disk around a point mass. This solution can be written in terms of variables $`\chi `$, $`\xi `$, $`\varphi `$ and a scaling parameter $`r_0`$ which are related to cylindrical coordinates via $`𝐫[r_0\xi (\chi ),\varphi ,r_0\chi ]`$. Here, $`\chi `$ is the coordinate along a field line, $`\xi (\chi )`$ is found as part of a self-consistent solution to the MHD equations, and $`r_0`$ is the field line footpoint on the disk. The flow velocity components are given by
$$𝐯=[\xi ^{}(\chi )f(\chi ),g(\chi ),f(\chi )]\sqrt{GM/r_0},$$
(2)
where a prime denotes differentiation with respect to $`\chi `$, and $`M`$ is the mass of the central black hole. The EBS provide for a non-trivial extension of Blandford & Payne solution for an arbitrary scaling of volume density $`nr_0^b`$ and magnetic field $`Br_0^{(b+1)/2}`$. At the base of the flow, the rotational velocity, $`v_\varphi `$, is Keplerian so $`v_\varphi r_0^{1/2}`$. The functions $`\xi (\chi )`$, $`f(\chi )`$ and $`g(\chi )`$ are chosen to satisfy the flow MHD equations subject to the above scalings of $`\rho `$, $`B`$ and $`v_\varphi `$. This implies that the Alfvén speed scales with the disk Keplerian velocity, and the specific angular momentum and energy in the flow will scale similarly to their Keplerian counterparts, while the disk mass loss per decade in radius scales as $`r_0^{(b1.5)}`$. The above velocity law is of course a non-Keplerian one at all finite distances from the disk (e.g., EBS and Paper 1).
EBS utilized an analytic approximation for $`\xi (\chi )`$ in which $`\xi (\chi )\chi ^{1/2}`$, and asymptotically tends to the Blandford & Payne solution. To attach the solution to the disk at $`r_0`$, $`\xi (\chi )`$ was constructed to have the form $`\xi (\chi )=\sqrt{\chi /c_2+1}`$, so that $`\xi (0)=1`$. The constant $`c_2=(\mathrm{tan}\theta _0)/2`$, where $`\theta _0`$ is the initial angle between the magnetic line and the disk. The MHD equations have been solved for $`f(\chi )`$ and $`g(\chi )`$, so as to be consistent with the analytic form of $`\xi (\chi )`$. The model parameters have been fixed in Paper 1 by modeling the BLR response to the variation of the central continuum (see section 3.2). We use $`b=1.5`$, so $`n`$, the particle density, scales as $`nR^{3/2}`$ and $`BR^{5/4}`$, where $`R`$ is the spherical radius from the central mass. We show below that this choice in $`b`$ is consistent with the variation in density inferred from observations, although the number of observational points is small. This result, therefore, should be taken with necessary caution. We have further neglected the effect of radiation pressure on the gas dynamics (e.g., de Kool & Begelman 1995), as it will not qialitatively change our results.
### 3.2 Constraining the MHD Flow with the BLR Spectrum
With an analytical solution for the volume emissivity EBS showed that realistic BLR emission line profiles can be produced. Paper 1 continued these efforts replacing the analytic emissivity function with an emissivity function obtained from a fit to an amalgam of optically thick clouds calculated by the photoionization code Cloudy (Ferland 1996). Emission line anisotropy and finite optical depth effects were introduced to model C iv emission from optically thick BLR clouds in the well studied object, NGC 5548. As part of the modeling, variations in C iv emission in response to observed variations of the continuum, including light travel time effects, were studied. A time series of synthetic C iv emission line profiles were generated and compared with the observed C iv emission line profiles from the HST (Korista et al. 1995), and a model fit to the data was thereby achieved. The BLR parameters in NGC 5548 deduced from the model were the central black hole mass ($`3\times 10^7M_{}`$), physical extent of the C iv emitting gas (1 to 24 light days) and the orientation of the observer line-of-sight relative to the axis of symmetry ($`40^{}`$). Here, we extend the results of Paper 1 and analyze the flow beyond the BLR and at larger latitudes above the disk, in particular as the flow crosses the observer’s line-of-sight.
In the ideal EBS flow, cold molecular gas is launched from a Keplerian disk and is flung along the magnetic field lines, like beads along a rotating wire. Most of the magnetic lines, however, are inclined unfavorably and hence will not be able to accelerate the gas. The centrifugally accelerated gas becomes illuminated by the central ionizing continuum at some latitude. The distribution of densities and ionization parameters, combined with the MHD model’s velocity field produces typically observed broad emission-line profiles. For NGC 5548, Paper 1 found that the BLR lies within a toroidal wedge $`\pm 30^{}`$ of the equatorial disk plane. The boundary of the wedge was interpreted as coming from optically-thick wind filaments experiencing thick-to-thin transition. In this section, we focus on the optically thin (to the ionizing continuum) flow, which is the continuation of the BLR flow, as it crosses the line-of-sight to the observer. Due to rather unfavorable conditions for loading and acceleration at the base of the wind, it is natural to expect a number of separate kinematic components along the line-of-sight. The five blueshifted UV absorbers reported in C99 and MEW99 for NGC 5548, while perhaps not fully distinct from one another, are interpreted here as kinematically distinct regions containing gas in an EBS-type magnetized wind. Moreover, each individual component is expected itself to be clumpy and is characterized by a volume filling factor, as we show below.
A point on a streamline is defined in terms of the footpoint radius $`r_0`$, the spherical radius $`R`$, and the observer’s aspect angle $`i`$ with respect to the disk axis, namely
$$r_0=Q(i,\theta _0)R,$$
(3)
where
$$Q(i,\theta _0)=\frac{\sqrt{\mathrm{cos}^2i+\mathrm{tan}^2\theta _0\mathrm{sin}^2i}\mathrm{cos}i}{\mathrm{tan}\theta _0}.$$
(4)
In Paper 1, for the best fit model of the BLR in NGC 5548, the launch angle $`\theta _0`$ and aspect angle $`i`$, were found to be $`20^{}`$ and $`40^{}`$, respectively. In this case, $`R10.4r_0`$. Figure 1 illustrates the model geometry for NGC 5548. A patch of gas filaments lifts off the disk between cylindrical radii $`r_1`$ and $`r_2`$ and rises along a helical trajectory, forming a parabola in the $`rz`$-plane. Gas that starts at $`r_1`$ eventually crosses the observer line-of-sight at a spherical radius $`R_1`$ and gas that starts at $`r_2`$ does this at $`R_2`$.
The number density of a total hydrogen along the flow line that starts at $`r_0`$ is prescribed by the model to be
$$n=\frac{n_A}{m}\left(\frac{r_0}{r_1}\right)^{1.5}.$$
(5)
Here $`n_A`$ is a normalization constant, setting the density $`n`$ on the innermost flow line at $`m=1.0`$, where $`m`$ is the Alfvén Mach number. The value of $`m`$ increases from unity at about $`17^{}`$ above the disk to about $`10^{3.3}`$ at $`85^{}`$ above the disk. The ionization parameter $`U1/nR^2`$, therefore, scales as $`UmQ^2(i,\theta _0)`$ along a flow line. At small latitudes, the effective ionization parameter for lower column density filaments will be even smaller, due to finite optical depth effects in the BLR (Paper 1). The change in radius and density causes the ionization parameter along any flow line to have a maximum at about $`30^{}`$ above the disk, where the BLR flow becomes optically thin. Here, the BLR flow is no longer self-shielded, becomes overionized and remains so at larger latitudes, where it crosses the observer’s line-of-sight. As a result, this gas does not contribute significantly to the UVX absorbing column. The gas filaments launched at larger radii than the BLR flow are less ionized and form the bulk of the UVX absorption. This absorbing gas is indicated by the thick line segment in Figure 1.
While the inner boundary of the wind is set by photoionization, the outer boundary can be fixed by considering the amplification of magnetic fields in the disk. An efficient generation of magnetic fields in the disk will cause its expulsion through the buoyancy effects and probable field reconnection above the disk (e.g., Galeev, Rosner & Vaiana 1979). The energy release during the reconnection can in principle inject disk material and load it onto the large-scale field treading the disk, resulting in the hydromagnetic wind. While the large-scale disk-threading poloidal field is probably governed by global considerations, the disk-generated fields are expected to be amplified by differential rotation. Such dynamo effect will operate only within the “radius of influence,” $`r_{BH}`$, of the central black hole (BH), because for $`r>r_{BH}`$ the inner galactic rotation is typically that of a solid body. Hence, the outer wind cutoff can plausibly result from a dramatic decrease in the shear outside $`r_{BH}`$, which prevents the magnetic field in the disk from reconnecting and loading the molecular material onto large-scale field lines. This radius can be estimated from equating the gravitational potentials of the BH and that of the host galaxy, namely, $`r_{BH}1.3\times 10^{19}m_7v_2^2`$ cm, where $`m_7M_{BH}/10^7\mathrm{M}_{}`$ and $`v_2v_K/100\mathrm{km}\mathrm{s}^1`$ is the rotational velocity in the galactic disk. Adopting the BH mass in NGC 5548 from Paper 1, leads to $`R_21.4\times 10^{20}`$ cm for a typical $`v_7=2.5`$. As we show below in $`\mathrm{\S }`$ 3.3.1 this is within a factor of $`23`$ from the outer boundary of the observed kinematic components.
### 3.3 Multi-Component Absorption Model for NGC 5548
#### 3.3.1 Location of Absorption Components
The positions of separate absorption components are calculated using the relationship between $`R`$ and $`v_{obs}`$, the line-of-sight component of wind velocity. The velocity $`v_{obs}`$ is obtained by projecting the velocity field onto the observer line-of-sight and is given by
$$v_{obs}=[\xi ^{}(\chi )f(\chi )\mathrm{sin}i+f(\chi )\mathrm{cos}i]\sqrt{GM/r_0}.$$
(6)
Substituting $`r_0=Q(40^{},20^{})R`$ and $`M=3\times 10^7\mathrm{M}_{}`$ for NGC 5548 (Paper 1), gives
$$R=7.3\times 10^{18}\left(\frac{v_{obs}}{10^3\mathrm{km}\mathrm{s}^1}\right)^2\mathrm{cm}.$$
(7)
Observed component velocities in the rest frame of NGC 5548 from C99 and the FWHM of each component which we use below are listed in Table 3b. Inserting the observed component velocities into Equation 7, yield component distances which range from about 2 to about 87 parsecs (Table 3b). The most important consequence is that the UVX gas in NGC 5548 appears to be positioned well outside its BLR, both in radii and in latitude. Hence, within the framework of the EBS model, which neglects the effects of radiation pressure, the the UVX absorbing gas originates in the outer part of the wind, beyond the BLR flow.
#### 3.3.2 Calculation of Possible Ionization States
The observational constraints on the ionic column densities and their ratios in section 2 are now used to find the limits on the ratios of the N v ion fraction, $`f(N\mathrm{v})`$, with the C iv ion fraction, $`f(C\mathrm{iv})`$. These estimates allow the determination of a set of possible gas ionization states, i.e., the particle density, electron temperature and various ion fractions in each of the UVX absorption components. We assume a total hydrogen number density $`n`$ as given by equation 5, so the column density $`N(X_i)`$ of ion $`X_i`$ is
$$N(X_i)=_{\mathrm{\Delta }R}ϵf(X_i)a_Xn𝑑Rϵf(X_i)a_X_{\mathrm{\Delta }R}n𝑑R,$$
(8)
where $`ϵ`$ is the volume filling factor in the gas column which covers the central continuum source, $`f(X_i)`$ is the ion fraction, $`a_X`$ is the abundance of element $`X`$, and $`\mathrm{\Delta }R`$ is the line-of-sight geometrical width of the cloud. If the ionization parameter is a slowly varying function of radius with little change over $`\mathrm{\Delta }R`$ and $`ϵ`$ is constant within each component, then $`ϵf(X_i)a_X`$ can be taken out of the integral. Then $`f(N\mathrm{v})/f(C\mathrm{iv})`$ for each component is given by
$$\frac{N(N\mathrm{v})}{N(C\mathrm{iv})}\frac{f(N\mathrm{v})}{f(C\mathrm{iv})}\frac{a_N}{a_C}.$$
(9)
Assuming solar abundances, $`a_N=9.33\times 10^5`$ and $`a_C=3.55\times 10^4`$, we obtain the ion fraction ratio
$$\frac{f(N\mathrm{v})}{f(C\mathrm{iv})}3.805\frac{N(N\mathrm{v})}{N(C\mathrm{iv})}.$$
(10)
The central values of $`\mathrm{log}[f(N\mathrm{v})/f(C\mathrm{iv})]`$ for components $`15`$, assuming the column densities listed in C99 and solar abundances, are listed in the final column of Table 1. In accord with section 2, we assign factors of two uncertainty to their values. Component 1 is a special case because the reported values for $`N(C\mathrm{iv})`$ provide a lower bound only.
The range of possible ionization states for each component was determined using Cloudy (version 90.04). The luminosity and spectral energy distribution (SED) of the continuum source in NGC 5548, combined with the distance of each component (Eq. 7), is used to calculate the incident flux $`\mathrm{\Phi }(H)`$ of hydrogen ionizing photons falling on each absorption component. We used the following continuum shape for the SED of NGC 5548, i.e.,
$$f_\nu \nu ^{\alpha _{UV}}e^{h\nu /kT_{BB}}e^{kT_{IR}/h\nu }+a\nu ^{\alpha _X}$$
(11)
and
$$\frac{f_\nu (2\mathrm{keV})}{f_\nu (2500\mathrm{\AA })}=403.3^{\alpha _{OX}},$$
(12)
where $`kT_{IR}=0.01`$ Ryd is the spectrum cutoff in the near-infrared, and $`a`$ in Equation 11 is chosen to satisfy Equation 12.<sup>1</sup><sup>1</sup>1Outside the range 1.36 eV$``$100 keV, $`a=0`$. Above 100 keV, we assume $`f_\nu \nu ^3`$. The input parameters are $`\mathrm{log}(T_{BB})=6.683`$, $`\alpha _{OX}=1.20`$, $`\alpha _{UV}=1.20`$ and $`\alpha _X=0.90`$, chosen to be in accord with IUE/ROSAT observations of the NGC 5548 continuum (Walter et al. 1994). Using the above SED, the ionizing luminosity of the continuum source is estimated at $`10^{44.3}\mathrm{erg}\mathrm{s}^1`$ ($`H_0=75\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$), based upon the mean continuum flux at 1350Å, obtained during the 1993 observing campaign (Korista et al. 1995).
With the above assumptions, the ionization parameter $`U`$ ($`\mathrm{\Phi }[H]/nc`$), which controls the ionization state of the gas, is tuned by adjusting the hydrogen number density $`n`$ in a series of Cloudy calculations for an optically-thin slab of width $`10^{10}\mathrm{cm}`$. In this manner, a set of synthetic photoionization gas states covering a range of possible $`f(N\mathrm{v})/f(C\mathrm{iv})`$ values is produced for each component.
In order to provide for a complete description of the UVX absorption in each kinematic component, it is necessary to know the column densities of different ions. Note that, if we replace $`N(N\mathrm{v})`$ with $`N(X_i)`$ and $`a_N`$ with $`a_X`$ in Equation 9, we obtain a generic expression for the column density of ion $`X_i`$. In the approximation of Equation 9, the quantity $`[f(X_i)/f(C\mathrm{iv})](a_X/a_C)`$ is constant, so $`N(X_i)N(C\mathrm{iv})`$. Thus for fixed $`f(N\mathrm{v})/f(C\mathrm{iv})`$, all ionic column densities scale directly with the C iv column density. The parameters $`f(N\mathrm{v})/f(C\mathrm{iv})`$ and $`N(C\mathrm{iv})`$ then map out the set of possible ionic column densities for each component. We use these two parameters to determine the electron temperature $`T_e`$ and column densities of N v, O vi, O vii, O viii, H i and H, on a grid of $`[`$f(N v)/f(C iv)$`,N(C\mathrm{iv})]`$ pairs that are consistent with the observational constraints.
The grid of solutions constrained by f(N v)/f(C iv) and N(C iv) were used to determine the range of possible values of the column densities of O vi, O vii, and O viii in each component. The factor of 2 uncertainty in the observational constraints on the C iv and N v column densities and their ratios translates into column density uncertainties of a factor of $`10`$ on O vi, a factor of $`100`$ on O vii and a factor of $`1,000`$ on O viii. The inferred ranges on the column densities of H and H i are no better. H i ranges over a factor of $`100`$ and H over a factor of $`1,000`$ (even more in component 1). We, therefore, obtain an optimal solution for UVX absorption in NGC 5548, within the five component sets of possible solutions. Since individual column densities of C iv and N v for each component are observed, but only total column densities are available for O vii and O viii, we search for a solution that minimizes the chi-squares, $`\chi _O^2`$, for oxygen ion column densities, namely
$$\chi _O^2=\frac{(_{j=1}^5N(O\mathrm{vii})_jN(O\mathrm{vii})_{obs})^2}{\sigma _{O\mathrm{vii}}^2}$$
$$+\frac{(_{j=1}^5N(O\mathrm{viii})_jN(O\mathrm{viii})_{obs})^2}{\sigma _{O\mathrm{viii}}^2}.$$
(13)
Here $`N(\mathrm{O}\mathrm{vii})_j`$ and $`N(\mathrm{O}\mathrm{viii})_j`$ are the calculated column densities of O vii and O viii in a component $`j`$ (see Table 3), $`N(\mathrm{O}\mathrm{vii})_{obs}`$ and $`N(\mathrm{O}\mathrm{viii})_{obs}`$ are the observed total column densities of O vii and O viii (see Table 1), and $`\sigma _{O\mathrm{vii}}`$ and $`\sigma _{O\mathrm{viii}}`$ are the adopted observational uncertainties (section 2) in the O vii and O viii column densities. Our search of the solution grid yields optimal $`\chi _O^2=0.13`$. The total column density of O vii is overpredicted by $`30\%`$ and the total column density of $`O\mathrm{viii}`$ is underpredicted by $`9\%`$, compared to target values in Table 1a, placing both well within the observational constraints.
#### 3.3.3 An Optimal Solution for UVX Absorption using the EBS Model
The model column densities of C iv, N v, O vi, O vii, O viii, H i and H in each component and their sum over all components are listed in Table 3a. Component C iv and N v column densities automatically satisfy the observational constraints because the optimization was done on a grid of values bounded by the constraints. The component column density values for O vi, O vii, O viii, and H in each component are predictions of our model. The good match between the predicted and observed total O vii, O viii, and hydrogen column densities, given present observational constraints, indicates that the soft X-ray WA and the UV resonance-line absorber are directly related. The EBS model constrains the bulk of the WA to lie within two of the kinematic “components” of the outflow, separated by about $`500\mathrm{km}\mathrm{s}^1`$ in line-of-sight velocity. While observations at the required resolution do not yet exist to test these predictions in detail, the recently launched X-ray satellite Chandra may be able to detect the O vii and O viii absorption edges of the individual kinematic absorption components in NGC 5548, and the recently launched FUSE satellite should at least detect the presence of absorption in O vi.
The predicted column densities of H i from our optimal solution do not match well their reported values (Table 1b). The observed component column density values are factors of $`428`$ below their counterparts in the optimal solution. However, as mentioned in section 2, the observationally inferred H I column densities are the least certain, and are lower limits to their true values, at best.
The hydrogen number density for each component in the optimal solution as well as $`T_e`$, $`U`$ and other physical values are listed in Table 3b. The density derived for each component using Cloudy depends on the kinematic structure of the EBS model via Equation 7, but is independent of the parameter $`b=1.5`$ used to describe the scaling of the EBS density in Equation 5. It is therefore possible to test whether the assumed value of $`b`$ used in Paper 1 and here in this paper is appropriate.
Equations 3–5 together imply that along any ray through the origin $`nR^{1.5}`$, and, therefore, the ionization parameter varies as $`UR^{0.5}`$. If individual absorption components obey this density prescription, then $`n`$ and $`U`$ from component to component should vary accordingly. As a test, we fit $`\mathrm{log}(n)`$ and $`\mathrm{log}(U)`$ from the optimal solution at the center of each absorption component, as functions of $`\mathrm{log}(R)`$ with a linear least squares fit. Figure 2 shows $`\mathrm{log}(n)`$, $`\mathrm{log}(U)`$ and the least squares fit of both. The range in uncertainty is $`\pm 0.2`$ and corresponds to our photoionization grid’s resolution in gas density. We find that the linear least squares fits to $`\mathrm{log}(n)`$ and $`\mathrm{log}(U)`$ vs. $`\mathrm{log}(R)`$ have slopes of $`1.57\pm 0.16`$ and $`0.45\pm 0.16`$, respectively. These are consistent with our specific EBS assignment of $`nR^{1.5}`$ and $`UR^{0.5}`$. Since $`U`$ is a slowly decreasing function of $`R`$, our simplification of the integral in Equation 8 is reasonable, assuming that there are no significant optical depth effects along the absorbing column and that in each component $`\mathrm{\Delta }R/R`$ is sufficiently small, so that there are no significant changes in $`U`$ over the width of a component either. The picture that emerges for the wind model of NGC 5548 is that of optically thin filaments originating over a large range of radii in the disk, from $`0.2`$ pc to about 9 pc. These filaments being centrifugally accelerated along the magnetic field lines, cross the line-of-sight about ten times further out. Along the line-of-sight, the ionization parameter drops as a weak function of distance ($`UR^{1/2}`$), but because the components are spread out over a factor of $`40`$ in radius, there is a factor of $`6`$ drop in ionization parameter so the wind is ionization-stratified along the line-of-sight. The thermal stability of kinematic components in NGC 5548 is analyzed in the Appendix.
#### 3.3.4 Absorption Component Widths and Volume Filling Factors
A more complete physical picture of the UVX absorption system is provided by determining the line-of-sight width, $`\mathrm{\Delta }R`$, of each absorption component. In this section we estimate the ratio $`\mathrm{\Delta }R/R`$, discuss the radial gas distribution, separation of individual components, their degree of clumpiness, variation of $`U`$, etc.
The line-of-sight extension of each kinematic component can be estimated from the FWHM of the absorption components (given in C99) and the electron temperature, provided by the optimal solution (Table 3). The FWHM is attributed to the effects of thermal Doppler broadening and velocity gradients in the MHD flow. The component due to velocity gradients, $`\mathrm{\Delta }v`$, is obtained by making a correction to the FWHM by subtracting the thermal broadening which we take to be $`v_s=\sqrt{2kT_e/m_i}`$, and by taking the ion mass $`m_i`$ to be the average mass of carbon and nitrogen atoms. This average is taken because the FWHMs listed in C99 were the result of averaging the FWHM determined from the N v and C iv absorption lines. Table 3b lists $`v_s`$ and $`\mathrm{\Delta }v`$, defined by $`\mathrm{\Delta }v=FWHMv_s`$. All of the $`\mathrm{\Delta }v`$ values are positive, therefore, some of the line width cannot be attributed solely to thermal Doppler broadening. Equation 7 is used to determine the spatial extent $`\mathrm{\Delta }R`$ of each component via
$$\mathrm{\Delta }R=R\left(v_{obs}\frac{\mathrm{\Delta }v}{2}\right)R\left(v_{obs}+\frac{\mathrm{\Delta }v}{2}\right)=$$
$$4R\left(\frac{\mathrm{\Delta }v}{2v_{obs}}\right)\left(1\left[\frac{\mathrm{\Delta }v}{2v_{obs}}\right]^2\right)^2.$$
(14)
Numerical values for $`\mathrm{\Delta }R`$ in each component are also listed in Table 3b. Figure 3 shows the width and location of each component for the optimal solution plotted in $`\mathrm{log}(R)`$. Gaps between the components indicate that they are well separated (except for possibly components 3 and 4), and that $`\mathrm{\Delta }R`$ values grow monotonically with $`R`$. For the first three components $`\mathrm{\Delta }R/R<0.32`$, but it is of order unity for components 4 and 5. According to the EBS model, the ionization parameter will vary across the component by $`U_{front}/U_{back}=\sqrt{(R+\mathrm{\Delta }R/2)/(R\mathrm{\Delta }R/2)}`$. The values of $`R`$ and $`\mathrm{\Delta }R`$ for each component (Table 3b) yield percentage differences in $`U`$ measured at the front and back sides of each component. These are $`11\%`$, $`6\%`$, $`17\%`$, $`71\%`$ and $`61\%`$ for components 1 through 5, respectively. Thus our approximation in Equations 8, 9 and 10 is reasonable only for the first three components. For components 4 and 5 we estimate that the effect will result in an error of about $`0.1`$ dex in the N v and C iv column densities, and as much as $`+0.4`$ dex uncertainty in the oxygen column densities. Fortunately, these errors are minor in those components where oxygen ions predominate (i.e., 1 and 3), and, therefore, the breakdown of our assumption of constant $`U`$ across individual components does not affect much the predicted integrated oxygen column densities.
An estimate for the volume filling factor is obtained from the column density. Using $`nR^{1.5}`$ as a density profile, the integral in Equation 8 is given by $`_{\mathrm{\Delta }R}n𝑑R=n(R)2R[(1\mathrm{\Delta }R/2R)^{1/2}(1+\mathrm{\Delta }R/2R)^{1/2}]`$, where $`n(R)`$ and $`R`$ are the density and radius at the center of each component of width $`\mathrm{\Delta }R`$. Hence
$$N(X_i)$$
$$ϵf(X_i)a_Xn(R)2R[(1\mathrm{\Delta }R/2R)^{1/2}(1+\mathrm{\Delta }R/2R)^{1/2}].$$
(15)
If $`n(X_i)`$ is the density of ion $`X_i`$ at $`R`$, then
$$n(X_i)=f(X_i)a_Xn(R).$$
(16)
Substituting Equation 16 into Equation 15, letting $`q=\mathrm{\Delta }R/2R`$ and solving for $`ϵ`$ gives
$$ϵ\frac{N(X_i)}{n(X_i)\mathrm{\Delta }R}\frac{q(1q^2)^{1/2}}{(1+q)^{1/2}(1q)^{1/2}}.$$
(17)
The quantity $`N(X_i)/n(X_i)\mathrm{\Delta }R`$ is the filling factor, when the density is constant throughout the component. The term involving $`q`$ makes a correction for the variation in density across a component. The correction factors, however, are close to unity. For components 1 through 5, they are 0.994, 0.998, 0.984, 0.844 and 0.871, respectively. This requires the choice of an ion to define $`ϵ`$, and we choose C iv since it is already used as one of the solution parameters. We note that N v or even a blend of the C iv and N v column densities could be used as well, but the differences in calculated $`ϵ`$ are not significant.
Calculated values of $`\mathrm{log}(ϵ)`$ for the optimal solution range from $`5.58`$ to $`1.37`$ (Table 3b). This means that an absorption component is clumpy and has a filamentary structure. In other words, each of the five distinct absorption components in NGC 5548 consists of filaments which move along similar trajectories and have probably a common origin in the disk. On the other hand, we presume that the distribution of points of origin of any given absorption component in the disk plane is random. The clumpiness of individual components combined with the geometry and non-radial nature of the MHD flow has an important implication for the longevity of UV absorption structures in AGNs, which we discuss in the next section.
#### 3.3.5 Temporal Changes in the UVX Absorbing Column
Additional information about the character of the absorbing column of gas is inferred from the temporal behavior (or lack thereof) of UV and X-ray absorption. Changes in absorption lines in principle can be due to the motion of material into or out of the observer’s line-of-sight, or it may be due to changes in the ionization structure of the gas brought about by changes in the incident continuum or both. The X-ray absorption line variability has been discussed in Reynolds & Fabian (1995), Reynolds (1997) and Otani et al. (1995), where ASCA observations of MCG 6-30-15 show apparent anti-correlation of the O viii edge depth with an increase in the continuum level. A similar conclusion was reached by Reynolds & Fabian (1995). Observations of UV absorption lines in AGNs also reveal that they change with time. Shull & Sachs (1993) used IUE data to show that the C iv absorption equivalent width in NGC 5548 responds to continuum changes in an anti-correlated fashion on a time scale less than about 4 days (this value is likely to be rather uncertain due to the fact that the temporal sampling of this monitoring campaign was four days). UV observations of other AGNs spaced over many months reveal that absorption lines change in depth with time, but the relationship with the continuum is not clear. While the depth of an absorption line may vary, there are no unambiguous observations of a change in the velocity centroid of the line (Weymann 1997), which argues against purely radial motion, since one would expect clouds moving only radially to accelerate along the line-of-sight, resulting in time with a shift in the velocity centroid. This is in agreement with our modeling of gas dynamics in the BLR of NGC 5548 (Paper 1) using the EBS wind, which ruled out a purely radial motion as well. Here we focus on the temporal phenomena associated with UV and X-ray absorption in AGNs.
Our analysis of the discrete absorption components in the previous section suggests that each component is an agglomeration of filaments moving along the same trajectory. The heuristic model is that the wind filaments are confined by the ambient magnetic field and are extended along the field lines. The exact details of gas loading onto the magnetic lines are subject to future work. Filaments which rotate out of the line-of-sight are replace by their neighbors, so the velocity centroid of the absorption component is not expected to change with time. The time scale $`t_c`$ over which we expect an absorption structure to survive is given by $`t_c=\mathrm{\Delta }R/v_\varphi `$, where $`\mathrm{\Delta }R`$ is the width and $`v_\varphi `$ is the rotational velocity of the absorption component. For the first UV absorption component in NGC 5548, where $`t_c`$ has the smallest value, we find $`v_\varphi 10^3\mathrm{km}\mathrm{s}^1`$ and $`\mathrm{\Delta }R10^{19}\mathrm{cm}`$, leading to $`t_c3,000`$ years. The proposed model can, therefore, explain the observed long constancy of the velocity centroids of UV absorption structures.
Given the long dynamical time scale, observed temporal changes in the absorbing column are attributed instead to changes in the photoionization structure of the absorbing gas, rather than a change in gas density along the line-of-sight, and are presumably due to fluctuations in the incident continuum. The response of the absorption within the EBS wind to continuum fluctuations is investigated by considering the effect of a sudden drop in the continuum level on the column density of C iv, O vii and O viii. C iv is chosen to compare with observation (Shull & Sachs 1993), and O vii and O viii are chosen to compare and contrast the response of our model of NGC 5548 with observations of MCG 6-30-15.
In the above scenario, ions will begin to recombine with the drop in the continuum intensity. However, the time scale for recombination as given by Equation 1 does not account simultaneously for the cascade into the population of $`X_i`$ ions from the population of $`X_{i+1}`$ ions, and the cascade out of the population of $`X_i`$ ions into the population of $`X_{i1}`$ ions. Equation 1 is hence misleading and is replaced for consistency by
$$t(X_i)=\frac{1}{\alpha (X_i)n_e[\frac{f(X_{i+1})}{f(X_i)}\frac{\alpha (X_{i1})}{\alpha (X_i)}]}.$$
(18)
The value of $`t(X_i)`$ in Eq. 18 establishes the minimum time, that gas will respond to a decrease in the continuum, and depends on the local density and on the ion population. As a single measure within a component, we take the average ion density-weighted recombination time which is
$$<t(X_i)>=\frac{_{\mathrm{\Delta }R}t(X_i)n_{X_i}ϵ𝑑R}{N_{X_i}}.$$
(19)
This definition weights more heavily recombination times in regions with higher column density contributions. Assuming the validity of Equations $`810`$, $`n_{X_i}`$,$`n_{X_{i+1}}`$ and $`n_e`$ are all proportional to $`R^{1.5}`$. In addition, $`T_e`$ within each component is approximately independent of $`R`$ and, therefore, $`\alpha (X_i)`$ and $`\alpha (X_{i1})`$ are also independent of $`R`$. The result is that $`t(X_i)n_{X_i}`$ is independent of $`R`$ and may be factored out of the integral. This gives
$$<t(X_i)>=\frac{t(X_i)_Rn_{X_i}(R)ϵ\mathrm{\Delta }R}{N_{X_i}},$$
(20)
where $`t(X_i)_R`$ and $`n_{X_i}(R)`$ are $`t(X_i)`$ and $`n_{X_i}`$, respectively, evaluated at $`R`$. Hence, $`<t(X_i)>`$ may be positive or negative, depending on whether or not the column density of ion $`X_i`$ is anti-correlated or correlated with a decrease in the ionizing continuum, respectively.
Calculated values of $`<t(X_i)>`$ for C iv, O vii and O viii in NGC 5548 are listed in Table 4. Not surprisingly, these recombination time scales differ from those calculated from Equation 1 — an issue sometimes ignored in the literature. For a fixed electron density, the ionic distributions as well as the effect of recombinations to lower ionization states can produce a wide range of the recombination time scales for the various metallic ions. In Table 4, the computed value of $`<t(X_i)>`$ for each ion differs from component to component. The recombination time scale for C iv spans a factor of $`5,000`$. The ranges for O vii and O viii, span factors of $`600`$ and $`50`$, respectively. Whether or not changes of ion $`X_i`$ in a single component have a significant effect on the total ionic column density over all components of ion $`X_i`$, depends on the relative contribution of $`X_i`$ in the single component. (All reported variations in ionic UVX column densities were measured over the total value, due to insufficient spectral resolution.) For this reason the percent contributions of C iv, O vii and O viii are also shown in Table 4. The values are used to define the column density-weighted mean recombination time over all components, $`<<t(X_i)>>`$ (Table 4). The individual recombination times in each component serve as lower bound response time scales to changes in the ionizing continuum flux, and $`<<t(C\mathrm{iv})>>`$ may be compared in a crude manner with observations. In our model, $`<<t(C\mathrm{iv})>>+3.7`$ days is reasonably consistent with the results of Shull & Sachs (1993), namely, that C iv absorption equivalent width is anti-correlated with continuum variations with a delay of roughly four days.
Long term X-ray observations of NGC 5548 do not yet exist, so the time scales reported in Table 4 for O vii and O viii cannot be tested. Instead we compare the results of this model to the observations of MCG 6-30-15. In our model of the UVX absorption flow in NGC 5548, the bulk of the contribution to the column density of these two oxygen ions resides in components 1 and 3. However, the values of $`<<t(O\mathrm{vii})>>`$ and $`<<t(O\mathrm{viii})>>`$ are considerably skewed by the extremely long recombination time scale of component 5. As a conservative measure, we, therefore, consider only components 1 and 3. In these components, the column density contributions are roughly equal for both O vii and O viii, but the recombination times in the first component are over 10 times shorter. These recombination times, which represent the minimum possible response time scale for a continuum drop, are $`<t(O\mathrm{vii})>3`$ days and $`<t(O\mathrm{viii})>50`$ days. (The O vii and O viii edge optical depths are correlated with the ionizing continuum intensity.) We contrast this with observations of MCG 6-30-15 (see section 2). During the second half of the $`4`$ day observation period of this object, the X-ray continuum dropped by a factor of 2. The response was that the O viii optical depth anti-correlated with the continuum on a time scale of $`10^4`$ s, but there was no observed change in the O vii optical depth.
A possible explanation for the observed behavior of the O viii edge optical depth in MCG 6-30-15 is that the absorbing gas in MCG 6-30-15 is more highly ionized than in NGC 5548. The observed continuum in MCG 6-30-15 is much harder than in NGC 5548, and the WA gas may, therefore, support a large population of O ix ions. Equation 18 allows then for a complex dependency of the recombination time scales, which might explain the differences between the predicted behavor of NGC 5548 and the observed behavior of MCG 6-30-15. Long looks by Chandra and other advanced X-ray telescopes should establish the relationships between the continuum and WA optical depths in AGNs.
## 4 A GENERIC MODEL FOR WARM ABSORBING GAS
The discrete UVX absorber model for NGC 5548 does not allow for its direct application to other objects. This is mainly because detailed kinematic data is not available, the model does not provide for the UVX absorption dependence on the aspect angle, and because it does not specify the appropriate boundary conditions not along the line-of-sight. To generalize the model, we proceed by spatially smoothing the contributions from individual UVX kinematic components of our model and by calculating the optical properties of the MHD flow along a number of aspect angles. The resulting generic model is used to estimate orientation effects on the soft X-ray absorption properties of AGNs. The general aspects of thermal stability of the WA gas in this model are discussed in the Appendix.
### 4.1 Column Densities and physical conditions
We first build a quasi-continuous model of MHD flow in NGC 5548 by smoothing the kinematic components along the line-of-sight, by imposing observational constraints on the total column density of hydrogen, and by retaining the $`R^{3/2}`$ density profile from the discrete component model. For simplicity, the outer edge of the continuous flow was set at $`\mathrm{log}R20.6`$ which coincides with the outer edge of the fifth component. Two free parameters, i.e., the volume filling factor and the inner radius, which we used as Cloudy input parameters, were tuned to fit observations of the column densities of $`N(C\mathrm{iv})`$, $`N(N\mathrm{v})`$, $`N_H`$, $`N(O\mathrm{vii})`$ and $`N(O\mathrm{viii})`$. A reasonable fit to the ionic column densities is found for $`\mathrm{log}(R_1)17.7`$ and $`ϵ10^3`$ (Table 2). The first plot in Figure 4 clearly illustrates ionization stratification, with the ion fraction O viii achieving a maximum at smaller radii, O vii at intermediate radii, and both C iv and N v peaking near the outer edge. The second plot in Figure 4 shows the density and electron temperature as a function of distance. Note the temperatures range from $`1.6\times 10^4\mathrm{K}`$ at larger radii to $`2\times 10^5\mathrm{K}`$ at smaller radii, where the bulk of the WA resides.
### 4.2 Orientation Effects in AGNs and Soft X-Ray Absorption
We now investigate the orientation effects in the continuous wind model of NGC 5548 by varying the viewer aspect angle. Figure 5 shows the column densities of $`N(N\mathrm{v})`$, $`N(C\mathrm{iv})`$, $`N(H)`$, $`N(O\mathrm{vii})`$ and $`N(O\mathrm{viii})`$ as a function of viewer aspect angle $`i`$. The density profile along each aspect angle was adjusted so that the MHD flow is normalized to the solution for NGC 5548 (Table 2), which corresponds to the aspect angle $`i=40^{}`$. Absorbing column densities seen by an observer with aspect angle $`i>60^{}`$ are within the angular wedge shadowed by the BLR, i.e., by hydrogen column densities in excess of $`10^{23}\mathrm{cm}^2`$. Although we restrict our discussion to angles at $`60^{}`$ or below, and to smaller column densities compatible with those in NGC 5548, larger column densities cannot be excluded for other objects. We adopt the generic model for NGC 5548 as a template, when applied to other Seyfert 1 galaxies.
Figure 5 shows that $`N(O\mathrm{vii})`$ and $`N(O\mathrm{viii})`$ decrease with decreasing $`i`$, making detection of the oxygen edges more difficult for small $`i`$. Using the instrument error bar in R97, and assuming a $`2\sigma `$ detection criteria, absorption edges of O vii and O viii will be lost for aspect angles less than $`i30^{}`$ with the disk axis. Since no objects will be seen for $`i>60^{}`$, due to obscuration, we restrict our analysis to angles $`i<60^{}`$. In this case, if all Seyferts are taken as randomly oriented templates of NGC 5548, then about $`73\%`$ of them will have detectable WA gas.
Figure 5 also shows that $`N(O\mathrm{vii})`$ drops faster than $`N(O\mathrm{viii})`$ when viewed at smaller aspect angles, and drops below $`N(O\mathrm{viii})`$ for angles less than $`50^{}`$. The fraction of Seyfert 1 galaxies with detectable WA showing $`N(O\mathrm{viii})>N(O\mathrm{vii})`$ can be estimated from the ratio of the solid angle subtended on a sphere between the detection limit $`30^{}`$ and $`50^{}`$, and the solid angle between $`30^{}`$ and $`60^{}`$. This ratio is about $`61\%`$, for current instrument sensitivity. For comparison, 8 out of 16 Seyfert 1 galaxies listed in R97 (50%) have $`N(O\mathrm{viii})>N(O\mathrm{vii})`$. (We used only Seyfert 1s from R97 that have definite error bounds on both the tabulated values of optical depth $`\tau _{O\mathrm{vii}}`$ and $`\tau _{O\mathrm{viii}}`$.) Finally, we note that the predicted range of $`N_H`$ varies from $`10^{21}\mathrm{cm}^2`$ to $`10^{23}\mathrm{cm}^2`$ within the observable range of aspect angles, and this compares favorably with the inferred range of $`N_H`$ in R97. Of course, one can question this approach, when one object is used as a template to describe the whole population of Seyferts, and the Seyfert sample in R97 is small. The model’s consistency with the observations in these respects is nevertheless interesting. Taken at face value, the ratio of optical depths of O vii and O viii ions can serve as a new diagnostic for AGN aspect angles.
### 4.3 Diffuse Emission from the UVX Absorber
The UVX absorbing gas reprocesses and scatters the incident continuum into diffuse emission across the UV and soft X-rays. We used our “generic” UVX absorber for NGC 5548 to estimate the magnitude of this emission. We computed Cloudy models for thermal local line widths and one, in which microturbulence broadened the local line width ($`\sigma _{turb}=100\mathrm{km}\mathrm{s}^1`$). In the latter scenario, optically thick lines are desaturated, and the cross-section for photon pumping of the line’s upper level (continuum resonance line scattering) is elevated, enhancing the line intensities. Photon pumping may be especially important for the soft X-ray lines because of their small Boltzmann factors in photoionized gas. However, even under the assumption that the observer can see the gas represented by our generic UVX absorber distributed over $`4\pi `$ steradians over the continuum source, the diffuse emission is relatively small. This is expected, given the reported small optical depths of the absorber in this object. We find that this gas contributes at most $``$ 20% to the observed UV narrow line fluxes of Ly$`\alpha `$, C iv, and N v in NGC 5548 (Goad & Koratkar 1998). Emission of this significance may be sufficient to partially fill in some of the absorption troughs. Depending upon the covering fraction and the contribution from photon pumping the diffuse soft X-ray line emission could be also be significant in comparison to the absorption edge depths (Netzer 1993). We will have to await the results from the upcoming missions of Chandra, XMM, and Constellation-X to get hard observational constraints from this portion of the spectrum.
## 5 SUMMARY
We have applied the EBS hydromagnetic wind model from a clumpy molecular accretion disk to the problem of UVX-ray absorption systems observed in Seyfert galaxies. We have studied the state of the UVX absorbing gas, with a particular emphasis on NGC 5548, where five discrete UV absorption systems seen in this object were modeled. Extending our previous work on the dynamics of the BLR gas in NGC 5548 (Paper 1), we first inferred the location of each component relative to the continuum source. By estimating the ratio of the ion fractions of N v to C iv from the observed column densities and the use of photoionization modeling, we have determined possible values for the density, electron temperature and the ion fractions of C iv, N v, O vi, O vii O viii, H and H i. Optimization over the set of possible values yielded the ion column densities in each kinematic component and allowed us to test the radial and polar density distributions of the WA in NGC 5548, within the EBS framework. We have also estimated the size and volume filling factor of each component, using the FWHMs and electron temperatures, and found that the flow in each component is clumpy. X-ray absorption by O vii and O viii column densities seen in NGC 5548 have been accounted for in the five UV absorption components, though main contribution to the X-ray absorption comes from two components. We, therefore, were able to explain the UVX absorption columns within the framework of the same dynamical model. An additional important point is that the model parameters used here are the best fitting parameters from Paper 1 to explain the broad emission-line variability in NGC 5548 during the 1989 and 1993 observing campaigns.
Secondly, we find that the WA gas in NGC 5548 lies at larger radii from the central source than the BLR gas and at larger altitudes above the disk. This means that WA exists in the outer parts of the hydromagnetic disk wind, and as such is not a continuation of the BLR flow. As a result we do not expect it to contribute to absorption along the line of sight unless the aspect angle is near the obscuring torus. With regard to the spatial extent of the WA gas we find it extends in both radial and polar directions from the central continuum source, and is ionization stratified.
Thirdly, we have modeled the WA in NGC 5548 also as a continuous flow, in order to investigate the generic properties of the UVX absorption outflows in Seyfert galaxies. We find that orientation effects may fully account for the fraction of objects with detectable WAs. Furthermore, the model predicts that the ratio of optical depths of O viii to O vii can serve as a new diagnostic of orientation of an AGN, namely of its rotation axis — an issue of a particular interest in AGN theory. In addition, we find that the UV line emission from the UVX flow probably accounts for no more than $`20\%`$ of the NLR emission in lines, such as Ly$`\alpha `$, C iv, N v, and O vi, though diffuse emission in the soft X-rays may be significant under some conditions.
A thermal stability analysis has been carried out for the model. We find that all of the five components in NGC 5548 are stable, though two components (1 and 3) lie near an unstable, to isobaric perturbations, region on the S-curve. By considering the effects of magnetic field on the thermal stability of the WA, we find that these components can be further thermally stabilized by a modest field, even in the presence of substantial continuum fluctuations, even when considering extremes in SED or metallicity. Such stabilization by magnetic fields may explain the ubiquity of WA gas in AGNs.
As more objects are observed with more sensitive instruments (e.g., Chandra and XMM coupled with the HST), better statistical tests of the wind model can be carried out, involving determination of the fraction of AGNs with detectable WA gas, the frequency of objects in which the O vii column density is larger than the O viii column density, and a determination of the range of $`N_H`$. Future higher resolution X-ray spectra (ASCA, XMM) will provide additional kinematic information, such as oxygen edge and resonance line absorption velocities and help to verify the link between X-ray and UV absorption.
###### Acknowledgements.
We gratefully acknowledge illuminating discussions with Roger Blandford, Mike Crenshaw, Arieh Königl, Richard Mushotzky and Chris Reynolds. We thank Gary Ferland for the use of Cloudy. This work was supported in part by NASA grants NAG5-3841, WKU-522762-98-06, WKU-521782-99-04 and HST AR-07982.01-96A.
## Appendix A THERMAL STABILITY OF WARM ABSORBING GAS
The WA gas is found in a wide range of densities, temperatures, ionization and metallicity. It seems plausible, that this ubiquity of WAs in AGNs is related to their thermal stability. Permeating magnetic fields which are expected to play an important dynamical role in AGNs, can affect the thermal stability of the absorbing gas. Therefore, along with Field (1965) and EBS, we discuss the general aspects of thermal stability of magnetized gas and its application to NGC 5548 and to a generic model of the WA, subject to AGN radiation field.
### A.1 The “S-curve” and Thermal Stability: General Considerations
For a gas in thermal equilibrium, the heating rate per unit volume, $`G(n,T)`$, is balanced by the cooling rate per unit volume, $`\mathrm{\Lambda }(n,T)`$. If $`H`$ is the net heating rate defined by
$$HG(n,T)\mathrm{\Lambda }(n,T),$$
(A1)
then equilibrium occurs when $`H=0`$. If gas is slightly perturbed, i.e., $`H0`$, the gas will either heat up or cool down. The change in the gas temperature defines whether it is thermally stable (Field 1965; Krolik, McKee & Tarter 1981). The equation $`H=0`$ is more conveniently treated when mapped to the $`\mathrm{log}T\mathrm{log}U/T`$ plane where the equilibrium curve has roughly an S-shape. Thermal stability of gas depends on the details of the S-curve, and the path that a perturbation is allowed to move along. Figure 6a displays a series of S-curves corresponding to different hydrogen gas densities for the SED of NGC 5548. The curves are relatively insensitive to the gas densities considered and become virtually indistinguishable above $`T10^5`$ K.
To analyze thermal stability of gas, one must be able to follow the path of the perturbed gas relative to the S-curve. Figure 7 shows a small segment of the S-curve for NGC 5548, having a negative slope in the vicinity of point P. To the right of the S-curve, the net heating $`H`$ is positive, i.e., perturbed gas always heats up. To the left of the S-curve, the net heating $`H`$ is negative, meaning that perturbed gas always cools down. Crossing the S-curve at the equilibrium point P are three paths (G, L and I), representing different alternatives for an evolving perturbation off the point P. Path I, a vertical line, represents an isobaric perturbation<sup>2</sup><sup>2</sup>2If luminosity and radius are fixed then $`U/T1/nT`$, therefore, $`U/T=constant`$ implies $`nT`$ is constant. Hence, the gas moving on vertical lines in the $`T`$ vs. $`U/T`$ plane evolves isobarically.. G and L represent perturbations constrained to paths with slopes greater (less negative) and with slope less (more negative) than the slope of the S-curve at P, respectively. Along I and L, a small perturbation off the S-curve leads to a dramatic change in temperature. Perturbations constrained to move along G, on the other hand, are thermally stable since they are driven back to P. Hence, the thermal stability of the gas depends on the slope of developing perturbation in $`\mathrm{log}T\mathrm{log}U/T`$ plane. The gas is thermally stable when this slope is greater (less negative) than the slope of the S-curve at P. For an ionized gas permeated by a magnetic field, as shown below, perturbations move along inclined paths in this diagram, e.g., as G or L in Figure 7. Since stability depends on the relative steepness of the S-curve, as compared to that of the the perturbation path, we first investigate the factors which influence the shape and, therefore, the slope of the S-curve.
### A.2 Factors Affecting the S-curve: SED and Metallicity
The overall shape of the S-curve is affected by the SED. Figure 6b, shows the S-curves for three different SEDs with labels corresponding to SED with photon index $`\mathrm{\Gamma }=1.8`$ and energy cutoffs between 13.6 eV and 40 keV (this SED was utilized in Reynolds 1995), to the SED used in our modeling of NGC 5548, and to the typical quasar SED deduced by Matthews & Ferland (1987). The sequence is in terms of decreasing hardness of the continuum and the greatest effect is on the upper branch of the S-curve. Here the temperature is mainly controlled by Compton heating and cooling. In the limit of large $`\mathrm{log}(U/T)`$, the S-curve is asymptotic to the Compton temperature $`T_C`$ which is given by $`T_C=h<\nu >/4k`$, where $`<\nu >`$ is the flux-averaged photon frequency. Higher $`<\nu >`$ corresponds to harder SED, and, therefore, higher $`T_C`$, explaining the sequence of decreasing upper branch temperatures.
The horizontal separation between S-curves (Fig. 6b) is due to decrease in the continuum hardness, when the percentage of high energy photons ($`h\nu >1`$ Ryd) that heat the gas per unit frequency interval decreases. Decreasing the hardness, allows the gas to cool down, unless compensated for by increasing the number of high energy photons, which is the same as increasing the ionization parameter. Thus, for a given value of $`T`$, the S-curve lies further to the right (i.e., it has higher $`U/T`$ values which correspond to higher values of $`U`$) as the hardness decreases. In spite of a wide range of hardness, the slopes of the S-curves vary little, particularly in regions where the slope is negative. We further note that in all cases considered here, the S-curve slopes is always less than $`1`$, in regions where the slope is negative. This is an important point because, as we show below, the path that gas perturbation treaded by a magnetic field follows, has slope close to $`1`$, resulting in thermal stabilization.
The bumps and wiggles in an S-curve are due to the metallicity of the gas. If AGNs are fueled with galactic ISM, its composition may reflect intense localized star formation which is typically concentrated within nuclear and circumnuclear regions and, therefore, uniformity of composition is not guaranteed. In addition, when grains are directly exposed to an intense central continuum, their mantles will be destroyed and only the graphite cores will be able to survive photo-desorption (Draine & Salpeter 1979) at distances and in the radiation field relevant for WA in NGC 5548. This introduces the possibility of metallicity gradients within the WA environment. Gas abundances are important, because at WA temperatures line radiation by heavier elements, such as oxygen, constitute the dominant coolant. Thus metallicity can influence the thermal structure and stability of the WA gas. This is demonstrated below.
Figure 6c shows S-curves for the SED of NGC 5548 for gas with three different metallicities. At temperatures above $`T10^5K`$, higher metallicity curves are shifted further to the left than lower metallicity curves in the $`\mathrm{log}T`$ vs. $`\mathrm{log}U/T`$ plot. This is because the presence of more metal ions gives the gas a net higher energy absorption cross section than that at a lower metallicity, so that a particular temperature is achieved with a lower ionization parameter. In addition, extra electrons provided by metal enhancement are able to recombine with $`H^+`$ (and $`He^{++}`$), thus the effective ionization of the gas is lower for a given ionization parameter. For example, along a line of $`\mathrm{log}T5.85`$, the ionization parameter increases from roughly 0.85 to 1.14 to 1.25, as the metallicity drops. Below $`T10^5`$ K, the order is reversed because the metal ions recombine and become more efficient coolers via line emission than heaters via absorption. Inspection of the Figure 6c reveals that the slopes of the S-curves for different metallicities are similar, and in the regions where the slopes are negative, they all are less than $`1`$. So, metallicity differences, even as large as a factor of 2 greater or less than solar, will not affect our conclusions about thermal stability of magnetically-dominated gas.
### A.3 Thermal Stability of Magnetized Gas
For isobaric perturbations, there are no dynamical effects and the resulting evolution of the perturbation is controlled by thermal effects only. In the absence of a magnetic field, an isobaric path corresponds to a vertical line (path I in Figure 7). For this kind of perturbation, regions where the slope of the S-curve is negative are thermally unstable. The location of the discrete UV/X-ray WA model is shown in the $`\mathrm{log}T`$ vs. $`\mathrm{log}U/T`$ plane (Fig. 6d). The temperature range of absorption components 1 and 3 lies close to a region which is thermally unstable to isobaric perturbations, so changes in continuum luminosity may drive these components into an unstable region.<sup>3</sup><sup>3</sup>3The temperature uncertainty for each component is estimated by searching for the largest and smallest values temperature values in the $`(n,N(C\mathrm{iv}))`$ solution grid that are adjacent to the optimal solution. Components 2, 4 and 5 are in a region where the S-curve slope is positive, so these components are thermally stable.
The mere existence of kinematic components 1 and 3 in NGC 5548 hints that the gas is thermally stable, despite being close to unstable region. Such a stabilization mechanism can be provided by magnetic fields permeating the gas (Field 1965; EBS). In the presence of magnetic field, isobaric perturbations are constrained to evolve along the curve defined by $`nkT+B^2/8\pi =A`$ where $`A`$ is a constant. To analyze thermal stability, we rewrite this expression in terms of $`T`$ and $`\frac{U}{T}`$ and show that when plotted in the $`\mathrm{log}T`$ vs. $`\mathrm{log}U/T`$ plane, the slope is approximately equal to $`1`$. Assuming magnetic flux freezing, the magnetic field scales with the gas density $`B^2=\varphi ^2n^2`$, where $`\varphi `$ is a constant. Substitution into the expression for total pressure gives $`nkT+\varphi ^2n^2/8\pi =A`$. We eliminate $`n`$ by solving
$$\frac{U}{T}=\frac{L_{ion}}{<E_{ion}>4\pi R^2cnT}$$
(A2)
for $`n`$. This gives
$$n=\frac{E}{(\frac{U}{T})T},$$
(A3)
where $`E=L_{ion}/<E_{ion}>4\pi R^2c`$. Substitution into the expression for total pressure gives
$$\frac{Ek}{\frac{U}{T}}+\frac{\varphi ^2E^2}{8\pi T^2(\frac{U}{T})^2}=A,$$
(A4)
and solving for $`T`$ yields
$$T=\frac{\frac{\varphi E}{\sqrt{8\pi }}}{\sqrt{A(\frac{U}{T})^2Ek(\frac{U}{T})}}.$$
(A5)
To thermally stabilize the gas requires
$$\frac{d\mathrm{log}(T)}{d\mathrm{log}(U/T)}>S,$$
(A6)
where $`S`$, the slope of the S-curve in the $`\mathrm{log}T`$ vs. $`\mathrm{log}U/T`$ plane, and the derivative on the left hand side of Equation A6 are evaluated at the point where Equation A5 crosses the S-curve, yielding
$$\frac{d\mathrm{log}T}{d\mathrm{log}(\frac{U}{T})}=1\frac{1}{2}\beta ,$$
(A7)
where $`\beta `$ is the ratio of gas to magnetic field pressures. Note, that when the magnetic field $`B0`$, then $`\beta \mathrm{}`$, and $`d\mathrm{log}T/d\mathrm{log}(\frac{U}{T})\mathrm{}`$, recovering the strictly isobaric case. At the other extreme lies the limiting case of the cold EBS flow, in which the magnetic pressure completely dominates the gas pressure ($`\beta <<1`$), so $`d\mathrm{log}T/d\mathrm{log}(\frac{U}{T})1`$.
The minimum value of $`B`$ required to stabilize the gas is found by equating the slope of the S-curve to Equation A7, and solving for $`\beta _{min}`$. Figure 8 shows $`1/\beta _{min}`$ values as a function of temperature for the two regions of instability, for the SED of NGC 5548. For stable regions, no magnetic field is needed, so $`1/\beta _{min}0`$. Components 1 and 3 are close to the thermally unstable temperature corresponding to the left peak in Figure 8. The maximum $`1/\beta _{min}`$ occurs at $`\mathrm{log}T=5.55`$, where $`1/\beta _{min}=0.395`$, giving
$$\frac{1}{\beta }=\frac{\frac{B^2}{8\pi }}{nkT}=0.395,$$
(A8)
where $`B`$ and $`n`$ are the values for $`\mathrm{log}T=5.55`$. Thus, according to Equation A8, if the value of the magnetic pressure is about $`40\%`$ of the gas pressure or larger the gas is stabilized. Cloudy calculations and our wind model yield a density of $`\mathrm{log}n=4.99`$ at the center of component 1. With this density a lower bound on the magnetic field required for stabilization can be estimated. Solving Equation A8 for $`B`$ gives
$$B=\sqrt{8\pi \times 0.395nkT}.$$
(A9)
In the regime of the left peak, the gas is stable at all possible temperatures, if $`B7.0\times 10^3`$ Gauss. It should be noted, however, that the bound on $`B`$ is determined only from the stability requirement. A more stringent requirement from the EBS model is that the flow is a “cold” MHD flow, which means $`1/\beta >>1`$, superseding the above criteria. The implication is that the slope given by Equation A7 is nearly equal to $`1`$. We have noted above that variations in SED or metallicity do not significantly change the slopes of the negative parts of the S-curve, and that the slopes are always less than $`1`$, corresponding to case G of Figure 7, in which the slope of the perturbation is larger than the slope of the S-curve. Thus, in the case of NGC 5548, as described by the EBS flow, regions of the S-curve that correspond to thermally unstable regions to isobaric perturbations are thermally stabilized by the presence of a magnetic field.
Table 1a
ASCA R97 X-ray Observational Summary NGC 5548
(Adopted Target Values<sup>a</sup>)
| $`\mathrm{log}(N_H)`$ | $`21.71`$ |
| --- | --- |
| $`\mathrm{log}(N_{O\mathrm{vii}})`$ | $`17.95`$ |
| $`\mathrm{log}(N_{O\mathrm{viii}})`$ | $`18.20`$ |
Table 1b
HST C99 UV Observational Summary NGC 5548
(Adopted Target Values<sup>a</sup>)
| comp. | $`\mathrm{log}[N(N\mathrm{v})]`$ | $`\mathrm{log}[N(C\mathrm{iv})]`$ | $`\mathrm{log}[N(H\mathrm{i})]`$ | $`\mathrm{log}[\frac{f(N\mathrm{v})}{f(C\mathrm{iv})}]`$ |
| --- | --- | --- | --- | --- |
| 1 | $`14.30<`$ | $`13.04<`$ | $`14.13^b`$ | $`2.14>`$ |
| 2 | $`13.78`$ | $`13.45`$ | $`13.88`$ | $`0.91`$ |
| 3 | $`14.59`$ | $`13.68`$ | $`14.25`$ | $`1.49`$ |
| 4 | $`14.81`$ | $`14.46`$ | $`14.52`$ | $`0.93`$ |
| 5 | $`14.04`$ | $`13.61`$ | $`13.11`$ | $`1.01`$ |
| $`\mathrm{\Sigma }`$ | $`15.17`$ | $`14.62`$ | $`14.87`$ | |
<sup>a</sup> All values have assigned error bars of $`\pm \mathrm{log}(2.0)`$ except for those with a $`<`$, which indicates a lower bound or $`>`$, which indicates an upper bound.
<sup>b</sup>STIS observation (Crenshaw & Kraemer 1999)
Table 2
Model Solution for the Continuously Distributed
Warm Absorber in NGC 5548
| parameter | log(value)<sup>a</sup> |
| --- | --- |
| $`\mathrm{R}_{\mathrm{min}}`$ | 17.75 |
| $`\mathrm{R}_{\mathrm{max}}`$ | 20.59 |
| $`\mathrm{n}_{\mathrm{min}}`$ | 6.455 |
| $`\mathrm{n}_{\mathrm{max}}`$ | 2.253 |
| N(H) | 21.76 |
| N(H i) | 15.94 |
| N(C iv) | 14.70 |
| N(N v) | 15.13 |
| N(O vii) | 17.94 |
| N(O viii) | 18.22 |
| $`ϵ`$ | -2.716 |
<sup>a</sup> values in cgs units.
Table 3a
Optimal EBS Solution Column Densities<sup>a</sup>
| comp. | C iv | N v | O vi | O vii | O viii | H | H i |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | $`13.50`$ | $`14.30`$ | $`16.15`$ | $`17.73`$ | $`17.84`$ | $`21.29`$ | $`15.57`$ |
| 2 | $`13.45`$ | $`13.48`$ | $`14.57`$ | $`14.51`$ | $`13.00`$ | $`18.14`$ | $`14.11`$ |
| 3 | $`13.62`$ | $`14.41`$ | $`16.25`$ | $`17.80`$ | $`17.88`$ | $`21.33`$ | $`15.66`$ |
| 4 | $`14.46`$ | $`14.51`$ | $`15.65`$ | $`15.65`$ | $`14.20`$ | $`19.22`$ | $`15.14`$ |
| 5 | $`13.49`$ | $`13.74`$ | $`15.11`$ | $`15.56`$ | $`14.55`$ | $`18.88`$ | $`14.35`$ |
| $`\mathrm{\Sigma }`$ | $`14.62`$ | $`14.94`$ | $`16.58`$ | $`18.07`$ | $`18.16`$ | $`21.61`$ | $`16.00`$ |
<sup>a</sup>Logarithims of column densities in units of $`\mathrm{cm}^2`$
Table 3b
Optimal EBS Solution Physical Parameters<sup>a</sup>
| comp. | $`\mathrm{v}_{\mathrm{obs}}^\mathrm{b}`$ | FWHM<sup>b</sup> | $`\mathrm{v}_\mathrm{s}`$ | $`\mathrm{\Delta }\mathrm{v}`$ | $`\frac{f(N\mathrm{v})}{f(C\mathrm{iv})}`$ | R | $`\mathrm{\Delta }\mathrm{R}`$ | n | T | U | $`ϵ`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | $`1060`$ | $`114`$ | $`9.8`$ | $`104.2`$ | $`1.39`$ | $`18.8`$ | $`18.1`$ | $`5.0`$ | $`4.87`$ | $`0.03`$ | $`1.8`$ |
| 2 | $`655`$ | $`45`$ | $`5.4`$ | $`39.5`$ | $`0.61`$ | $`19.2`$ | $`18.3`$ | $`5.4`$ | $`4.37`$ | $`1.29`$ | $`5.6`$ |
| 3 | $`518`$ | $`90`$ | $`9.6`$ | $`80.4`$ | $`1.37`$ | $`19.4`$ | $`18.9`$ | $`3.8`$ | $`4.86`$ | $`0.05`$ | $`1.4`$ |
| 4 | $`344`$ | $`156`$ | $`5.4`$ | $`150.6`$ | $`0.63`$ | $`19.8`$ | $`19.8`$ | $`4.2`$ | $`4.36`$ | $`1.23`$ | $`4.8`$ |
| 5 | $`165`$ | $`74`$ | $`5.9`$ | $`68.7`$ | $`0.83`$ | $`20.4`$ | $`20.4`$ | $`2.6`$ | $`4.44`$ | $`0.85`$ | $`4.1`$ |
<sup>a</sup>Velocities are given in $`\mathrm{km}\mathrm{s}^1`$. All other entries are given as logarithm of the value expressed in cgs.
<sup>b</sup>Values are from C99.
Table 4
Component Ion Recombination Times
and Percentage Total Column Contributions
for NGC 5548
| n | C iv<sup>a</sup> | %<sup>b</sup> | O vii | % | O viii | % |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | $`2.71`$ | $`7.59`$ | $`5.39`$ | $`45.71`$ | $`6.62`$ | $`47.86`$ |
| 2 | $`4.23`$ | $`6.76`$ | $`5.46`$ | $`0.03`$ | $`5.54`$ | $`>0.01`$ |
| 3 | $`3.96`$ | $`10.00`$ | $`6.60`$ | $`53.70`$ | $`7.80`$ | $`52.48`$ |
| 4 | $`5.30`$ | $`69.18`$ | $`6.59`$ | $`0.38`$ | $`6.57`$ | $`0.01`$ |
| 5 | $`6.39`$ | $`7.41`$ | $`8.14`$ | $`0.31`$ | $`8.30`$ | $`0.02`$ |
| $`\mathrm{\Sigma }^c`$ | $`5.51`$ | | $`6.43`$ | | $`7.54`$ | |
<sup>a</sup> Values are the $`\mathrm{log}t(s)`$ of the absolute value of the mean component recombination time. All C iv responses are anticorrelated and all O vii and O viii responses are correlated with a continuum drop.
<sup>b</sup> Percentage of the total ionic column density.
<sup>c</sup> Values are the $`\mathrm{log}t(s)`$ of the absolute value of the mean recombination time averaged over all components and weighted by their relative column density contributions. |
warning/0001/cond-mat0001317.html | ar5iv | text | # Magnetic and Thermodynamic Properties of the Collective Paramagnet-Spin Liquid Pyrochlore Tb2Ti2O7
## I Introduction
In magnetic systems, competition between magnetic interactions, combined with certain local lattice symmetries involving triangles, give rise to the notion of geometric frustration . Geometrically frustrated antiferromagnets are currently attracting much interest within the condensed matter community . The main reason for this interest is that geometric frustration can cause sufficiently large zero-temperature quantum spin fluctuations as to drive a system into novel types of intrinsically quantum mechanical magnetic ground states with no classical equivalent .
Among three-dimensional systems, the pyrochlore lattice of corner-sharing tetrahedra (see Fig. 1) with antiferromagnetic nearest-neighbor exchange interaction is particularly interesting. For this system, theory and Monte Carlo simulations show that for classical Heisenberg magnetic moments interacting with a nearest-neighbor antiferromagnetic coupling, there is no transition to long-range magnetic order at finite temperature. This is unlike the two-dimensional kagomé lattice antiferromagnet where a thermally-driven order-by-disorder of spin nematic order occurs. Villain coined the name “collective paramagnetic” to describe the classical state of the pyrochlore lattice at low temperatures .
Because of their low propensity to order even for classical spins, antiferromagnetic materials based on a pyrochlore lattice appear to be excellent systems in which to seek exotic quantum mechanical ground states. For example, numerical calculations suggest that the $`S=1/2`$ pyrochlore Heisenberg antiferromagnet may be fully quantum disordered, giving rise to a state that is commonly referred to as “spin liquid” . Both terms “collective-paramagnet” and “spin liquid” are meant to emphasize that despite such a system remaining in a paramagnetic phase down to absolute zero temperature, the properties of such a state involve very strong and nontrivial short-range spin correlations, analogous to the nontrivial position-position correlations present in an ordinary atomic or molecular fluid.
A number of experimental studies on insulating pyrochlore materials have been reported in the past ten years. Interestingly, it has been found that such systems do not typically form such a spin liquid state that remains paramagnetic down to zero temperature. Most often, these systems either display long-range antiferromagnetic order; such as FeF<sub>3</sub> , Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> , ZnFe<sub>2</sub>O<sub>4</sub> , and ZnCr<sub>2</sub>O<sub>4</sub> , or enter a spin-glass-like state below some nonzero spin-freezing temperature as exhibited by 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> , as well as the disordered CsNiCrF<sub>6</sub> pyrochlore .
Recently, several studies of the pyrochlore rare-earth titanates, R<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, have been published . In these compounds, the trivalent rare earth ions, R<sup>3+</sup>, occupy the 16d sites of the Fd$`\overline{3}`$m space and form a pyrochlore lattice (Fig. 1). The behaviors displayed in this family of pyrochlores are much varied indeed. Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> develops true long-range order at a critical temperature of about 1K . Tm<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> possesses a trivial non-magnetic (i.e. spin singlet) ground state separated by an energy gap of about 120 K to the next crystal field level . Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is well described by an Ising doublet . In that system, it was originally argued that the nearest-neighbor exchange interaction is weakly ferromagnetic , and that the strong Ising-like single ion anisotropy along $`(111)`$ directions frustrates the development of long range ferromagnetic order . This material also exhibits low temperature spin dynamics reminiscent of Pauling’s “ice model” , an equivalence proposed by Harris and co-workers . Recently, it has been found that Dy<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is also a very good example of “spin ice” , and that the application of a magnetic field can restore much of the ground-state entropy and drive magnetic phase transitions. Most recently, den Hertog and Gingras have argued that the spin ice physics in both Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and Dy<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is not driven by nearest-neighbor ferromagnetic exchange, but is rather due to the long-range $`1/r^3`$ nature of magnetic dipole-dipole interactions .
In contrast to the long-range ordered or spin glass states mentioned above, strong evidence for collective paramagnetism, or spin liquid behavior, was recently observed in the insulating pyrochlore Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . It was found using neutron scattering and muon spin relaxation methods that this material remains paramagnetic down to (at least) 70 mK despite the fact that the paramagnetic Curie-Weiss temperature, $`\theta _{\mathrm{CW}}`$, is -19 K, and that short-range antiferromagnetic correlations begin to develop at $``$ 50 K. At first sight, one could argue that it is “pleasing” to have found at last the spin liquid state anticipated by theory for a highly frustrated pyrochlore antiferromagnet . However, the situation for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is not as simple as it might naively appear.
In Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> the Tb<sup>3+</sup> ions have a partially filled $`{}_{}{}^{7}F_{6}^{}`$ shell, and one most first understand their crystal field level scheme and, in particular, the nature of the single-ion magnetic ground state before constructing a correct effective spin-spin Hamiltonian for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. Indeed, we show below that crystal field anisotropy renders the description of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> in terms of an isotropic Heisenberg antiferromagnetic model completely inappropriate.
If one neglects the axial oxygen distortion around the Tb<sup>3+</sup> sites, and assumes that the local environment of the Tb<sup>3+</sup> is perfectly cubic, one would expect, based on point charge calculations, that the ground state of both Tb<sup>3+</sup> and Tm<sup>3+</sup> should either be a singlet or a nonmagnetic doublet . For example, as mentioned above, experimental evidence for a nonmagnetic singlet ground state has been found in Tm<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . Based on this naive picture, one can see that the experimental evidence of a moment for Tb<sup>3+</sup> in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is therefore a nontrivial issue that needs to be understood.
A simple possibility is that corrections beyond the point-charge approximation and/or the known axial oxygen distortions around each of the 16d sites cause the Tb<sup>3+</sup> cations to acquire a permanent magnetic moment. Another and more interesting possibility, is that the moment on the Tb<sup>3+</sup> site in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is induced by a collective bootstrapping of the magnetic (exchange and/or dipolar) interactions as occurs in the tetragonal LiTbF<sub>4</sub> material . In Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, a priori, it is theoretically possible that there could be no moment on the Tb<sup>3+</sup> site for a concentration $`x`$ of Tb<sup>3+</sup> less than some critical concentration $`x_c`$ in (Tb<sub>x</sub>Y<sub>1-x</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, as occurs in LiTb<sub>x</sub>Y<sub>1-x</sub>F<sub>4</sub> . This is an important issue. Indeed, one could imagine that for the highly-frustrated pyrochlore lattice, the collective development of a permanent ground-state moment would not give rise to homogeneous moments on the Tb<sup>3+</sup> sites, but to a kind of “modulated moment structure”. This idea is conceptually similar to what is found in the frustrated tetragonal TbRu<sub>2</sub>Ge<sub>2</sub> material , but where for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> there might be instead a quantum-disordered state “intervening” between a trivial singlet ground state and a long-range ordered one, with the quantum-disordered state extending all the way to $`x=1`$.
In the case where a permanent moment does exist on Tb<sup>3+</sup> even in absence of interaction (i.e. the limit $`x0`$ in (Tb<sub>x</sub>Y<sub>1-x</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>), the important issue is to determine the wavefunction decomposition of the ground state in terms of $`|J,M_J`$ states and the symmetry, Heisenberg or otherwise, of the resulting effective spin variable. The goal of such a programme is to construct a low-energy effective spin Hamiltonian in order to tackle theoretically why Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> does not order at nonzero temperature. Consequently, it is very important to know in more details what the magnetic nature of the Tb<sup>3+</sup> single-ion ground state in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is.
The main purpose of this paper is to examine the magnetic nature of the Tb<sup>3+</sup> ion in the Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> pyrochlore in order to assess whether or not there is indeed a permanent moment at the Tb site as the temperature goes to zero, and determine the nature of this moment (e.g. level of effective spin anisotropy). We present in Section II experimental evidence, based on results from d.c. susceptibility, heat capacity and powder inelastic neutron studies that show there is a permanent moment at the Tb site, but that its approximate 5$`\mu _\mathrm{B}`$ value is less than the value of 9.4$`\mu _\mathrm{B}`$ estimated from the d.c. susceptibility measurements above 200K , or the 9.72$`\mu _\mathrm{B}`$ $`{}_{}{}^{7}F_{6}^{}`$ free ion value. To complement the experimental work, results from ab-initio crystal field calculations that take into account covalent and electrostatic effects are presented in Section III and Appendix B. We discuss in Section IV the possibility that dipole-dipole interactions and extra perturbative exchange couplings beyond nearest-neighbor may be responsible for the lack of ordering in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
## II Experimental Method & Results
### A Sample Preparation
Samples of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> were prepared in the form of polycrystalline pellets by high temperature solid state reaction. Starting materials, Tb<sub>2</sub>O<sub>3</sub>, Y<sub>2</sub>O<sub>3</sub> and TiO<sub>2</sub> were taken in stoichiometric proportions, mixed thoroughly, pressed into pellets and heated in an alumina crucible at 1400<sup>o</sup>C for 12 hours in air. Tb<sub>2</sub>O<sub>3</sub> was prepared by hydrogen reduction of Tb<sub>4</sub>O<sub>7</sub>. The powder x-ray diffraction patterns of the samples, obtained with a Guinier-Hagg camera, indicate that they are single phase with cubic unit cell constants, a<sub>0</sub>, of 10.491$`\AA `$ for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and 10.104$`\AA `$ for (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. The value for the concentrated sample is in excellent agreement with previous reports .
Some of this high quality polycrystalline material was then used as starting material for a successful single crystal growth using an optical floating zone image furnace. Details of the crystal growth are described elsewhere .
### B d.c. Magnetic Susceptibility Measurements
As a first step towards determining the magnetic nature of the electronic ground state of the Tb<sup>3+</sup> cations in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, we have investigated the d.c. magnetic susceptibility of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. The d.c. magnetic susceptibility was measured using a SQUID magnetometer(Quantum Design, San Diego) in the temperature range 2-300 K. The inverse susceptibility, $`\chi ^1`$, of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> measured at an applied field of 0.01 Tesla is shown in Fig. 2. A fit of the data to the Curie-Weiss (CW) law above 200 K gives an effective paramagnetic moment of 9.4$`\mu _\mathrm{B}`$/Tb<sup>3+</sup> and an effective (high-temperature) CW temperature, $`\theta _{\mathrm{CW}}`$ = -19 K, which indicates the dominance of antiferromagnetic interactions (the Curie-Weiss fits are done above 200 K). Deviation from the Curie-Weiss law sets in at a rather high temperature, $`T`$70 K. This is consistent with elastic neutron scattering results reported previously, which showed evidence for short range magnetic correlations at temperatures up to 50 K .
Recognizing that for Tb<sup>3+</sup>, an <sup>7</sup>F<sub>6</sub> even electron and non-S-state ion, there may be “crystal field” as well as exchange contributions to the experimentally determined $`\theta _{\mathrm{CW}}`$, a magnetically dilute sample, (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, was also studied. The data for this sample are shown in Fig. 3 along with a Curie-Weiss fit giving again a value close to the free ion value for the effective paramagnetic moment, 9.6$`\mu _\mathrm{B}`$/Tb<sup>3+</sup>, and $`\theta _{\mathrm{CW}}6`$ K. This finite value is in contrast to the essentially zero $`\theta _{\mathrm{CW}}`$ value obtained for a similarly diluted sample of Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> which contains the isotropic “spin only” <sup>8</sup>S<sub>7/2</sub> Gd<sup>3+</sup> ion , and therefore indicates that a significant crystal field contribution to $`\theta _{\mathrm{CW}}`$ exists in the Tb$`{}_{}{}^{3+}`$based material. Thus, to a first approximation one can estimate that the portion of $`\theta _{\mathrm{CW}}`$ for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> which can be attributed to magnetic interactions is
$`\theta _{\mathrm{CW}}`$$`\{`$Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> $`\}`$ \- $`\theta _{\mathrm{CW}}`$$`\{`$(Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> $`\}`$ $``$ -13 K,
a value similar to the $`\theta _{\mathrm{CW}}`$ -10 K found for Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . This approach can be made more rigorous by noting that in a high-temperature series expansion, one finds that the magnetic susceptibility is $`\chi =C_1(1/T+C_2/T^2)`$ where $`C_2\theta _{\mathrm{CW}}`$ can be “decomposed” as a simple sum of terms that are ascribed to exchange interactions, dipolar interactions and crystal-field terms (see Appendix A). Note also that, down to a temperature of $`T=2`$ K, neither Fig. 2 nor Fig. 3 show any sign of a singlet ground state which would manifest itself as an approach to a constant susceptibility with decreasing temperature, as found in the single ground-state of the Tm<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . A more detailed analysis concerning this issue is presented in Section III.
While it is tempting to use the magnetic interaction contribution ($`\theta _{\mathrm{CW}}\theta _{\mathrm{CW}}^{\mathrm{cf}})13`$ K, (with $`\theta _{\mathrm{CW}}19`$ K and $`\theta _{\mathrm{CW}}^{\mathrm{cf}}6`$ K), to extract the approximate value of the nearest neighbor exchange, an estimate of the nearest neighbor dipole-dipole interaction for $`\mathrm{Tb}^{3+}`$ ions indicates an energy scale of about 1 K. Given the long range nature of dipolar forces, it becomes clear that the classical nearest neighbor exchange constant cannot be obtained from a measurement of the Curie-Weiss temperature until the effects of long range dipolar interactions on $`\theta _{\mathrm{CW}}`$ have been understood. We find via a high temperature series expansion analysis of the long range dipolar contributions, $`\theta _{\mathrm{CW}}^{\mathrm{dip}}`$, to the Curie-Weiss temperature (see Appendix A), that the estimated upper bound on $`\theta _{\mathrm{CW}}^{\mathrm{dip}}`$ is ferromagnetic and $`1.2`$ K (for needle-shaped powder crystallites), while the lower bound is antiferromagnetic and $`2.4`$ K (for slab-shaped powder crystallites). Consequently, we find that antiferromagnetic exchange interactions are predominantly responsible for the ($`\theta _{\mathrm{CW}}\theta _{\mathrm{CW}}^{\mathrm{cf}})=13`$ K value determined above $`T200`$ K, with a resulting $`\theta _{\mathrm{CW}}^{\mathrm{exchange}}[14.2,10.6]`$ K.
### C Neutron Scattering Experiments
Inelastic neutron scattering allows us to determine with reasonable precision the values of the electronic energy levels of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. The inelastic neutron scattering measurements were carried out on a 50g sample of polycrystalline Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, loaded in a sealed Al cell with a helium exchange gas present. The cell was mounted in a closed cycle helium refrigerator with a base temperature of 12 K. The sample was the same as used in reference . Measurements were performed on the C5 triple axis spectrometer at the Chalk River Laboratories in constant scattered neutron energy mode. Two spectrometer configurations were employed, appropriate for relatively high and low energy resolution, respectively. Both configurations employed pyrolitic graphite (PG) as both monochromator and analyser. The low resolution measurements, appropriate for relatively high energy transfers, were performed using $`\frac{E^{}}{h}`$=3.52 THz (1THz=48 K), open-60-80-open collimation, and a PG filter in the scattered beam. The high resolution configuration used $`\frac{E^{}}{h}`$=1.2 THz, open-40-60-open collimation, and a cooled Be filter in the scattered beam. These results clearly indicate excitations at $`E`$ 0.35THz, 2.5THz and a broad neutron group centered at 3.5THz (corrsponding to 16.8 K, 120 K, and 168 K, respectively).
The low energy-resolution inelastic neutron scattering measurements at 12 K with $`\frac{E^{}}{h}`$=3.52 THz revealed the presence of two $`Q`$-independent modes with frequencies $`\nu `$ 2.5 and 3.5 THz. Representative neutron groups, as well as the dispersion relation for these two excitations are shown in the top panel of Fig. 4. These excitations are identified as being magnetic in origin due to their temperature and Q dependence. Their flat dispersion indicates that they are crystal electric field levels for Tb<sup>3+</sup> in the environment appropriate for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
The high energy-resolution inelastic neutron measurements at 12 K with $`\frac{E^{}}{h}`$=1.2 THz shows the presence of a low lying magnetic excitation near $`\nu `$ 0.35 THz. This mode is also dispersionless above a temperature of $``$ 25 K, but partially softens in energy at the wavevector which characterizes the very short range spin correlations, which develop below 25 K. The development of this interesting dispersion has been described previously . The dispersion of the low lying excitation is shown in the lower panel of Fig. 4. One can see that this partial softening of the excitation branch occurs only for the lowest lying mode. As we show in the next section, this is manifested in the heat capacity measurements as broad features that result of a broadening of the single-ion energy levels via these magnetic correlation effects.
