jackkuo commited on
Commit
92153fc
·
verified ·
1 Parent(s): 955a718

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. -tFIT4oBgHgl3EQf9Cv1/content/tmp_files/2301.11405v1.pdf.txt +1366 -0
  2. -tFIT4oBgHgl3EQf9Cv1/content/tmp_files/load_file.txt +0 -0
  3. .gitattributes +45 -0
  4. 09FKT4oBgHgl3EQfOS13/content/2301.11758v1.pdf +3 -0
  5. 0NFRT4oBgHgl3EQfkTex/content/2301.13595v1.pdf +3 -0
  6. 0NFRT4oBgHgl3EQfkTex/vector_store/index.pkl +3 -0
  7. 0tAyT4oBgHgl3EQfoPjL/content/tmp_files/2301.00505v1.pdf.txt +419 -0
  8. 0tAyT4oBgHgl3EQfoPjL/content/tmp_files/load_file.txt +343 -0
  9. 19E2T4oBgHgl3EQfiwfE/vector_store/index.pkl +3 -0
  10. 1NAyT4oBgHgl3EQfbff_/content/tmp_files/2301.00266v1.pdf.txt +0 -0
  11. 1NAyT4oBgHgl3EQfbff_/content/tmp_files/load_file.txt +0 -0
  12. 1NAzT4oBgHgl3EQfRftD/content/2301.01216v1.pdf +3 -0
  13. 1NAzT4oBgHgl3EQfRftD/vector_store/index.faiss +3 -0
  14. 1NAzT4oBgHgl3EQfRftD/vector_store/index.pkl +3 -0
  15. 3tAzT4oBgHgl3EQfD_po/content/2301.00985v1.pdf +3 -0
  16. 3tAzT4oBgHgl3EQfD_po/vector_store/index.faiss +3 -0
  17. 49E0T4oBgHgl3EQfegBh/content/tmp_files/2301.02391v1.pdf.txt +1561 -0
  18. 49E0T4oBgHgl3EQfegBh/content/tmp_files/load_file.txt +0 -0
  19. 49E1T4oBgHgl3EQfAwK8/vector_store/index.pkl +3 -0
  20. 4dAyT4oBgHgl3EQf2Pkf/content/2301.00746v1.pdf +3 -0
  21. 4dAyT4oBgHgl3EQf2Pkf/vector_store/index.pkl +3 -0
  22. 4tE4T4oBgHgl3EQf1A0X/content/2301.05286v1.pdf +3 -0
  23. 6NA0T4oBgHgl3EQfN_-e/content/tmp_files/2301.02155v1.pdf.txt +1206 -0
  24. 6NA0T4oBgHgl3EQfN_-e/content/tmp_files/load_file.txt +0 -0
  25. 7NE2T4oBgHgl3EQf7giQ/content/tmp_files/2301.04210v1.pdf.txt +1080 -0
  26. 7NE2T4oBgHgl3EQf7giQ/content/tmp_files/load_file.txt +0 -0
  27. 7tFAT4oBgHgl3EQfoR3-/content/tmp_files/2301.08634v1.pdf.txt +2775 -0
  28. 7tFAT4oBgHgl3EQfoR3-/content/tmp_files/load_file.txt +0 -0
  29. 9tAzT4oBgHgl3EQfFPrO/vector_store/index.pkl +3 -0
  30. A9E1T4oBgHgl3EQfVgQy/vector_store/index.pkl +3 -0
  31. AdAzT4oBgHgl3EQfTPx3/vector_store/index.pkl +3 -0
  32. CdFAT4oBgHgl3EQftB4M/content/2301.08661v1.pdf +3 -0
  33. DtAyT4oBgHgl3EQfSPej/content/tmp_files/2301.00083v1.pdf.txt +1482 -0
  34. DtAyT4oBgHgl3EQfSPej/content/tmp_files/load_file.txt +0 -0
  35. ENE1T4oBgHgl3EQfEQP0/vector_store/index.pkl +3 -0
  36. EdE3T4oBgHgl3EQfVQpt/content/tmp_files/2301.04458v1.pdf.txt +1536 -0
  37. EdE3T4oBgHgl3EQfVQpt/content/tmp_files/load_file.txt +0 -0
  38. F9AzT4oBgHgl3EQfHPuf/vector_store/index.faiss +3 -0
  39. FNAzT4oBgHgl3EQfw_7Q/content/2301.01732v1.pdf +3 -0
  40. FNAzT4oBgHgl3EQfw_7Q/vector_store/index.faiss +3 -0
  41. FNE2T4oBgHgl3EQf-Anw/content/tmp_files/2301.04235v1.pdf.txt +0 -0
  42. FNE2T4oBgHgl3EQf-Anw/content/tmp_files/load_file.txt +0 -0
  43. GNE4T4oBgHgl3EQfgA3B/vector_store/index.faiss +3 -0
  44. INAzT4oBgHgl3EQfU_y9/content/tmp_files/2301.01277v1.pdf.txt +1185 -0
  45. INAzT4oBgHgl3EQfU_y9/content/tmp_files/load_file.txt +0 -0
  46. IdE0T4oBgHgl3EQfRwCH/content/tmp_files/2301.02212v1.pdf.txt +0 -0
  47. IdE0T4oBgHgl3EQfRwCH/content/tmp_files/load_file.txt +0 -0
  48. ItE1T4oBgHgl3EQfYARl/content/2301.03133v1.pdf +3 -0
  49. ItE1T4oBgHgl3EQfYARl/vector_store/index.pkl +3 -0
  50. N9E3T4oBgHgl3EQfZQq9/content/2301.04496v1.pdf +3 -0