These measurements place constraints on any calculations for the energy eigenstates of Tb<sup>3+</sup> as they set both the energy spacing of the levels, and require that magnetic dipole matrix elements must connect the ground state with these levels in order that they be visible to the inelastic neutron scattering experiment. In particular, this indicates nonzero $`0|J^+|1`$, $`0|J^{}|1`$, or $`0|J^z|1`$ matrix elements connecting the ground state, $`|0`$, and the first excited state, $`|1`$, at an energy $`0.35`$ THz $``$ 17 K above the ground state. In other words, there must be large $`|J,M_J`$ components in $`|0`$ and $`|1`$ where some of the $`M_J`$ involved for the ground state and the excited state differ by 0, $`\pm 1`$.
### D Specific Heat Measurements
Low-temperature specific-heat measurements on Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> were performed using a thermal-relaxation microcalorimeter. The single crystal sample was mounted on a sapphire holder which was isolated from the bath by four copper-gold alloy wires. The relative precision and absolute accuracy of the calorimeter were confirmed by measuring copper and gold standards. In principle, specific heat-measurements on a dilute (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> sample would be useful. Unfortunately, this is not technically easily feasible as the magnetic contribution to the total specific heat would be too small to be determined accurately.
The total specific heat, $`C_p`$, of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> was measured from 0.4 K to 30 K at applied fields of 0, 2 and 5 T (see Fig. 5). The zero field data exhibits two broad peaks centered at about 1.5 K and 6 K. The data are in agreement with those of Ref. , above a temperature of 0.4 K. Hyperfine contributions to the specific heat become important below 0.4 K for Tb$``$based compounds, as found for example in the Tb<sub>2</sub>(GaSn)O<sub>7</sub> pyrochlore , and this is presumably the reason for the sharp increase of $`C_p(T)/T`$ found in Fig. 4 of Ref. below 0.5 K. The solid line in Fig. 5 corresponds to the estimated lattice heat capacity, $`C_l`$, for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, determined by scaling the heat capacity for Y<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, which is insulating, non-magnetic, and is isotructural to Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> .
The magnetic specific heat, $`C_m`$, obtained by subtracting $`C_l`$ from $`C_p`$, is shown in Fig. 6 for the three applied fields. With the application of a 2 T field the magnitude of the lower temperature peak diminishes and moves to a slightly higher temperature. In contrast, the position of the second peak does not change but increases in magnitude as the lower temperature feature begins to overlap with it. At 5 T the low temperature feature disappears completely and the remaining peak is shifted to higher temperatures.
The crystal field calculations described in Section III and Appendix B (see Table 2) indicate a level scheme consisting of a ground state doublet with another doublet as the first excited state. Attempts to fit the $`C_m`$ data to Schottky anomalies using a unique doublet$``$doublet level scheme for the Tb<sup>3+</sup> ions and varying the ground state$``$excited state energy splitting failed, since important magnetic short-range correlations are present in this sytem, as discussed above in neutron scattering . Consequently, we interpret the anomaly at $``$ 6 K as a remnant of the excited doublet that is “broadened” by exchange correlation fields, while the 1.5K anomaly is presumably due to these same correlation effects, but now acting on the single-ion ground state doublet. In other words, the build-up of short-range correlations in the low-temperature sector of the effective Hamiltonian for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> results in a specific heat anomaly at 1.5 K. This low-temperature anomaly at 1.5K that results from correlations is akin to the broad specific heat bump at $``$ 2 K found in Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . However, Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is an $`{}_{}{}^{8}S_{7/2}^{}`$ spin-only ion, and there are no crystal field levels at high energy, nor are there correlation remnants of crystal field levels above the ground state such as those that cause the specific heat anomaly at 6 K in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. Another distinction between Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and Gd<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is that the latter shows a very sharp specific heat anomaly at 0.9 K , and therefore presumably a transition to long range order at that temperature as suggested by recent neutron scattering experiments . Down to 0.4 K, no such sharp specific heat anomaly is found in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. From the fit of the d.c. susceptibility and the crystal field calculations presented in Section III and Appendix B, good evidence is obtained that the magnetic moment in the ground state and the first excited state is $``$ 5 $`\mu _\text{B}`$ and $`6`$ $`\mu _\text{B}`$, respectively. Given a doublet-doublet energy gap of about 17 K, we can estimate the strength of the magnetic field where the separation between the doublets is equal to the magnetic field energy, and find a magnetic field of about 5 T. For an applied field of that strength, the ground state and excited states merge and are strongly mixed. This explains the disappearance of the low-temperature specific heat anomaly in Fig. 6 for a fuekd $`H=`$ 5 T.
It is usual to estimate the magnetic entropy, $`S_m(T)`$, in a system by integrating $`C_m(T)/T`$ between the lowest temperature reached and the temperature, $`T`$, of interest. Although it is straightforward to integrate $`C_m(T)/T`$, the interpretation of the results for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is difficult. The main reasons for this are:
* The hyperfine interaction is large for Tb<sup>3+</sup> and the nuclear specific heat contribution becomes significant with respect to the magnetic contribution near 0.4K .
* There is a doublet crystal field excitation at an energy of approximately 17 K. Hence, one can hardly integrate $`C_m(T)/T`$ above 5 K without “already” embedding in the resulting entropy a contribution from excitations to the first excited doublet.
* From our neutron results and crystal field calculations presented in Appendix B, it is known that there are other levels at an energy $`10^2`$ K. Hence one cannot integrate $`C_m(T)/T`$ to obtain $`S_m(T)`$ up to a high enough temperature without having to consider the specific heat contribution from the states at $`O(10^2)`$ K.
* By integrating $`C_m(T)/T`$ up to $`30`$ K one enters a regime where the lattice contribution to the specific heat, $`C_l(T)`$, becomes sizeable (see Fig. 5). In that case the subtraction of $`C_l(T)`$ from the total $`C_p(T)`$ using rescaled results for the isostructural non-magnetic Y<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> leads to inherent uncertainties which increase dramatically above 10 K (see Fig. 5).
With these provisions in mind, we have determined $`S_m(T)`$ (see Fig. 7). We find that the recovered entropy, $`S_m(T)`$, at 15 K is already larger than that expected for a singlet$``$doublet energy level scheme, $`S_m(15\mathrm{K})>\mathrm{R}\mathrm{ln}(3)`$. Since 15 K is much less than the excited states at $`100`$ K, there should be little contribution to $`S_m(15\mathrm{K})`$ coming from states at $`T`$ 100 K. Consequently, the results presented in this figure support further the picture that the two lowest energy levels in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> consists of two doublets. However, it is interesting to note that at a temperature of 30 K, the recovered entropy is not yet equal to R$`\mathrm{ln}(4)`$, the total entropy for two doublets. Either 30 K is not yet at high enough temperature to have recovered the full doublet$``$doublet entropy, or there exists macroscopic entropy in the ground state as occurs in Dy<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . For the four reasons mentioned above, it is difficult to make the discussion about the recovered entropy above 30 K in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> much more quantitative.
In summary the magnetic specific heat data are consistent with the inelastic neutron scattering results in that the two lowest-lying energy levels for Tb<sup>3+</sup> in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> consist of two doublet energy levels separated by an excitation energy of $``$ 15 $``$ 20 K. As we will show in the next Section and Appendix B, our crystal field calculations strongly suggest that these two lowest lying levels are magnetic (Ising) doublets.
## III Single ion Properties
As discussed in the introduction, it is very important to determine whether the existence of a moment at the Tb<sup>3+</sup> site in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is intrinsic, or driven by magnetic interactions. Consequently, we have investigated in detail the problem of single-ion properties of Tb<sup>3+</sup>, both theoretically and via d.c. susceptibility measurements, of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and of the dilute (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> material where the Tb$``$Tb interactions should be negligible. The main conclusion from this investigation is that the Tb<sup>3+</sup> cation does indeed carry an intrinsic moment inherent to its environment in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, and that moment is not due to a bootstrapping effect from interactions.
The results from inelastic neutron scattering and specific heat measurements presented in the previous section already provide some evidence for a doublet ground state and an excited doublet state at an energy of $``$ 17K. The purpose of this section is to investigate further from a theoretical point of view the question of the existence of a magnetic doublet ground state for Tb<sup>3+</sup> in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. In short, both a simple point charge calculation and a more sophisticated ab-initio method confirm that a doublet$``$doublet scheme is the most likely low energy level structure for Tb<sup>3+</sup>. We have constructed a van Vleck equation based on such a doublet$``$doublet scheme in order to parametrize the d.c. suceptibility of the dilute (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> sample, and to extract the value of the magnetic moment for both the ground state and excited state doublets. We find that experimental results for the d.c. susceptibility of (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> are in reasonably good agreement with the ab-initio calculations.
### A Crystal Field Effects
Experimental evidence of a doublet-doublet structure for the low temperature crystal field levels of $`\mathrm{Tb}^{3+}`$ in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> can be understood by considering the crystal field environment surrounding the ion. Pyrochlore oxides $`\mathrm{A}_2\mathrm{B}_2\mathrm{O}_7`$ are described in space group Fd$`\overline{3}`$m with A<sup>3+</sup>, the trivalent rare earth in 16d, B<sup>4+</sup>, the tetravalent transition metal ion, in 16c, O1 in 48e and O2 in 8b. The A or Tb<sup>3+</sup> site in this case is coordinated to six O1 ions at about 2.5$`\AA `$ in the form of a puckered ring and to two O2 ions at a distance of 2.2$`\AA `$ in the form of a linear O2-Tb-O2 chain oriented normal to the mean plane of the O1 ring. The O2-Tb-O2 units are parallel to the $`(111)`$ directions within the cubic unit cell. Overall, the local geometry at the Tb<sup>3+</sup> site can be described as a severe trigonal compression along the body diagonal of a simple cube.
Based on symmetry considerations, the cubic plus axial distortion surrounding the $`\mathrm{Tb}^{3+}`$ ion may be expressed by a crystal field Hamiltonian of the general form
$`H^{\mathrm{cf}}`$ $`=`$ $`B_0^2C_0^2+B_0^4C_0^4+B_3^4C_3^4`$ (1)
$`+`$ $`B_0^6C_0^6+B_3^6C_3^6+B_6^6C_6^6,`$ (2)
where the $`B_q^k`$’s are yet to be determined crystal field parameters, and the $`C_q^k`$’s are tensorial operators defined as $`C_q^k=[4\pi /(2k+1)]^{1/2}Y_k^q`$, where $`Y_q^k`$ is a normalized spherical harmonic. In general, the $`B_q^k`$’s represent an effective one-body potential which lifts the degeneracy of the angular momentum states in question. In practice, they may be determined experimentally by spectroscopic or thermodynamic probes, or theoretically within various levels of approximation.
In the simplest approximation of the crystal field interactions, one often uses the so-called point charge (PC) approximation where the crystal field is simply assumed to be caused by the Coulomb field of point charges situated at neighboring sites. In such a picture, the crystal field eigenstates and eigenvalues have been determined for a number of rare earths by Lea et al. for systems with cubic symmetry (i.e. without trigonal distortion) . For $`\mathrm{Tb}^{3+}`$, the lowest three energy levels in a cubic environment are a $`\mathrm{\Gamma }_3`$ singlet, a non-magnetic $`\mathrm{\Gamma }_2`$ doublet and a $`\mathrm{\Gamma }_5^{(2)}`$ triplet, with their precise ordering in terms of lowest energy states dependent upon variation of the point charge crystal field parameters . In general, the addition of the trigonal distortion splits the $`\mathrm{\Gamma }_5^{(2)}`$ triplet into a singlet and doublet, while the $`\mathrm{\Gamma }_3`$ and $`\mathrm{\Gamma }_2`$ states are preserved. Hence, we expect that the crystal field ground state of the $`\mathrm{Tb}^{3+}`$ ion in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> will be a competition between two doublets and two singlets.
Some notion regarding the difficulty of determining the precise ordering of these states and the size of their associated magnetic moments may be obtained by diagonalizing the crystal field Hamiltonian $`H^{\mathrm{cf}}`$ using Stevens’ operator equivalents of the $`C_q^k`$ within a fixed $`J`$ manifold, and by using the point charge approximation for the Coulomb effects of the surrounding oxygen ions. The resulting crystal field point charge Hamiltonian, H$`{}_{}{}^{\mathrm{cf}}{}_{\mathrm{pc}}{}^{}`$, with quantization axis along the appropriate $`111`$ direction can be expressed as,
$`H_{\mathrm{pc}}^{\mathrm{cf}}`$ $`=`$ $`\alpha _J\stackrel{~}{B}_2^0(r^31)O_2^0`$ (3)
$`+`$ $`\beta _J\stackrel{~}{B}_4^0\left[{\displaystyle \frac{(27r^5+1)}{28}}O_4^020\sqrt{2}O_4^3\right]`$ (4)
$`+`$ $`\gamma _J\stackrel{~}{B}_6^0\left[{\displaystyle \frac{(188+324r^7)}{512}}O_6^0+{\displaystyle \frac{35\sqrt{2}}{4}}O_6^3+{\displaystyle \frac{77}{8}}O_6^6\right],`$ (5)
where $`r=R_1/R_2`$ and $`R_1,R_2`$ are the Tb-O distances for oxygen ions situated on the puckered ring and on the distortion axis respectively. The $`O_m^n`$ represent crystal field operators as discussed by Hutchings , while $`\alpha _J`$, $`\beta _J`$, and $`\gamma _J`$ are the Steven’s coefficients . Trivalent Tb<sup>3+</sup> is an $`{}_{}{}^{7}F_{6}^{}`$ ion and thus the fixed $`J`$ manifold is $`J=6`$ ($`L=3,S=3,J=L+S=6`$) for the operator equivalent point charge calculation. The precise relationship between the point charge parameter set $`\{\stackrel{~}{B}_q^k\}`$ and the more general $`\{B_q^k\}`$ is discussed by Kassman .
Although a point charge estimation of the crystal field parameters is in most cases unreliable in predicting the actual crystal field level spacing of rare earth ions, we find that varying the point charge parameter set $`\{\stackrel{~}{B}_q^k\}`$ indicates that the ground state can be confirmed to be a competition between two singlets close in energy and two magnetic doublets, which are well separated from the other crystal field states. These levels are indeed the remnants of the $`\mathrm{\Gamma }_2,\mathrm{\Gamma }_3`$ and $`\mathrm{\Gamma }_5^{(2)}`$ states of the cubic environment eluded to earlier.
In general, we find that in a large region of the crystal field parameter space, the two doublets form the lowest energy levels, although their precise ordering may change. Although the weight of each angular momentum component of a crystal field eigenstate varies with the values of the crystal field parameters, some other general features do emerge. In particular, one singlet contains only $`|\pm 6`$ and $`|\pm 3`$ states while the other is a combination of $`|\pm 6,|\pm 3`$ and $`|0`$. On the other hand both sets of doublets are magnetic (e.g they have a nonzero quantum expectation value of $`J^z`$), and also contain components of exclusively different $`J^z`$ values (the $`J^z`$ operator does not connect the ground state to the excited state). One doublet has large $`|\pm 4`$ and $`|\pm 1`$ components, while the other has large $`|\pm 5`$ and $`|\pm 2`$ components. Hence, the two doublets have $`M_J`$ components that differ by $`\pm 1`$, and a neutron spin-flip induced transition from one to the other is allowed, consistent with what is found in the inelastic neutron scattering results presented above.
We have confirmed these conclusions based on our point charge analysis by performing a more sophisticated first principles calculation that take into account both electrostatic and covalency effects as well as the intra atomic and configurational interactions. This approach, described in Appendix B, does not restrict the decomposition of the electronic energy levels into a fixed $`|J,M_J`$ manifold as is usually done using the Steven’s operator equivalents, as discussed above. In the results presented in Appendix B, we find that the two lowest energy doublets have a leading $`M_J=\pm 4`$ and $`M_J\pm 5`$ components, respectively (close to 90% of the weight). Table 2 in Appendix B lists three very similar energy level structures given for slightly different constraints on the crystal field parameters. The theoretically determined energy levels (mean values: 0, 13, 60 and 83 cm<sup>-1</sup>, that is 0, 19, 86, and 119 K, respectively, be compared with the experimental levels determined by inelastic neutron diffraction (0, 17, 115 and 168 K). The corresponding wavefunctions allow for a good estimate of the magnetic susceptibility of the diluted and concentrated compounds as shown by Table 3 and Figs. 2, 3, and 8.
From these results, a picture of the low temperature single-ion properties of Tb<sup>3+</sup> can be deduced. Considering the structure of the eigenstates of the two lowest doublets, a calculation of their $`g`$tensors indicates extremely strong Ising like anisotropy along the appropriate $`(111)`$ axes (the axis formed by joining the two centers of the tetrahedra that the ion belongs to) for each Tb<sup>3+</sup> ion at low temperatures. In summary, based on both point charge and ab inition calculations, we have strong evidence of a doublet$``$doublet scheme for Tb<sup>3+</sup>, and a substantial single-ion anisotropy in $`\mathrm{Tb}_2\mathrm{Ti}_2\mathrm{O}_7`$ making the the Tb<sup>3+</sup> moment effectively Ising$``$like for $`TO(10^1)`$ K. The consequences of this result will be discussed in Section IV.
### B Susceptibility
The zero field susceptibility measurements of the powder Tb-diluted compound (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> shown in Fig. 3 enables us to gain a more concise understanding of the nature of the crystal field levels of $`\mathrm{Tb}^{3+}`$ in $`\mathrm{Tb}_2\mathrm{Ti}_2\mathrm{O}_7`$, and the single ion properties of Tb<sup>3+</sup>. Indeed, the dilute concentration of $`\mathrm{Tb}^{3+}`$ions in the system removes the effect of magnetic interactions and in principle leaves only crystal field contributions to the magnetic susceptibility. Having obtained strong evidence for a doublet$``$ doublet scheme for the Tb<sup>3+</sup> ions we now proceed further and analyse the d.c. magnetic susceptibility by constructing a phenomenological expression for the susceptibility based on a van Vleck equation for such a doublet$``$ doublet energy level structure . This allows us to extract the size of the magnetic moments of both doublets from experiment, as well as check for consistency with the conclusions based on our crystal field calculations of Appendix B.
Due to the powder nature of our (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> sample, the random orientation of the grains can lead to both parallel and transverse contributions to the susceptibility at first and second order. For the doublet-doublet structure at low temperature eluded to earlier, the van Vleck equation for the susceptibility has the general form
$$\chi =\frac{g^2\mu _B^2N_{\mathrm{Tb}}}{3k_\mathrm{B}}\left(\frac{a/T+b+e^{\mathrm{\Delta }/T}(c/Td)}{1+e^{\mathrm{\Delta }/T}}\right),$$
(6)
where $`g`$ is the Landé factor, equal to 3/2 for $`\mathrm{Tb}^{3+}`$, $`\mu _B`$ is the Bohr magneton and $`N_{\mathrm{Tb}}=0.04N_0`$, where $`N_0`$ is Avogadro’s constant. The adjustable parameters $`a,b,c,d`$ are defined through second order perturbation theory to be
$`\begin{array}{cc}a={\displaystyle \underset{\alpha ,n_0,m_0}{}}|n_0|J^\alpha |m_0|^2,& c={\displaystyle \underset{\alpha ,n_1,m_1}{}}|n_1|J^\alpha |m_1|^2,\\ b=2{\displaystyle \underset{\alpha ,n_0,m_{i0}}{}}{\displaystyle \frac{|n_0|J^\alpha |m_i|^2}{\mathrm{\Delta }_{0,i}}},& d=2{\displaystyle \underset{\alpha ,n_1,m_{i1}}{}}{\displaystyle \frac{|n_1|J^\alpha |m_i|^2}{\mathrm{\Delta }_{1,i}}},\end{array}`$
where $`n_0,m_0`$ label states within the ground state doublet, $`n_1,m_1`$ label states in the excited doublet while the index $`i`$ defines any state from the $`i`$th crystal field level. The $`\mathrm{\Delta }`$’s represent crystal field energy level differences (in Kelvin) and $`\alpha =x,y`$ or $`z`$. The fitting parameters $`a`$ and $`c`$ are due to first order terms in perturbation theory while $`b`$ and $`d`$ are from second order terms and give rise to temperature independent van Vleck paramagnetism contributions to $`\chi `$.
We have performed a least squares fit to the susceptibility data of (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> up to a temperature $`30`$ K, which is approximately where thermal contributions from a crystal field level at $``$ 100 K become non-negligible. Because of the narrow energy spacing between the ground state and excited state doublet, all four adjustable parameters $`a,b,c,d`$ are important in fitting the susceptibility data. Based on the crystal field calculations of the previous section, the specific heat analysis of the doublet-doublet gap in the concentrated Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> sample, and the evidence of an anisotropy gap of $``$ 17 K observed in inelastic neutron measurements on the same sample , we have carried out the fit of the low temperature behavior of the susceptibility using a doublet-doublet gap $`\mathrm{\Delta }`$ ranging from $`[1224]`$ K. We find that the goodness of fit is quite flat in this range for $`\mathrm{\Delta }`$. However the magnitude of the adjustable parameters do not deviate strongly and can be determined to a reasonable degree of accuracy over this interval. Using values of the gap outside of this interval yields a noticeably poorer goodness of fit. In Fig. 8, we show the fit to susceptibility data for the dilute (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> sample using an anisotropy gap of 17 K as well as the results for the ab-initio crystal field calculations.
We can interpret the fitted results for the susceptibility data by making use of our crystal field results and Eq. (6). Due to the strong Ising like nature of the $`g`$tensors of the two theoretically calculated doublets, transverse terms such as $`n_i|J^\pm |m_i`$ between two states within a doublet are negligible compared to $`n_i|J^z|n_i`$. Thus, we expect that $`a|n_0|J^z|n_0|^2`$ and $`c|n_1|J^z|n_1|^2`$ and consequently, both the $`a`$ and $`c`$ terms represent permanent moment contributions to the susceptibility. For the same reason, the $`g`$tensor characterizing the ground state is extremely anisotropic with essentially only a $`g_{}`$ component along the local $`111`$ direction with very little $`g_{}`$ component. As a result, and this is the most important point of the paper:
At a temperature $`T<O(10^1)`$ K, Tb<sup>3+</sup> ions can be considered to a very good approximation as (classical) Ising magnetic moments confined to point parallel or antiparallel to their local $`(111)`$ directions.
Over the range mentioned above for the doublet-doublet gap $`\mathrm{\Delta }`$, we have found that the magnitude of the moment in the doublet ground state to be $`5.10\pm 0.3\mu _B`$. Overall, we find that our best fit for the susceptibility data gives $`a=11.6\pm .1,b=1.53\pm .04\mathrm{K}^1,c=15.7\pm 4.0`$ and $`d=.71\pm .05\mathrm{K}^1`$. The value of $`c`$ gives a magnitude for the moment of the first excited doublet of $`5.9\pm .8\mu _B`$.
Both fitted moments are compatible with the eigenstate structures of the doublets determined from the crystal field calculations presented in Appendix B. Additionally, the values of the paramagnetic terms $`b`$ and $`d`$ in the susceptibility are also consistent with our crystal field results. In particular, from the $`J^z`$ components of our calculated low energy doublet and singlet eigenstates, there will be predominant contributions to $`b`$ and $`d`$ coming from transverse angular momentum matrix elements connecting the two doublets, as well as additional contributions involving transverse matrix elements between the doublets and the higher energy singlet states at $`100`$ K. The doublet-doublet coupling will give equal (in magnitude) contributions to $`b`$ and $`d`$ while coupling to the singlets will give further positive contributions to $`b`$ while reducing the value of $`d`$. This can be simply understood in terms of the signs of the denominators for each of these virtual excitation processes in our definitions of $`b`$ and $`d`$.
In summary, we are able to successfully fit the (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> susceptibility measurements at low temperature using a doublet-doublet picture consistent with our crystal field calculations. We find reasonable agreement in terms of calculated moments and paramagnetic contributions between that expected from theory and our fitted values. In particular, we find a permanent moment in the ground state of approximately $`5.1\mu _\mathrm{B}`$. This moment is intrinsic to the Tb<sup>3+</sup> ion and is not driven by exchange and/or dipolar interactions as occurs in LiTbF<sub>4</sub> . This value is also compatible to what is estimated from the limiting low-temperature muon spin relaxation rate $`1/T_1`$ found in Ref. , assuming a dipole coupling between a positive muon $`\mu ^+`$ bounded to an oxygen at $`2.5\AA `$ away from a $`5\mu _\mathrm{B}`$ Tb<sup>3+</sup> moment (see Appendix C).
## IV Discussion
Combining together our results from d.c. susceptibility measurements, specific heat data, inelastic neutron scattering data, and crystal field calculations, the following picture emerges:
* The Tb<sup>3+</sup> ion in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> carries a permanent magnetic moment of approximately $`5\mu _\mathrm{B}`$ . That moment is intrinsic to the Tb<sup>3+</sup> ion and is not driven by magnetic correlation from exchange and/or dipolar interactions.
* The ground state is well described as an Ising doublet with extremely anisotropic $`g`$tensor. In other words, the moments in the single-ion ground state are predominantly confined to point along the local $`(111)`$ directions.
* The first excited state is at an energy of approximately 18 $`\pm 1`$ K above the ground state, which is also characterized by a very anisotropic Ising-like $`g`$tensor.
* Consequently, in the absence of interactions, the Tb<sup>3+</sup> ions should be very well modeled below $`TO(10^1)`$ K as effective classical Ising spins confined to point along the local $`(111)`$ directions.
From d.c. susceptibility measurements at high temperature, we know that the magnetic interactions are predominantly antiferromagnetic, with $`\theta _{\mathrm{CW}}^{\mathrm{exchange}}13`$ K. Although the interactions are not small compared to the first excitation energy gap of 18 K, let us momentarily ignore the exchange coupling between the ground state and the excited doublet, and consider only the Ising-like ground state doublet. We then have a classical model with effective Ising spins pointing along their local $`(111)`$ directions. In other words, spins that can only point either inward or outward on a tetrahedron.
Because of the open structure of the pyrochlore lattice, we expect that the nearest-neighbor exchange interactions predominate. If we make the further approximation that only nearest-neighbors contribute to the antiferromagnetic interactions, we arrive at a scenario where Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> is effectively described by classical Ising spins pointing along the $`(111)`$ directions and coupled via nearest-neighbor antiferromagnetic exchange. As discussed in Refs. , such a model is not frustrated. Indeed, if we pick one spin, and chose either an “in” or “out” orientation for it, then the three other spins on the same tetrahedron must also take the same “in” or “out” configuration to minimize the exchange interactions. Since the pyrochlore lattice can be described as an FCC lattice with a tetrahedron as the basis cell, either an “all in” or “all out” configuration repeats identically on all basis units; one refers to such a magnetically long-range ordered structure as a $`Q=0`$ ground state .
The observation that Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> remains paramagnetic down to at least 70 mK, while sustaining important short range magnetic correlations, is therefore very puzzling. We believe that the most plausible explanation is that the interactions in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> involve further than nearest-neighbor interactions, $`J_{nn}`$. The presence of these interactions, as well as the long-range dipolar interactions, possibly “reintroduce” large frustration in the spin Hamiltonian describing Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, and conspire to destroy the “would-be” long-range Néel $`Q=0`$ ground state, and give rise to a collective paramagnet$``$spin liquid, ground state.
## V Conclusion
In conclusion, we have presented results from d.c. susceptibility, specific heat, inelastic neutron scattering, and crystal field calculations for the pyrochlore lattice antiferromagnet Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. We have obtained strong evidence that the Tb<sup>3+</sup> magnetic moment on the 16d site at $`T<2`$ K is intrinsic and is not induced by magnetic (exchange and/or dipolar) interactions or correlation effects such as found in LiTbF<sub>4</sub> , and is of the order of $`5\mu _\mathrm{B}`$ . All evidence points towards a very strong Ising like anisotropy for the doublet ground state which forces the resulting classical Tb<sup>3+</sup> Ising moments to point either parallel or antiparallel to their local $`111`$ direction. For antiferromagnetic exchange interactions, such strong anisotropy largely removes all local ground state spin degeneracy, and should naively force the system to possess an Ising-like long range order ground state with all spins in or out on a tetrahedron basis cell. The reason for the failure of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> to order at a temperature $`[10^010^1]`$ K set by the exchange part of the Curie-Weiss temperature remains unresolved at this time. The results presented here point towards the need to consider the role that exchange interactions beyond nearest-neighbor and dipolar interactions play in Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
## VI Acknowledgements
It is a pleasure to acknowledge contributions from B. Canals, S. Dunsiger, R. Kiefl and Z. Tun to these studies. We thank S. Bramwell, B. Ellman, P. Holdsworth, and D. Schmitt for very useful and stimulating discussions. This research was funded by NSERC of Canada via operating grants and also under the Collaborative Research Grant, Geometrically Frustrated Magnetic Materials. M.G. acknowledges the Research Corporation for a Research Innovation Award and a Cottrell Scholar Award, and the Province of Ontario for a Premier Research Excellence Award.
APPENDIX A:
Dipole Contributions to $`\theta _{\mathrm{CW}}`$
The contribution from dipolar interactions to $`\theta _{\mathrm{CW}}`$ may be determined from a high temperature series expansion of the long range dipole-dipole Hamiltonian to first order in $`\beta `$. That is to say, given the additive property of crystal field effects, exchange and dipolar interactions to $`\theta _{\mathrm{CW}}`$, i.e. $`\theta _{\mathrm{CW}}=\theta _{\mathrm{CW}}^{\mathrm{cf}}+\theta _{\mathrm{CW}}^{\mathrm{ex}}+\theta _{\mathrm{CW}}^{\mathrm{dip}}`$, we may determine $`\theta _{\mathrm{CW}}^{\mathrm{dip}}`$ from an expansion of
$`\chi ={\displaystyle \frac{g^2\mu _B^2}{kT}}{\displaystyle \underset{ij}{}}J_i^zJ_j^z_{H^{\mathrm{dip}}}`$
where $`z`$ defines some global direction and we assume $`T\theta _{CW}^{\mathrm{cf}}`$, i.e. isotropic spins. The angular brackets reflect an expectation value with respect to the dipole-dipole Hamiltonian $`H^{\mathrm{dip}}`$ defined by
$`H^{\mathrm{dip}}={\displaystyle \frac{1}{2}}g^2\mu _B^2{\displaystyle \underset{ij}{}}\left({\displaystyle \frac{𝐉_i.𝐉_j}{|𝐫_{ij}|^3}}{\displaystyle \frac{3(𝐉_i𝐫_{ij})(𝐉_j𝐫_{ij})}{|𝐫_{ij}|^5}}\right).`$
To first order in $`\beta `$ this yields
$`\chi ={\displaystyle \frac{g^2\mu _B^2NJ(J+1)}{3kT}}\left(1{\displaystyle \frac{g^2\mu _B^2J(J+1)}{3NkT}}\mathrm{\Lambda }\right),`$
and consequently that
$`\theta _{\mathrm{CW}}^{\mathrm{dip}}={\displaystyle \frac{g^2\mu _B^2J(J+1)}{3Nk}}\mathrm{\Lambda },(\mathrm{A}.1)`$
where
$`\mathrm{\Lambda }={\displaystyle \underset{ij}{}}{\displaystyle \frac{1}{|𝐫_{ij}|^3}}{\displaystyle \frac{3|r_{ij}^z|^2}{|𝐫_{ij}|^5}}.(\mathrm{A}.2)`$
We see from the above analysis that the evaluation of $`\theta _{\mathrm{CW}}^{\mathrm{dip}}`$ involves the summation of a conditionally convergent series (a lattice sum), and thus special care must be taken in its evaluation. In general, it can be treated in a controlled manner by the use of a rapid convergence factor via the Ewald method . Within the Ewald approach, the sum is split into two rapidly converging sums, one over the real space lattice and one in reciprocal space. Additionally, a surface (shape dependent) term also arises (which is interpreted as a demagnetization factor).
Indeed, if we approximate the powder grains of the (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> as spherical, then the sum Eq. (A.2) is identically zero. This can be understood by imagining Eq. (A.2) as a sum over parallel (along the $`z`$ axis) dipoles moments of magnitude unity. For a spherical object, it can be shown that such a sum must be identically zero for a system with cubic symmetry. Accordingly there would be no dipolar contribution to the measured value of $`\theta _{\mathrm{CW}}`$. On the other hand, we do not expect that the geometry of the powder grains is in fact spherical, (additionally there is also the possibility of effects from inter-granular interactions). Therefore to gain an estimate on the upper bound of the dipolar contribution, we carry out the lattice sum Eq. (A.2) for an infinitely long cylinder (needle shape) along the $`z`$ direction where surface effects are zero. This allows us to gain an approximate upper bound on the dipolar contribution as there are effectively no demagnetization effects. This calculation can be carried out rather simply by noting that for a spherical sample, Eq. (A.2), can be written using the Ewald method as :
$`\mathrm{\Lambda }_{\mathrm{sphere}}`$ $`=`$ $`\mathrm{\Lambda }^{bulk}+\mathrm{\Lambda }_{\mathrm{sphere}}^{surface},(\mathrm{A}.3)`$
$`\mathrm{where}`$
$`\mathrm{\Lambda }^{bulk}`$ $`=`$ $`{\displaystyle \frac{M}{L^3}}{\displaystyle \underset{ij}{}}[{\displaystyle \underset{𝐧}{}}{\displaystyle \frac{\alpha H(\alpha |𝐑_{ij}(𝐧)|)+(2\alpha /\sqrt{\pi })e^{\alpha ^2|𝐑_{ij}(𝐧)|^2}}{|𝐑_{ij}(𝐧)|^2}}`$
$``$ $`{\displaystyle \underset{𝐧}{}}{\displaystyle \frac{3\alpha |𝐑_{ij}^z(n^z)|^2H(\alpha |𝐑_{ij}(𝐧)|)e^{\alpha ^2|𝐑_{ij}(𝐧)|^2}}{|𝐑_{ij}(𝐧)|^4}}`$
$``$ $`{\displaystyle \underset{𝐧}{}}{\displaystyle \frac{(2\alpha /\sqrt{\pi })|𝐑_{ij}^z(n^z)|^2(3+2\alpha ^2|𝐑_{ij}(𝐧)|^2)e^{\alpha ^2|𝐑_{ij}(𝐧)|^2}}{|𝐑_{ij}(𝐧)|^4}}`$
$`+`$ $`{\displaystyle \underset{𝐧0}{}}4\pi (n^z/|𝐧|)^2e^{\pi ^2|𝐧|^2/\alpha ^2}e^{2\pi i𝐧.𝐫_{ij}/L}],(\mathrm{A}.4)`$
$`\mathrm{\Lambda }_{\mathrm{sphere}}^{surface}`$ $`=`$ $`{\displaystyle \frac{M}{L^3}}{\displaystyle \underset{ij}{}}{\displaystyle \frac{4\pi }{3}}.(\mathrm{A}.5)`$
The $`𝐑_{ij}(𝐧)=𝐫_{ij}/L+𝐧`$, $`i,j`$ label the dipole ($`\mathrm{Tb}^{3+}`$) sites within the cubic unit cell, $`L`$ is the length of the conventional cubic cell for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and $`M`$ is the number of cells in the sample. The function $`H(y)=(2/y\sqrt{\pi })_y^{\mathrm{}}e^{x^2}𝑑x`$ while $`𝐧=(k,l,m)`$ such that $`k,l,m`$ are integers and $`\alpha `$ is suitably chosen to ensure rapid convergence of the summations over $`𝐧`$. Based on the previously mentioned result that $`\mathrm{\Lambda }_{spher}`$ should vanish for a spherical sample, this implies that the first four terms in the above expression, $`\mathrm{\Lambda }^{bulk}`$, must sum to the opposite value of the last term ($`\mathrm{\Lambda }_{\mathrm{sphere}}^{surface}`$, the surface term), and indeed we have verified that this is the case numerically. For an infinitely cylindrical sample, the surface term, $`\mathrm{\Lambda }_{\mathrm{cylinder}}^{surface}=0`$, and thus one finds
$`\mathrm{\Lambda }_{\mathrm{cylinder}}`$ $`=`$ $`\mathrm{\Lambda }^{bulk}+\mathrm{\Lambda }_{\mathrm{cylinder}}^{surface}`$
$`=`$ $`\mathrm{\Lambda }^{bulk}`$
$`=`$ $`{\displaystyle \frac{M}{L^3}}{\displaystyle \underset{ij}{}}{\displaystyle \frac{4\pi }{3}}`$
$``$ $`1005{\displaystyle \frac{M}{L^3}}.(\mathrm{A}.6)`$
On the other hand, for a slab geometry, we have
$`\mathrm{\Lambda }_{\mathrm{slab}}`$ $`=`$ $`\mathrm{\Lambda }^{bulk}+\mathrm{\Lambda }_{\mathrm{slab}}^{surface}.(\mathrm{A}.7)`$
where
$`\mathrm{\Lambda }_{\mathrm{slab}}^{surface}`$ $`=`$ $`{\displaystyle \frac{M}{L^3}}{\displaystyle \underset{ij}{}}4\pi ,`$
thus arriving at
$`\mathrm{\Lambda }_{\mathrm{slab}}`$ $`=`$ $`{\displaystyle \frac{M}{L^3}}{\displaystyle \underset{ij}{}}{\displaystyle \frac{8\pi }{3}}.(\mathrm{A}.8)`$
Combining Eqs.(A.1), (A.6) and (A.8), and using a cubic cell length of $`L10.104\mathrm{\AA }`$, we arrive at upper and lower bounds for dipolar contributions to the Curie-Weiss temperature, namely $`2.4\theta _{\mathrm{CW}}^{\mathrm{dip}}+1.2`$ K.
APPENDIX B:
Crystal Field Calculations
Our aim in this Appendix is to pursue in more details and using a more sophisticated (ab-initio) approach the calculations of the magnetic susceptibilites of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. To do so, we need to determine the electronic structure of the ground <sup>7</sup>F<sub>6</sub> level of the Tb<sup>3+</sup> ($`4f^8`$ configuration) in these two compounds, deduce the wavefunctions and from there infer the value of the magnetic moment of the ground state level. The determination of the $`4f^8`$ electronic configuration is obtained by diagonalizing the following Hamiltonian for a generic $`f^n`$ configuration without making the fixed $`J`$ manifold approximation used in Section III.A:
$`H(f^n)`$ $`=`$ $`{\displaystyle \underset{k}{}}F^k(ff)f^k+\zeta (f)A_{so}+\alpha L(L+1)+`$
$`\beta C(\mathrm{G}_2)+\gamma C(\mathrm{R}_7)+{\displaystyle \underset{kq}{}}B_q^kC_q^k.(\mathrm{B}.1)`$
Let us explain first what are the different terms in $`H(f^n)`$.
* The $`F^k`$’s ($`k=2,4,6`$) are the electrostatic integrals (Slater’s parameters) which splits the $`4f^n`$ configurations into terms $`{}_{}{}^{2S+1}L`$ where $`S`$ is the total spin and $`L`$ is the total orbital angular momentum. The $`f^k`$’s are the associated two-electron operators .
* $`\zeta (f)`$ is the spin-orbit interaction integral which splits the terms into $`{}_{}{}^{2S+1}L_{J}^{}`$ levels. $`A_{so}`$ is the associated one-electron spin-orbit operator .
* $`\alpha `$, $`\beta `$ and $`\gamma `$ are parameters associated with effective two-body correction terms for inter-configuration interaction . $`C`$(G<sub>2</sub>) and $`C`$(R<sub>7</sub>) are the Casimir’s operators for groups G<sub>2</sub> and R<sub>7</sub>. When $`2<n<12`$, there can be several terms $`{}_{}{}^{2S+1}L`$ with the same $`S`$ and $`L`$ values in the f<sup>n</sup> configuration. For instance there are three <sup>5</sup>G terms in 4f<sup>8</sup> while there exists only one term <sup>7</sup>F. The states may differ by the way they are built from the parent configuration f<sup>n-1</sup>. An additional classification of the states is therefore necessary. It is done according to the irreducible representations of the the groups G<sub>2</sub> and R<sub>7</sub> and bestows additional quantum numbers to the states.
* the $`B_q^k`$ are the coefficients of the one-electron crystal field interaction which acts between $`|^{2S+1}LJM_J`$ sublevels. They can be theoretically predicted or extracted from fits of the energy levels (spectral lines) from experiments. In the point charge electrostatic model, their expression is :
$`B_q^k=(4\pi /2k+1)^{1/2}r^k\mathrm{\Sigma }_j(Q_j/R_j^{k+1})Y_k^q(\theta _j\varphi _j)`$
Where $`r^k`$ is a $`4f`$ electron radial integral, $`Q_j`$ is the point charge of ligand $`j`$, and $`R_j`$, $`\theta _j`$, and $`\varphi _j`$, are the polar coordinates of ligand $`j`$. The derivation of the $`B_q^k`$ for the covalent interactions is much more involved. The $`C_q^k`$ = (4$`\pi `$/2k+1)$`{}_{}{}^{1/2}[[𝐘_k^q]]`$, are the tensorial one-electron crystal field operators.
The evaluation of the matrix elements of i) the electrostatic interaction, ii) the spin-orbit interaction, iii) the free-ion configuration interaction, and iv) the crystal field interaction, between the states of the basis set chosen for the $`f^n`$ configurations is necessary in order to determine the eigenvalues and eigenvectors of the latter. The matrix elements are calculated by the means of tensorial algebra . Besides, if coupled $`|^{2S+1}LJM_J`$ states are chosen as the basis set, the one- or two-electron operators which are involved in the Hamiltonian cannot act directly on them. The calculation requires intermediary mathematical quantities known as reduced matrix elements which are tabulated for standard $`f^n`$ configurations ( ). Once evaluated, the complete matrix elements are multiplied by the associated parameters before diagonalization. The parameters are then determined by trial and error by successive diagonalizations and comparison of the eigenvalues with experimental energy levels. In the present case our main concern is the structure of the Tb<sup>3+</sup> ground level, so that the only specific material dependent parameters which have to be determined before diagonalizing $`H(f^n)`$ are the $`B_q^k`$ crystal-field interaction parameters. As already mentioned , the most convenient way to deduce the electronic structure of a rare earth ion in a solid compound, and hence determine the crystal-field parameters, is usually via analysis of its electronic spectrum by fitting the $`B_q^k`$ to match the frequency of the observed transitions. The strongest lines are due to electronic transitions partly allowed by the mixing of the ground configuration with opposite parity configurations (Judd-Ofelt mechanism) . However, in the case of pyrochlores, the site symmetry at the rare earth site is centrosymmetrical. The odd parity crystal field parameters vanish, the mixing of opposite parity configurations is impossible, hence no electric dipole transitions are detectable in the spectrum. Indeed, only the weak <sup>7</sup>F<sub>1</sub>$``$<sup>5</sup>D<sub>0</sub> and <sup>7</sup>F<sub>0</sub>$``$<sup>5</sup>D<sub>1</sub> magnetic dipole transitions were observed previously in the absorption spectra of the pyrochlore compounds Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and Eu<sub>2</sub>Sn<sub>2</sub>O<sub>7</sub> . Hence a complete set of “phenomenological” crystal field parameters (CFP) cannot be determined from optical absorption or emission spectra. However, recent inelastic neutron scattering experiments have been able to give information on the lowest electronic levels of Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . From these neutron results, some CFP’s can be deduced. CFP’s can also be calculated ab-initio from the compound structure and the atomic data of the constituents. Therefore, in what follows, two approaches are used to determine the CFP’s of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>:
* A full ab-initio calculation of the CFP’s utilizing the structural data of the compounds.
* The fit of the CFP’s from the Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> data and a transposition to the Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> compounds.
The next two steps for the calculations of the magnetic susceptibility are:
* The calculation of the 4$`f^8`$ electronic configuration utilizing plausible free ion parameters, and the fitted or calculated ab-initio values of the CFP’s.
* The calculation of the magnetic susceptibility of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> utilizing the wavefunctions derived from the previous step II.
These steps are detailed in the following :
IA) Ab-initio calculation of the CFPs
An ab-initio determination of the CFP’s is obtained by adding an electrostatic and a covalent contribution along lines similar to the ones developed in Refs. for oxygen ligands. The crystal structure, the ionic charges and the ionization energies of the ligands are used. In Ref. , experimental and predicted values of the parameters calculated by the “covalo-electrostatic” model were compared for ten compounds. The mean deviation between experimental and calculated values:
$`\mathrm{\Delta }B^k/B^k`$ $`=`$ $`\left[{\displaystyle \underset{kqk}{}}(B_{qe}^kB_{qc}^k)^2/{\displaystyle \underset{kqk}{}}(B_{qe}^k)^2\right]^{1/2}`$
$`=`$ $`[1+1/(S^k)^22\mathrm{cos}(R^k)/S^k]^{1/2}`$
where $`S^k`$ and $`R^k`$ are the scale and reliability factors listed in Table 7 of Reference . $`\mathrm{\Delta }B^k/B^k`$ is found to be equal to 52, 30 and 23% for $`k=2`$, $`k=4`$ and $`k=6`$, respectively. Such is the uncertainty which can be expected from a “blind eyed” prediction of the CFP’s of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
As mentioned earlier, the space group of rare earth titanates with the pyrochlore structure is Fd$`\overline{3}`$m. The eight oxygen first neighbours form a distorted cubic polyhedron. Two oxygens occupy ideal positions on opposite summits of the cubic threefold axis. The three \[\[ sides \]\] of the cube originating from each of these two summits are equally elongated. The $`a`$ cubic lattice parameter is equal to 10.15 and 10.09 $`\AA `$, and $`x`$ the positional parameter for the six displaced oxygens to 0.3 and 0.2968 for the dense and dilute compound respectively .
The site symmetry at the rare earth site is reduced from O<sub>h</sub> (cubic) to D<sub>3d</sub>. The remaining threefold order symmetry axis imposes for the crystal field parameters the condition $`q=0`$, modulo 3, so that the non-zero even $`k`$ CFP’s are $`B_0^2`$, $`B_0^4`$, $`B_3^4`$, $`B_0^6`$, $`B_3^6`$ and $`B_6^6`$. The predicted CFP values are reported in Table 1. The distances between the Tb<sup>3+</sup> ion and the oxygens on the threefold axis is short (2.20 and 2.18 $`\AA `$ for the dense and the dilute compound, respectively) while the distances to the six peripheral oxygens are much larger (2.52 and 2.49 $`\AA `$, respectively). This explains why the “axial” $`B_0^k`$ parameters are much larger than the “azimuthal” $`B_q^k`$’s .
IB) Experimental determination of the CFPs.
The CFPs can be determined by fitting the energy levels of the $`f^n`$ configuration either using spectroscopic or inelastic neutron scattering data from other pyrochlore compounds. Here, we can use the $`B_0^2`$ from spectroscopic data on Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and the other $`B_q^k`$ from neutron scattering data on Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> .
* Determination of $`B_0^2`$ from spectroscopic data on Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>
The <sup>7</sup>F<sub>1</sub>$``$<sup>5</sup>D<sub>0</sub> and <sup>7</sup>F<sub>0</sub>$``$<sup>5</sup>D<sub>1</sub> magnetic dipole transitions were observed previously in the electronic absorption spectrum of Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> . In the latter, the <sup>7</sup>F<sub>1</sub> and <sup>5</sup>D<sub>1</sub> splittings amount to 291 K and 51.9 K, respectively. A fit in $`4f^6`$(Eu<sup>3+</sup>) yields $`B_0^2`$ = 684 K. The transposition to Tb<sup>3+</sup> is made assuming the crystal field parameter is scaled by the ratio of radial integrals: $`B_0^2`$(Tb<sup>3+</sup>) = $`B_0^2`$(Eu<sup>3+</sup>) $`\times `$$`r^2`$(Tb<sup>3+</sup>)/$`r^2`$(Eu<sup>3+</sup>) = $`B_0^2`$(Eu<sup>3+</sup>) $`\times `$ 0.91 = 622 K. This “experimental” value, listed in Table 1, and referred to as (a), has the same sign, but it is about half of the predicted value. As pointed out hereabove, the uncertainty on the calculated $`k=2`$ parameters is large.
* CFP determination in Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
Siddharthan et al. recently reported results from ineleastic neutron scattering at low temperature in Ho<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. They determined the six lowest E irreducible representations of the <sup>5</sup>I<sub>6</sub> ground state level. Utilizing their experimental values, we fitted the CFP’s of Ho<sup>3+</sup> while maintaining the ratio between CFP’s with the same $`k`$ value close to the theoretical ratio. The program ATOME was used for the refinement . In this program, the basis set is composed of Slater determinants which makes unnecessary the use of tables of reduced matrix elements . The evaluation of the matrix elements is straightforward, but the configuration cannot be truncated. Indeed, each eigenvector being a linear combination of a large number of Slater determinants, none of the latter can be omitted. As a consequence, all the 1001 states of 4f<sup>10</sup> configuration are included in the diagonalization matrix. The basis is large (1001) but still tractable. The final mean deviation between experimental and calculated levels was equal to 7.8 K. The fitted CFP’s are : $`B_0^4`$= 3173 K, $`B_3^4`$= -1459 K, $`B_0^6`$= 1343 K, $`B_3^6`$= 1292 Kand $`B_6^6`$= 609 K. As pointed out hereabove, the CFP’s were then scaled according to the ratio between the radial integrals of Ho<sup>3+</sup> and Tb<sup>3+</sup> to give the experimental CFP’s for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. Namely, $`B_q^k`$(Tb<sup>3+</sup>) = $`B_q^k`$(Ho<sup>3+</sup>) $`\times `$$`r^k`$(Tb$`{}_{}{}^{3+})`$/$`r^k`$(Ho<sup>3+</sup>). These “experimental values” are are listed in Table 1, and referred to as (b). The experimental $`k=4`$ and $`k=6`$ CFP’s are 1.7 and 1.9 times larger than the theoretical values which is somewhat unusual.
III) Calculation of $`4f^8`$ electronic configuration.
The calculation of the $`4f^8`$ electronic configuration is done by the means of program $`f^n`$ utilizing the Hamiltonian $`H(f^n)`$ in B.1 acting on coupled states <sup>2S+1</sup>L<sub>J</sub>. Contrary to program ATOME previously mentioned, program f<sup>n</sup> can work on a truncated basis, which is necessary to resolve the $`4f^8`$configuration of Tb<sup>3+</sup>with a large number of states (3003 states). The interaction matrix is built on a 387x387 basis set comprising the following <sup>2S+1</sup>L terms of the Tb<sup>3+</sup>( $`4f^8`$) configuration: <sup>7</sup>F, <sup>5</sup>D(1,2,3), <sup>5</sup>F(1,2), <sup>5</sup>G(1,2,3), <sup>3</sup>P(1,2,3,4,5,6) and <sup>1</sup>S(1,2,3,4). The conjugate configuration of Tb<sup>3+</sup> is that of Eu<sup>3+</sup> with $`4l+28=6`$ electrons. The $`4f^6`$ (n=6) configuration of Eu<sup>3+</sup> contains exactly the same number of basis states (e.g. \[4$`l`$+2\]!/\[\[4$`l`$+2-n\]!n!\] = 3003 states) as Tb<sup>3+</sup>, the same terms, and the same levels. The interactions involving an odd number of electrons have a reverse sign in 4f<sup>6</sup> and 4f<sup>8</sup>. For instance the <sup>2S+1</sup>L<sub>J</sub> levels appear in a reverse order, and so do the crystal field sub-levels. In addition, the terms determined by the electrostatic interaction (two electron interaction) appear up in the same order for Eu<sup>3+</sup> and Tb<sup>3+</sup>.
For Eu<sup>3+</sup> the above quoted basis had proved large enough to allow a simulation of the levels up to <sup>5</sup>D<sub>2</sub> (30219 K) without drastic truncation effects . The $`F^k`$’s were assigned the Gd<sup>3+</sup> values given in Ref. , that is 147289 K, \[\[ 102479 K \]\], and 55868 K for $`k=`$ 2, 4 and 6, respectively. $`\alpha `$, $`\beta `$, $`\gamma `$ were ascribed the Nd<sup>3+</sup> values fitted in Ref. , that is 30.98 K, -1005.03 K, and 2510.48 K, respectively. The spin-orbit coupling constant $`\zeta (f)`$ was set equal to 2446.30 K which is a standard value for Tb<sup>3+</sup> . $`B_0^2`$ was assigned the transposed value quoted hereabove, and the other CFP’s the values listed in Table 1 obtained after rescaling the $`B_q^k`$ extracted from fits to the levels of Ho<sup>3+</sup>. The diagonalization of the interaction matrix gives the energy levels and the corresponding leading eigenvectors listed in Table 2. The lowest levels are two doublets. Both states are rather Ising-like, with nearly exclusive ($`0.95`$) $`M_J=\pm 4`$ and $`M_J=\pm 5`$ components. This is caused by the very “axial” crystal field parameters. In other words, the $`M_J=\pm 4`$ and $`M_J=\pm 5`$ carry, respectively, about 90% of the weight of the ground state and first excited state doublet wavefunctions.
III) Calculation of the magnetic susceptibility of Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
The d.c. magnetic susceptibility is by Van Vleck’s formula using the eigenvectors determined in the previous step.
$`\chi ={\displaystyle \frac{N\mu _\mathrm{B}^2}{_ie^{E_i^{(0)}/k_\mathrm{B}T}}}\left[{\displaystyle \underset{i}{}}{\displaystyle \frac{(ϵ_i^{(1)})^2}{k_\mathrm{B}T}}2ϵ_i^{(2)}\right]e^{E_i^{(0)}/k_\mathrm{B}T}`$
where $`N`$ is the number of moles of Tb<sup>3+</sup>, $`\mu _\mathrm{B}`$ is the Bohr magneton, $`k_\mathrm{B}`$ is the Boltzmann constant, $`E_i^{(0)}`$ is the energy of the $`i^{th}`$ level. Besides,
$`ϵ_i^{(1)}=\psi _i|\stackrel{}{L}+\stackrel{}{2S}|\psi _i`$
$`ϵ_i^{(2)}={\displaystyle \underset{ji}{}}{\displaystyle \frac{(\psi _i|\stackrel{}{L}+\stackrel{}{2S}|\psi _j)^2}{E_i^{(0)}E_j^{(0)}}}`$
The results as a function of the temperature are given in Table 3. The values for $`\mu (T)`$ are very similar for the Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> compounds. Therefore any large experimentally stated difference between the two compounds at low temperature cannot be accounted for by a difference in the individual characteristics of Tb<sup>3+</sup> in the dense Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and the dilute (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>.
TABLE 1
Empirical CFP’s deduced by the covalo-electrostatic model for= Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, and “experimental” CFP’s for Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. The latter are obtained using two approaches. (a) refers to the value of $`B_0^2`$ obtained by transposing to Tb<sup>3+</sup> the value of $`B_0^2`$ determined from spectroscopic data of Eu<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>. (b) refers to the values $`B_q^{k4}`$ obtained by transposing to Tb<sup>3+</sup> the values of $`B_q^{k4}`$ determined from inelastic neutron data on Eu<sub>2</sub>Ti<sub>2</sub> O<sub>7</sub>. All values in K. (PCEM = point charge electrostatic model.)
| | | $`B_0^2`$ | $`B_0^4`$ | $`B_3^4`$ | $`B_0^6`$ | $`B_3^6`$ | $`B_6^6`$ |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | PCEM | 471 | 708 | -187 | | | |
| Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> | Cov. | 610 | 1599 | -227 | 1261 | 314 | 482 |
| | Total | 1081 | 2307 | -414 | 1261 | 314 | 482 |
| | PCEM | 407 | 731 | -210 | | | |
| (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> | Cov. | 609 | 1711 | -288 | 1324 | 389 | 571 |
| | Total | 1016 | 2442 | -498 | 1324 | 389 | 571 |
| “Experimental” | | 622<sup>a</sup> | 3691<sup>b</sup> | -1698<sup>b</sup> | 1731<sup>b</sup> | 1665<sup>b</sup> | 784<sup>b</sup> |
.
TABLE 2
Lowest energy levels (in K), irreducible representations in (Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, and leading compositions of the corresponding eigenvectors. Lines labelled (a) and (b) are for the predicted projections for the dilute Tb<sub>0.02</sub>Y<sub>0.98</sub>)<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> and dense Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub>, respectively, using the CFP’s listed in Table 1. Lines labelled (x3) are for the predicted projections for the dense Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> using the “experimental” CFP’s listed in Table 1.
| (E) | En | <sup>7</sup>F<sub>6</sub> , +/-4 | <sup>7</sup>F<sub>6</sub> ,+/-1 | <sup>7</sup>F<sub>6</sub>,-/+5 | <sup>5</sup>G<sub>6</sub>(1), +/-4 | <sup>5</sup>G<sub>6</sub>(3) , +/-4 | |
| --- | --- | --- | --- | --- | --- | --- | --- |
| (a) | 0. | -0.97 | +/-0.06 | -/+0.06 | +0.15 | -0.14 | |
| (b) | 0. | -0.97 | +/-0.05 | -/+0.06 | +0.15 | -0.14 | |
| (c) | 0. | -0.95 | +/-0.13 | -/+0.13 | +0.14 | -0.13 | |
| (E) | En | <sup>7</sup>F<sub>6</sub> , +/-5 | <sup>7</sup>F<sub>6</sub> ,+/-2 | <sup>7</sup>F<sub>6</sub>,-/+4 | <sup>5</sup>G<sub>6</sub>(1), +/-5 | <sup>5</sup>G<sub>6</sub>(3) , +/-5 | |
| (a) | 18.7 | -0.96 | +/-0.13 | -/+0.07 | +0.14 | -0.13 | |
| (b) | 15.9 | -0.97 | +/-0.11 | -/+0.06 | +0.14 | -0.14 | |
| (c) | 21.6 | -0.92 | +/-0.28 | -/+0.14 | +0.14 | -0.13 | |
| (A2) | En | <sup>7</sup>F<sub>6</sub>,-3 | <sup>7</sup>F<sub>6</sub>,3 | <sup>7</sup>F<sub>6</sub>,-6 | <sup>7</sup>F<sub>6</sub>,6 | <sup>5</sup>G<sub>6</sub>(1),-3 | <sup>5</sup>G<sub>6</sub>(3),3 |
| (a) | 86.5 | +0.67 | +0.67 | +0.16 | -0.16 | -0.10 | -0.10 |
| (b) | 87.9 | +0.67 | +0.67 | +0.16 | -0.16 | -0.10 | -0.10 |
| (c) | 85.0 | +0.65 | +0.65 | +0.22 | -0.22 | -0.10 | -0.10 |
| (A1) | En | <sup>7</sup>F<sub>6</sub>,-3 | <sup>7</sup>F<sub>6</sub>,3 | <sup>7</sup>F<sub>6</sub>,-6 | <sup>7</sup>F<sub>6</sub>,6 | <sup>7</sup>F<sub>6</sub>,0 | |
| (a) | 121.1 | +0.66 | -0.66 | +0.21 | +0.21 | +0.07 | |
| (b) | 118.2 | +0.66 | -0.66 | +0.20 | +0.20 | +0.06 | |
| (c) | 119.6 | +0.64 | -0.64 | +0.25 | +0.25 | +0.11 | |
TABLE 3
Ionic magnetic moment (in Bohr magnetons) and inverse molar magnetic susceptibility as a function of temperature.
| $`T`$ (K) | $`\mu `$ ($`\mu _\mathrm{B}`$) | $`1/\chi `$ (mole/emu) | $`1/\chi `$ (mole/emu) |
| --- | --- | --- | --- |
| | Tb<sub>0.02</sub>Y<sub>0.98</sub>Ti<sub>2</sub>O<sub>7</sub> | Tb<sub>0.02</sub>Y<sub>0.98</sub>Ti<sub>2</sub>O<sub>7</sub> | Tb<sub>2</sub>Ti<sub>2</sub>O<sub>7</sub> |
| 1 | 5.90 | 5.8 | 0.11 |
| 5 | 6.84 | 21.4 | 0.41 |
| 10 | 7.58 | 34.8 | 0.67 |
| 15 | 7.97 | 47.3 | 0.92 |
| 20 | 8.20 | 59.5 | 1.17 |
| 30 | 8.47 | 83.5 | 1.65 |
| 300 | 9.22 | 706 | 14.09 |
. |
warning/0001/astro-ph0001226.html | ar5iv | text | # Wind properties of Wolf-Rayet stars at low metallicity: Sk 41 (SMC) Based on observations collected at the European Southern Observatory, La Silla, Chile. Proposal Nos. 61.D–0680 and 63.H–0683.
## 1 Introduction
Galaxies containing the youngest starbursts, with ages of a few Myr, are known as Wolf-Rayet galaxies (Vacca & Conti 1992) since they show the spectroscopic signature of large numbers of WR stars. At present, in excess of 140 WR galaxies have been discovered (Schaerer et al. 1999), including examples at very low metallicity, $`Z`$, such as I Zw18 with $`Z`$=0.02$`Z_{}`$ (De Mello et al. 1998), probably typical of star forming galaxies in the early universe.
The O star content of starburst galaxies is derived from nebular Balmer line fluxes, while WR populations are obtained from comparing broad He ii $`\lambda `$4686 and C iv $`\lambda `$5801-12 emission line fluxes with calibrations of individual Galactic and/or LMC WR stars (Schaerer & Vacca 1998).
Radiative driven wind theory predicts that $`\dot{M}Z^{0.5}`$ (Kudritzki et al. 1989). Incorporating this effect into massive star evolutionary models implies that the minimum initial mass star reaching the WR phase through single star evolution is predicted to be a function of $`Z`$; 25$`M_{}`$ in the Galaxy, 35$`M_{}`$ in the LMC and 45$`M_{}`$ in the SMC (Maeder 1997). Although WR winds are considered to be driven by radiation pressure, their mass-loss rates are assumed to be independent of metal content in current evolutionary calculations.
Do WR stars have comparable mass-loss properties and emission line strengths within various metallicity regions? Attempts to establish differences between the properties of O and WR stars in the Galaxy and LMC have proved inconclusive (e.g. Puls et al. 1996; Crowther & Smith 1997), since the metal content of the LMC differs from the Solar neighbourhood only by a factor of $``$2.
Consequently, the SMC represents our closest neighbour in which to establish whether the stellar properties of O and WR stars are affected by low metal content (12+log O/H=8.1; Russell & Dopita 1990). Walborn et al. (1995), Puls et al. (1996) and Prinja & Crowther (1998) demonstrated that SMC O-type stars indeed reveal lower mass-loss rates and slower winds, except amongst very early O giants and supergiants. Comparisons for Wolf-Rayet stars are more problematic, since they are very rare in the SMC, and almost entirely binaries. Omitting O3 If/WN stars, the LMC contains 125 bona-fide WR stars (Breysacher et al. 1999), in contrast to only eight WR stars in the SMC, which are listed in Table 1. Crowther (1999) discusses the influence of metallicity dependent WR mass-loss rates on stellar spectra and ionizing flux distributions.
HD 5980 has recently received considerable interest because of a Luminous Blue Variable (LBV)-type eruption in one of its component stars (Barba et al. 1995; Koenigsberger et al. 1998; Moffat et al. 1998), although its variability and multiplicity hinders the reliability of spectroscopic analysis.
In order to avoid uncertainties caused by binarity, a study of the only known single WR star in the SMC, Sk 41 (AB4), as discussed by Moffat (1988), is presented here, using the non-LTE line blanketed model atmosphere code of Hillier & Miller (1998). Its stellar properties are compared with counterparts in the Galaxy and LMC, and the role of metal content on spectral appearance is investigated. In addition, the line luminosities of He ii $`\lambda `$4686 and C iv $`\lambda \lambda `$5801-12 in SMC WR stars are compared with Galactic and LMC WR stars, and a new calibration is derived.
## 2 Observations
UV, optical and near-IR spectroscopy of Sk 41 has been obtained from the International Ultraviolet Exporer (IUE) archive, the 3.9m Anglo-Australian Telescope (AAT) and the European Southern Observatory 3.5m New Technology Telescope (NTT), respectively. Following Smith et al. (1996), a spectral classification of WN5h is obtained for Sk 41, in comparison to previous classifications of WN6-A (Walborn 1986) or WN4.5 (Conti et al. 1989). Although it could be claimed that a WN5h+abs classification is more appropriate from the Smith et al. definition, we adopt WN5h because of the close similarity of Sk 41 with other WN5h stars, such as HD 65865 (WR10) and R136a1 (BAT99-108).
### 2.1 Optical spectroscopy
An optical spectrogram of Sk 41 was obtained at the AAT during 1992 November 3–6, using the RGO spectrograph, 25cm camera, Tektronix CCD (1024 $`\times `$ 1024, 24$`\mu `$m pixels), 1200V and 1200B gratings, plus a slit width of 2<sup>′′</sup>. The measured spectral resolution in the extracted spectra is 1.6–1.8Å, using the FWHMs of the Cu-Ar arc spectra. Four settings covered 3670–6005Å, plus 6455–7263Å. A standard reduction was carried out as discussed by Crowther & Smith (1997). Subsequent analysis made use of dipso (Howarth et al. 1998). In addition, a low resolution flux calibrated spectrum of Sk 41 from Torres-Dodgen & Massey (1988) has been used.
### 2.2 Near-IR spectroscopy
Long slit, near-IR spectroscopy of Sk 41 was acquired with the NTT, using the Son OF Isaac (SOFI) instrument, a 1024x1024 pixel NICMOS detector, and low resolution IJ (GRB) and HK (GRR) gratings on 1 Sept 1999. The spectral coverage was 0.94–1.65$`\mu `$m and 1.50–2.54$`\mu `$m, respectively, with dispersions of 7.0Å/pix and 10.2Å/pix. The 0.6 arcsec slit provided a 2 pixel spectral resolution of 14–20Å. The total integration time was 960 sec at each grating setting. Atmospheric calibration was achieved by observing HD 10747 (B3V) immediately before or after Sk 41, at an close airmass (within 0.03). Similar observations of HD 2002 (F5V) permitted a relative flux correction, using a $`T`$=6,500K model atmosphere normalized to V=8.13 mag.
A standard extraction and wavelength calibration was carried out with iraf, while figaro (Shortridge et al. 1999) and dipso were used for the atmospheric and flux calibration, first artificially removing stellar hydrogen features from the B3V spectrum. Convolving our fluxed spectra with suitable filters indicates J$``$13.6, H$``$13.5, and K$``$13.3 mag.
### 2.3 UV spectroscopy
A short wavelength (SWP48325), high resolution (HIRES) IUE observation of Sk 41 was retrieved from the World Data Centre at the Rutherford Appleton Laboratory. This was obtained on 7 August 1993, with an exposure time of 20,340 sec and was reduced using iuedr (Giddings et al. 1996). In addition, three final archive (IUEFA), low resolution, large aperture datasets were obtained from STScI. Two short wavelength (SWP6195, SWP107270) and one long wavelength (LWR5359) spectra were obtained between 15 Aug 1979 and 2 Dec 1980, with exposure times of 900, 1200 and 960 sec, respectively.
## 3 Spectroscopic analysis
The model calculations are based on the iterative technique of Hillier (1987, 1990) which solves the transfer equation in the co-moving frame subject to statistical and radiative equilibrium, assuming an expanding, spherically-symmetric, homogeneous and static atmosphere. Allowance is made for line blanketing and clumping following the formulation of Hillier & Miller (1998).
Calculations consider detailed model atoms of H i, He i-ii, C iv, N iii-v, O iii-vi, Si iv and Fe iv-vii (see Dessart et al. (2000) for details of the source of atomic data). Weak transitions of iron ($`gf10^4`$) have been excluded without affecting the emergent spectrum. In total, 1027 full levels and 7680 non-LTE transitions are simultaneously considered. These are combined into 275 ‘super levels’, with solely the populations of the super level calculated in the rate equations. Populations of individual atomic levels are then determined by assuming that it has the same departure coefficient as the super level to which it belongs.
In line with recent iron determinations for O stars in the SMC (Haser et al. 1998), abundances of elements other than hydrogen and helium are fixed at 0.2 $`\times `$ Solar (Si=0.02% by mass, Fe=0.03% by mass) or 0.2 $`\times `$ Solar CNO-equilibrium values (C=0.008%, N=0.3%, O=0.005% by mass).
The analysis technique follows that of Crowther et al. (1995), such that diagnostic lines of He i ($`\lambda `$10830), He ii ($`\lambda `$5412) and H i (H$`\beta `$+He ii $`\lambda `$4859) are chosen to derive the stellar temperature, mass-loss rate, luminosity and hydrogen content. The mass-loss rate is actually derived as the ratio $`\dot{M}/\sqrt{f}`$, where $`f`$ is the volume filling factor. This can be constrained by fits to the electron scattering wings of the helium line profiles. A wind velocity of $``$1300 km s<sup>-1</sup> is obtained from IUE/HIRES C iv $`\lambda `$1548-51 observations, although the S/N of this dataset is low. A standard $`\beta `$=1 velocity law is adopted, in order to provide consistency with recent analyses. However, recent evidence indicates that early WR winds accelerate more slowly (e.g. Lepine & Moffat 1999), so we have also investigated the effect of a slow $`\beta `$10 law on derived stellar parameters. For Sk 41, a slow velocity law reveals a mass-loss rate that is systematically lower by $``$25%, with the stellar temperature and luminosity barely affected.
### 3.1 Stellar properties of Sk 41
Rectified optical (AAT/RGO) and near-IR (NTT/SOFI) spectroscopy of Sk 41 is compared with our synthetic spectra in Fig. 1. Overall, agreement is excellent for most H i, He i-ii, N iii-iv and C iv transitions. The fit to the He ii $`\lambda `$4686 electron scattering wing is excellent using a clumped model with a volume filling factor of $`f`$=0.1. Note that N v $`\lambda `$4603–20 emission is not predicted, while the H$`\delta `$ region, containing Si iv $`\lambda \lambda `$4088-4116 and N iii $`\lambda \lambda `$4097–4103 is underestimated. Consistency between optical and near-IR fits in Sk 41 is good, although near-IR He ii transitions, such as 1.01$`\mu `$m and 2.19$`\mu `$m, are $``$25% too strong, based on fits to optical He ii transitions. Similar results were obtained for strong-lined WNE stars by Crowther & Smith (1996).
Fig. 2 compares the dereddened energy distribution of Sk 41 with model predictions, using a distance of 60.2 kpc to the SMC (Westerlund 1997). The Seaton (1979) extinction law is adopted for the foreground extinction, assumed to be $`E_{BV}`$=0.03 mag. In addition, the Bouchet et al. (1985) SMC law as parameterised by Calzetti (priv. comm.) is used for internal SMC extinction. $`E_{BV}=0.07`$ mag is required to match the synthetic model (Fig. 2, inset box). SMC interstellar Lyman$`\alpha `$ absorbtion has been accounted for, assuming $`n`$(H i)=10<sup>21.8</sup> cm<sup>-2</sup> (Fitzpatrick 1985).
Synthetic filter photometry from Torres-Dodgen & Massey (1988) indicates $`b`$=13.22 mag for Sk 41, so the absolute b-magnitude of Sk 41 is $`M_b=`$6.1 mag. Agreement between the IUE dataset and the synthetic spectra (degraded to the IUE/LORES resolution) is reasonable, although imperfect in the $`\lambda \lambda `$1250–1500Å region, dominated by transitions of Fe v.
The derived stellar parameters for the WN5h star are $`T_{}`$=42kK, $`R_{}`$=13.6$`R_{}`$, log($`L/L_{}`$)=5.7, H/He=2.25 by number, and $`\dot{M}/\sqrt{f}`$=2.8$`\times `$10<sup>-5</sup>$`M_{}`$yr<sup>-1</sup>. The bolometric correction of this model is $``$3.9 mag, while the predicted number of ionizing photons shortward of H i $`\lambda `$911 and He i $`\lambda `$504 are log $`Q_0`$=49.45 s<sup>-1</sup> and log $`Q_1`$=48.77 s<sup>-1</sup>, respectively.
### 3.2 Comparison with WN stars in the Galaxy and LMC
Historically, WN5–6 spectral types have been considered solely as early WN (WNE) stars. In contrast, we assign a WNL status for those WN5–6 stars which contain atmospheric hydrogen following Smith et al. (1996), for consistency with evolutionary definitions. Consequently, the stellar properties of Sk 41 are shown in the upper panel of Fig. 3 (cross), together with a large sample of Galactic and LMC WN5–9 stars. Stellar parameters of the other stars are taken from Hamann et al. (1993, 1995), Crowther et al. (1995), Crowther & Smith (1997) and Crowther & Dessart (1998). Previous approaches were identical except that line blanketing and clumping were neglected, and nitrogen stellar diagnostics were employed by Crowther & Dessart (1998). The effect of selecting nitrogen, rather than helium diagnostics is indicated as an arrow for HD 38282 (BAT99-118, WN6h) in Fig. 3, although from the previous discussion no shift is appropriate for Sk 41.
The mass-loss rate of Sk 41 compares closely to other WNL stars with similar luminosities. A homogeneous model implies a wind performance number, $`\text{ .}Mv_{\mathrm{}}/(L_{}/c)`$, of 3.6 for Sk 41, which compares closely with the three luminous WN5h stars in R136a (Crowther & Dessart 1998). Considering a volume filling factor of $`f`$=0.1 reduces the wind performance number for Sk 41 to approximately unity. Although the wind velocity of Sk 41 is a factor of two lower than that of the R136a stars, it is comparable to the sole Galactic WN5h star HD 65865 (WR10), for which $`v_{\mathrm{}}`$=1500 km s<sup>-1</sup> (Hamann et al. 1993).
It remains to be successfully demonstrated whether WR winds in high metallicity environments can be driven solely by multiple scattering (Schmutz 1997). Therefore, the wind properties of Sk 41 provide an excellent challenge at low metallicities.
### 3.3 Evolutionary model predictions at low metallicity
Since Sk 41 is apparently single, how do its properties compare with theoretical expectations for single massive stars at SMC metallicities (Meynet et al. 1994)? In the lower panel of Fig. 3 we show the luminosity and hydrogen content (in mass fraction) for a large sample of Galactic and LMC WNL stars, as well as Sk 41 (again indicated by cross). The latest evolutionary tracks for metallicities appropriate to the SMC (0.2$`Z_{}`$, Meynet et al. 1994) are superimposed with initial masses of 40, 60, 85 and 120$`M_{}`$. The minimum initial mass that is predicted to progress through to the WN phase has 45–50$`M_{}`$, with log($`L/L_{})`$5.9, and a WN lifetime of $``$100,000 yr. Therefore, Sk 41 has a stellar luminosity which is 50% lower than the minimum that is predicted by evolutionary models. Since it is unlikely that Sk 41 is a disrupted binary (its radial velocity is typical of SMC stars), our results appear to identify a deficiency in present evolutionary models at low metallicity. Either less massive stars can advance to the WR stage, or the stellar luminosity decreases as the star reaches advanced evolutionary phases. These are completely determined by the mixing and the previous evolutionary phases, which remain poorly known.
## 4 Effect of metal content on spectral types of WN stars
If the mass-loss properties of WN stars in different environments are unaffected by metallicity, do differences in metal content have any effect on the emergent spectral appearance? To investigate this, we have fixed the stellar properties of Sk 41, and calculated additional models in which the metal content is varied by a factor of five, to solar (1.0$`Z_{}`$) or 0.04$`Z_{}`$, equivalent to twice the metal content of I Zw18 which is known to host WR stars (e.g. De Mello et al. 1998).
CNO elements are fixed at equilibrium WN values, appropriate for each environment (Meynet et al. 1994), such that the nitrogen content varies from 0.06% to 1.5% by mass. The atmospheric structures of these models are identical, except that since their outer wind temperatures are dependent on the cooling of the wind through metal resonance lines (Hillier 1988). At a radius of 10$`R_{}`$, equivalent to $`\tau _{\mathrm{Ross}}`$0.01, the wind temperature is $`T_e^{\mathrm{wind}}`$=23kK, 18.5kK and 16kK, for 0.04$`Z_{}`$, 0.2$`Z_{}`$ and 1.0$`Z_{}`$, respectively.
Fig. 4 shows synthetic UV and optical WN spectra for each case. In the UV, differences between the models are dominated by the strength of Fe v-vi features and the appearance of N iii $`\lambda \lambda `$1748–52 in the Solar metallicity model. Differences in blanketing play a minor role in the Lyman ionizing flux distribution of these models, although shortward of the O<sup>+</sup> edge (353Å) the higher blanketing of the 1.0$`Z_{}`$ model predicts a factor of three times fewer ionizing photons than the 0.04$`Z_{}`$ case.
From Fig. 4, the optical region reveals that solely low excitation lines, such as N iii $`\lambda `$4634–41 and He i $`\lambda `$4471, are enhanced at high metallicities. He ii $`\lambda `$4686 is essentially unaffected, as is N iv $`\lambda `$4058 due to its complex line formation mechanism, despite the factor of 25 decrease in nitrogen content. Differences for N iii $`\lambda `$4634–41 are principally due to abundance effects. Lines formed in the outer wind, such as He i, are sensitive to wind cooling, such that the emission equivalent width of He i $`\lambda `$10830 decreases from 59Å at 1.0$`Z_{}`$, to 35Å at 0.2$`Z_{}`$ and 26Å at 0.04$`Z_{}`$.
Therefore, although identical physical parameters are adopted in each WN model, earlier spectral types are obtained at lower metallicity. Following the scheme of Smith et al. (1996), whose spectral classification diagnostics were specifically chosen to avoid metallicity effects, the spectral type resulting from our sample of models ranges from WN6 at 1.0$`Z_{}`$ to WN3–4 at 0.04$`Z_{}`$ (the latter depending on the selection of diagnostic lines). This effect may contribute to the trend towards early WN types at low metallicities. Recall that 100% of SMC WN stars have spectral types of WN2–5, in contrast to 78% in the LMC and 46% in the Galaxy.
## 5 Line strengths of Wolf-Rayet stars
Unfortunately, definitive results are not possible from our comparison between the physical parameters of Sk 41 with the wide range of parameters observed in Galactic and LMC counterparts. Conti et al. (1989) have compared the emission equivalent widths of SMC WR stars with those of the Galaxy and LMC, which revealed that their emission lines are relatively weak. Massey et al. (1987) and Armandroff & Massey (1991) extended these comparisons to other Local Group galaxies. However, their conclusions are affected by binarity, or line-of-sight companions. Instead, absolute line luminosities are presented here, since they are unaffected by a binary companion, and also used to determine the WR stellar content of starburst galaxies. Comparisons with absolute visual magnitudes provide a superior indication of wind emission strengths. Conti & Massey (1989) provide a comparison between line fluxes and absolute visual magnitudes for LMC and Galactic WN stars, while Massey & Johnson (1998) present a similar comparison for LMC, SMC and M33 WN stars.
### 5.1 Sample of Wolf-Rayet stars
He ii $`\lambda `$4686 (for WN2–9) and C iv $`\lambda `$5801-12 (for WC4–6 and WO) line luminosities have been measured in a large sample of Galactic, LMC and SMC WR systems, restricting the former to stars of known distance, generally through cluster/association membership. These two lines were selected since they are amongst the strongest features in all spectral types, and are commonly used to assess WR populations in starburst regions (Schaerer & Vacca 1998).
The majority of our measurements were taken from the Torres-Dodgen & Massey (1988) atlas, except where superior resolution fluxed datasets are available to us, obtained with either AAT/RGO, MSO 74inch/coude or MSO 2.3m/DBS during Dec 1991–Dec 1997 (e.g. Crowther & Smith 1997).
Absolute visual magnitudes have been calculated as follows. For WN stars, interstellar reddenings were generally obtained from weighted averages of Schmutz & Vacca (1991), Morris et al. (1993), Hamann et al. (1993), Hamann & Koesterke (1998) and our own determinations based on comparing dereddened optical and UV (IUE) datasets with theoretical WN flux distributions. Interstellar extinction laws follow Seaton (1979) for the Galaxy, Howarth (1983) for the LMC, and Bouchet et al. (1985) for the SMC. WC interstellar reddenings were weighted averages of Smith et al. (1990ab), Morris et al. (1993), Koesterke & Hamann (1995), Kingsburgh et al. (1995), Gräfener et al. (1998), and again our own determinations. Individual measurements are available upon request to the author. Spectral types are taken from Smith et al. (1990b, 1996) and Crowther et al. (1998). Magellanic Cloud distances are taken from Westerlund (1997), namely 51.2kpc and 60.2kpc for the LMC and SMC, respectively.
Absolute magnitudes of WR stars in binaries have been corrected for the presence of their OB companion. Where possible, previously estimated light ratios are adopted (e.g. Smith et al. 1996). Otherwise, standard OB calibrations are used. In some cases, such as AzV 2a (WN3+O4), O-type calibrations (Conti et al. 1983; Crowther & Dessart 1998) exceed the systemic absolute magnitude, so either the spectral type of their companions is in error, or they are sub-luminous. For these systems, we generally adopt $`M_\mathrm{V}`$(O)=$``$4.0 mag, with the exception of MDV1, for which no correction was applied (Table 1).
### 5.2 Line luminosities
In Figs. 56 we present line luminosities of Galactic, LMC and SMC WN and WC stars. Although the scatter is large, we confirm the trend towards higher line luminosities for WN stars with increasing $`M_v`$, previously identified by Conti & Massey (1989) for the LMC, and find a similar correlation for WCE stars. Absolute magnitudes of single stars (open symbols) are more reliable than members of binaries (filled-in symbols).
Table 2 presents mean line luminosities for Galactic, LMC and SMC WN and WC stars, and the combined sample. Note that the formal standard deviations are poor indicators, given the small numbers involved and non-gaussian distribution. From the Table, the line luminosities of WN stars in the SMC are apparently a factor of 2–3 times lower than their LMC and Galactic counterparts. However, line luminosities of SMC WNE stars do not do not fall systematically below Galactic or LMC WNE stars with comparable (low) absolute magnitudes as shown in Fig. 5. Similarly, Sk 41 has a luminosity that is within $``$20% of mean Galactic and LMC values, if the remarkable WN5–6 stars in LMC and Galactic giant H ii regions are excluded (e.g. Crowther & Dessart 1998).
HD 5980 is omitted from the SMC sample shown in Table 2, since its line luminosity is from multiple components, highly variable and indeed of uncertain origin<sup>1</sup><sup>1</sup>1Moffat et al. (1998) attribute pre-outburst line emission to material formed in the shocked region between two early type components, rather than a single WNE star. The He ii $`\lambda `$4686 luminosity of HD 5980 was 5.59$`\times `$10<sup>36</sup> erg s<sup>-1</sup> in 1981, prior to outburst, when its spectral type was WN3–4. Observations obtained in Dec 1994, shortly after outburst, when the spectral type was WN8, revealed an even more remarkable He ii line luminosity of 2.26$`\times `$10<sup>37</sup> erg s<sup>-1</sup>, which is a factor of four greater than any other WN system from our sample, albeit with an equally impressive absolute visual magnitude (Table 1).
Amongst WC/WO stars, comparisons between the SMC and other galaxies are hindered by the absence of WC stars in the SMC. Amongst the WO stars with known distances, Sand 5 (WR142, WO2) in the Galaxy, Sand 2 (BAT99-123, WO3) in the LMC, and Sand 1 (AB8, WO3+O) in the SMC, it is Sand 1 which has the largest C iv $`\lambda `$5801 luminosity. However, attempts to draw conclusions are severely hindered by the very low numbers involved. To illustrate this, DR1 (WO3) in the SMC-like IC 1613 has a C iv line luminosity which is a factor of three times lower than Sand 1 (Kingsburgh & Barlow 1995).
Therefore, WO and WN stars in the SMC do not have systematically lower line luminosities than their higher metallicity counterparts. Therefore, mean values of our entire sample may be determined, and are indicated in Table 2. How do mean line luminosities compare with previous determinations? Smith et al. (1990a) obtained 3.3$`\times `$10<sup>36</sup> erg s<sup>-1</sup> from observations of 5 LMC WC4 stars, which is confirmed here, based on a much larger sample of 25 Galactic and LMC WCE stars. More recently, Schaerer & Vacca (1998) obtained (5.2$`\pm `$2.7)$`\times `$10<sup>35</sup> erg s<sup>-1</sup> from 26 WNE stars and (1.6$`\pm `$1.5)$`\times `$10<sup>36</sup> erg s<sup>-1</sup> from 19 WNL stars. From 59 WNE stars, a 20% higher mean results here, while 43 WNL stars indicate a 30% lower calibration for late WN stars.
Note that application of these mean luminosities will lead to an overestimate of the true WN population in starburst regions by a factor of $``$2–3 if luminous WN5–6h stars dominate the WR signature, as is the case in 30 Dor and NGC 3603. Conversely, if the properties of constituent WNE stars are more typical of SMC stars, where close binary evolution probably plays an important role, the actual population would be underestimated by a similar factor.
### 5.3 Line widths of Galactic, LMC and SMC WR stars
If WR stars in the SMC have slower winds than their Galactic and LMC counterparts due to lower metallicity, we expect that their line widths will also be systematically lower.
However, in contrast to line fluxes, FWHM are sensitive to observational resolution. For example, AB7 has FWHM(He ii $`\lambda `$ 4686)=32Å from low resolution spectroscopy of Torres-Dodgen & Massey (1988), in contrast to 26Å from medium resolution AAT/RGO data. This observational limitation also complicates the Smith et al. classification scheme, which assigns ‘b’ for broad lined stars if FWHM(He ii $`\lambda 4686)`$30Å (see also Conti 1999). Consequently, we have measured FWHM(He ii) for a WR sample obtained with a uniform medium resolution of $``$2Å). Where these are unavailable, corrections to measurements from Torres-Dodgen & Massey (1988) datasets are made, using stars in common to both datasets as calibrators.
From Table 2, we find that the mean FWHM(He ii $`\lambda `$4686) for WNE stars does decrease at lower metallicity, from 33Å in the Galaxy to 25Å in the SMC. However, since the SMC WR population is so low, taking into consideration the dependence of FWHM on individual spectral types, no unambiguous conclusions may be drawn. For example, LMC WCE stars reveal systematically higher FWHM (C iv $`\lambda `$5801–12) than Galactic stars since the LMC WC population is solely represented by WC4 stars, which are rare in our Galaxy. The situation is further complicated in our Galaxy since Schild et al. (1990) and Armandroff & Massey (1991) identified a correlation between FWHM(C iv $`\lambda `$5801–12) and galacto-centric distance for WCE stars in our Galaxy and M33, attributed to a metallicity gradient. Amongst WO stars, we find an apparent trend towards narrower lines at lower metallicity, again based on a very small sample.
## 6 Conclusions
Contrary to expectations, it appears that both the stellar properties and line luminosities of Wolf-Rayet stars in the SMC do not differ significantly from their counterparts in higher metallicity galaxies. Therefore, the reliability of studies of WR starburst galaxies at low metallicity based on template Galactic or LMC stars is supported here. However, individual spectral types may differ by up to a factor of $``$5 from mean WN, WC or WO line luminosity calibrations. Therefore, the question of what Wolf-Rayet flavours are produced in different star forming regions becomes relevant.
One added complication is that close binary evolution probably plays a major role in the formation of SMC WR stars, which may affect their physical properties. This is probably not the case for regions undergoing powerful bursts of massive star formation. It would be very useful to compare stellar properties and line luminosities of individual WR stars in low metallicity starbursts with the present calibrations, for which IC10 represents an ideal candidate (Massey & Armandroff 1995).
###### Acknowledgements.
Thanks to John Hillier for providing his current stellar atmospheric code, and to Werner Schmutz for permitting the use of new SOFI datasets in this work. I appreciate Daniela Calzetti providing a parameterization of the Bouchet et al. SMC law, and Peter Conti, Nolan Walborn and Orsola De Marco for a careful reading of this manuscript. PAC is funded by a University Research Fellowship of the Royal Society. The support of the staff from the Anglo-Australian, European Southern and Mount Stromlo Observatories is greatly appreciated. This work has made use of the SIMBAD database, operated at the CDS, Strasbourg, France. |
warning/0001/cond-mat0001272.html | ar5iv | text | # Theory of Coherent 𝑐-Axis Josephson Tunneling between Layered Superconductors
## I Introduction
One of the most interesting features of superconductivity is the Josephson effect. This occurs when flat surfaces of two superconductors are brought together, forming a uniform junction. A supercurrent flows without a voltage drop across the junction, provided that the properties of the superconductivity in the two superconductors are compatible. For conventional superconductors in which both the normal state and superconducting properties are isotropic, it doesn’t matter which particular crystal surfaces are employed to form the junction.
However, for anisotropic superconductors, the junction orientation can be very important. Even if the superconducting order parameter (OP) is isotropic, the intrinsic anisotropic normal state properties of a layered superconductor make the properties of Josephson junctions involving one or more layered superconductors different from those formed from two isotropic materials. For example, Josephson junctions between an isotropic, conventional superconductor and a layered superconductor can differ greatly, depending upon whether the junction is on the top or an edge of the layered superconductor. This is especially true if the layered superconductor has an anisotropic OP. In addition, the properties of Josephson junctions formed between two layered superconductors depend strongly upon the junction orientation, especially if the OPs are anisotropic.
In this paper, we consider the case of a Josephson junction formed between two layered superconductors stacked on top of each other along the $`c`$-axis. Our results can be easily modified to include the related problems in which one or both of the superconductors is bulk, rather than layered. By treating the two superconductors as layered, the surface states that form near the junction appear naturally in the calculation. These surface states affect the Josephson current results, even when the limit of tunneling between two isotropic superconductors is taken. With one exception, these surface states have previously been neglected. In addition, when the quasiparticle dispersions of the two superconductors are different, such as for different materials or identical tight-binding materials that are rotated with respect to one another, an impedance mismatch occurs, reducing the coherent critical current.
We consider the Josephson tunneling along the $`c`$-axis between two distinct layered superconductors, assuming all of the tunneling processes can be taken to be purely coherent, as pictured in Fig. 1. In Section II, we solve for the Green’s function in each layer, keeping the quasiparticle dispersions and OP symmetries fully general. The effects of surface states are explicitly included in the calculations, and the various tunneling strengths are included exactly. In Section III, we derive the temperature ($`T`$) dependence of the superconducting gaps under three specific assumptions about the quasiparticle band structures and three different OP symmetries in each layered superconductor. These quasiparticle dispersions are chosen to model the experimental cases of underdoped, optimally doped, and overdoped Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O$`_{8_+\delta }`$, (BSCCO), respectively, with respective Fermi surfaces that we denote as FS1, FS2, and FS3. However, the dispersions for overdoped samples can also be made to fit the three-dimensional free-quasiparticle dispersions of conventional bulk superconductors. The OP symmetries chosen are the isotropic $`s`$-wave and the tight-binding $`d_{x^2y^2}`$-wave and “extended-$`s`$”-wave OP functions, respectively. In Section IV, we solve for the gaps for the symmetric case of two identical superconductors. Then, in Section V, we calculate the Josephson tunneling current across the two layered superconductors, assuming that the coupling across the junction is not stronger than the intrinsic coupling within each layered superconductor. Our final results for the tunneling current can be used to calculate the $`c`$-axis Josephson critical current $`I_c`$, and can be generalized to the standard case of two bulk, conventional superconductors, for which the usually neglected surface states are explicitly included. Then, in Section VI, we present detailed $`I_c(\varphi _0)`$ results for coherent $`c`$-axis twist junctions between identical coherent layered superconductors. Finally, we discuss our results in Section VII.
## II Model and Procedure for Solving it
We assume that two layered superconductors are placed with the upper ($`U`$) one on top of the lower ($`L`$), and that the contact between them is sufficiently strong that quasiparticle tunneling between them occurs. For simplicity, we assume each superconductor consists of $`N1`$ layers separated an equal distance $`s`$ apart. We index the layers with $`n,m`$, where $`N+1n,mN`$. In the $`U`$ half space, $`1n,mN`$, and in the $`L`$ half space, $`N+1n,m0`$, as pictured in Fig. 1. Within each layer in the $`\eta =L,U`$ half space, the quasiparticles propagate with dispersion $`\xi _{0\eta }(𝐤)`$ and have the gap function $`\mathrm{\Delta }_\eta (𝐤)`$, where $`𝐤=(k_x,k_y)`$ is a two-dimensional wavevector. We assume $`\mathrm{\Delta }_\eta `$ is independent of layer index. This assumption will be checked in Section III, and is found to be usually valid. Between adjacent layers in each half space, the quasiparticles tunnel with matrix element $`J_\eta /2`$. At the junction between layers 0 and 1 where the $`L`$ and $`U`$ half spaces meet, the quasiparticles tunnel with matrix element $`J/2`$. Since we are not interested in spin-dependent effects, we assume the quasiparticles are spinless fermions, only taking account of the spin values in counting the number of quasiparticles. We set $`k_B=c=\mathrm{}=1`$.
We shall focus on the general procedure for evaluating $`\widehat{G}`$. The details are given in the Appendix. In order to calculate the Josephson current across the junction between the two half spaces, we first find the form of the finite temperature Greens’ functions matrix $`\widehat{G}`$. This matrix is the product of two matrices, one of rank $`2N`$, with elements indexed by the layers $`n,m`$, and the other the Nambu matrix of rank 2, with elements $`G`$, $`F`$, $`G^{}`$, and $`F^{}`$ in the usual cyclic order beginning with the upper left hand position. This Nambu matrix is represented by the Pauli matrices $`\tau _i`$ for $`i=1,2,3`$, plus the rank two identity matrix $`\tau _0`$. We let $`\omega `$ represent the Matsubara frequencies.
We first begin by constructing the Green’s function matrix $`\widehat{𝒢}`$ for a bulk layered superconductor. We then add a perturbation with the particular form that decouples the Green’s functions in each half space from each other. The resulting Green’s functions $`\widehat{g}^\eta `$ of two single half spaces have parameters appropriate for each half space. We then couple these two half-space Green’s functions together.
The Green’s function matrix $`\widehat{𝒢}^\eta `$ for a bulk superconductor of type $`\eta `$ satisfies
$$[i\omega \xi _{0\eta }\tau _3\mathrm{\Delta }_\eta \tau _1+\widehat{𝒥}^\eta \tau _3]\widehat{𝒢}^\eta =\widehat{1},$$
(1)
where the layer space matrix $`\widehat{𝒥}^\eta `$ has the elements
$$𝒥_{mn}^\eta =\frac{J_\eta }{2}\left(\delta _{m,n+1}+\delta _{m,n1}\right).$$
(2)
Terms not containing a Pauli matrix are implicitly proportional to $`\tau _0`$. In a bulk layered superconductor, it is then possible to Fourier transform this expression, as we did in the Appendix, with the result differing only from that of a bulk, three dimensional superconductor by the corrugated cylinder form of the quasiparticle dispersion, $`\xi _\eta (𝐤,k_z)=\xi _{0\eta }(𝐤)J_\eta \mathrm{cos}(k_zs)`$. However, for a half space, such Fourier transformation is not permissible, and so we must keep the layer indices $`n,m`$ explicitly.
To construct $`\widehat{g}^\eta `$, we first remove the tunneling matrix elements across the junction between the two superconductors by adding the perturbation $`\widehat{V}^\eta \tau _3`$, with elements
$$V_{mn}^\eta =\frac{J_\eta }{2}(\delta _{m0}\delta _{n1}+\delta _{m1}\delta _{n0}),$$
(3)
taking account of the restrictions $`n,m0`$ for $`\eta =L`$ and $`n,m1`$ for $`\eta =U`$. The perturbation $`\widehat{V}^\eta `$ “cuts” the bonds connecting the two identical half spaces which terminate at layers 0 and 1. We then solve for the elements $`g_{mn}^\eta `$ of $`\widehat{g}^\eta `$ satisfying
$$\widehat{g}^\eta =\widehat{𝒢}^\eta +\widehat{𝒢}^\eta \widehat{V}^\eta \tau _3\widehat{g}^\eta .$$
(4)
Solutions for each $`\widehat{g}^\eta `$ are given in the Appendix.
So far, we have found the expressions for two distinct half-space Green’s functions, which are electronically uncoupled from each other, since no quasiparticle propagation from one half space to the other has yet been introduced. We thus couple $`\widehat{g}^U`$ and $`\widehat{g}^L`$ together via the local perturbation $`\widehat{𝒥}`$ with matrix elements
$$𝒥_{mn}=\frac{J}{2}(\delta _{m0}\delta _{n1}+\delta _{m1}\delta _{n0}).$$
(5)
For tunneling strength comparisons, we then let
$$\gamma =\frac{J^2}{J_LJ_U}.$$
(6)
The exact solution to this problem of coupled half spaces then yields the full Green’s function matrix $`\widehat{G}`$, with matrix elements $`G_{mn}`$, which is given in the Appendix.
## III Gap Equation
In order to obtain the temperature dependence of the Josephson critical current, we need to solve a gap equation for the temperature dependence of the gap. For this we make a simple assumption of a BCS-like equation
$$\mathrm{\Delta }_{\eta n}(𝐤,T)=\frac{T}{2}\underset{\omega }{}\underset{𝐤^{}}{}\lambda _\eta (𝐤,𝐤^{})\mathrm{Tr}[(\tau _1+i\tau _2)G_{nn}(𝐤^{},\omega )].$$
(7)
This layer-dependent gap function is in contradiction to the layer-independent gap function used to calculate the Green’s functions. This must therefore be regarded as the first correction to the assumed constant gap, which determines the Green’s function on the right hand side of this equation. Ideally, this should be a small correction, if our initial assumption is justifiable.
In the Appendix, we found an expression for $`G_{00}\frac{J_L}{2}\tau _3`$. Using this result, letting
$$\mathrm{\Xi }_\eta =\frac{2}{\xi _{0\eta }+i\mathrm{\Omega }_\eta +[(\xi _{0\eta }+i\mathrm{\Omega }_\eta )^2J_\eta ^2]^{1/2}},$$
(8)
where $`\mathrm{\Omega }_\eta =\sqrt{\omega ^2+|\mathrm{\Delta }_\eta |^2}`$, and defining $`\mathrm{\Xi }_\eta ^{}=\mathrm{}(\mathrm{\Xi }_\eta )`$ and $`\mathrm{\Xi }_\eta ^{\prime \prime }=\mathrm{}(\mathrm{\Xi }_\eta )`$, the equation for the gap at the interface in the lower superconductor is
$$\mathrm{\Delta }_{L0}(𝐤,T)=T\underset{\omega }{}\underset{𝐤^{}}{}\frac{\lambda _L(𝐤,𝐤^{})}{D(𝐤^{},\omega )}f(𝐤^{},\omega ),$$
(9)
where
$$f=\frac{\mathrm{\Delta }_L\mathrm{\Xi }_L^{\prime \prime }}{\mathrm{\Omega }_L}+\frac{\mathrm{\Xi }_U^{\prime \prime }|\mathrm{\Xi }_L|^2J^2\mathrm{\Delta }_U}{4\mathrm{\Omega }_U},$$
(10)
$`D`$ $`=`$ $`{\displaystyle \frac{J^2}{2}}[\left({\displaystyle \frac{\omega ^2+\mathrm{\Delta }_L\mathrm{\Delta }_U\mathrm{cos}(\varphi _L\varphi _U)}{\mathrm{\Omega }_U\mathrm{\Omega }_L}}\right)\mathrm{\Xi }_U^{\prime \prime }\mathrm{\Xi }_L^{\prime \prime }`$ (13)
$`\mathrm{\Xi }_U^{}\mathrm{\Xi }_L^{}]+1+{\displaystyle \frac{J^4}{16}}|\mathrm{\Xi }_U|^2|\mathrm{\Xi }_L|^2,`$
and $`\varphi _L\varphi _U`$ is the phase difference of the OPs across the junction.
As a quick check of this result, we take $`L=U`$ and $`\gamma =1`$, and find that this reduces to
$`\mathrm{\Delta }_L(𝐤,T)`$ $`=`$ $`T{\displaystyle \underset{\omega }{}}{\displaystyle \underset{𝐤^{}}{}}\lambda _L(𝐤,𝐤^{}){\displaystyle \frac{\mathrm{\Delta }_L(𝐤^{},\omega )}{\mathrm{\Omega }_L(𝐤^{},\omega )}}\times `$ (16)
$`\times \mathrm{}\left[\left([\xi _{0L}(𝐤^{})+i\mathrm{\Omega }_L(𝐤^{},\omega )]^2J_L^2\right)^{1/2}\right],`$
in agreement with the result for a homogeneous coherent layered superconductor.
Note that in Eq. (9), the proximity effect couples the gap function at the interface in $`L`$ to the gap function in $`U`$. This disappears in the limit $`\gamma 1`$, wherein $`\mathrm{\Delta }_0(𝐤,T)`$ becomes the gap function at the surface of $`L`$:
$`\mathrm{\Delta }(𝐤,T)_{L,surface}`$ $`=`$ $`T{\displaystyle \underset{\omega }{}}{\displaystyle \underset{𝐤^{}}{}}\lambda _L(𝐤,𝐤^{})\times `$ (19)
$`\times {\displaystyle \frac{\mathrm{\Delta }_L(𝐤^{},\omega )\mathrm{\Xi }_L^{\prime \prime }(𝐤^{},\omega )}{\mathrm{\Omega }_L(𝐤^{},\omega )}},`$
in agreement with previous results.
The gap functions which go into the right hand side of (9) are the spatially constant (“zeroth order”) bulk gap functions, obtained from (16) for $`L`$, and an analogous equation for $`U`$. We use these to calculate $`\mathrm{\Delta }_\eta (𝐤,T)`$ and $`\mathrm{\Delta }(𝐤,T)_{\eta ,surface}`$ for $`\eta =L,U`$, and compare the magnitudes of these to the bulk results. Ideally, the difference in magnitudes between the bulk gaps and the interface gaps should be small. In the following section, we will investigate this for the special case of a symmetric junction.
## IV Gaps for the Symmetric Case
In this section, we assume that $`L=U`$ and solve for the bulk gap, and the gap at the interface with $`\gamma 1`$. To find the bulk gap, we fix $`T_c`$ at 9 meV, and solve Eq. (16) for the temperature dependent gap. We assume each superconductor has a single OP, which for simplicity, we limit to $`s`$, $`d_{x^2y^2}`$, or extended-$`s`$-wave symmetry. We index these OPs by $`\zeta =s,d,es`$. For each OP, we take the pairing interaction on the $`\eta =L,U`$ superconductor to have the form
$$\lambda _{\zeta \eta }(𝐤,𝐤^{})=\lambda _{\zeta \eta }\mathrm{\Psi }_\zeta (𝐤)\mathrm{\Psi }_\zeta (𝐤^{}),$$
(20)
where $`\mathrm{\Psi }_s(𝐤)=1`$, $`\mathrm{\Psi }_d(𝐤)=\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)`$, and $`\mathrm{\Psi }_{es}(𝐤)=\{[\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)]^2+ϵ^2\}^{1/2}`$, where $`ϵ1`$.
By fixing $`T_c`$, we determine the value of $`\lambda _{\zeta \eta }`$. These then determine the $`T`$ dependence of the s-wave or d-wave gaps, respectively, via Eq. (16), and the extended-s-wave gap with $`ϵ0`$ has an identical $`T`$ dependence to that of the d-wave gap. We take the in-plane dispersion to be
$`\xi _{\eta 0}(𝐤)`$ $`=`$ $`J_\eta J_{\eta ||}[\mathrm{cos}(k_xa)+\mathrm{cos}(k_ya)`$ (23)
$`\nu \mathrm{cos}(k_xa)\mathrm{cos}(k_ya)\mu ],`$
where the part proportional to $`J_{\eta ||}`$ is chosen to approximate the in-plane dispersion relation for BSCCO. The value of $`J_{\eta ||}`$ is thus taken to be 500 meV. We choose three sets of parameters $`\nu `$ and $`\mu `$, which determine the details of the quasiparticle dispersion, and the resulting shape of the two-dimensional Fermi surface, pictured in Fig. 2. For a heavily underdoped sample, we choose $`\mu =1.3`$, $`\nu =1.3`$, with Fermi surface denoted FS1 in Fig. 2. For the tight-binding dispersion appropriate for an optimally doped sample of BSCCO, we take $`\nu =1.3`$ and $`\mu =0.6`$. This dispersion has the Fermi surface denoted FS2 in Fig. 2. In addition, for a heavily overdoped sample, we choose $`\nu =0`$ and $`\mu =1.0`$, with the Fermi surface FS3 in Fig. 2. The only remaining free parameter is the value of $`J_\eta /2`$, the interlayer overlap integral. We therefore perform our calculations for three different values: $`J_\eta =25,50,100`$ meV. In each case, $`J_\eta =25`$ meV gives results which closely approximate the weak hopping limit $`J_\eta 0`$, as has been verified by explicitly checking our results with $`J_\eta =1`$ meV. However, for that small an interlayer hopping parameter, one needs to use a much finer grid for the Brillouin zone integration, in order to obtain sufficient accuracy.
For each OP symmetry $`\zeta `$, we calculate $`\mathrm{\Delta }_\zeta (T)`$ from the symmetric gap equation. We note that $`\mathrm{\Delta }_{es}(T)=\mathrm{\Delta }_d(T)`$ for $`T_{cd}=T_{ces}`$, which occurs for $`\lambda _{es}=\lambda _d`$. In the case $`\gamma =1`$, the combined junction of two layered superconductors is just the same as a single layered superconductor, as long as the two halves are not twisted with respect to each other. However, for $`\gamma <1`$, the central junction is different than the intrinsic ones in each layered superconductor. In this case, surface states can arise, and the gap $`\mathrm{\Delta }_\mu `$ can in principle depend upon the layer index. This would be particularly true in the case of two incompatible OP components, which has been considered in detail for a cylindrical Fermi surface previously.
For the Fermi surface FS2 and only one OP component, if $`\gamma 1`$ and $`J_\eta =0`$, then each layer would be completely isolated from each other, and the gap wouldn’t depend upon the layer index, reducing to the two-dimensional mean-field value of independent layers. Hence, the question of whether the gap depends upon the layer index can be more easily answered by checking whether the gap depends upon the intrinsic interlayer hopping strength $`J_\eta `$ and on the impedance mismatch parameter $`\gamma `$. In Fig. 3a, we plotted $`\mathrm{\Delta }_s(0)`$ and $`\mathrm{\Delta }_d(0)`$ for $`J_\eta =25,50,100`$ meV, as functions of $`\gamma `$ ($`0\gamma 1`$), for the optimally doped Fermi surface FS2, evaluated with $`\nu =1.3`$ and $`\mu =0.6`$, pictured in Fig. 2. Although there is a weak $`\gamma `$ dependence of $`\mathrm{\Delta }_d(0)`$ for $`J_\eta =100`$ meV, all other cases have essentially no significant $`\gamma `$ dependence, so that the gap values are essentially the same as for independent layers. The slightly greater $`\gamma `$ dependence of $`\mathrm{\Delta }_d(0)`$ for $`J_\eta =100`$ meV apparently arises because the Fermi energy $`E_F`$ is not much further than $`J_\eta `$ from the top of the quasiparticle band. In Ref. (3), it was also found that for a free particle dispersion within the layers that the gap at the surface $`(\gamma =0`$) was the same as that in the bulk ($`\gamma =1`$). Thus, our results for a finite in-plane bandwidth closely approximate that infinite bandwidth limit.
We also studied $`\mathrm{\Delta }_s(0)`$ as a function of $`J_\eta `$ at $`\gamma 1`$ for Fermi surfaces FS1 and FS2, and plotted the results in Fig. 3b. Again, we found very little $`J_\eta `$ dependence to the $`s`$-wave gap magnitudes for $`\gamma 1`$, so that it is generally safe to take the gap to be the bulk value calculated far from the central junction location. We also found that $`\mathrm{\Delta }_\zeta (T)/\mathrm{\Delta }_\zeta (0)`$ for $`\zeta =s,d`$ with Fermi surface FS2 were rather independent of $`J_\eta `$ for $`J_\eta =1,25,50,100`$ meV, each of the curves differing only slightly from the ordinary BCS curves for $`\mathrm{\Delta }(T)`$.
## V Josephson Tunneling Current
The tunneling current across the junction is given by
$$I=ieJ\underset{𝐤}{}c_0^{}(𝐤)c_1(𝐤)c_1^{}(𝐤)c_0(𝐤)$$
(24)
where the $`c_i(𝐤),c_i^{}(𝐤)`$ are creation and annihilation operators for electrons with wavevector $`𝐤`$ in the $`i^{th}`$ layer. The angular brackets indicate a thermodynamic equilibrium average. In terms of the full space Green’s functions, this is
$$I=ieJT\underset{𝐤}{}\underset{\omega }{}\mathrm{Tr}[\frac{1}{2}(\tau _0+\tau _3)(G_{10}G_{01})],$$
(25)
where we have suppressed the $`(𝐤,\omega )`$ dependence of the Green’s functions.
In the Appendix, the matrices $`G_{01}`$ and $`G_{10}`$ are given, and the trace evaluated. We find
$$I=eT\underset{\omega }{}\underset{𝐤}{}\frac{J^2\mathrm{\Delta }_L\mathrm{\Delta }_U\mathrm{\Xi }_L^{\prime \prime }\mathrm{\Xi }_U^{\prime \prime }}{\mathrm{\Omega }_L\mathrm{\Omega }_UD}\mathrm{sin}(\varphi _L\varphi _U),$$
(26)
where $`\mathrm{\Xi }_\eta `$ is given by Eq. (8) and $`D`$ is given by Eq. (13).
The equations derived above relate the supercurrent through the interface between two layered superconductors to the phase difference across this interface. We can find the critical current $`I_c^J`$ of this interface by varying the phase difference until the maximum current is obtained. We can vary the value of $`\gamma `$, which we term “the impedance match parameter”, to obtain $`I_c(\gamma )`$. We can also vary $`T`$ to obtain $`I_c(T)`$. Finally, we can investigate all of the above for $`s`$-, and $`d`$-, and extended-$`s`$-wave gaps, and mixtures thereof.
To order $`J^2`$, we can set $`D1`$, and $`I_c^J`$ obtained from Eq. (26) reduces precisely to our previous result, Eq. (8) of Ref. (4), derived using the tunneling Hamiltonian, provided only that one neglects the incoherent tunneling, and replaces $`J`$ with $`2𝒯_0`$ in the coherent tunneling part. However, this result differs greatly from that found by Tanaka and Kashiwaya (TK). TK did not correctly derive the coherent $`c`$-axis tunneling between two layered superconductors, but instead attempted to approximate the tight-binding quasiparticle dispersion in the $`c`$-axis direction by treating the corrugated Fermi surface as a narrow belt around a spherical Fermi surface. This procedure leads to the uncontrolled approximation of dividing zero by zero. Thus, Eq. (100) of TK, obtained in the weak tunneling limit of their calculation, is not correct for coherent tunneling. Instead, it happens (from compensating errors) to be correct for purely incoherent tunneling, as in the model of Ambegaokar-Baratoff (AB). However, the subsequent Eq. (101) of TK is still not quite correct for any type of tunneling, and Eqs. (102)-(104) of TK are completely wrong for both incoherent and coherent tunneling. A modified version of Eq. (101), correct for incoherent tunneling between a conventional and an unconventional superconductor, was used appropriately in fits to $`c`$-axis tunneling between Pb and BSCCO.
In this notation, we can also investigate the case of the coherent tunneling matrix elements depending upon the in-plane wavevectors, $`J_\eta (𝐤_\eta ,𝐤_\eta )=J_\eta \phi ^2(𝐤_\eta )`$, and $`J(𝐤_L,𝐤_U)=J\phi (𝐤_L)\phi (𝐤_U)`$, where $`\phi (𝐤)=|\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)|`$, as suggested from band structure calculations of Liechtenstein et al. This form, along with the wavevector independent model, is useful for studying the coherent critical currents across a $`c`$-axis twist junction. We note that $`\gamma (𝐤_L,𝐤_U)=\gamma `$ is independent of the $`𝐤_\eta `$.
As long as $`\gamma 1`$, the overall critical current $`I_c`$ will be given by the above $`I_c^J`$. For both strong $`\gamma =1`$ and weak $`\gamma 1`$ coupling with Fermi surface FS2 and $`J_\eta =1,25,50,1000`$ meV, we calculated $`I_c(T)`$, normalized to its $`T=0`$ value, for each of the OP symmetries considered. For $`\gamma 1`$, the $`s`$-wave $`I_c(T)/I_c(0)`$ curves are almost indistinguishable from the standard AB curve, as for Fig. 2 of Ref. (4). The other results are plotted in Fig. 4. Note that the $`d`$-wave and extended-$`s`$-wave curves are identical, but they differ substantially from the $`s`$-wave results. However, as shown in Fig. 4a, the $`d`$-wave (and extended-$`s`$-wave) $`I_c(T)/I_c(0)`$ curves are strongly dependent upon $`J_\eta `$, in a non-monotonic fashion, unlike Fig. (3) of Ref. (4). Those curves in Ref. (4) were evaluated using the free-particle quasiparticle dispersion within the layers. In addition, in Figs. 4b and 4c, we plotted $`I_c(T)/I_c(0)`$ for the strong coupling case $`\gamma =1`$ for $`\zeta =s,d`$. respectively. In this case, the non-monotonicity of the $`d`$-wave (and extended-$`s`$-wave) case is similar to that in the $`\gamma 1`$ limit, but the $`J_\eta `$ variation of $`I_c(T)/I_c(0)`$ is stronger in Fig. 4c for $`\gamma =1`$ than for $`\gamma 1`$ in Fig. 4a. Similarly, a much stronger $`J_\eta `$ variation of $`I_c(T)/I_c(0)`$ is seen in Fig. 4b for $`\gamma =1`$ than for the (not pictured) BCS-like results obtained with $`\gamma 1`$.
We remark that Eq. (26) is a fully general expression for the $`c`$-axis tunneling between two layered superconductors, assuming that only one OP component exists in each superconductor, and that all tunneling processes are completely coherent. That is, the $`L`$ and $`R`$ superconductors can have different quasiparticle dispersions (both parallel and normal to the junction), different OP symmetries, and different $`T_c`$ values. Thus, Eq. (26) also describes the tunneling between a three-dimensional, conventional superconductor placed on the top of a layered superconductor with an unknown OP symmetry. It also applies to the case of coherent tunneling between two three-dimensional superconductors. When one of the superconductors is a three-dimensional material, one simply modifies the quasiparticle dispersion $`\xi _\eta (𝐤,k_z)=\xi _{0\eta }(𝐤)J_\eta \mathrm{cos}(k_zs)`$ far from the junction, by setting $`J_{\eta ||}`$ sufficiently large, and $`J_\eta `$ comparable to $`J_{\eta ||}`$. Except for the case of real-space pairing with a component normal to the junction, the fact that the pairing was assumed to take place only within the layers is not important, since the layer index is dropped, so that the pairing strength is constant throughout each superconductor. Thus, a superconductor with an isotropic, three-dimensional quasiparticle dispersion would have $`\xi _\eta (𝐤,k_z)=(𝐤^2+k_z^2)/(2m)`$ far from the junction, where $`m`$ is the quasiparticle effective mass. This form is obtained from Eq. (2) by setting $`J_{\eta ||}=(ma^2)^1`$ and $`J_\eta =(ms^2)^1`$, and requiring $`J_\eta ,J_{\eta ||}>>ϵ_F`$. An anisotropic three-dimensional superconductor might have either $`J_\eta `$ or $`J_{\eta ||}`$ different from these values.
We remark additionally that for the two-dimensional tight-binding Fermi surfaces, one restricts the wavevector integration to the first Brillouin zone (BZ), which for a tetragonal material has $`|k_x|,|k_y|\pi /a`$. In the three-dimensional limit, one can use the free-particle form for the quasiparticle dispersion afr from the junction, and remove the BZ limits on the components $`k_x,k_y`$ of the wavevectors parallel to the junction. Note that one still does not integrate over $`k_z`$, the wavevector normal to the junction, which is not a good quantum number in the presence of the junction. Thus, even in the limit of Josephson tunneling between two conventional bulk superconductors, surface states are present, and affect the tunneling. Except for the tunneling Hamiltonian limit $`\gamma <<1`$, for which we had previously included the effects of these surfaces states, our present calculation is thus the first one to correctly include the surface states in coherent Josephson tunneling between two superconductors, whether conventional bulk, or layered.
It is important to note that when the quasiparticle dispersions in the two superconductors are not identical, there is an intrinsic impedance mismatch between the two materials. Since in coherent tunneling, the wavevector parallel to the junction is preserved exactly, if for some wavevectors the Fermi surfaces in the two half spaces are not identical, it is not possible for the tunneling at those places in the BZ to be elastic. That is, one cannot preserve both the momenta and the energy. The result of this impedance mismatch is that the amplitude of the coherent tunneling, and hence the critical current, is reduced from what it would be if this impedance mismatch did not occur. This impedance mismatch should actually occur in coherent tunneling between all inequivalent superconductors. For example, in tunneling between Nb and Pb, the Fermi surfaces, which are both three-dimensional in nature, are not identical, and some amount of coherent tunneling would be suppressed by this effect. The amount of the suppression ought to depend strongly upon the particular crystal surfaces studied at the junction location.
## VI $`c`$-axis twist junctions
We now consider the case of purely coherent $`c`$-axis tunneling between identical layered superconductors twisted an angle $`\varphi _0`$ about the $`c`$-axis with respect to each other. This is a special case of coherent tunneling between two different layered superconductors. However, the OP symmetry must be the same on each superconductor, with the only caveat that the OP has to arise from the local pairing interaction, Eq. (20). The only difference is that the crystal orientation is rotated by $`\pm \varphi _0/2`$ in the two half spaces, and this leads to similarly rotated quasiparticle wavevectors. Thus, in the two superconducting half spaces, we take $`𝐤_\eta =(k_{x\eta },k_{y\eta })`$, which are rotated by $`\pm \varphi _0/2`$ for $`\eta =L,U`$, respectively. For any OP symmetry, we thus have $`\xi _{0\eta }(𝐤)=\xi _0(𝐤_\eta )`$ and $`\mathrm{\Delta }_\eta (𝐤)=\mathrm{\Delta }(𝐤_\eta )`$. Although for an isotropic $`s`$-wave pairing interaction, the twist orientation doesn’t matter as far as the OP $`\mathrm{\Delta }`$ is concerned, a non-vanishing $`\varphi _0`$ still causes the quasiparticle dispersions to be different in the two superconductors. This leads to a strong impedance mismatch for all $`\varphi _0`$ values not too close to $`0^{},90^{}`$.
In Figs. 5-8, we have presented our results for $`I_c(\varphi _0)/I_c(0)`$ at $`T/T_c=0.5`$ for different values of the material parameters. In Fig. 5, we presented our results for the optimally doped tight-binding quasiparticle dispersion, Eq. (23), in which $`\mu =0.6`$, $`\nu =1.3`$, and $`J_{||}=500`$ meV, with Fermi surface FS2. We include curves for each of the three OP symmetries, and for $`J_\eta =25,50,100`$ meV. For the extended-$`s`$-wave OP, we set $`ϵ=0`$, so that the wavevector dependence of the OP has the $`|\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)|`$ form. Results for the extended-$`s`$-wave OP with $`ϵ>0`$ are intermediate to the extended-$`s`$ and $`s`$-wave results shown.
In each case, there are both direct and Umklapp tunneling processes. In the direct processes, a quasiparticle undergoing tunneling across the twist junction has wavevectors $`𝐤_L`$ and $`𝐤_U`$ both within the first BZ. However, for the case in which either the initial or the final wavevectors is outside of the respective first BZ, Umklapp processes can occur, due to the periodic nature of the quasiparticle dispersions assumed. To investigate whether these Umklapp processes are important, we treat the two limiting cases, either (1) that they can be completely neglected, or (2) that they are equal in weight to the direct processes.
In Fig. 5a, we set $`\gamma 1`$, the tunneling Hamiltonian limit. In this case, we included the Umklapp processes with the same weighting as for direct tunneling processes within the first BZ on each side of the twist junction. Preliminary versions of Fig. 5a with slightly different parameters were presented earlier. For comparison, in Fig. 5b, we used the same parameters as in Fig. 5a, but the Umklapp processes were completely excluded. We note that for this quasiparticle dispersion, not very much of the intersection of the Fermi surfaces is excluded by neglecting the Umklapp processes for a twisted junction. In addition, in Fig. 5c, we presented our results for $`\gamma =1`$, including the Umklapp processes. We note that comparing these results with those of Fig. 5b, the main differences occur for the larger $`J_\eta `$ values, especially $`J_\eta =100`$ meV. Otherwise, for small $`J_\eta `$, there is very little difference between them.
In addition, we note that the main differences between the $`d_{x^2y^2}`$-wave and extended-$`s`$-wave $`I_c(\varphi _0)`$ results appear for $`\varphi _0`$ close to 45. As $`\varphi _045^{}`$, $`I_c(\varphi _0)/I_c(0)0`$ for the $`d`$-wave OP, whereas for the extended-$`s`$-wave OP, $`I_c(\varphi _0)/I_c(0)`$ remains finite and flattens out, becoming only weakly dependent on $`\varphi _0`$. Note that for Fermi surface FS2, even the isotropic $`s`$-wave OP leads to an anisotropic $`I_c(\varphi _0)/I_c(0)`$, for each of the $`J_\eta `$ values shown, reflecting the impedance mismatch across the twist junction. The greatest $`\varphi _0`$ variation occurs for the smallest $`J_\eta `$ values, as the quasiparticle dispersions are the most two-dimensional, with the greatest impedance mismatch occurring for $`\varphi _0`$ far from $`0^{},90^{}`$.
In Fig. 6, we again plotted $`I_c(\varphi _0)/I_c(0)`$ at $`T/T_c=0.5`$ fro $`c`$-axis twist junctions between optimally doped sampels with Fermi surface FS2 in the tunneling Hamiltonian limit $`\gamma 1`$, including the Umklapp processes. But, we now used the wavevector-dependent coherent interlayer tunneling similar to that suggested by Liechtenstein et al. We note that this wavevector-dependent interlayer tunneling gives rise to a stronger $`\varphi _0`$ dependence of $`I_c`$ than does the constant, wavevector-independent tunneling model. In addition, for this model, the extended-$`s`$ and $`d`$-wave OPs give rise to nearly identical $`I_c(\varphi _0)/I_c(0)`$ curves, except for $`\varphi _0`$ near to $`45^{}`$, of course, where $`I_c(\varphi _0)0`$ for the $`d`$-wave case, but not for the other two OPs.
In Fig. 7, we plotted $`I_c(\varphi _0)/I_c(0)`$ at $`T/T_c=0.5`$ in the tunneling Hamiltonian limit ($`\gamma 1`$) for $`c`$-axis twist junctions between heavily underdoped samples with Fermi surface FS1, including Umklapp processes, and kept $`J_\eta `$ fixed at 100 meV. Although curves are shown for each of the three OP symmetries studied, for this quasiparticle dispersion, it is difficult to distinguish them, as the $`I_c(\varphi _0)/I_c(0)`$ values all either vanish (for the $`d`$-wave OP) or become very small for $`\varphi _045^{}`$.
Finally, in Fig. 8, we presented plots of $`I_c(\varphi _0)/I_c(0)`$ at $`T/T_c=0.5`$ in the tunneling Hamiltonian limit ($`\gamma 1`$) for $`c`$-axis twist junctions between heavily overdoped samples with Fermi surface FS3. Results for $`J_\eta =25,50,100`$ meV are shown. In this case, there is very little impedance mismatch for the isotropic $`s`$-wave OP, since the rotation induced by the twist junction nearly maps the Fermi surface onto itself. One of these curves (the $`s`$-wave OP curve with $`J_\eta =100`$ meV) is actually consistent with the data of Li et al., although the quasiparticle dispersion assumed is very metallic and three-dimensional, completely inconsistent with that expected for BSCCO. The $`s`$-wave curve with $`J_\eta =50`$ meV is only marginally consistent with the data at best, and the $`s`$-wave curve with smaller $`J_\eta `$ values can be ruled out. Similarly, the extended-$`s`$ and especially the $`d`$-wave curves are all inconsistent with the experimental data.
## VII Conclusions
We calculated the Josephson current for $`c`$-axis coherent tunneling between two layered superconductors exactly. We assumed the intralayer quasiparticle dispersion of the tight-binding form, Eq. (23). Since the parameters in this model are arbitrary, our results also apply to the cases in which one or both of the superconductors is a conventional, bulk material. In these cases, one simply lets one or both of the $`J_\eta J_{\eta ||}\mathrm{}`$. For $`\gamma 1`$, the tunneling properties across the junction are different than the intrinsic tunneling between adjacent layers far from the junction, and the role of surface states on the sides of the junction becomes important.
Such surface states were first investigated for superconductors in the case of a vacuum interface. Then, the role of surface states upon the coherent Josephson tunneling between layered superconductors was investigated within the tunneling Hamiltonian limit. Our present results comprise the first calculation of the strong coherent Josephson tunneling between two superconductors which properly takes account of the surfaces states. The earlier results of Tanaka and Kashiwaya, which were claimed to give the correct results for this problem, are completely incorrect. In those calculations, the quasiparticle dispersions were not of the proper layered, tight-binding form, and the approximations used were not correct in any limit.
Our results also apply to the case of tunneling between two superconductors in other orientations. For instance, in the case of an isotropic bulk superconductor, the surface states present along the sides of a junction normal to the $`c`$-axis are precisely the same as that along the sides of a junction in another orientation. Thus, the calculations done by others, including Tanaka and Kashiwaya, for the Josephson tunneling within the $`ab`$-plane of layered superconductors, do not correctly take these surface states into account. In addition, since they always assume a circular Fermi surface cross-section (similar to FS3 in Fig. 2), all effects of impedance mismatching present in our investigation of the $`c`$-axis twist junctions are neglected. As shown in Section VI, such effects can be very strong, especially when the OP is highly anisotropic.
Although we have not introduced the effects of incoherent $`c`$-axis tunneling, we found that coherent tunneling between two superconductors that are not precisely the same gives rise to strong impedance mismatch effects. For coherent tunneling between identical layered superconductors twisted about the $`c`$-axis with respect to each other, the impedance mismatch effects are strong. Even when the OP is isotropic, one expects a strong twist angle $`\varphi _0`$ dependence of the Josephson critical current. Although Umklapp processes are present in tunneling between superconductors twisted about the $`c`$-axis, they make only a very small correction to the actual critical current.
We calculated the Josephson critical current across $`c`$-axis twist junctions, for OPs of the isotropic $`s`$-wave, the tight-binding $`d_{x^2y^2}`$-wave, $`\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)`$, and the tight-binding ‘extended-$`s`$-wave’, $`\{[\mathrm{cos}(k_xa)\mathrm{cos}(k_ya)]^2+ϵ^2\}^{1/2}`$, forms. We studied the cases of relative weak tunneling ($`\gamma 1`$) and strong tunneling ($`\gamma =1`$), for various $`c`$-axis dispersion bandwidths. Except for the simple case of an isotropic OP on a circular Fermi surface (akin to FS3), we conclude that coherent tunneling cannot possibly explain the data of Li et al.. Moreover, since the $`d_{x^2y^2}`$-wave OP cannot possibly fit the data of Li et al. under any circumstances, we conclude strongly that the OP is not $`d`$-wave near to $`T_c`$. Since there is widespread agreement that the Fermi surface of BSCCO is unlikely to have a circular cross-section, our calculations thus support the contention of Li et al. that the $`c`$-axis tunneling intrinsic in BSCCO must be largely incoherent. A brief discussion of the role of incoherent tunneling has been presented, supporting this argument. A more thorough treatment of the role of incoherent tunneling will be discussed in detail elsewhere.
## VIII acknowledgments
The authors would like to thank A. Bille, Qiang Li, and K. Scharnberg for useful discussions. This work was supported by the USDOE-BES through Contract No. W-31-109-ENG-38.
## IX Appendix: Exact Solutions for Green’s Functions in the Layers
To obtain the elements $`𝒢_{mn}`$ of the bulk tight-binding Green’s function matrix $`\widehat{𝒢}`$, we can first Fourier series transform in the layer index, letting
$$𝒢_{mn}^\eta (𝐤,\omega )=_{\pi /s}^{\pi /s}\frac{sdk_z}{2\pi }\mathrm{exp}[ik_zs(mn)]𝒢^\eta (𝐤,k_z,\omega ).$$
(27)
Eq. (1) can then be solved algebraically to give
$$𝒢^\eta (𝐤,k_z,\omega )=\frac{i\omega +\mathrm{\Delta }_\eta \tau _1+\xi _\eta \tau _3}{\omega ^2+|\mathrm{\Delta }_\eta |^2+\xi _\eta ^2},$$
(28)
where the quasiparticle dispersion
$$\xi _\eta (𝐤,k_z)=\xi _{0\eta }(𝐤)J_\eta \mathrm{cos}k_zs.$$
(29)
For notational simplicity in the following, we shall omit the $`𝐤`$ and $`\omega `$ dependencies of the various functions.
Performing the integral in Eq. (27) and multiplying by $`(J_\eta /2)\tau _3`$ on the right, we write
$`𝒢_{mn}^\eta {\displaystyle \frac{J_\eta }{2}}\tau _3`$ $`=`$ $`\beta _{mn}^\eta +(ϵ_\eta \tau _3\delta _\eta \tau _2)\alpha _{mn}^\eta `$ (30)
$``$ $`\beta _{mn}^\eta +\stackrel{}{\alpha }_{mn}^\eta \stackrel{}{\tau },`$ (32)
where
$$\alpha _{mn}^\eta =(\varphi _{mn}^{\eta +}\varphi _{mn}^\eta )/2,$$
(33)
$$\beta _{mn}^\eta =(\varphi _{mn}^{\eta +}+\varphi _{mn}^\eta )/2,$$
(34)
and
$$\varphi _{mn}^{\eta \pm }=\frac{\pm i\mathrm{exp}(\pm i\theta _{\eta \pm }|nm|)}{2\mathrm{sin}(\theta _{\eta \pm })},$$
(35)
with
$$\mathrm{cos}(\theta _{\eta \pm })=\frac{\xi _{0\eta }\pm i\mathrm{\Omega }_\eta }{J_\eta }.$$
(36)
The parameter $`\theta _{\eta +}`$ is defined to have a positive imaginary part, and $`\theta _\eta `$ is defined to have negative imaginary part. For convenience, we have defined $`\mathrm{\Omega }_\eta =\sqrt{\omega ^2+|\mathrm{\Delta }_\eta |^2}`$, $`ϵ_\eta =\omega /\mathrm{\Omega }_\eta `$, and $`\delta _\eta =\mathrm{\Delta }_\eta /\mathrm{\Omega }_\eta `$. For comparison with previous notation, we then write
$$\mathrm{exp}(i\theta _{\eta +})=\frac{J_\eta }{2}\mathrm{\Xi }_\eta ,$$
(37)
where $`\mathrm{\Xi }_\eta `$ is given by Eq. (8).
We then find the half-space Green’s functions, using the procedure of Eqs. (3) and (4). For the $`L`$ half space defined by $`m,n0`$, we have
$$g_{mn}^L=𝒢_{mn}^L𝒢_{m1}^L\frac{J_L}{2}\tau _3g_{0n}^L,$$
(38)
which is easily solved for $`g_{0n}^L`$ by setting $`m=0`$, yielding
$$g_{mn}^L=𝒢_{mn}^L(𝒢_{m1}^L\frac{J_L}{2}\tau _3)[1+𝒢_{01}^L\frac{J_L}{2}\tau _3]^1𝒢_{0n}^L.$$
(39)
Similarly, for the $`U`$ half space defined by $`m,n1`$, we find
$$g_{mn}^U=𝒢_{mn}^U(𝒢_{m0}^U\frac{J_U}{2}\tau _3)[1+𝒢_{10}^U\frac{J_U}{2}\tau _3]^1𝒢_{1n}^U.$$
(40)
After some straightforward algebra, one finds for each half space
$`g_{mn}^\eta {\displaystyle \frac{J_\eta }{2}}\tau _3`$ $``$ $`\overline{g}_{mn}^\eta =b_{mn}^\eta +(ϵ_\eta \tau _3\delta _\eta \tau _2)a_{mn}^\eta `$ (43)
$`b_{mn}^\eta +\stackrel{}{a}_{mn}^\eta \stackrel{}{\tau },`$
where
$$a_{mn}^\eta =\frac{1}{2}(\chi _{mn}^{\eta +}\chi _{mn}^\eta )$$
(44)
and
$$b_{mn}^\eta =\frac{1}{2}(\chi _{mn}^{\eta +}+\chi _{mn}^\eta )$$
(45)
for $`\eta =L,U`$. For $`nm0`$, we have
$$\chi _{mn}^{L\pm }=\frac{\mathrm{exp}[i\theta _{L\pm }(n1)]\mathrm{sin}[\theta _{L\pm }(m1)]}{\mathrm{sin}(\theta _{L\pm })}.$$
(46)
For $`1mn`$,
$$\chi _{mn}^{U\pm }=\frac{\mathrm{exp}(\pm i\theta _{U\pm }n)\mathrm{sin}(\theta _{U\pm }m)}{\mathrm{sin}(\theta _{U\pm })}.$$
(47)
To allow for different OP phases in the two sides of the junction, we introduce a phase factor multiplying the order parameter in the left hand side, letting
$$\delta _L=|\delta _L|\mathrm{exp}[i(\varphi _L\varphi _U)].$$
(48)
The function $`\delta _U`$ thus does not include such a phase factor.
So far, we have found expressions for the two distinct half space Green’s functions, which are electronically uncoupled from each other, since no quasiparticle propagation from one half space to the other has yet been introduced. We thus couple them together via the local perturbation $`\widehat{𝒥}`$ with matrix elements given by Eq. (5).
The exact solution to this problem of coupled half spaces then yields the full Green’s function matrix $`\widehat{G}`$, with matrix elements $`G_{mn}`$. In the $`L`$ half space, $`n,m0`$, the $`G_{mn}`$ satisfy
$$G_{mn}\frac{J_L}{2}\tau _3=\overline{g}_{mn}^L+\gamma \overline{g}_{m0}^L[(\overline{g}_{11}^U)^1\gamma \overline{g}_{00}^L]^1\overline{g}_{0n}^L.$$
(49)
Now the gap equation, Eq. (7) in Section III can be found from $`G_{nn}`$. On the $`L`$ side of the junction, we then need $`G_{00}`$. From Eq. (49) we find that the exact Green’s function at the interface and in the lower superconductor is
$$G_{00}\frac{J_L}{2}\tau _3=\frac{\overline{g}_{00}^L\gamma |\mathrm{exp}(i\theta _{U+}+i\theta _{L+})|^2(\overline{g}_{11}^U)^1}{D},$$
(50)
where
$`D`$ $`=`$ $`1+\gamma ^2|\mathrm{exp}(i\theta _{U+}+i\theta _{L+})|^2`$ (55)
$`2\gamma \{[ϵ_Lϵ_U+\delta _U\delta _L\mathrm{cos}(\varphi _L\varphi _U)]A_UA_L`$
$`+B_UB_L\},`$
$$A_\eta =[\mathrm{exp}(i\theta _{\eta +})\mathrm{exp}(i\theta _\eta )]/2$$
(56)
and
$$B_\eta =[\mathrm{exp}(i\theta _{\eta +})+\mathrm{exp}(i\theta _\eta )]/2.$$
(57)
Then, the trace in Eq. (7) can be evaluated by first multiplying Eq. (50) by $`2\tau _3/J_L`$ on the right side, Leading to
$`\mathrm{Tr}[(\tau _1+i\tau _2)G_{00}]`$ $`=`$ $`{\displaystyle \frac{2i}{J_LD}}[A_L\delta _L`$ (60)
$`+\gamma |\mathrm{exp}(i\theta _{L+})|^2A_U\delta _U].`$
This then leads to Eq. (9) in the text.
In order to calculate the Josephson tunneling current, we need to find $`G_{01}`$ and $`G_{10}`$ explicitly. Solving for these functions yields the exact results
$$G_{01}\frac{J}{2}\tau _3=\gamma \overline{g}_{00}^L[(\overline{g}_{11}^U)^1\gamma \overline{g}_{00}^L]^1$$
(61)
and
$$G_{10}\frac{J}{2}\tau _3=\gamma [(\overline{g}_{11}^U)^1\gamma \overline{g}_{00}^L]^1\overline{g}_{00}^L.$$
(62)
Combining, and evaluating the trace, we find
$$I=4e\gamma T\underset{𝐤}{}\underset{\omega }{}\frac{|\delta _L||\delta _U|A_UA_L}{D}\mathrm{sin}(\varphi _U\varphi _L).$$
(63)
From this expression, it is elementary to obtain Eq. (26) of the text.
In the tunneling Hamiltonian limit, $`\gamma <<1`$, this reduces to
$`I`$ $`=`$ $`4e\gamma T{\displaystyle \underset{𝐤}{}}{\displaystyle \underset{\omega }{}}|\delta _L||\delta _U|\mathrm{}[\mathrm{exp}(i\theta _{U+})]\times `$ (66)
$`\times \mathrm{}[\mathrm{exp}(i\theta _{L+})]\mathrm{sin}(\varphi _U\varphi _L).`$
This result agrees with the tunneling Hamiltonian result for the coherent tunneling limit, as expected. |
warning/0001/astro-ph0001277.html | ar5iv | text | # High Resolution Rotation Curves of Low Surface Brightness Galaxies
## 1. Introduction
The rotation curves of high surface brightness (HSB) spiral galaxies rise fairly steeply to reach an extended, approximately flat part, well within the optical disk (Bosma 1978, 1981a,b; Rubin, Ford, & Thonnard 1978, 1980). The discovery that the rotation curves of these galaxies are more or less flat out to one or two Holmberg radii has been one of the key pieces of evidence for the existence of dark matter outside the optical disk (see also van Albada et al. 1985). Within the optical disk, the observed rotation curves can in most cases be explained by the stellar components alone (Kalnajs 1983; Kent 1986).
The rotation curves of so-called low surface brightness (LSB) galaxies have been studied only recently (de Blok, McGaugh, & van der Hulst 1996, hereafter BMH; see also Pickering et al. 1997). These rotation curves, derived from H I observations, were found to rise more slowly than those of HSB galaxies of the same luminosity, if the radii are measured in kpc. At the outermost measured point, they were often still rising. McGaugh & de Blok (1998) noted that, with radii expressed in disk scale lengths, the rotation curve shapes of LSB and HSB galaxies become more similar, but not necessarily identical. Based on mass modeling of these H I rotation curves, de Blok & McGaugh (1997, hereafter BM) concluded that LSB galaxies were dominated by dark matter and that the contribution of the stellar disk to the rotation curve, even if scaled to its maximum possible value, could not explain the observed rotation curve in the inner parts.
The H I rotation curves of LSB galaxies have received a great deal of attention, because they provide additional constraints on theories of galaxy formation and evolution, and dark halo structure (e.g., Dalcanton, Spergel, & Summers 1997; Mihos, McGaugh, & de Blok 1997; Hernandez & Gilmore 1998; Kravtsov et al. 1998; McGaugh & de Blok 1998). Unfortunately, most of the galaxies studied in BMH and BM are only poorly resolved, making their results sensitive to the effects of beam smearing. For five of the galaxies in their sample, high resolution H$`\alpha `$ rotation curves, which have been obtained in order to eliminate beam smearing and to investigate the rotation curve shapes in the inner parts of LSB galaxies, are presented in this Letter.
## 2. Sample, observations and data reduction
The galaxies presented here were selected from the sample of LSB galaxies of BMH. Only galaxies were chosen that satisfied the criteria given in BM to define their useful rotation curves. An overview of the properties of the galaxies is given in Table LABEL:tabprops, which lists the name of the galaxy (1), the adopted distance in Mpc, for $`H_0=75`$ km s<sup>-1</sup> Mpc<sup>-1</sup> (2), the central surface brightness in mag arcsec<sup>-2</sup> (3), the disk scale length in kpc (4), the inclination angle (5), the position angle (6), the absolute magnitude (7) and the systemic velocity (8).
The observations were carried out at Palomar Observatory with the $`200^{\prime \prime }`$ Hale telescope<sup>4</sup><sup>4</sup>4The Palomar $`200^{\prime \prime }`$ telescope is operated in a joint agreement among the California Institute of Technology, the Jet Propulsion Laboratory and Cornell University, on November 20 1998. The FWHM velocity resolution was 54 km s<sup>-1</sup>, the pixel size in the spatial direction was $`0.5^{\prime \prime }`$. Each galaxy spectrum consisted of a single 1800s exposure. The slit was oriented along the major axis, at the position angle derived by BMH (see Table LABEL:tabprops). Despite their low surface brightnesses, all galaxies showed up clearly on the slit-viewing monitor. The slit could therefore accurately be aligned with the center of the galaxy by eye. The data were reduced using standard procedures in iraf and the resulting H$`\alpha `$ position-velocity diagrams are presented in Fig. 2.
Table 1
Properties of the galaxies<sup>a</sup> Name $`D`$ $`\mu _0^B`$ $`h`$ $`i`$ P.A. $`M_B`$ v$`{}_{}{}^{b}{}_{sys}{}^{}`$ (Mpc) (mag/$`^2`$) (kpc) () () (mag) (km s<sup>-1</sup>) (1) (2) (3) (4) (5) (6) (7) (8) F563-V2 61 22.1 2.1 29 148 $`18.2`$ $`4310\pm 4`$ F568-1 85 23.8 5.3 26 13 $`18.1`$ $`6524\pm 6`$ F568-3 77 23.1 4.0 40 169 $`18.3`$ $`5911\pm 3`$ F568-V1 80 23.3 3.2 40 136 $`17.9`$ $`5769\pm 7`$ F574-1 96 23.3 4.3 65 90<sup>c</sup> $`18.4`$ $`6889\pm 6`$
<sup>a</sup> Data from de Blok et al. (1996) and de Blok & McGaugh (1997).
<sup>b</sup> This Letter.
<sup>c</sup> A P.A. of $`90^{}`$ was used, derived from the optical image in BMH.
## 3. The high resolution rotation curves
To derive the rotation curves, we started by making Gaussian fits to the line profiles at each position along the major axis to obtain the radial velocities. These fits and their errors are overlayed on the H$`\alpha `$ position-velocity diagram in Fig. 2. The positions of the galaxy centers were determined from the peak of the continuum light along the slit. All galaxies were suffi-
ciently bright to allow to determine the position of the center along the slit with an accuracy of better than $`1^{\prime \prime }`$.
To obtain a larger radial coverage of the rotation curves, the H$`\alpha `$ data were combined with the H I data presented in BMH. To this end, the derived H$`\alpha `$ velocities were plotted on the H I position-velocity diagrams. Both sets of data were found to agree well with each other if the effects of beam smearing on the H I data are taken into account (cf. Swaters 1999). Therefore, we have used the H I data to determine rotation velocities beyond radii where we found H$`\alpha `$ emission. Note that in most cases the H I extends only little beyond the H$`\alpha `$. Next, the rotation curves derived for the approaching and the receding sides were combined. The H$`\alpha `$ points were sampled every $`2^{\prime \prime }`$, the H I points every $`7.5^{\prime \prime }`$ (approximately two points per beam). The errors on the rotation velocities were estimated from the differences between the two sides and the uncertainties in the derived velocities. We will refer to these rotation curves as the high resolution rotation curves (HRC).
The derived HRCs are shown in Fig. LABEL:figure2 together with the H I rotation curves presented in BM. Probably because BM did not correct for beam smearing, the H I rotation curves systematically underestimate the inner slopes of the rotation curves, especially for F568-V1 and F574-1. Both these galaxies have a central depression in the H I distribution, as can be seen in the H I maps presented in BMH. The spatial smearing of H I from larger radii into the observed central depression leads to an apparent solid body-like rotation curve, in particular for the highly inclined galaxy F574-1.
## 4. Mass modeling
For the mass modeling presented here we have used the same parameters for the thickness of the gaseous and stellar disks as BM have done. The stellar disk was assumed to have a vertical sech<sup>2</sup> distribution, with a scale height $`z_0=h/6`$. $`R`$-band light profiles, presented in de Blok, van der Hulst, & Bothun (1995) and BMH, were used to calculate the contribution of the stellar disk to the rotation curve. The H I was assumed to reside in an infinitely thin disk. The only difference with the mass models presented in BM is that we have decomposed the light profile of F568-1 into a disk and a central component, and fitted these components to the rotation curve separately. In the other galaxies no significant central component is present. For the dark matter component a pseudo-isothermal halo was used, following BM, which has a rotation curve given by:
$$v_{\mathrm{halo}}^2(r)=4\pi G\rho _0r_c^2\left[1\frac{r_c}{r}\mathrm{arctan}\left(\frac{r}{r_c}\right)\right],$$
where $`r_c`$ is the halo core radius and $`\rho _0`$ is the central density.
One of the major uncertainties in fitting mass models to rotation curves, in absence of an independent measurement of the stellar mass-to-light ratio $`\mathrm{{\rm Y}}_{}`$, is the uncertainty in the contribution of the stellar disk to the rotation curve. However, lower and upper limits on $`\mathrm{{\rm Y}}_{}`$, and hence on the dark matter content, can be obtained by assuming that the contribution of the stellar disk to the rotation curve is either minimal or maximal.
In the maximum disk mass models, the contribution of the stellar disk to the rotation curve is scaled to explain most of the inner parts of the rotation curve. The resulting rotation curve fits are shown in the top panel of Fig. 1. What immediately strikes the eye is that, in contrast with the findings of BM, the
Table 2
Mass model parameters max. disk no disk LSBC $`\mathrm{{\rm Y}}_{}`$ $`r_c`$ $`\rho _0`$ $`r_c`$ $`\rho _0`$ name M$`{}_{}{}^{}/\mathrm{L}_{R,}`$ kpc $`\mathrm{M}_{}\mathrm{pc}^3`$ kpc $`\mathrm{M}_{}\mathrm{pc}^3`$ F563-V2<sup>a</sup> 5.4 $`\mathrm{}`$ $`\mathrm{}`$ 0.94 0.283 F568-1<sup>b</sup> 17.2 $`\mathrm{}`$ $`\mathrm{}`$ 1.5 0.181 F568-3 1.5 3.0 0.027 2.5 0.48 F568-V1 9.3 6.7 0.005 1.2 0.188 F574-1 3.7 3.4 0.003 1.5 0.92
<sup>a</sup> No $`R`$-band data are available, mass modeling is based $`B`$-band data.
<sup>b</sup> For F568-1 a central component was fitted separately, which has $`\mathrm{{\rm Y}}_{,\mathrm{bulge}}=14.4`$ in the maximum disk fit.
inner parts of the rotation curves can be explained almost entirely by the contribution of the stellar disk in all of these LSB galaxies, with the exception of F568-3. The dark halo parameters (see Table LABEL:tabfits) are ill-defined for most galaxies in our sample, because most of the HRCs do not extend to large radii. Nonetheless, it is clear that in these maximum disk fits the contribution of the dark halo will only become important outside the optical disk, as is also the case for HSB galaxies.
The required stellar mass-to-light ratios for the maximum disk fits (listed in Table LABEL:tabfits), may be high, up to 17 in the $`R`$-band. Most of these are well outside the range of what current population synthesis models predict (e.g., Worthey 1994). If these high values of $`\mathrm{{\rm Y}}_{}`$ are to be explained solely by a stellar population, the stellar content and the processes of star formation in LSB galaxies need to be very different from those in HSB galaxies. Alternatively, these high mass-to-light ratios may indicate the presence of an additional baryonic component that is associated with the disk, as has been suggested by e.g. Pfenniger, Combes, & Martinet (1994). On the other hand, the fact that stellar disk can be scaled to explain the observed rotation curve may simply reflect the possibility that the luminous and dark mass have a similar distribution within the optical galaxy.
The other extreme for the contribution of the stellar disk to the rotation curve is to assume that its contribution is negligible. The dark halo parameters for this minimum disk fit are listed in Table LABEL:tabfits. High central densities of dark matter and small core radii are required to explain the observed steep rise in the HRCs. From Fig. 1 it is clear that the minimum disk mass models fit the rotation curves equally well as the maximum disk models. In fact, a good fit can be obtained with any mass-to-light ratio lower than the maximum disk mass-to-light ratio, demonstrating that the degeneracy that exists in the mass modeling for HSB galaxies (e.g. van Albada et al. 1985) also exists for LSB galaxies. Irrespective of the contribution of the stellar disk to the rotation curve, the similarity between the shapes of the observed rotation curves and those of the stellar disks implies that the total mass density and the luminous mass density are coupled within the region of the optical disk.
## 5. Discussion
The HRCs derived for the LSB galaxies rise steeply in the inner parts, and reach a flat part beyond about two disk scale lengths, as is found for HSB galaxies. In Fig. 4, the rotation curves for LSB galaxies (dotted lines) are compared with those of three typical late-type HSB galaxies from Begeman (1987), NGC 2403, NGC 3198 and NGC 6503. All the galaxies in Fig. 4 have no bulges, or only weak ones. In the top panel of Fig. 4 the radii are given in kpc. In these units, the rotation curves of LSB galaxies rise more slowly than those of HSB galaxies, indicating that these galaxies not only have lower central surface brightnesses, but also lower central mass densities, as was found by de Blok & McGaugh (1996) as well.
A different picture emerges in the bottom panel of Fig. 4, in which the rotation curves are scaled by their optical disk scale lengths and normalized to the velocity at two disk scale lengths. The normalization probably does not introduce systematic effects because the maximum difference in absolute magnitude between the galaxies in Fig. 4 is only 1.5 mag. With radii expressed in disk scale lengths, the rotation curves of the LSB galaxies presented here and those of HSB galaxies have almost identical shapes. This is consistent with the concept of a ‘universal rotation curve’ (Persic, Salucci, & Stel 1996, see also Rubin et al. 1985).
The similarity between the LSB and HSB rotation curve shapes suggests that rotation curve shapes are linked to the distribution of light in the stellar disks, independent of central disk surface brightness. Although such a link is most easily understood if the stellar disks are close to maximal, independent of surface brightness, the required high mass-to-light ratios seem to favor a picture in which LSB galaxies are dominated by dark matter within the optical disk, and HSB galaxies more by the stellar disk. The relative contribution of the stellar disk to the rotation curve may change continuously from LSB to HSB, or perhaps LSB and HSB galaxies constitute discrete galaxy families, as has been suggested by Tully & Verheijen (1997).
We thank Roelof Bottema for valuable discussions, and Erwin de Blok for making available the H I data and for useful comments. RS thanks IPAC for its hospitality during his visits which were funded in part by a grant to BFM as part of the NASA Long-Term Space Astrophysics Program. |
warning/0001/hep-ph0001068.html | ar5iv | text | # 1 Introduction
## 1 Introduction
The phenomena in the long-distance (or the low-momentum) domain of hadronic interactions are known to be dominated by non-perturbative mechanisms of QCD. While the lattice QCD simulations are widely accepted to provide the most direct description of these interactions, they contain the inherent difficulties connected with a finite lattice size and the violation of translational and rotational invariance. Therefore, those nonperturbative methods which maintain the mentioned symmetries are also needed, and the potential model either in the nonrelativistic or relativistic wave equation context belong to this category. Furthermore, the continued accumulation of data on availability of hadrons with the same quantum numbers, e.g., the vector mesons produced in the $`e^+e`$-annihilation, $`\tau `$-lepton decays, photo- and electro-production reactions as well as continued and more elaborated treatments of data , and phenomenological analyses of hadron form factors seem to favour the use of more sophisticated mixing schemes of constituent configurations in the considered resonance states. So, the mixing of quarkonia $`(q\overline{q})`$, the multiquark, e.g. $`(q^2\overline{q}^2)`$, and hybrid $`(q\overline{q}g)`$-states is observed as a really important problem of hadron spectroscopy. We believe that in solving this problem, those models will be advantageous which deal with translation-invariant wave functions depending on a correct number of degrees of freedom in the corresponding configuration space of two ( or more) particles. It seems natural to expect that the use of the same approach based on 2-, 3-, or 4-body relativistic equations with effective potential interactions between the constituents can present a sufficiently consistent approximate scheme to consider the indicated problem before resorting to more reliable but much more sophisticated numerical approaches to the genuine nonperturbative lattice QCD. In this paper, we outline some features of the approach, where we are going to keep as close as possible to known methods of dealing with one-particle relativistic and few-body nonrelativistic wave equations while discussing the spectroscopy of hadrons on the basis of Schrödinger-like relativized wave equations for bound quark-gluon systems.
## 2 The orbital and radial trajectories of light mesons
The valence quark model is known to be very successful in the description of mesons and baryons as $`q\overline{q}`$\- and $`q^3`$-states, where quarks interact via effective potentials . The principally important ingredient of these interactions is the linearly rising potential of confinement. This result follows from the ”quenched” approximation of the lattice QCD giving the flavour-independent force between static colour sources combined in the colour singlet states. The ”unquenching” procedure, i.e., inclusion of vacuum polarization due to the light quark loops can result both in the system- and state - dependence of effective potentials and in the modification or even termination of the linear behaviour of the confinement potential. A relevant way to locate these effects is to study the properties of particle states with the highest orbital and radial excitation of hadrons.
In this section, we report some results on mass spectra and radial structure parameters of higher excitations of quarkonia following the formulation proposed earlier for the potential approach to relativistic bound quark systems. A specific feature of the formulated approach is that the equations constructed include the squared forms of the Dirac hamiltonians of each particle interacting with all other particles of a system and additional variation conditions that define the one-particle energies via the total energy of a system, the only spectral parameter entering into the wave equation.
Here we concentrate on the topic of Regge–trajectories for light quarkonia to emphasize the special role of squared forms of long-range potentials of confinement in evaluation of characteristics of such states. It is just the squared form of the linear world-scalar potential of confinement that provides linearity of the Regge trajectories with the slope which can be made close to that following from the relativistic string theory. To illustrate this, we write first a general structure of our relativistic two-body equation in the form
$$\frac{1}{2W}(W^2+\stackrel{}{P}^2+M^2(p^2,r^2,..))\mathrm{\Psi }(1,2)=[\underset{i=1,2}{}\frac{1}{2\epsilon _i}(\epsilon _i^2+\stackrel{}{p}_{i}^{}{}_{}{}^{2}+m_i^2)+\widehat{V}]\mathrm{\Psi }(1,2)=W\mathrm{\Psi }(1,2).$$
(1)
$`\stackrel{}{P}=\stackrel{}{p}_1+\stackrel{}{p}_2,W=\epsilon _1+\epsilon _2.`$
The solution of the eigenvalue problem, i.e. the finding of the eigenmass $`M`$ for a given mass-operator $`M^2`$ in (1), is seen to correspond to zero of the inverse of the one-particle propagator written in the covariant form
$`P_\mu P^\mu M^2=0`$
The interaction kernel $`V`$ in (1) is defined when in the equation for noninteracting particles in their c.m.s.
$$W_0\mathrm{\Psi }_0(1,2)=[\underset{i=1,2}{}\frac{1}{2\epsilon _i}(\epsilon _i^2+\widehat{h}_0^2(i)]\mathrm{\Psi }_0(1,2)=[\underset{i=1,2}{}\frac{1}{2\epsilon _i}(\epsilon _i^2+\stackrel{}{p}_{i}^{}{}_{}{}^{2}+m_i^2)]\mathrm{\Psi }_0(1,2)$$
(2)
we alternately replace $`\widehat{h}_0^2(i)`$ by the squared forms of Dirac operator $`\widehat{h}(i)`$ for either particle in the field of another one
$$W\mathrm{\Psi }(1,2)=\{[\omega _1(\frac{1}{2\epsilon _1}(\epsilon _1^2+\widehat{h}^2(1))+\frac{1}{2\epsilon _2}(\epsilon _2^2+\widehat{h}_0^2(2))]+[12]\}\mathrm{\Psi }(1,2),$$
(3)
$$\widehat{h}^2(i)=\stackrel{}{p}_i^2+m_i^2+2m_iV_s(r_{ij})+V_s(r_{ij})^2+2\epsilon _iV_v(r_{ij})V_{v}^{}{}_{}{}^{2}(r_{ij})+\text{spin-dependent terms},$$
(4)
where the weight factors $`\omega _i`$ are normalized to unity:$`\omega _1+\omega _2=1`$.If $`m_1=m_2`$, then, by symmetry arguments, one can expect $`\omega _1=\omega _2=\frac{1}{2}`$. When $`m_1m_2`$, we suggest to use the simple relation $`\omega _1/\omega _2=\epsilon _2/\epsilon _1`$ that reproduces, for the static limit $`m_j\mathrm{}`$, the correct form of the corresponding one-particle equation in the field of a fixed center, e.g., the Klein-Gordon-Fock equation for spinless particles. In our case, the long-range confinement ”potential” is seen to be, in fact, the mass term parametrically dependent on parton configuration coordinates. As usual, in the spin-independent part of our interaction kernel, we retain a world-scalar part and the 0-th component of the vector interaction potential which survive in nonrelativistic approximation.
To be phenomenologically acceptable, our relativistic equation with a given Lorentz structure of the confining kernel should possess stable physical solutions, i.e., the binding energy should be real, and the state should be localized in a finite region of the coordinate space. As the presumed scalar and vector parts of the confining interaction squared enter into the equation with opposite signs, the scalar part is required to be stronger than the vector confinement potential: $`V_vV_s`$. The squared Coulomb potential gives the most singular part of the interaction. Hence for typical scales of the ground state and higher radial states of quarkonia with $`l=0`$, the effective $`\alpha _s(r)`$ values should be bounded from above:
$`Im[\sqrt{(l+{\displaystyle \frac{1}{2}})^2{\displaystyle \frac{1}{2}}({\displaystyle \frac{4}{3}}\alpha _s)^2}]=0,\alpha _s{\displaystyle \frac{3}{4\sqrt{2}}}.53.`$ (5)
This means that the effective coupling constant of QCD should be taken as ”freezing” at the value $`.5`$ by one of the proposed nonperturbative mechanisms (see, e.g., ) in the infrared region.
Below we are going to invoke also some ideas of the string-approximated QCD, or rather, a string-like solution of the bag model , to further specify the structure of the confining interaction and then to compare emerging results with the model-independent constraints on general relativistic bound states. It will be shown that the string-like structure of QCD, resulting e.g., from a semiclassical consideration of the rotating and deformed (”fat-string-like”) bag , can be represented by a much more simple potential model of mesons that is able to approximately reproduce some important features and results of the ”microscopic” field-theoretical models.
The famous constant $`B`$ of the MIT-bag model is known to define the volume energy of the space region inside an extended hadron, where nonperturbative vacuum fluctuations of coloured fields are at least partially suppressed, and it defines also the balance of the inward and outward pressure of vacuum fields and the coloured fields created by partons of a given hadron. This constant is considered to be connected with the contribution to the total hadron mass of the ”abnormal” part (i.e., with the nonzero trace) of the QCD energy-momentum tensor. The general model-independent statement stressed by Ji is that the ratios of contributions to the mass of the ”abnormal”($`\overline{T}_{\mu \nu }`$) and traceless ($`\widehat{T}_{\mu \nu }`$) parts of the energy-momentum tensor are
$`\overline{M}/\widehat{M}=1/3,`$ (6)
$`\widehat{M}/M_{tot}=3/4.`$ (7)
The effective string-potential, or rather, the energy of the string-like configuration of two static $`\overline{q}q`$-sources of a colour field, is found, following , to be
$`V_s(r)=a_{tot}r=(\overline{a}+\widehat{a})r=({\displaystyle \frac{128}{3}}\pi \alpha _sB)^{\frac{1}{2}},`$ (8)
where $`\overline{a}=\widehat{a}`$ represents, respectively, the contribution of the chromoelectric field of quarks (i.e., the traceless part of the gluon energy-momentum tensor), and the bag energy is connected with the gluon condensate, that is the trace-anomaly of QCD.
It was shown in that the classic consideration of rotation of the chromoelectric ”flux-tube” leads to a Regge-type relation between the classic nonquantized angular momentum $`J`$ and mass $`M_{string}`$
$`J={\displaystyle \frac{1}{2\pi a}}M_{string}^2`$ (9)
where two parts of $`M_{string},\widehat{M}_{string}`$, and $`\overline{M}_{string}`$ verify the general relations (6) and (7).
We use $`V_s`$ in the quantum-mechanical (quasi)potential two-body equation treating it alternately as the effective mass of one or another quark that parametrically depends on the interquark distance. According to the given definition, we obtain for massless quarks, with the neglect of spin-dependent terms,
$`(\stackrel{}{p}^2+{\displaystyle \frac{1}{2}}a^2r^2{\displaystyle \frac{1}{4}}W^2)\mathrm{\Psi }(1,2)=0`$ (10)
The solution gives a simple dependence of the mass $`M=M(l,n_r)`$ on the quantized orbital ($`l`$) and radial ($`n_r`$) quantum numbers
$`W^2M_{tot}^2=4\sqrt{2}a(l+2n_r+{\displaystyle \frac{3}{2}}).`$ (11)
characteristic of a harmonic oscillator. The harmonic quasipotential is obtained, however, after squaring the linear world–scalar ”potential” of confinement. The adopted way of inclusion of the interaction kernel into our relativistic two-body equation resembles the prescription of the so-called ”spectator”- type equation , where, alternately, one of the particles is put off-mass-shell, while the other is taken to be on-mass-shell.
According to (11), the Regge-trajectory slope $`\alpha ^{}(0)=(4\sqrt{2}a)^1`$ is within $`10\%`$ of the value $`\alpha ^{}(0)=(2\pi a)^1`$ following from the relativistic string theory. Replacing formally $`a`$ in (10) by $`\widehat{a}=(1/2)a`$, we keep in $`M^2`$ only contributions of the traceless $`\widehat{T}_{\mu \nu }`$-part of massless quarks and gluons. Therefore, we have
$`{\displaystyle \frac{\widehat{M}^2}{M_{tot}^2}}={\displaystyle \frac{1}{2}}`$ (12)
which is again within $`10\%`$ of the model-independent ratio (7).
The intercept $`\alpha _J(0)`$ of the leading $`J=l+1`$-trajectory is $`.5`$ according to (11), hence unrealistic. To calculate the realistic intercepts of Regge-trajectories, one should include the spin-dependent potentials, presumably the spin-orbit interaction induced by a scalar confinement ”potential”.
In this case, one can eliminate a still unknown parameter, using the value of mass and known quantum numbers of a certain state to evaluate masses of the states with other quantum numbers but with the same quark content. We note also that by analogy with the nonrelativistic case one can calculate the mean value of the commutator of the mass-operator $`M^2(p^2,r^2,..)`$ with the scalar product $`(\stackrel{}{p}\stackrel{}{r})`$
$`\mathrm{\Psi }_{l,n_r}|[M^2(p^2,r^2,..),(\stackrel{}{p}\stackrel{}{r})]_{}|\mathrm{\Psi }_{l,n_r}=0`$ (13)
to get the ”virial theorem”, demonstrating that the contribution of the mean value of the kinetic energy of two massless valence quarks to the value of the total mass of a highly excited meson is equal to the corresponding contribution of the mean value of the ”potential energy” that is connected with the energy integrated over gluon degrees of freedom of a given hadron. It is further tempting to interpret this fact as giving a hint for approximate equipartition, on the scale pertinent to a given bound state, of the total momentum of a hadron in the ”infinite momentum” frame between the valence quark and gluon-sea partons, the fact, following from the known moment of the nucleon structure function measured on the scales of deep inelastic lepton-hadron scattering.
Further, we discuss further in brief some characteristics of higher radial excitation of light vector quarkonia. This question is especially timely in view of recent development of the Vector Meson Dominance model applications to the analysis of nucleon electromagnetic form factors both in the spacelike and timelike regions of the transferred momenta $`Q^2`$. With the approximate equation (10), we obtain ”asymptotic” relations between masses of resonances with high spins $`J`$, lying on the same trajectory, and masses of higher radial excitations with the same spin
$`m^2(J,n_r)m^2(J^{^{}},n_r)=\alpha ^{^{}}(0)^1(JJ^{^{}})`$ (14)
$`m^2(J,n_r)m^2(J,n_r^{^{}})=2\alpha ^{^{}}(0)^1(n_rn_r^{^{}})`$ (15)
If we take $`\rho (2130)`$ as one of these higher radial states, then the next two are $`\rho (2600)`$ and $`\rho (3000)`$ according to (15) and the value $`\alpha ^{^{}}(0).9GeV^2`$. Curiously enough, a $`\rho `$ \- type resonance with a mass close to $`2.6`$ GeV was suggested in on the basis of analysis of nucleon form factors. The isoscalar partner $`\omega (3000)`$ of the second state, presumably, degenerated with it, is seen to be near in mass to the $`J/\psi `$ \- resonance, and it can play a role in the enhancement of certain strong decays of $`J/\psi `$. This situation deserves a more detailed study.
In the approach constructed, the energy functional has a recognizable quasi-nonrelativistic form, therefore, the perspective is open up to combine the accumulated experience in approximate solutions of the spectral problems in the nonrelativistic (NR) domain with the description of relativistic motion of hadron constituents.To this end, one should to have preferably an analytic, although approximate, solution of the corresponding NR problem. For example, in our case, one can simply estimate the dependence of the ”psi-at-zero” values
$`\psi (0)={\displaystyle \frac{m_{red}}{2\pi }}<V^{}(r)>`$ (16)
on masses or quantum numbers of corresponding meson states. Making use of the analogy of our equation (10) with the nonrelativistic Schrödinger equation, we obtain approximate scaling relations for the ”psi-at-zero” in the case of highly relativistic radially-excited states, hence, for leptonic widths of the corresponding resonances
$`\psi (0)^2={\displaystyle \frac{a^2}{4\pi }}<r>={\displaystyle \frac{am_{V_n}}{\sqrt{6\pi ^3}}}`$ (17)
$`\mathrm{\Gamma }_{ee}(V_n)\psi (0)^2/m_{V_n}^2(n1/4)^{\frac{1}{2}}`$ (18)
where $`m_{V_n}`$ is the mass of the radially excited $`nS`$-state of the vector resonance; $`n=n_r+1`$ , the principal quantum number; $`n_r`$, the radial quantum number. To get the absolute values of these leptonic widths, one should consistently include the QCD radiative corrections into the amplitude of $`\overline{q}qe^+e^{}`$ transition. The estimation of total widths of higher radial excitations of vector mesons can proceed as follows. With the meson mass of order $`2GeV`$ or larger, many decay channels are open, and one can rely on the quark-hadron duality idea while assuming the dominant role of the initial quark-parton stage of the reaction that defines the width dependence on quantum numbers and mass of the resonance. The total $`q\overline{q}`$ cross-section in the $`S`$-state is assumed to scale as $`m_{V_n}^2`$, where $`m_{V_n}`$ is the mass of the $`V_{n_r}`$-meson. At the hadronization stage, we adopt the simplest phase-space correction, assuming its form from a presumably main decay mode. For the isovector $`\rho _n`$-type resonances, the apparently conspicuous or, at least, important decay channel is the $`\rho (.77GeV)\pi \pi `$-channel. With adopted assumptions, one can get
$`\mathrm{\Gamma }_{tot}(V_{n+1})\mathrm{\Gamma }_{tot}(V_n){\displaystyle \frac{m_{V_{n+1}}f(\alpha _{n+1})}{m_{V_n}f(\alpha _n)}}`$ (19)
$`f(\alpha _{\rho _n})=1\alpha ^4+4\alpha ^2log\alpha `$ (20)
where $`\alpha _n=m_\rho /m_{\rho _n}`$, and the phase-space formula for the decay $`\rho _n(m_n)\rho (.77)2\pi `$ has been obtained for massless pions. So, with the input $`(m=2.6GeV;\mathrm{\Gamma }=.6GeV)`$ for mass and width of the heaviest claimed $`\rho `$-type resonance, we get for the next two resonances lying below the open charm threshold: $`(mass;width)(2.98;.77)`$ and $`(3.34;.92)`$. The very large and, seemingly, overestimated widths suggest, nevertheless, that hadronic corrections can be important in calculations of masses of high radially-excited states. This problem is still to be considered.
Concerning perspectives of the study of higher excited states within the outlined approach, one can notice that the lattice QCD simulations with the unquenched light quarks seem to be consistent with the linear behaviour of the confinement potential up to the distance of $`r2fm`$ . With the help of the virial theorem (2.13) one can relate the mean distance $`\sqrt{r_{q\overline{q}}^2}2fm`$ between the $`q\overline{q}`$-pair with the corresponding quantum numbers ($`l15;n_r=0`$), or ($`n_r9;l=0`$) of a given bound state which are considerably larger than the corresponding values $`l6÷7`$ or $`n_r4÷5`$ of the known meson resonances. Hence, it seems that many resonances could still be explored theoretically on the basis of relativistic wave equations with the linear confinement term, playing the role of part of effective quark (or gluon) mass, parametrically dependent on the distance between corresponding force centers.
## Acknowledgments
The author is grateful to the Organizing Committee of the NUPPAC’99 Conference, 13-17 November 1999, Cairo, Egypt, for the invitation and partial support. |
warning/0001/hep-ph0001276.html | ar5iv | text | # REFERENCES
ON THE $`\tau (a_1h)^{}\nu _\tau `$ DECAYS
K.R.Nasriddinov, B.N.Kuranov, U.A.Khalikov
Institute of Nuclear Physics, Academy of Sciences of Uzbekistan,
pos. Ulugbek, Tashkent,702132 Uzbekistan
T.A. Merkulova
Joint Institute for Nuclear Research, Dubna, 141980 Russia
> The $`\tau (a_1h)^{}\nu _\tau `$ decays of the $`\tau `$-lepton are studied using the method of phenomenological chiral Lagrangians. The expression of weak hadronic currents between pseudoscalar and axial-vector meson states is obtained. Calculated partial widths for these decays are compared with the available experimental data.
> PACS number(s): 13.35.Dx, 12.39.Fe
In this paper the $`\tau (a_1h)^{}\nu _\tau `$ decays of the $`\tau `$-lepton are studied using the method of phenomenological chiral Lagrangians (PCL’s) . The main uncertainty in the study of these decay channels is connected with weak hadron currents. Therefore such decay channels are a unique ”laboratory” for verification of weak hadron currents between pseudoscalar and axial-vector meson states and investigations of these decays are of interest. Note that the hadron decays of the $`\tau `$-lepton up to three pseudoscalar mesons in the final state and also $`\tau VP\nu _\tau `$ decays have been studied in the framework of this method.
In the PCL, the weak interaction Lagrangian, has the form
$$L_W=\frac{G_F}{\sqrt{2}}J_\mu ^hl_\mu ^++H.c.,$$
(1)
where $`G_F10^5/m_P^2`$ is the Fermi constant,
$`l_\mu =\overline{u_l}\gamma _\mu (1+\gamma _5)u_{\nu _l}`$ is the lepton current, and hadron currents have the form
$$J_\mu ^h=J_\mu ^{1+i2}\mathrm{cos}\mathrm{\Theta }_c+J_\mu ^{4i5}\mathrm{sin}\mathrm{\Theta }_c,$$
where $`\mathrm{\Theta }_c`$ is the Cabibbo angle.
Weak hadron currents between pseudoscalar and axial-vector meson states are obtained by including the gauge fields of these mesons in covariant derivatives :
$$_\mu _\mu +igv_\mu V+iga_\mu A,$$
(2)
here $`v_\mu ^i`$ and $`a_\mu ^i`$ are the fields of the $`1^{}`$ and $`1^+`$ \- mesons, $`V_i=\lambda _iI/2`$, and $`A_i=V_i\gamma _5`$ are the vector and axial-vector generators of the $`SU(3)\times SU(3)`$ group, respectively.
In this method the hadron currents are defined as
$$i\lambda ^iJ_\mu ^i=F_\pi ^2e^{i\xi A}(_\mu +igv_\mu V+iga_\mu A)e^{i\xi A},$$
(3)
here $`\xi =\frac{1}{F_\pi }\lambda ^i\phi ^i`$, $`F_\pi `$ = 93 MeV, $`\phi ^i`$ represent the fields of the $`0^{}`$ mesons, and $`g`$ is the ”universal” coupling constant, which is fixed from the experimental $`\rho \pi \pi `$ decay width
$$\frac{g^2}{4\pi }3.2.$$
The weak hadron currents between pseudoscalar and axial-vector meson states obtained in this way have the form
$$J_\mu ^i=F_\pi ga_\mu ^b\phi ^cf_{bci}.$$
(4)
Axial-vector and vector meson currents are defined as
$$J_\mu ^i=\frac{m_v^2}{g}v_\mu ^i+\frac{m_a^2}{g}a_\mu ^i,$$
(5)
where $`m_v`$ and $`m_a`$ are the masses of vector and axial-vector mesons , respectively.
The strong interaction Lagrangian of axial-vector mesons with vector and pseudoscalar mesons is obtained also by this way and has the form
$$L_S(1^+,1^{},0^{})=F_\pi g^2f_{klm}a_\mu ^kv_\mu ^l\phi ^m.$$
(6)
The decay amplitudes for these channels can be written as
$$M(\tau (k_\tau )a_1(p)h(p_1)\nu _\tau (k_\nu ))=G_Fϵ_\mu ^\lambda \overline{U}(k_\nu )\gamma _\mu [f_1+g_1\gamma _5+\widehat{p}(f_2+g_2\gamma _5)$$
$`+\widehat{p}_1(f_3+g_3\gamma _5)]U(k_\tau ),`$
where $`ϵ_\mu ^\lambda `$ is the polarization vector of $`1^\pm `$ mesons, $`f_i`$ and $`g_i`$ are the form factors that depend on the final state momenta; $`q=k_\tau k_\nu =p+p_1`$, and $`k_\tau `$, $`k_\nu `$ are the lepton four-momenta ($`\widehat{p}_ip_{i\mu }\gamma ^\mu `$).
Using these Lagrangians we calculated the partial widths of the $`\tau (a_1h)^{}\nu _\tau `$ decays by means of the TWIST code . The results are shown in the Table I. In columns I and II are listed the results without $`1^{}`$ contributions, and with the vector $`1^{}`$-meson contributions, respectively.
These decay channels get contributions from the $`\rho (770)`$-, $`\rho (1450)`$-, and $`\rho (1700)`$\- vector intermediate meson states which have widths of 150, 310, and 240 MeV, respectively. Note that the contribution of the $`\rho (1450)`$\- and $`\rho (1700)`$\- mesons to the partial widths dominate those of the $`\rho (770)`$ ones.
Table I shows that the result obtained for the $`\mathrm{\Gamma }(\tau ^{}(a_1h)^{}\nu _\tau )=0.79\times 10^{10}sec^1`$ decay channels without taking into account the vector $`1^{}`$ contributions are in good agreement with available experimental data :
$`\mathrm{\Gamma }(\tau ^{}(a_1h)^{}\nu _\tau )<6.9\times 10^{10}sec^1`$.
Note that the calculated partial widths with taking into account $`1^{}`$ contributions lie above this experimental value. In these calculations we used, as in Ref.s \[2-5\], the same $`g`$-coupling constant, according to Eq. (2), for all the vector intermediate meson states. Indeed, it is a rough appoximation and it was more appreciable in study of such $`\tau `$ lepton rare decays than in Ref.s \[2-5\]. Therefore, it would be expedient to present $`g`$-coupling constant in Eq. (2) in a matrix form so that various decay channels have their own coupling constants. Though at present we have shortage of experimental data on the vector intermediate mesons, but there are some theoretical attempts to determine coupling constants of such mesons (see Ref.s ). And taking into account corresponding coupling constants in future would allow us to describe these decays more correctly compared to these calculations.
Note that according to Eq.(4) the partial widths of the $`\tau ^{}a_1^{}\eta \nu _\tau `$ and $`\tau ^{}a_1^{}\eta ^{}\nu _\tau `$ decays are equal to zero in the PCL method; as in Ref., these decay channels can be realized via effects of secondary importance .
Thus, the expression of weak hadronic currents between pseudoscalar and axial-vector meson states Eq. (4) obtained by including the gauge fields of axial-vector and vector mesons in covariant derivatives allow us to describe the $`\tau (a_1^{}h)^{}\nu _\tau `$ decays in satisfactory agreement with available experimental data. Determination of corresponding coupling constants in Eq. (4) would allow us to calculate these decay probabilities with high accuracy. Probably, contributions from the vector intermediate mesons which are very sensitive to $`g`$ would be in satisfactory agreement with the experimental data.
We would like to thank F. Hussain, M. Fabbrichesi, D. V. Sao, S. Azakov, M.M.Musakhanov, and A.Rakhimov for interest in this study and for useful discussions. One of authors (KRN) would like to thank Prof. S. Randjbar-Daemi for hospitality at the ICTP during the course of this work.
Table 1. The partial widths (in $`sec^1`$) for the $`\tau ^{}(a_1h)^{}\nu _\tau `$ decays.
| Decays | I | II | Experiment |
| --- | --- | --- | --- |
| $`\tau ^{}a_1^0\pi ^{}\nu _\tau `$ | $`0.39\times 10^{10}`$ | $`0.5\times 10^{11}`$ | $``$ |
| $`\tau ^{}a_1^0K^{}\nu _\tau `$ | $`0.73\times 10^5`$ | $`0.92\times 10^4`$ | $``$ |
| $`\tau ^{}a_1^{}\pi ^0\nu _\tau `$ | $`0.40\times 10^{10}`$ | $`0.52\times 10^{11}`$ | $``$ |
| $`\tau ^{}a_1^{}\overline{K}^0\nu _\tau `$ | $`0.92\times 10^5`$ | $`0.19\times 10^5`$ | $``$ |
Fig.1. Diagrams for the $`\tau ^{}(a_1h)^{}\nu _\tau `$ decays, here W and S are the vertices of weak and strong interactions, respectively. (a) is without the pole conrtibution of $`1^{}`$ mesons and (b) includes these pole contributions. |
warning/0001/cond-mat0001255.html | ar5iv | text | # Ground and Metastable States in 𝛾-Ce from Correlated Band Theory
## Abstract
Multiple energy minima of the LDA+U energy functional are obtained for $`\gamma `$-Ce when it is implemented in a full potential, rotationally invariant scheme including spin-orbit coupling, and different starting local configurations are chosen. The lowest energy solution leads to a fully spin polarized 4f state and the lattice constant of $`\gamma `$-Ce. The higher energy local minima (additional self-consistent solutions) are shown to be strongly indicative of crystal electric field and multiplet excitations.
The isostructural $`\alpha `$ \- $`\gamma `$ phase transition in Ce, and the character of the phases separately, have been a subject of continious theoretical efforts for decades. Experimentally, the low temperature fcc $`\alpha `$ phase of Ce transforms to the fcc $`\gamma `$ phase at the high temperatures with a large increase of the volume. When pressure is applied, the crystal collapses back into the $`\alpha `$ phase in a first-order phase transition. The room temperature $`\gamma \alpha `$ phase transition occurs at $``$ 7 kbar pressure with the decrease of the volume of $``$ 15%.
Several theoretical models have been suggested and they are described in some detail in Ref. . The difference between the models arises from the different treatments for the Ce 4f electrons. The promotional model assumed depopulation of the 4f level upon compression, but does not agree with the results of various experiments which indicate little change in the 4f occupation. Band structure calculations also indicate that promotion is too high in energy to drive the transition, and the 4f occupation remains near unity. The Mott transition model assumes the transformation of localized to band-like 4f state with the decrease of volume and is similar to the Mott-Hubbard metal-insulator transition. First principles calculations using the self interaction correction (SIC) to the LDA produce total energy minima for localized ($`\gamma `$) and band ($`\alpha `$) states with formally the same SIC total energy variational functional, and are in accord with the Mott transition model. However, the use of atomic sphere approximation (ASA), which sphericalizes the potential and density within large atomic spheres, limits the ability of these calculations to produce quantitative total energy description of the system with highly non-spherically symmetric 4f electron charge/spin densities.
The other viable theoretical model is the Kondo volume collapse (KVC) model . The essence of the KVC is an assumption of localized f states in the both $`\alpha `$ and $`\gamma `$ phases of Ce. The $`\alpha `$ phase consists of a mixed valence 4f state while the $`\gamma `$ phase has almost integer 4f occupancy. The phase transition is due to the entropy contribution of the localized 4f state. The relation between KVC and Mott transition model was analyzed recently in Ref. and quantitative arguments in favor of KVC model were provided.
The local (spin) density approximation (LDA) has had tremendous success in the quantitative description of a wide varietly of solids, but the rare earths are extreme cases where the LDA description is inadequate. The “correlated band theory” (LDA+U) approach has had much success in the treatment of correlated magnetic insulators where LDA results are incorrect. In this paper we explore not the $`\gamma \alpha `$ transition per se, but the $`\gamma `$ phase itself, by an application of the LDA+U approach that includes the new features (1) a full potential method is used, which enables the novel features we uncover, and (2) the LDA+U method is applied to a metal where the interaction of the local orbitals with the conduction states is a central part of the physics. For the first time, multiple LDA+U energy minima (ground plus metastable states) are obtained within a specified broken symmetry, and we discuss how the various states are related to 4f excitations in the $`\gamma `$ phase.
The LDA+U method is based on an energy which is a functional of the spin densities {$`\rho ^s,s=\pm 1`$}, and the occupation matrices {$`n_{m,m^{}}^s`$} of the 4f orbitals labelled by their azimuthal projections $`mm_{\mathrm{}}`$. There is every reason to expect that, within the allowed values of $`\rho ^s`$ and $`n^s`$, there can be local energy minima as well as the absolute minimum, which represents the ground state energy of the system. (Such local minima are rarely found within LDA except in magnetic systems.) The formal meaning of these local minima, as well as the formal underpinnings of the LDA+U approach itself, remain to be settled, but – like the Kohn-Sham eigenvalues (the band structure) which have little formal meaning but immense practical importance – these minima will be shown to bear a close relationship to local excitations of the 4f shell.
Our results are based upon the full-potential linearized augmented plane wave method (LAPW) as the basis for total energy calculations with the rotationally invariant LDA+U functional , with spin-orbit coupling (SOC) is included self-consistently . Literature values are used for the on-site repulsion $`U=\mathrm{\hspace{0.33em}6.1}`$ eV and exchange $`J=\mathrm{\hspace{0.33em}0.7}`$ eV (Slater integrals, $`F_0=`$ 6.10 eV, $`F_2=`$ 8.34 eV, $`F_4=`$ 5.57 eV, $`F_6=`$ 4.12 eV ). Ferromagnetic spin alignment is assumed (allowed, not imposed) since local moments are present in paramagnetic high temperature $`\gamma `$ phase of Ce .
The key to finding the various solutions (local minima) is to start from atomic densities and different ‘guesses’ for the 4f occupation matrices $`n^s`$ and then obtain self-consistently both spin/charge densities and occupation matrices. As an intuitive guess for 4f occupations we use the fully spin-polarized state (Tr $`n^{}`$=1, Tr $`n^{}`$=0) and choose various orbital characters for $`n_{m,m^{}}^{}`$.
Ground State. The calculated equilibrium lattice constant (cf. Table I) is very close to the experimental value of $`\gamma `$-Ce at zero pressure and room temperature . The calculated bulk modulus is about 25% larger than the experimental value, similar to other calculations that have treated the localized character of the 4f state in $`\gamma `$-Ce.
For the $`\gamma `$ phase volume (lattice constant a=9.76 a.u.) we find the lowest solution to be fully spin polarized with the f occupation of 1.04. This state is primarily $`m_l`$=-2 ($`m_j=\frac{3}{2})`$, with some mixing in of the spin-majority $`m_l=2`$ state. The LDA+U yields significant enhancement for the absolute values of both spin and orbital magnetic moments (anti-aligned in accord with the 3rd Hund’s rule) in comparison with LDA results.
The electronic density of states for the $`\gamma `$-Ce ground state is shown in Fig. 1 in comparison with the result of relativistic (with SOC) LDA calculations. The f-majority peak at the bottom of valence band ($``$ 2.5 eV below the Fermi level) indicates the position of the localized 4f state, in quantitative agreement with the resonant photoemission measurements . The difference between LDA+U and LDA is due to the localization of the 4f state, which removes both majority and minority 4f states from the vicinity of the Fermi level. Since the 4f state is not well separated from the valence band (cf. Fig. 1) we conclude that in spite of its localized character the 4f state in $`\gamma `$-Ce cannot be treated as core-like.
Metastable States. As mentioned above, several self-consistent solutions corresponding to different (local) minima of the LDA+U energy functional are possible. Different minima must be searched for by exploring various regions of the underlying space. Our initial starting points consisted of fully spin polarized 4f<sup>1</sup> with different orbital characters {$`m_l`$}. Our initial and final (self-consistent) majority spin occupation matrix elements are shown in Table II. In the LDA+U calculations the spin polarization enters as an effective magnetic field providing the spin quantization axis and the direction of the quantization axis is chosen along z().
The LDA+U minima can be considered as mean-field-like projections (in the Fock space) of many-body wavefunctions on the single-particle angular basis set. The interpretation is much simpler for an f<sup>1</sup> ion than it would be for a multi-electron ion because the $`\{m_l,m_s\}`$ and $`\{J,J_z\}`$ representations are unitarily related. In the simple crystal-electric-field (CEF) model for the paramagnetic Ce<sup>3+</sup> ion the ground state is formed by one of the $`\mathrm{\Gamma }_7`$ and $`\mathrm{\Gamma }_8^{1,2}`$ doublets from $`J=5/2`$ multiplet . In order to compare the CEF model with the LDA+U results, we apply to the LDA+U solutions the unitary transformation from $`\{m_l,m_s\}`$ representation to $`\{J,J_z\}`$ representation with $`J=5/2,7/2`$. The states from Table III are then classified as follows:
Solution (1) - Ground State: It has 96% $`m_l=2`$, with 3% $`m_l=2`$ mixed in. The transformation to $`\{J,J_z\}`$ representation yields 69% of $`|5/2,3/2>`$ and 1% of $`|5/2,5/2>`$ states from $`J=5/2`$ multiplet, plus 27% of $`|7/2,3/2>`$ and 3% of $`|7/2,5/2>`$ states from $`J=7/2`$. We conclude that the Solution 1 consists of 70% states from the $`J=5/2`$ multiplet which are the linear combination of the CEF levels $`\mathrm{\Gamma }_7`$ and $`\mathrm{\Gamma }_8^1`$, and 30% states from the $`J=7/2`$ multiplet.
Solution (2): Roughly equal amounts of $`m_l=3`$ and $`m_l=+1`$) in this solution indicates coupled $`m_j=5/2`$ and $`m_j=3/2`$ states. It consists of 55% states from $`J=5/2`$ multiplet (combinations of the CEF levels $`\mathrm{\Gamma }_7`$ and $`\mathrm{\Gamma }_8^1`$) and 45% of states from $`J=7/2`$.
Solution (3), involving admixture of $`m_l=1`$ with 25% $`m_l=+3`$, which transforms to an admixture of 42% of $`|5/2,1/2>`$ ($`|\mathrm{\Gamma }_8^2>`$) state with the 58% of $`|7/2,7/2>`$.
Solution (4), pure $`m_l=0`$, corresponds to the combination of 42% of $`|5/2,1/2>`$ ($`|\mathrm{\Gamma }_8^2>`$) state with the 58% of $`|7/2,1/2>`$, and lies 232 meV above the ground state.
These correlated band structure calculations are producing metastable solutions that contain the CEF states from the conventionally presumed lowest multiplet for Ce<sup>3+</sup>. In addition, there is a considerable fraction of the states from the first excited $`J_{7/2}`$ multiplet that must be understood.
There are two important reasons why our solutions do not correspond directly to CEF levels as normally considered. First, the CEF picture assumes ideal cubic symmetry, whereas a 4f state in $`\gamma `$-Ce is actually surrounded by twelve atoms whose own moments are oriented randomly, hence breaking cubic symmetry. In addition the non-cubic nature of the 4f density perturbes the conduction electron density, which leads to a non-cubic local field. Secondly, in our calculations we have artificially ordered the spins of the magnetic ions and allowed orbital moments, which reduces out site symmetry to tetragonal.
The comparison between LDA+U calculated splitting scheme with the results of inelastic neutron scattering experiments is shown in Fig. 2. It is clearly seen that Solutions 1-3 lie well within the range of low-energy excitations peak and reflect the mixed CEF and spin-orbit excitations. The energy position of Solution 4 (232 meV) agrees quantitatively with the experimental inelastic peak at 260 meV. This peak is usually interpreted as the spin-orbit excitation $`{}_{}{}^{2}F_{5/2}^{}^2F_{7/2}`$ . Our calculations suggest that this interpretation is oversimplified since the states from both lower $`{}_{}{}^{2}F_{5/2}^{}`$ and first excited $`{}_{}{}^{2}F_{7/2}^{}`$ multiplets are mixed in the ground and excited states of $`\gamma `$-Ce. The spectrum from present LDA+U calculations are in better agreement with the experiment than previously reported results of SIC calculations (86-100 meV and 130 meV).
In order to classify the LDA+U solutions we transform them from the compex to the cubic spherical harmonics. Solution (1) involves admixture of $`xyz`$ and $`(x^2y^2)z`$ states; Solutions (2) and (3) consist of mixed $`(x^2z^2)y`$ and $`(y^2z^2)x`$ states; and Solution (4) has $`(5z^23r^2)z`$ symmetry. All the LDA+U solutions have the tetragonal symmetry assumed in the calculations. The localized states without SOC are formed by CEF states in $`\{L,L_z\}`$ representation and for $`L=3`$ these are $`\mathrm{\Gamma }_2,\mathrm{\Gamma }_4,\mathrm{\Gamma }_5`$ states . In fully spin-polarized case the spin-orbit coupling $`\xi (\stackrel{}{l}\stackrel{}{s})`$ is reduced to $`\xi (\widehat{l}_z\widehat{s}_z)`$ and the eigenvalues of localized “CEF+SOC” model Hamiltonian are easily obtained. The “CEF+SOC” coupling scheme is presented in Fig. 3 and shows how the LDA+U solutions result from CEF eigenstates coupled by spin-orbit coupling. From the total energy differences between LDA+U solutions we then derive the CEF splitting parameters $`\mathrm{\Delta }_{2,5}=\mathrm{\Gamma }_2\mathrm{\Gamma }_5`$ = 90 meV and $`\mathrm{\Delta }_{4,2}=\mathrm{\Gamma }_4\mathrm{\Gamma }_2`$ = 107 meV and SOC parameter $`\xi `$ of 66 meV (SOC splitting = $`(7/2)\xi `$ = 231 meV).
The scheme in Fig. 3 allows more eigenstates than the calculated local minima of the LDA+U total energy functional. In each case the occupation matrices $`n_{m_l,m_l^{}}`$ for these additional states have the non-zero elements in the same $`\{m_l,m_l^{}\}`$ sub-space as one of the metastable states. As a result the minimization procedure yields the lowest total energy state in given $`\{m_l,m_l^{}\}`$ sub-space and the higher states do not appear among variational LDA+U solutions.
The physical origin of the CEF splitting can be understood to be due to the anisotropy of the mixing interaction between conduction band and localized f-states . We obtain the CEF splitting for $`\gamma `$-Ce of the order of 100 meV comparable with the SOC splitting of 231 meV. In both ground and excited states the CEF levels are then mixed by SOC yielding low-energy excitations of the order of 20-50 meV and high-energy excitations about 230 meV.
The results of LDA+U calculations allow us to conclude that the interaction between conduction band and localized f-states plays an important role for the both ground and excited states in $`\gamma `$Ce. As a result, the Ce f-states are not core-like and should be treated in terms of quantum impurity (Anderson) model rather than as localized states of Ce<sup>3+</sup> ion in the cubic CEF. The present LDA+U calculations correspond to the numerical solution of the lattice Anderson model in static mean-field approximation with the assumed ferromagnetic order. This simplification does not allow us to describe quantitatively the paramagnetic (disordered local moment) high-temperature state of $`\gamma `$Ce observed experimentally. The treatment of this paramagnetic state requires the use of dynamic mean-field parametrization for the self-energy.
Recently, Solovyev et al. proposed a correction to the LDA+U total energy functional when SOC is included and non-collinear magnetic configurations are considered. This correction accounts for additional contributions to the exchange energy due to non-zero spin-off-diagonal elements of the occupation matrix {$`n_{m,m^{}}^{s,s}`$}. We performed LDA+U calculations with this spin-off-diagonal correction and found that the spin-off-diagonal occupations are very small and have minute effects on the values of spin and orbital magnetic moments, and the total energies (less than 2 meV) for ground and metastable states (cf. Table II) in $`\gamma `$Ce.
To summarize, we have obtained the ground and three metastable states from relativistic (with spin-orbit coupling) spin-polarized full-potential LDA+U calculations for fcc Ce. The ground state has equilibrium lattice constant of $`\gamma `$-Ce. Our calculations reproduce the localized character of 4f states in $`\gamma `$-Ce, which however cannot be treated as a part of atomic core. Analysis of various LDA+U solutions allows us to make a comparison between correlated band theory results and CEF model. The excitation energies calculated from the total energy differences between LDA+U solutions are in reasonable agreement with the experimental data.
We are grateful to D. L. Cox for helpful comments. This research was supported by National Science Foundation Grant DMR-9802076, and by NSF Grant PHY-94-07194 while A.I.L and W.E.P. were in residence at the Institute of Theoretical Physics, UCSB. |
warning/0001/hep-ph0001114.html | ar5iv | text | # Relic neutralinos – update on neutralino–nucleon scalar cross-section
## Abstract
We discuss the effect induced on the neutralino–nucleon scalar cross–section by the present uncertainties in the values of the quark masses and of the quark scalar densities in the nucleon. We examine the implications of this aspect on the determination of the neutralino cosmological properties, as derived from measurements of WIMP direct detection. We show that, within current theoretical uncertainties, the DAMA annual modulation data are compatible with a neutralino as a major dark matter component, to an extent which is even larger than the one previously derived.
Experiments of direct search for WIMPs have remarkably improved their sensitivity in the last years, allowing the exploration of sizeable regions of the physical parameter space of specific particle candidates for dark matter. This is the case of the neutralino , for which some direct detection experiments are already capable of investigating significant features in domains of the parameter space which are also under current exploration at LEP2.
Currently, the most sensitive direct search experiments are probing, or are starting to probe, a range of $`\rho _\chi ^{0.3}\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ from about a few $``$ $`10^9`$ to $`110^8`$ nbarn, where $`\rho _\chi ^{0.3}`$ denotes the local dark matter density normalized to 0.3 GeV cm<sup>-3</sup> and $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ is the neutralino-nucleon scalar cross section . This goal has already been achieved by the DAMA experiment , which has reported the indication of an annual modulation effect in its counting rate, compatible with
$$310^9\text{nbarn}<\rho _\chi ^{0.3}\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}<110^8\text{nbarn},$$
(1)
for values of the WIMP mass which extend over the range 30 GeV $`<m_\chi <`$ 130 GeV . The region in the $`m_\chi `$$`\rho _\chi ^{0.3}\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ plane, singled out by the DAMA experiment at 2–$`\sigma `$ C.L., is the one depicted in Fig. 1. We also notice that the uncertainties in the local total dark matter density: 0.1 GeV cm$`{}_{}{}^{3}\rho _l`$ 0.7 GeV cm<sup>-3</sup>, imply for $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ the range:
$$110^9\text{nbarn}<\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}<310^8\text{nbarn},$$
(2)
Another experiment of WIMP direct detection which is now entering the upper left corner of the region in Fig. 1 is the CDMS experiment .
Once a given range for $`\rho _\chi \sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ is singled out by an experiment, the implications for specific particle candidates rely on the theoretical calculation of the $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ cross-section. In the case of neutralino, this quantity usually takes dominant contributions from interaction processes where neutralinos and quarks inside the nucleon interact by exchange of Higgs particles or squarks. The relevant couplings involve the use of quark masses $`m_q`$ and quark scalar-densities inside the nucleon $`N|\overline{q}q|N`$, whose values can be related to some physical observables which can be identified with the pion–nucleon sigma term $`\sigma _{\pi N}`$, the fractional strange–quark content of the nucleon $`y`$ and the strange–to–light-quark mass ratio $`r=2m_s/(m_u+m_d)`$ (see Ref.).
Actually, the quantities $`\sigma _{\pi N}`$ and $`y`$ have recently been the object of various evaluations, based mainly on chiral perturbation theory and on QCD simulations on a lattice; however, the present situation, which is reviewed in Ref., is still far from being clear. Thus, the Higgs–quark–quark and squark–quark–neutralino couplings are still plagued by significant uncertainties, and this reflects on large uncertainties on the calculations of $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$. From the experimentally allowed values of the relevant quantities $`\sigma _{\pi N}`$, $`y`$ and $`r`$ , we select three sets of values as representative of the uncertainties which can affect the calculation of the neutralino–nucleus cross section. These three sets are reported in Table 1: set 1 is our reference set, while set 2 and set 3 are representative of choices which can provide enhanced values for $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$. We wish to remark that all these choices are well inside their allowed intervals coming from the determination of the quantities $`\sigma _{\pi N}`$, $`y`$ and $`r`$ .
Fig. 1 shows the scatter plot of the quantity $`\rho _\chi ^{0.3}\sigma _{\mathrm{s}calar}^{(nucleon)}`$ calculated in the minimal supersymmetric standard model (MSSM), for a general scan of its parameter space (for details, see Refs. ) and for set 1. We see that the annual modulation region is fully compatible with the hypothesis of a relic neutralino as a dark matter component .
In order to discuss the astrophysical and cosmological properties of a relic neutralino in order for it to be compatible with the indication coming from the DAMA experiment, in Fig. 2 we show a scatter plot of $`\rho _\chi `$ versus $`\mathrm{\Omega }_\chi h^2`$, obtained as follows: i) $`\rho _\chi `$ is evaluated as $`\rho _\chi =[\rho _\chi \sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}]_{R_m}/\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$, where $`[\rho _\chi \sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}]_{R_m}`$ denotes the set of experimental values of $`\rho _\chi \sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ inside the DAMA annual modulation region and $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ is calculated in the MSSM; ii) To each value of $`\rho _\chi `$, which then pertains to a specific susy configuration, one associates the corresponding value of $`\mathrm{\Omega }_\chi h^2`$, calculated as indicated, for example, in Ref. and references therein. With this procedure, we determine the values of $`\rho _\chi `$ which, for each calculated $`\sigma _{\mathrm{scalar}}^{(\mathrm{nucleon})}`$ satisfy the DAMA annual modulation data.
The effect induced by the choice of different sets for the parameters $`\sigma _{\pi N}`$, $`y`$ and $`r`$ is shown in Fig. 2 by the solid lines. The three lines delimit the region covered by the scatter plots in the different cases, and show to which extent the region moves downward when these parameters are varied.
The most interesting feature of Fig. 2 is that it shows that the set of susy configurations selected by the DAMA data has a significant overlap with the region of main cosmological interest: $`\mathrm{\Omega }_\chi h^2>0.02`$ and 0.1 GeV cm$`{}_{}{}^{3}\rho _\chi 0.7`$ GeV cm<sup>-3</sup>. The extent of this overlap is increasingly larger for set 2 and set 3. By way of example, for set 3 one has that, at $`\rho _\chi =0.3`$ GeV cm<sup>-3</sup>, $`\mathrm{\Omega }_\chi h^2`$ may reach the value 0.3. Therefore, these results reinforce our conclusions of Ref. , i.e. that the DAMA annual modulation data are compatible with a neutralino as a major component of dark matter, on the average in the Universe and in our Galaxy.
The same figure shows that different situations may also occur. For instance, for configurations which fall inside the band delimited by the slant dot–dashed lines, the neutralino would provide only a fraction of the cold dark matter both at the level of local density and at the level of the average $`\mathrm{\Omega }`$, a situation which would be possible if the neutralino is not the unique cold dark matter particle component. Clearly, configurations above the upper horizontal line are incompatible with the upper limit on the local density of dark matter in our Galaxy and must be disregarded.
Acknowledgements. This work was supported by the Spanish DGICYT under grant number PB95–1077 and by the TMR network grant ERBFMRXCT960090 of the European Union. |
warning/0001/gr-qc0001039.html | ar5iv | text | # ROTATING 5D-KALUZA-KLEIN SPACE-TIMES FROM INVARIANT TRANSFORMATIONS
## 1 Introduction
In recent years, dilaton fields have been proposed as a strong candidate for describing dark matter. At a cosmological level it has been used to explain the Large Scale Structure of the Universe . At a galactic level, scalar fields seem to play a crucial role in explaining the curves of rotational velocities vs. radius, observed in all the galaxies , and it is thought that it will also play an important role in several physical phenomena at a local level , $`i.e.,`$ in the realm of compact objects.
From a theoretical point of view, dilaton fields coupled to Einstein-Maxwell fields, naturally appear in the low energy limit of string theory, and as a result of a dimensional reduction of the Kaluza-Klein Lagrangian. Therefore, the study of the Einstein-Maxwell-Dilaton theory is of importance to investigate the properties of compact objects involving these fields and for the understanding of more general theories.
On the other hand, if the scalar fields are so important in physics, why they have not been yet detected? As we just mentioned, in several models they play an important role, but due to the fact that they interact very weakly with matter , in most of the observational tests the results at most just can not exclude them. It is expected that the scalar fields will have an important measurable signature of their presence in regions with strong gravitational fields . Thus, it is necessary to have exact analytical solutions to the Einstein-Maxwell-Dilaton theory, not only perturbative solutions, and then compare with the observations the predictions made using those exact solutions. The problem with this approach, is that the field equations are very complicated to be solved exactly and one must recur to mathematical methods which usually prove to be very cumbersome. In this work we want to give some simple formulas which allow us to derive exact rotating dilatons solutions, starting from static ones, and which avoids many of the mathematical difficulties usually encountered in deriving exact solutions from seed ones. In this work we derive three expressions which can be used to genereate families of solutions, starting from known seed ones. Out of these expressions, ony one has been previously obtained in reference
In order to do so, let us start from the Lagrangian
$$=\sqrt{g}[R+2(\varphi )^2+e^{2\alpha \varphi }F^2]$$
(1)
This Lagrangian contains very interesting limits. For $`\alpha ^2=3`$ Lagrangian (1) contains the Kaluza-Klein theory; for $`\alpha ^2=1`$, equation (1) represents the effective Lagrangian for the low energy limit of super-strings theory; finally, equation (1) contains the Einstein-Maxwell theory with a minimally couple scalar field for $`\alpha ^2=0`$. This Lagrangian is also very convenient because after a conformal transformation of the metric, one can obtain a equivalent Lagrangian for an almost arbitrary scalar-tensor theory of gravity (with a non trivial electromagnetic-scalar interaction which can be avoided setting $`F^2=0`$). The field equations derived from Lagrangian (1) are give by
$`_\mu (e^{2\alpha \varphi })F^{\mu \nu }`$ $`=`$ $`0;`$
$`^2\varphi +{\displaystyle \frac{\alpha }{2}}e^{2\alpha \varphi }F^2`$ $`=`$ $`0;`$
$`R_{\mu \nu }`$ $`=`$ $`2_\mu _\nu \varphi +2e^{2\alpha \varphi }(F_{\mu \rho }F_{\nu }^{}{}_{}{}^{\rho }{\displaystyle \frac{1}{2}}g_{\mu \nu }e^{2\alpha \varphi }F^2).`$ (2)
There exist several exact solutions of equations (2) (see reference for solutions with $`\alpha ^2=3`$ and for their generalization to $`\alpha `$ arbitrary). Some of them could be models for the exterior space-time of an astrophysical compact object or of a black hole with a scalar field interaction . Let us give two examples of such metrics. The first space-time we deal with behaves gravitationally like the Schwarzschild solution for $`\alpha 0`$ and contains an arbitrary magnetic field. This metric reads
$$ds^2=e^{2k_s}\text{g}^\gamma \frac{dr^2}{1\frac{2m}{r}}+\text{g}^\gamma r^2(e^{2k_s}d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)\frac{1\frac{2m}{r}}{\text{g}^\gamma }dt^2$$
$$A_{03,z}=Q\rho \tau _{,z},A_{03,\overline{z}}=Q\rho \tau _{,\overline{z}},e^{2\alpha \varphi _0}=\frac{k_1^2}{(1\frac{2m}{r})\text{g}^\beta }$$
(3)
where a subindex $`0`$ stands for a seed solution and
$$\text{g}=a_1\tau +1,\text{and}e^{2k_s}=\left(1+\frac{m^2\mathrm{sin}^2\theta }{r^2(1\frac{2m}{r})}\right)^{1/\alpha ^2},$$
In this work we use the coordinates $`z=\rho +i\zeta =\sqrt{r^22mr}\mathrm{sin}\theta +i(rm)\mathrm{cos}\theta `$. A$`{}_{0}{}^{}=A_{0\mu }dx^\mu ,`$(with $`\mu =1\mathrm{}4),`$ is the electromagnetic four potential, $`m`$ the mass parameter, $`\gamma =2/(1+\alpha ^2)`$, $`\beta =2\alpha ^2/(1+\alpha ^2)`$; $`Q`$ and $`a_1`$ are constants related by
$$2\gamma a_1^2k_1^2Q^2=0$$
Solution (3) can be interpreted as a magnetized Schwarzschild solution in dilaton gravity for $`\alpha 0`$. For $`\alpha =0`$ the construction of dipoles is different and the form of the metric is not similar to the Schwarzschild solution any more . In what follows, we will assume $`\alpha 0`$. The function $`\tau =\tau (\rho ,\zeta )`$ is a harmonic parameter in a two dimensional flat space, $`i.e.`$, it is a solution of the Laplace equation
$$\frac{1}{2\rho }[(\rho \tau _{,z})_{,\overline{z}}+(\rho \tau _{,\overline{z}})_{,z}]=\tau _{\rho \rho }+\frac{1}{\rho }\tau _{,\rho }+\tau _{\zeta \zeta }=0.$$
(4)
This metric represents the exterior field of a gravitational object with an arbitrary magnetic field coupled to a scalar field. The metric is singular for $`r=2m`$ and for an interior radius determined by the magnetic field. For a pulsar with magnetic and scalar fields in the region where $`r>2m`$, this metric could be an static model of an astrophysical object with magnetic and scalar fields, metric (3) is always regular in that region.
The second metric we will deal with is given by
$$ds^2=\frac{1}{f_0}\left[e^{2k_0}\left(d\rho ^2+d\zeta ^2\right)+\rho ^2d\phi ^2\right]f_0dt^2,$$
(5)
where
$`f_0`$ $`=`$ $`{\displaystyle \frac{e^\lambda }{(a_1\mathrm{\Sigma }_1+a_2\mathrm{\Sigma }_2)^\gamma }},`$
$`e^{2\alpha \varphi _0}`$ $`=`$ $`\kappa _0^2=\kappa _1^2(a_1\mathrm{\Sigma }_1+a_2\mathrm{\Sigma }_2)^\beta e^{\lambda \tau _0\tau },`$
$`\mathrm{\hspace{0.17em}\; 2}A_{04}`$ $`=`$ $`\psi _0={\displaystyle \frac{a_3\mathrm{\Sigma }_1+a_4\mathrm{\Sigma }_2}{a_1\mathrm{\Sigma }_1+a_2\mathrm{\Sigma }_2}},`$ (6)
where $`a_1,\mathrm{},\kappa _1`$, and $`\tau _0`$ are constants and $`\beta ,\gamma `$ are again functions of $`\alpha `$ defined as $`\gamma =2/(1+\alpha ^2)`$, $`\beta =2\alpha ^2/(1+\alpha ^2)`$; $`\tau =\tau (\rho ,\zeta )`$ and $`\lambda =\lambda (\rho ,\zeta )`$ are harmonic functions which satisfy again the Laplace equation (observe that we have defined a new parameter $`\lambda `$ with respect to the one defined in ). $`\mathrm{\Sigma }_1`$ and $`\mathrm{\Sigma }_2`$ are functions given in terms of $`\tau `$ and the equation (6) contains two subclasses determined by the functions $`\mathrm{\Sigma }_1`$ and $`\mathrm{\Sigma }_2`$. For the first subclass we have $`\tau _0=0`$ and
$$\mathrm{\Sigma }_1=\tau \mathrm{\Sigma }_2=1$$
(7)
with the relation between the constants
$$4a_{1}^{}{}_{}{}^{2}\kappa _1^2(1+\alpha ^2)(a_1a_4a_2a_3)^2=0.$$
(8)
For the second subclass we have
$$\mathrm{\Sigma }_1=e^{q_1\tau }\mathrm{\Sigma }_2=e^{q_2\tau },$$
(9)
where $`q_1,q_2`$ are constants and $`\tau _0`$ satisfy the relation $`\tau _0=q_1+q_2`$; the condition for the constants in this case is given by:
$$4a_1a_2+\kappa _{1}^{}{}_{}{}^{2}(1+\alpha ^2)(a_1a_4a_2a_3)^2=0.$$
(10)
One of he most interesting solution contained in this class is the Gibbons-Maeda black-hole
$$ds^2=\text{g}^\gamma \frac{dr^2}{1\frac{2m}{r}}+\text{g}^\gamma r^2(e^{2k_s}d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)\frac{1\frac{2m}{r}}{\text{g}^\gamma }dt^2$$
(11)
where
$`e^{2\alpha \varphi _0}`$ $`=`$ $`\left(1+{\displaystyle \frac{r_{}}{r}}\right)^{\frac{2\alpha ^2}{1+\alpha ^2}}`$
g $`=`$ $`\left(1+{\displaystyle \frac{r_{}}{r}}\right);`$
$`r_{}r_+`$ $`=`$ $`2m`$
(we have used $`rr+r_{}`$ from the original solution). This metric could represent a static charged black hole containing a scalar field $`\varphi _0`$ (for a study of this metric see ). Observe that in both metrics (3) and (11), the space-time is qualitatively different only for $`\alpha =0`$, but for $`\alpha 0`$ the qualitative behavior is very similar for any $`\alpha `$ and many of the main features of the metrics can be obtained for a specific $`\alpha `$.
On the other hand, real astrophysical objects rotate. For an object like a pulsar, taking into account the rotation is very important in order to understand its space-time configuration. Therefore, if we want to model an astrophysical object with scalar field, we must find the corresponding rotating metrics of (3) and (11) to obtain the solution which we want to use for modeling them.
Using the potential space formalism for 5D gravity introduced by Neugebauer , we were able to find a set of formulas valid for $`\alpha ^2=3`$ in order to obtain rotating exact solutions from a static one, and without making any complicated integration. We will introduce this formalism in section two.
In section three we start from the axial-symmetric stationary field equations derived from equation (1) for the specific case $`\alpha ^2=3`$. Using the formalism mentioned above, we will derive new solutions in section four, and finally show that the Kerr space-time, the Gibbons-Maeda , the Frolov-Zelnikov and also their NUT generalizations are special cases of one of these new solutions.
## 2 The Potential-Space Formalism
For $`\alpha ^2=3`$ the field equations can be derived from a five-dimensional space-time action. We will deal with a five-dimensional space-time possessing a Killing vector field $`X`$ with close orbits. We will work with stationary space-times, this symmetry implies the existence of a second Killing vector field $`Y`$ with close orbits as well. Thus, we start with a five-dimensional space-times possessing two commuting Killing vectors fields; a space-like one $`X`$, representing the inner symmetry and a time-like one $`Y`$, representing stationarity. The potential formalism consists in defining covariantly five potentials in terms of the Killing vectors $`X`$ and $`Y`$. The five potentials are given by
$`I^2`$ $`=`$ $`\kappa ^{4/3}=X_AX^A;f=IY_AY^A+I^1(X^AY_A)^2`$
$`\psi `$ $`=`$ $`I^2X_AY^A;ϵ_{,A}=ϵ_{ABCDE}X^BY^BX^{D;E}`$
$`\chi _{,\alpha }`$ $`=`$ $`ϵ_{ABCDE}X^BY^BX^{D;E}`$ (12)
($`A,B,..=1\mathrm{}5`$), where $`f,ϵ,\psi ,\chi `$ and $`\kappa `$ respectively are the gravitational, rotational, electrostatic, magnetostatic and scalar potentials; $`ϵ_{ABCDE}`$ is the five-dimensional Levi-Civita pseudo-tensor, $`X=X^A/x^A=/x^5,`$ and $`Y=Y^A/x^A=/t`$. We will work with spaces possessing axial symmetry as well, which is a realistic assumption for a star. Thus for the axial symmetric stationary case we have another Killing vector $`Z=Z^A/x^A=/\phi `$, representing this symmetry. The field equations (2) in terms of the five potentials $`\mathrm{\Psi }^A=(f,ϵ,\psi ,\chi ,\kappa )`$ read
$`\widehat{D}^2\kappa +\left({\displaystyle \frac{\widehat{D}\rho }{\rho }}{\displaystyle \frac{\widehat{D}\kappa }{\kappa }}\right)\widehat{D}\kappa +{\displaystyle \frac{3\kappa ^3}{4f}}(\widehat{D}\psi ^2{\displaystyle \frac{1}{\kappa ^4}}\widehat{D}\chi ^2)`$ $`=`$ $`0,`$
$`\widehat{D}^2\psi +\left({\displaystyle \frac{\widehat{D}\rho }{\rho }}+{\displaystyle \frac{2\widehat{D}\kappa }{\kappa }}{\displaystyle \frac{\widehat{D}f}{f}}\right)\widehat{D}\psi {\displaystyle \frac{1}{\kappa ^2f}}(\widehat{D}ϵ\psi \widehat{D}\chi )\widehat{D}\chi `$ $`=`$ $`0,`$
$`\widehat{D}^2\chi +\left({\displaystyle \frac{\widehat{D}\rho }{\rho }}{\displaystyle \frac{2\widehat{D}\kappa }{\kappa }}{\displaystyle \frac{\widehat{D}f}{f}}\right)\widehat{D}\chi +{\displaystyle \frac{\kappa ^2}{f}}(\widehat{D}ϵ\psi \widehat{D}\chi )\widehat{D}\psi `$ $`=`$ $`0,`$
$`\widehat{D}^2f+\left({\displaystyle \frac{\widehat{D}\rho }{\rho }}{\displaystyle \frac{\widehat{D}f}{f}}\right)\widehat{D}f+{\displaystyle \frac{1}{f}}(\widehat{D}ϵ\psi \widehat{D}\chi )^2{\displaystyle \frac{\kappa ^2}{2}}\left(\widehat{D}\psi ^2+{\displaystyle \frac{1}{\kappa ^4}}\widehat{D}\chi ^2\right)`$ $`=`$ $`0,`$
$`\widehat{D}^2ϵ\widehat{D}\psi \widehat{D}\chi \psi \widehat{D}^2\chi +\left({\displaystyle \frac{\widehat{D}\rho }{\rho }}{\displaystyle \frac{2\widehat{D}f}{f}}\right)(\widehat{D}ϵ\psi \widehat{D}\chi )`$ $`=`$ $`0`$ (13)
where $`\widehat{D}`$ is the differential operator $`\widehat{D}=(_\rho ,_\zeta )`$. The field equations (13) can be derived from the Lagrangian
$$=\frac{\rho }{2f^2}[f_{,i}f^{,i}+(ϵ_{,i}\psi \chi _{,i})(ϵ^{,i}\psi \chi ^{,i})]+\frac{\rho }{2f}\left(\kappa ^2\psi _{,i}\psi ^{,i}+\frac{1}{\kappa ^2}\chi _{,i}\chi ^{,i}\right)\frac{2\rho }{3\kappa ^2}\kappa _{,i}\kappa ^{,i},$$
(14)
with $`i=(\rho ,\zeta )`$. The next step is to look for the invariant transformations of Lagrangian (14), which were found in . The invariance group of the Lagrangian (14) is $`SL(3,\text{I}\text{R})`$. We can write these transformations in a very simple form as
$$hCh_0C^T,$$
(15)
where $`h`$ and $`C`$ are elements of $`SL(3,\text{I}\text{R})`$. One parameterization of matrix $`h`$ is given by
$$h=\frac{1}{f\kappa ^{2/3}}\left(\begin{array}{ccc}f^2+ϵ^2f\kappa ^2\psi ^2& ϵ& ϵ\chi +f\kappa ^2\psi \\ ϵ& 1& \chi \\ ϵ\chi +f\kappa ^2\psi & \chi & \chi ^2\kappa ^2f\end{array}\right)$$
(16)
In terms of matrix $`h,`$ it is possible to write down the field equations (13) in a non-linear $`\sigma `$-model form
$$(\rho h_{,z}h^1)_{,\overline{z}}+(\rho h_{,\overline{z}}h^1)_{,z}=0.$$
(17)
Thus, we can define an abstract Riemannian space using the standard metric of the group defined by
$`ds^2`$ $`=`$ $`{\displaystyle \frac{1}{4}}tr(dhdh^1)`$ (18)
$`=`$ $`{\displaystyle \frac{\rho }{2f^2}}[df^2+(dϵ\psi d\chi )^2]{\displaystyle \frac{\rho \kappa ^2}{2f}}(d\psi ^2+{\displaystyle \frac{1}{\kappa ^4}}d\chi ^2)+{\displaystyle \frac{2\rho }{3\kappa ^2}}d\kappa ^2`$
This Riemannian space defines a five-dimensional symmetric space (the covariant derivative of the Riemannian tensor vanishes), with a isometry group $`SL(3,\text{I}\text{R})`$.
In what follows we will write explicitly the potentials $`\mathrm{\Psi }^A`$ in terms of the metric components. In order to do so, we recall that the five-dimensional space-time metric in terms of the four-dimensional one and the electromagnetic and scalar fields reads
$$ds_{5}^{}{}_{}{}^{2}=\widehat{g}_{\mu \nu }dx^\mu dx^\nu +I^2(A_\mu dx^\mu +dx^5)(A_\nu dx^\nu +dx^5),$$
(19)
where $`\widehat{g}_{\mu \nu }`$; $`\mu ,\nu =1,\mathrm{},4`$ are the 4-dimensional metric components of the five-dimensional space-time, $`I`$ is the scalar potential and $`A_\mu `$ is the electromagnetic four potential. For the axial symmetric stationary case $`I,A_\mu `$ and $`\widehat{g}_{\mu \nu }`$ depend only on $`\rho `$ and $`\zeta `$. The five-dimensional metric and its inverse can be written as
$$\widehat{g}_{AB}=\left(\begin{array}{cc}\widehat{g}_{\mu \nu }+I^2A_\mu A_\nu & I^2A_\mu \\ I^2A_\mu & I^2\end{array}\right)$$
(20)
$$\widehat{g}^{AB}=\left(\begin{array}{cc}\widehat{g}^{\nu \tau }& A^\nu \\ A^\nu & A^2+I^3\end{array}\right)$$
(21)
Due to the symmetries we are working with, its is convenient to write the four-dimensional metric in the Papapetrou parameterization
$$ds_{4}^{}{}_{}{}^{2}=g_{\mu \nu }dx^\mu dx^\nu =\frac{1}{f}(e^{2k}dzd\overline{z}+\rho ^2d\phi ^2)f(\omega d\phi +dt)^2$$
(22)
In terms of this parameterization, and recalling that $`ds_{5}^{}{}_{}{}^{2}=\widehat{g}_{AB}dx^Adx^B=\frac{1}{I}ds_{4}^{}{}_{}{}^{2}+I^2(A_\mu dx^\mu +dx^5)^2,`$ the metric coefficients $`\widehat{g}_{AB}`$ can be written as
$$\widehat{g}_{AB}=\left(\begin{array}{ccccc}0& \frac{1}{2If}e^{2k}& 0& 0& 0\\ \frac{1}{2If}e^{2k}& 0& 0& 0& 0\\ 0& 0& \frac{\rho ^2}{If}\frac{f\omega ^2}{I}+I^2A_{3}^{}{}_{}{}^{2}& \frac{f\omega }{I}+I^2A_3A_4& I^2A_3\\ 0& 0& \frac{f\omega }{I}+I^2A_3A_4& \frac{f}{I}+I^2A_{4}^{}{}_{}{}^{2}& I^2A_4\\ 0& 0& I^2A_3& I^2A_4& I^2\end{array}\right)$$
(23)
Using the expressions for the metric given in equation (22), it is straightforward to calculate the gravitational, electrostatic and scalar potentials. Recalling that the Killing vectors components $`X^A`$ and $`Y^A`$ satisfy the relations $`X^A=\delta _{}^{A}{}_{5}{}^{}`$ and $`Y^A=\delta _{}^{A}{}_{4}{}^{}`$, one finds that $`Y^AY_A=\widehat{g}_{44}`$; and $`X^AY_A=\widehat{g}_{54}.`$ Now, substituting these relations into the definition for $`f=I\widehat{g}_{44}+I^1(\widehat{g}_{54})^2`$ and using the relations (23) we obtain
$$f=Ig_{44}$$
(24)
In similar way one obtains for the electrostatic and the scalar potentials:
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`\widehat{g}_{55}=I^2,`$
$`\psi `$ $`=`$ $`A_4.`$ (25)
For the magnetostatic and rotational potentials the corresponding expressions can be reduced to
$`ϵ_{,\mu }`$ $`=`$ $`ϵ_{45\gamma \delta \mu }\widehat{g}^{\tau \delta }\widehat{g}^{\gamma \theta }\widehat{g}_{4\theta ,\tau },`$
$`\chi _{,\mu }`$ $`=`$ $`ϵ_{54\gamma \delta \mu }\widehat{g}^{\tau \delta }\widehat{g}^{\gamma \theta }\widehat{g}_{5\theta ,\tau }.`$ (26)
Using now the relations (20), and (12) in (26), we find:
$`ϵ_{,\overline{z}}`$ $`=`$ $`{\displaystyle \frac{I^2}{\rho }}[(g_{34}g_{44,\overline{z}}g_{44}g_{34,\overline{z}})]+\psi \chi _{,\overline{z}},`$
$`ϵ_{,z}`$ $`=`$ $`{\displaystyle \frac{I^2}{\rho }}[(g_{34}g_{44,z}g_{44}g_{34,z})]+\psi \chi _z,`$
$`\chi _{,\overline{z}}`$ $`=`$ $`{\displaystyle \frac{I^4}{\rho }}(g_{34}A_{4,\overline{z}}g_{44}A_{3,\overline{z}}),`$
$`\chi _{,z}`$ $`=`$ $`{\displaystyle \frac{I^4}{\rho }}(g_{34}A_{4,z}g_{44}A_{3,z}).`$
The potentials are written in terms of $`g_{34}`$ and $`g_{44}`$ components of the four-dimensional metric tensor as well as of the $`A_3`$ and $`A_4`$ components of the electromagnetic four potential. From (26) and (23) we arrive at the final expressions
$`A_{3,z}`$ $`=`$ $`{\displaystyle \frac{\rho }{f\kappa ^2}}\chi _{,z}+{\displaystyle \frac{g_{34}I}{f}}\psi _{,z}`$
$`A_{3,\overline{z}}`$ $`=`$ $`{\displaystyle \frac{\rho }{f\kappa ^2}}\chi _{,\overline{z}}+{\displaystyle \frac{g_{34}I}{f}}\psi _{\overline{z}}`$
$`\left({\displaystyle \frac{g_{34}}{g_{44}}}\right)_{,z}`$ $`=`$ $`{\displaystyle \frac{ϵ_{,z}\psi \chi _{,z}}{f^2}}`$ (27)
In the following sections we will use these expressions for obtaining exact solutions of the field equations.
## 3 Calculations and Solutions
In this section we will apply the previous results for finding exact solutions of the field equations. Let us consider $`f_0,I_0,\psi _0,\mathrm{},`$ etc. as seed solutions, $`i.e.`$, as components of the matrix $`h_0`$ in (15). We proceed as follows. First, using the inverse matrix $`h^1,`$ and the $`SL(3,\text{I}\text{R})`$ invariance of the field equations, from (15) we obtain the $`h`$ components in terms of the $`\mathrm{\Psi }_0^A`$ potentials. Finally, using a particular matrix $`C`$ in (15) we integrate the new potentials $`\mathrm{\Psi }^A`$ in general for this particular cases.
The inverse matrix of (16) reads
$$h^1=\frac{\kappa ^{2/3}}{f}\left(\begin{array}{ccc}1& ϵ\chi \psi & \psi \\ ϵ\chi \psi & f^2+(ϵ\chi \psi )^2f\chi ^2\kappa ^2& \kappa ^2f\chi +\psi (ϵ\chi \psi )\\ \psi & \chi f\kappa ^2+\psi (ϵ\chi \psi )& \kappa ^2f+\psi ^2\end{array}\right).$$
(28)
Then, we can write the potentials $`\mathrm{\Psi }^A`$ in terms of the components of matrices $`h`$ and $`h^1`$
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`{\displaystyle \frac{h_{11}^{}{}_{}{}^{1}}{h_{22}}};f^2={\displaystyle \frac{1}{h_{11}^{}{}_{}{}^{1}h_{22}}};\chi ={\displaystyle \frac{h_{23}}{h_{22}}}`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{h_{13}^{}{}_{}{}^{1}}{h_{11}^{}{}_{}{}^{1}}};ϵ={\displaystyle \frac{h_{12}}{h_{22}}}`$ (29)
where we have used the notation $`h_{ij}^{}{}_{}{}^{1}=(h^1)_{ij}`$. Using expressions (29) we can straightforwardly calculate the potentials $`\mathrm{\Psi }^A`$ from the $`h`$ components. We will take each case separately.
### 3.1 Case $`\mathrm{\Psi }_{0}^{}{}_{}{}^{A}=(f_0,0,\psi _0,0,\kappa _0)`$
In this case we start from a solution with electrostatic, scalar and gravitational potentials, for this case matrix $`h_0`$ reads
$$h_0=\frac{1}{\kappa _{0}^{}{}_{}{}^{2/3}f_0}\left(\begin{array}{ccc}f_{0}^{}{}_{}{}^{2}f_0\kappa _{0}^{}{}_{}{}^{2}\psi _0& 0& f_0\kappa _{0}^{}{}_{}{}^{2}\psi _0\\ 0& 1& 0\\ f_0\kappa _{0}^{}{}_{}{}^{2}\chi _0& 0& \kappa _0^2f_0\end{array}\right)$$
(30)
The inverse matrix is given by
$$h_{0}^{}{}_{}{}^{1}=\frac{\kappa _{0}^{}{}_{}{}^{2/3}}{f_0}\left(\begin{array}{ccc}1& 0& \psi _0\\ 0& f_0^2& 0\\ \psi _0& 0& \kappa _0^2f_0+\psi _{0}^{}{}_{}{}^{2}\end{array}\right)$$
(31)
Now we take the invariance equation $`h=ChC^T`$ taking the constant matrix $`C`$ arbitrary as
$$C=\left(\begin{array}{ccc}a& b& c\\ d& e& j\\ i& h& k\end{array}\right)$$
(32)
and its inverse as
$$C^1=\left(\begin{array}{ccc}q& p& t\\ u& v& w\\ s& y& z\end{array}\right).$$
(33)
Substituting it into (29) we arrive at
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`{\displaystyle \frac{U}{V}}`$
$`f^2`$ $`=`$ $`{\displaystyle \frac{f_{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{\frac{8}{3}}}{UV}}`$
$`\chi `$ $`=`$ $`{\displaystyle \frac{idf_{0}^{}{}_{}{}^{2}(idij)f_0\kappa _{0}^{}{}_{}{}^{2}\psi _{0}^{}{}_{}{}^{2}+(dk\psi _0kj)f_0\kappa _{0}^{}{}_{}{}^{2}+eh}{V\kappa _0^{\frac{2}{3}}}}`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{\kappa _{0}^{}{}_{}{}^{2}[tq(ts+zq)\psi _0wuf_{0}^{}{}_{}{}^{2}sz\psi _{0}^{}{}_{}{}^{2}]+szf_0}{U}}`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{f_0[daf_0+\kappa _{0}^{}{}_{}{}^{2}\psi _{0}^{}{}_{}{}^{2}(dcda)+\kappa _{0}^{}{}_{}{}^{2}(ja\psi _0jc)]+be}{V\kappa _0^{\frac{2}{3}}}}`$ (34)
with $`U=\kappa _{0}^{}{}_{}{}^{2}(q^2+2qs\psi _0+u^2f_{0}^{}{}_{}{}^{2}+s^2\psi _{0}^{}{}_{}{}^{2})s^2f_0`$ and $`V=\kappa _{0}^{}{}_{}{}^{\frac{2}{3}}[df_{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{2}(f_0\psi _{0}^{}{}_{}{}^{2}2djf_0\psi _0+j^2f_0)+e^2]`$. In order to perform a total integration of the metric components, we take the matrix $`C`$ as
$$C=\left(\begin{array}{ccc}1& 0& 0\\ 0& v& w\\ 0& w& v\end{array}\right)$$
(35)
with inverse
$$C^1=\left(\begin{array}{ccc}1& 0& 0\\ 0& v& w\\ 0& w& v\end{array}\right)$$
(36)
Then equations (34) reduce to the simple expressions
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`{\displaystyle \frac{\kappa _{0}^{}{}_{}{}^{\frac{4}{3}}}{v^2w^2\kappa _{0}^{}{}_{}{}^{2}f_0}}`$
$`f^2`$ $`=`$ $`{\displaystyle \frac{f_{0}^{}{}_{}{}^{2}}{v^2w^2\kappa _{0}^{}{}_{}{}^{2}f_0}}`$
$`\chi `$ $`=`$ $`{\displaystyle \frac{vw(1f_0\kappa _{0}^{}{}_{}{}^{2})}{v^2w^2\kappa _{0}^{}{}_{}{}^{2}f_0}}`$
$`\psi `$ $`=`$ $`v\psi _0`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{wf_0\psi _0\kappa _{0}^{}{}_{}{}^{2}}{v^2w^2\kappa _{0}^{}{}_{}{}^{2}f_0}}`$ (37)
keeping in mind that matrix $`C`$ fulfills the condition$`detC=1,`$ $`i.e.`$ $`v^2w^2=1`$, from equations (37), we obtain the following relation:
$$\frac{ϵ_{,l}\psi \chi _{,l}}{f^2}=\frac{w\kappa _{0}^{}{}_{}{}^{2}\psi _{0,l}}{f_0},$$
(38)
with $`l=z,\overline{z}`$. Then, from these last expressions (38), and from the last pair of equations (27), using equation (37), we have that
$$\left(\frac{g_{34}I}{f}\right)_{,z}=\frac{w\rho \kappa _{0}^{}{}_{}{}^{2}\psi _{0,z}}{f_0};\left(\frac{g_{34}I}{f}\right)_{,\overline{z}}=\frac{w\rho \kappa _{0}^{}{}_{}{}^{2}\psi _{0,\overline{z}}}{f_0}.$$
(39)
We start using the solution (6) and (7), with $`\alpha ^2=3`$ as seed solution and substituting it into equations (37) to obtain
$`f^2`$ $`=`$ $`{\displaystyle \frac{e^{2\lambda }}{(a_1\tau +a_2)^{1/2}(v^2w^2k_{1}^{}{}_{}{}^{2}(a_1\tau +a_2)e^{2\lambda })}}`$
$`\chi `$ $`=`$ $`{\displaystyle \frac{vw(1k_{1}^{}{}_{}{}^{2}(a_1\tau +a_2)e^{2\lambda })}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1\tau +a_2)e^{2\lambda }}}`$
$`\psi `$ $`=`$ $`v{\displaystyle \frac{a_3\tau +a_4}{a_1\tau +a_2}}`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{wk_{1}^{}{}_{}{}^{2}(a_3\tau +a_4)}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1\tau +a_2)e^{2\lambda }}}`$
$`\kappa ^{4/3}`$ $`=`$ $`\kappa _{0}^{}{}_{}{}^{4/3}{\displaystyle \frac{(a_1\tau +a_2)e^{2/3\lambda })}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1\tau +a_2)e^{2\lambda }}}`$ (40)
and substituting this seed solution (6) and (7) together with the restriction (8) into expressions (39), we arrive at:
$$\left(\frac{g_{34}I}{f}\right)_{,z}=wa_1k_1\rho \tau _{,z};\left(\frac{g_{34}I}{f}\right)_{,\overline{z}}=wa_1k_1\rho \tau _{,\overline{z}}.$$
(41)
The integrability of the right hand side of expression (41) is guaranteed because $`\tau `$ is harmonic and fulfills the Laplace equation (4). The explicitly form of the function $`g_{34}`$ depends on $`\tau `$. In is presented a list of expressions of the rhs of (39) for different $`\tau .`$ In terms of $`g_{34}`$ and the solution given in (40), we can write the final metric (19) as
$`ds_{5}^{}{}_{}{}^{2}`$ $`=`$ $`{\displaystyle \frac{B}{\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}}}e^{2\kappa \frac{4\lambda }{3}}dzd\overline{z}+\left(\rho ^2e^\lambda m_{1}^{}{}_{}{}^{\frac{1}{2}}{\displaystyle \frac{\kappa _1^{\frac{2}{3}}m_1g_{34}^{}{}_{}{}^{2}}{e^{\frac{2\lambda }{3}}}}\right){\displaystyle \frac{e^{\frac{4\lambda }{3}}}{\kappa _{1}^{}{}_{}{}^{\frac{4}{3}}m_{1}^{}{}_{}{}^{2}}}d\phi ^2+2g_{34}d\phi dt`$ (42)
$``$ $`{\displaystyle \frac{e^{\frac{2\lambda }{3}}}{\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}m_1}}dt^2+\kappa _1m_1^{\frac{1}{2}}e^{\frac{\lambda }{3}}(A_3d\phi v{\displaystyle \frac{a_3\tau +a_4}{m_1}}dt+dx^5)^2`$
with
$$A_{3,z}=\frac{\rho }{w\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}e^{\frac{4\lambda }{3}}}(\mathrm{ln}B)_{,z}+\frac{ve^{\frac{2\lambda }{3}}}{\kappa ^{\frac{5}{3}}m_1}g_{34}\tau _{,z}$$
(43)
and $`B=v^2w^2e^{2\lambda }\kappa _1m_1`$, $`m_1=a_1\tau +a_2.`$ That is, for a given $`\tau ,`$ we are able to obtain a rotating exact solution, with scalar, magnetic and gravitational fields using formulas (39) and (43).
For the second subclass we use the solution (6) with (10) with $`\alpha ^2=3`$ into (37) to obtain
$`f^2`$ $`=`$ $`{\displaystyle \frac{e^{2\lambda }}{(a_1e^{q_1\tau }+a_2e^{q_2\tau })(v^2w^2k_{1}^{}{}_{}{}^{2}(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2\lambda +\tau _0\tau })}}`$
$`\chi `$ $`=`$ $`vw{\displaystyle \frac{1k_{1}^{}{}_{}{}^{2}(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2\lambda +\tau _0\tau }}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2\lambda +\tau _0\tau }}}`$
$`\psi `$ $`=`$ $`v{\displaystyle \frac{a_3e^{q_1\tau }+a_4e^{q_2\tau }}{a_1e^{q_1\tau }+a_2e^{q_2\tau }}}`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{wk_1^2(a_3e^{q_1\tau }+a_4e^{q_2\tau })e^{2\lambda +\tau _0\tau }}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2\lambda +\tau _0\tau }}}`$
$`\kappa ^{4/3}`$ $`=`$ $`\kappa _0^{4/3}{\displaystyle \frac{(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2/3(\lambda +\tau _0\tau )}}{v^2w^2k_{1}^{}{}_{}{}^{2}(a_1e^{q_1\tau }+a_2e^{q_2\tau })e^{2\lambda +\tau _0\tau }}}`$ (44)
In this case, relations (39) with the seed solution (6) with (10) and the constraints among the constant, equation(9), give us
$`\left({\displaystyle \frac{g_{34}I}{f}}\right)_{,z}`$ $`=`$ $`{\displaystyle \frac{w(q_1q_2)a_1\kappa _1}{\sqrt{a_1a_2}}}\rho \tau _{,z};`$
$`\left({\displaystyle \frac{g_{34}I}{f}}\right)_{,\overline{z}}`$ $`=`$ $`{\displaystyle \frac{w(q_1q_2)a_1\kappa _1}{\sqrt{a_1a_2}}}\rho \tau _{,\overline{z}}.`$ (45)
As in the last case, we can now obtain the line element in terms of the function $`g_{34}`$, which can be integrated from these last pair of equations once a $`\tau `$ is given. The metric then reads
$`ds_{5}^{}{}_{}{}^{2}`$ $`=`$ $`{\displaystyle \frac{B}{\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}m_1^{\frac{1}{2}}}}e^{2\kappa (\frac{4\lambda }{3}+(q_1+q_2)\tau )}dzd\overline{z}+(\rho ^2e^\lambda m_{1}^{}{}_{}{}^{\frac{1}{2}}{\displaystyle \frac{\kappa _1^{\frac{2}{3}}m_1g_{34}^{}{}_{}{}^{2}}{e^{\frac{\lambda }{3}+(q_1+q_2)\tau }}})\times `$ (46)
$`{\displaystyle \frac{e^{2\frac{\lambda }{3}+2(q_1+q_2)\tau }}{\kappa _{1}^{}{}_{}{}^{\frac{4}{3}}m_{1}^{}{}_{}{}^{2}}}d\phi ^2+2g_{34}d\phi dt{\displaystyle \frac{e^{\frac{\lambda }{3}+(q_1+q_2)\tau }}{\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}m_1^{\frac{3}{2}}}}dt^2`$
$`+\kappa _{1}^{}{}_{}{}^{\frac{2}{3}}m_1e^{\frac{2\lambda }{3}}(A_3d\phi v{\displaystyle \frac{a_3e^{q_1\tau }+a_4e^{q_2\tau }}{m_1}}dt+dx^5)^2.`$
The function $`A_3`$ again depends on the harmonic function $`\tau `$ as
$$A_{3,z}=\frac{\rho }{\omega \kappa _1^{\frac{2}{3}}e^{\frac{4\lambda }{3}}}(\mathrm{ln}B)_{,z}+\frac{va_1(q_1q_2)e^{\frac{\lambda }{3}+2(q_1+q_2)\tau }}{\kappa ^{\frac{5}{3}}m_1}g_{34}\tau _{,z}$$
(47)
where $`B=v^2\omega ^2e^{2\lambda +(q_1+q_2)\tau }\kappa _1m_1`$ and $`m_1=a_1e^{q_1\tau }+a_2e^{q_2\tau }.`$ With solutions (42) and (46) we are now able to obtain rotating exact solutions which represent rotating monopoles, dipoles, etc. coupled to a dilaton field.
### 3.2 Case $`\mathrm{\Psi }_{0}^{}{}_{}{}^{A}=(f_0,0,0,\chi _0,\kappa _0)`$
We start from a static solution with magnetostatic, scalar and gravitational potentials. In this case, the matrix $`h_0`$ is
$$h_0=\frac{1}{\kappa _0^{2/3}f_0}\left(\begin{array}{ccc}f_0^2& 0& 0\\ 0& 1& \chi _0\\ 0& \chi _0& \chi _0^2\kappa _{0}^{}{}_{}{}^{2}f_0\end{array}\right)$$
(48)
with inverse
$$h_{0}^{}{}_{}{}^{1}=\frac{\kappa _0^{2/3}}{f_0}\left(\begin{array}{ccc}1& 0& 0\\ 0& f_0^2+\chi _0^2\kappa _0^2f_0+f_0^2& f_0\chi _0\kappa _0^2\\ 0& f_0\chi _0\kappa _0^2& f_0\kappa _0^2\end{array}\right).$$
(49)
Using the invariance equation (15) and substituting expressions (48) and (49) into the set given by equation (29), with the matrix $`C`$ given by equation (32), we obtain:
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`{\displaystyle \frac{1}{\kappa _0^{2/3}}}{\displaystyle \frac{B}{A}}`$
$`f^2`$ $`=`$ $`{\displaystyle \frac{f_{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{2}}{AB}}`$
$`\chi `$ $`=`$ $`{\displaystyle \frac{1}{A}}\{dif_{0}^{}{}_{}{}^{2}+e(h+k\chi _0)+j(h\chi _0+k(\chi _{0}^{}{}_{}{}^{2}f_0\kappa _{0}^{}{}_{}{}^{2}))\}`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{1}{B}}\{qt\kappa _{0}^{}{}_{}{}^{2}+uw(\chi _{0}^{}{}_{}{}^{2}f_0+\kappa _{0}^{}{}_{}{}^{2}f_{0}^{}{}_{}{}^{2})+\chi _0f_0(sw+uz)szf_0\}`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{1}{A}}\{adf_{0}^{}{}_{}{}^{2}+be+\chi _0(bj+ce)+cj(\chi _{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{2}f_0)\}`$ (50)
with $`A=d^2f_{0}^{}{}_{}{}^{2}+e(e+2j\chi _0)+j^2(\chi _{0}^{}{}_{}{}^{2}\kappa _0f_0)`$ and $`B=q^2\kappa _{0}^{}{}_{}{}^{2}+u^2(f_{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{2}f_0\chi _{0}^{}{}_{}{}^{2})+sf_0(2u\chi _0s).`$
Next, we take for the matrix $`C`$ the following particular form:
$$C=\left(\begin{array}{ccc}q& 0& s\\ 0& 1& 0\\ s& 0& q\end{array}\right),$$
(51)
and its inverse
$$C^1=\left(\begin{array}{ccc}q& 0& s\\ 0& 1& 0\\ s& 0& q\end{array}\right),$$
(52)
thus $`q^2s^2=1`$. With this particular form of $`C`$, the potentials reduce to the following expressions:
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`\kappa _0^{4/3}(q^2s^2f_0\kappa _{0}^{}{}_{}{}^{2})`$
$`f`$ $`=`$ $`{\displaystyle \frac{f_0}{\sqrt{q^2s^2f_0\kappa _{0}^{}{}_{}{}^{2}}}}`$
$`\chi `$ $`=`$ $`q\chi _0`$
$`ϵ`$ $`=`$ $`s\chi _0`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{sq(1f_0\kappa _{0}^{}{}_{}{}^{2})}{q^2s^2f_0\kappa _{0}^{}{}_{}{}^{2}}}.`$ (53)
Using the first differential equation of expressions (27) for $`A_3`$ in terms of the seed potentials, (for this case $`ϵ_0=\psi _0=0`$), we have that
$`A_{03,z}`$ $`=`$ $`{\displaystyle \frac{\rho }{f_0I_0^3}}\chi _{0,z}`$
$`A_{03,\overline{z}}`$ $`=`$ $`{\displaystyle \frac{\rho }{f_0I_0^3}}\chi _{0,\overline{z}}.`$ (54)
On the other hand, from the potentials given in (53), recalling that $`\kappa _{0}^{}{}_{}{}^{2}=I_{0}^{}{}_{}{}^{3}`$, we obtain that in this case:
$$\frac{ϵ_{,z}\psi \chi _{,z}}{f^2}=\frac{s}{f_0I_{0}^{}{}_{}{}^{3}}\chi _{0,z}.$$
(55)
Thus, using the last differential equation from (27) we find that
$$\frac{1}{\rho }\left(\frac{g_{34}}{g_{44}}\right)_{,z}=\frac{s}{f_0I_0^3}\chi _{0,z}.$$
(56)
From equation (54) and this last one, we obtain that
$$\left(\frac{g_{34}}{g_{44}}\right)_{,z}=sA_{03,z},$$
(57)
which implies that (up to a constant)
$$\left(\frac{g_{34}}{g_{44}}\right)=sA_{03},$$
(58)
that is
$$g_{34}=g_{44}sA_{03}=\frac{f}{I}sA_{03},$$
(59)
$`i.e.`$, for this case we do not need to perform any extra integration for generating a new rotating solution starting from the seed one. In this way, we finally obtain the following expression for the target metric
$`ds_5^2`$ $`=`$ $`{\displaystyle \frac{1}{I}}\{T^{\frac{1}{2}}{\displaystyle \frac{e^{k_0}}{f_0}}dzd\overline{z}+\left[T^{\frac{1}{2}}{\displaystyle \frac{\rho ^2}{f_0}}{\displaystyle \frac{s^2A_{03}^{}{}_{}{}^{2}f_0}{T^{\frac{1}{2}}}}\right]d\phi ^2{\displaystyle \frac{sA_{03}f_0}{T^{\frac{1}{2}}}}d\phi dt{\displaystyle \frac{f_0}{T^{\frac{1}{2}}}}dt^2\}`$ (60)
$`+I^2\left({\displaystyle \frac{qA_{03}}{T}}d\phi {\displaystyle \frac{qs(1f_0\kappa _{0}^{}{}_{}{}^{2})}{T}}dt+dx^5\right)^2`$
where $`T=q^2s^2f_0\kappa _{0}^{}{}_{}{}^{2}`$, $`A_3=qA_{03}/(q^2s^2f_0\kappa _o^2)`$ and $`A_4=qs(1s^2f_0\kappa _o^2)/(q^2s^2f_0\kappa _o^2)`$. Thus, we have generated again a new exact solution to the Einstein-Maxwell-dilaton theory, for $`\alpha ^2=3`$, in which all the fields are non trivially involved.
### 3.3 case $`\mathrm{\Psi }_{0}^{}{}_{}{}^{A}=(f_0,ϵ_0,0,0,\kappa _0)`$
This case was studied in , we presented it here somewhat more detailed in order to have all the cases together. In this case we take as initial solution one without electrostatic and magnetostatic fields. The matrix $`h_0`$ is
$$h_0=\frac{1}{\kappa ^{2/3}f_0}\left(\begin{array}{ccc}f_0^2+ϵ_0^2& ϵ_0& 0\\ ϵ_0& 1& 0\\ 0& 0& \kappa _{0}^{}{}_{}{}^{2}f_0\end{array}\right)$$
(61)
and its inverse is
$$h_{0}^{}{}_{}{}^{1}=\frac{\kappa ^{2/3}}{f_0}\left(\begin{array}{ccc}1& ϵ_0& 0\\ ϵ_0& f_0^2+ϵ_0^2& 0\\ 0& 0& \frac{f_0}{\kappa _0^2}\end{array}\right).$$
(62)
In a similar way as in the other cases presented in this work, we take the equations relating the components of the matrices given in equations (16) and (28), and using (32) and (33) in the invariance equation (15), we arrive at
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`{\displaystyle \frac{U}{\kappa _{0}^{}{}_{}{}^{\frac{2}{3}}V}}`$
$`f^2`$ $`=`$ $`{\displaystyle \frac{f_{0}^{}{}_{}{}^{2}\kappa _{0}^{}{}_{}{}^{2}}{UV}}`$
$`\chi `$ $`=`$ $`{\displaystyle \frac{idf_{0}^{}{}_{}{}^{2}+ϵ_0(idϵ_0iehd)+hej\kappa \kappa _0^2f_0}{V\kappa _0^{\frac{2}{V}}}}`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{\kappa _{0}^{}{}_{}{}^{2}[tq+wuf_{0}^{}{}_{}{}^{2}+ϵ_0(uwϵ_0+tu+qw)]szf_0}{U}}`$
$`ϵ`$ $`=`$ $`{\displaystyle \frac{f_0(daf_0cj\kappa _0^2)+ϵ_0(adϵ_0bdae)+be}{V}}`$
where $`U=\kappa _0(q^2+2quϵ_0+(uf_0)^2+(uϵ_0)^2)s^2f_0`$, and $`V=(df_0)^2+(dϵ_0)^22deϵ_0+e^2(j\kappa _0)^2f_0`$. In order to integrate the metric, it is again necessary to consider a simpler matrix $`C.`$ We take
$$C=\left(\begin{array}{ccc}q& 0& s\\ 0& 1& 0\\ s& 0& q\end{array}\right),$$
(63)
and its inverse
$$C^1=\left(\begin{array}{ccc}q& 0& s\\ 0& 1& 0\\ s& 0& q\end{array}\right),$$
(64)
then, recalling that again $`q^2s^2=1`$, the potentials read
$`\kappa ^{\frac{4}{3}}`$ $`=`$ $`B\kappa _{0}^{}{}_{}{}^{\frac{4}{3}}`$
$`f^2`$ $`=`$ $`f_{0}^{}{}_{}{}^{2}B^1`$
$`\chi `$ $`=`$ $`sϵ_0`$
$`\psi `$ $`=`$ $`{\displaystyle \frac{sq[1\kappa _0^2f_0]}{B}}`$
$`ϵ`$ $`=`$ $`qϵ_0`$ (65)
where $`B=q^2s^2\kappa _0^2f_0.`$ Integrating equation (27), and substituting in it expression (65), we obtain:
$$\frac{1}{\rho }\left(\frac{g_{34}I}{f}\right)_{,z}=\frac{q}{\rho }\left(\frac{g_{034}I_0}{f_0}\right)_{,z},$$
(66)
which implies that the expression $`\frac{1}{\rho }\left(\frac{g_{34}I}{f}\right)_{,z}`$ remains invariant (up to a constant) for the seed solution and the generated one, $`i.e.`$
$$g_{34}=\frac{q}{B}g_{034}.$$
(67)
Similarly we find that
$$A_3=s\frac{g_{034}I_0}{B},$$
(68)
$`i.e.`$, again, for this case we do not need to perform any extra integration for generating a new rotating solution starting from the seed one. Substituting the solution into the Papapetrou metric (19) we arrive at
$`ds_{5}^{}{}_{}{}^{2}`$ $`=`$ $`{\displaystyle \frac{1}{I_0f_0}}e^{2k}dzd\overline{z}B\left[g_{033}+{\displaystyle \frac{g_{034}^{}{}_{}{}^{2}I_0}{f_0}}\left(1{\displaystyle \frac{q^2}{B}}\right)\right]d\phi ^2+2{\displaystyle \frac{q}{B}}g_{034}d\phi dt`$
$`{\displaystyle \frac{f_0}{BI_0}}dt^2+I^2\left({\displaystyle \frac{sg_{034}I_0}{\kappa _0^2B}}d\phi {\displaystyle \frac{sq[1\kappa _0^2f_0]}{B}}dt+dX^5\right)^2`$
Here we must start from a static exact solution coupled to a scalar field. If there are no extra fields besides the scalar one, the Einstein equations decouple from the scalar one which satisfy a harmonic equation. The field equation for the scalar field can be integrated independently from the Einstein equations. As an exact solution for the scalar field equation we take the function $`\kappa _0=\left[(rm+\sigma )/(rm\sigma )\right]^\delta .`$ For the Einstein equations we take as seed metric the Kerr-NUT space-time, $`i.e.`$ as seed solution we have
$$\kappa _0=\left(\frac{rm+\sigma }{rm\sigma }\right)^\delta ;ϵ_0=\frac{2(\omega L_+lr)}{\omega };f_0=\frac{\omega 2mr2lL_+}{\omega },$$
(69)
with
$$L_+=a\mathrm{cos}\theta +l;L_{}=a\mathrm{cos}\theta l;\omega =r^2+(a\mathrm{cos}\theta +l)^2$$
(70)
where $`r`$ and $`\theta `$ are the Boyer-Lindsquit coordinates, $`\rho =\sqrt{r^2+2mr+a^2l^2}\mathrm{sin}\theta `$ and $`\zeta =(rm)\mathrm{cos}\theta `$; $`a`$, $`m`$ and $`l`$ respectively are the rotation, mass and NUT parameters, $`\sigma `$, and $`\delta `$ are integration constants. The resulting target solution is an exact axial symmetric stationary solution of $`5D`$ gravity, with electromagnetic and scalar fields. It reads
$`ds_{5}^{}{}_{}{}^{2}`$ $`=`$ $`{\displaystyle \frac{\omega }{\omega 2mr2lL_+}}\left({\displaystyle \frac{rm+\sigma }{rm\sigma }}\right)^{\frac{2\delta }{3}}(r^22mr+L_+L_{})e^{2k_s}\left({\displaystyle \frac{dr^2}{\mathrm{\Delta }}}+d\theta ^2\right)`$ (71)
$`+{\displaystyle \frac{1}{D}}\left({\displaystyle \frac{rm+\sigma }{rm\sigma }}\right)^{\frac{4\delta }{3}}\{(\omega 2mr2lL_+)dt^2`$
$`(4qa(mr+l)\mathrm{sin}^2\theta 4l\mathrm{cos}\theta \mathrm{\Delta }){\displaystyle \frac{\omega 2mr2lL_+}{r^22mr+L_{}L_+}}dtd\phi `$
$`+[{\displaystyle \frac{\omega }{\omega 2mr2lL_+}}\mathrm{\Delta }\mathrm{sin}^2\theta D\left({\displaystyle \frac{rm+\sigma }{rm\sigma }}\right)^{2\delta }`$
$`q^2(\omega 2mr2lL_+)\left({\displaystyle \frac{2a\mathrm{sin}^2\theta (mr+l)+2l\mathrm{cos}\theta \mathrm{\Delta }}{r^22mr+L_+L_{}}}\right)^2]d\phi ^2\}`$
$`+\left({\displaystyle \frac{rm+\sigma }{rm\sigma }}\right)^{\frac{2\delta }{3}}{\displaystyle \frac{D}{\omega }}(A_3d\phi +A_4dt+dX^5)^2`$
where
$$A_3=\left(\frac{rm+\sigma }{rm\sigma }\right)^{2\delta }\left(\frac{2a(mr+l)\mathrm{sin}^2\theta +2l\mathrm{\Delta }\mathrm{cos}\theta }{r^22mr+L_+L_{}}\right)\frac{\omega 2mr+2lL_+}{D}$$
(72)
$$A_4=qs\frac{\omega (\frac{rm+\sigma }{rm\sigma })^{2\delta }(\omega 2mr+2lL_+)}{D}$$
(73)
$`D`$ $`=`$ $`\omega \left({\displaystyle \frac{rm+\sigma }{rm\sigma }}\right)^{2\delta }q^2s^2(\omega 2mr+2lL_+);`$
$`\mathrm{\Delta }`$ $`=`$ $`r^22mr+a^2l^2`$
$$e^{2k_s}=\left[\frac{(\sqrt{\rho ^2+(\zeta m)^2}+\sqrt{\rho ^2+(\zeta +m)^2})^24m^2}{4\sqrt{(\rho ^2+(\zeta m)^2)(\rho ^2+(\zeta +m)^2)}}\right]^{\frac{8}{3}\delta ^2}$$
(74)
Exact solution (71) was first obtained in (see also ) and contains a great amount of well-known solutions of $`5D`$ gravity. We will list three of the most important ones. In order to do so, we start setting $`\delta =l=0`$ in (71), obtaining for the 4-dim space-time (see )
$`ds_4^2`$ $`=`$ $`{\displaystyle \frac{\sqrt{D\omega }}{\mathrm{\Delta }}}dr^2+\sqrt{D\omega }d\theta ^2{\displaystyle \frac{\mathrm{\Delta }a^2\mathrm{sin}\theta }{\sqrt{D\omega }}}dt^2{\displaystyle \frac{4qamr\mathrm{sin}^2\theta }{\sqrt{D\omega }}}dtd\phi `$ (75)
$`+{\displaystyle \frac{\mathrm{sin}^2\theta }{\sqrt{D\omega }(\mathrm{\Delta }a^2\mathrm{sin}^2\theta )}}[\mathrm{\Delta }(D\omega )4q^2a^2m^2r^2\mathrm{sin}^2\theta ]d\phi ^2`$
where $`D=\omega q^2s^2(\omega 2mr);`$ $`\mathrm{\Delta }=r^22mr+a^2;`$ $`\omega =r^2+a^2\mathrm{cos}^2\theta .`$ For the scalar field the solution reduces to
$$\kappa ^{\frac{4}{3}}=q^2s^2\left(1\frac{2mr}{r^2+a^2\mathrm{cos}^2\theta }\right).$$
The following known solutions are contained as particular cases of the generated target solution whose $`ds_4^2`$ part is given by equation (22):
Frolov-Zelnikov Solution
This solution is a charged rotating black hole, obtained by Frolov and Zelnikov in 1987 . The $`ds_4^2`$ part of the metric is given by
$`ds_4^2`$ $`=`$ $`{\displaystyle \frac{1z}{B}}dt^22a\mathrm{sin}^2\theta {\displaystyle \frac{1}{\sqrt{1v^2}}}{\displaystyle \frac{z}{B}}dtd\phi `$
$`+\left[B(r^2+a^2)+a^2\mathrm{sin}^2\theta {\displaystyle \frac{z}{B}}\right]\mathrm{sin}^2\theta d\phi ^2+B{\displaystyle \frac{\mathrm{\Sigma }}{\mathrm{\Delta }}}dr^2+B\mathrm{\Sigma }d\theta ^2,`$
where $`B=\sqrt{(1v^2+v^2z)/(1v^2)}`$, $`z=2mr/\mathrm{\Sigma }`$, and $`\mathrm{\Delta }=r^2+a^22mr.`$ Comparing it with equation (3.3), we find that this solution corresponds to $`\delta =l=0`$, $`a0`$; $`\mathrm{\Sigma }=\omega `$, $`D=B^2\omega `$ and $`q=1/\left(1v^2\right)`$.
Gibbons-Maeda-Horner-Horowitz Solution
This solution describes a dilatonic static charged black hole. If we set $`a=0`$ in (75) we get the metric
$$ds_4^2=\frac{A}{\mathrm{\Delta }}dr^2\frac{\mathrm{\Delta }}{A}dt^2+r\sqrt{D}(d\theta ^2+\mathrm{sin}^2\theta d\phi ^2)$$
(76)
where $`\omega =r^2;`$ $`D=\omega q^2s^2(r^22mr)`$; $`\mathrm{\Delta }=r^22mr`$; $`A=\sqrt{D\omega }.`$ Metric (76) can be rewritten as
$$ds^2=\frac{1}{f}dr^2fdt^2+r\sqrt{D}d\mathrm{\Omega }^2$$
(77)
where
$$f=\frac{1\frac{2m}{r}}{\sqrt{q^2s^2\frac{12m}{r}}}.$$
(78)
With the restriction $`q^2s^2=1,`$ the function $`f`$ transform into
$$f=\frac{1\frac{2m}{r}}{\sqrt{1+\frac{2m}{r}s^2}}$$
(79)
If we set $`r_+r_{}=2m;`$ $`r_{}=2ms^2;`$ $`r_+=2m(1s^2)`$ we get $`f=(1+(r_{}r_+)/r)/\sqrt{1+r_{}/r};`$ $`r\sqrt{D}=R^2=r^2\sqrt{1+r_{}/r}`$ which corresponds just to the Gibbons-Maeda-Horner-Horowitz solution . The charge and the mass parameters can be written as $`Q=ms\sqrt{1s^2};`$ $`M=m\frac{1}{2}ms.`$ From the case $`s=0`$ we find the Schwarzschild solution $`(Q=0`$ and $`M=m).`$
Finally, we want to stress the fact that with the procedure of solution generation presented in this work, the seed solution is also included within the target solution as a particular case, thus we have:
Kerr solution
To recover the Kerr solution setting $`\delta =l=0;`$ $`s=0`$, $`q=1`$ and $`a0`$ in (75), we get
$`ds_4^2`$ $`=`$ $`\left({\displaystyle \frac{\mathrm{\Delta }a^2\mathrm{sin}^2\theta }{\omega }}\right)dt^2{\displaystyle \frac{2a\mathrm{sin}^2\theta (r^2+a^2\mathrm{\Delta })}{\omega }}dtd\phi +`$
$`\mathrm{sin}^2\theta d\phi ^2+{\displaystyle \frac{\omega }{\mathrm{\Delta }}}dr^2+\omega d\phi ^2`$
with $`\mathrm{\Delta }=r^22mr+a^2`$ and $`\omega =r^2+a^2\mathrm{cos}^2\theta `$.
NUT parameter
We can also obtain the NUT solution taking $`a=0`$, $`\delta =0`$ and $`l0`$
$$ds_4^2=\frac{r^2+l^2}{\mathrm{\Delta }^2}dr^2+(r^2+l^2)d\theta ^2+(r^2+l^2)\mathrm{sin}^2\theta d\phi ^2\frac{\mathrm{\Delta }}{r^2+l^2}dt^2$$
$`\mathrm{\Delta }=r^22mrl^2`$ where $`l`$ and $`m`$ respectively are the NUT parameter and the mass.
## 4 Conclusions
In this work we have given a series of formulae to obtain rotating exact solutions of the Einstein-Maxwell-Dilaton field equations generated from seed static ones. The examples we gave for the application of these formulae, consist on start from a seed solutions in terms of harmonic maps, $`i.e.`$, in terms of two functions which fulfill the Laplace equation. The static seed solutions represent gravitational fields coupled to a scalar field and to a magnetostatic (electrostatic) monopoles, dipoles, quadrupoles, etc. The new solutions generated using our formulae represent the rotating version of the seed ones. The new solutions contain induced electric (magnetic) fields generated by the rotation of the body. Some of the seed solutions model the exterior field of a pulsar containing a scalar field in the slow rotation limit. The scalar fields could be fundamental or generated by spontaneous scalarization . The new rotating version of the solution generated using our formulae are in this sense more realistic and could represent the exterior field of a pulsar with fast rotation. We suggest that this solutions could be used as theoretical models for testing the strong gravitational regime of the Einstein theory or the most important generalizations of general relativity near a pulsar containing a scalar field. Because of the presence of the electromagnetic field, our solutions could give also a light to the understanding of such strong effects like the origin of jets and maybe of the origin of the QPOs, where approximated and numerical methods could be not completely trustable.
## 5 Acknowledgments
This work has been supported by DGAPA-UNAM, project IN105496, and by CONACYT, Mexico, project 3697E. TM thanks the hospitality from the relativity group in Jena, Germany, and the DAAD support while this work was partially done. |
warning/0001/hep-th0001060.html | ar5iv | text | # Non-perturbative scaling in the scalar theory
## Abstract
A new approach to study the scaling behavior of the scalar theory near the Gaussian fixed point in $`d`$-dimensions is presented. For a class of initial data an explicit use of the Green’s function of the evolution equation is made. It is thus discussed under which conditions non-polynomial relevant interactions can be generated by the renormalization group flow.
Recent studies have discussed the possibility that at the Gaussian Fixed Point (GFP) of the $`O(N)`$ invariant scalar theory new relevant eigen-directions exist. In particular in Halpern and Huang have found a class of non-polynomial relevant interactions at the GFP by means of Wilson’s renormalization group (RG) transformation with the “sharp” momentum shell integration, but their result is still much debated .
In fact an interesting question to study is the general features of the scaling around the GFP. In particular it would be important to have a global understanding of all the possible solutions in order to discuss the structure of the continuum limit.
A necessary condition to describe this limit is to select a set of marginal and relevant scaling eigen-operators at this fixed point. A scaling eigen-operator is a solution of the linearised fixed point equation of the form
$$u=\left(\frac{\mathrm{\Lambda }}{k}\right)^\nu h(x)$$
(1)
where $`x`$ is the dimensionless field, $`\mathrm{\Lambda }`$ is the UV overall cut-off, $`k`$ is the running cut-off and $`\nu `$ is a scaling exponent. While in perturbation theory only a finite number of polynomial relevant and marginal interactions are considered, it is still not clear if new non-perturbative non-polynomial relevant scaling interactions can be generated by the RG flow. Were this true, it would mean for instance that the polynomial interactions cannot span all the continuum physics.
The aim of this work is to clarify some aspects of this question. Instead of using the methods discussed in , the Green’s function of the linearised renormalization group equation is constructed. It will be thus shown the role played by the boundary conditions in determining the structure of the scaling fields and it will be shown that is possible to generate relevant non-polynomial interactions not belonging to the same universality class of the polynomial interactions.
The most convenient way to address such a question is to use the Wilsonian formulation of the RG transformation. In particular we consider the local potential approximation (LPA) of Wegner-Houghton equation as discussed in for the $`N=1`$ components scalar theory, but we shall not restrict the potential to a polynomial.
The flow equation for the scaling field $`u(x,t)`$ reads
$$\frac{u}{t}=du\frac{d2}{2}x\frac{u}{x}+a_d\frac{^2u}{x^2}$$
(2)
where $`d>2`$ is the dimension of the spacetime, $`\mathrm{\Lambda }`$ is the UV cut-off, $`a_d=1/(2\sqrt{\pi })^d\mathrm{\Gamma }(d/2)`$, and $`t\mathrm{ln}\mathrm{\Lambda }/k`$. The GFP is at $`u0`$. In studying the UV region near the GFP one is interested in finding the unique function $`u(x,t)`$ which is continuous in the closed upper half plane $`\mathrm{}<x<\mathrm{}`$, $`0t`$ and satisfies eq.(2) with initial condition $`u(x,0)=f(x)`$, $`\mathrm{}<x<\mathrm{}`$ with $`f(x)C^0`$. We shall leave open the possibility that $`f`$ is not bounded since there are no physical reasons to assume $`|u(\pm \mathrm{},t)|<\mathrm{}`$ in discussing the scaling interactions.
It is convenient to make the transformation
$$u(x,t)=v(x,t)e^{(3d2)t/2}$$
(3)
which brings (2), in the equivalent Fokker-Planck form
$$\frac{v}{t}=\frac{d2}{2}(v+x\frac{v}{x})+a_dv_{xx}$$
(4)
After the rescaling $`xx\sqrt{(d2)/2a_d}`$ and $`tt(d2)/2\tau `$, it reads
$`{\displaystyle \frac{v}{\tau }}=L_{\mathrm{FP}}[v],L_{\mathrm{FP}}{\displaystyle \frac{}{x}}\left[{\displaystyle \frac{}{x}}x\right]`$ (5)
where we have introduced the Fokker-Planck operator $`L_{\mathrm{FP}}`$. It is not difficult to build a Green’s function for such an operator. Let us define the formally self-adjont operator
$$L=e^{x^2/4}L_{\mathrm{FP}}e^{x^2/4}=\left[\frac{^2}{x^2}+\frac{x^2}{4}+\frac{1}{2}\right]H_{\mathrm{ho}}\frac{1}{2}$$
(6)
being $`H_{\mathrm{ho}}`$ the Hamiltonian of the quantum mechanical harmonic oscillator with $`\mathrm{}=\omega =1`$ and mass $`m=1/2`$. We shall refer to the description in terms of such an operator and eigenfunctions as the quantum-mechanical (QM) frame, as opposed to the Fokker-Planck (FP) frame in eq.(5). The eigenfunctions of $`L`$ have the asymptotic behavior $`e^{\pm x^2/4}`$ as $`|x|\mathrm{}`$. For $`L`$ to be self-adjoint only the damped exponential must be chosen, and the eigenfunctions are know from Quantum Mechanics. In particular the spectrum of $`L`$ satisfies $`L\phi _n=(n+1)\phi _n`$ with $`n=0,1,2,\mathrm{}`$ and all the eigenvalues of $`L`$ are eigenvalues of $`L_{\mathrm{FP}}`$ with eigenfunctions $`L_{\mathrm{FP}}\varphi _n=(n+1)\varphi _n=e^{x^2/4}\phi _n`$ Therefore a complete and orthonormal set of eigenfunctions for the $`L_{\mathrm{FP}}`$ operator reads
$$\varphi _n(x)=\frac{1}{{}_{}{}^{4}\sqrt{2\pi }\sqrt{2^nn!}}H_n(x/\sqrt{2})$$
(7)
where $`H_n`$ are the Hermite polynomials, with the orthonormality condition
$$\varphi _n|\varphi _m=_{\mathrm{}}^+\mathrm{}e^{x^2/2}\varphi _n\varphi _m𝑑x=\delta _{mn}$$
(8)
Thus the linear space generated by the set $`\varphi _n`$ is $`^2()`$ with the norm $`f=\sqrt{f|f}`$. It is now straightforward to write down the Green’s function $`G(x,\tau |y,0)`$ of the Fokker-Planck operator from the well known expression of the Green’s function of the harmonic oscillator. We have
$`G(x,\tau |y,0)=x|e^{\tau L_{\mathrm{FP}}}|y=e^{y^2/2}{\displaystyle \underset{n}{}}e^{\tau (n+1)}\varphi _n(x)\varphi _n(y)`$ (9)
$`=e^{y^2/2}e^{x^2/4}{\displaystyle \underset{n}{}}e^{\tau (n+1)}\phi _n(x)\phi _n(y)e^{y^2/4}`$ (12)
$`=e^{x^2/4}x|e^{\tau (H_{\mathrm{ho}}+\frac{1}{2})}|ye^{y^2/4\tau /2}`$
$`=\sqrt{{\displaystyle \frac{e^\tau }{4\pi \mathrm{sinh}\tau }}}\mathrm{exp}\left({\displaystyle \frac{e^\tau }{4\mathrm{sinh}\tau }}(xe^\tau y)^2\right)`$
This precisely what we find in with a summation formula of the Hermite polynomials. Therefore the unique solution of (5) which is continuous and satisfies the following initial conditions $`v(x,0)=f(x)`$ with $`|\varphi _n|f|<\mathrm{}`$ is given by
$$v(x,\tau )=_{\mathrm{}}^+\mathrm{}G(x,\tau ;y,0)f(y)𝑑y$$
(13)
We now come to the main questions of our investigation, namely to study the subspace of relevant and marginal interactions present at the fixed point. It is not difficult to see that the solutions of eq.(2) that $`^2()`$ which are not bounded as $`t+\mathrm{}`$ have $`n2d/(d2)`$ in the spectrum of $`L_{\mathrm{FP}}`$.
The proof is a consequence of the spectral properties of the flow evolution operator: given $`f^2()`$ as an initial condition, from (3) and (13) one has
$`u(x,t)=e^{(3d2)t/2}{\displaystyle _{\mathrm{}}^+\mathrm{}}G(x,\tau |y,0)f(y)`$ (14)
$`=e^{(3d2)t/2}{\displaystyle _{\mathrm{}}^+\mathrm{}}{\displaystyle \underset{n}{}}e^{\tau (n+1)}\varphi _n(x)\varphi _n(y)f(y)e^{\frac{y^2}{2}}dy`$ (15)
$`={\displaystyle \underset{n}{}}e^{(d(d2)n/2)t}\varphi _n(x)\varphi _n|f`$ (16)
where we have used eq.(12) and the uniform convergence of the sum. From the last line we see that in order to have a solution that grows for $`t>0`$ it must be $`n2d/(d2)`$. If we set $`f=\varphi _m`$ we recover the subspace of marginal and relevant eigenvectors spanned by the first $`n2d/(d2)`$ Hermite polynomials, as it has already found in the previous literature . Non-polynomial interactions may be evolved by means of the Green’s function provided $`|\varphi _n|f|<\mathrm{}`$, but these will not be of the form (1) and therefore cannot be considered as scaling fields. More specifically if we try to evolve an initial datum that behaves like $`\mathrm{exp}(qx^2)`$ for $`x\mathrm{}`$ with $`q<1/2`$, we find $`u(x,t)e^{(3d2)t/2}\mathrm{exp}(e^{(d2)t}qx^2)`$ as $`t\mathrm{}`$ (note that $`|\varphi _n|f|=\mathrm{}`$ if $`q1/2`$. ).
We want now to show that the space where all the possible physically meaningful scaling interactions live is greater than $`^2()`$. The key point to be noticed is that the QM frame does not reproduce all the possible solutions present in the FP frame. This is because $`L`$ is self-adjoint for the boundary conditions at infinity of the type $`\phi _ne^{x^2/4}`$. In this case from Sturm-Liouville theory we know that the spectrum is bounded and the number of relevant and marginal interactions is thus finite . However $`L_{\mathrm{FP}}`$ is not self-adjoint and its spectrum is greater than the self-adjoint extension of $`L`$. In fact, although there are no non-stationary solutions of the linearised equation in the QM frame, a non-trivial zero mode is present in the spectrum of the $`L_{\mathrm{FP}}`$ operator, namely
$$v_0=e^{x^2/2}$$
(17)
which of course $`^2()`$ and it corresponds to the following non-stationary solution of the original equation (2)
$$u(x,t)=e^{(3d2)t/2}e^{(d2)x^2/2a_d}$$
(18)
where we have inserted the factor $`\sqrt{(d2)/2a_d}`$ in the definition of $`x`$. This is a perfectly well defined and regular scaling field which is “relevant” because it grows in the IR, with scaling dimension $`\nu =(3d2)/2`$, and it does not belong to the subspace spanned by the $`\varphi _n`$ eigenfunctions previously constructed. It is therefore a non-perturbative scaling field (the connection of this solutions with the type discussed in is not immediately clear to me although their asymptotic behavior for large $`x`$ is similar to what discussed in .). In Quantum Mechanics or in Ornstein-Uhlenbeck diffusion processes one would discard such a solution because of the boundary conditions at infinity. In our context there is no specific reason for not considering scaling fields of this type.
Given this particular zero mode of the $`L_{\mathrm{FP}}`$ operator it is immediate to write down the general solution of the homogeneous equation $`L_{\mathrm{FP}}[v]=0`$ which is of the form
$$v(x,\tau )=e^{x^2/2}\left(A+B\mathrm{erf}(\frac{x}{\sqrt{2}})\right)$$
(19)
being $`A`$ and $`B`$ two arbitrary constants of integration. More generally if we set $`v_n(x,\tau )=e^{\lambda \tau }h_\lambda (x)`$ eq.(5) reads
$$h^{\prime \prime }(x)xh^{}(x)(\lambda +1)h(x)=0$$
(20)
For integer and non-negative values of $`\lambda `$ a particular solution of the homogeneous equation (20) can always be found with the Laplace method. One explicitly finds for $`\lambda =1,2,3,4`$
$`h_1=e^{x^2/2}xh_2=e^{x^2/2}(1+x^2)`$ (22)
$`h_3=e^{x^2/2}(3x+x^3)h_4=e^{x^2/2}(3+6x^2+x^4)`$
It should be stressed that solutions of the type (22) which are even function of $`x`$ are bounded from below and a linear combination may develop non-trivial minima at some value of the field. This is the case of $`h_4e^{4\tau }+ch_2e^{2\tau }`$ for some negative values of the constant $`c`$. The general solution is
$$v_n(x,\tau )=e^{n\tau }\left(Ah_n(x)+Bh_n(x)^xe^{s^2/2}h_n(s)^2𝑑s\right)$$
(23)
where $`h_n`$ is obtained with the Laplace method and $`n=0,1,2,\mathrm{}`$.
The presence of solutions of the flow equation with entirely different physical implications should not come as a surprise. In $`d=2`$ eq.(2) admits the following solutions, as it can be checked by direct substitution,
$`u(x,t)=e^{(2\beta ^2/4\pi )t}\mathrm{cos}(\beta x)`$ (24)
$`u(x,t)=e^{(2+\gamma /4\pi )t}e^{\sqrt{\gamma }x}`$ (25)
where $`\beta `$ and $`\gamma `$ are arbitrary constants. The first one describes the physics of the Sine-Gordon theory, while the second one defines the Liouville theory . Also in this case the same fixed point may describe entirely different physical theories.
The physical relevance (if any) of these solutions is not clear to us. Although one may speculate about their possible applications to the Standard Model or to Cosmology, all we can say for the moment is that the structure of the GFP for a simple scalar theory is richer than previously discussed. In particular, the universality class spanned by the polynomial eigen-potentials in the LPA does not contain all the possible relevant interactions.
One final remark is the following. We have used the linearised Wegner-Hougton equation in the LPA approximation with sharp cut-off. This equation is the same of the Polchinski equation obtained with the smooth cut-off, apart for the unimportant constant $`a_d`$ . Thus our result is quite robust since these new scaling scaling exponents does not depend on the cut-off function used in the derivation.
I am very much indebted with Martin Reuter and Dario Zappalà for very enlightening discussions and encouragements. |
warning/0001/gr-qc0001017.html | ar5iv | text | # General analysis of self-dual solutions for the Einstein-Maxwell-Chern-Simons theory in (1+2) dimensions
## Abstract
The solutions of the Einstein-Maxwell-Chern-Simons theory are studied in $`(1+2)`$ dimensions with the self-duality condition imposed on the Maxwell field. We give a closed form of the general solution which is determined by a single function having the physical meaning of the quasilocal angular momentum of the solution. This function completely determines the geometry of spacetime, also providing the direct computation of the conserved total mass and angular momentum of the configurations.
The $`(1+2)`$-dimensional general relativity has attracted considerable attention recently (see, e.g., and references therein). This is explained by two main reasons. Firstly, since the discovery of the BTZ black hole solutions , the three-dimensional gravity became a helpful laboratory for the study of geometrical, statistical and thermodynamics properties of black holes. Secondly, the quantization of these models may give new insights into the general quantum gravity problem.
A number of generalizations of BTZ solution to the case of nontrivial electromagnetic field source were developed previously . The aim of our present paper is to give a new general analysis of the self-dual Einstein-Maxwell solutions in three dimensions.
The Lagrangian 3-form,
$$L=\frac{1}{2}1\lambda 1\frac{1}{2}FF\frac{\mu }{2}AF,$$
(1)
contains the Einstein-Hilbert term, the cosmological constant $`\lambda `$, and the standard Maxwell field $`F=dA`$ Lagrangian along with the Chern-Simons term with the coupling constant $`\mu `$, . Variation of $`L`$ with respect to the coframe field $`\vartheta ^\alpha `$ and the electromagnetic potential $`A`$ yields the system of field equations:
$`G_{\alpha \beta }\vartheta ^\beta +\lambda \vartheta _\alpha `$ $`=`$ $`\mathrm{\Sigma }_\alpha ,`$ (2)
$`dF+\mu F`$ $`=`$ $`0.`$ (3)
Here $`\mathrm{\Sigma }_\alpha =\frac{1}{2}[(e_\alpha F)FFe_\alpha F]`$ is the Maxwell field energy-momentum 2-form, and $`G_{\alpha \beta }`$ is the Einstein tensor.
In the study of the “spherically”-symmetric solutions, we choose the local coordinates $`(t,r,\varphi )`$ and make the general ansatz for the coframe 1-form,
$$\vartheta ^0=fdt,\vartheta ^1=gdr,\vartheta ^2=h(d\varphi +adt),$$
(4)
and for the Maxwell field
$$F=E\vartheta ^0\vartheta ^1+B\vartheta ^1\vartheta ^2,$$
(5)
Here $`f,g,h,a`$ and $`E,B`$ are the functions of the radial coordinate $`r`$.
Without any loss of generality it will be convenient to absorb the metric function $`g(r)`$ by the simple redefinition of the radial coordinate:
$$\rho =g(r)𝑑r(\mathrm{hence}\vartheta ^1=d\rho ).$$
(6)
¿From now on, the derivatives w.r.t. new coordinate $`\rho `$ will be denoted by prime.
After all these preliminaries, the Einstein field equations read explicitly
$`{\displaystyle \frac{1}{2}}\beta ^{}\beta \gamma `$ $`=`$ $`EB,`$ (7)
$`\gamma ^{}+\gamma ^2+{\displaystyle \frac{1}{4}}\beta ^2+\lambda `$ $`=`$ $`{\displaystyle \frac{1}{2}}(E^2+B^2),`$ (8)
$`\alpha ^{}+\alpha ^2{\displaystyle \frac{3}{4}}\beta ^2+\lambda `$ $`=`$ $`{\displaystyle \frac{1}{2}}(E^2+B^2),`$ (9)
$`\alpha \gamma {\displaystyle \frac{1}{4}}\beta ^2\lambda `$ $`=`$ $`{\displaystyle \frac{1}{2}}(E^2B^2),`$ (10)
and this system is supplemented by the (modified) Maxwell equations:
$`B^{}\alpha B+\beta E+\mu E`$ $`=`$ $`0,`$ (11)
$`E^{}\gamma E+\mu B`$ $`=`$ $`0.`$ (12)
Here we introduced the functions
$$\alpha =\frac{f^{}}{f},\beta =\frac{a^{}h}{f},\gamma =\frac{h^{}}{h}.$$
(13)
which actually describe the Levi-Civita connection coefficients. The remarkable feature is that the complete Einstein-Maxwell system (7)-(12) involves no metric functions (i.e., $`f,g,h,a`$), but only the connection combinations $`\alpha ,\beta ,\gamma `$.
Let us assume “self-duality” of the electromagnetic field:
$$E=kB,\mathrm{with}k^2=1.$$
(14)
Substituting this into (11)-(12) and (10), we find that the two unknown functions are expressed in terms of the third:
$$\alpha =\frac{k}{2}\beta +\mathrm{},\gamma =\frac{k}{2}\beta +\mathrm{}.$$
(15)
Here we denote $`\mathrm{}:=\pm \sqrt{\lambda }`$.
Taking into account the algebraic relations (14) and (15), we are left with two essential equations for determining the functions $`\beta `$ and $`B`$. Explicitly, the equations (7) and (12) are reduced to
$`\beta ^{}k\beta ^2+2\mathrm{}\beta `$ $`=`$ $`2kB^2,`$ (16)
$`(B^2)^{}k\beta B^2+2\mathrm{}B^2`$ $`=`$ $`2k\mu B^2.`$ (17)
This system of nonlinear coupled equations is simplified with the help of the substitution
$$\beta =\frac{1}{\omega },B^2=\frac{k}{2\omega }\frac{\phi ^{}}{\phi },$$
(18)
which yields for the new functions $`\phi `$ and $`\omega `$ the linear equations:
$`\phi ^{\prime \prime }=2k\mu \phi ^{}.`$ (19)
$`\omega ^{}+\left({\displaystyle \frac{\phi ^{}}{\phi }}2\mathrm{}\right)\omega +k=0.`$ (20)
Multiplying (20) by $`\phi e^{2\mathrm{}\rho }`$, we easily obtain the general solution
$$\omega =\frac{k\mathrm{\Omega }}{\phi e^{2\mathrm{}\rho }},\mathrm{with}\mathrm{\Omega }:=c_0\stackrel{\rho }{}𝑑\stackrel{~}{\rho }\phi (\stackrel{~}{\rho })e^{2\mathrm{}\stackrel{~}{\rho }}.$$
(21)
Note that in fact it is not necessary to know the explicit form of $`\phi `$ when solving (20). At the same time, of course, the equation (19) is straightforwardly integrated. Depending on $`\mu `$, it admits two solutions:
$`\phi `$ $`=`$ $`\rho +\rho _0,\mathrm{when}\mu =0,`$ (22)
$`\phi `$ $`=`$ $`1+u_0e^{2k\mu \rho },\mathrm{when}\mu 0.`$ (23)
Here $`c_0,\rho _0,u_0`$ are integration constants. It is worthwhile to note that an overall constant factor is irrelevant for $`\phi `$ because this function appears everywhere only through the ratio (18).
Quite remarkably, however, we will not need the explicit form of $`\phi `$ till the very end of our analysis. Such a formulation is extremely convenient since it makes it possible to treat the cases of standard Maxwell theory with $`\mu =0`$, and the Maxwell-Chern-Simons with $`\mu 0`$ simultaneously.
It remains to integrate the equations for the metric functions (13). This is straightforward, and using (21) in (15), we find:
$`f`$ $`=`$ $`f_0e^\mathrm{}\rho \mathrm{\Omega }^{\frac{1}{2}},`$ (24)
$`h`$ $`=`$ $`h_0e^\mathrm{}\rho \mathrm{\Omega }^{\frac{1}{2}},`$ (25)
$`a`$ $`=`$ $`{\displaystyle \frac{kf_0}{h_0}}\mathrm{\Omega }^1a_0.`$ (26)
For completeness, the magnetic field reads:
$$B^2=\frac{\phi ^{}}{2}e^{2\mathrm{}\rho }\mathrm{\Omega }^1.$$
(27)
Here $`f_0,h_0,a_0`$ are integration constants.
The first main result which we learned in our study, is that a general “spherically”-symmetric (rotating, for nontrivial $`a`$) solution
$$ds^2=\left(\vartheta ^0\right)^2+\left(\vartheta ^1\right)^2+\left(\vartheta ^2\right)^2$$
(28)
of the Einstein-Maxwell (with or without Chern-Simons term) field equations is always represented solely in terms the function $`\phi `$.
Because of such an important role played by $`\phi `$, it would be interesting to find out its physical meaning. The latter is revealed in the analysis of the quasilocal mass and angular momentum which characterize our general solution.
We refer the reader to for a comprehensive discussion of the conserved quantities for gravitating systems within the framework of Hamiltonian formulation of general relativity theory. As a first step, let us use the coordinate freedom and replace $`\rho `$ by a new radial coordinate defined by
$$r=h(\rho ).$$
(29)
Then a nontrivial metric function $`g`$ will reappear in the coframe (4) \[and hence in the metric (28)\]. Using (25) we find explicitly
$$g=\frac{d\rho }{dr}=\left(\mathrm{}r\frac{h_0^2}{2r}\phi \right)^1.$$
(30)
Now we can write the quasilocal angular momentum at a distance $`r`$, which reads
$$j(r)=\frac{g^1r^3}{f}\frac{da}{dr},$$
(31)
in our notations. Using (24)-(26) and (29)-(30), we find
$$j(r)=kh_0^2\phi .$$
(32)
Clearly, one should invert (29) and use $`\rho =\rho (r)`$ in (32), or alternatively, one can consider the angular momentum $`j`$ as a function of $`\rho `$.
The quasilocal energy is given, in our notations, by the difference
$$E(r)=g_0^1g^1,$$
(33)
where the first term describes the contribution of the background “empty” spacetime. The latter, as usually, is given by $`g_0^1=\mathrm{}r`$. Making use of (30), we obtain explicitly
$$E(r)=\frac{h_0^2}{2r}\phi =\frac{k}{2r}j(r).$$
(34)
Finally, the quasilocal mass is determined by the expression
$$m(r)=2fE(r)ja.$$
(35)
Substituting (24), (26), (32) and (34), we arrive at the result:
$$m(r)=a_0j(r).$$
(36)
We thus have demonstrated that the function $`\phi `$, which determines the spacetime geometry via (21) and (24)-(26), is also determining all the quasilocal quantities of the gravitating system: its energy, mass and angular momentum. They turn out to be proportional to each other, describing a sort of extremal configuration. Because of the relation (32), one can say that the angular momentum $`j(r)`$ underlies the construction of self-dual Einstein-Maxwell equations: given this function, the metric and electromagnetic field are described by (24)-(27) with $`j(r)`$ inserted.
The total angular momentum and mass are defined by the limits $`J:=j|_r\mathrm{}`$ and $`M:=m|_r\mathrm{}`$, respectively. In order to find these quantities, one does not need to obtain the explicit exact form of the inverse coordinate transformation $`\rho (r)`$ from (29). It is sufficient to investigate the approximate behaviour of $`\phi (r)`$ and $`\mathrm{\Omega }(r)`$ for large values of $`r`$, which is always clear directly from the inspection of (19)-(21).
In particular, one can immediately verify that the limiting value $`\mathrm{\Omega }|_r\mathrm{}`$ is equal either infinity or $`c_0`$, depending on the values of $`\mu `$ and $`\mathrm{}`$. Consequently, the integration constant $`a_0`$ should be equal either $`0`$, or $`\frac{kf_0}{h_0c_0}`$, providing the required asymptotic vanishing of the metric function $`a(r)`$. Correspondingly, one finds that the quasilocal mass $`m`$ vanishes for many configurations.
The quasilocal angular momentum $`j(r)`$ (or the function $`\phi (r)`$) diverges, in general, for $`r\mathrm{}`$. However, the direct analysis of (19)-(21) shows that $`J`$ is finite for all the solutions with $`k\mu <0`$. Actually, there are two large classes of such configurations: $`(A)`$ $`k\mu <0,\mathrm{}=0`$, then $`J=kh_0^2`$ and $`M=0`$, and $`(B)`$ $`k\mu <0,\mathrm{}>0`$, then $`J=kh_0^2`$ and $`M=\frac{f_0h_0}{c_0}`$. Imposing the standard asymptotic condition $`f^2g^2|_r\mathrm{}=1`$, one finds $`a_0=k\mathrm{}`$, and thus the solutions of the class $`(B)`$ are all characterized (irrespectively of the value of the Chern-Simons coupling constant $`\mu `$) by $`M^2=\mathrm{}^2J^2`$. This class also contains the extremal BTZ solution, as a particular case when the electromagnetic field is absent. \[The general non-extremal BTZ solution cannot be recovered because of the algebraic relations (15) which necessarily hold for the self-dual electromagnetic field\].
Summarizing, we have obtained a general solution of the Einstein-Maxwell-Chern-Simons theory in $`(1+2)`$ dimensions which covers all the particular cases studied previously. The form of the solution (24)-(27), (21) is transparent and easy to analyse: everything is determined by a single function $`\phi (\rho )`$ which has a clear physical meaning as the quasilocal angular momentum of the gravitational field configuration. The computation of the total mass and angular momentum is straightforward and it involves only the analysis of the asymptotic behaviour of $`\phi `$.
The authors are grateful to TUBITAK for the support of this research. Y.N.O. is also grateful to the Department of Physics, Middle East Technical University, for the warm hospitality. |
warning/0001/cond-mat0001135.html | ar5iv | text | # First and second order transition of frustrated Heisenberg spin systems
## 1 INTRODUCTION
The critical properties of frustrated spin systems are still under discussion . In particular no consensus exists about the nature of the phase transition of an antiferromagnet on a stacked triangular lattice with nearest neighbor interactions. The symmetry group for a Heisenberg antiferromagnet at high temperatures $`SO_3`$ changes to $`SO_2`$ for the usual antiferromagnet, but the symmetry is completely broken in the low temperature phase of the frustrated antiferromagnet. This difference in the breakdown of symmetry between the frustrated and non frustrated cases should lead to different classes of critical behavior. The controversial point for the stacked triangular antiferromagnet (STA) is whether the phase transition is of second order with a new chiral universality class or whether its true nature is a weak first order change.
Monte Carlo studies favor a new universality class whereas renormalization group studies indicate a first order transition, since no stable fixed point can be found for the Heisenberg case in order $`ϵ^2`$ with $`ϵ=4d`$ (d is the dimension of space) and also for an expansion up to three loops in $`d=3`$ . The results for the critical exponents of the numerous Monte Carlo simulations are listed in table 1. They agree reasonably well and also with the experimental values in table 2. One notices that the new critical indices are quite distinct from the standard ones for the Heisenberg model with symmetry breaking $`SO_3/SO_2`$ listed in table 3. The only flaw is the small negative values of $`\eta `$, since this exponent should always be positive .
The discrepancy between the renormalization group analysis and the Monte Carlo simulation can be removed by assuming that the first order phase transition occurs with a correlation length larger than the diameter of the cluster studied. Since this size cannot be substantially increased with the present technical means and skills we have tried to investigate modified lattice spin systems which should belong to the same universality class. The idea is to shorten the correlation length by these modifications in order to reveal the first order nature of the transition. This strategy was successful for the $`x`$$`y`$ STA–model we investigated before and we could show that in fact the phase transition is of first order in this case.
Following Zumbach one can analyze such a weak first order transition in terms of the renormalization group approach as due to a fixed point in the complex parameter space (More precisely it is only necessary to have a minimum in the RG flow). A basin of attraction for the real parameters will be generated and “mimic” a second order transition with slightly changed scaling relations. The unphysical negative values of $`\eta `$ in table 1 could be corrected using Zumbach’s approach.
Investigations about the phase transitions of the STA–systems and similar helimagnetic systems are not only of interest for the field of magnetism including the experimental studies. Phase transitions in superfluids, in type II superconductors, and in smectic–A liquid crystals should be similar in nature to the ones of frustrated Heisenberg magnets.
The numerical simulations are closely connected to a similar study for the $`x`$$`y`$ spins . We study the classical Heisenberg spin system on the stacked triangular lattice by fixing for selected triangles the spin direction to a 120–order which would correspond to the ideal antiferromagnetic order on a triangular lattice as in Fig. 4. The common orientation of a 120–cluster is still free. The principle behind this construction is that modes removed by the 120–rigidity are “irrelevant” close to the critical temperature. What is relevant is the common direction and orientation. The antiferromagnet STA and the rigid antiferromagnet STAR should therefore be in the same universality class. These considerations can also be applied to Stiefel’s $`V_{3,2}`$ or dihedral model . A third model studied is the right handed trihedral model. It is constructed by adding a third vector given by the vector product of the two of the dihedron. As we do not add new degrees of freedom the model should again belong to the same universality class.
That the behavior close to the phase transition should not change if one modifies the models by constraints can only be expected from a second order transition. However, if one assumes that the transition is of first order, but not visible since the barriers between the two phases are too low for the spin clusters one can simulate, then building in constraints in the models and simulations should make a difference since the barriers are more difficult to overcome.
In section II we present a review of the RG studies. In the following section the three models studied are presented. The details of the simulations and the finite size scaling analysis are described in section IV. Results are given in section V. The discussion is in section VI where we will review also the experimental studies and the other RG studies in $`2+ϵ`$ expansions.
## 2 Renormalization group studies and complex fixed points
### 2.1 Renormalization group studies in $`d=3`$ and $`d=4ϵ`$
The renormalization group studies have motivated for the re-analysis of the frustrated Heisenberg antiferromagnet with the Monte Carlo technique. Therefore we repeat briefly the main points. For more details we refer to Antonenko and Sokolov .
The Hamiltonian one considers is
$$H=\frac{1}{2}d^3x\left[r_0^2\varphi _\alpha \varphi _\alpha ^{}+\varphi _\alpha \varphi _\alpha ^{}+\frac{u_0}{2}\varphi _\alpha \varphi _\alpha ^{}\varphi _\beta \varphi _\beta ^{}+\frac{w_0}{2}\varphi _\alpha \varphi _\alpha \varphi _\beta ^{}\varphi _\beta ^{}\right]$$
(1)
with $`\varphi _\alpha `$ a complex vector order parameter field. Summation of repeated indices $`\alpha ,\beta =1,\mathrm{}N`$ is implied with $`N=3`$ for the Heisenberg case. The deviation from the transition temperature is given by $`r_0^2`$ and there are two coupling constants $`u_0`$ and $`w_0`$, both positive for a non–collinear ground state, that is the two vectors forming the complex vector $`\varphi `$ should not be parallel. For the non frustrated case there is only one field and one coupling constant, but for the superfluid <sup>3</sup>He the above complexity also arises .
Renormalization group calculations for $`d=4ϵ`$ or also directly for $`d=3`$ show that four fixed points exist:
* The Gaussian fixed G point at $`u^{}=v^{}=0`$ with mean field critical exponents.
* The $`O(2N)`$ fixed point H at $`w^{}=0`$ and $`u^{}=u_H0`$ with $`O(2N)`$ exponents (see table 3).
* Two fixed points $`F_+`$ and $`F_{}`$ at location ($`u_{F_+},w_{F_+})`$ and $`(u_F_{},w_F_{}`$) different from zero. These are the fixed points associated with a new ”chiral” universality class.
The existence and stability of the fixed points depend on the number of components $`N`$:
* $`N>N_c`$: four fixed points are present but three are unstable ($`G`$, $`H`$, $`F_{}`$) and the stable one is $`F_+`$. Therefore the transition belongs to a new universality class different from the standard $`O(N)`$ class. If the initial point for the RG flow is to the left of the line ($`G,F_{}`$), see fig. 2 a. the flow is unstable and the transition will be of first order.
* $`N=N_c`$: the fixed points $`F_{}`$ and $`F_+`$ coalesce to a marginally stable fixed point. One would think that the transition is “tricritical” but the exponents are different and not given by the tricritical mean field values. The reason is that there are two quartic coupling constants not zero, see Fig. 2 b, in contrast to the tricritical $`O(N)`$ point where the quartic term disappears and a sextic term takes over .
* $`N<N_c`$: F<sub>-</sub> and $`F_+`$ move into the complex parameter space, see fig. 2 c and d. The absence of stable fixed points is interpreted as a signature of a first order transition.
The difficulty is to find a reliable value for $`N_c`$. The dependence on $`d`$ has been calculated to first order in $`ϵ=4d`$ as
$$N_c=\mathrm{\hspace{0.17em}21.8}23.4ϵ.$$
(2)
In repeating the analysis for the triangular antiferromagnet and choosing $`ϵ=1`$ for $`d=3`$ Kawamura has argued that the $`XY`$ and Heisenberg cases should be above $`N_c`$, motivated by Monte Carlo calculations which supported a second order transition of a new universality class with critical exponents for the Heisenberg case $`\nu =0.59`$, $`\beta =0.29`$ (see table 1) different from the standard ones.
Of course the results linear or higher in $`ϵ`$ are at best asymptotic . Antonenko, Sokolov and Varnashev extended the analysis to $`ϵ^2`$ . For an analysis directly in three dimensions see . For the critical number of components after resummation
$$N_c=3.91(1)$$
(3)
is obtained by the direct method and $`N_c=3.39`$ with the $`ϵ`$ expansion to second order . In Fig. 2 the results of the RG studies are depicted. There is a line which separates a region of first order transition for high dimensions and low $`N`$ (Fig. 2c) from a region of second order transition for low dimensions and high $`N`$ (Fig. 2a). The transition on this line is special (Fig. 2b) but is not the standard tricritical.
Similar considerations have been applied to the normal to superconducting type II transition. Starting point is the Ginzburg–Landau Hamiltonian with also two coupling constants, a quartic coupling and the coupling to the magnetic field. Expanding to first order in $`ϵ=4d`$ Halperin et al. and Chen et al. arrived at results almost identical to the ones for frustrated magnets. There are also four fixed points with one stable one if the number of order parameters $`N`$ exceeds $`N_c=365.9`$ for a fictitious superconductor in 4 dimensions. The unstable $`0(2N)`$ fixed point is replaced by an $`XY`$ fixed point. For $`N<N_c`$ only the unstable Gaussian and this unstable $`XY`$ fixed point are present, and the question whether the physical relevant $`N=2`$ case is of second order or not can also not be answered, although the value of $`N_c`$ in two loop order has been calculated : $`N_c=365.90640.76ϵ+O(ϵ^2)`$. If we put $`ϵ=1`$ we obtain $`N_c(d=3)275`$ which is less than two and the transition should be of second order. The nematic-to-smectic-A transition in liquid crystals is described by a model similar to the Ginzburg–Landau Hamiltonian transition and again the problem of second or first order transition arises .
The superfluid phase transitions of He<sup>3</sup> and the phase transitions of helimagnets were first studied with the field theoretical approach (1) by Love, More, Jones and Bailin and by Garel and Pfeuty respectively. In principle all the considerations should also be applicable to these systems, however for He<sup>3</sup> the critical region is not really accessible. Returning to the frustrated antiferromagnets and accepting the value $`N_c=3.91`$ for $`d=3`$ of eq. 3, the $`XY`$ and Heisenberg systems and also the helimagnets should have a first order transition in three dimensions.
### 2.2 Almost second order transitions and complex fixed points
The discrepancy between the results of RG studies which indicate a first order transition and the results of Monte Carlo study which favor a second order transition (see table 1) can be removed by using complex fixed points in the RG analysis. Indeed, as $`N_c`$ is very close to $`N=3`$ the fixed points $`F_+`$ and $`F_{}`$ will not be far off from the real space of parameters $`u^{}`$ and $`v^{}`$ at $`N_c`$ as depicted in fig. 2 d. Zumbach has studied in detail the influence of a complex fixed point on the RG analysis. The basin of attraction of such a complex fixed point will mimic a second order transition with modified scaling relations
$`2\beta `$ $`=`$ $`\nu (d2+\eta c_\beta )`$ (4)
$`\gamma `$ $`=`$ $`\nu (2\eta +c_\gamma ).`$ (5)
The constants $`c_\beta `$ and $`c_\gamma `$ are corrections of the scaling relations absent if the fixed point is real. Using these relations to determine $`\eta `$ without corrections we get a negative value for $`\eta `$. Since $`\eta `$ must be positive or at least zero we have an estimate for $`c_{\beta ,\gamma }`$ and an indication that the transition is not a real second order one. We have used this criterion before for $`x`$$`y`$ case . The essential point is the smallness of $`\eta `$, approximately $`\eta 0.03`$ for magnets, so that the correction terms $`c_{\beta ,\gamma }`$ are clearly visible.
A negative $`\eta `$ is unphysical since unitarity would be violated in the corresponding quantum field theory . For spin glasses this restriction no longer holds and indeed a negative $`\eta `$ can be obtained . Also for superconductors a negative $`\eta `$ have been found influenced by the choice of gauge . The critical exponents must be gauge-independent and in a recent study it has been shown that $`\eta `$ is positive for superconductors.
Using the local potential approximation for the RG the critical exponent $`\nu `$ can be calculated for real and complex fixed points (more precisely for a minimum in the flow) following Zumbach . In this approximation $`\eta `$ is zero and the result will have errors of a few percents, for example 8% for the Heisenberg ferromagnet . For the fixed point $`F_+`$ of the frustrated systems Zumbach obtains $`\nu 0.63`$ for $`N=3`$. This result is compatible with $`\nu =0.59(1)`$ obtained by MC calculations for Heisenberg spins on a stacked triangular lattice (see table 1).
By the same approximation Zumbach estimated the interval where the systems should be under the influence of a complex fixed point as
$$2.58<N<4.7.$$
(6)
Outside of these limits for $`N<2.58`$ the transition is of first order and for $`N>4.7`$ the transition belongs to the new universality class. The two values $`N_c=4.7`$ for the local potential approximation and $`N_c=3.91(1)`$ found by Antonenko and Sokolov by field theoretical means are not so different. The minimum limit 2.58 is in agreement with our simulations for the Stiefel model. The transition is indeed of first order for $`N=2`$ and it will be shown here that the $`N=3`$ case has a pseudo second order behavior.
## 3 Models and simulations
In this section we introduce briefly different models that we use in this work. A more complete presentation can be found in .
### 3.1 The STAR model
For the stacked triangular antiferromagnet (STA) we take the simplest Hamiltonian with one exchange interaction constant $`J>0`$ (antiferromagnetic).
$$H=\underset{(ij)}{}J_{ij}𝐒_i.𝐒_j,$$
(7)
where $`𝐒_i`$ is a three component spin vector of unit length, and the sum is over all neighbor spin pairs of the lattice. There are six nearest neighbor spins in the plane and two in adjacent planes. In the ground state the spins are in a planar arrangement with the three spins at the corners of each triangle forming a 120 structure (see Fig. 4).
For one triangle the spin vectors obey the equation
$$𝐒_1+𝐒_2+𝐒_3=\mathrm{𝟎}.$$
(8)
where the indices refer to the three corners. To get the STAR model one imposes this equation as a restriction for the spin directions valid at all temperatures. In this theory the local fluctuations violating this constraint become modes with a gap. Thus they do not contribute to the critical behavior and can be neglected. However, one can do this only for selected triangles, for instance the shaded ones in fig. 4, otherwise the spin configuration would be completely frozen.
The lattice is partitioned into interacting triangles which do not have common corners, so that each spin belongs only to one triangle. For the Monte Carlo simulation one spin direction (two degrees of freedom) is chosen and then the direction of the second spin selected in the cone of 120 around the first one. The third is determined by the constraint (8). All the spins of the supertriangles (the shaded ones) have, only in the ground state, the same orientation. At finite temperatures there is a perfect order within a supertriangle, but fluctuations of the orientation between these triangles occur.
The Monte Carlo updating for the state of the supertriangle is done the following way. First two orthogonal unit vectors are chosen at random in three dimensions. One needs three Euler angles to do this, the first $`\theta _0`$ must be chosen with probability $`\mathrm{sin}(\theta _0)d\theta _0`$ and the other two $`\theta _{1,2}`$ with probability $`\theta _{1,2}`$ so that
$`𝐮`$ $`=`$ $`\left(\begin{array}{c}\mathrm{sin}(\theta _0)\mathrm{cos}(\theta _1)\\ \mathrm{sin}(\theta _0)\mathrm{sin}(\theta _1)\\ \mathrm{cos}(\theta _0)\end{array}\right)`$ (9)
$`𝐯`$ $`=`$ $`\left(\begin{array}{c}\mathrm{sin}(\theta _1)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _0)\mathrm{cos}(\theta _1)\mathrm{sin}(\theta _2)\\ \mathrm{cos}(\theta _1)\mathrm{cos}(\theta _2)\mathrm{cos}(\theta _0)\mathrm{sin}(\theta _1)\mathrm{sin}(\theta _2)\\ \mathrm{sin}(\theta _0)\mathrm{sin}(\theta _2)\end{array}\right)`$ (10)
Then the three spin vectors on each supertriangle are determined by
$`𝐒_\mathrm{𝟏}`$ $`=`$ $`𝐮`$
$`𝐒_\mathrm{𝟐}`$ $`=`$ $`\mathrm{cos}(120^{})𝐮+\mathrm{sin}(120^{})𝐯`$ (11)
$`𝐒_\mathrm{𝟑}`$ $`=`$ $`\mathrm{cos}(240^{})𝐮+\mathrm{sin}(240^{})𝐯.`$
The interaction energy between the spins of this supertriangle with the spins of the neighboring ones is calculated in the usual way. We follow the standard Metropolis algorithm to update one supertriangle after the other. One million MC–steps for equilibration are carried out and up to six million steps were used for the largest sizes to obtain reliable averages. Long enough simulations are necessary because the critical slowing down is strong.
The number of spins is given by $`N=L\times L\times L_z`$, where $`L\times L`$ gives the number of spins in one plane and $`L_z=2L/3`$ the number of planes. Simulations have been done for $`L=18,\mathrm{\hspace{0.17em}21},\mathrm{\hspace{0.17em}24},\mathrm{\hspace{0.17em}30},\mathrm{\hspace{0.17em}36},\mathrm{\hspace{0.17em}42}`$, where L must be a multiple of 3 in order to use periodic boundary conditions and to avoid frustration effects in the planes.
The order parameter $`M`$ used in the calculations is
$$M=\frac{1}{N}\underset{i=1}{\overset{3}{}}|M_i|$$
(12)
where the magnetization $`M_i`$ is defined on one of the three sublattices. This definition generalizes the one used for two collinear sublattices.
### 3.2 The Stiefel model
The Stiefel model $`V_{3,2}`$, that is a “Zweibein” (or dihedral model) in three dimensional space is a further abstraction of the constrained three spin system discussed in the previous section. The three spins at the corner of a triangle can be taken as a planar or degenerate “Dreibein” where the third leg is a linear combination of the other two and can be left out. The energy is given by
$$H=J\underset{ij}{}\left[𝐞_1(i)𝐞_1(j)+𝐞_2(i)𝐞_2(j)\right]$$
(13)
where the mutual orthogonal three component unit vectors $`𝐞_1(i)`$ and $`𝐞_2(i)`$ at lattice site $`i`$ interact with the next pair of vectors at the neighboring sites $`j`$. The interaction constant is here negative to favor alignment of the vectors at different sites. Also a cubic lattice instead of a hexagonal one can be taken (for further details see ).
We use a single Monte-Carlo cluster algorithm . A cluster of connected spins is constructed and updated in the standard way. The first site of the cluster $`o`$ is chosen at random together with a random reflection r. Then all neighbors $`j`$ are visited and added to the cluster with probability
$$P=1\mathrm{exp}(\mathrm{min}\{0,E_1+E_2\})$$
(14)
where
$$E_{1,2}=\mathrm{\hspace{0.17em}2}\beta (𝐫𝐞_{1,2}(o))(𝐫𝐞_{1,2}(j))$$
(15)
and $`\beta =J/T`$ until the process stops. Finally all spins of the cluster are “reflected” with respect to the plane $``$ to $`𝐫`$, that is
$$𝐞_{1,2}^{new}=𝐞_{1,2}^{old}2(𝐞_{1,2}^{old}𝐫)𝐫.$$
(16)
It has been demonstrated that this cluster method is very efficient in reducing the critical slowing down for the $`O(N)`$ ferromagnet . We thought it should be useful in our case since the order is also ferromagnetic, contrary to the original STA model where the spins have a 120 structure so that the cluster algorithm described does not work. However, we did not observe a decreasing of the critical slowing down. Indeed on general grounds Sokal et al. have argued that Wolff’s cluster algorithm cannot reduce the critical slowing down unless the manifold for the order parameter is a sphere as for $`O(N)`$ ferromagnet.
In each simulation 5 million measurements were made after enough single cluster updatings of 1 million steps for equilibration. Cubic systems with linear dimension $`L=15,\mathrm{\hspace{0.17em}20},\mathrm{\hspace{0.17em}25},\mathrm{\hspace{0.17em}30},\mathrm{\hspace{0.17em}35},\mathrm{\hspace{0.17em}40}`$ were simulated. In comparison with the STAR model $`L`$ should be multiplied by $`\sqrt{3}`$ since one site of the Stiefel model represents three spins. The equivalent sizes would be 25 to 70.
The order parameter $`M`$ for this model is
$$M=\frac{1}{2N}\underset{i=1}{\overset{2}{}}\left|M_i\right|$$
(17)
where $`M_i`$ is the total magnetization given by the sum of the vectors $`𝐞_i`$ over all sites and $`N=L^3`$ is the total number.
### 3.3 The right handed trihedral model
If one adds a third vector given by the vector product
$$𝐞_3(i)=𝐞_1(i)\times 𝐞_2(i).$$
(18)
and takes an energy formally identical to the previous one
$$H=J\underset{(ij)}{}\left[𝐞_1(i)𝐞_1(j)+𝐞_2(i)𝐞_2(j)+𝐞_3(i)𝐞_3(j)\right]$$
(19)
has one constructed a really different model? Since introducing this new vector $`𝐞_3`$ no degree of freedom has been added and therefore this system should belong to the same universality class as the original $`V_{3,2}`$ Stiefel model. It is not the $`V_{3,3}`$ Stiefel model since only right handed systems are constructed. An additional Ising symmetry is absent contrary to the $`x`$$`y`$ or planar case where the “chirality” as an additional Ising variable is present.
In the work of Azaria et al. it has been shown that the interaction between the third vector, which is absent in the original STA model, will be automatically generated by the renormalization group in $`d=2+ϵ`$ expansion and, at the fixed point, should be equal to the first two.
This system has been studied by Loison and Diep before. Here we will confirm that the transition is strongly first order. The Monte Carlo procedure is similar to the one used for the STAR model. In addition to the two orthogonal vectors u and v of eq. 9-10 a third vector given by the vector product $`𝐰=𝐮\times 𝐯`$ is constructed. The interaction energy between the spins of this trihedral with the spins of the neighboring trihedral is calculated in the usual way and we follow the standard Metropolis algorithm to update one trihedral after the other. In each simulation 10 000 Monte Carlo steps were made for equilibration and for taking the averages. Cubic system of linear dimension up to $`L=30`$ were simulated. The order parameter $`M`$ is similar to the previous case
$$M=\frac{1}{3N}\underset{i=1}{\overset{3}{}}\left|M_i\right|$$
(20)
but one has to take the average of three $`M_i`$ instead of two.
## 4 Numerical method
### 4.1 Finite size scaling for second order transitions
We use the histogram MC technique developed by Ferrenberg and Swendsen and divide the energy range into 30,000 intervals, for more details see . The errors are determined with the help of the Jackknife procedure .
For each temperature $`T`$ we calculate the internal energy per site $`\overline{E}=E/N`$ and the specific heat $`C=(E^2E^2)/(NT^2)`$, where $`\mathrm{}`$ indicate the average and $`N`$ is the number of sites. Similarly we determine the averages of the order parameter or staggered magnetization $`\overline{M}=M`$ and the corresponding susceptibility $`\chi =N\left(M^2M^2\right)/T`$. The quantities needed besides $`\overline{M}`$ for the finite size analysis are defined below
$`\chi _2`$ $`=`$ $`{\displaystyle \frac{N}{T}}M^2`$ (21)
$`\chi _4`$ $`=`$ $`{\displaystyle \frac{N^3}{T^3}}\left(M^4\mathrm{\hspace{0.17em}3}M^2^2\right)`$ (22)
$`U`$ $`=`$ $`\mathrm{\hspace{0.17em}1}{\displaystyle \frac{M^4}{3M^2^2}}`$ (23)
$`V_1`$ $`=`$ $`{\displaystyle \frac{ME}{M}}E`$ (24)
where $`\chi _2`$ is the magnetic susceptibility and $`\chi _4`$ the fourth derivative of the free energy with respect to the magnetic field in the high temperature region where the order parameter is zero. The cumulant $`V_1`$ is used to obtain the critical exponent $`\nu `$, and the fourth order cumulant $`U`$ to determine the critical temperature.
According to the FSS theory for a second order transition the various quantities just defined should scale for a sufficiently large system at the critical temperature $`T_c`$ as
$`\chi _2`$ $`=`$ $`g_{\chi _2}L^{\gamma /\nu }`$ (25)
$`\chi _4`$ $`=`$ $`g_{\chi _4}L^{\gamma _4/\nu }`$ (26)
$`V_1`$ $`=`$ $`g_{V_1}L^{1/\nu }`$ (27)
$`\overline{M}`$ $`=`$ $`g_ML^{\beta /\nu }`$ (28)
where $`\beta =1/T`$ and $`g`$ are constants not dependent on size $`L`$. We will not use $`\chi `$ to determine $`\gamma /\nu `$ but $`\chi _2`$ (25) and $`\chi _4`$ (26) using
$$\gamma _4/\nu =d+\mathrm{\hspace{0.17em}2}\gamma /\nu ,$$
(29)
since the errors are smaller.
To find the critical temperature $`T_c`$ we record the variation of $`U`$ with $`T`$ for various system sizes and then locate $`T_c`$ as the intersection of these curves , since the ratio of $`U`$ for two different lattice sizes $`L`$ and $`L^{}=bL`$ should be 1 at $`T_c`$, that is
$$\frac{U_{bL}}{U_L}|_{T=T_c}=\mathrm{\hspace{0.25em}1}.$$
(30)
Due to the presence of residual corrections to finite size scaling, actually one has to extrapolate the results taking the limit (ln$`b`$)$`{}_{}{}^{1}0`$ (Fig. 6, 6, 8 and 8).
### 4.2 First order transitions
A first order transition has a different scaling behavior .
* The histogram $`P(E)`$ should have a double peak.
* Magnetization and energy should show a hysteresis.
* The minimum of the fourth order energy cumulant $`W=1E^4/(3E^2^2)`$ varies as
$$W=W^{}+bL^d\mathrm{and}W^{}\frac{2}{3}.$$
(31)
A double peak in P(E) means that at least two states with different energies coexist at the same temperature. As a consequence the fourth order energy cumulant cannot be $`\frac{2}{3}`$. This fact was employed for the smaller sizes in a preliminary study . Hysteresis effects have to be expected in the simulation of larger systems where the two peaks are well separated since the transition time from one state to the other grows exponentially with the system size. For frustrated antiferromagnets we are studying these criteria are not very helpful. Only for the trihedral model the hysteresis is clearly visible in Fig. 10, indicating a strong first order transition.
## 5 Results
### 5.1 STAR model
Using (30) we first determine $`T_c`$. $`U`$ is plotted for different sizes from $`L=18`$ to $`L=24`$ as a function of temperature in Fig. 6. From the intersections we extrapolate $`T_c`$ as
$$T_c=1.43122(12),$$
(32)
see Fig. 6. The estimate for the universal quantity $`U^{}`$ at the critical temperature is
$$U^{}=0.6269(10).$$
(33)
With the value of $`T_c`$ we determine the critical exponents by log–log fits. We obtain $`\nu `$ from $`V_1`$ (Fig. 10), $`\gamma /\nu `$ and $`\gamma _4/\nu `$ from $`\chi _2`$ and $`\chi _4`$ (Fig. 10, not shown), and $`\beta /\nu `$ from $`\overline{M}`$ (Fig. 10):
$`\nu `$ $`=`$ $`0.504(10)`$ (34)
$`\gamma /\nu `$ $`=`$ $`2.131(13)`$ (35)
$`\gamma _4/\nu `$ $`=`$ $`7.252(29)`$ (36)
$`\beta /\nu `$ $`=`$ $`0.439(8).`$ (37)
The uncertainty of $`T_c`$ is included in the estimation of the errors. The value of $`\gamma /\nu =2.126(14)`$ found from $`\gamma _4/\nu `$ using (29) is compatible with the one found directly. Combining these results we obtain $`\beta =0.221(9)`$, $`\gamma =1.074(29)`$ and from the scaling relation
$$\gamma /\nu =2\eta $$
(38)
we obtain $`\eta =0.131(13)`$. The results are summarized in table 4.
A negative value of $`\eta `$ is impossible, for a second order phase transition it should always be positive . A slightly negative value was already discernible in some of the results of the unconstrained STA–model, see table 1. In making use of our previous analysis of the $`x`$$`y`$ case we conclude that the critical behavior should be described by the renormalization flow of a fixed point in the complex parameter space. The scaling relation for $`\eta `$ above is then modified to give a positive value for this critical exponent, see previous section. Moreover this large negative value cannot come from only the presence of a complex fixed point. Thus we interpret the negative value as a crossover from the basin of attraction of the complex fixed point to the true first order transition, see Fig. 12
In a previous study of the STAR model, Dobry and Diep obtained for the exponent $`\nu =0.440(20)`$ which disagrees with ours 0.504(10). However this seems to be a misprint since the inverse value is also given and $`\nu =1/2.08=0.48`$ is within the statistical errors. However the other exponents seem wrong. We repeated these calculations for larger clusters and with better statistics in order to have evidence that this system shows a weak first order transition, since unambiguously $`\eta `$ is really negative
### 5.2 Stiefel’s $`V_{3,2}`$ model
Following the same procedure as for the STAR model we determine first $`T_c`$. The cumulant $`U`$ plotted as a function of temperature for different system sizes from $`L=15`$ to $`L=40`$ is shown in Fig.8. The extrapolation, shown in Fig. 8, for $`L=20`$ and $`L=25`$ agrees rather well; the little difference for $`L=15`$ indicates that this lattice size is not yet sufficient for a strictly linear extrapolation. We obtain
$$T_c=1.5312(1)$$
(39)
and for the universal quantity $`U`$ at $`T_c`$
$$U^{}=0.6326(12).$$
(40)
In plotting the logarithm of $`V_1`$, $`\chi _2`$, $`\chi _4`$ and $`\overline{M}`$ as function of the logarithm of temperature difference $`(TT_c)/T_c`$, shown in Fig. 10, one finds from the slope
$`\nu `$ $`=`$ $`0.507(8)`$ (41)
$`\gamma /\nu `$ $`=`$ $`2.240(10)`$ (42)
$`\gamma _4/\nu `$ $`=`$ $`7.480(22)`$ (43)
$`\beta /\nu `$ $`=`$ $`0.381(6).`$ (44)
From $`\gamma _4/\nu `$ using (29) we get $`\gamma /\nu =2.240(11)`$ which is the same value found directly from $`\chi _2`$. Further we obtain for $`\beta =0.193(4)`$ and $`\gamma =1.136(23)`$ and with the scaling relation (38) an even more negative value $`\eta =0.240(10)`$. See for a listing table 4, where also the specific heat exponent $`\alpha =2d\nu `$ has been added. The errors given include the uncertainty of the estimate of $`T_c`$.
The results of Kunz and Zumbach are also given there and are compatible with ours. Their $`\alpha `$ is actually determined by fitting specific heat data in the high temperature region and $`\nu `$ with the hyperscaling relation. They noticed also that $`\eta `$ is negative, apparently their value $`0.10(5)`$ is too small, compared to $`0.24(1)`$ we found. We cannot share their conclusion that the negative $`\eta `$ is due to a strong finite size effect not properly taken care of. We rather take it as an indication that an analysis as a genuine second order transition leads to inconsistencies and interpret the negative value as a crossover from the basin of attraction of the complex fixed point to the true first order transition, see Fig. 12.
### 5.3 Direct trihedral model
It is known from that the energy distribution $`P(E)`$ near the critical temperature has a double peak structure. A rough estimate of the correlation length $`\xi _0`$ by $`\frac{1}{3}`$ of the smallest size where the two peaks are well separated by a region of zero probability gives $`\xi _0=6`$. Here we determine the energy cumulant at the transition. Extrapolating $`W`$ to a system of infinite size we obtain
$$W^{}=\mathrm{\hspace{0.17em}0.623}(1)$$
(45)
which of course deviates from the value $`\frac{2}{3}`$ for a second order transition. It is not difficult to see hysteresis effects. In Fig. 10 hysteresis for the energy $`E`$ is clearly visible.
Therefore the behavior of a system of trihedrals differs from the one of dihedrals. In weakening the interaction between the third components of the trihedrals one could of course recover the apparent second order nature of the dihedral system. The first order character of the transition of pure dihedrals is impossible to show directly. This system remains always under the influence of the virtual or complex fixed point for the sizes of systems one can simulate, while the presence of the third vector allows the system of trihedrals to stay outside of the neighborhood of this fixed point and the “true” first order behavior is seen directly.
## 6 Discussion
### 6.1 Summary
The STAR and the Stiefel model $`V_{3,2}`$ have slightly different critical exponents. Looking at table 4, most noticeable are the differences for $`\beta `$ and $`\gamma `$: $`\beta _{STAR}=0.221(9)`$ and $`\beta _{Stie}=0.193(4)`$, outside the estimated statistical and systematic errors. The difference to the values of $`\beta 0.286`$ for the original stacked triangular antiferromagnet in table 1 is even larger. Also the non–collinear antiferromagnet on the bct lattice has a different $`\beta `$–value (see table 1). Using the scaling relation $`\eta =2d+2\beta /\nu `$ the exponent $`\eta `$ is always negative. Moreover, a member of the same universality class, the right handed trihedral model, has a strong first order transition. This does not correspond to the standard behavior of a universality class one expects from the seemingly clear evidence for a second order transition as visible in Fig. 10 for the STAR and the Stiefel model.
In taking the point of view that the simulations show an extremely weak first order transition it is very natural to use a field theoretical description with the usual renormalization group scheme supplemented by complex fixed points, see Fig. 2. Such complex fixed points exist if the number of components $`N`$ is lower than a critical one $`N_c`$, and according to field theoretical analysis the critical dimension is $`N_c4`$. As for a real fixed point the renormalization flow will be attracted and the system imitates a second order behavior, however it can escape if the system size $`L`$ is larger than the largest correlation length $`\xi _0`$ possible and then the crossover to first order region is reached. This crossover phenomenon is known to occur in the two dimensional Potts model with $`q=5`$ components .
In the analysis based on complex fixed points Zumbach showed that the scaling relations should be modified by correction terms eqs. (4-5). Leaving out these terms the scaling relations for $`\eta `$ produces negative values, for $`XY`$ spins of the STA model it is slightly negative ($`\eta 0.06`$ ) and for the models studied here the effect is two or three times larger, see table 4. Another possibility of obtaining negative values for $`\eta `$ is to visualize the spin system in the crossover region between a second to first order transition. Using the same scaling relations as before the exponent $`\eta `$ will tend to the value $`\eta =1`$ of a “weak first order transition” (table 3). Our MC simulations of the STAR and the Stiefel model are in such a crossover region otherwise the very negative values for $`\eta 0.2`$ cannot be understood. The same applies for MC simulations of helimagnetics on bct lattices .
In contrast to the $`XY`$ case with $`N=2`$ the frustrated Heisenberg case is closer to the critical dimension. That is, the STAR and the Stiefel model appear at first sight as second order transitions whereas in the $`XY`$ case they show really first order behavior . The stacked triangular antiferromagnet STA follows the same trend with a small negative $`\eta `$ in the $`XY`$ case and an only slightly negative one in the Heisenberg case table 1. The smaller sphere of influence or basin of attraction of the complex fixed point in the $`XY`$ case compared to the Heisenberg case is depicted in Fig. 12 and Fig. 12.
Another interpretation following Kawamura takes the diagram in Fig. 2a as basis, where an attractive $`F_+`$ and an unstable fixed point $`F_{}`$ are neighbors. The initial points in the RG flow for the STAR and the Stiefel model are placed outside the domain of attraction of the fixed point $`F_+`$, that is, to the left of the line ($`G,F_{}`$) in Fig. 2a while the STA model should be to the right of this line. Therefore the STAR and the dihedral model have a first order transition and the STA model a standard second order transition. The difficulty is the critical dimension $`N_c>3`$ according to Antonenko and Sokolov and actually Fig 2c should be the starting point. Beyond all question is that all the MC simulations give a negative $`\eta `$ impossible to explain with a real fixed point . One observes also the experimental crossover behavior from second order to first order in systems belonging to the $`XY`$ class like the holmium, dysprosium or $`CsCuCl_3`$ (see ) which is in disfavor of this interpretation.
### 6.2 Experiments
In real systems because of the omnipresence of planar or axial anisotropies the behavior of frustrated Heisenberg antiferromagnets will be difficult to observe. Moreover such systems are usually quasi one or two dimensional, that is a succession of crossovers will occur from 2d to 3d and from Heisenberg to Ising or $`XY`$ behavior. To observe the first order behavior the temperature must be very close to the critical temperature in order to notice that the correlation length is limited in the basin of attraction of the complex fixed point $`F`$. For $`XY`$ spin systems, in spite of this limitation the crossover has been observed in certain materials like Ho , Dy , CuCoCl<sub>3</sub> (see also ).
Before the first order region is reached in Heisenberg systems the crossover to Ising or $`XY`$ behavior prevents a first order transition of Heisenberg type. In Fig. 13 a typical scheme is drawn for the crossovers and transitions of a system with Ising anisotropy, for more details see and for other crossovers .
Nevertheless the second order transition can be studied. A list of results is in Table 2: for VCl<sub>2</sub> , for VBr<sub>2</sub> , for Cu(HCOO)<sub>2</sub>2CO(ND<sub>2</sub>)<sub>2</sub>2D<sub>2</sub> and for Fe\[S<sub>2</sub>CN(C<sub>2</sub>H<sub>5</sub>)<sub>2</sub>\]<sub>2</sub>Cl . For the last two examples the observed exponents might be influenced by the crossover between 2d to 3d Ising behavior. The experimental results agree quite well with the MC simulation in table 1.
### 6.3 Renormalization group studies in 2+$`ϵ`$
For the STA model studied in $`d=2+ϵ`$ expansion using the corresponding nonlinear $`\sigma `$–model Azaria et al. obtained for all $`N`$ a stable fixed point $`F_{2+ϵ}`$ at a small distance $`ϵ`$ from $`d=2`$. In comparing results in lowest order of $`1/N`$ for the $`2+ϵ`$ to the $`4ϵ`$ expansion they show that the fixed points found in the different approaches are equivalent for large $`N`$.
Specially for $`N=3`$ a clear evidence by symmetry arguments can be given that the transition near $`d=2`$ should be of the standard ferromagnetic $`O(4)`$ class. If the transition is of second order also for $`d=3`$ it must therefore belong to the $`O(4)`$ class with the exponent $`\gamma _{O4}=1.54`$ (table 3) very different from the value of the STA model $`\gamma _{STA}=1.18`$ found by the MC method (table 1). This is not compatible with the Monte Carlo simulations and experimental studies (see tables).
We think that we can rule out the possibility of a ferromagnetic $`O(4)`$ class for $`N=3`$ if the relevant operators are the same for large and low $`N`$. The $`1/N`$ expansion for $`\gamma `$ (similar arguments hold for all exponents) for both the ferromagnetic non frustrated (NF) case and frustrated case (F) are:
$`\gamma _{NF}`$ $`=`$ $`{\displaystyle \frac{2}{d2}}\left[\mathrm{\hspace{0.17em}1}{\displaystyle \frac{6}{N}}S_d\right]`$ (46)
$`\gamma _F`$ $`=`$ $`{\displaystyle \frac{2}{d2}}\left[\mathrm{\hspace{0.17em}1}{\displaystyle \frac{9}{N}}S_d\right]`$ (47)
where
$$S_d=\frac{\mathrm{sin}(\pi (d2)/2)\mathrm{\Gamma }(d1)}{2\pi \mathrm{\Gamma }(d/2)^2}$$
(48)
is a positive quantity. Thus for $`N`$ large $`\gamma _{NF}>\gamma _F`$. The reason is the different breakdown of symmetry from $`O(N)`$ to $`O(NP)`$ with $`P=1`$ in the ferromagnetic case and $`P=2`$ in the frustrated case. When $`N`$ decreases the ratio $`P/N`$ increases also the difference between $`\gamma _{NF}`$ and $`\gamma _F`$ will increase and one should have $`\gamma _{NF}>\gamma _F`$ for any $`N`$ and in particular for $`N=3`$.
However it is possible that irrelevant operators in the $`4ϵ`$ and $`1/N`$ expansions become relevant near dimension two and low $`N`$. In this case for $`N=3`$ the transition should be of first order for dimension three and a little less than three. Below this limit the transition could be of the ”chiral” universality class and in approaching two dimensions it should become a $`O_4`$ transition. This hypothesis has found some support . We note that at exactly two dimensions, systems are known to be driven by the $`O(4)`$ symmetry, at least at low temperature .
Other RG studies near two dimensions have found more than a single transition governed by the $`O(4)`$ fixed point with an intermediate “nematic” phase between the disordered high temperature region and the ordered $`120^{}`$ structure at low temperatures. This double transition should also occur for the right handed trihedral model studied here when the interaction between the third components are larger than between the two others. The ordering of the third vector occurs before the ordering of the other two vectors and both transitions are of second order, the first of Heisenberg type and the second of $`XY`$ type. The intermediate “nematic” phase can also be interpreted as a ferromagnetic phase and the two transitions characterized as paramagnetic–collinear and collinear–noncollinear.
Another possibility discussed in the literature is that the transition is influenced by the presence of topological defects which are not visible in the continuum and perturbation theory as in RG ($`2+ϵ`$, $`4ϵ`$ and directly in $`d=3`$) . In this work we have assumed that the effects of topological defects are irrelevant for the critical behavior .
## 7 Conclusion
We have tried to discuss the phase transition of frustrated Heisenberg spin systems in general terms. The starting point is Monte Carlo simulations of systems for which the condition of local rigidity is imposed. From the finite size analysis we suggest that the transition is of first order. Our result is that one can rely on the renormalization group studies and the true behavior is indeed first order. However one should use the concept of complex fixed points to describe the almost second order behavior of frustrated Heisenberg spin systems. In the strict sense there is no “chiral universality class” but in practical terms it exists. We have given reasons why in the usual Monte Carlo simulations and in experiments this first order transition is difficult or even impossible to observe.
## 8 Acknowledgments
This work was supported by the Alexander von Humboldt Foundation. The authors are grateful to Professors B. Delamotte, G. Zumbach, H.T. Diep, A. Dobry, Y. Holovatch and F. Nogueira for discussions. We thank A.I. Sokolov, E. Brezin and J. Zinn-Justin for the reference of the proof of $`\eta 0`$, S. Thoms for the reference of the two loop order calculation for superconductors and A. Schakel for the reference of the gauge dependence of $`\eta `$ for superconductors. Moreover we want to acknowledge L. Beierlein and M.E. Myer to a carefully reading of the manuscript. |
warning/0001/math0001077.html | ar5iv | text | # Irreducible degenerations of primary Kodaira surfaces
## Introduction
A smooth compact complex surface with trivial canonical bundle is a K3 surface, a 2-dimensional complex torus, or a primary Kodaira surface. Normal crossing degenerations of such surfaces have attracted much attention. For example, Kulikov analyzed projective degenerations of K3 surfaces . His results were generalized by Persson and Pinkham to degenerations of surfaces with trivial canonical bundle whose central fiber has algebraic components . Conversely, Friedman characterized the singular K3 surfaces from Kulikov’s list that deform to smooth K3 surfaces by his notion of $`d`$-semistability .
For non-Kähler degenerations no general results seem to be known. We chose primary Kodaira surfaces as our object of study because they lack some complications that the more interesting classes of tori and K3 surfaces have. On the other hand, Kodaira surfaces have a lot in common with tori and, via the Kummer construction, with K3 surfaces. The phenomena one sees in degenerations of Kodaira surfaces should therefore also be observable in the other classes of surfaces with trivial canonical bundle.
By a result of Borcea the moduli space $`𝔎`$ of smooth Kodaira surfaces is isomorphic to a countable union of copies of $`\times \mathrm{\Delta }^{}`$ . The parameters correspond to the $`j`$-invariant of the elliptic base and a refined $`j`$-invariant of the fiber. As $`\mathrm{\Delta }^{}`$ can not be completed to a closed Riemann surface by adding finitely many points, it is clear that by studying degenerations we can at most hope for a partial completion of $`𝔎`$. The explicit form of the moduli space also suggests the existence of families leading to a degeneration of the base or the fiber to a nodal elliptic curve. We will see that this is indeed the case. This leads to a partial completion $`\times \mathrm{\Delta }^{}^1\times \mathrm{\Delta }(\mathrm{},0)`$ of each component of $`𝔎`$. However, as elliptic curves can degenerate to any $`k`$-cycle of rational curves, this is not the full story. Rather we obtain a whole hierarchy of such completions, that are linked by non-Hausdorff phenomena at the boundary. Moreover, we will see that at the most interesting point $`(\mathrm{},0)`$ the picture becomes complicated. We were not able to fully clarify what happens there. If one restricts to normal crossing surfaces one should certainly blow up this point. On the other hand, we will also make the presumably not so surprising observation that, as in the case of abelian varieties , it does not suffice to restrict to normal crossing varieties. We did however find families of generalized Kodaira surfaces mapping properly to $`\mathrm{\Delta }\times ^1`$, whose singularities are at most products of normal crossing singularities.
The bulk of the paper is concerned with a classification of irreducible, $`d`$-semistable, locally normal crossing surfaces $`X`$ with $`K_X=0`$. Three different types occur. Our main result is a description of the resulting completion $`𝔎\overline{𝔎}`$, derived by deformation theory. The three different types correspond to three different parts in the boundary $`𝔅=\overline{𝔎}𝔎`$. Each part is a countable union of copies of $`\mathrm{\Delta }^{}`$ or $``$. Locally along the boundary divisor the completion $`𝔎\overline{𝔎}`$ looks like the blowing-up of $`(\mathrm{},0)^1\times \mathrm{\Delta }`$, with two points on the exceptional divisor removed.
Some examples for degenerations of Kodaira surfaces have previously been given by Friedman and Shepherd-Barron in .
This article is divided into six sections. The first section contains general facts about smooth Kodaira surfaces. In the second section we describe their potential degenerations and show that three types are possible. Sections 3–5 contain an analysis of each type. In the final section we assemble our results in terms of moduli spaces, complemented by some examples of smoothable surfaces with singularities that are products of normal crossings.
We thank the referee for suggestions concerning Theorem 3.10 and the interpretation of the completed moduli space after Proposition 6.1.
This paper is dedicated to Hans Grauert on the occasion of his 70th birthday. The second author wants to take this opportunity to express his gratitude for the support and mathematical stimulus he received from him as one of his last students. It was a great pleasure to learn from him.
## 1. Smooth Kodaira surfaces
In this section we collect some facts on primary Kodaira surfaces. Suppose $`B,E`$ are two elliptic curves, and endow $`E`$ with a group structure. Let $`f:XB`$ be a holomorphic principal $`E`$-bundle. The canonical bundle formula gives $`K_X=0`$, so the Kodaira dimension is $`\kappa (X)=0`$. As a topological space, $`X`$ is the product of the $`1`$-sphere with a $`1`$-sphere-bundle $`g:MB`$. Let $`e(g)H^2(B,)`$ be its Euler class. The homological Gysin sequence
$$H_2(B,)\stackrel{e(g)}{}H_0(B,)H_1(M,)H_1(B,)0$$
and the Künneth formula yield $`H_1(X,)=^3/d`$ with $`d=e(g)[B]`$. We call the integer $`d0`$ the *degree* of $`X`$. Bundles of degree $`d=0`$ are 2-dimensional complex tori. A smooth compact complex surface $`X`$ with an elliptic bundle structure of degree $`d>0`$ is called a *primary Kodaira surface*. For simplicity, we refer to such surfaces as Kodaira surfaces.
Kodaira surfaces have three invariants. The first invariant is the degree $`d>0`$. It determines the underlying topological space. The Universal Coefficient Theorem gives
$$H^1(X,)=^3\text{and}H^2(X,)=^4/d.$$
Hence the degree $`d`$ is also the order of the torsion subgroup of the Néron-Severi group $`\mathrm{NS}(X)=\mathrm{Pic}(X)/\mathrm{Pic}^0(X)`$.
The second invariant is the $`j`$-invariant of $`B`$. It depends only on $`X`$: Since $`b_1(X)=3`$ is odd, $`X`$ is nonalgebraic. Moreover, since $`f:XB`$ has connected fibers it is the algebraic reduction and so the fibration structure does not depend on choices.
The third invariant is an element $`\alpha \mathrm{\Delta }^{}`$. Here $`\mathrm{\Delta }^{}=\{z0<|z|<1\}`$ is the punctured unit disk. The Jacobian $`\mathrm{Pic}_X^0`$ has dimension $`h^1(𝒪_X)=2`$, and the quotient $`\mathrm{Pic}_X^0/\mathrm{Pic}_B^0`$ is isomorphic to $`^{}`$. We have $`h^2(𝒪_X(X_b))=h^0(𝒪_X(X_b))=1`$ for each $`bB`$. So the map on the left in the exact sequence
$$H^1(X,𝒪_X)H^1(X_b,𝒪_{X_b})H^2(X,𝒪_X(X_b))H^2(X,𝒪_X)0$$
is surjective. Thus $`\mathrm{Pic}_X^0/\mathrm{Pic}_B^0\mathrm{Pic}_{X_b}^0`$ is an epimorphism. By semicontinuity, the kernel equals $`\mathrm{NS}(B)=`$ and is generated by a well-defined element $`\alpha \mathrm{\Delta }^{}`$.
According to , p. 145, the principal $`E`$-bundle $`X`$ is an associated fibre bundle $`X=(L0)\times _{^{}}^{}/\alpha `$ for some line bundle $`LB`$ of degree $`d`$. One can check that this $`\alpha `$ agrees with the invariant $`\alpha `$ above. It follows that the isomorphism classes of Kodaira surfaces correspond bijectively to the triples $`(d,j,\alpha )_{>0}\times \times \mathrm{\Delta }^{}`$. Moreover, each Kodaira surface with invariant $`(d,j,\alpha )`$ is the quotient of a properly discontinuous free $`^2`$-action on $`^{}\times ^{}`$ defined by
$$\mathrm{\Phi }(z_1,z_2)=(\beta z_1,z_1^dz_2)\text{and}\mathrm{\Psi }(z_1,z_2)=(z_1,\alpha z_2).$$
Here $`\beta \mathrm{\Delta }^{}`$ is defined as follows: Consider the $`j`$-invariant as a function $`j:H`$ on the upper half plane. It factors over $`\mathrm{exp}:H\mathrm{\Delta }^{}`$, $`\tau \mathrm{exp}(2\pi \sqrt{1}\tau )`$. Now $`\beta =\mathrm{exp}(\tau )`$ with $`j(\tau )=j`$. This uniformization illustrates Borcea’s result on the moduli space of Kodaira surfaces . Moreover, it suggests the application of toric geometry for the construction of degenerations.
The following observation will be useful in the sequel:
###### Lemma 1.1.
Suppose $`XB`$ is a principal $`E`$-bundle of degree $`d0`$. Let $`G`$ be a finite group of order $`w`$ acting on $`X`$. If the induced action on $`B`$ is free, then $`w`$ divides $`d`$, and the quotient $`X/G`$ is a principal $`E`$-bundle of degree $`d/w`$.
###### Proof.
Set $`X^{}=X/G`$ and $`B^{}=B/G`$. Then $`B^{}`$ is an elliptic curve, $`BB^{}`$ is Galois and the induced fibration $`X^{}B^{}`$ is a principal $`E`$-bundle. To calculate its degree $`d^{}`$, we use a characterization of $`d`$ in terms of $`\pi _1(X)`$. Since the universal covering of $`E`$ is contractible the higher homotopy groups of $`E`$ vanish. The beginning of the homotopy sequence of the fibrations $`XB`$, $`X^{}B^{}`$ thus leads to the following diagram of central extensions:
$$\begin{array}{ccccccccc}0& & \pi _1(E)& & \pi _1(X)& & \pi _1(B)& & 0\\ & & \mathrm{id}& & & & & & \\ 0& & \pi _1(E)& & \pi _1(X^{})& & \pi _1(B^{})& & 0.\end{array}$$
Let $`h_1,h_2\pi _1(X^{})`$ map to generators of $`\pi _1(B^{})`$. Since $`[\pi _1(B^{}):\pi _1(B)]=w`$ there exists an integral matrix $`\left(\genfrac{}{}{0pt}{}{ab}{cd}\right)`$ with determinant $`w`$ such that
$$g_1=ah_1+bh_2,g_2=ch_1+dh_2$$
map to generators of $`\pi _1(B)`$. Then $`g_1,g_2\pi _1(X)\pi _1(X^{})`$ and $`[g_1,g_2]`$, $`[h_1,h_2]`$ are generators for the commutator subgroups $`[\pi _1(X),\pi _1(X)]`$ and $`[\pi _1(X^{}),\pi _1(X^{})]`$ respectively. Now the degrees $`d,d^{}`$ being the orders of the torsion subgroups of $`H_1(X,)=\pi _1(X)/[\pi _1(X),\pi _1(X)]`$ and of $`H_1(X^{},)=\pi _1(X^{})/[\pi _1(X^{}),\pi _1(X^{})]`$ they can be expressed as divisibility of a generator of the commutator subgroup. Write $`[h_1,h_2]=d^{}x`$ for some primitive $`x\pi _1(E)`$. Then
$$[g_1,g_2]=[ah_1+bh_2,ch_1+dh_2]=(adbc)[h_1,h_2]=wd^{}x$$
shows $`d=wd^{}`$ as claimed.∎
## 2. $`D`$-semistable surfaces with trivial canonical class
Our objective is the study of degenerations of smooth Kodaira surfaces. We consider the following class of singular surfaces:
###### Definition 2.1.
A reduced compact complex surface $`X`$ is called *admissible* if it is irreducible, has locally normal crossing singularities, is $`d`$-semistable, and satisfies $`K_X=0`$.
The sheaf of first order deformations $`𝒯_X^1=\text{xt}^1(\mathrm{\Omega }_X^1,𝒪_X)`$ is supported on $`D=\mathrm{Sing}(X)`$. Following Friedman , we call a locally normal crossing surface $`X`$ *d-semistable* if $`𝒯_X^1𝒪_D`$. This is a necessary condition for the existence of a global smoothing with smooth total space.
Smooth admissible surfaces are either K3 surfaces, 2-dimensional complex tori, or Kodaira surfaces. Throughout this section, $`X`$ will be a *singular* admissible surface. We will see that three types of such surfaces are possible. A finer classification is deferred to subsequent sections.
Let $`\nu :SX`$ be the normalization, $`CS`$ its reduced ramification locus, $`DX`$ the reduced singular locus, and $`\phi :CD`$ the induced morphism. Then $`S`$ is smooth. The surface $`X`$ can be recovered from the commutative diagram
$$\begin{array}{ccc}C& & S\\ \phi & & \nu & & \\ D& & X,\end{array}$$
which is cartesian and cocartesian. It gives rise to a long exact Mayer-Vietoris sequence
$$\mathrm{}H^p(X,𝒪_X)H^p(S,𝒪_S)H^p(D,𝒪_D)H^p(C,𝒪_C)\mathrm{}.$$
The ideals $`𝒪_S(C)𝒪_S`$ and $`_D=\nu _{}(𝒪_S(C))𝒪_X`$ are the conductor ideals of the inclusion $`𝒪_X\nu _{}(𝒪_S)`$. They coincide with the relative dualizing sheaf
$$\nu _{}(\omega _{S/X})=\mathrm{Hom}(\nu _{}(𝒪_S),𝒪_X).$$
Hence $`K_S=C`$; in particular, the Kodaira dimension is $`\kappa (S)=\mathrm{}`$. The Enriques-Kodaira classification of surfaces (, Chap. VI) tells us that $`S`$ is either ruled or has $`b_1(S)=1`$. In the latter case, one says that $`S`$ is a surface of *class VII*.
Following Deligne and Rapoport we call the seminormal curve obtained from $`^1\times /n`$ by the relations $`(0,i)(\mathrm{},i+1)`$, a Néron polygon.
###### Lemma 2.2.
Each connected component of $`CS`$ is an elliptic curve or a Néron polygon. For each singular connected component $`D^{}D`$, the number of irreducible components in $`\nu ^1(D^{})C`$ is a multiple of 6.
###### Proof.
The adjunction formula gives $`\omega _C=𝒪_C`$. Obviously, $`C`$ has ordinary nodes. By , Lemma 1.3, each component of $`C`$ is an elliptic curve or a Néron polygon. Let $`s`$ be the number of singularities and $`c`$ the number of irreducible components in $`D^{}`$. The preimage $`\nu ^1(D^{})S`$ is a disjoint union of Néron polygons. It has $`2c`$ irreducible components and $`3s`$ singularities. For Néron polygons, these numbers coincide. ∎
We will use below the following triple point formula for $`d`$-semistable normal crossing surfaces:
###### Lemma 2.3.
(, Cor.2.4.2) Let $`D^{}D`$ be an irreducible component and $`C_1^{}C_2^{}C`$ its preimage in $`S`$. Then $`(C_1^{})^2(C_2^{})^2`$ is the number of triple points of $`D^{}`$.
Let $`g:SS^{}`$ be a minimal model and $`C^{}=g(C)`$. Then $`K_S^{}=C^{}`$, so $`C^{}S^{}`$ has ordinary nodes, and $`g:SS^{}`$ is a sequence of blowing-ups with centers over $`\mathrm{Sing}(C^{})`$. Recall that $`S`$ is a *Hopf surface* if its universal covering space is isomorphic to $`^20`$.
###### Proposition 2.4.
Suppose $`S`$ is nonalgebraic. Then $`S`$ is a Hopf surface. The curve $`C=C_1C_2`$ is a disjoint union of elliptic curves, and $`X`$ is obtained by gluing them together.
###### Proof.
Since $`S`$ has algebraic dimension $`a(S)<2`$, no curve on $`S`$ has positive selfintersection. Suppose $`C`$ contains a Néron polygon $`C^{}=C_1\mathrm{}C_m`$. As $`X`$ is $`d`$-semistable, the triple point formula Lemma 2.3 implies $`C_i^2=1`$. So $`(C^{})^2=C_i^2+_{ij}C_iC_j=m+2m>0`$, contradiction. Hence $`C`$ is the disjoint union of elliptic curves. Consider the exact sequence
$$H^0(S,𝒪_S)H^0(C,𝒪_C)H^1(S,\omega _S).$$
By the Kodaira classification, $`b_1(S)=1`$, consequently $`h^1(𝒪_S)=1`$, thus $`C`$ has at most two components. Suppose $`\phi :CD`$ induces a bijection of irreducible components. Then each component of $`C`$ double-covers its image in $`D`$. So the map on the left in the exact sequence
$$H^1(S,𝒪_S)H^1(D,𝒪_D)H^1(C,𝒪_C)H^2(X,𝒪_X)0$$
is surjective, contradicting $`K_X=0`$. Consequently $`C=C_1C_2`$ consists of two elliptic curves which are glued together in $`X`$. Let $`n0`$ be the number of blowing-ups in $`SS^{}`$. The normal bundle of $`C`$ has degree $`K_S^2=K_S^{}^2n=n`$. Again by the triple point formula it follows $`n=0`$. Hence $`S`$ is minimal. By , Prop. 3.1, $`S`$ is a Hopf or Inoue surface. The latter case is impossible here because Inoue surfaces contain at most one elliptic curve (, Prop. 1.5).∎
###### Proposition 2.5.
Suppose $`S`$ is nonrational algebraic. Then $`S`$ is a $`^1`$-bundle over an elliptic curve $`B`$. The curve $`C=C_1C_2`$ is the disjoint union of two sections, and $`X`$ is obtained by gluing them together.
###### Proof.
By the Enriques classification, the minimal model $`S^{}`$ is a $`^1`$-bundle over a nonrational curve $`B`$. By Lüroth’s Theorem, $`C`$ does not contain Néron polygons, hence is the disjoint union of elliptic curves. The Hurwitz formula implies that $`B`$ is also an elliptic curve. Since $`K_S`$ has degree $`2`$ on the ruling, we infer $`C=K_S`$ has at most two irreducible components. As in the preceding proof, we conclude that $`C`$ consists of two components, which are identified in $`X`$. Let $`n0`$ be the number of blowing-ups in $`SS^{}`$. The normal bundle of $`C`$ has degree $`K_S^2=K_S^{}^2n=n`$. Since $`X`$ is $`d`$-semistable, $`n=0`$ follows. Hence $`S`$ is a $`^1`$-bundle. ∎
###### Proposition 2.6.
Suppose $`S`$ is a rational surface. Then it is the blowing-up of a Hirzebruch surface in two points $`P_1,P_2`$ in disjoint fibers $`F_1,F_2`$. The curve $`C`$ is a Néron 6-gon, consisting of the two exceptional divisors, the strict transforms of $`F_1,F_2`$, and the strict transforms of two disjoint sections whose union contains $`P_1`$ and $`P_2`$. The surface $`X`$ is obtained by identifying pairs of irreducible components in $`C`$.
###### Proof.
The minimal model $`S^{}`$ is either $`^2`$ or a Hirzebruch surface of degree $`e1`$. Since each representative of $`K_S^{}`$ is connected the curve $`C`$ must be connected. Suppose $`C`$ is an elliptic curve. Then the map on the left in the exact sequence
$$H^1(S,𝒪_S)H^1(D,𝒪_D)H^1(C,𝒪_C)H^2(X,𝒪_X)0$$
is surjective, contradicting $`K_X=0`$. Consequently, $`C=C_1\mathrm{}C_m`$ is a Néron polygon. The triple point formula gives $`C_i^2=m`$, hence
$$K_S^2=C_i^2+\underset{ij}{}C_iC_j=m+2m=m.$$
Let $`n0`$ be the number of blowing-ups in $`SS^{}`$, so $`K_S^2=K_S^{}^2n`$. For $`S^{}=^2`$ this gives $`m=9n`$. If $`S^{}`$ is a Hirzebruch surface, $`m=8n`$ holds instead. According to Lemma 2.2, the natural number $`m`$ is a multiple of $`6`$. The only possibilities left are $`S^{}=^2`$ and $`C^{}`$ a Néron 3-gon, or $`S^{}`$ a Hirzebruch surface and $`C^{}`$ a Néron 4-gon. From this it easily follows that $`X`$ is obtained from a blowing-up of a Hirzebruch surface as stated. ∎
###### Remark 2.7.
For the results of this section one can weaken the hypothesis that $`X`$ is $`d`$-semistable. It suffices to assume that the invertible $`𝒪_D`$-module $`𝒯_X^1`$ is numerically trivial. One might call such surfaces *numerically d-semistable*. They will occur as locally trivial deformations of $`d`$-semistable surfaces.
## 3. Hopf surfaces
In this section we analyze the geometry and the deformations of admissible surfaces $`X`$ whose normalization $`S`$ is *nonalgebraic*. By Proposition 2.4, such $`S`$ is a Hopf surface containing a union of two disjoint isomorphic elliptic curves $`C=C_1C_2`$.
As a topological space, Hopf surfaces are certain fibre bundles over the 1-sphere, whose fibres are quotients of the 3-sphere (see , Thm. 9, and ). By Hartog’s Theorem, the action of $`\pi _1(S)`$ on the universal covering space $`^20`$ extends to $`^2`$, fixing the origin. We call a Hopf surface *diagonizable* if $`\pi _1(S)`$ is contained in the maximal torus $`^{}\times ^{}\mathrm{GL}_2()`$ up to conjugacy inside the group of all biholomorphic automorphisms of $`^2`$ fixing the origin. This is precisely the class of Hopf surfaces we are interested in:
###### Proposition 3.1.
A Hopf surface $`S`$ is diagonizable if and only if it contains two disjoint elliptic curves $`C_1,C_2S`$ with $`K_S=C_1+C_2`$. Moreover, a diagonizable Hopf surface has fundamental group $`\pi _1(S)=/n`$.
###### Proof.
Suppose $`\pi _1(S)\mathrm{GL}_2()`$ consists of diagonal matrices. Let $`C_1,C_2S`$ be the images of $`^{}\times 0`$ and $`0\times ^{}`$, respectively. The invariant meromorphic 2-form $`dz_1/z_1dz_2/z_2`$ on $`^2`$ yields $`K_S=C_1+C_2`$.
Conversely, assume the condition $`K_S=C_1+C_2`$. An element $`\alpha \pi _1(S)`$ is called a *contraction* if $`lim_n\mathrm{}\alpha ^n(U)=\left\{0\right\}`$ holds for the unit ball $`U`$ in $`^2`$. According to a classical result , equation 44, in suitable coordinates a contraction takes the form
$$\alpha (z_1,z_2)=(\alpha _1z_1+\lambda z_2^m,\alpha _2z_2)$$
with $`0<|\alpha _1||\alpha _2|<1`$ and $`\lambda (\alpha _1\alpha _2^m)=0`$. We claim that each contraction $`\alpha \pi _1(S)`$ has $`\lambda =0`$. Suppose not. The quotient of $`^20`$ by $`\alpha `$ is a primary Hopf surface and defines a finite étale covering $`f:S^{}S`$. Let $`C^{}S^{}`$ be the image of $`^{}\times 0`$. Then , p. 696, gives $`K_S^{}=(m+1)C^{}`$. On the other hand,
$$K_S^{}=f^{}(K_S)=f^1(C_1)+f^1(C_2)$$
holds. Consequently, a nonempty subcurve of $`(m+1)C^{}`$ is base point free, so $`S^{}`$ is elliptic. Now according to , Theorem 31, ellipticity of $`S^{}=(^20)/(\alpha )`$ implies $`\lambda =0`$, contradiction.
Next we claim that $`\pi _1(S)`$ is contained in $`\mathrm{GL}_2()`$, at least up to conjugacy. Suppose not. A nonlinear fundamental group $`\pi _1(S)`$ is necessarily abelian (, Thm. 47). By , p. 231, there is a contraction $`\alpha \pi _1(S)`$ with $`\lambda 0`$, contradiction.
Now we can assume $`\pi _1(S)\mathrm{GL}_2()`$. Write $`\pi _1(S)=GZ`$ as a semidirect product, where $`G=\{\gamma \pi _1(S)|det(\gamma )|=1\}`$, and $`Z`$ is generated by a contraction $`\alpha =(\alpha _1,\alpha _2)`$, acting diagonally. If $`\alpha _1\alpha _2`$ then , p. 24, ensures that $`S`$ is diagonizable. It remains to treat the case $`\alpha _1=\alpha _2`$. Then $`Z`$ is central and we only have to show that $`G`$ is abelian. Let $`S^{}`$ be the quotient of $`^20`$ by $`\alpha `$. The canonical projection $`^20^1`$ induces an elliptic bundle $`S^{}^1`$. This defines an elliptic structure $`g:S=S^{}/G^1/G`$. Suppose $`^1^1/G`$ has more than $`2`$ ramification points. Let $`F_b`$ be the reduced fiber over $`g^1/G`$ with multiplicity $`m_b`$ and $`C_i=F_{b_i}`$; the canonical bundle formula yields
$`0`$ $`=`$ $`C_1+C_2+K_S=F_{b_1}+F_{b_2}2F+{\displaystyle \underset{b^1/G}{}}(m_b1)F_b`$
$`=`$ $`(1m_{b_1})F_{b_1}+(1m_{b_2})F_{b_2}+{\displaystyle \underset{b^1/G}{}}(m_b1)F_b>0,`$
contradiction. Consequently there are only 2 ramification points. Therefore $`G`$ is a subgroup of $`^{}`$, hence abelian. This finishes the proof of the equivalence.
As for the fundamental group, it must have rank 1 for $`b_1=1`$. If there were more than one generator needed for the torsion part the action could not be free on $`(^{}\times 0)(0\times ^{})`$. ∎
Suppose that $`S`$ is a diagonizable Hopf surface with $`\pi _1(S)/n\mathrm{GL}_2()`$ consisting of diagonal matrices. The free part of $`\pi _1(S)`$ is generated by a contraction $`\alpha =(\alpha _1,\alpha _2)`$ with $`\alpha _1,\alpha _2\mathrm{\Delta }^{}`$. The torsion part is generated by a pair $`\zeta =(\zeta _1,\zeta _2)`$ of primitive $`n`$-th roots of unity. Note that they must be primitive, again since the action is free on the coordinate axes minus the origin. Choose $`\tau _i`$ and $`n_i`$ with $`\alpha _i=\mathrm{exp}(2\pi \sqrt{1}\tau _i)`$ and $`\zeta _i=\mathrm{exp}(2\pi \sqrt{1}n_i/n)`$. Consider the lattice
$$\mathrm{\Lambda }_i=\tau _i,n_i/n,1=\tau _i+1/n.$$
The images $`C_1,C_2S`$ of $`^{}\times 0`$ and $`0\times ^{}`$ take the form
$$C_i=^{}/\alpha _i,\zeta _i=/\mathrm{\Lambda }_i.$$
Without loss of generality we can assume $`C=C_1C_2`$. In fact, if $`S`$ is not elliptic then $`C_1,C_2`$ are the only curves on $`S`$ at all. Otherwise the elliptic fibration comes from a rational function of the form $`z_1^a/z_2^b`$ on $`^20`$. Then $`|ab|1`$ iff the elliptic fibration $`S^1`$ has multiple fibers. In this case the canonical bundle formula implies $`C=C_1C_2`$. Otherwise we can apply a linear coordinate change on $`^2`$ to achieve that the preimages of the components of $`C`$ are $`^{}\times 0`$ and $`0\times ^{}`$.
Now assume there is an isomorphism $`\psi :C_2C_1`$. Set $`D=C_1`$, $`C=C_1C_2`$, and $`\phi =\mathrm{id}\psi :CD`$. The cocartesian diagram
$$\begin{array}{ccc}C& & S\\ \phi & & \nu & & \\ D& & X\end{array}$$
defines a normal crossing surface $`X`$. Since each translation of the elliptic curve $`C_2`$ extends to an automorphism of $`S`$ fixing $`C_1`$, we can assume that $`\psi :C_2C_1`$ respects the origin. Hence $`\psi `$ is defined by a homothety $`\mu :`$ with $`\mu \mathrm{\Lambda }_2=\mathrm{\Lambda }_1`$. In other words, there is a matrix
(3.1)
$$\begin{array}{c}\hfill \left(\begin{array}{cc}a& c\\ b& d\end{array}\right)\mathrm{SL}_2()\end{array}\text{with}\begin{array}{c}\hfill \mu \tau _2=a\tau _1+b/n\\ \hfill \mu /n=c\tau _1+d/n.\end{array}$$
The following two propositions describe the properties of the surface $`X`$ in terms of $`\pi _1(S)\mathrm{GL}_2()`$ and $`\mu `$.
###### Proposition 3.2.
The condition $`K_X=0`$ holds if and only if $`a=d=1`$ and $`c=0`$.
###### Proof.
As $`\nu ^{}(K_X)=0`$ in any case $`K_X`$ is numerically trivial. The condition $`K_X=0`$ is thus equivalent to $`H^2(X,𝒪_X)0`$. Assume $`K_X=0`$. Then the map on the left in the exact sequence
$$H^1(S,𝒪_S)H^1(D,𝒪_D)H^1(C,𝒪_C)H^2(X,𝒪_X)0$$
is not surjective. Hence $`H^1(𝒪_D)`$ and $`H^1(𝒪_S)`$ have the same image in $`H^1(𝒪_C)`$. According to the commutative diagram
$$\begin{array}{ccccc}H^1(S,𝒪_S)& & H^1(C,𝒪_C)& \stackrel{\phi ^{}}{}& H^1(D,𝒪_D)\\ \text{bij}& & \text{inj}& & \text{inj}& & \\ H^1(S,)& & H^1(C,)& \stackrel{\phi ^{}}{}& H^1(D,)\end{array}$$
from Hodge theory, the image of $`H^1(D,)`$ in $`H^1(C,)`$ contains the image of $`H^1(S,)`$. So the composition
$$H^1(S,)H^1(C,)\stackrel{(\mathrm{id},\psi ^{})^1}{}H^1(D,)H^1(D,)$$
factors over the diagonal $`H^1(D,)H^1(D,)^2`$. Dually, the composition
$$H_1(D,)H_1(D,)\stackrel{(\mathrm{id},\psi _{})^1}{}H_1(C,)H_1(S,)$$
factors over the addition map $`H_1(D,)^2H_1(D,)`$. In other words, the composition
$$\mathrm{\Lambda }_1\stackrel{(\mathrm{id},\mathrm{id})}{}\mathrm{\Lambda }_1\mathrm{\Lambda }_1\stackrel{\mathrm{id}\left(\genfrac{}{}{0pt}{}{dc}{ba}\right)}{}\mathrm{\Lambda }_1\mathrm{\Lambda }_2\stackrel{(1,0,1,0)}{}$$
is zero. It is described by $`(1d,c)`$; thus $`d=1`$, $`c=0`$, and in turn $`a=1`$. Hence the condition is necessary. The converse is shown in a similar way. ∎
###### Remark 3.3.
Inserting the above conditions into equation 3.1, one sees that the homothety $`\mu :`$ must be the identity, and then $`\tau _2=\tau _1+b/n`$. In particular $`S`$ is elliptically fibered.
###### Proposition 3.4.
Suppose $`K_X=0`$. Then $`X`$ is $`d`$-semistable if and only if the congruences $`(n_1n_2)^20`$ and $`b(n_1n_2)0`$ modulo $`n`$ hold.
###### Proof.
Let $`N_i`$ be the normal bundle of $`C_iS`$. We can identify $`𝒯_X^1`$ with $`N_2\psi ^{}(N_1)`$. The bundle $`N_1`$ can be obtained as quotient of the normal bundle of $`^{}\times 0`$ in $`^{}\times ^{}`$, which is trivial, by the induced action of $`\pi _1(S)=/n`$: $`N_1=^{}\times ^{}/(\alpha ,\zeta )`$ with
$$\alpha (z_1,v)=(\alpha _1z_1,\alpha _2v)\text{and}\zeta (z_1,v)=(\zeta _1z_1,\zeta _2v).$$
After pull back along $`^{},z\mathrm{exp}(2\pi \sqrt{1}z)`$, the bundle $`N_1`$ can also be seen as the quotient of the trivial bundle $`\times `$ by $`\pi _1(C_1)=\mathrm{\Lambda }_1`$ acting via
$$\tau _1(w,v)=(w+\tau _1,\alpha _2v),n_1/n(w,v)=(w+n_1/n,\zeta _2v),1(w,v)=(w+1,v).$$
Choose $`m_1`$ with $`m_1n_11`$ modulo $`n`$. Then $`1/n\mathrm{\Lambda }_1`$ acts via $`1/n(w,v)=(w+1/n,\zeta _2^{m_1}v)`$. An analogous situation holds for $`N_2`$. Inserting $`\mathrm{\Psi }=\left(\genfrac{}{}{0pt}{}{\mathrm{1\; 0}}{b1}\right)`$ we obtain that $`𝒯_X^1=N_2\psi ^{}(N_1)`$ is the quotient of $`\times `$ by the action
$$\tau _2(w,v)=(\tau _2+w,\alpha _1\alpha _2\zeta _2^{bm_1}v)\text{and}1/n(w,v)=(1/n+w,\zeta _1^{m_2}\zeta _2^{m_1}v).$$
The isomorphism class of $`𝒯_X^1=N_2\psi ^{}(N_1)`$ depends only on the element
$$\tau _2\alpha _1\alpha _2\zeta _2^{bm_1},1/n\zeta _1^{m_2}\zeta _2^{m_1}$$
in $`\mathrm{Hom}_{}(\mathrm{\Lambda }_2,^{})`$. To check for triviality of $`𝒯_X^1`$, consider the commutative diagram
$$\begin{array}{ccccc}\mathrm{Hom}_{}(\mathrm{\Lambda }_2,)& \stackrel{\mathrm{pr}}{}& \mathrm{Hom}_{}(,\overline{})& \stackrel{}{}& H^1(C_2,𝒪_{C_2})\\ \mathrm{exp}& & & & \mathrm{exp}& & \\ \mathrm{Hom}_{}(\mathrm{\Lambda }_2,^{})& & \mathrm{Pic}(C_2)& \underset{}{}& H^1(C_2,𝒪_{C_2}^{}),\end{array}$$
explained in , Sect. I.2. Here $`\mathrm{pr}`$ is the projection of $`\mathrm{Hom}_{}(\mathrm{\Lambda }_2,)\mathrm{Hom}_{}(,)`$ onto the $``$-antilinear homomorphisms $`\mathrm{Hom}_{}(,\overline{})`$. Lifting the homomorphism defining $`N_2\phi ^{}(N_1)`$ from $`\mathrm{Hom}_{}(\mathrm{\Lambda }_2,^{})`$ to $`\mathrm{Hom}_{}(\mathrm{\Lambda }_2,)`$, we obtain
$$\tau _2\tau _1+\tau _2+\frac{bm_1n_2}{n},1/n\frac{m_2n_1+m_1n_2}{n}.$$
Using the diagram, we infer that $`X`$ is $`d`$-semistable if and only if this homomorphism is $``$-linear up to an integral homomorphism. The latter condition is equivalent to the equation
$$(\tau _1+\tau _2+\frac{bm_1n_2}{n}+e)\tau _2n(\frac{m_2n_1+m_1n_2}{n}+f)=0$$
for certain integers $`e,f`$. Now proceed as follows: Substitute $`\tau _1=\tau _2b/n`$; the coefficients of $`\tau _1`$ and $`1`$ give two equations; since $`e`$ and $`f`$ are variable this leads to two congruences modulo $`n`$; finally use $`n_1m_11`$, $`n_2m_21`$ modulo $`n`$ to deduce the stated congruences. ∎
We seek a coordinate free description of $`X`$. We call an automorphism of a Néron 1-gon a *rotation* if the induced action on the normalization $`^1`$ fixes the branch points (compare , Sect. 3.6).
###### Proposition 3.5.
Let $`X`$ be a complex surface. The following are equivalent:
1. $`X`$ is admissible with nonalgebraic normalization $`S`$.
2. There is an elliptic principal bundle $`X^{}B^{}`$ of degree $`e>0`$ over the Néron 1-gon $`B^{}`$ and a (cyclic) Galois covering $`g:X^{}X`$ such that the Galois group $`G`$ acts effectively on $`B^{}`$ via rotations.
###### Proof.
Suppose (ii) holds. The normalization $`S^{}`$ of $`X^{}`$ is an elliptic principal bundle of degree $`e>0`$ over $`^1`$, hence nonalgebraic. Since $`g:X^{}X`$ induces a finite morphism $`g^{}:S^{}S`$, the surface $`S`$ is nonalgebraic as well. It is easy to check $`K_X=0`$ using the assumption that $`G`$ acts via rotations.
Conversely, assume (i). As we have seen, $`S`$ is a diagonizable Hopf surface. The exact sequence
$$0\pi _1(S^{})\pi _1(S)\mathrm{PGL}_2()$$
determines another diagonizable Hopf surface $`S^{}`$ and a Galois covering $`S^{}S`$. The Galois group $`G=\pi _1(S)/\pi _1(S^{})`$ is cyclic of order $`m=n/\mathrm{gcd}(n,n_1n_2,b)`$, and the order of the torsion subgroup in $`\pi _1(S^{})`$ is $`n^{}=\mathrm{gcd}(n,n_1n_2)`$, as simple computations show.
The projection $`^20^1`$ induces an elliptic principal bundle $`f^{}:S^{}^1`$. The Galois group $`G`$ acts effectively on $`^1`$ and fixes $`0,\mathrm{}`$. Let $`C_1^{},C_2^{}S^{}`$ be the fibres over the fixed points. The quotient $`^1/G`$ is a projective line, and we obtain an induced elliptic structure $`f:S^1/G`$. Note that $`f`$ has two multiple fibres of order $`m`$, whose reductions are $`C_1,C_2`$. Next, the canonical isomorphism $`\psi ^{}:C_2^{}C_1^{}`$ gives a normal crossing surface $`X^{}`$. Clearly, it is an elliptic principal bundle of degree $`e=n^{}>0`$ over the Néron 1-gon $`B^{}`$ obtained from $`^1`$ by the relation $`0\mathrm{}`$. The congruences $`(n_1n_2)^20b(n_1n_2)`$ modulo $`n`$ are equivalent to $`n^{}(n_1n_2)0n^{}b`$. A straightforward calculation shows that this ensures that the free $`G`$-action on $`S^{}`$ descends to a (free) action on $`X^{}`$. The quotient is $`X=X^{}/G`$. ∎
###### Remark 3.6.
At this point we are in position to classify the admissible surfaces with nonalgebraic normalization. The elliptic principal bundles $`X^{}B^{}`$ are in one-to-one correspondence with pairs $`(\alpha ,e)\mathrm{\Delta }^{}\times `$, while up to isomorphism the action of $`G`$ on $`B^{}`$ only depends on $`|G|`$. A further discrete invariant belongs to possibly non-isomorphic lifts of the $`G`$-action to $`X^{}`$.
We call the invariants $`e`$ and $`w=|G|`$ of $`X`$ the *degree* and the *warp* of $`X`$ respectively. The congruences in Proposition 3.4 imply that $`w`$ divides $`e`$. This is important for the construction of smoothings:
###### Theorem 3.7.
Suppose $`X`$ is an admissible surface (Definition 2.1) with nonalgebraic normalization, of degree $`e`$ and warp $`w`$. Then $`X`$ deforms to a smooth Kodaira surface of degree $`e/w`$.
###### Proof.
Let $`X^{}X`$ be the Galois covering from Proposition 3.5. It suffices to construct a $`G`$-equivariant deformation of $`X^{}`$.
The deformation of $`X^{}`$ will be obtained as infinite quotient of a toric variety belonging to an infinite fan. Our construction fibers over a similar construction of a smoothing of the Néron 1-gon due to Mumford (, Ch.I,4) that we now recall for the reader’s convenience. Mumford considered the fan generated by the cones
$$\overline{\sigma }_m=(m,1),(m+1,1),m.$$
There is a $``$-action on the associated toric variety $`W`$ belonging to the linear (right-) action by $`\left(\genfrac{}{}{0pt}{}{\mathrm{1\; 0}}{\mathrm{1\; 1}}\right)`$ on the fan. Let $`q:W`$ be the morphism coming from the projection $`\mathrm{pr}_2:^2`$. Note that the central fiber is an infinite chain of rational curves (a Néron $`\mathrm{}`$-gon), while the general fiber is isomorphic to $`^{}`$. Now $``$ acts fiberwise and the action is proper and free over the preimage of $`\mathrm{\Delta }`$. Moreover, $``$ acts transitively on the set of irreducible components of the central fiber. The quotient is the desired deformation of the Néron 1-gon.
For our purpose, consider the infinite fan in $`^3`$ generated by the two-dimensional cones
$$\sigma _m=(m,e\left(\genfrac{}{}{0pt}{}{m}{2}\right),1),(m+1,e\left(\genfrac{}{}{0pt}{}{m+1}{2}\right),1),m.$$
Let $`V`$ be the corresponding 3-dimensional smooth toric variety and $`V`$ the toric morphism belonging to the projection $`\mathrm{pr}_3:^3`$. Now the special fibre $`V_0`$ is isomorphic to a $`^{}`$-bundle over the Néron $`\mathrm{}`$-gon. As neighbouring cones $`\sigma _m`$, $`\sigma _{m+1}`$ are not coplanar, its degree over each irreducible component of the Néron $`\mathrm{}`$-gon is nonzero. A simple computation shows that its degree is $`e`$. This time the $``$-action on the fan is given by the automorphism
$$\mathrm{\Phi }=\left(\begin{array}{ccc}1& e& 0\\ 0& 1& 0\\ 1& 0& 1\end{array}\right)\mathrm{SL}_3().$$
Again the induced $``$-action on the preimage $`UV`$ of $`\mathrm{\Delta }`$ is proper and free. Choose $`\alpha \mathrm{\Delta }^{}`$ such that the fibres of $`X^{}B^{}`$ are isomorphic to $`^{}/\alpha ^w`$. Let $`\mathrm{\Psi }:VV`$ be the automorphism extending the action of $`(1,\alpha ,1)(^{})^3`$. Then $`𝔛^{}=U/\mathrm{\Phi },\mathrm{\Psi }^w`$ is a smooth 3-fold, endowed with a projection $`𝔛^{}`$. The general fibres $`𝔛_t`$, $`t\mathrm{\Delta }^{}`$, are smooth Kodaira surfaces of degree $`e`$ with fibres $`^{}/\alpha ^w`$ and basis $`^{}/t`$. The special fibre $`𝔛_0^{}`$ is isomorphic to $`X^{}`$.
Finally we extend the $`G`$-action on $`X^{}`$ to $`𝔛^{}`$. The automorphism $`\mathrm{\Psi }:VV`$ descends to a free $`G`$-action on $`𝔛^{}`$. Replacing $`\alpha `$ by another primitive $`w`$-th root of $`\alpha ^w`$ if necessary, the induced action on $`𝔛_0^{}`$ coincides with the given action on $`X^{}`$. Consequently, $`𝔛=𝔛^{}/G`$ is the desired smoothing of $`X=𝔛_0`$. The action of $`\mathrm{\Psi }`$ on $`B^{}`$ is free. Therefore, according to Lemma 1.1, the general fibres $`𝔛_t`$ are smooth Kodaira surfaces of degree $`e/w`$. ∎
The next task is to determine the versal deformation of $`X`$. We have to calculate the relevant cohomology groups:
###### Lemma 3.8.
For a locally normal crossing surface $`X`$ with $`K_X=0`$ and non-algebraic normalization it holds $`h^0(\mathrm{\Theta }_X)=1`$, $`h^1(\mathrm{\Theta }_X)=1`$ and $`h^2(\mathrm{\Theta }_X)=0`$.
###### Proof.
The Ext spectral sequence together with a local computation gives
$$H^p(X,\mathrm{\Theta }_X)=\mathrm{Ext}^p(\mathrm{\Omega }_X^1/\tau _X,𝒪_X),$$
see , page 88ff. Here $`\tau _X\mathrm{\Omega }_X^1`$ denotes the torsion subsheaf. In view of Serre duality
$$\mathrm{Ext}^p(\mathrm{\Omega }_X^1/\tau _X,𝒪_X)^{}H^{2p}(X,\mathrm{\Omega }_X^1/\tau _X)$$
it thus suffices to compute the cohomology of $`\mathrm{\Omega }_X^1/\tau _X`$. Consider the exact sequence
$$0\mathrm{\Omega }_X^1/\tau _X\mathrm{\Omega }_S^1\mathrm{\Omega }_D^10$$
of $`𝒪_X`$-modules (, Prop. 1.5). Note that the map on the right is an *alternating sum*. The inclusion $`H^0(X,\mathrm{\Omega }_X^1/\tau _X)H^0(S,\mathrm{\Omega }_S^1)=0`$ yields $`h^2(\mathrm{\Theta }_X)=0`$. Moreover,
$$0H^0(D,\mathrm{\Omega }_D^1)H^1(X,\mathrm{\Omega }_X^1/\tau _X)H^1(S,\mathrm{\Omega }_S^1)$$
and $`h^{1,1}(S)=0`$ gives $`h^1(\mathrm{\Theta }_X)=1`$. With $`b_3(S)=1`$, $`h^1(\omega _S)=h^1(𝒪_S)=1`$ and degeneracy of the Fröhlicher spectral sequence for smooth compact surfaces (, IV, Thm. 2.7) we have $`h^{1,2}(S)=0`$. Now the exact sequence
$$0H^1(D,\mathrm{\Omega }_D^1)H^2(X,\mathrm{\Omega }_X^1/\tau _X)H^2(S,\mathrm{\Omega }_S^1)$$
gives $`h^0(\mathrm{\Theta }_X)=1`$. ∎
###### Proposition 3.9.
Let $`X`$ be as in Theorem 3.7. Then $`dimT_X^0=1`$, $`dimT_X^1=2`$ and $`dimT_X^2=1`$.
###### Proof.
Since $`X`$ is locally a complete intersection the $`E_2`$ term of the spectral sequence $`E_2^{p,q}=H^p(𝒯_X^q)T_X^{p+q}`$ has at most one non-trivial differential, which is $`H^0(𝒯_X^1)H^2(\mathrm{\Theta }_X)`$. The previous proposition shows first degeneracy at $`E_2`$ level and in turn gives the stated values for $`dimT_X^i`$. ∎
###### Theorem 3.10.
$`X`$ is an admissible surface with nonalgebraic normalization, of degree $`e`$ and warp $`w`$. Then the semiuniversal deformation $`p:𝔛V`$ of $`X`$ has a smooth, 2-dimensional base $`V`$. The locally trivial deformations are parameterized by a smooth curve $`V^{}V`$, and $`VV^{}`$ corresponds to smooth Kodaira surfaces of degree $`e/w`$.
###### Proof.
Since $`h^1(\mathrm{\Theta }_X)=1`$ and $`h^2(\mathrm{\Theta }_X)=0`$, the locally trivial deformations are unobstructed, and $`V^{}`$ is a smooth curve. Since $`dimT_X^1=2`$ and since $`X`$ deforms to smooth Kodaira surfaces of degree $`e/w`$, which move in a 2-dimensional family, the base $`V`$ is a smooth surface. We saw in the proof of the previous proposition that the restriction map $`T^1H^0(X,𝒯_X^1)`$ is surjective. According to , Proposition 2.5, the total space $`𝔛`$ is smooth. Now Sard’s Lemma implies that the projection $`𝔛V`$ is smooth over $`VV^{}`$, at least after shrinking $`V`$. ∎
###### Remark 3.11.
The referee pointed out that $`V^{}`$ should have an interpretation as versal deformation space of the elliptic curve $`DX`$. This is indeed the case: Since the restriction of the Kodaira-Spencer map to $`T_V^{}`$ generates $`H^1(\mathrm{\Theta }_X)`$ it suffices to show that the composition
(3.2) $`H^1(\mathrm{\Theta }_X)H^1(\mathrm{\Theta }_X𝒪_D)H^1(\mathrm{\Theta }_D)`$
is surjective. This statement is stable under étale covers, so we may assume that there is an elliptic fibration $`p:XB`$ over the Néron 1-gon (Proposition 3.5). The double curve $`D`$ is the fiber over the node of $`B`$, and by relative duality $`R^1p_{}(\mathrm{\Theta }_{X/B})=𝒪_B`$. The base change theorem implies that the restriction map $`R^1p_{}(\mathrm{\Theta }_{X/B})H^1(\mathrm{\Theta }_D)`$ is surjective. On global sections this map is nothing but the composition of
$$=H^0(B,R^1p_{}(\mathrm{\Theta }_{X/B}))H^1(X,\mathrm{\Theta }_X)$$
with (3.2). Therefore the latter map is surjective too.
## 4. Ruled surfaces over elliptic curves
In this section we study admissible surfaces $`X`$ whose normalization $`S`$ is *algebraic and nonrational*, as in Proposition 2.5.
Let $`B^{}`$ be an elliptic curve and $`f:SB^{}`$ a $`^1`$-bundle with two disjoint sections $`C_1,C_2S`$. Put $`C=C_1C_2`$ and $`D=C_1`$. Let $`\psi :C_2C_1`$ be an isomorphism and $`\phi =\mathrm{id}\psi :CD`$ the induced double covering. The cocartesian diagram
$$\begin{array}{ccc}C& & S\\ \phi & & \nu & & \\ D& & X\end{array}$$
defines a normal crossing surface $`X`$. By abuse of notation we write $`\psi `$ also for the induced automorphism $`(f|_{C_1})\psi (f|_{C_2})^1`$ of $`B^{}`$.
###### Proposition 4.1.
We have $`K_X=0`$ if and only if $`\psi :B^{}B^{}`$ is a translation.
###### Proof.
As in the proof of Proposition 3.2, $`K_X`$ is trivial iff the images of $`H^1(S,𝒪_S)`$ and $`H^1(D,𝒪_D)`$ in $`H^1(C,𝒪_C)`$ coincide. Notice that the map $`\varphi ^{}:H^1(𝒪_D)H^1(𝒪_C)H^1(𝒪_D)H^1(𝒪_D)`$ is the diagonal embedding. It does not depend on $`\varphi `$.
Suppose $`\psi `$ is a translation. Then the images of $`H^1(𝒪_S)`$ and $`H^1(𝒪_D)`$ both agree with $`(f|_C)^{}H^1(𝒪_B^{})`$.
Conversely, assume that $`\psi `$ is not a translation. Then it acts on $`H^1(𝒪_E)`$ as a nontrivial root of unity. Consequently, $`H^1(𝒪_S)`$ and $`H^1(𝒪_C)`$ have different images in $`H^1(𝒪_D)`$. ∎
The next task is to calculate the sheaf of first order deformations $`𝒯_X^1`$. We call the degree $`e=\mathrm{min}\left\{A^2AS\text{ a section}\right\}`$ of $`S`$ also the *degree* of $`X`$. In our case $`e0`$, since $`SB^{}`$ has two disjoint sections. Let $`w`$ be the order, possibly $`0`$, of the automorphism $`\psi :B^{}B^{}`$. Call it the *warp* of $`X`$.
###### Proposition 4.2.
Suppose $`K_X=0`$. Then $`X`$ is $`d`$-semistable if and only if the warp $`w`$ divides the degree $`e`$.
###### Proof.
We identify $`C_1,C_2`$ with $`B^{}`$ via $`f:SB^{}`$. Set $`_i=f_{}(_{C_i}/_{C_i}^2)`$. Then $`\mathrm{deg}(_i)=C_i^2`$ and $`_1_2=𝒪_B^{}`$. On the other hand, $`𝒯_X^1`$ is the dual of $`_1\psi ^{}(_2)`$. The kernel of the homomorphism
$$\varphi _{}:B^{}\mathrm{Pic}^0(B^{}),bT_b^{}()^1$$
is the subgroup $`B_e^{}B^{}`$ of $`e`$-torsion points (, Lem. 4.7). Choose an origin $`0B^{}`$ and some $`bB^{}`$ such that $`\psi `$ is the translation $`T_b:B^{}B^{}`$. So $`𝒯_X^1`$ is trivial if and only if $`bB_e^{}`$, which means $`w|e`$. ∎
###### Remark 4.3.
Here we see that a complete set of invariants of admissible $`X`$ with nonrational algebraic normalization and positive degree is: The $`j`$-invariant of $`B^{}`$, the degree $`e>0`$, and a translation $`\psi `$ of $`B^{}`$ of finite order. In fact, $`^1`$ bundles over $`B^{}`$ of degree $`e>0`$ with two disjoint sections are of the form $`(𝒪_B^{}L^{})`$ with $`L`$ a line bundle of degree $`e`$. Up to pull-back by translations the latter are all isomorphic.
We come to the construction of smoothings:
###### Theorem 4.4.
Suppose $`X`$ is an admissible surface (Definition 2.1) with algebraic, nonrational normalization, of degree $`e`$ (possibly $`0`$) and warp $`w`$. Then $`X`$ deforms to an elliptic principal bundle of degree $`e/w`$ over an elliptic curve.
###### Proof.
In order to construct the desired smoothing, we first pass to a Galois covering of $`X`$. The ruling $`f:SB^{}`$ yields a bundle $`XB`$ over the isogenous elliptic curve $`B=B^{}/\psi `$ whose fibres are Néron $`w`$-gons. Let $`G\mathrm{Pic}^0(B^{})\mathrm{Aut}(B^{})`$ be the group of order $`w`$ generated by $`\psi `$. Consider the surface $`S^{}=S\times G`$ and the isomorphisms
$$\psi _j:C_2\times \left\{\psi ^j\right\}C_1\times \left\{\psi ^{j+1}\right\},(s,\psi ^j)(\psi (s),\psi ^{j+1})$$
for $`j/w`$. The corresponding relation defines a normal crossing surface $`X^{}`$ with $`w`$ irreducible components. Clearly, $`X^{}`$ is $`d`$-semistable with $`K_X^{}=0`$. The surface $`S^{}`$ is endowed with the free $`G`$-action $`\psi :(s,\psi ^j)(s,\psi ^{j+1})`$. It descends to a free $`G`$-action on $`X^{}`$ with quotient $`X=X^{}/G`$.
We proceed similarly as in Theorem 3.7. Let $`V`$ be the smooth toric variety corresponding to the infinite fan in $`^3`$ generated by the cones
$$\sigma _n=(0,n,1),(0,n+1,1),n.$$
The projection $`\mathrm{pr}_3:^3`$ defines a toric morphism $`V`$. The special fibre $`V_0`$ is the product of a Néron $`\mathrm{}`$-gon with $`^{}`$. The automorphism
$$\mathrm{\Psi }=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 1& 1\end{array}\right)\mathrm{SL}_3().$$
acting from the right on row vectors maps $`\sigma _n`$ to $`\sigma _{n+1}`$. By abuse of notation we also write $`\mathrm{\Psi }`$ for the induced automorphism of $`V`$. Choose $`\alpha \mathrm{\Delta }^{}`$ with $`B^{}=^{}/\alpha ^w`$ and $`B=^{}/\alpha `$. The automorphism $`(t_1,t_2,t_3)(\alpha t_1,t_1^{e/w}t_2,t_3)`$ of $`(^{})^3`$ extends to another automorphism $`\mathrm{\Phi }`$ of $`V`$. The action of $`\mathrm{\Phi },\mathrm{\Psi }`$ on the preimage $`UV`$ of $`\mathrm{\Delta }`$ is proper and free. Hence $`𝔛^{}=U/\mathrm{\Phi }^w,\mathrm{\Psi }^w`$ is a smooth 3-fold, endowed with a projection $`𝔛^{}\mathrm{\Delta }`$. The general fibres $`𝔛_t^{}`$, $`t\mathrm{\Delta }^{}`$, are smooth Kodaira surfaces of degree $`e`$. The special fibre is $`𝔛_0^{}=X^{}`$. The action of $`(\mathrm{\Phi }\mathrm{\Psi })^w`$ descends to a free $`G`$-action on $`𝔛^{}`$, which coincides with the given action on $`𝔛_0^{}=X^{}`$. Hence $`𝔛=𝔛^{}/G`$ is the desired smoothing. ∎
We head for the calculation of the versal deformation of $`X`$.
###### Proposition 4.5.
Suppose $`K_X=0`$. Then the following holds:
1. If $`e>0`$, then $`h^0(\mathrm{\Theta }_X)=1`$, $`h^1(\mathrm{\Theta }_X)=2`$, and $`h^2(\mathrm{\Theta }_X)=1`$.
2. If $`e=0`$, then $`h^0(\mathrm{\Theta }_X)=2`$, $`h^1(\mathrm{\Theta }_X)=3`$, and $`h^2(\mathrm{\Theta }_X)=1`$.
###### Proof.
As in the proof of Lemma 3.8 we use $`H^p(\mathrm{\Theta }_X)H^{2p}(\mathrm{\Omega }_X^1/\tau _X)`$ and the exact sequence
$$0\mathrm{\Omega }_X^1/\tau _X\mathrm{\Omega }_S^1\mathrm{\Omega }_D^10.$$
Recall $`h^{1,0}(S)=h^{1,0}(D)=1`$. The map on the right in
$$0H^0(X,\mathrm{\Omega }_X^1/\tau _X)H^0(S,\mathrm{\Omega }_S^1)H^0(D,\mathrm{\Omega }_D^1)$$
is zero, since it is an alternating sum, so $`h^2(\mathrm{\Theta }_X)=1`$. Next we consider
$$0H^0(D,\mathrm{\Omega }_D^1)H^1(X,\mathrm{\Omega }_X^1/\tau _X)H^1(S,\mathrm{\Omega }_S^1)H^1(D,\mathrm{\Omega }_D^1).$$
We have $`h^{1,1}(S)=2`$ and $`h^{1,1}(D)=1`$. The class of a fibre $`FS`$ in $`H^{1,1}(S)`$ maps to zero in $`H^{1,1}(D)`$. For $`e=0`$, this also holds for the class of the section $`C_1S`$, and $`h^1(\mathrm{\Theta }_X)=3`$ follows. For $`e>0`$, the image of the section does not vanish, and $`h^1(\mathrm{\Theta }_X)=2`$ holds instead. Finally, the sequence
$$H^1(S,\mathrm{\Omega }_S^1)H^1(D,\mathrm{\Omega }_D^1)H^2(X,\mathrm{\Omega }_X^1/\tau _X)H^2(S,\mathrm{\Omega }_S^1)0$$
is exact. Now $`h^{1,2}(S)=1`$ yields $`h^0(\mathrm{\Theta }_X)=2`$ for $`e=0`$, and $`h^0(\mathrm{\Theta }_X)=1`$ for $`e>0`$. ∎
###### Corollary 4.6.
Suppose $`K_X=0`$ and $`e=0`$. Then $`X`$ does not deform to a smooth Kodaira surface.
###### Proof.
Write $`S=(𝒪_B^{})`$ for some invertible $`𝒪_B^{}`$-module $``$ of degree 0. Moving the gluing parameter $`\psi `$ and the isomorphism class of $`B^{}`$ and $``$ gives a 3-dimensional locally trivial deformation of $`X`$. Since $`h^1(\mathrm{\Theta }_X)=3`$, the space parameterizing the locally trivial deformations in the semiuniversal deformation $`p:𝔛V`$ of $`X`$ is a smooth 3-fold. Now each fibre $`𝔛_t`$ deforms to an elliptic principal bundle of degree 0 (Thm. 4.4), and the embedding dimension of $`V`$ is $`dimT_X^1=4`$. This shows that $`V`$ is a smooth 4-fold with an open dense set parameterizing complex tori. Hence no fibre $`𝔛_t`$ is isomorphic to a smooth Kodaira surface. ∎
###### Proposition 4.7.
Let $`X`$ be as in Theorem 4.4 with $`e>0`$. Then $`dimT_X^0=1`$, $`dimT_X^1=3`$ and $`dimT_X^2=2`$.
###### Proof.
In view of Proposition 4.5 we only have to show degeneracy of the spectral sequence of tangent cohomology $`E_2^{p,q}=H^p(𝒯_X^q)T_X^{p+q}`$ at $`E_2`$ level, cf. Proposition 3.9. This is the case iff $`T_X^1H^0(𝒯_X^1)`$ is surjective. Now by $`d`$-semistability $`H^0(𝒯_X^1)`$ is one-dimensional and any generator has no zeros. Geometrically degeneracy of the spectral sequence therefore means the existence of a deformation of $`X`$ over $`\mathrm{\Delta }_\epsilon :=\mathrm{Spec}[\epsilon ]/\epsilon ^2`$ that is not locally trivial. This is what we established by explicit construction in Theorem 4.4. More precisely, let $`𝔛\mathrm{\Delta }`$ be a deformation of $`X`$. For $`PD`$ the image of the Kodaira-Spencer class $`\kappa T_X^1`$ of this deformation in $`𝒯_{X,P}^1\text{xt}^1(\mathrm{\Omega }_{X,P}^1,𝒪_{X,P})`$ is the Kodaira-Spencer class of the induced deformation of the germ of $`X`$ along $`D`$. In appropriate local coordinates such deformations have the form $`xy\epsilon f(z)=0`$ with $`f`$ inducing the section of $`\text{xt}^1(\mathrm{\Omega }_{X,P}^1,𝒪_{X,P})`$. Therefore $`f0`$ iff the total space $`𝔛`$ is smooth at $`P`$. ∎
###### Theorem 4.8.
Suppose $`X`$ is admissible with algebraic, nonrational normalization, of degree $`e>0`$ and warp $`w`$. Let $`p:𝒳V`$ be the semiuniversal deformation of $`X`$. Then $`V=V_1V_2`$ has two irreducible components, and the following holds:
1. $`V_2`$ is a smooth surface and parameterizes the locally trivial deformations.
2. $`V_1V_2`$ is a smooth curve and parameterizes the $`d`$-semistable locally trivial deformations.
3. $`V_1`$ is a smooth surface, and $`V_1V_1V_2`$ parameterizes smooth Kodaira surfaces of degree $`e/w`$.
###### Proof.
A similar situation has been found by Friedman in his study of deformations of $`d`$-semistable K3 surfaces. We follow the proof of , Theorem 5.10 closely.
Deformation theory provides a holomorphic map $`h:T_X^1T_X^2`$ with $`V=h^1(0)`$, whose linear term is zero, and whose quadratic term is given by the Lie bracket $`1/2[v,v]`$ (cf. e.g. ). Moving the two parameters $`j(E)`$, $`\psi B^{}`$ and using $`h^1(\mathrm{\Theta }_X)=2`$, we see that the locally trivial deformations are parameterized by a smooth surface $`V_2T_X^1`$. Let $`h_2:T_X^1`$ be a holomorphic map with $`V_2=h_2^1(0)`$, and $`h_1:T_X^1T_X^2`$ a holomorphic map with $`h=h_1h_2`$. Let $`L_1:T_X^1T_X^2`$ and $`L_2:T_X^1`$ be the corresponding tangential maps and set $`V_1=h_1^1(0)`$.
We proceed by showing that $`V_1V_2`$ is a smooth curve, or equivalently that the intersection of $`\mathrm{kern}(L_1)`$ with $`H^1(\mathrm{\Theta }_X)=\mathrm{kern}(L_2)`$ is 1-dimensional. The smoothing of $`X`$ constructed in Theorem 4.4 obviously has a smooth total space. As in the proof of Proposition 3.9 we see that its Kodaira-Spencer class $`kT_X^1`$ generates $`𝒯_X^1`$.
Let $`L:H^1(\mathrm{\Theta }_X)H^1(𝒯_X^1)`$ be the linear map $`v[v,k]`$. By , Proposition 4.5, its kernel are the first order deformations which remain $`d`$-semistable. Since we can destroy $`d`$-semistability by moving the gluing $`\psi \mathrm{Pic}^0(E)`$, this kernel is 1-dimensional. As in , page 109, one shows that for $`vH^1(\mathrm{\Theta }_X)`$:
$$L(v)=[v,k]=1/2[v+k,v+k]=L_1(v+k)L_2(v+k)=L_1(v)L_2(k).$$
Using $`L_2(k)0`$ we infer $`\mathrm{kern}L_1H^1(\mathrm{\Theta }_X)=\mathrm{kern}L`$. It follows that $`V_1V_2`$ is a smooth curve, which parameterizes the $`d`$-semistable locally trivial deformations, and $`V_1`$ must be a smooth surface. Our local computation together with the interpretation of $`j(E)`$ as coordinate on $`V_1V_2`$ shows that $`V_1`$ is nothing but the base of the smoothing of $`X`$ constructed in Theorem 4.4. In particular, $`V_1V_1V_2`$ parameterizes smooth Kodaira surfaces. ∎
## 5. Rational surfaces and honeycomb degenerations
The goal of this section is to analyze rational admissible surfaces. According to Proposition 2.6, the ramification curve $`CS`$ of the normalization $`\nu :SX`$ is a Néron 6-gon $`C=C_0\mathrm{}C_5`$ with $`K_S=C`$. We suppose $`C_iC_{i+1}\mathrm{}`$, regarding the indices as elements in $`/6`$. The singular locus $`DX`$ is the seminormal curve with normalization $`^1\times /3`$ and imposed relations $`(0,i)(0,j)`$ and $`(\mathrm{},i)(\mathrm{},j)`$ for $`i,j/3`$. The gluing $`\phi :CD`$ identifies pairs of irreducible components in $`C`$. Since the resulting surface should be normal crossing the fibers of $`SX`$ have cardinality at most $`3`$. There are two possibilities left. The first alternative is $`\phi (C_i)=\phi (C_{i+3})`$ and $`\phi (C_iC_{i+1})=\phi (C_{i+2}C_{i+3})`$ for all $`i/6`$. We call it untwisted gluing. The second alternative is $`\phi (C_i)=\phi (C_{i+3})`$, $`\phi (C_{i+1})=\phi (C_{i1})`$, $`\phi (C_{i+2})=\phi (C_{i2})`$ and $`\phi (C_iC_{i+1})=\phi (C_{i1}C_{i2})`$ for some $`i/6`$. Call it *twisted* gluing. The distinction is important for the canonical class:
###### Proposition 5.1.
The condition $`K_X=0`$ holds if and only if the gluing $`\phi :CD`$ is untwisted.
###### Proof.
As in Propositions 3.2 and 4.1 the condition $`K_X=0`$ is equivalent to $`H^2(𝒪_X)0`$. By the exact sequence
$$H^1(S,𝒪_S)H^1(D,𝒪_D)H^1(C,𝒪_C)H^2(X,𝒪_X)0$$
and with $`h^{0,1}(S)=0`$ this holds precisely if $`\phi ^{}:H^1(𝒪_D)H^1(𝒪_C)`$ vanishes. The bicoloured graphs attached to the curves $`C,D`$ (, section 3.5) are
Here the white vertices are the irreducible components, the black vertices are the singularities, and the edges are the ramification points on the normalization. Since the curves in question are seminormal, the map $`\phi ^{}:H^1(𝒪_D)H^1(𝒪_C)`$ coincides with $`\phi ^{}:H^1(\mathrm{\Gamma }(D),)H^1(\mathrm{\Gamma }(C),)`$. A direct calculation left to the reader shows that it vanishes if and only if $`\phi `$ is untwisted. ∎
From now on we assume that the gluing $`\phi :CD`$ is untwisted. According to Proposition 2.6, the surface $`S`$ is obtained from a Hirzebruch surface $`f^{}:S^{}^1`$ by blowing-up twice. Our conventions are that $`C_0,C_3S`$ are sections of the induced fibration $`f:S^1`$, that $`C_1C_2`$, $`C_4C_5`$ are the fibres over $`0,\mathrm{}^1`$, and that $`C_1,C_4`$ are exceptional for the contraction $`r:SS^{}`$. Let us also assume that the image of $`C_0`$ in $`S^{}`$ is a minimal section. If $`e0`$ is the degree of the Hirzebruch surface $`S^{}`$, this means $`C_0^2=e1`$ and $`C_3^2=e1`$.
Set $`D_i=\phi (C_i)`$ for $`i/3`$. The space of all untwisted gluings $`\phi :CD`$ is a torsor under
(5.1)
$$\mathrm{Aut}^0(D)=\mathrm{Aut}^0(D_i)(^{})^3.$$
The action of $`\varphi \mathrm{Aut}^0(D)`$ is given by composing $`\phi C_0C_1C_2`$ with $`\varphi `$. We call $`\phi C_0C_3`$ the *vertical* gluing. The ruling $`f:S^1`$ yields a preferred vertical gluing, which identifies points of the same fibre. Every other vertical gluing differs from the preferred one by a vertical gluing parameter $`\zeta ^{}`$. We call its order $`w0`$ the *warp* of $`X`$. The warp is important for the calculation of $`𝒯_X^1`$:
###### Proposition 5.2.
The surface $`X`$ is $`d`$-semistable if and only if the warp $`w`$ divides the degree $`e`$.
###### Proof.
The inclusions $`D_iD_jD`$ give an injection $`\mathrm{Pic}(D)_{ij}\mathrm{Pic}(D_iD_j)`$. We proceed by calculating the class of $`𝒯_X^1D_0D_1`$. Consider the normal crossing surface $`\overline{S}`$ defined by the cocartesian diagram
$$\begin{array}{ccc}C_2C_5& & S\\ & & & & \\ D_2& & \overline{S}\end{array}$$
Let $`\overline{C}_i\overline{S}`$ be the images of $`C_iS`$. The ideals $`_{01},_{34}𝒪_{\overline{S}}`$ of the Weil divisors $`\overline{C}_{01}=\overline{C}_0\overline{C}_1`$ and $`\overline{C}_{34}=\overline{C}_3\overline{C}_4`$ are invertible. A local computation shows that
$$(𝒯_X^1D_0D_1)^{}_{01}/_{01}^2\psi ^{}(_{34}/_{34}^2),$$
where $`\psi :\overline{C}_{01}\overline{C}_{34}`$ is the isomorphism obtained from the induced gluing $`\overline{\phi }:\overline{C}_{01}\overline{C}_{34}D_0D_1`$. Another local computation gives
$$_{01}/_{01}^2\overline{C}_0_{C_0}/_{C_0}^2(C_0C_1),_{01}/_{01}^2\overline{C}_1_{C_1}/_{C_1}^2(C_0C_1).$$
It follows that $`𝒯_X^1D_0D_1`$ is trivial if and only if $`\zeta ^e=1`$. A similar argument applies to $`D_0D_2`$ and $`D_1D_2`$. ∎
###### Remark 5.3.
The previous two propositions yield the following classification of admissible $`X`$ with rational normalization and positive degree $`e`$: According to Equation 5.1, the isomorphism class of $`X`$ is determined by $`e>0`$ and 3 gluing parameters. By Proposition 5.2 the vertical one is an $`e`$-th root of unity $`\zeta `$. For $`(e,\zeta )`$ fixed the automorphisms of $`S`$ act on the remaining two horizontal gluing parameters with quotient isomorphic to $`^{}`$.
We come to the existence of smoothings:
###### Theorem 5.4.
Suppose $`X`$ is an admissible surface (Definition 2.1) with rational normalization, of degree $`e`$ (possibly $`0`$) and warp $`w`$. Then $`X`$ deforms to an elliptic principal bundle of degree $`e/w`$ over an elliptic curve.
###### Proof.
It should not be too surprising that the construction is a modification of both the constructions in Theorems 3.7 and 4.4.
First, we simplify matters by passing to a Galois covering. Let $`G^{}`$ be the group of order $`w`$ generated by the vertical gluing parameter $`\zeta ^{}`$. Since $`w|e`$, the $`G`$-action $`z\zeta z`$ on $`^1`$ lifts to a $`G`$-action on the Hirzebruch surface $`S^{}`$, and hence to the blowing-up $`S`$. The diagonal action on $`\stackrel{~}{S}=S\times G`$ is free. Let $`\psi _i:C_iC_{i+3}`$ be the isomorphisms induced by the gluing $`\phi :CD`$. This gives $`G`$-equivariant isomorphisms
$$\psi _{ij}:C_i\times \left\{\zeta ^j\right\}C_{i+3}\times \left\{\zeta ^{j+1}\right\},(s,\zeta ^j)(\psi _i(s),\zeta ^{j+1}).$$
Let $`X^{}`$ be the normal crossing surface obtained from $`\stackrel{~}{S}`$ modulo the relation imposed by the $`\psi _{ij}`$. Then $`G`$ acts freely on $`X^{}`$ with quotient $`X=X^{}/G`$.
The next task is to construct a $`G`$-equivariant smoothing of $`X^{}`$. Again toric geometry enters the scene. Consider the infinite fan in $`^4`$ generated by the cones
$`\sigma _{m,n}`$ $`=`$ $`(m,e\left(\genfrac{}{}{0pt}{}{m}{2}\right),1,0),(m+1,e\left(\genfrac{}{}{0pt}{}{m+1}{2}\right),1,0),(0,n,0,1),(0,n+1,0,1)`$
for $`m,n`$. Let $`V`$ be the corresponding 4-dimensional smooth toric variety. The projection $`\mathrm{pr}_{34}:^4^2`$ defines a toric morphism $`\mathrm{pr}_{34}:V^2`$. The special fibre $`V_0`$ is isomorphic to a bundle of Néron $`\mathrm{}`$-gons over the Néron $`\mathrm{}`$-gon. Let $`f:\widehat{}^2^2`$ be the blowing-up of the origin $`0^2`$ with the closed toric orbits removed and $`E\widehat{}^2`$ the exceptional set. It is isomorphic to $`^{}`$. The cartesian diagram
$$\begin{array}{ccc}\widehat{V}& & V\\ & & & & \\ \widehat{}^2& & ^2\end{array}$$
defines a smooth toric 4-fold $`\widehat{V}`$, which is an open subset of the blowing-up of $`V`$ with center $`V_0V`$. The exceptional divisor $`\widehat{V}_E=\widehat{V}_0`$ of $`\widehat{V}V`$ is isomorphic to $`E\times V_0`$. The exceptional divisor $`E`$ will be a parameter space of the whole construction via the $`^{}`$-bundle $`\widehat{}E`$ induced by the canonical map $`^20^1`$.
Let $`Z\widehat{V}`$ be the set of 1-dimensional toric orbits. Each fibre $`Z_t`$, $`tE`$, consists of the discrete set of non-normal-crossing singularities in $`\widehat{V}_tV_0`$, as illustrated by Figure 1. Let $`\overline{V}\widehat{V}`$ be the blowing-up with center $`Z\widehat{V}`$. The exceptional divisor $`\overline{Z}\overline{V}`$ is easy to determine. Each fibre $`\widehat{V}_t`$, $`tE`$, is smooth, except for the points in $`Z_t`$. At such points a local equation for $`\widehat{V}_t`$ is $`T_1T_2=tT_3T_4`$, which is an affine cone over a smooth quadric in $`^2`$. Hence $`\overline{Z}`$ is a disjoint union of smooth quadrics in $`^3`$, each isomorphic to $`^1\times ^1`$. The whole exceptional divisor is $`\overline{Z}E\times Z\times ^1\times ^1`$. The picture over a fixed $`tE`$ is depicted in Figure 2.
Fig. 1 Fig. 2
Here the octagons lie on the strict transform of $`V_t`$, and the squares are contained in the exceptional divisor $`\overline{Z}_t`$.
We seek to contract $`\overline{Z}\overline{V}`$ along one of the two rulings of $`^1\times ^1`$. Let $`F\overline{Z}`$ be a $`^1`$-fibre. A local calculation shows that $`𝒪_F(\overline{Z})`$ has degree $`1`$, and that the Cartier divisors $`K_{\overline{V}}`$ and $`𝒪_{\overline{V}}(\overline{Z})`$ coincide in a neighborhood of $`F\overline{V}`$. So the Nakano contraction criterion applies, and there exists a contraction $`\overline{V}\stackrel{~}{V}`$ which restricts to the projection
$$\mathrm{pr}_{123}:\overline{Z}=E\times Z\times ^1\times ^1E\times Z\times ^1.$$
The special fibre $`\stackrel{~}{V}_0`$ resembles an infinite system of honeycombs:
Fig. 3
Finally, we want to arrive at compact surfaces, so we seek to divide out a cocompact group action. Consider the two commuting automorphisms
$$\mathrm{\Phi }=\left(\begin{array}{ccccc}1& e& 0& 0& 0\\ 0& 1& 0& 0& 0\\ 1& 0& 1& 0& 0\\ 0& 0& 0& 1& 0\\ 0& 1& 0& 0& 1\end{array}\right)\text{and}\mathrm{\Psi }=\left(\begin{array}{ccccc}1& 0& 0& 0& 0\\ 0& 1& 0& 0& 0\\ 0& 0& 1& 0& 0\\ 0& 1& 0& 1& 0\\ 0& 0& 0& 0& 1\end{array}\right)$$
of $`^5`$ acting from the right on row vectors. We regard our fan in $`^4`$ as a fan in $`^5=^4`$. Torically, the trivial $``$-factor amounts to going over to $`V\times ^{}`$, which will give a horizontal gluing parameter. Since $`\mathrm{\Phi }(\sigma _{m,n})=\sigma _{m+1,n}`$ and $`\mathrm{\Psi }(\sigma _{m,n})=\sigma _{m,n+1}`$, we get an induced action of the discrete group $`^2=\mathrm{\Phi },\mathrm{\Psi }`$ on $`V\times ^{}`$. Let $`UV`$ be the preimage of $`\mathrm{\Delta }^2^2`$. The action is proper and free on $`U\times ^{}`$. Let $`\stackrel{~}{U}\stackrel{~}{V}`$ be the corresponding preimage. It is easy to see that the action on $`V\times `$ induces an action on $`\stackrel{~}{V}\times ^{}`$ which is proper and free on $`\stackrel{~}{U}\times ^{}`$. So the quotient $`𝔛^{}=(\stackrel{~}{U}\times ^{})/\mathrm{\Phi },\mathrm{\Psi }^w`$ is a smooth complex 5-fold. The action on the special fibre is indicated in Figure 3 by the arrows. The general fibres $`𝔛_t^{}`$, $`(t,\lambda )f^1(\mathrm{\Delta }^{}\times \mathrm{\Delta }^{})\times ^{}`$ are elliptic bundles of degree $`e`$ over the elliptic curve $`^{}/t_1`$ with fibre $`^{}/t_2`$, where $`f(t)=(t_1,t_2)`$. Some special fibre $`𝔛_t^{}`$, $`tE\times ^{}`$, is isomorphic to $`X^{}`$, since $`t_5`$ moves through all horizontal gluings.
It remains to extend the $`G`$-action on $`X^{}`$ to a $`G`$-action on $`𝔛^{}`$. The automorphism
$$(^{})^5(^{})^5,(t_1,t_2,t_3,t_4,t_5)(\zeta t_1,t_2t_4,t_3,t_4,t_5)$$
of the torus extends to an automorphism of the torus embedding $`V`$. It induces a free $`G`$-action on $`𝔛^{}`$, which is the desired extension. Set $`𝔛=𝔛^{}/G`$. General fibres $`𝔛_t`$, $`t(^20)\times ^{}`$ are elliptic bundles of degree $`e/w`$ over an elliptic curve. Some special fibre $`𝔛_t`$, $`tE\times ^{}`$ is isomorphic to $`X`$. ∎
We turn our attention to the versal deformation of $`X`$ and calculate the relevant cohomology groups:
###### Proposition 5.5.
Suppose $`K_X=0`$. Then the following holds:
1. If $`e>0`$, then $`h^0(\mathrm{\Theta }_X)=1`$, $`h^1(\mathrm{\Theta }_X)=2`$, and $`h^2(\mathrm{\Theta }_X)=0`$.
2. If $`e=0`$, then $`h^0(\mathrm{\Theta }_X)=2`$, $`h^1(\mathrm{\Theta }_X)=3`$, and $`h^2(\mathrm{\Theta }_X)=0`$.
###### Proof.
As in Lemmas 3.8 and 4.5 we use $`H^p(X,\mathrm{\Theta }_X)H^{2p}(X,\mathrm{\Omega }_X^1/\tau _X)`$ and the exact sequence
$$0\mathrm{\Omega }_X^1/\tau _X\mathrm{\Omega }_S^1\mathrm{\Omega }_{\stackrel{~}{D}}^10.$$
Here $`\stackrel{~}{D}`$ is the normalization of $`D`$. The inclusion $`H^0(\mathrm{\Omega }_X^1/\tau _X)H^0(\mathrm{\Omega }_S^1)`$ and $`b_1(S)=0`$ gives $`h^2(\mathrm{\Theta }_X)=0`$. Next consider the exact sequence
$$0H^1(X,\mathrm{\Omega }_X^1/\tau _X)H^1(S,\mathrm{\Omega }_S^1)H^1(\stackrel{~}{D},\mathrm{\Omega }_{\stackrel{~}{D}}^1)$$
The task is to determine the rank of the map on the right. This is done as in , p. 91: One has to compute the images in $`H^{1,1}(\stackrel{~}{D})`$ of the classes of $`C_iS`$ using intersection numbers on $`S`$. For $`e>0`$ this gives $`h^1(\mathrm{\Theta }_X)=2`$, whereas for $`e=0`$ the result is $`h^1(\mathrm{\Theta }_X)=3`$. We leave the actual computation to the reader. Finally, the exact sequence
$$H^1(S,\mathrm{\Omega }_S^1)H^1(\stackrel{~}{D},\mathrm{\Omega }_{\stackrel{~}{D}}^1)H^2(X,\mathrm{\Omega }_X^1/\tau _X)H^2(S,\mathrm{\Omega }_S^1),$$
together with the preceding observations and $`b_3(S)=0`$ gives the stated values for $`h^0(\mathrm{\Theta }_X)`$. ∎
###### Corollary 5.6.
Suppose $`K_X=0`$ and $`e=0`$. Then $`X`$ does not deform to a smooth Kodaira surface.
###### Proof.
The locally trivial deformations of $`X`$ are unobstructed since $`h^2(\mathrm{\Theta }_X)=0`$. They define a smooth 3-fold $`V^{}V`$ in the base of the semiuniversal deformation $`𝔛V`$. It is easy to see that it is given by the three gluing parameters in $`\phi :CD`$. Since each fibre $`𝔛_t`$ deforms to a complex torus (Thm. 5.4), $`V`$ is smooth of dimension $`4`$ and no fibre can be isomorphic to a smooth Kodaira surface. ∎
###### Proposition 5.7.
Let $`X`$ be as in Theorem 5.4 with $`e>0`$. Then $`dimT_X^0=1`$, $`dimT_X^1=3`$ and $`dimT_X^2=2`$.
###### Proof.
This is shown as the similar statement of Proposition 4.7. ∎
###### Theorem 5.8.
Suppose $`X`$ is admissible with rational normalization, of degree $`e>0`$ and warp $`w`$. Let $`p:𝔛V`$ be the semiuniversal deformation of $`X`$. Then $`V=V_1V_2`$ consists of two irreducible components, and the following holds:
1. $`V_2`$ is a smooth surface parameterizing the locally trivial deformations.
2. $`V_1V_2`$ is a smooth curve parameterizing the locally trivial deformations which remain $`d`$-semistable.
3. $`V_1V_1V_2`$ parameterizes smooth Kodaira surfaces of degree $`e/w`$.
###### Proof.
The argument is as in the proof of Theorem 4.8. ∎
## 6. The completed moduli space and its boundary
In this final section, we take up a global point of view and analyze degenerations of smooth Kodaira surfaces in terms of moduli spaces. Here we use the word “moduli space” in the broadest sense, namely as a topological space whose points correspond to Kodaira surfaces. We do not discuss wether it underlies a coarse moduli space or even an analytic stack or analytic orbispace.
Let $`𝔎_d=\times \mathrm{\Delta }^{}`$ be the moduli space of smooth Kodaira surfaces of degree $`d>0`$, and $`𝔎=_{d>0}𝔎_d`$ their union. The points of $`𝔎`$ correspond to the isomorphism classes $`[X]`$ of smooth Kodaira surfaces. The topology on $`𝔎`$ is induced by what we suggest to call the *versal topology* on the set $`𝔐`$ of isomorphism classes of all compact complex spaces. The versal topology is the finest topology on $`𝔐`$ rendering continuous all maps $`V𝔐`$ defined by flat families $`𝔛V`$ which are versal for all fibres. The complex structure on $`𝔎`$ comes from Hodge theory: One can view $`𝔎_d`$ as the period domain of polarized pure Hodge structures of weight 2 on $`H^2(X,)/\text{torsion}`$ divided by the automorphism group of this lattice. According to , the induced structure of a ringed space on $`𝔎_d`$ is the usual complex structure on $`\times \mathrm{\Delta }^{}`$.
Let $`𝔎\overline{𝔎}`$ be the space obtained by adding all admissible surfaces in the closure of $`𝔎𝔐`$. The surfaces parameterized by the boundary $`𝔅=\overline{𝔎}𝔎`$ are called *d-semistable* Kodaira surfaces. These are nothing but the admissible surfaces deforming to smooth Kodaira surfaces. The boundary decomposes into three parts
$$𝔅=𝔅^h𝔅^r𝔅^e$$
according to the three types of admissible surfaces. Here $`𝔅^h`$ refers to the surfaces whose normalization are Hopf surfaces, $`𝔅^r`$ to the surfaces with rational normalization, and $`𝔅^e`$ to surfaces whose normalization is ruled over an elliptic base. We refer to these parts of $`\overline{𝔎}`$ as Hopf, rational and elliptic ruled stratum respectively.
###### Proposition 6.1.
The irreducible components of the boundary $`𝔅`$ are smooth complex curves. The components in $`𝔅^h`$ are isomorphic to $`\mathrm{\Delta }^{}\left\{\mathrm{}\right\}\times \mathrm{\Delta }^{}`$, the components in $`𝔅^r`$ are isomorphic to $`^{}`$, and the components in $`𝔅^e`$ are isomorphic to $`\times \left\{0\right\}`$.
###### Proof.
This follows from Remarks 3.6, 4.3 and 5.3. ∎
Locally, the completion $`𝔎\overline{𝔎}`$ is isomorphic to the blowing-up of $`(\mathrm{},0)^1\times \mathrm{\Delta }`$, but with the points $`0,\mathrm{}^1`$ on the exceptional divisor removed. This follows in particular from the construction in Theorem 5.4 of a family over the blow up of $`^2`$ with $`2`$ points removed, with fiber over $`(t_1,t_2)^{}\times ^{}`$ a smooth Kodaira surface with invariants $`(j,\alpha )=(\mathrm{exp}(t_1),t_2)`$. To complete further we need to enlarge the class of generalized Kodaira surfaces. The least one would hope for is that any family $`𝔛^{}\mathrm{\Delta }^{}`$ of generalized Kodaira surfaces, that can be completed to a proper family over $`\mathrm{\Delta }`$, can be completed by a generalized Kodaira surface.
We first show that this is impossible if we consider only normal crossing surfaces.
###### Theorem 6.2.
Let $`𝔛\mathrm{\Delta }`$ be a degeneration of $`d`$-semistable Kodaira surfaces with elliptically ruled normalization. Assume that $`𝔛`$ is bimeromorphic to a Kähler manifold, and that the $`j`$-invariant of the base of $`𝔛_t`$ tends to $`0`$ with $`t\mathrm{\Delta }`$. Then $`𝔛_0`$ is not of normal crossing type.
###### Proof.
Let $`\stackrel{~}{𝔛}𝔛`$ be the normalization. Let $`𝔅`$ be the component of the Douady space of holomorphic curves in $`\stackrel{~}{𝔛}`$ that contain a fiber of $`\stackrel{~}{𝔛}_tB_t`$ for general $`t\mathrm{\Delta }`$. Since these curves are contained in fibers of $`\stackrel{~}{𝔛}\mathrm{\Delta }`$ there is a map $`𝔅\mathrm{\Delta }`$. By the Kähler assumption this map is proper . Let $`\stackrel{~}{𝔛}^{}𝔅`$ be the universal family. The universal map $`\stackrel{~}{𝔛}^{}\stackrel{~}{𝔛}`$ is an isomorphism over $`\mathrm{\Delta }^{}`$, hence bimeromorphic. By desingularisation we may dominate this map by successively blowing up $`\stackrel{~}{𝔛}`$ in smooth points and curves. This can be arranged to keep the property “normal crossing”. In other words, we can assume that $`\stackrel{~}{𝔛}^{}=\stackrel{~}{𝔛}`$ is a fibration by rational curves over a degenerating family $`𝔅\mathrm{\Delta }`$ of elliptic curves.
Let $`^{},^{\prime \prime }\stackrel{~}{𝔛}`$ be the two components of the preimage of the singular locus of $`𝔛`$ mapping onto $`\mathrm{\Delta }`$. For any $`b𝔅`$ the corresponding rational curve $`F_b`$ intersects $`^{},^{\prime \prime }`$ in one point each. Let $`b_0𝔅_0`$ be a node of the nodal elliptic curve over $`0\mathrm{\Delta }`$. Then each intersection $`^{}F_{b_0}`$ and $`^{\prime \prime }F_{b_0}`$ gives a point of multiplicity at least $`2`$ on $`\stackrel{~}{𝔛}_0`$. As $`^{},^{\prime \prime }`$ are identified under $`\stackrel{~}{𝔛}𝔛`$ this leads to a point of multiplicity at least $`4`$ on $`𝔛_0`$. ∎
There are however various completions if we admit *products of normal crossing singularities*. In dimension $`2`$ the only such singularity that is not normal crossing is a point $`(X,x)`$ of multiplicity $`4`$. It has as completed local ring
$$𝒪_{X,x}^{}=[[T_1,\mathrm{},T_4]]/(T_1T_2,T_3T_4).$$
In particular, it is a complete intersection and hence still Gorenstein. We will call $`(X,x)`$ a *quadrupel point*. The singular locus $`CX`$ consists of the four coordinate lines $`C_1,\mathrm{},C_4`$, and $`C_i,C_j`$ lie on the same irreducible component iff $`ij2`$ modulo $`4`$. The embedding dimension of $`(X,x)`$ is $`4`$. It is therefore not embeddable into a smooth 3-fold. Its appearance here is perhaps not so surprising, as it occurs also in certain compactifications of moduli of polarized abelian surfaces .
We were nevertheless unable to select a natural class of generalized Kodaira surfaces that would satisfy the mentioned completeness property. Therefore we content ourselves to define two surfaces with quadrupel points, connecting the elliptic ruled stratum to the Hopf stratum and to the rational stratum respectively. A surface connecting the Hopf stratum with the rational stratum could not be found, although such surfaces should probably exist.
Fix an integer $`e>0`$. Let $`S`$ be the Hirzebruch surface of degree $`e`$. Choose two sections $`C_0,C_2S`$ with $`C_0^2=e`$, $`C_2^2=e`$, and let $`C_1,C_3`$ be the fibers over $`0,\mathrm{}^1`$. We define the surface $`X_1`$ by gluing $`C_1,C_3`$ by any isomorphism (all choices give isomorphic results), and $`C_0,C_2`$ by an isomorphism of finite order $`w`$ over $`^1`$. Note that $`X_1`$ is normal crossing except at one quadrupel point. As in the proof of Proposition 5.1 one checks $`K_{X_1}=0`$.
###### Proposition 6.3.
Suppose $`w`$ divides $`e`$. There is a family $`𝔛^2`$ with the following properties:
1. $`𝔛_0X_1`$
2. the fibers over $`^{}\times \{0\}`$ are $`d`$-semistable Kodaira surfaces of warp $`w`$ and degree $`e`$ with elliptic ruled normalization
3. the fibers over $`\{0\}\times ^{}`$ are $`d`$-semistable Kodaira surfaces of warp $`w`$ and degree $`e`$ with nonalgebraic normalization
###### Proof.
In the proof of Theorem 5.4 we constructed a toric morphism $`V^2`$ with central fiber a bundle of Néron $`\mathrm{}`$-gons over the Néron $`\mathrm{}`$-gon. The restriction of $`\mathrm{\Phi },\mathrm{\Psi }\mathrm{GL}(^5)`$ defined there to $`^4^4\{0\}`$ defines an action of $`^2`$ on the fan defining $`V`$. The induced action on $`V`$ is proper and free, and commutes with the map to $`^2`$. We obtain a proper family $`𝔛^{}=V/^2^2`$. By construction of $`V`$ there is also an action of $`G/w`$ on $`𝔛^{}/^2`$. The family $`𝔛^{}/G^2`$ has the desired properties. ∎
Finally we study a surface connecting the rational and the elliptic stratum, under the expense of changing the warp. Again we fix an integer $`e>0`$. Let $`S=S^{}S^{\prime \prime }`$ be the disjoint union of two Hirzebruch surfaces of degrees $`e+1`$ and $`e1`$ respectively. Choose two sections $`C_0^{},C_2^{}S^{}`$ with $`(C_0^{})^2=(e+1)`$ and $`(C_2^{})^2=e+1`$, and let $`C_1^{},C_3^{}`$ be the fibres over $`0,\mathrm{}^1`$. Make the analogous choices $`C_0^{\prime \prime },C_2^{\prime \prime }`$ etc. for $`S^{\prime \prime }`$. We glue $`C_i^{}`$ to $`C_{i+2}^{\prime \prime }`$ in a way that preserves the intersection points indicated in the following figure by circles and crosses.
Fig. 4
For odd $`i`$ the isomorphism $`C_i^{}C_i^{\prime \prime }`$ can be chosen arbitrarily; for even $`i`$ we glue compatibly with the projections $`S^{}^1`$, $`S^{\prime \prime }^1`$. The result is a reduced surface $`X_2`$ that is normal crossing except at two points corresponding to the circles and the crosses in the figure. It is a bundle of Néron $`1`$-gons over $`^1`$. Again as in the proof of Proposition 5.1, one verifies that $`K_{X_2}=0`$ holds.
###### Proposition 6.4.
1. $`X_2`$ deforms to $`d`$-semistable Kodaira surfaces with rational normalization of degree $`e`$ and warp $`1`$.
2. $`X_2`$ deforms to $`d`$-semistable Kodaira surfaces with elliptic ruled normalization of degree $`2e`$ and warp $`2`$.
###### Proof.
(i) The idea is to modify a fiberwise degeneration of Hirzebruch surfaces. Let $`\stackrel{~}{S}^{}`$, $`\stackrel{~}{S}^{\prime \prime }`$ be Hirzebruch surfaces of degree $`e`$. Define curves $`\stackrel{~}{C}_i^{}\stackrel{~}{S^{}}`$ and $`\stackrel{~}{C}_i^{\prime \prime }\stackrel{~}{S^{\prime \prime }}`$ as above.
Let $`p:\stackrel{~}{}\mathrm{\Delta }`$ be a flat family whose general fibres $`\stackrel{~}{}_t`$, $`t0`$, are Hirzebruch surfaces of degree $`e`$; the closed fiber $`\stackrel{~}{}_0`$ is the union of $`\stackrel{~}{S}^{}`$ and $`\stackrel{~}{S}^{\prime \prime }`$ with $`\stackrel{~}{C}_2^{}`$ and $`\stackrel{~}{C}_0^{\prime \prime }`$ identified, analogously to the left-half of Figure 4. Such a family can be constructed either by toric methods or as follows. Let $`L^1`$ be the $`^{}`$-bundle of degree $`e`$. Let $`\mathrm{\Delta }`$ be a versal deformation of the nodal rational curve $`xy=0`$ in $`^2`$. The action $`t(z,w)=(tz,t^1w)`$ of $`^{}`$ on $`_0`$ extends to an action on $``$ over $`\mathrm{\Delta }`$. We may take $`\stackrel{~}{}=(L\times )/^{}`$ with $`^{}`$ acting diagonally.
The points $`\stackrel{~}{C}_0^{}\stackrel{~}{C}_1^{}`$ and $`\stackrel{~}{C}_2^{\prime \prime }\stackrel{~}{C}_3^{\prime \prime }`$ on $`\stackrel{~}{}_0`$ can be lifted to sections of $`\stackrel{~}{}\mathrm{\Delta }`$. Let $`\stackrel{~}{}`$ be the blowing-up of these sections. The strict transform of $`\stackrel{~}{C}_0^{}\stackrel{~}{C}_2^{\prime \prime }`$ is a Cartier divisor on $`_0`$. It can be extended to an effective Cartier divisor on $``$, whose associated line bundle we denote by $``$. Consider the factorization $`\mathrm{\Delta }\times ^1`$ of $`\mathrm{\Delta }`$. The relative base locus of $``$ over $`\mathrm{\Delta }\times ^1`$ is obviously finite. According to the Fujita-Zariski Theorem (, Thm. 1.10) the corresponding invertible $`𝒪_{}`$-module $``$ is relatively semiample over $`^1\times \mathrm{\Delta }`$. Moreover, $``$ is relatively ample over $`^1\times \mathrm{\Delta }^{}`$. Let $`𝔖`$ be the corresponding contraction. The exceptional locus of this contraction is the strict transform of $`\stackrel{~}{C}_1^{}\stackrel{~}{C}_3^{\prime \prime }`$.
Now $`𝔖\mathrm{\Delta }`$ has as general fiber the blowing up of a Hirzebruch surface of degree $`e`$ in two points as needed for the construction of $`d`$-semistable Kodaira surfaces with rational normalization. The central fiber consists of a union of two Hirzebruch surfaces of degrees $`e+1`$ and $`e1`$ respectively, with a section of degree $`e+1`$ glued to a section of degree $`(e1)`$. So this is a partial normalization of $`X_2`$. The gluing morphism $`\phi :CD`$ on the special fibre $`𝔖_0=S`$ extends to a gluing morphism over $`\mathrm{\Delta }`$. The corresponding cocartesian diagram gives the desired flat family $`𝔛\mathrm{\Delta }`$ with $`𝔛_0=X_2`$.
(ii) For the elliptic case consider the partial normalization $`Y`$ of $`X_2`$ obtained from $`S`$ by identifying $`C_1^{}`$ with $`C_3^{\prime \prime }`$ and $`C_3^{}`$ with $`C_1^{\prime \prime }`$, see the right half of Figure 4. It is the projective closure of a line bundle over the Néron 2-gon. In this picture the two disjoint sections $`C_0=C_0^{}C_0^{\prime \prime }`$, $`C_2=C_2^{}C_2^{\prime \prime }`$ are the zero section and the section at infinity. Let $`𝔅\mathrm{\Delta }`$ be a smoothing of $`B`$. Since the line bundle defining $`Y`$ extends to $`𝔅`$ there exists an extension $`𝔜𝔅`$ of $`YB`$ with disjoint sections $`_0,_2𝔜`$. The gluing $`C_0C_2`$ is given by an automorphism of $`B`$ of order $`2`$. This automorphism can be extended to an automorphism of $`𝔅/\mathrm{\Delta }`$ of the same order. An appropriate identification of the sections now yields the desired deformation of $`X_2`$. ∎ |
warning/0001/math0001075.html | ar5iv | text | # Numerical investigation of a dipole type solution for unsteady groundwater flow with capillary retention and forced drainage
## 1. Introduction.
In the present work two problems from the theory of filtration through a horizontal porous stratum are considered. First we study a short, but intense, flooding followed by natural outflow through the vertical face of an aquifer. Further, we consider the possibility to control the spreading of the water mound by use of forced drainage at the boundary.
An important practical example of such a problem is groundwater mound formation and extension following a flood, after a breakthrough of a dam, when water (possibly contaminated) enters and then slowly extends into a river bank.
Consider an aquifer that consists of a long porous stratum with an impermeable bed at the bottom and a permeable vertical face on one side (Fig. 1). The space coordinate $`x`$ is directed along the horizontal axis with $`x=0`$ at the vertical face. A water reservoir is located in the region $`x<0`$. We assume that the flow is homogeneous in the y-direction. The height of the resulting mound is denoted by $`z=h(x,t)`$. The initial level of water in the stratum is assumed to be negligible.
The problem is formulated as follows. At some time $`t=\tau <0`$, the water level at the wall begins to rise rapidly, and water enters the porous medium. By time $`t=0`$, the water level at the vertical face returns to the initial one.We assume that the distribution at time $`t=0`$ is given by $`h(x,0)=h_0(x)`$ and is concentrated over a finite region $`[0,d]`$ (compactly supported). We also assume that $`h_0(x)`$ is concave down and $`h_0^2(x)`$ is gently sloping.
In problem 1, water naturally seeps through the boundary back into the reservoir, giving the boundary condition $`h(0,t)=0`$. Inclusion of the effects of capillary retention into the model distinguishes our case from the well know dipole-type problem. The numerical and asymptotic solutions for the source-type boundary conditions were obtained in . Most recently, the dipole-type problem with capillary retention was studied numericaly and analytically, using Lie-group techniques, by B. Wagner in .
Analysis and numeric computations show that in the case of natural outflow, the water mound is not extinguished in finite time. The outflow rate cannot be further increased by lowering the level at the boundary. In problem 2, in order to control the spreading of the water mound, forced drainage is introduced. The problem formulation was proposed in , where a complete mathematical derivation and rigorous analysis can be found. The forced drainage can be implemented, for instance, by drilling a number of holes or horizontal wells near the impermeable bottom. In this way, an additional discharge rate is created, and the fluid level becomes zero on some interval $`[0,x_l(t)]`$.
These are certainly highly idealized problems, but their solutions allow one to extract the qualitative properties and to check the numerical methods in solving more realistic problems.
## 2. Porous medium equation with capillary retention.
In the case of seepage and gently sloping profile $`h^2(x,t)`$ and in the absence of capillary retention, the model of flow in a porous stratum is described by the Boussinesq equation ( see also ,):
(2.1)
$$_th=\kappa _{xx}h^2.$$
Here $`\kappa =k\rho g/2\mu m`$, $`k`$ is the permeability of the medium, $`m`$ its porosity (the fraction of the volume in the stratum which is occupied by the pores), $`\rho `$ the fluid density, $`\mu `$ its dynamic viscosity, and $`g`$ the acceleration of gravity. According to the hydrostatic law, water pressure $`p=\rho g(hz)`$. Then, the total head $`H=p+\rho gz=\rho gh`$ is constant throughout the height of the mound. Under the assumption of seepage and gently sloping profiles $`h^2`$, Darcy law is used to obtain the relation for the total flux $`q=\frac{k}{\mu }Hh`$.
Mathematical properties of the Boussinesq equation are well known . An essential feature of this equation is the finite speed of disturbance propagation given a finite (compactly supported) initial distribution. Another important feature of this equation is the existence of special self-similar solutions. The graphs of such a solution for any two times $`t_0`$ and $`t_1`$ are related via a similarity transformation . The special solutions, themselves corresponding to certain, sometimes artificial, initial and boundary conditions, are important because they provide intermediate asymptotics for a wide class of initial value problems. For these problems, the details of the initial distribution affect the solution only in the beginning; after some time, the solution approaches a self-similar asymptotics. The Boussinesq equation has been studied extensively and a number of self-similar solutions, for different boundary conditions, have been constructed (, ).
Following , , the Boussinesq equation can be modified to incorporate the effects of capillary retention into the model. If we exclude the possibility of water reentering the region that was filled with water at some earlier time and assume that initially the stratum is empty, we have the following situation: when water enters a pore, it occupies the entire volume, allowed by active porosity; when water leaves the pore, a fraction $`\delta `$ of the pore volume remains occupied by the trapped water. We assume that $`\delta `$ is constant. Let us denote the initial active porosity by $`m`$. Then, when water is entering previously unfilled pores, the effective porosity is $`m`$; when water is leaving previously water-filled pores, the effective porosity becomes $`m(1\delta )`$. Hence, in the presence of capillary retention, porosity depends on the sign of $`_th(x,t)`$. Notice, that permeability can be assumed unaffected, as the effect of capillary forces on permeability is significant only for small and/or dead-end pores, whose contribution to the total flux, in the first approximation, can be neglected.
The rate of change in the amount of water $`\mathrm{\Delta }V`$ inside a volume element (Fig. 1) is equal to:
(2.2)
$$\mathrm{\Delta }V=\{\begin{array}{cc}m\frac{h}{t}\mathrm{\Delta }x\hfill & \text{if }\frac{h}{t}>0\text{ },\hfill \\ m(1\delta )\frac{h}{t}\mathrm{\Delta }x\hfill & \text{if }\frac{h}{t}<0.\hfill \end{array}$$
On the other hand, the rate of change in the volume of water due to the flux through the faces of a volume element $`(x,x+\mathrm{\Delta }x)`$ is equal to
(2.3)
$$\mathrm{\Delta }x_x(\frac{k}{\mu }_xHh)=\mathrm{\Delta }x\frac{k\rho g}{2\mu }_{xx}(h^2).$$
We denote $`\kappa _1=\frac{k\rho g}{2\mu m}`$ and $`\kappa _2=\frac{k\rho g}{2\mu m(1\delta )}`$. Then, using the continuity of flux (no sources inside the water mound) and the balance of mass we obtain:
(2.4)
$$\frac{h}{t}=\{\begin{array}{cc}\kappa _1_{xx}(h^2)\hfill & \text{if }\frac{h}{t}>0\text{ },\hfill \\ \kappa _2_{xx}(h^2)\hfill & \text{ if }\frac{h}{t}<0.\hfill \end{array}$$
This is a nonlinear parabolic partial differential equation with discontinuous coefficients, also known as the generalized porous medium equation , .
Continuity of the flux $`q=\frac{\rho gk}{\mu }h_xh`$ implies that at the mound tip $`x_r`$, where mound height is zero, the flux is also zero. For problem 1, these considerations lead to the following initial and boundary conditions to supplement equation (2.4):
$`h(x,0)`$ $`=`$ $`h_0(x)0\text{ (where }h_0(x)=0\text{ for }xd\text{)},`$
(2.5) $`h(x,t)`$ $`=`$ $`0\text{}_xh^2(x,t)=0\text{ at }x=x_r,`$
$`h(0,t)`$ $`=`$ $`0.`$
The second line in (2) corresponds to the free boundary conditions on the right boundary, $`x_r(t)`$, which is unknown a priori.
It should be noted that for the solution of equation (2.1) (but not for (2.4)) with boundary conditions (2) the dipole moment is constant:
(2.6)
$$Q=\underset{0}{\overset{\mathrm{}}{}}xh(x,t)𝑑x=C.$$
We call equation (2.4) with boundary conditions (2) a dipole-type problem. A similar problem, for source type initial and boundary conditions was considered in , see also .
For problem 2, the boundary conditions are changed to include the forced drainage condition. The discharge rate $`q_0(t)`$, which is a quantity that should be specified, determines the boundary condition at the left free boundary $`x_l`$.
$`h(x,0)`$ $`=`$ $`h_0(x)0\text{ (where }h_0(x)=0\text{ for }xd\text{)},`$
(2.7) $`h(x,t)`$ $`=`$ $`0,_xh^2(x,t)=0\text{ at }x=x_r,`$
$`h(x,t)`$ $`=`$ $`0,_xh^2(x,t)={\displaystyle \frac{q_0(t)}{m\kappa }}\text{ at }x=x_l\text{ }.`$
The second and third lines in (2) define, respectively, the free boundary condition on the right boundary and the forced drainage condition on the left boundary. Equation (2.4) together with boundary conditions (2) define problem 2.
## 3. Dimensional analysis of problem 1.
The parameters in the problem are $`h,x,t,\kappa _1,\kappa _2`$, $`d`$ \- the initial width of the water mound, and $`Q=Q(0)`$ \- the initial dipole moment. We can take the dimensions as follows: $`[h]=H`$, $`[x]=L`$, $`[t]=T`$. Then from equation (2.4) we have $`[\kappa ]=\frac{L^2}{TH}`$. For the remaining parameters $`[d]=L`$, $`[Q]=HL^2`$. The dimensions for $`h`$ and $`x`$ are set to be independent. This can be done because the differential equation (2.4) is invariant with respect to the following group of transformations:
(3.1)
$$x^{}=\alpha x,\text{ }t^{}=\frac{\alpha ^2}{\gamma }t,\text{ }h^{}=\gamma h.$$
The invariance insures that we can scale the units of measurement for $`h`$, while keeping the units for $`x`$ unchanged.
The following dimensionless quantities can be obtained from these parameters:
$$\mathrm{\Pi }_1=\frac{x}{(Q\kappa _1t)^{1/4}},\mathrm{\Pi }_2=\frac{d}{(Q\kappa _1t)^{1/4}},\mathrm{\Pi }_3=\frac{\kappa _1}{\kappa _2},\text{and }\mathrm{\Pi }=h(\frac{\kappa _1t}{Q})^{1/2}.$$
It follows that $`\mathrm{\Pi }=F(\mathrm{\Pi }_1,\mathrm{\Pi }_2,\mathrm{\Pi }_3)`$.
Since for large times, $`t\frac{\mathrm{\Delta }x^4}{Q\kappa _1}`$, the parameter $`\mathrm{\Pi }_21`$, it would seem natural to set $`\mathrm{\Pi }=f(\mathrm{\Pi }_1,\mathrm{\Pi }_3)`$, as in the case of $`\kappa _1=\kappa _2`$, and look for a solution of the form:
(3.2)
$$h=(\frac{Q^2}{\kappa _1t})^{1/2}f(z,\frac{k_1}{k_2}),\text{ where }z=\frac{x}{(Q\kappa _1t)^{1/4}}.$$
However, this leads to a contradiction when we consider an ordinary differential equation obtained from (2.4):
(3.3)
$$(2f+\frac{df}{dz}z)=\{\begin{array}{cc}4\frac{d(f^2)}{dz^2}\hfill & \text{if }2f+\frac{df}{dz}z<0,\hfill \\ 4\frac{\kappa _1}{\kappa _2}\frac{d(f^2)}{dz^2}\hfill & \text{if }2f+\frac{df}{dz}z>0.\hfill \end{array}$$
Multiplying both sides by $`z`$ we obtain an equation in total differentials, which is readily solved:
(3.4)
$$fz^2=\{\begin{array}{cc}4(z\frac{df^2}{dz}f^2)+C_1\hfill & \text{if }2f+\frac{df}{dz}z<0,\hfill \\ 4\frac{\kappa _1}{\kappa _2}(z\frac{df^2}{dz}f^2)+C_2\hfill & \text{if }2f+\frac{df}{dz}z>0.\hfill \end{array}$$
Observe that near $`z=z_r`$, where the height of the mound vanishes, the first equation holds. At $`z=z_r`$, $`h`$ vanishes along with the flux, which is proportional to $`\frac{h^2}{x}`$. From the first equation at $`z_r`$, we obtain that $`C_1=0`$. Similarly, evaluating the second expression at $`z=0`$, where $`h=0`$, we find that $`C_2=0`$. Next, evaluating the two expressions at $`z_1`$, we obtain:
(3.5)
$$(\frac{df^2}{dz}(z_1)z_1f^2(z_1))(1\frac{\kappa _1}{\kappa _2})=0,$$
and using $`2f+\frac{df}{dz}z=0`$ we obtain $`5f^2(z_1)(1\frac{\kappa _1}{\kappa _2})=0`$. For $`\kappa _2=\kappa _1`$, the solution can be found and thus the assumption of complete similarity in $`\mathrm{\Pi }_2`$ is correct. However, in the case of $`\kappa _2\kappa _1`$, we have $`f(z_1)=0`$. This is a contradiction, because the change in sign of $`\frac{h(x,t)}{t}`$ should occur inside the mound, where the height is positive. Hence, the assumption of complete similarity for $`\kappa _2\kappa _1`$ does not hold.
We next solve the problem numerically and study the asymptotic behavior of the solution.
## 4. Numerical solution of the partial differential equation and further analysis for problem 1.
In order to simplify the numerical solution for equation (2.4) with free boundary conditions (2), we use a change of variables: $`\xi =\frac{x}{x_r}`$. We set $`H(\xi ,t)=h(x_r\xi ,t)`$, and equation (2.4) is transformed:
(4.1)
$$_tH=\{\begin{array}{cc}\frac{1}{x_r^2}(\kappa _1_{\xi \xi }H^2(\xi ,t)+\kappa _1\xi _\xi H(1,t)_\xi H(\xi ,t))\hfill & \text{if }_tH(\xi ,t)>0,\hfill \\ \frac{1}{x_r^2}(\kappa _2_{\xi \xi }H^2(\xi ,t)+\kappa _1\xi _\xi H(1,t)_\xi H(\xi ,t))\hfill & \text{if }_tH(\xi ,t)<0;\hfill \end{array}$$
with boundary conditions $`H(0,t)=H(1,t)=0`$. This effectively fixes the right boundary at $`\xi =1`$.
The location of the free boundary $`x_r(t)`$ can be obtained in the course of the numerical solution in the following way. We assume that the solution is nearly stationary near the tip $`x_r`$ and $`h(x,t)H(xvt)`$. Here, $`v`$ denotes the instantaneous speed of mound extension, which changes slowly as a function of $`t`$. Then $`_thv_xh`$ near $`x_r`$.
Considering equation (2.4) near the boundary $`x_r`$ of the mound, where $`h`$ is small, we have:
$$_th=\kappa _1_{xx}h^2(x,t)=2\kappa _1((_xh(x,t))^2+h_{xx}h(x,t))2\kappa _1(_xh(x,t))^2$$
so that
(4.2) $`v(t)=2\kappa _1_xh(x_r,t),x_r(t)={\displaystyle _0^t}v(t)𝑑t+x_r(0).`$
We solve the new boundary value problem numerically by using a forward-in-time, centered-in-space finite-difference approximation, where $`u_i^n`$ is an approximation to the solution of (4.1) at the grid point $`(x_i,t_n)`$:
$`\kappa _i^n`$ $`=`$ $`\{\begin{array}{cc}\kappa _1\hfill & \text{if }[(u_{i1}^{n1})^22(u_i^{n1})^2+(u_{i+1}^{n1})^2]>0,\hfill \\ \kappa _2\hfill & \text{if }[(u_{i1}^{n1})^22(u_i^{n1})^2+(u_{i+1}^{n1})^2]<0;\hfill \end{array}`$
$`u_i^{n+1}`$ $`=`$ $`u_i^n+{\displaystyle \frac{\mathrm{\Delta }t}{\mathrm{\Delta }x^2}}\{\kappa _i^n[(u_{i1}^n)^22(u_i^n)^2+(u_{i+1}^n)^2]`$
$``$ $`\kappa _1\xi _i(u_i^nu_{i1}^n)(u_N^nu_{N1}^n)\}/(x_r^n)^2,`$
$`x_r^{n+1}`$ $`=`$ $`x_r^n2\kappa _1{\displaystyle \frac{\mathrm{\Delta }t}{\mathrm{\Delta }x}}{\displaystyle \frac{u_Nu_{N1}}{x_r^n}}.`$
In the numerical computation we start with an initial distribution of the source type, localized near $`x=0`$ (Fig.2). Before the left free boundary reaches the point $`x=0`$, the solution is of the source type and we can compare our numerical results to those in . After some time $`t`$ the left free boundary reaches $`x=0`$, where it is thereafter fixed (Fig. 2).
Now we consider the scaled solution:
(4.3)
$$\frac{H(\xi ,t)}{\mathrm{max}_\xi H(\xi ,t)}=f(\xi ,\frac{\kappa _1}{\kappa _2}),\xi =\frac{x}{x_r}.$$
We can see in Figs. 6 and 7 that as time $`t`$ increases the numerical solution approaches a self-similar regime, so that the graphs of the scaled solution for different times “collapse” into a single curve. Moreover, Figs. 3 and 4 show a power-law dependence on time for both $`\mathrm{max}_\xi H`$ and $`x_r`$ in the self-similar regime.
In part 3, we have shown that a self-similar solution of the first kind does not exist for this problem. To explain what happened we return to the dimensional analysis and look now for a generalized self-similar solution.
We have determined that the variables in the problem are related as follows: $`\mathrm{\Pi }=F(\mathrm{\Pi }_1,\mathrm{\Pi }_2,\mathrm{\Pi }_3)`$, where
$$\mathrm{\Pi }_1=\frac{x}{(Q\kappa _1t)^{1/4}},\mathrm{\Pi }_2=\frac{d}{(Q\kappa _1t)^{1/4}},\mathrm{\Pi }_3=\frac{\kappa _1}{\kappa _2},\text{and }\mathrm{\Pi }=h(\frac{\kappa _1t}{Q})^{1/2}.$$
Our numerical investigation shows that for large $`t`$, as $`\mathrm{\Pi }_20`$ :
(4.4)
$$F(\mathrm{\Pi }_1,\mathrm{\Pi }_2,\mathrm{\Pi }_3)=\mathrm{\Pi }_2^\gamma f_2(\mathrm{\Pi }_1\mathrm{\Pi }_2^\epsilon ,\mathrm{\Pi }_3),$$
where $`\gamma `$ and $`\epsilon `$ are constants. In fact, this is the next simplest situation after complete self-similarity and it is referred to as self-similarity of the second kind in $`\mathrm{\Pi }_2`$ (see ).
Indeed, from the analysis above:
$`{\displaystyle \frac{\mathrm{\Pi }_1}{\mathrm{\Pi }_2^\epsilon }}`$ $`=`$ $`xd^\epsilon (t\kappa _1Q)^{\frac{\epsilon 1}{4}},`$
$`\mathrm{\Pi }_2^\gamma `$ $`=`$ $`{\displaystyle \frac{d^\gamma }{(t\kappa _1Q)^{\gamma /4}}},`$
(4.5) $`h(x,t)`$ $`=`$ $`At^\alpha f({\displaystyle \frac{x}{Bt^\beta }},{\displaystyle \frac{\kappa _1}{\kappa _2}}),`$
$`x_r(t)`$ $`=`$ $`Bt^\beta ,`$
where
$$A=(\kappa _1Q)^{\gamma /4}d^\gamma ,B=d^\epsilon (\kappa _1Q)^{\frac{\epsilon 1}{4}},\alpha =\frac{\gamma +2}{4},\text{and }\beta =\frac{1\epsilon }{4}.$$
The parameters $`\alpha `$ and $`\beta `$ depend on the ratio $`\frac{\kappa _1}{\kappa _2}`$. They cannot be determined on the basis of dimensional analysis alone and have to be computed as a part of the solution. We will see that there is, actually, only one unknown parameter involved, since the differential equation provides an additional relation between $`\alpha `$ and $`\beta `$.
## 5. Derivation and numerical solution of a nonlinear eigenvalue problem.
The numerical solution of partial differential equation showed that there is indeed an intermediate asymptotic solution of the form (4). Now, we can obtain such a self-similar solution by transforming the problem of solving partial differential equation (2.4) with boundary conditions (2) into a nonlinear eigenvalue problem.
We substitute (4) into (2.4) and normalizing $`f`$ so that $`\frac{A}{B^2}=1`$ we get:
(5.1) $`_th`$ $`=`$ $`t^{(\alpha +1)}A(\alpha f+\beta f^{}\xi ),`$
(5.2) $`_{xx}h^2`$ $`=`$ $`{\displaystyle \frac{A^2t^{2\alpha }(f^2(\xi ))^{\prime \prime }}{B^2t^{2\beta }}},`$
(5.3) $`\alpha f(\xi )+\beta f^{}(\xi )\xi `$ $`=`$ $`\kappa ^{}(f^2(\xi ))^{\prime \prime }t^{\alpha 2\beta +1},`$
where
$$\kappa ^{}=\{\begin{array}{cc}\kappa _1\hfill & \text{if }((12\beta )f+\beta f^{}\xi )<0,\hfill \\ \kappa _2\hfill & \text{if }((12\beta )f+\beta f^{}\xi )>0.\hfill \end{array}$$
Since equation (5.3) cannot depend on time explicitly, $`\alpha 2\beta +1=0`$. Finally, we get an ordinary differential equation:
(5.4)
$$(f^2)^{\prime \prime }=\{\begin{array}{cc}(\beta f^{}\xi +(12\beta )f)\hfill & \text{if }((12\beta )f+\beta f^{}\xi )<0,\hfill \\ \frac{\kappa _2}{\kappa _1}(\beta f^{}\xi +(12\beta )f)\hfill & \text{if }((12\beta )f+\beta f^{}\xi )>0.\hfill \end{array}$$
The boundaries $`x_0=0`$ and $`x_r=Bt^\beta `$, in the new space variable, correspond to $`\xi =0`$ and $`\xi =1`$. The boundary condition at $`\xi =0`$ becomes:
(5.5)
$$f(0)=0.$$
For the right boundary, $`\xi =1`$, we have:
(5.6)
$$f(1)=0\text{and}(f^2)^{}(1)=0.$$
From 5.4 and 5.6 it follows that $`2\kappa _1(f^{}(1))^2+\beta \xi f^{}(1)=0`$ and the tip conditions become:
(5.7)
$$f(1)=0\text{and}f^{}(1)=\frac{\beta }{2\kappa _1}.$$
The second order ODE (5.4) with three boundary conditions (5.7 and 5.5) constitutes a non-linear eigenvalue problem, which we now have to solve numerically. For each value of $`\frac{\kappa _1}{\kappa _2}`$, we find a value of $`\beta `$ such that the boundary conditions are satisfied.
We use a high order, Taylor-expansion-based method to start the integration at $`\xi =1`$ followed by a 4th order Runge-Kutta method and an iterative procedue to arrive at the value for $`\beta `$ such that the third condition $`f(0)=0`$ is satisfied. For computational convenience, we transform the differential equation by changing variables: $`g(\xi )=f^2(\xi )`$, so that $`g(\xi )`$ does not have a singularity at $`\xi =0`$. In this manner, we obtain the dependence of $`\beta `$ on $`\frac{\kappa _1}{\kappa _2}`$ (Fig. 5).
## 6. Comparison of the results for problem 1.
In logarithmic coordinates, we obtain from (4):
(6.1) $`\mathrm{log}(u(1/2,t))`$ $`=`$ $`\alpha \mathrm{log}(t)+\mathrm{log}(Af(1/2,{\displaystyle \frac{\kappa _1}{\kappa _2}}))`$
(6.2) $`\mathrm{log}(x_r(t))`$ $`=`$ $`\beta \mathrm{log}(t)+\mathrm{log}(B)`$
i.e., straight lines with slopes $`\alpha `$ and $`\beta `$.
From plots in Figs. 3 and 4, we can observe that after some initial time $`t`$ both graphs approach straight lines. We repeat the calculations for a range of values of $`\frac{\kappa _1}{\kappa _2}`$.
Comparison of the results of the numerical solution of the nonlinear eigenvalue problem with the results obtained from the numerical solution to the partial differential equation (Fig. 5), shows that the two agree with high precision.
Also, the exact solution for the case $`\kappa _1=\kappa _2`$ gives the value $`\beta =.25`$, which coincides with the results of the numerical computations with good accuracy.
## 7. Numerical solution of the partial differential equation for problem 2.
Although problems 1 and 2 are similar, the numerical treatment of problem 2 is more complicated. Time evolution of the left boundary in problem 2 makes rescaling, which was used in the numerical solution of problem 1, infeasible. Instead, we solve equation (2.4) on a grid, taking into account that the left and right boundaries may not fall onto gridpoints. We determine new positions of the boundaries from the numerical solution at each timestep.
Equation (2.4) is discretized using a forward-in-time, centered-in-space finite-difference scheme:
$`\kappa _i^n`$ $`=\{\begin{array}{cc}\kappa _1\hfill & \text{if }[(u_{i1}^{n1})^22(u_i^{n1})^2+(u_{i+1}^{n1})^2]>0,\hfill \\ \kappa _2\hfill & \text{if }[(u_{i1}^{n1})^22(u_i^{n1})^2+(u_{i+1}^{n1})^2]<0;\hfill \end{array}`$
$`u_i^{n+1}`$ $`=u_i^n+{\displaystyle \frac{\mathrm{\Delta }t}{\mathrm{\Delta }x^2}}\{\kappa _i^n[(u_{i1}^n)^22(u_i^n)^2+(u_{i+1}^n)^2]\},`$
(7.1) $`u_l^{n+1}`$ $`=u_l^n+{\displaystyle \frac{2\mathrm{\Delta }t}{\mathrm{\Delta }x+\mathrm{\Delta }x_l}}\kappa _2^n\{{\displaystyle \frac{(u_{l+1}^n)^2(u_l^n)^2}{\mathrm{\Delta }x}}q^n\},`$
$`u_r^{n+1}`$ $`=u_r^n+{\displaystyle \frac{2\mathrm{\Delta }t}{\mathrm{\Delta }x+\mathrm{\Delta }x_r}}\kappa _1^n\{{\displaystyle \frac{(u_{r1}^n)^2(u_r^n)^2}{\mathrm{\Delta }x}}{\displaystyle \frac{(u_r^n)^2}{\mathrm{\Delta }x_r}}\}.`$
Here $`u_l`$ and $`u_r`$ are the nonzero values of $`u`$ on the grid, adjacent to the left and right boundaries respectively, $`q`$ is the drainage flux, $`\mathrm{\Delta }x_l`$ and $`\mathrm{\Delta }x_r`$ are distances from the left and right boundaries to the grid points. We treat the values $`u_l`$ and $`u_r`$ separately in order to incorporate the boundary conditions and improve precision.
The location of the left boundary is obtained from the values of $`u`$:
(7.2)
$$x_l^{n+1}=x_l^n+\mathrm{\Delta }x_l\frac{(u_l^n)^2}{q}.$$
The right boundary location is obtained by extrapolation from the values of $`u^{n+1}`$.
We check the numerical method for $`\kappa _1=\kappa _2`$ by comparing the numerical solution with a known analytic solution. The exact self-similar solutions for the problems with forced drainage are given in . We choose a value of $`\beta .25`$ and then solve an ordinary differential equation (5.3) with the initial condition (5.7). For $`\beta <.25`$ the solution $`f(\xi )`$ of the ordinary differential equation intersects the $`x`$-axis at some point $`\xi =\lambda >0`$ and at $`\xi =1`$.
From the solution of the ordinary differential equation we obtain a self-similar solution:
(7.3)
$$h(x,t)=Bt^{(12\beta )}f(\frac{x}{At^\beta })$$
of the partial differential equation. The locations of the free boundaries are given by $`x_r(t)=At^\beta `$, $`x_l(t)=\lambda At^\beta `$, and the drainage flux is given by $`q(t)=mBAt^{2+3\beta }(f^2)^{}(\lambda )`$ (see ). We use the self-similar solution at some time $`t_0`$ as an initial value for the numerical solver, set drainage flux on the left boundary to be $`q(t)`$, and compute the solutions until time $`t_1`$. As in the analysis in section 6, the graphs of $`x_r(t)`$, $`x_l(t)`$ should be straight lines in logarithmic coordinates, and the graphs of the scaled solution for different times $`t`$ should collapse into one curve. That’s what we observe in Figs. 9 and 10.
Now, we try to model the conditions of a flood followed by forced drainage, as described in the introduction. We begin by computing the solution to problem 1 until some time $`t_0`$, which corresponds to the flood followed by natural drainage through the boundary of the aquifer. After $`t_0`$, we set a constant drainage flux $`q(t)=q_0`$ at the left boundary. In particular, we set $`q_0`$ to equal twice the natural drainage flux at time $`t_0`$. As we see in Fig. 12, the water mound, that has appeared after the flood, is completely extinguished in finite time.
## 8. Conclusion.
1. The numerical simulations of two problems involving drainage and capillary retention of the fluid a in porous medium were presented. It was shown that the problem with dipole type initial and boundary conditions has a self-similar intermediate asymptotics in the case of a porous medium with capillary retention.
2. A problem of control of the water mound extension by forced drainage was considered. The possibility of extinguishing the propagating water mound by creating a forced drainage flux $`q(t)`$ at the left boundary was confirmed numerically. Using our results, it should be possible to derive a cost efficient drilling regime and to localize the mound and contain the contamination inside a prescribed region.
It would be interesting to extend the numerical investigation above to the case of a fissurized porous medium.
## 9. Acknowledgements.
The authors are grateful to Professor G.I. Barenblatt, without whose direction and advice this work would not have been possible. The authors use this occasion to thank Professor A. Chorin for many helpful discussions of this work and for his constant attention and encouragement.
This work was supported in part by the Computational Science Graduate Fellowship Program of the Office of Scientific Computing in the Department of Energy, NSF grant contract DMS-9732710, and the Office of Advanced Scientific Computing Research, Mathematical, Information, and Computational Sciences Division, Applied Mathematical Sciences Subprogram, of the U.S. Department of Energy, under Contract No. DE-AC03-76SF00098.
## References
* G.I. Barenblatt, *Scaling, self-similarity, and intermediate asymptotics*, first ed., Cambridge University Press, New York, 1996.
* G.I. Barenblatt, V.M. Entov, and V.M. Ryzhik, *Theory of fluid flows through natural rocks*, first ed., Kluwer Academic Publishers, Dordrecht, 1990.
* G.I. Barenblatt and J.L. Vasquez, *A new free boundary problem for unsteady flows in porous media*, Euro. Jnl of Applied Mathematics 9 (1998), 37–54.
* C.W. Fetter, *Applied hydrogeology*, third ed., Macmillan College Publishing Company, New York, 1988.
* A.S. Kalashnikov, *Some problems of qualitative theory of the non-linear second-order parabolic equations*, Russian Math. Surveys (1987), no. 42, 169–222.
* I.N. Kochina, N.N. Mikhailov, and M.V. Filinov, *Groundwater mound damping.*, Int. J. Engng Sci 21 (1983), no. 4, 413–421.
* B.A. Wagner, *Perturbation techniques and similarity analysis for the evolution of interfaces in diffusion and surface tension driven problems*, Zentrum Mathematik, TU Munchen, 1999. |
warning/0001/cs0001019.html | ar5iv | text | # PushPush is NP-hard in 2D
## 1 Introduction
There are a variety of “sliding blocks” puzzles whose time complexity has been analyzed. One class, typified by the 15-puzzle so heavily studied in AI, permits an outside agent to move the blocks. Another class falls more under the guise of motion planning. Here a robot or internal agent plans a path in the presence of movable obstacles. This line was initiated by a paper of Wilfong \[Wil91\], who proved NP-hardness of a particular version in which the robot could pull as well as push the obstacles, which were not restricted to be squares. Subsequent work sharpened the class of problems by weakening the robot to only push, never pull obstacles, and by restricting all obstacles to be unit squares. Even this version is NP-hard when some blocks may be fixed to the board (made unpushable) \[DO92\].
One theme in this research has been to establish stronger degrees of intractability, in particular, to distinguish between NP-hardness and PSPACE-completeness, the latter being the stronger claim. The NP-hardness proved in \[DO92\] was strengthened to PSPACE-completeness in an unfinished manuscript \[BOS94\]. More firm are the results on Sokoban, a computer game that restricts the pushing robot to only push one block at a time, and requires the storing of (some or all) blocks into designated “storage locations.” This game was proved NP-hard in \[DZ95\], and PSPACE-complete by Culberson \[Cul98\].
Here we emphasize another theme: finding a nontrivial version of the game that is not intractable. To date only the most uninteresting versions are known to be solvable in polynomial time, for example, where the robot’s path must be monotonic \[DO92\]. We explore a different version, again inspired by a computer game, PushPush.<sup>1</sup><sup>1</sup>1 The earliest reference we can find to the game is a version written for the Macintosh by Alan Rogers and C.M. Mead III, Copyright 1994, http://www.kidsdomain.com/down/mac/pushpush.html. Another version for the Amiga was written by Luigi Recanatese in 1997, http://de.aminet.net/aminet/dirs/game\_think.html. The key difference is that when a block is pushed, it necessarily slides the full extent of the available empty space in the direction in which it was shoved. This further weakens the robot’s control, and the resulting puzzle has certain polynomial characteristics. It was established in \[OS99\] that the problem is intractable in 3D, but its status in 2D was left open in that paper. Here we settle the issue by extending the reduction to 2D.
## 2 Problem Classification
The variety of pushing-block puzzles may be classified by several characteristics:
1. Can the robot pull as well as push?
2. Are all blocks unit squares, or may they have different shapes?
3. Are all blocks movable, or are some fixed to the board?
4. Can the robot push more than one block at a time?
5. Is the goal for the robot to move from $`s`$ to $`t`$, or is the goal for the robot to push blocks into storage locations?
6. Do blocks move the minimal amount, exactly how far they are pushed, or do they slide the maximal amount of their free range?
7. The dimension of the puzzle: 2D or 3D?
If our goal is to find the weakest robot and most unconstrained puzzle conditions that still lead to intractability, it is reasonable to consider robots who can only push (1), and to restrict all blocks to be unit squares (2), as in \[DO92, DZ95, Cul98\], for permitting robots to pull, and permitting blocks of other shapes, makes it relatively easy to construct intractable puzzles. It also makes sense to explore the goal of simply finding a path (5) as in \[Wil91, DO92\], rather than the more challenging task of storing the blocks as in Sokoban \[DZ95, Cul98\].
Restricting attention to these choices still leaves a variety of possible problem definitions. If the robot can only move one block at a time, then the distinction between all blocks movable and some fixed essentially disappears, because $`2\times 2`$ clusters of blocks are effectively fixed to a robot who can only push one. If all blocks are movable and the robot can push more than one at a time, then the blocks should be confined to a rectangular frame.
The version explored in this paper superficially seems that it might lend itself to a polynomial-time algorithm: the robot can only push one block (4), all blocks are pushable (3), and finally, the robot’s control over the pushing is further weakened by condition (6): once pushed, a block slides (as without friction) the maximal extent of its free range in that direction. Allowing the robot to move in 3D gives it more “power” than it has in 2D, so the natural question after \[OS99\] is to explore the weaker option of condition (7): PushPush in 2D.
Because our proof is a direct extension of the proof for 3D in \[OS99\], we repeat through Section 6 the 3D construction in that paper (with a minor simplification), and in Section 7 show a similar reduction holds in 2D as well. (Both reductions are from SAT, i.e., satisfiability of formulas in conjunctive normal form.) A summary of related results is presented in the final section.
## 3 Elementary Gadgets
First we observe, as mentioned above, that any $`2\times 2`$ cluster of movable blocks is forever frozen to a PushPush robot, for there is no way to chip away at this unit. This makes it easy to construct “corridors” surrounded by fixed regions to guide the robot’s activities. We will only use corridors of width 1 unit, with orthogonal junctions of degree two, three, or four. We can then view a particular PushPush puzzle as an orthogonal graph, whose edges represent the corridors, understood to be surrounded by sufficiently many $`2\times 2`$ clusters to render any movement outside the graph impossible. We will represent movable blocks in the corridors or at corridor junctions as circles.
We start with three elementary gadgets.
### 3.1 One-Way Gadget
A One-Way gadget is shown in Fig. 1a. It has these obvious properties:
###### Lemma 1
In a One-Way gadget, the robot may travel from point $`x`$ to point $`y`$, but not from $`y`$ to $`x`$. (After travelling from $`x`$ to $`y`$, however, the robot may subsequently return from $`y`$ to $`x`$.)
Proof: The block at the degree-three junction may be pushed into the storage corridor when approaching from $`x`$, as illustrated in Fig. 1b, but the block may not be budged when approaching from $`y`$ (Fig. 1c). $`\mathrm{}`$
### 3.2 Fork Gadget
The Fork gadget<sup>2</sup><sup>2</sup>2 This is a simplification of the functionally equivalent gadget used in \[OS99\]. shown in Fig. 2a presents the robot with a binary choice, the proverbial fork in the road:
###### Lemma 2
In a Fork gadget, the robot may travel from point $`x`$ to $`y`$, or from $`x`$ to $`z`$, but if it chooses the former it cannot later move from $`y`$ to $`z`$, and if it chooses the latter it cannot later move from $`z`$ to $`y`$. (In either case, the robot may reverse its original path.)
Proof: Fig. 2b shows the only way for the robot to pass from $`x`$ to $`y`$. Now the corridor to $`z`$ is permanently sealed off by block $`A`$. Fig. 2c shows the only way to move from $`x`$ to $`z`$, which similarly seals off the path to $`y`$. $`\mathrm{}`$
Note that in both these gadgets, the robot may reverse its path, a point to which we will return in Section 7.
### 3.3 3D Crossover Gadget
Crossovers are trivial in 3D, as shown in Fig. 3.
## 4 Variable-Setting Component
The robot first travels through a series of variable-setting components, each of which follows the structure shown in Fig. 4: a Fork gadget, followed by two paths, labeled t and f, each with attached wires exiting to the right, followed by a re-merging of the t and f paths via One-Way gadgets. 3D crossovers are illustrated in this and subsequent figures by broken-wire underpasses.
###### Lemma 3
The robot may travel from $`a`$ to $`b`$ only by choosing either the t-path, or the f-path, but not both. Whichever t/f-path is chosen allows the robot to travel down any wires attached to that path, but down none of the wires attached to the other path.
Proof: The claims follow directly from Lemma 2 and Lemma 1. $`\mathrm{}`$
## 5 Clause Component
The clause component shown in Fig. 5a cannot be traversed unless one or more blocks (“keys”) are pushed in from the left along the attached horizontal wires.
###### Lemma 4
The robot may only pass from $`x`$ to $`y`$ of a clause component if at least one block is pushed into it along an attached wire ($`a`$, $`b`$, or $`c`$ in Fig. 5a).
Proof: Block $`A`$ is necessarily pushed by the robot starting at $`x`$. This block will clog exit at $`y`$ (Fig. 5b) unless its sliding is stopped by a block pushed in on an attached wire. $`\mathrm{}`$
The basic mechanism that gives a clause component its functionality will be reused in several guises in Section 7, so we pause to redescribe it. Essentially the component is a lock with three keys, which we identify with the three blocks on the $`a`$, $`b`$, and $`c`$ wires. A necessarily-pushed block $`A`$ is characteristic of locks, as is an alternate path around the spot(s) where the key(s) come to rest.
## 6 Complete SAT Reduction
The complete construction for four clauses $`C_1C_2C_3C_4`$ is shown in Fig. 6. Two versions of the clauses are shown in the figure: an unsatisfiable formula (the dark lines), and a satisfiable formula (including the shaded $`x_2`$ wire):
$`(x_1x_2)(x_1x_2)(x_1x_3)(x_1x_3)`$ (1)
$`(x_1x_2)(x_1x_2)(x_1x_2x_3)(x_1x_3)`$ (2)
Here we are using $`x`$ to represent the negation of the variable $`x`$.
A path from $`s`$ to $`t`$ in the satisfiable version is illustrated in Fig. 7.
###### Theorem 1
PushPush is NP-hard in 3D.
Proof: The construction clearly ensures, via Lemmas 3 and 4, that if the simulated Boolean expression is satisfiable, there is a path from $`s`$ to $`t`$, as illustrated in Fig. 7. For the other direction, suppose the expression is unsatisfiable. Then the robot can reach $`t`$ only by somehow “shortcutting” the design. The design of the variable components ensures that only one of the t/f paths may be accessed. The crossovers ensure there is no “leakage” between wires. The only possible thwarting of the design would occur if the robot could travel from a clause component back to set a variable to the opposite Boolean value. But each variable-clause wire contains a block that prevents any such leakage. $`\mathrm{}`$
## 7 2D Crossovers
We now modify the 3D SAT reduction to a 2D SAT reduction by replacing the 3D crossovers with appropriate 2D crossovers; it is only here that we deviate substantively from \[OS99\]. Note that there are two distinct types of crossovers used in the construction:
1. FT-crossovers: A horizontal wire from an f-wire in a variable unit crosses the vertical t-wire of the same variable unit.
2. VC-crossovers: A horizontal wire from some variable unit to a clause unit crosses a vertical wire from some other variable unit to some clause unit.
The FT-crossovers are significantly different from the VC-crossovers in that the former are traversed in one direction or the other but never both. This is because once the robot chooses the t-wire, it can never get into the f-wire, and vice versa (Lemma 3). Thus a limited crossover suffices here.
There is another approach to handling the FT-crossovers: reduction from “Planar 3-SAT” \[Lic82\] (as used, e.g., in Dor and Zwick’s NP-hardness proof \[DZ95\]) permits eliminating this type of crossover entirely. However, we do not pursue that tack, for use of Planar 3-SAT introduces additional crossovers in the final clause-threading path. Instead we develop a limited crossover gadget for FT-crossovers, which may serve as an introduction to the more demanding VC-crossovers in Section 7.2.
### 7.1 XOR Crossover
We call the limited crossover gadget an XOR Crossover; it is shown in Fig. 8(a).
###### Lemma 5
An XOR Crossover gadget permits either horizontal, rightward passage without leakage into the vertical channel, or vertical, downward passage without leakage into the horizontal channel, but not both. In either case, the passage may afterwards be traversed an arbitrary number of times in the same direction.
Proof: Entering the unit from the f-wire to the left pushes block $`A`$ into the vertical channel (Fig. 8(b)), which, together with block $`B`$, disallows entry into the vertical $`T`$-wire ((c) of the figure). Similarly, entering the unit from the t-wire at the top pushes block $`B`$ to clog the horizontal channel (d), preventing leakage into either the f-wire or the clause unit (e). $`\mathrm{}`$
### 7.2 Locking Door Unit
VC-crossovers do not have the exclusive-or property that makes FT-crossovers so easy to handle. They may need to be traversed in either direction. However, note that VC-crossovers need only be traversed once in each of the four directions: from a variable component $`x_i`$, to deposit a key into a clause component $`C_j`$, and returning back to $`x_i`$; and later from $`x_k`$, $`ki`$, to some $`C_l`$, and back to $`x_k`$. The design of such a crossover is the most complex part of our construction, and will proceed in several stages.
The most important gadget used to build this crossover is one that permits passage in one direction, but then prevents return in the other direction. We call this a Locking Door Unit; it is shown in Fig. 9(a). Like a One-Way gadget, it may only be traversed in one direction. But unlike that gadget, once traversed it becomes a permanently locked door with respect to both directions. A more accurate name for the gadget might be “unidirectional, single-use, self-closing and self-locking door.”
Note that the unit contains a lock mechanism (similar to that used in a clause component) centered around point $`b`$, which requires the key block $`B`$ to permit passage.
###### Lemma 6
Upon first encounter, the robot may pass from $`x`$ to $`y`$ through a locking door unit, and not $`y`$ to $`x`$; but once through, the unit becomes impassable in either direction.
Proof: We first argue that the unit may be passed through as claimed. The robot first moves from $`x`$ to below block $`A`$, which it pushes upward to point $`a`$. This seals off return passage. It then pushes block $`B`$ down to point $`b`$, as shown in Fig. 9(b). $`B`$ serves as a key to an exit door. It then “unlocks” that door by pushing block $`C`$ to abut against $`B`$. It may then travel around $`CB`$, push block $`D`$ to the right, and reach the exit $`y`$. See (c) of the figure. Note that it now impossible for the robot to return from $`y`$ to $`x`$; moreover it is impossible for the robot to retraverse the unit from $`x`$ to $`y`$ again, in both cases because of block $`A`$ at point $`a`$.
Next we argue that the above is the only way to traverse the unit. It may not be traversed from $`y`$ to $`x`$ because of the One-Way gadget represented by block $`D`$. We consider the four options of what to do with block $`A`$ after entering at $`x`$:
1. Push $`A`$ upward. This leads to the solution just described.
2. Do not move $`A`$. Then $`B`$ is unreachable, and the only option is to push block $`C`$ rightwards. But without the key $`B`$ at $`b`$, $`C`$ slides over to abut $`D`$ and clog the exit. See (d) of the figure.
3. Push $`A`$ downward. Then $`A`$ slides down to touch $`C`$. Now the robot can only push the key $`B`$ down to $`b`$ ((e) of the figure), but then the door cannot be “unlocked” with $`C`$ as that is now inaccessible.
4. Push $`A`$ rightward. Then $`A`$ hits $`B`$, and prevents the use of the key at $`b`$. Now, similar to (d) of the figure, $`C`$ slides all the way to $`D`$ and prevents exit.
This exhausts the options and establishes the claims of the lemma. $`\mathrm{}`$
We will use the symbols illustrated in Fig. 10(a-b) to represent a locking door before (a) and after (b) passage.
### 7.3 Double Lock Unit
We next detail a gadget that behaves like a locking door, in that passage is permitted once in one direction, but which requires an external key to operate. We call this a Double Lock Unit. It is illustrated in Fig. 11(a).
###### Lemma 7
The robot may pass from $`x`$ to $`y`$ through a double lock unit only if a block $`A`$ (the external key) is first pushed down the entrance $`x`$-wire. A double lock may not be traversed from $`y`$ to $`x`$, and once traversed forwards from $`x`$ to $`y`$, the unit becomes impassable in either direction.
Proof: We call the two locks in the unit the $`A`$\- and $`B`$-locks. The former consists of the key block $`A`$, block $`C`$, and the wires near point $`a`$; the latter consists of the key block $`B`$, block $`D`$, and the wires near point $`b`$. Passage through the unit when there is an external block $`A`$ is shown in Fig. 11(b). Key block $`A`$ is pushed to point $`a`$, and then the robot goes through the locking door unit to reach $`B`$. This $`B`$-key is pushed to point $`b`$. Now both locks have keys. The robot then pushes block $`C`$ leftwards and passes through the $`A`$-lock, pushes block $`D`$ rightwards and passes through the $`B`$-lock, pushes block $`E`$ out of the way, and finally exits at $`y`$.
Suppose there is no external $`A`$ block, and the robot attempts to traverse the unit. To reach $`y`$, the robot must pass through the $`B`$-lock. If it approaches from the left without first pushing down the $`B`$-key, then block $`D`$ will be pushed rightwards and clog exit at point $`d`$. So the key block $`B`$ must be pushed down to point $`b`$. The only way to reach $`B`$ is via the locking door above it (because block $`C`$ prevents access from the left). But passing through the locking door leaves the robot no alternative but to push $`C`$ leftwards, as in in Fig. 11(c). This blocks access to the $`B`$-lock.
The initial position of block $`E`$ ensures that the unit cannot be traversed from $`y`$ to $`x`$ upon first encounter. That it cannot be traversed in either direction after forward passage is clear by inspection of Fig. 11(b). $`\mathrm{}`$
We will use the symbol shown in Fig. 10(c) to represent a double lock unit prior to passage,<sup>3</sup><sup>3</sup>3 The circular “bite” taken out of the arrow is to remind us that passage requires an external key. and again the ‘X’ in (b) to represent the unit after passage.
### 7.4 Unidirectional Crossover Gadget
We are finally prepared to design a gadget that removes the exclusive-or limitation of the XOR crossover gadget. We call this a Unidirectional Crossover gadget; see Fig. 12(a).
Call the directed horizontal path $`(x_1,x_2)`$ the forward $`x`$-path, and the directed vertical path $`(y_1,y_2)`$ the forward $`y`$-path. The backward paths are the reverse of these. The point where the paths meet we call the junction of the gadget. The forward $`x`$-path consists of a locking door followed by a key block $`X`$ prior to the junction and a double lock after the junction; the $`y`$-path is structured similarly.
We denote the unidirectional crossover gadget by the symbol in Fig. 10(d).
###### Lemma 8
A unidirectional crossover gadget may be traversed along the forward $`x`$-path (but not the backward $`x`$-path), and along the forward $`y`$-path (but not the backward $`y`$-path), in either order. Once either path is traversed, that path (but not the other) becomes impassable in either direction. If the junction is approached from $`x_1`$, then the robot may not from there reach either $`y_1`$ or $`y_2`$ without first passing through the $`x`$-path to $`x_2`$; and symmetrical claims hold after approaching the junction from $`y_1`$.
Proof: The design is symmetrical, so we need only establish its properties for the $`x`$-path. If the robot is at $`x_1`$, it can only enter the unit by passing through the locking door, which then is permanently sealed behind it. The robot must then push the $`X`$ block into the double lock unit on the $`x`$-path. $`X`$ serves as the external key for that unit, and permits passage as in Lemma 7. Thus passage along the forward $`x`$-path is possible. The reverse passage is not possible, neither initially (because a double lock cannot be traversed backwards), nor after forward passage (because the double lock then becomes closed).
Unlike the XOR crossover, passage along the $`x`$-path does not affect the ability to later traverse the forward $`y`$-path. Finally, note that when the robot reaches the junction starting from $`x_1`$, it may not “leak” into the $`y`$-path: its upward movement is stopped by the locking door, and its downward movement is prevented by the inability to pass through the double lock without the $`Y`$-key—the robot is on the “wrong side” of the key. Similarly, if the robot first traverses the $`x`$-path, and then reaches the junction via the $`y`$-path, then it may not “leak” into the $`x`$-path, as Fig. 12(b) illustrates. $`\mathrm{}`$
### 7.5 Bidirectional Crossover Gadget
We can finally construct a VC-crossover, which we call a Bidirectional Crossover Gadget; see Fig. 13.
Again call the directed horizontal path $`(x_1,x_2)`$ the forward $`x`$-path, and similarly for the forward $`y`$-path and the backward paths.
###### Lemma 9
A bidirectional crossover may be traversed as many as four times with any combination, in any order, of the forward $`x`$-path, the backwards $`x`$-path, the forward $`y`$-path, and the backwards $`y`$-path. Furthermore, no leakage between an x-path and y-path is possible.
Proof: The claim follows immediately from the construction and the properties of the unidirectional crossover gadget established in Lemma 8. For example, the forward $`x`$-path is traversed by passing through the bottom two unidirectional gadgets. Later the reverse $`y`$-path can be traversed by passing through the left two unidirectional gadgets. This still leaves it possible to later traverse the forward $`y`$-path and the reverse $`x`$-path. $`\mathrm{}`$
### 7.6 2D Construction
Replacing each FT-crossover by an XOR crossover gadget (Fig. 8), and each VC-crossover by a bidirectional crossover gadget (Fig. 13) completes the 2D construction. (Note we could replace the FT-crossovers by bidirectional crossover units, although the extra complexity of this gadget is not needed here.) The entire construction may be traversed from $`s`$ to $`t`$ iff the represented Boolean formula is satisfiable. We have proved:
###### Theorem 2
PushPush is NP-hard in 2D.
We leave it an open question whether this theorem can be strengthened in either direction: either by proving PushPush is in NP, in which case it is NP-complete, or by showning that PushPush is PSPACE-complete.
## 8 Summary
We conclude by summarizing in Table 1 previous work according to the classification scheme offered in Section 2. The first four lines show previous results. The next two are the results from \[OS99\]. (The 2D storage result is, incidentally, not difficult.) The boldface line of the table is the result of this paper.
The last two lines list open problems. The penultimate line reposes the open question from \[DO92\]: Is the problem where all blocks are movable and the robot can push $`k`$ blocks, sliding the minimal amount, intractable in 2D? The last line presents a new open problem, which we call Push-1: Is the problem where all blocks are movable, the robot can only push $`1`$ block at time, pushing it the minimal amount, intractable in 2D? This differs from the problem addressed in this paper only in altering the sliding from max to min. Both of these open problems remain candidates for being solvable in polynomial time, the former because it seems difficult to construct gadgets when the robot can destroy them, the latter because without sliding the maximal extent, it seems difficult to design gadgets with the requisite functionality.
#### Acknowledgements.
We thank Therese Biedl for helpful discussions. The third author acknowledges many insights from meetings of the Smith Problem Solving Group.<sup>4</sup><sup>4</sup>4 Beenish Chaudry, Sorina Chircu, Elizabeth Churchill, Alexandra Fedorova, Judy Franklin, Biliana Kaneva, Haley Miller, Anton Okmianski, Irena Pashchenko, Ileana Streinu, Geetika Tewari, Dominique Thiébaut, Elif Tosun. |
warning/0001/cond-mat0001264.html | ar5iv | text | # 1 Introduction
## 1 Introduction
Consider a ferromagnetic model, without quenched randomness, evolving from a disordered initial state, according to some dynamics at fixed temperature $`T`$. In the high-temperature paramagnetic phase $`(T>T_c)`$, the system relaxes exponentially to equilibrium. At equilibrium, two-time quantities such as the autocorrelation function $`C(t,s)`$ or the response function $`R(t,s)`$ only depend on the time difference $`\tau =ts`$, where $`s`$ (waiting time) is smaller than $`t`$ (observation time), and both quantities are simply related to each other by the fluctuation-dissipation theorem
$$R_{\mathrm{eq}}(\tau )=\frac{1}{T}\frac{\mathrm{d}C_{\mathrm{eq}}(\tau )}{\mathrm{d}\tau }.$$
(1.1)
In the low-temperature phase $`(T<T_c)`$ the system undergoes phase ordering. In this non-equilibrium situation, $`C(t,s)`$ and $`R(t,s)`$ are non-trivial functions of both time variables, which only depend on their ratio at late times, i.e., in the self-similar domain growth (or coarsening) regime . This behavior is usually referred to as aging . Moreover, no such simple relation as eq. (1.1) holds between correlation and response, i.e., $`R(t,s)`$ and $`C(t,s)/s`$ are no longer proportional. It is then natural to characterize the distance to equilibrium of an aging system by the so-called fluctuation-dissipation ratio
$$X(t,s)=\frac{TR(t,s)}{{\displaystyle \frac{C(t,s)}{s}}}.$$
(1.2)
In recent years, several works have been devoted to the study of the fluctuation-dissipation ratio for systems exhibiting domain growth, or for aging systems such as glasses and spin glasses, showing that in the low-temperature phase $`X(t,s)`$ turns out to be a non-trivial function of its two arguments. In particular, for domain-growth systems, analytical and numerical studies indicate that the limit fluctuation-dissipation ratio,
$$X_{\mathrm{}}=\underset{s\mathrm{}}{lim}\underset{t\mathrm{}}{lim}X(t,s),$$
(1.3)
vanishes throughout the low-temperature phase .
However, to date, only very little attention has been devoted to the response function $`R(t,s)`$, and fluctuation-dissipation ratio $`X(t,s)`$, for non-equilibrium systems at criticality. From now on, we will only have in mind ferromagnetic systems without quenched randomness. For instance one may wonder whether there exists, for a given model, a well-defined limit $`X_{\mathrm{}}`$ at $`T=T_c`$, different from its trivial value $`X_{\mathrm{}}=0`$ in the low-temperature phase, and to what extent $`X_{\mathrm{}}`$ is universal. Indeed a priori, for a system such as a ferromagnet, quenched from infinitely high temperature to its critical point, the limit fluctuation-dissipation ratio $`X_{\mathrm{}}`$ at $`T=T_c`$ (if it exists) may take any value between $`X_{\mathrm{}}=1`$ ($`T>T_c`$: equilibrium) and $`X_{\mathrm{}}=0`$ ($`T<T_c`$: domain growth).
The only cases of critical systems for which the fluctuation-dissipation ratio has been considered are, to our knowledge, the models of ref. (random walk, free Gaussian field, and two-dimensional X-Y model at zero temperature) which share the limit fluctuation-dissipation ratio $`X_{\mathrm{}}=1/2`$, and the backgammon model, a mean-field model for which $`T_c=0`$, where it has been shown that $`X_{\mathrm{}}=1`$, up to a large logarithmic correction, for both energy fluctuations and density fluctuations .
In a recent companion paper , we have determined the fluctuation-dissipation ratio $`X(t,s)`$ for the Glauber-Ising chain, another model for which $`T_c=0`$. In particular, the limit fluctuation-dissipation ratio was found to be $`X_{\mathrm{}}=1/2`$ (see also ref. ). In the present work we investigate the non-equilibrium correlation and response functions and the associated fluctuation-dissipation ratio in generic ferromagnetic models at their critical point. We first present (in section 2) an analytical study of the spherical model in arbitrary dimension. We then turn (in section 3) to a scaling analysis of the generic case, and to numerical simulations on the two-dimensional Ising model. One salient outcome of these joint works is the realization that the limit fluctuation-dissipation ratio $`X_{\mathrm{}}`$ is a novel universal characteristic of critical dynamics, intrinsically related to non-equilibrium initial situations.
The present paper is written in a self-contained fashion. For the spherical model, though our main intention lies in the study of non-equilibrium dynamics at the critical point, we shall present the three situations $`T>T_c`$, $`T=T_c`$, and $`T<T_c`$ in parallel. The latter case has already been the subject of a number of investigations, for both correlation and response . Results on the scaling behavior of the two-time autocorrelation function at $`T_c`$ can be found in ref. . For the two-dimensional Ising model, several numerical works have already been devoted to its non-equilibrium dynamics in the low-temperature phase, concerning both correlations and response . We will therefore restrict our numerical study to the dynamics at the critical point.
## 2 The spherical model
### 2.1 Langevin dynamics
The ferromagnetic spherical model was introduced by Berlin and Kac , as an attempt to simplify the Ising model. It is solvable in any dimension, yet possesses non-trivial critical properties . Consider a lattice of points of arbitrary dimension $`D`$, chosen to be hypercubic for simplicity, with unit lattice spacing. The spins $`S_𝐱`$, situated at the lattice vertices $`𝐱`$, are real variables subject to the constraint
$$\underset{𝐱}{}S_𝐱^2=N,$$
(2.1)
where $`N`$ is the number of spins in the system. The Hamiltonian of the model reads
$$=\underset{(𝐱,𝐲)}{}S_𝐱S_𝐲,$$
(2.2)
where the sum runs over pairs of neighboring sites.
Throughout the following, we assume that the system is homogeneous, i.e., invariant under spatial translations. This holds for a finite sample with periodic boundary conditions, and (at least formally) for the infinite lattice. We also assume that the initial state of the system at $`t=0`$ is the infinite-temperature equilibrium state. This state is fully disordered, in the sense that spins are uncorrelated. The dynamics of the system is given by the stochastic differential Langevin equation
$$\frac{\mathrm{d}S_𝐱}{\mathrm{d}t}=\underset{𝐲(𝐱)}{}S_𝐲\lambda (t)S_𝐱+\eta _𝐱(t).$$
(2.3)
The first term, where $`𝐲(𝐱)`$ denotes the $`2D`$ first neighbors of the site $`𝐱`$, is equal to the gradient $`/S_𝐱`$, while $`\lambda (t)`$ is a Lagrange multiplier ensuring the constraint (2.1), which we choose to parameterize as
$$\lambda (t)=2D+z(t),$$
(2.4)
and $`\eta _𝐱(t)`$ is a Gaussian white noise with correlation
$$\eta _𝐱(t)\eta _𝐲(t^{})=2T\delta _{𝐱,𝐲}\delta (tt^{}).$$
(2.5)
Equation (2.3) can be solved in Fourier space. Defining the spatial Fourier transform by the formulas
$$f^\mathrm{F}(𝐪)=\underset{𝐱}{}f_𝐱\mathrm{e}^{\mathrm{i}𝐪.𝐱},f_𝐱=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}f^\mathrm{F}(𝐪)\mathrm{e}^{\mathrm{i}𝐪.𝐱},$$
(2.6)
where
$$\frac{\mathrm{d}^D𝐪}{(2\pi )^D}=_\pi ^\pi \frac{\mathrm{d}q_1}{2\pi }\mathrm{}_\pi ^\pi \frac{\mathrm{d}q_D}{2\pi }$$
(2.7)
is the normalized integral over the first Brillouin zone, we obtain
$$\frac{S^\mathrm{F}(𝐪,t)}{t}=[\omega (𝐪)+z(t)]S^\mathrm{F}(𝐪,t)+\eta ^\mathrm{F}(𝐪,t),$$
(2.8)
where
$$\omega (𝐪)=2\underset{a=1}{\overset{D}{}}(1\mathrm{cos}q_a)\underset{𝐪\mathrm{𝟎}}{}𝐪^2,$$
(2.9)
and
$$\eta ^\mathrm{F}(𝐪,t)\eta ^\mathrm{F}(𝐪^{},t^{})=2T(2\pi )^D\delta ^D(𝐪+𝐪^{})\delta (tt^{}).$$
(2.10)
The solution to eq. (2.8) reads
$$S^\mathrm{F}(𝐪,t)=\mathrm{e}^{\omega (𝐪)tZ(t)}\left(S^\mathrm{F}(𝐪,t=0)+_0^t\mathrm{e}^{\omega (𝐪)t_1+Z(t_1)}\eta ^\mathrm{F}(𝐪,t_1)dt_1\right),$$
(2.11)
with
$$Z(t)=_0^tz(t_1)dt_1.$$
(2.12)
### 2.2 Equal-time correlation function
Our first goal is to compute the equal-time correlation function
$$C_{𝐱𝐲}(t)=S_𝐱(t)S_𝐲(t),$$
(2.13)
which is a function of the separation $`𝐱𝐲`$, by translational invariance. We have in particular
$$C_\mathrm{𝟎}(t)=S_𝐱(t)^2=1,$$
(2.14)
because of the spherical constraint (2.1), and
$$C_𝐱(t=0)=\delta _{𝐱,\mathrm{𝟎}},$$
(2.15)
reflecting the absence of correlations in the initial state. In eq. (2.13), the brackets denote the average over the ensemble of infinite-temperature initial configurations and over the thermal histories (realizations of the noise).
In Fourier space the equal-time correlation function is defined by
$$S^\mathrm{F}(𝐪,t)S^\mathrm{F}(𝐪^{},t)=(2\pi )^D\delta ^D(𝐪+𝐪^{})C^\mathrm{F}(𝐪,t).$$
(2.16)
Using the expression (2.11), averaging it over the white noise $`\eta ^\mathrm{F}(𝐪,t)`$ with variance given by eq. (2.10), and imposing the condition
$$C^\mathrm{F}(𝐪,t=0)=1$$
(2.17)
implied by eq. (2.15), we obtain
$$C^\mathrm{F}(𝐪,t)=\mathrm{e}^{2\omega (𝐪)t2Z(t)}\left(1+2T_0^t\mathrm{e}^{2\omega (𝐪)t_1+2Z(t_1)}dt_1\right).$$
(2.18)
At this point, we are naturally led to introduce two functions, $`f(t)`$ and $`g(T,t)`$, which play a central role in the following developments.
The function $`f(t)`$ is explicitly given by
$$f(t)=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}\mathrm{e}^{2\omega (𝐪)t}=\left(\mathrm{e}^{4t}I_0(4t)\right)^D\underset{t\mathrm{}}{}(8\pi t)^{D/2},$$
(2.19)
where
$$I_0(z)=\frac{\mathrm{d}q}{2\pi }\mathrm{e}^{z\mathrm{cos}q}\underset{z\mathrm{}}{}(2\pi z)^{1/2}\mathrm{e}^z$$
(2.20)
is the modified Bessel function.
The function
$$g(T,t)=\mathrm{e}^{2Z(t)}$$
(2.21)
is related to $`f(t)`$ by the constraint (2.14), namely
$$\frac{\mathrm{d}^D𝐪}{(2\pi )^D}C^\mathrm{F}(𝐪,t)=\frac{1}{g(T,t)}\left(f(t)+2T_0^tf(tt_1)g(T,t_1)dt_1\right)=1,$$
(2.22)
which yields a linear Volterra integral equation for $`g(T,t)`$ , namely
$$g(T,t)=f(t)+2T_0^tf(tt_1)g(T,t_1)dt_1.$$
(2.23)
This equation can be solved using temporal Laplace transforms, denoted by
$$f^\mathrm{L}(p)=_0^{\mathrm{}}f(t)\mathrm{e}^{pt}dt.$$
(2.24)
We obtain
$$g^\mathrm{L}(T,p)=\frac{f^\mathrm{L}(p)}{12Tf^\mathrm{L}(p)},$$
(2.25)
with
$$f^\mathrm{L}(p)=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}\frac{1}{p+2\omega (𝐪)}.$$
(2.26)
The dependence of $`g^\mathrm{L}(T,p)`$ on temperature appears explicitly in eq. (2.25).
We now present an analysis of the long-time behavior of the function $`g(T,t)`$, considering successively the paramagnetic phase ($`T>T_c`$), the ferromagnetic phase ($`T<T_c`$), and the critical point ($`T=T_c`$). To do so, we shall extensively utilize eq. (2.25). We therefore investigate first the function $`f^\mathrm{L}(p)`$, as given in eq. (2.26). This function has no closed-form expression, except in one and two dimensions:
$`D=1:`$ $`f^\mathrm{L}(p)={\displaystyle \frac{1}{\sqrt{p(p+8)}}},`$ (2.27)
$`D=2:`$ $`f^\mathrm{L}(p)={\displaystyle \frac{2}{\pi |p+8|}}𝐊\left({\displaystyle \frac{8}{|p+8|}}\right),`$ (2.28)
where $`𝐊`$ is the complete elliptic integral. Together with the definition (2.9), eq. (2.26) implies that $`f^\mathrm{L}(p)`$ is analytic in the complex $`p`$-plane cut along the real interval $`[8D,0]`$. The behavior of $`f^\mathrm{L}(p)`$ in the vicinity of the branch point at $`p=0`$ can be analyzed heuristically as follows. The asymptotic behavior of $`f(t)`$ given in eq. (2.19) suggests that its Laplace transform has a universal singular part:
$$f_{\mathrm{sg}}^\mathrm{L}(p)\underset{p0}{}(8\pi )^{D/2}\mathrm{\Gamma }(1D/2)p^{D/21},$$
(2.29)
while there also exists a regular part of the form
$$f_{\mathrm{reg}}^\mathrm{L}(p)=A_1A_2p+A_3p^2+\mathrm{},$$
(2.30)
where
$$A_k=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}\frac{1}{(2\omega (𝐪))^k}$$
(2.31)
are non-universal (lattice-dependent) numbers, given in terms of integrals which are convergent for $`D2k>0`$. For instance $`A_1`$ only exists for $`D>2`$, and so on. Equations (2.29) and (2.30) jointly determine the small-$`p`$ behavior of $`f^\mathrm{L}(p)`$, as a function of the dimensionality $`D`$:
$`D<2:`$ $`f^\mathrm{L}(p)(8\pi )^{D/2}\mathrm{\Gamma }(1D/2)p^{(1D/2)},`$ (2.32)
$`2<D<4:`$ $`f^\mathrm{L}(p)A_1(8\pi )^{D/2}|\mathrm{\Gamma }(1D/2)|p^{D/21},`$ (2.33)
$`D>4:`$ $`f^\mathrm{L}(p)A_1A_2p.`$ (2.34)
These expressions can be justified by more systematic studies (see e.g. ref. ): $`f^\mathrm{L}(p)`$ possesses an asymptotic expansion involving only powers of the form $`p^n`$ and $`p^{D/21+n}`$, for $`n=0,1,\mathrm{}`$ Whenever $`D=2,4,\mathrm{}`$ is an even integer, the two sequences of exponents merge, giving rise to logarithmic corrections, which shall be discarded throughout the following.
In low enough dimension ($`D<2`$), $`f^\mathrm{L}(p)`$ diverges as $`p0`$. As a consequence, for any finite temperature, $`g^\mathrm{L}(T,p)`$ has a pole at some positive value of $`p`$, denoted by $`1/\tau _{\mathrm{eq}}`$, away from the cut of $`f^\mathrm{L}(p)`$. Hence
$$g(T,t)\underset{t\mathrm{}}{}\mathrm{e}^{t/\tau _{\mathrm{eq}}},$$
(2.35)
and therefore, as further analyzed below, the system relaxes exponentially fast to equilibrium, with a finite relaxation time $`\tau _{\mathrm{eq}}`$. The latter diverges as the zero-temperature phase transition is approached, as
$$\tau _{\mathrm{eq}}\underset{T0}{}\left(2(8\pi )^{D/2}\mathrm{\Gamma }(1D/2)T\right)^{2/(2D)}.$$
(2.36)
In high enough dimension ($`D>2`$), $`f^\mathrm{L}(p=0)=A_1`$ is finite, so that the pole of $`g^\mathrm{L}(T,p)`$ hits the cut of $`f^\mathrm{L}(p)`$ at $`p=0`$ at a finite critical temperature
$$T_c=\frac{1}{2A_1}=\left(\frac{\mathrm{d}^D𝐪}{(2\pi )^D}\frac{1}{\omega (𝐪)}\right)^1.$$
(2.37)
As $`TT_c^+`$, the relaxation time $`\tau _{\mathrm{eq}}`$ diverges according to
$`2<D<4:`$ $`\tau _{\mathrm{eq}}\underset{TT_c^+}{}\left({\displaystyle \frac{2(8\pi )^{D/2}|\mathrm{\Gamma }(1D/2)|T_c^2}{TT_c}}\right)^{2/(D2)},`$ (2.38)
$`D>4:`$ $`\tau _{\mathrm{eq}}\underset{TT_c^+}{}{\displaystyle \frac{2A_2T_c^2}{TT_c}}.`$ (2.40)
Note that these equations can be recast into the form $`\tau _{\mathrm{eq}}(TT_c)^{\nu z_c}`$, where $`\nu `$ is the critical exponent of the correlation length, equal to $`1/(D2)`$ for $`2<D<4`$ and to $`1/2`$ for $`D>4`$ , while $`z_c`$ is the dynamic critical exponent, equal to 2 in the present case.<sup>1</sup><sup>1</sup>1A summary of the values of static and dynamical exponents appearing in this work is given in Table 1.
We now discuss the asymptotic behavior of the function $`g(T,t)`$ according to temperature. Throughout the following, we will assume that $`D>2`$, so that the model has a ferromagnetic transition at a finite $`T_c`$, given by eq. (2.37).
* In the paramagnetic phase $`(T>T_c)`$, $`g(T,t)`$ still grows exponentially, according to eq. (2.35).
* In the ferromagnetic phase $`(T<T_c)`$, a careful analysis of eq. (2.25) yields
$$g(T,t)\underset{t\mathrm{}}{}\frac{f(t)}{M_{\mathrm{eq}}^4}\frac{(8\pi t)^{D/2}}{M_{\mathrm{eq}}^4},$$
(2.41)
where the spontaneous magnetization $`M_{\mathrm{eq}}`$ is given by
$$M_{\mathrm{eq}}^2=1\frac{T}{T_c}.$$
(2.42)
* At the critical point $`(T=T_c)`$, we obtain
$`2<D<4:`$ $`g(T_c,t)\underset{t\mathrm{}}{}(D2)(8\pi )^{D/21}\mathrm{sin}[(D2)\pi /2]{\displaystyle \frac{t^{(2D/2)}}{T_c^2}},`$ (2.43)
$`D>4:`$ $`g(T_c,t)\underset{t\mathrm{}}{}{\displaystyle \frac{1}{4A_2T_c^2}}.`$ (2.44)
Finally, eqs. (2.25), (2.26), (2.34), and (2.37) yield the following identities:
$`{\displaystyle _0^{\mathrm{}}}f(t)dt`$ $`=`$ $`{\displaystyle \frac{1}{2T_c}},`$ (2.45)
$`{\displaystyle _0^{\mathrm{}}}f(t)\mathrm{e}^{t/\tau _{\mathrm{eq}}}dt`$ $`=`$ $`{\displaystyle \frac{1}{2T}}(T>T_c),`$ (2.47)
$`{\displaystyle _0^{\mathrm{}}}g(T,t)dt`$ $`=`$ $`{\displaystyle \frac{1}{2T_cM_{\mathrm{eq}}^2}}(T<T_c).`$ (2.49)
We are now in a position to discuss the temporal behavior of the equal-time correlation function in the different phases. Its expression (2.18) in Fourier space reads
$$C^\mathrm{F}(𝐪,t)=\frac{\mathrm{e}^{2\omega (𝐪)t}}{g(T,t)}\left(1+2T_0^t\mathrm{e}^{2\omega (𝐪)t_1}g(T,t_1)dt_1\right),$$
(2.50)
using the definition (2.21) of $`g(T,t)`$. We shall consider in particular the dynamical susceptibility
$$\chi (t)=\frac{1}{T}\underset{𝐱}{}S_\mathrm{𝟎}(t)S_𝐱(t)=\frac{C^\mathrm{F}(𝐪=\mathrm{𝟎},t)}{T},$$
(2.51)
for which eq. (2.50) yields
$$\chi (t)=\frac{1}{g(T,t)}\left(\frac{1}{T}+2_0^tg(T,t_1)dt_1\right).$$
(2.52)
The asymptotic expressions (2.35), (2.41), and (2.44) of $`g(T,t)`$ lead to the following predictions.
* In the paramagnetic phase $`(T>T_c)`$, the correlation function converges exponentially fast to its equilibrium value, which has the Ornstein-Zernike form
$$C_{\mathrm{eq}}^\mathrm{F}(𝐪)=\frac{T}{\omega (𝐪)+\xi _{\mathrm{eq}}^2},$$
(2.53)
where the equilibrium correlation length $`\xi _{\mathrm{eq}}`$ is given by
$$\xi _{\mathrm{eq}}^2=2\tau _{\mathrm{eq}}.$$
(2.54)
The corresponding value of the equilibrium susceptibility is $`\chi _{\mathrm{eq}}=\xi _{\mathrm{eq}}^2=2\tau _{\mathrm{eq}}`$. Eq. (2.53) implies an exponential and isotropic fall-off of correlations, of the form $`C_{𝐱,\mathrm{eq}}\mathrm{e}^{|𝐱|/\xi _{\mathrm{eq}}}`$, at large distances and for $`\xi _{\mathrm{eq}}`$ large, i.e., $`T`$ close enough to $`T_c`$.
* In the ferromagnetic phase $`(T<T_c)`$, using the third of the identities (2.49), we obtain a scaling form for the correlation function, namely
$$C^\mathrm{F}(𝐪,t)M_{\mathrm{eq}}^2(8\pi t)^{D/2}\mathrm{e}^{2𝐪^2t},$$
(2.55)
or equivalently,
$$C_𝐱(t)M_{\mathrm{eq}}^2\mathrm{e}^{𝐱^2/(8t)},$$
(2.56)
in the regime where $`𝐱`$ is large (i.e., $`𝐪`$ is small) and $`t`$ is large. Both the Gaussian profile of the correlation function, and its scaling law involving one single diverging length scale
$$L(t)t^{1/2},$$
(2.57)
reflect the diffusive nature of the coarsening process. The growing length $`L(t)`$ can be interpreted as the characteristic size of an ordered domain. The dynamical susceptibility,
$$\chi (t)\frac{M_{\mathrm{eq}}^2}{T}(8\pi t)^{D/2},$$
(2.58)
grows as $`\chi (t)L(t)^D`$, or else as the volume explored by a diffusive process.
* At the critical point $`(T=T_c)`$, the equilibrium correlation function reads
$$C_{\mathrm{eq}}^\mathrm{F}(𝐪)\frac{T_c}{𝐪^2},$$
(2.59)
i.e.,
$$C_{𝐱,\mathrm{eq}}\frac{\mathrm{\Gamma }(D/21)}{4\pi ^{D/2}}\frac{T_c}{|𝐱|^{D2}}.$$
(2.60)
These limiting expressions are reached according to scaling laws of the form
$`C^\mathrm{F}(𝐪,t)`$ $``$ $`C_{\mathrm{eq}}^\mathrm{F}(𝐪)\mathrm{\Phi }(𝐪^2t),`$ (2.61)
$`C_𝐱(t)`$ $``$ $`C_{𝐱,\mathrm{eq}}\mathrm{\Psi }(𝐱^2/t),`$ (2.62)
with
$`2<D<4:`$ $`\mathrm{\Phi }(x)=2x{\displaystyle _0^1}\mathrm{e}^{2x(1z)}z^{D/22}dz,`$ (2.64)
$`\mathrm{\Psi }(y)=\mathrm{e}^{y/8},`$
$`D>4:`$ $`\mathrm{\Phi }(x)=1\mathrm{e}^{2x},`$ (2.67)
$`\mathrm{\Psi }(y)={\displaystyle \frac{1}{\mathrm{\Gamma }(D/21)}}{\displaystyle _{y/8}^{\mathrm{}}}\mathrm{e}^zz^{D/22}dz.`$
The second expression of eq. (2.62) has the general scaling form for the equal-time correlation function (see eq. (3.9)), with the known value of the static exponent of correlations $`\eta =0`$ for the spherical model , and with $`z_c=2`$, already found above.
The dynamical susceptibility grows linearly with time, as $`\chi (t)\mathrm{\Phi }^{}(0)t`$, i.e.,
$`2<D<4:`$ $`\chi (t){\displaystyle \frac{4}{D2}}t,`$ (2.68)
$`D>4:`$ $`\chi (t)2t.`$ (2.70)
### 2.3 Two-time correlation function
We now consider the two-time correlation function
$$C_{𝐱𝐲}(t,s)=S_𝐱(t)S_𝐲(s),$$
(2.71)
with $`0s`$ (waiting time) $`t`$ (observation time). Its Fourier transform $`C^\mathrm{F}(𝐪,t,s)`$ is defined as in eq. (2.16). Using eq. (2.11), we obtain
$$C^\mathrm{F}(𝐪,t,s)=\frac{\mathrm{e}^{\omega (𝐪)(t+s)}}{\sqrt{g(T,t)g(T,s)}}\left(1+2T_0^s\mathrm{e}^{2\omega (𝐪)t_1}g(T,t_1)dt_1\right),$$
(2.72)
or else
$$C^\mathrm{F}(𝐪,t,s)=C^\mathrm{F}(𝐪,s)\mathrm{e}^{\omega (𝐪)(ts)}\sqrt{\frac{g(T,s)}{g(T,t)}},$$
(2.73)
using the expression (2.50) for $`C^\mathrm{F}(𝐪,s)`$.
In the following, we shall be mostly interested in the two-time autocorrelation function
$$C(t,s)C_\mathrm{𝟎}(t,s)=S_𝐱(t)S_𝐱(s)=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}C^\mathrm{F}(𝐪,t,s),$$
(2.74)
for which eq. (2.72) yields
$$C(t,s)=\frac{1}{\sqrt{g(T,t)g(T,s)}}\left[f\left(\frac{t+s}{2}\right)+2T_0^sf\left(\frac{t+s}{2}t_1\right)g(T,t_1)dt_1\right].$$
(2.75)
The autocorrelation with the initial state assumes the simpler form
$$C(t,s=0)=\frac{f(t/2)}{\sqrt{g(T,t)}}.$$
(2.76)
The asymptotic expressions (2.19), (2.35), (2.41), and (2.44) of the functions $`f(t)`$ and $`g(T,t)`$ lead to the following predictions.
* In the paramagnetic phase $`(T>T_c)`$, as $`s\mathrm{}`$ with $`\tau =ts`$ fixed, the system converges to its equilibrium state, where the correlation function only depends on $`\tau `$:
$$C(s+\tau ,s)\underset{s\mathrm{}}{}C_{\mathrm{eq}}(\tau )=T_\tau ^{\mathrm{}}f(\tau _1/2)\mathrm{e}^{\tau _1/(2\tau _{\mathrm{eq}})}d\tau _1.$$
(2.77)
This equilibrium correlation function decreases exponentially to zero as $`\mathrm{e}^{\tau /(2\tau _{\mathrm{eq}})}`$ when $`\tau \mathrm{}`$. The initial value $`C_{\mathrm{eq}}(\tau =0)=1`$ is ensured by the second identity of eq. (2.49).
* In the ferromagnetic phase $`(T<T_c)`$, two regimes need to be considered. In the first regime ($`s\mathrm{}`$ and $`\tau `$ fixed, i.e., $`1\tau s`$), using again the identities (2.49), we obtain
$$C(s+\tau ,s)M_{\mathrm{eq}}^2+(1M_{\mathrm{eq}}^2)C_{\mathrm{eq},\mathrm{c}}(\tau ),$$
(2.78)
where we have set
$$C_{\mathrm{eq},\mathrm{c}}(\tau )=T_c_\tau ^{\mathrm{}}f(\tau _1/2)d\tau _1.$$
(2.79)
This function, which corresponds to the $`TT_c`$ limit of eq. (2.77), decreases only algebraically to zero when $`\tau \mathrm{}`$, as
$$C_{\mathrm{eq},\mathrm{c}}(\tau )\underset{\tau \mathrm{}}{}\frac{2(4\pi )^{D/2}}{D2}T_c\tau ^{(D/21)},$$
(2.80)
as implied by eq. (2.19). The first identity of eq. (2.49) ensures that $`C_{\mathrm{eq},\mathrm{c}}(\tau =0)=1`$.
In the second regime, where $`s`$ and $`t`$ are simultaneously large (i.e., $`1s\tau `$), with arbitrary ratio
$$x=\frac{t}{s}=1+\frac{\tau }{s}1,$$
(2.81)
the correlation function obeys a scaling law of the form
$$C(t,s)M_{\mathrm{eq}}^2\left(\frac{4ts}{(t+s)^2}\right)^{D/4}M_{\mathrm{eq}}^2\left(\frac{4x}{(x+1)^2}\right)^{D/4}.$$
(2.82)
When $`x1`$, this expression behaves as
$$C(t,s)AM_{\mathrm{eq}}^2x^{\lambda /2},$$
(2.83)
which can be recast into
$$C(t,s)M_{\mathrm{eq}}^2\left(\frac{L(t)}{L(s)}\right)^\lambda ,$$
(2.84)
where $`L(t)`$ is the length scale defined in eq. (2.57), and $`\lambda `$ is the autocorrelation exponent (see section 3), which is equal to $`D/2`$ in the present case, in agreement with the result found in the $`n\mathrm{}`$ limit of the $`O(n)`$ model (see ref. , p. 386, and references therein).
Between these two regimes, the correlation function takes a plateau value
$$q_{\mathrm{EA}}=\underset{\tau \mathrm{}}{lim}\underset{s\mathrm{}}{lim}C(s+\tau ,s)=M_{\mathrm{eq}}^2=1\frac{T}{T_c},$$
(2.85)
known as the Edwards-Anderson order parameter (see e.g. ref. ).
Hereafter we shall refer to the first regime ($`1\tau s`$) as the stationary regime, and to the second one ($`1s\tau `$) as the scaling (or aging) regime. In the former, the system becomes stationary, though without reaching thermal equilibrium, because the system is coarsening. In the latter regime, as said above, the system is aging. It is possible to match these two kinds of behavior, corresponding respectively to eq. (2.78) and (2.82), into a single expression:
$$C(t=s+\tau ,s)(1M_{\mathrm{eq}}^2)C_{\mathrm{eq},\mathrm{c}}(\tau )+M_{\mathrm{eq}}^2\left(\frac{4ts}{(t+s)^2}\right)^{D/4},$$
(2.86)
which is the sum of a term corresponding to the stationary contribution, and a term corresponding to the aging one. Let us finally recall that, in the context of glassy dynamics, in a low-temperature phase, the first regime, where $`C(t,s)>q_{\mathrm{EA}}`$, is usually referred to as the $`\beta `$ regime, while the second one, where $`C(t,s)<q_{\mathrm{EA}}`$, is referred to as the $`\alpha `$ regime .
* At the critical point $`(T=T_c)`$, the same two regimes are to be considered. However their physical interpretation is slightly different, since the order parameter $`M_{\mathrm{eq}}`$ vanishes, and symmetry between the phases is restored.
In the first regime $`(1\tau s)`$, the system again becomes stationary, the autocorrelation function behaving as the $`TT_c`$ limit of eq. (2.77), that is
$$C(s+\tau ,s)\underset{s\mathrm{}}{}C_{\mathrm{eq},\mathrm{c}}(\tau ),$$
(2.87)
which decreases algebraically to zero when $`\tau \mathrm{}`$ (cf. eq. (2.80)). In the second regime ($`1s\tau `$), the correlation function obeys a scaling law of the form
$$C(t,s)T_cs^{(D/21)}F(x),$$
(2.88)
where the scaling function $`F(x)`$ reads
$`2<D<4:`$ $`F(x)={\displaystyle \frac{4(4\pi )^{D/2}}{(D2)(x+1)}}x^{1D/4}(x1)^{1D/2},`$ (2.89)
$`D>4:`$ $`F(x)={\displaystyle \frac{2(4\pi )^{D/2}}{D2}}\left((x1)^{1D/2}(x+1)^{1D/2}\right).`$ (2.91)
In this regime the system is still aging, in the sense that $`C(t,s)`$ bears a dependence in both time variables. However, the scaling of expression (2.88) is different from that found in the low-temperature phase (see eq. (2.82)), which depends on the ratio $`x=t/s`$ only. The presence in eq. (2.88) of an additional $`s`$-dependence through the factor $`s^{(D/21)}`$ can be interpreted as coming from the anomalous dimension of the field $`S_𝐱`$ at $`T_c`$. In the critical region one has indeed $`M_{\mathrm{eq}}(TT_c)^\beta \xi _{\mathrm{eq}}^{\beta /\nu }`$. Replacing $`\xi _{\mathrm{eq}}`$ by $`s^{1/z_c}`$ implies the replacement of $`M_{\mathrm{eq}}^2`$ by $`s^{2\beta /\nu z_c}s^{(D2+\eta )/z_c}`$. With $`\eta =0`$ and $`z_c=2`$, the factor $`s^{(D/21)}`$ is thus recovered. Note that the static hyperscaling relation $`2\beta /\nu =D2+\eta `$ holds for $`D<4`$, while it is violated for $`D>4`$ (see Table 1).
Two limiting regimes are of interest. First, for $`x1`$, i.e., $`1\tau s`$, eq. (2.88) matches eq. (2.80). Second, for $`x1`$, i.e., $`1st`$, one gets
$$F(x)Bx^{\lambda _c/z_c},$$
(2.92)
where the autocorrelation exponent $`\lambda _c`$ (see section 3) is equal to $`3D/22`$ if $`2<D<4`$, and to $`D`$ above four dimensions, in agreement with the result found in ref. .
We also quote for later reference the scaling law of the derivative
$$\frac{C(t,s)}{s}T_cs^{D/2}F_1(x),$$
(2.93)
with
$$F_1(x)=\frac{D2}{2}F(x)xF^{}(x),$$
(2.94)
i.e.,
$`2<D<4:`$ $`F_1(x)=(4\pi )^{D/2}{\displaystyle \frac{(D2)(x+1)^2+2(x1)^2}{(D2)(x+1)^2}}x^{1D/4}(x1)^{D/2},`$ (2.95)
$`D>4:`$ $`F_1(x)=(4\pi )^{D/2}\left((x1)^{D/2}+(x+1)^{D/2}\right).`$ (2.97)
### 2.4 Two-time response function
Suppose now that the system is subjected to a small magnetic field $`H_𝐱(t)`$, depending on the site $`𝐱`$ and on time $`t0`$ in an arbitrary fashion. This amounts to adding to the ferromagnetic Hamiltonian (2.2) a time-dependent perturbation of the form
$$\delta (t)=\underset{𝐱}{}H_𝐱(t)S_𝐱(t).$$
(2.98)
The dynamics of the model is now given by the modified Langevin equation
$$\frac{\mathrm{d}S_𝐱}{\mathrm{d}t}=\underset{𝐲(𝐱)}{}S_𝐲\lambda (t)S_𝐱+H_𝐱(t)+\eta _𝐱(t).$$
(2.99)
Causality and invariance under spatial translations imply that we have, to first order in the magnetic field $`H_𝐱(t)`$,
$$S_𝐱(t)=_0^tds\underset{𝐲}{}R_{𝐱𝐲}(t,s)H_𝐲(s)+\mathrm{}$$
(2.100)
This formula defines the two-time response function $`R_{𝐱𝐲}(t,s)`$ of the model. A more formal definition reads
$$R_{𝐱𝐲}(t,s)=\frac{\delta S_𝐱(t)}{\delta H_𝐲(s)}|_{\{H_𝐱(t)=0\}}.$$
(2.101)
The solution to eq. (2.99) reads, in Fourier space,
$$S^\mathrm{F}(𝐪,t)=\mathrm{e}^{\omega (𝐪)tZ(t)}\left(S^\mathrm{F}(𝐪,t=0)+_0^t\mathrm{e}^{\omega (𝐪)t_1+Z(t_1)}\left[H^\mathrm{F}(𝐪,t_1)+\eta ^\mathrm{F}(𝐪,t_1)\right]dt_1\right).$$
(2.102)
It can be checked that the Lagrange function $`\lambda (t)`$, and hence $`z(t)`$ and $`Z(t)`$, remain unchanged, to first order in the magnetic field. As a consequence, the two-time response function reads, in Fourier transform,
$$R^\mathrm{F}(𝐪,t,s)=\frac{\delta S^\mathrm{F}(𝐪,t)}{\delta H^\mathrm{F}(𝐪,s)}|_{\{H_𝐱(t)=0\}}=\mathrm{e}^{\omega (𝐪)(ts)}\sqrt{\frac{g(T,s)}{g(T,t)}}$$
(2.103)
(cf. eq. (2.73)). In the following, we shall be mostly interested in the diagonal component of the response function, corresponding to coinciding points:
$$R(t,s)R_\mathrm{𝟎}(t,s)=\frac{\delta S_𝐱(t)}{\delta H_𝐱(s)}|_{\{H_𝐱(t)=0\}}=\frac{\mathrm{d}^D𝐪}{(2\pi )^D}R^\mathrm{F}(𝐪,t,s).$$
(2.104)
With the notations (2.19), (2.21), eq. (2.103) yields
$$R(t,s)=f\left(\frac{ts}{2}\right)\sqrt{\frac{g(T,s)}{g(T,t)}}.$$
(2.105)
The response function at zero waiting time assumes the simpler form (cf. eq. (2.76))
$$R(t,s=0)=C(t,s=0)=\frac{f(t/2)}{\sqrt{g(T,t)}}.$$
(2.106)
The asymptotic expressions (2.19), (2.35), (2.41), and (2.44) of $`F(t)`$ and $`g(T,t)`$ lead to the following predictions.
* In the paramagnetic phase $`(T>T_c)`$, at equilibrium, the response function only depends on $`\tau `$, according to
$$R_{\mathrm{eq}}(\tau )=f(\tau /2)\mathrm{e}^{\tau /(2\tau _{\mathrm{eq}})}.$$
(2.107)
Moreover, it is related to the equilibrium correlation function $`C_{\mathrm{eq}}(\tau )`$ of eq. (2.77) by the fluctuation-dissipation theorem (1.1), as it should.
* In the ferromagnetic phase $`(T<T_c)`$, the two regimes defined in the previous section for the case of the autocorrelation function are still to be considered. In the stationary regime ($`1\tau s`$), the response function behaves as the $`TT_c`$ limit of eq. (2.107), namely
$$R_{\mathrm{eq},\mathrm{c}}(\tau )=f(\tau /2)=\frac{1}{T_c}\frac{\mathrm{d}C_{\mathrm{eq},\mathrm{c}}(\tau )}{\mathrm{d}\tau },$$
(2.108)
so that the fluctuation-dissipation theorem is valid.
On the contrary, in the scaling regime ($`1st`$), the response function has the form
$$R(t,s)(4\pi (ts))^{D/2}(t/s)^{D/4}=(4\pi s)^{D/2}(x1)^{D/2}x^{D/4},$$
(2.109)
which, when compared to the corresponding expression (2.82) for the autocorrelation function, demonstrates the violation of the fluctuation-dissipation theorem (see section 2.5).
* At the critical point $`(T=T_c)`$, in the stationary regime $`(1\tau s)`$, the response function still behaves as in eq. (2.108), so that the fluctuation-dissipation theorem still holds. In the scaling regime ($`1st`$), the response function obeys a scaling law of the form
$$R(t,s)s^{D/2}F_2(x),$$
(2.110)
where the scaling function $`F_2(x)`$ reads
$`2<D<4:`$ $`F_2(x)=(4\pi )^{D/2}x^{1D/4}(x1)^{D/2},`$ (2.111)
$`D>4:`$ $`F_2(x)=(4\pi )^{D/2}(x1)^{D/2}.`$ (2.112)
Again two limiting regimes are of interest. For $`x1`$, the scaling result (2.112) matches eq. (2.108). For $`x1`$, one finds the same power-law fall-off for the functions $`F(x)`$, $`F_1(x)`$, and $`F_2(x)`$, that is
$$F(x)F_1(x)F_2(x)x^{\lambda _c/z_c}$$
(2.113)
(see sections 2.5 and 3).
### 2.5 Fluctuation-dissipation ratio
As already mentioned in the introduction, the violation of the fluctuation-dissipation theorem (1.1) out of thermal equilibrium can be characterized by the fluctuation-dissipation ratio $`X(t,s)`$, defined in eq. (1.2). In the case of the spherical model, the results derived so far yield at once the following predictions.
* In the paramagnetic phase $`(T>T_c)`$, the system converges to an equilibrium state, where the fluctuation-dissipation theorem holds. In other words, the fluctuation-dissipation ratio converges toward its equilibrium value
$$X_{\mathrm{eq}}=1.$$
(2.114)
* In the ferromagnetic phase $`(T<T_c)`$, the fluctuation-dissipation theorem (1.1) is only valid in the stationary regime ($`1\tau s`$). On the contrary, in the scaling regime ($`1s\tau `$), the results (2.82) and (2.109) imply that the fluctuation-dissipation ratio falls off as
$$X(t,s)\frac{(8\pi )^{D/2}}{D}\frac{4T}{M_{\mathrm{eq}}^2}\left(\frac{x+1}{x1}\right)^{D/2+1}s^{(D/21)}.$$
(2.115)
In particular, the limit fluctuation-dissipation ratio introduced in eq. (1.3) reads
$$X_{\mathrm{}}=0.$$
(2.116)
* At the critical point $`(T=T_c)`$, the scaling laws (2.93) and (2.110) imply that the fluctuation-dissipation ratio $`X(t,s)`$ becomes asymptotically a smooth function of the time ratio $`x=t/s`$:
$$X(t,s)\underset{t,s\mathrm{}}{}𝒳(x)=\frac{F_2(x)}{F_1(x)},$$
(2.117)
i.e., explicitly,
$`2<D<4:`$ $`𝒳(x)={\displaystyle \frac{1}{1+{\displaystyle \frac{2}{D2}}\left({\displaystyle \frac{x1}{x+1}}\right)^2}},`$ (2.118)
$`D>4:`$ $`𝒳(x)={\displaystyle \frac{1}{1+\left({\displaystyle \frac{x1}{x+1}}\right)^{D/2}}}.`$ (2.120)
The scaling law (2.117) interpolates between the equilibrium behavior
$$𝒳(x)\underset{x1}{}X_{\mathrm{eq}}=1$$
(2.121)
in the stationary regime of relatively short time differences, and a non-trivial limit value
$$𝒳(x)\underset{x\mathrm{}}{}X_{\mathrm{}}$$
(2.122)
at large time differences, given by
$`2<D<4:`$ $`X_{\mathrm{}}={\displaystyle \frac{D2}{D}},`$ (2.123)
$`D>4:`$ $`X_{\mathrm{}}={\displaystyle \frac{1}{2}}.`$ (2.125)
Further comments on the scaling behavior of the fluctuation-dissipation ratio will be made in section 3.2.
## 3 The generic situation
### 3.1 Aging below $`T_c`$
Let us first briefly sketch the description of the dynamical behavior of a ferromagnetic system quenched from a disordered initial state to a temperature $`T<T_c`$ .
In the scaling regime ($`1st`$), the autocorrelation $`C(t,s)`$ is expected to be a function of the ratio $`L(t)/L(s)`$ only, where the length scale $`L(t)t^{1/z}`$ is the characteristic size of an ordered domain, and $`z`$ is the growth exponent, equal to 2 for non-conserved dynamics. More precisely,
$$C(t,s)=M_{\mathrm{eq}}^2f(t/s),$$
(3.1)
where the scaling function $`f`$ is temperature independent. Furthermore we have, for $`x=t/s1`$, i.e., $`1st`$,
$$f(x)Ax^{\lambda /z},$$
(3.2)
where $`\lambda `$ is the autocorrelation exponent . For the spherical model, eqs. (2.82) and (2.84) match eqs. (3.1) and (3.2), with $`\lambda =D/2`$ and $`z=2`$.
As a consequence, we have
$$\frac{C(t,s)}{s}\frac{M_{\mathrm{eq}}^2}{s}f_1(x),$$
(3.3)
with $`f_1(x)=xf^{}(x)`$, so that, when $`x1`$,
$$f_1(x)A_1x^{\lambda /z},$$
(3.4)
with $`A_1=A\lambda /z`$.
Although the situation of the response $`R(t,s)`$ is less clear-cut, it is however reasonable to make the scaling assumption
$$R(t,s)s^{1a}f_2(x),$$
(3.5)
where $`a>0`$ is an unknown exponent, and again with the behavior
$$f_2(x)A_2x^{\lambda /z}$$
(3.6)
when $`x1`$.
The scaling law (3.5) holds for the spherical model, with $`a=D/21`$, as can be seen from eq. (2.109). Furthermore, for non-conserved dynamics, at least in the case of a discrete broken symmetry, like e.g. in the Ising model, it has been argued that the integrated response $`\rho (t,s)`$ (to be defined in eq. (3.19)), scales as $`\rho (t,s)L(s)^1\phi (L(t)/L(s))`$. This corresponds to eq. (3.5) with $`a=1/z=1/2`$.
The scaling laws (3.3), (3.5) imply
$$X(t,s)s^ag(x),$$
(3.7)
with $`g(x)=(T/M_{\mathrm{eq}}^2)f_2(x)/f_1(x)`$, in agreement with eq. (2.115) for the spherical model, and especially
$$X_{\mathrm{}}=0.$$
(3.8)
### 3.2 Aging at $`T_c`$
Let us now turn to the situation where a ferromagnetic system is quenched from a disordered initial state to its critical point.
In such a circumstance, spatial correlations develop in the system, just as in the critical state, but only over a length scale which grows like $`t^{1/z_c}`$, where $`z_c`$ is the dynamic critical exponent. For example the equal-time correlation function has the scaling form
$$C_𝐱(t)=|𝐱|^{2\beta /\nu }\varphi \left(|𝐱|/t^{1/z_c}\right),$$
(3.9)
where $`\beta `$ and $`\nu `$ are the usual static critical exponents. The scaling function $`\varphi (x)`$ goes to a constant for $`x0`$, while it falls off exponentially to zero for $`x\mathrm{}`$, i.e., on scales smaller than $`t^{1/z_c}`$ the system looks critical, while on larger scales it is disordered. This behavior is illustrated in the case of the spherical model by eq. (2.62), corresponding to $`2\beta /\nu =D2`$ and $`z_c=2`$ in eq. (3.9).
In the scaling region of the two-time plane, where both times $`s`$ and $`t`$ are large and comparable ($`1st`$), with arbitrary ratio $`x=t/s`$, the two-time autocorrelation function $`C(t,s)`$ is expected to obey a scaling law of the form (see the discussion below eq. (2.91), and ref. )
$$C(t,s)s^{2\beta /\nu z_c}F(x).$$
(3.10)
When both time scales are well separated $`(1st`$, i.e., $`x1`$), the scaling function $`F(x)`$ falls off as
$$F(x)Bx^{\lambda _c/z_c},$$
(3.11)
where $`\lambda _c`$ is the critical autocorrelation exponent , related to the (magnetization) initial-slip critical exponent $`\mathrm{\Theta }_c`$ by $`\lambda _c=Dz_c\mathrm{\Theta }_c`$.
We thus have
$$\frac{C(t,s)}{s}s^{12\beta /\nu z_c}F_1(x),$$
(3.12)
with $`F_1(x)=(2\beta /\nu z_c)F(x)xF^{}(x)`$, so that, when $`x1`$,
$$F_1(x)B_1x^{\lambda _c/z_c},$$
(3.13)
with
$$B_1=\frac{\nu \lambda _c2\beta }{\nu z_c}B.$$
(3.14)
For the spherical model, eqs. (2.88), (2.92), and (2.93) respectively match eqs. (3.10), (3.11), and (3.12), with $`\lambda _c=3D/22`$ if $`D<4`$, and $`\lambda _c=D`$ if $`D>4`$ (see Table 1).
The similarity between the results (2.93) and (2.110), obtained in the case of the spherical model, strongly suggests that $`C(t,s)/s`$ and $`R(t,s)`$ behave similarly in the generic case, i.e., one is lead to hypothesize that a scaling law of the form (3.12), with the same power-law fall-off (3.13), holds for the response, that is
$$R(t,s)\frac{1}{T_c}s^{12\beta /\nu z_c}F_2(x),$$
(3.15)
with, when $`x1`$,
$$F_2(x)B_2x^{\lambda _c/z_c}.$$
(3.16)
The scaling laws (3.12) and (3.15) imply then that the fluctuation-dissipation ratio only depends on the time ratio $`x`$ throughout the scaling region:
$$X(t,s)𝒳(x)=\frac{F_2(x)}{F_1(x)},$$
(3.17)
where the scaling function $`𝒳(x)`$ is universal. It appears indeed as a dimensionless combination of scaling functions. In turn, the hypothesis (3.16) implies that the limit fluctuation-dissipation ratio reads
$$X_{\mathrm{}}=𝒳(\mathrm{})=\frac{B_2}{B_1}.$$
(3.18)
This number thus appears as a dimensionless amplitude ratio, in the usual sense of critical phenomena. It is therefore a novel universal quantity of non-equilibrium critical dynamics, as already claimed in ref. .
In the case of the spherical model, the analytical treatment of section 2 corroborates the above analysis, and yields the quantitative predictions (2.120) and (2.125).
In order to perform a numerical evaluation of $`X_{\mathrm{}}`$, one needs to measure the response. A convenient way to do so is to measure instead the dimensionless integrated response function
$$\rho (t,s)=T_0^sR(t,u)du.$$
(3.19)
By eq. (2.100), this quantity is proportional to the thermoremanent magnetization $`M_{\mathrm{TRM}}`$, i.e., the magnetization of the system at time $`t`$ obtained after applying a small magnetic field $`h`$, uniform and constant, between $`t=0`$ and $`t=s`$:
$$M_{\mathrm{TRM}}(t,s)\frac{h}{T}\rho (t,s).$$
(3.20)
The thermoremanent magnetization is a natural quantity to measure experimentally in spin glasses , and it is also accessible to numerical simulations, for systems with and without quenched randomness (see section 3.3).
The scaling law (3.15) for the response function implies
$$\rho (t,s)s^{2\beta /\nu z_c}F_3(x),$$
(3.21)
with $`F_2(x)=(2\beta /\nu z_c)F_3(x)xF_3^{}(x)`$, so that, when $`x1`$,
$$F_3(x)B_3x^{\lambda _c/z_c},$$
(3.22)
with
$$B_3=\frac{\nu z_c}{\nu \lambda _c2\beta }B_2.$$
(3.23)
A clear representation of the evolution of $`X(t,s)`$ in time is provided by the parametric plot of $`\rho (t,s)`$ against $`C(t,s)`$, obtained by varying $`t`$ at fixed $`s`$ . For well-separated times in the scaling regime (i.e., $`1st`$), the common power-law behavior (3.11), (3.13), (3.16), and (3.22) implies that the limit fluctuation-dissipation ratio has the alternative expression
$$X_{\mathrm{}}=\frac{B_3}{B},$$
(3.24)
which is equivalent to eq. (3.18), due to eqs. (3.14) and (3.23). In other words, the relationship (1.2) also holds in integral form, that is
$$\rho (t,s)X_{\mathrm{}}C(t,s),$$
(3.25)
in the regime $`1st`$. The limit fluctuation-dissipation ratio can thus be measured as the slope of the parametric plot in the scaling region, i.e., near the origin of the $`C\rho `$ plane. Eq. (3.25) is expected to hold as long as $`C`$ and $`\rho `$ are much smaller than the crossover scale
$$C^{}(s)=C(2s,s)s^{2\beta /\nu z_c},$$
(3.26)
corresponding to $`\tau =s`$. This quantity provides a measure of the size of the critical region, giving thus a quantitative definition of the critical analogue of $`M_{\mathrm{eq}}^2`$, involved in the discussion below eq. (2.91).
### 3.3 The two-dimensional Ising model: numerical simulations
In order to check the validity of the scaling analysis made in the previous section, beyond the case of the spherical model, we have performed numerical simulations on the ferromagnetic Ising model on the square lattice, evolving under heat-bath (Glauber) dynamics at its critical temperature $`T_c=2/\mathrm{ln}(1+\sqrt{2})2.2692`$, starting from a disordered initial state. The rules of the dynamics are as follows. Consider a finite system, consisting of $`N=L^2`$ spins $`\sigma _𝐱=\pm 1`$ situated at the vertices $`𝐱`$ of a square lattice, with periodic boundary conditions. The Ising Hamiltonian reads
$$=\underset{(𝐱,𝐲)}{}\sigma _𝐱\sigma _𝐲,$$
(3.27)
where the sum runs over pairs of neighboring sites. Heat-bath dynamics consists in updating the spins $`\sigma _𝐱(t)`$ according to the stochastic rule
$$\sigma _𝐱(t)\{\begin{array}{cc}+1& \text{with prob.}\frac{1+\mathrm{tanh}(h_𝐱(t)/T_c)}{2},\\ \\ 1& \text{with prob.}\frac{1\mathrm{tanh}(h_𝐱(t)/T_c)}{2},\end{array}$$
(3.28)
where the local field $`h_𝐱(t)`$ acting on $`\sigma _𝐱(t)`$ reads
$$h_𝐱(t)=\underset{𝐲(𝐱)}{}\sigma _𝐲(t),$$
(3.29)
with $`𝐲(𝐱)`$ denoting the four neighbors of site $`𝐱`$.
Let us give a brief summary of known facts on the dynamics of the Ising model. For $`T<T_c`$, numerical studies have shown that the scaling forms (3.1) and (3.2) hold, with $`z=2`$ (non-conserved dynamics) and $`\lambda 1.25`$ . The integrated response function (in another form, known as the ZFC magnetization) has been measured in ref. . At $`T=T_c`$, the dynamic critical exponent reads $`z_c2.17`$ , and the autocorrelation exponent $`\lambda _c1.59`$ .
Our aim is now to verify the hypotheses made in section 3.2, especially the scaling laws (3.15) and (3.21) for the response function, and to demonstrate the existence of a non-trivial limit $`X_{\mathrm{}}`$.
Computing $`C(t,s)`$ with good statistics is rather easy, while the computation of $`\rho (t,s)`$ requires more effort. We have followed the lines of the method introduced in ref. . In order to isolate the diagonal component of the response function, a quenched, spatially random magnetic field, is applied to the system from $`t=0`$ to $`t=s`$. This magnetic field is of the form $`H_𝐱=h_0\epsilon _𝐱`$, with a constant small amplitude $`h_0`$, and a quenched random modulation, $`\epsilon _𝐱=\pm 1`$ with equal probability, independently at each site $`𝐱`$. The heat-bath dynamical rules are modified by adding up the magnetic field $`H_𝐱`$ to the local field $`h_𝐱(t)`$ of eq. (3.29). We have then
$$\overline{\epsilon _𝐱\sigma _𝐱(t)}=h_0_0^sR(t,u)du=\frac{h_0}{T}\rho (t,s)=M_{\mathrm{TRM}}(t,s),$$
(3.30)
where the bar means an average with respect to the distribution of the modulation $`\epsilon _𝐱`$ of the magnetic field.
Figure 1: Log-log plot of the critical autocorrelation function $`C(t,s)`$ of the two-dimensional Ising model, against time ratio $`x=t/s`$, for several values of the waiting time $`s`$. Data are multiplied by $`s^{2\beta /\nu z_c}`$, in order to demonstrate collapse into the scaling function $`F(x)`$ of eq. (3.10). Straight line: exponent $`\lambda _c/z_c0.73`$ of the fall-off at large $`x`$.
We have first checked the validity of the scaling laws (3.10), and especially (3.21). Figures 1 and 2 respectively show log-log plots of the autocorrelation function $`C(t,s)`$ and of the corresponding integrated response function $`\rho (t,s)`$, against the time ratio $`x=t/s`$, for several values of the waiting time $`s`$. For each value of $`s`$, the simulations are run up to $`t/s=10`$, and data are averaged over at least 500 independent samples of size $`300\times 300`$. For the response function, the amplitude of the quenched magnetic field reads $`h_0=0.05`$. Multiplying the data by $`s^{2\beta /\nu z_c}`$, with $`2\beta /\nu z_c0.115`$, gives good data collapse, thus producing a plot of the scaling functions $`F(x)`$ and $`F_3(x)`$. The data follow a power-law fall-off at large values of $`x`$, with a slope in good agreement with the value $`\lambda _c/z_c0.73`$, shown on the plots as a straight line.
We then turned to an investigation of the parametric plot of these data in the $`C\rho `$ plane. At the qualitative level, this plot, shown in Figure 3 for several values of the waiting time $`s`$, confirms our expectations. The stationary regime $`(1\tau s`$, i.e., roughly speaking, $`C>C^{}(s))`$, corresponds to the right part of the plot. The symbols show the data for small integer values of the time difference, $`\tau =ts=0,\mathrm{},8`$, illustrating the fast decay of correlation and integrated response in the stationary regime. The rightmost points, corresponding to $`\tau =0`$, i.e., $`C=C(s,s)=1`$, are compatible with the scaling law $`1\rho (s,s)C^{}(s)s^{2\beta /\nu z_c}`$. The validity of the fluctuation-dissipation theorem is testified by the unit slope of this part of the plot, shown as a full straight line. The aging regime $`(1st`$, i.e., roughly speaking, $`C<C^{}(s))`$, corresponds to the left part of the plot. As expected, the data crossover toward a non-trivial slope, equal to the limit fluctuation-dissipation ratio $`X_{\mathrm{}}`$. The dashed line shows the slope $`X_{\mathrm{}}=0.26`$, obtained by the analysis described below.
Figure 2: Log-log plot of the critical integrated response function $`\rho (t,s)`$ of the two-dimensional Ising model, against time ratio $`x=t/s`$, for several values of the waiting time $`s`$. Data are multiplied by $`s^{2\beta /\nu z_c}`$, in order to demonstrate collapse into the scaling function $`F_3(x)`$ of eq. (3.21). Straight line: exponent $`\lambda _c/z_c0.73`$ of the fall-off at large $`x`$.
In order to obtain a quantitative prediction of the limit fluctuation-dissipation ratio $`X_{\mathrm{}}`$, we have followed two approaches. Figure 4 depicts the local slope of the plot of Figure 3, i.e., the ratio $`\rho /C`$, against $`C`$, in the aging regime. The data for the largest available waiting time $`s=200`$ have been discarded from the analysis because they appear as too noisy on that scale. The data look pretty linear all over the range presented in the plot. This precocious scaling is due to the fact that the exponent $`2\beta /\nu z_c0.115`$ is small. Hence the size of the critical region, given by the estimate (3.26), is very large, at least for waiting times $`s`$ accessible to computer simulations. We have indeed, for example, $`C^{}(100)=C(200,100)0.24`$. The straight lines show a constrained least-square fit of the three series of data, imposing a common intercept. The value of this intercept yields the prediction $`X_{\mathrm{}}0.262`$.
Figure 3: Parametric plot of the integrated response $`\rho (t,s)`$ against the autocorrelation $`C(t,s)`$, using the data of Figures 1 and 2. Symbols: data for integer time differences $`\tau =ts=0,\mathrm{},8`$. Full line: unit slope corresponding to the fluctuation-dissipation theorem in the stationary regime. Dashed line: limit slope $`X_{\mathrm{}}=0.26`$ (see text and Figures 4 and 5).
We have also followed an alternative approach, aiming at subtracting most of the deviations of the ratio $`\rho /C`$ with respect to its limit $`X_{\mathrm{}}`$ at $`C0`$. This can be done by incorporating the known limit of the stationary regime, i.e., $`\rho 1`$ as $`C1`$, into a quadratic phenomenological formula: $`\rho X_{\mathrm{}}C+(1X_{\mathrm{}})C^2`$. This formula can be rewritten as $`X_{\mathrm{}}(\rho C^2)/(C(1C))`$, suggesting to plot $`(\rho C^2)/(C(1C))`$ against $`C`$, instead of the mere ratio $`\rho /C`$. This has been done in Figure 5. As expected, the vertical scale has been considerably enlarged. In return this procedure increases the statistical noise on the data points. The straight lines again show a constrained least-square fit, yielding $`X_{\mathrm{}}0.260`$.
We can conclude from this numerical analysis that we have
$$X_{\mathrm{}}=0.26\pm 0.01$$
(3.31)
for the ferromagnetic Ising model in two dimensions.
Figure 4: Parametric plot of the ratio $`\rho /C`$ against $`C`$. Straight lines: constrained least-square fit with common intercept, yielding $`X_{\mathrm{}}0.262`$.
## 4 Discussion
In the present work we dealt with the dynamics of ferromagnetic spin systems quenched from infinite temperature to their critical state. This study, exemplified by the exact analysis of the spherical model in any dimension $`D>2`$, and by numerical simulations on the two-dimensional Ising model, complements that of the Glauber-Ising chain, presented in a companion paper . The main results obtained in this work can be summarized as follows.
In such a non-equilibrium situation, these systems are aging in the sense that their correlation and response functions depend non-trivially on the waiting time $`s`$ as well as on the observation time $`t`$, whenever these two times are simultaneously large. The corresponding scaling laws (see eqs. (3.10), (3.12), (3.15), and (3.21)), involve powers of $`s`$, related to the static anomalous dimension of the magnetization, and universal scaling functions of the ratio $`x=t/s`$. In the regime of large time separations, i.e., $`1st`$ (or $`x1`$), these scaling functions fall off algebraically with the common exponent $`\lambda _c/z_c`$.
The fluctuation-dissipation ratio $`X(t,s)`$, characterizing the violation of the fluctuation-dissipation theorem, has a universal scaling form $`𝒳(x)`$, and, for well-separated times in the aging regime, it assumes a limit value $`X_{\mathrm{}}`$ equal to a dimensionless amplitude ratio (see eqs. (3.18) and (3.24)). Therefore, as announced in ref. , $`X_{\mathrm{}}`$ is a novel universal characteristic of critical dynamics, which is intrinsically related to the non-equilibrium initial condition of a critical quench from a disordered state.
Figure 5: Parametric plot of the combination $`(\rho C^2)/(C(1C))`$ against $`C`$. Straight lines: constrained least-square fit yielding $`X_{\mathrm{}}0.260`$.
The ferromagnetic models studied in the present work turn out to have values of $`X_{\mathrm{}}`$ in the range
$$0X_{\mathrm{}}\frac{1}{2}.$$
(4.1)
We have indeed $`X_{\mathrm{}}=12/D`$ if $`2<D<4`$, and $`X_{\mathrm{}}=1/2`$ for $`D>4`$, for the spherical model, and $`X_{\mathrm{}}0.26\pm 0.01`$ for the two-dimensional Ising model. Let us mention that preliminary simulations on the three-dimensional Ising model yield $`X_{\mathrm{}}0.40`$. The backgammon, for which $`X_{\mathrm{}}=1`$ , thus belongs to another class of models.
The mean-field value
$$X_{\mathrm{}}^{\mathrm{MF}}=\frac{1}{2},$$
(4.2)
obtained for the spherical model in dimension $`D>4`$, also holds for a variety of models which are not mean-field-like, including the Glauber-Ising chain and the two-dimensional X-Y model at zero temperature .
Let us finally discuss a few open questions. It would be interesting to know whether there is an analogue for the present case of the results found for models with discontinuous spin-glass transitions, where the violation of the fluctuation-dissipation theorem is related to the configurational entropy . One would also like to know the status of the quantity $`X_{\mathrm{}}`$ for non-equilibrium systems with quenched disorder, or for systems defined by dynamical rules without detailed balance.
In principle the limit fluctuation-dissipation ratio $`X_{\mathrm{}}`$ could be calculated by field-theoretical renormalization-group methods, generalizing the computations done for universal amplitude ratios in usual static critical phenomena , as series in either $`\epsilon =4D`$, or in $`1/n`$ for the $`n`$-component Heisenberg model, the spherical model corresponding to the $`n\mathrm{}`$ limit. The dimensionless time ratio $`x=t/s`$, appearing in the two-time autocorrelation and response functions and fluctuation-dissipation ratio, is a temporal analogue of aspect ratios, which play an important role in static critical phenomena and finite-size scaling theory . One may therefore wonder whether the latter, and especially its latest developments involving conformal and modular invariance, could be used in order to put constraints on non-equilibrium critical dynamics. Generalized symmetry groups, such as those introduced in ref. , may also play a role in this issue.
#### Acknowledgements
It is a pleasure for us to thank A.J. Bray and S. Franz for interesting discussions.
## Table and caption
| exponent | spherical $`(2<D<4)`$ | spherical $`(D>4)`$ | Ising $`(D=2)`$ |
| --- | --- | --- | --- |
| $`\eta `$ | $`0`$ | $`0`$ | $`1/4`$ |
| $`\beta `$ | $`1/2`$ | $`1/2`$ | $`1/8`$ |
| $`\nu `$ | $`1/(D2)`$ | $`1/2`$ | $`1`$ |
| $`z`$ | $`2`$ | $`2`$ | $`2`$ |
| $`\lambda `$ | $`D/2`$ | $`D/2`$ | $`1.25`$ |
| $`z_c`$ | $`2`$ | $`2`$ | $`2.17`$ |
| $`\lambda _c`$ | $`3D/22`$ | $`D`$ | $`1.59`$ |
| $`\mathrm{\Theta }_c`$ | $`1D/4`$ | $`0`$ | $`0.19`$ |
Table 1: Static and dynamical exponents of the ferromagnetic spherical model and of the two-dimensional Ising model. First group: usual static critical exponents $`\eta `$, $`\beta `$, and $`\nu `$ (equilibrium). Second group: zero-temperature dynamical exponents $`z`$ and $`\lambda `$ (coarsening below $`T_c`$). Third group: dynamic critical exponents $`z_c`$, $`\lambda _c`$, and $`\mathrm{\Theta }_c`$ (non-equilibrium critical dynamics). |
warning/0001/cond-mat0001430.html | ar5iv | text | # Pairing in Inhomogeneous Superconductors
## 1 Introduction
Understanding the high-temperature superconducting cuprates remains a major goal in quantum many-body physics. Every conventional approach has failed to adequately explain their normal and superconducting phases. This fact, together with some new experimental data, may indeed point to the need for a new conceptual frame. In this work we propose a scenario based on three basic assumptions: 1\- The superconducting state is inhomogeneous. 2\- At the inhomogeneities (stripe segments), the spin-rotational symmetry is broken, providing a background for the charge carriers to form bound pairs. 3\- These pairs Josephson-tunnel between stripes. In this scenario, there are (at least) two different energy scales: A lower one related to the phase coherence of the superconducting state (and therefore to $`T_c`$) and another related to the pairing of holes. In this regard, there are some similarities to granular superconductors.
Recent neutron, X-ray, Raman and phonon-measurements , strongly suggest that at low temperatures and moderate doping the system is spatially inhomogeneous. The simplest realization of these inhomogeneities are termed “stripes.” In these stripes, charge clusters into nanoscale one-dimensional (1D) structures while the rest of the material displays strong antiferromagnetic correlations. There is no phase separation: the stripes are spatially separated. Note that this scenario is quite different from the BCS one, where the formation of a homogeneous superconducting state is described with a homogeneous superfluid density.
Remarkably, there is a subtle interplay between magnetism and superconductivity. In the undoped case, neutron scattering experiments show a peak at $`𝐤=𝐐=(\pi ,\pi )`$. At finite hole doping, this peak splits into four (at a distance $`(\pm \delta ,\pm \delta )`$ from $`𝐐`$), indicating the formation of antiferromagnetic domains. $`\delta `$ increases with hole concentration, suggesting that the stripes come closer. (It is believed that the concentration of holes in the stripe is nearly constant and equal to 1/2.) In the insulating state the position of these four peaks is such that the stripes run along a diagonal. Upon increasing the hole concentration further, these peaks rotate by $`\pi /2`$ near the superconducting transition. This is evidence that the two features (superconductivity and stripes) are inter-related. There is also experimental evidence showing that the stripes are 1D objects. The relation between spin incommensuration and charge ordering has been experimentally shown in Ref. where, for doping $`x=\frac{1}{8}`$, X-rays diffraction displays the same four peaks (with incommensuration $`2\delta `$). On the other hand, $`T_c`$ and $`\delta `$ seem to be linearly related in LSCO and YBCO: $`k_B`$ $`T_c=\mathrm{}v^{}\delta `$, where $`v^{}`$ defines a material-dependent velocity scale. Therefore, the only dependence of $`T_c`$ on $`x`$ is through $`\delta (x)`$. There is also experimental evidence supporting the existence of a spin-gap in these compounds.
These experimental facts suggest quasi-1D objects, rich in holes, separating $`\pi `$-shifted antiferromagnetic domains. There has been considerable theoretical work attempting to prove that a stripe state is the low-energy state of homogeneous $`t`$-$`J`$ or Hubbard models , i.e., a broken symmetry state of doped Mott insulators. There is no general consensus on this issue; it is most probable that the stripe state is an excited state of those homogeneous models . Here, we adopt a different strategy, namely introducing explicitly inhomogeneous terms in the model (which break the translational symmetry). We conclude that inhomogeneous terms breaking spin-rotational invariance locally are the most efficient way to produce a bound state of two holes .
Next we will introduce our microscopic model, discuss pairing and magnetic properties of the model, and finally will make an attempt to explain the phase-locked superconducting state using a phenomenological Josephson-spaghetti model.
## 2 Microscopic Model
Our microscopic scenario starts from a homogeneous $`t`$-$`J`$ model as background:
$$H_{tJ}=t\underset{i,j,\sigma }{}c_{i\sigma }^{}c_{j\sigma }^{}+J\underset{i,j}{}(𝐒_i𝐒_j\frac{1}{4}\overline{n}_i\overline{n}_j),$$
(1)
where $`c_{i\sigma }^{}`$ creates a fermion in the space with double occupancy forbiden, $`𝐒_i`$ is the spin operator and $`\overline{n}_i=c_{i\sigma }^{}c_{i\sigma }`$ is the number operator. To mimic the stripe segments, we add inhomogeneous magnetic interactions. These inhomogeneous terms break translational invariance and spin-rotational $`SU(2)`$ symmetry locally:
$$H_{\mathrm{inh}}=\underset{\alpha ,\beta }{}\delta J_zS_\alpha ^zS_\beta ^z+\frac{\delta J_{}}{2}\left(S_\alpha ^+S_\beta ^{}+S_\alpha ^{}S_\beta ^+\right),$$
with $`\delta J_{}\delta J_z`$, representing the magnetic perturbation of a static local Ising anisotropy, locally lowering spin symmetry (named $`t`$-$`JJ_z`$ model). Only a few links $`\alpha ,\beta `$ (at the stripes) have this lowered spin symmetry.
We have studied the binding energy of two holes ($`E_b=(E_{2\mathrm{holes}}E_{0\mathrm{hole}})2(E_{1\mathrm{hole}}E_{0\mathrm{hole}})`$) for different 1D and 2D lattices and extrapolated to the thermodynamic limit (Fig.1 Left). We conclude that only the $`t`$-$`JJ_z`$ model leads to considerable binding (we have also tried one-band Hubbard models with many different inhomogeneous terms). In the bound state, depending upon the value of $`t`$, the holes pair in the same or on different stripe segments (in both cases the binding energy is appreciable). It has been suggested that homogeneously breaking the spin-rotational symmetry stabilizes the stripe state . Here, we argue that doing it inhomogeneously also gives an excellent hole binding mechanism. One should note that the pair is not bound to an inhomogeneity: the wave function is spread over the whole system, but more concentrated around the stripes.
This model can also explain the magnetic properties outlined above. For instance, we have calculated the magnetic structure factor in an 8$`\times `$2 cluster, in which we have placed two stripes by breaking the spin-symmetry in a $`Y`$-link every 4 sites (the stripes are perpendicular to the $`X`$-axis). Although we cannot perform a good scaling here, in all cases studied the binding energy extrapolates to a significant value. The spin structure factor is shown in Fig.1 Right. The experimentally observed incommensuration appears for sufficiently high kinetic energy, $`t`$.
Contrary to the homogeneous $`t`$-$`J`$ and inhomogeneous $`t`$-$`JJ^{}`$ (where a link is weakened without breaking the spin-rotational symmetry) models, the $`t`$-$`JJ_z`$ model displays a spin-gap, as seen experimentally.
For a concentration near optimal doping, the stripe segments are close enough to losing their identity, suggesting the mechanism for the decrease of $`T_c`$.
## 3 Phenomenological approach to inhomogeneous superfluidity
To understand the linear relation between $`T_c`$ and $`\delta (x)`$, based on the microscopic model phenomenology, we introduce a Josephson Spaghetti model. This linear relation could be explained by connecting the superconducting mechanism to stripe fluctuations. In the following we consider Josephson tunneling of pairs between stripe segments. The simplest mean-field model involving only the phase $`\varphi (r_i)=\varphi _i`$ of the order parameter is
$`={\displaystyle \underset{ij}{}}J_{ij}\mathrm{exp}[i(\varphi _i\varphi _j)],\text{where}J_{ij}=J(r_{ij})=t_0/r_{ij}^\alpha .`$ (2)
The indices $`i`$ and $`j`$ stand for coarsed grained regions where the phase is well defined (around the stripes). The Josephson coupling is an inter and intra stripe distance dependant quantity $`J(r)`$, and the distance $`r`$ is a variable with a certain distribution $`P(r)`$. The mean-field $`T_c`$ depends upon the Josephson coupling $`J(r)`$. For simplicity we will take $`P(r)`$ as the “box” distribution depicted in Fig. 2. We find
$$J(r)=d^2rP(r)J(r)=\frac{2\pi t_0C}{2\alpha }a_1\mathrm{}^{2\alpha },r=\frac{2\pi C}{3}a_2\mathrm{}^3,$$
(3)
with the constants $`a_1`$ and $`a_2`$ $`𝒪(1)`$ numbers. Thus, for $`\alpha =1`$, we obtain the experimentally observed relation: $`T_c(x)J(r)[r]^1=\delta (x)`$. |
warning/0001/hep-ph0001183.html | ar5iv | text | # 1 Low energy dynamics
## 1 Low energy dynamics
Low energy hadron dynamics is determined by the structure of the QCD vacuum. Once we thought this ground state was empty, but now we know it is a seething cauldron of quarks, antiquarks and gluons. Indeed, so strong are their interactions that they form all measure of condensates: $`q\overline{q}`$, $`𝒢𝒢`$, $`\overline{q}𝒢q`$, etc. It is the nature of this vacuum and the scale of these condensates that determine low energy hadron physics. It is this that we can learn about at DA$`\mathrm{\Phi }`$NE.
Most importantly, the vacuum determines the spectrum of hadrons and its scale. Thus the nucleon has a mass of 1 GeV, while conventional $`\overline{q}q`$ mesons, like the $`\rho `$, are two-thirds of this mass, but pions merely 140 MeV. Indeed, it is these masses squared that determines dynamics and having $`m_\pi ^2=\mathrm{\hspace{0.17em}0.02}`$ GeV<sup>2</sup> makes pions by far the lightest of all hadrons. But why? This is a question for which we have had a good idea of the answer for many decades, but in fact its only now that we are on the threshold of definitively testing.
We begin with QCD with two flavours of quark, for simplicity. Then we have kinetic energy and interaction terms in the Lagrangian for both the up quark and the down. Since the proton and neutron masses are almost equal, we can imagine that $`m_u=m_d`$. Then QCD has an exact $`SU(2)_F`$ symmetry, which is reflected at the hadron level by the proton and neutron having the same strong interactions. Now, the scale of QCD is fixed on renormalization by the momentum scale at which the strong interaction is strong. This $`\mathrm{\Lambda }_{QCD}100200`$ MeV. The current masses of the $`u`$ and $`d`$ quarks are very small. At a few MeV, they are much smaller than $`\mathrm{\Lambda }_{QCD}`$ and so almost massless. If we separate the quark wavefunction into left-handed and right-handed components, it is only the mass term in the Lagrangian that couples left and right. So, if $`m_u=m_d=0`$, there is no such coupling and we can make $`SU(2)_F`$ transformations in the right-handed space, quite independently of those on the left. While this is a symmetry at the quark level, it is not there at the hadron level. Scalar and pseudoscalars, vectors and axial vectors are not degenerate in mass. Thus this symmetry must be broken at the hadron level.
A way to picture this was introduced by Nambu nearly forty years ago, long before the advent of QCD. Consider a hadron world with just scalar, $`\sigma `$, and pion fields $`\pi `$ . Then we can imagine that the form of the potential generated by their interactions, though symmetric in $`\sigma `$ and $`\pi `$, may not be simply parabolic with a minimum when these fields are zero, but could be Mexican hat shaped, Fig. 1. Then the ground state chosen by nature is where the $`\sigma `$ field has a non-zero vacuum expectation value, Fig. 1. The particles we observe correspond to quantum fluctuations about this ground state. The scalar field is massive as the fluctuations go up and down the sides of the hat, while the Goldstone fluctuations are massless corresponding to motion round the hat. This is an illustration of Goldstone’s theorem with Nambu’s ferromagnetic analogy. It is natural to identify these Goldstone modes with the pions, or in a world with 3 massless quark flavours with the $`\pi `$, $`K`$ and $`\eta _8`$.
This is only a model of the hadron world . What does QCD tell us is happening at the quark-gluon level? We can calculate the fully dressed quark propagator by summing all possible gluonic emissions and absorptions. If the current mass of the quarks is zero, then it is well known that the fermion mass function remains zero at all orders in perturbation theory. Thus any breaking of chiral symmetry must be non-perturbative. This we might calculate on the lattice, but zero mass particles don’t fit on a finite sized lattice. Consequently, we most naturally calculate the quark propagator in the continuum using the Schwinger-Dyson equations. These provide a genuinely non-perturbative set of integral equations in which the behaviour of the quark propagator is determined in terms of the dressed gluon (and ghost) propagators, and quark-gluon interactions . Such calculations have reached a stage of maturity that we can be certain that provided the effective quark-gluon coupling is large at infrared momenta less than $`\mathrm{\Lambda }_{QCD}`$, then a non-zero mass function is generated even though the current mass is zero. This mass function, Fig. 2 , produced by gluonic clouds and the $`q\overline{q}`$ sea is 350–400 MeV at low Euclidean momenta. Thus a quark’s constituent mass is dynamically generated. Importantly, this is merely an effective low energy mass: quarks being confined have no poles in their propagators and are never on-shell. From the large momentum behaviour of this mass function we learn this corresponds to the formation of $`q\overline{q}`$ condensates of size
$$q\overline{q}_0=(240\mathrm{MeV})^3.$$
A value totally consistent with QCD sum-rule phenomenology.
With the light quark propagator determined, one can consider the $`q\overline{q}`$ bound states with pseudoscalar and scalar quantum numbers, Fig. 3. The pseudoscalar bound state is massless, as one would expect for an explicit realisation of Goldstone’s theorem. In contrast, the scalar is massive, with a mass of roughly twice that of the dressed (or constituent) quark. These scalars are the Higgs bosons of the strong interaction, responsible for the masses of all light hadrons .
Though these QCD calculation and Nambu’s modelling are quite consistent in their phenomenology, there has been the proposal of Jan Stern and collaborators that chiral symmetry breaking need not be the result of such a large $`q\overline{q}`$ condensate. Instead this condensate could be small, perhaps $`(100\mathrm{MeV})^3`$, as in an antiferromagnetic analogy, and other condensates could be important in breaking chiral symmetry. Since physics is an experimental subject, we must test this hypothesis. The study of low energy $`\pi \pi `$ scattering provides just such an opportunity.
Consider the amplitude $``$ for the process $`\pi ^+\pi ^{}\pi ^0\pi ^0`$ in the $`t`$channel . If one of the pions is massless, chiral symmetry requires that in the limit when the 4–momentum of this massless pion goes to zero, the amplitude vanishes. Thus we have at the symmetry point of the Mandelstam triangle $`(s=t=u=m_\pi ^2)=0`$ — this is known as the Adler condition. This means that, though $`\pi \pi `$ scattering is by definition a strong interaction process, it is actually weak at low energies. It is then natural that we can make a Taylor series expansion of the low energy amplitude about the Adler point in powers of momenta and the mass of the off-shell pion. The scale for this expansion is supplied by $`m_\rho ^2`$ in hadron dynamics, or more naturally by $`32\pi f_\pi ^2`$ in chiral dynamics terms. Then at the symmetry point of the Mandelstam triangle with all the pions having their physical mass, we have the prediction
$$\left(s=t=u=\frac{4}{3}m_\pi ^2\right)=\frac{\alpha m_\pi ^2}{32\pi f_\pi ^2}+\mathrm{},$$
(1)
where the parameter $`\alpha `$ depends of the size of $`q\overline{q}`$. In fact, $`\alpha 1`$ for Standard Chiral Perturbation Theory ($`S\chi PT`$ with a large condensate and $`\alpha 4`$ with a small condensate , as is possible in Generalised $`\chi PT`$. Thus,
$`\left(s=t=u=\frac{4}{3}m_\pi ^2\right)`$ $``$ $`0.02\mathrm{in}\mathrm{S}\chi \mathrm{PT}`$ (2)
$`\left(s=t=u=\frac{4}{3}m_\pi ^2\right)`$ $``$ $`0.09\mathrm{in}\mathrm{G}\chi \mathrm{PT}.`$ (4)
Though the size of the $`q\overline{q}`$condensate can engender a factor of 4 variation in the amplitude at the centre of the Mandelstam triangle, both versions of the Chiral Expansion agree in the physical regions, where for instance at the $`\rho `$ their parameters are fixed by experimental data, Fig. 4. Thus only very close to threshold is there any testable difference between the large and small condensate predictions.
The closest to the symmetry point that one can get experimentally is by considering the lifetime of pionic atoms. This provides a direct measurement of the amplitude $``$ at $`t=4m_\pi ^2,s=u=0`$, Fig. 4. The decay rate, which is the inverse of the lifetime, is proportional to the modulus squared of this amplitude, with proportionality constants that have recently been corrected by higher orders in the electromagnetic coupling, Fig. 5, and for higher orders in the chiral expansion, among other considerations . There is now a consensus on the relationship . PS212 at CERN, the DIRAC experiment , is designed to measure the lifetime of such pionic atoms. DIRAC has had a successful run this year and the expectation is that with further running scheduled for 2000, they will be able to determine the scattering length for $`\pi ^+\pi ^{}\pi ^0\pi ^0`$ to an accuracy of 10%, with the ultimate hope of $`\pm 5\%`$ in later years.
Another experiment that we will hear about at this meeting involves $`K_{e4}`$ decays, both in E865 at Brookhaven and here with the KLOE detector . With a branching ratio of $`4.10^5`$, kaons decay into a dilepton pair, mainly $`e\nu `$, and a charged dipion pair. Measuring the dependence of the decay distribution on its five kinematic variables, one can in principle determine the relative phase of the $`\pi \pi `$ system in an $`S`$ and $`P`$wave. By the famous theorem of final state interactions of Watson, this phase difference is exactly the same phase difference as in $`\pi \pi `$ scattering. The pions scatter universally, independently of the way they are produced. The results for this phase difference from previous experiments by the University of Pennsylvania with 7000 decays and by the Geneva-Saclay group with 30000 are shown in Fig. 6, together with the predictions of S$`\chi `$PT and for G$`\chi `$PT with $`\alpha =2`$ and 3. One sees that the error bars from the current experiments have to be substantially smaller than those in Fig. 6 if we are going to be able to discriminate between values of $`\alpha `$ and hence the size of the $`q\overline{q}`$condensate. However, the correlation with energy is just as important as individual data-points for this determination. This we will hear about too at this meeting .
## 2 $`\eta \eta ^{}`$ mixing
Another topic that addresses the nature of the QCD vacuum is the study of $`\eta \eta ^{}`$ mixing. We are all familiar with the mixing among the vector and tensor mesons, where the octet and singlet components mix ideally to give physical states of definite quark flavour. We know the $`\omega `$ and $`\varphi `$ are within $`3^o`$ of this ideal situation. The same pattern emerges whether one considers the meson masses, their hadronic decays, their production in $`e^+e^{}`$, etc. However, it has been known for some time that $`\eta \eta ^{}`$ mixing depends critically on the quantity considered . As recognised by Leutwyler , two mixing angles are in fact required: one for the octet sector and the other for the singlet. These differ because the singlet couples to the vacuum.
First we specify decay constants $`f_P^a`$ for the nine pseudoscalars $`P`$ (with momentum $`p`$) by analogy with that of the pion from
$$\mathrm{\hspace{0.17em}0}A_\mu ^aP(p)=i\sqrt{2}f_P^ap_\mu ,$$
(5)
where $`a=0,1,\mathrm{},8`$. We can then define two mixing angles $`\theta _0`$ and $`\theta _8`$ by
$$f_\eta ^8/f_\eta ^{}^8=\mathrm{cot}\theta _8,f_\eta ^0/f_\eta ^{}^0=\mathrm{tan}\theta _0.$$
(6)
These angles are defined so that in the limit in which they vanish the $`\eta _8\eta `$ and $`\eta _0\eta ^{}`$. In the limit of exact $`SU(3)`$ flavour, $`\theta _0=\theta _8`$. Now, Leutwyler has shown how one can expand the difference in these angles as
$$\mathrm{sin}\left(\theta _0\theta _8\right)=\frac{2\sqrt{2}\left(f_K^2f_\pi ^2\right)}{4f_K^2f_\pi ^2}+\mathrm{}.$$
(7)
As the physical kaon decay constant $`f_K1.22f_\pi `$, this implies that $`\theta _0\theta _816^o`$. We are thus far from the $`SU(3)_F`$ symmetry in this sector — the singlet couples to the vacuum. In a recent comprehensive survey of $`\eta \eta ^{}`$ mixing, Feldmann has shown how the mixing in different situations for different quantities can all be reconciled by introducing these two mixing angles, which differ by $`16^o\pm 2^o`$.
The reason for the interest in the $`\eta `$ and $`\eta ^{}`$ is because of the way these states are intimately tied to the structure of the QCD vacuum. Clearly, the $`\eta `$ is only a little heavier than the kaons, but the $`\eta ^{}`$ at 958 MeV is much heavier, particularly as it is the mass squareds that are relevant. In the $`SU(3)`$ chiral limit, when the masses of the $`u`$, $`d`$ and $`s`$ quarks all go to zero, the 8 flavour components of the octet axial vector current of Eq. (3) are all conserved and the $`\eta _8`$ is massless. However, the divergence of the singlet current is given by the topological charge density and according to the Adler-Bell-Jackiw anomaly is non-zero. In contrast, in the limit of large $`N_c`$, all nine components of the divergence of the axial vector current vanish and the $`\eta _0`$ becomes a ninth Goldstone boson. Thus the mixing in this sector teaches us how close nature is to the chiral or to the large $`N_c`$ limits of QCD.
From the quark line diagrams shown in Fig. 7, it was recognised long ago by Deshpande and Eilam , and then by Rosner , that $`\varphi `$ radiatively decaying to the $`\eta `$ and $`\eta ^{}`$ can test their strange quark content, the $`\varphi `$ acting as an $`s\overline{s}`$ source. At that time the mixing was assumed to be simply governed by one mixing angle. Now we know better. Nevertheless, these decays provide a key test of the mixing, fixing parameters such as $`\mathrm{\Lambda }_3`$, the coupling to the Wess-Zumino term . The branching ratio for $`\varphi \gamma \eta `$ is reasonably well determined to be $`(1.26\pm 0.06)`$. From this, one predicts that the corresponding ratio for $`\varphi \gamma \eta ^{}`$ is between 4 and 8 times $`10^5`$ depending on the value of $`0.3\mathrm{\Lambda }_30.1`$. The 1997 experimental value from CMD2 gave a far bigger range from the branching ratio of $`(12\pm 7).10^5`$. However, the more recent results from VEPP-2M show that with much improved statistics, DA$`\mathrm{\Phi }`$NE should be able not only to test our current understanding of $`\eta \eta ^{}`$ mixing, but to fix the parameters in such a scheme.
## 3 Scalar mesons
The nature of the lightest scalar mesons has long been an enigma . There are far more than can fit into one quark model multiplet, Table 1. Some have claimed that there is a $`\kappa `$ at 900 MeV , as well as at 1430 MeV, and so perhaps there are two nonets. The LASS data rules this out , just as a re-analysis of the CERN-Munich polarised target data rules out the oft-claimed narrow $`\sigma (750)`$, see for example . Even so, there are too many isoscalar states, Table 1. One may be a glueball, or several may be mixed with glue .
A key aspect of the scalars is their dressing by hadrons. All mesons, of course, spend part of their time in a multi-meson continuum, $`\rho `$ in $`\pi \pi `$, $`\varphi `$ in $`K\overline{K}`$, etc., it is through these channels that these unstable particles decay. However, these hadronic dressings turn out to be a small perturbation on the underlying $`q\overline{q}`$ states. However, in the case of scalars, this is not the case. It is a phenomenological fact that their couplings to mesons are much larger. Their lifetimes shorter. They are looser systems. Consequently, they readily form dimeson systems. The details may depend on the calculational scheme . Nevertheless, even if the underlying scalar quark model multiplet is centred at 1400 MeV, the isotriplet $`a_0`$ and one of the isosinglet $`f_0`$ states automatically has such strong couplings to $`K\overline{K}`$, that they spend an appreciable fraction of their lifetime as $`K\overline{K}`$ systems that the masses of the physical hadrons are pulled to 1 GeV . That the $`f_0(980)`$ likes to couple strongly to $`K\overline{K}`$, or equivalently $`s\overline{s}`$ systems, is observed in its distinct appearance in channels with hidden strangeness. While the $`f_0(980)`$ appears as a sharp dip, or a shoulder, in $`\pi \pi \pi \pi `$ and in central production, such as $`pppp\pi \pi `$, it appears as a pronounced peak in $`J/\psi \varphi \pi \pi `$, as observed long ago . This behaviour is reproduced by the $`\pi \pi `$ spectrum in $`\varphi `$ radiative decay as shown in Fig. 8. Hopefully improved statistics at DA$`\mathrm{\Phi }`$NE will sharpen up its signal.
Once again these mesons illuminate our knowledge of the QCD vacuum. In the $`1/N_c`$ expansion, we expect processes in which flavour is annihilated to be suppressed. This we recognise as the OZI-rule, Fig. 9.
Even though this perturbative QCD argument is not expected to be as good for light flavours with $`N_c=3`$, nevertheless within the large $`N_c`$ framework, this still applies. However, the appearance of the $`f_0(980)`$ (and $`a_0(980)`$) mesons just at $`K\overline{K}`$ threshold counters this suppression. The OZI-rule not merely fails, but is reversed. As emphasised by Moussallam , this tests the flavour dependence of the QCD condensates.
It has long been advertised that, depending on the nature of these mesons, whether they are simple $`q\overline{q}`$, $`4`$quark as $`qq\overline{qq}`$ or as $`K\overline{K}`$molecule , there are significant differences in the branching ratio for $`\varphi `$ radiative decays to $`f_0`$ and $`a_0`$, see Tables 2, 3.
In reality, the choices are not likely to be so stark. In the calculation of Tornqvist (and Boglione and myself) the physical $`a_0(980)`$ is only 20% $`q\overline{q}`$ and 70% $`K\overline{K}`$. An outstanding calculation is to deduce what this inevitable mixture means for the predictions for their two photon decays and for their appearance in $`\varphi `$radiative decays. That is a challenge for theorists.
## 4 Challenge for DA$`\mathrm{\Phi }`$NE
The challenge for DA$`\mathrm{\Phi }`$NE is to produce high statistics results on $`Ke\nu \pi ^+\pi ^{}`$, and on all of $`\varphi \gamma \eta `$, $`\gamma \eta ^{}`$, $`\gamma f_0`$ and $`\gamma a_0`$. Then we have the hope of understanding the nature of the QCD vacuum a little more clearly: a vacuum that shapes the world of hadrons, of hadronisation and all other aspects of confinement — an exciting prospect.
## 5 Acknowledgements
I would like to thank my many colleagues in the DA$`\mathrm{\Phi }`$NE physics community for their enthusiasm for this project. I acknowledge travel support from the EEC-TMR Programme, Contract No. CT98-0169, EuroDA$`\mathrm{\Phi }`$NE. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.