-tFIT4oBgHgl3EQf9Cv1/content/tmp_files/2301.11405v1.pdf.txt ADDED
@@ -0,0 +1,1366 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
2
+ Zhongwen Zhang Yuri Boykov
3
+ University of Waterloo
4
+ {z889zhan, yboykov}@uwaterloo.ca
5
+ Abstract
6
+ Maximization of mutual information between the
7
+ model’s input and output is formally related to
8
+ “decisiveness” and “fairness” of the softmax pre-
9
+ dictions (Bridle et al., 1991), motivating such un-
10
+ supervised entropy-based losses for discrimina-
11
+ tive neural networks. Recent self-labeling meth-
12
+ ods based on such losses represent the state of
13
+ the art in deep clustering. However, some impor-
14
+ tant properties of entropy clustering are not well-
15
+ known or even misunderstood. For example, we
16
+ provide a counterexample to prior claims about
17
+ equivalence to variance clustering (K-means) and
18
+ point out technical mistakes in such theories.
19
+ We discuss the fundamental differences between
20
+ these discriminative and generative clustering ap-
21
+ proaches. Moreover, we show the susceptibility of
22
+ standard entropy clustering to narrow margins and
23
+ motivate an explicit margin maximization term.
24
+ We also propose an improved self-labeling loss;
25
+ it is robust to pseudo-labeling errors and enforces
26
+ stronger fairness. We develop an EM algorithm
27
+ for our loss that is significantly faster than the
28
+ standard alternatives. Our results improve the
29
+ state-of-the-art on standard benchmarks.
30
+ 1. Background and motivation
31
+ Entropy-based loss functions, e.g. decisiveness and fairness,
32
+ were proposed for network training (Bridle et al., 1991;
33
+ Krause et al., 2010) and regularization (Grandvalet & Ben-
34
+ gio, 2004) and are commonly used for unsupervised and
35
+ weakly-supervised classification problems (Ghasedi Dizaji
36
+ et al., 2017; Hu et al., 2017; Ji et al., 2019; Asano et al.,
37
+ 2020; Jabi et al., 2021). In particular, the state-of-the-art in
38
+ unsupervised classification (Asano et al., 2020; Jabi et al.,
39
+ 2021) is achieved by self-labeling methods using extensions
40
+ of decisiveness and fairness.
41
+ The community pursues challenging applications of unsu-
42
+ pervised classification using deep neural networks, but as
43
+ we show in this paper, some important basic properties of
44
+ entropy-based clustering are not well-understood or even
45
+ examples of linear decision functions over X ∈ R2
46
+ kµ(X) = arg mink ∥X − µk∥
47
+ σv(X) = soft-max(v⊤X)
48
+ (a) variance clustering
49
+ (b) entropy clustering
50
+ Figure 1. Variance vs entropy clustering - binary example (K = 2)
51
+ for 2D data {Xi} ⊂ RN (N = 2) comparing linear methods of
52
+ similar parametric complexity: (a) K-means [µk ∈ RN] and (b)
53
+ entropy clustering based on a linear classifier using K-columns lin-
54
+ ear discriminator matrix v = [vk ∈ RN] and soft-max predictions.
55
+ Red and green colors in (a) and (b) illustrate optimal linear decision
56
+ regions over X ∈ R2 produced by the decision functions kµ(X),
57
+ σv(X) for parameters µ and v minimizing two losses: (a) com-
58
+ pactness/variance of clusters �
59
+ i ∥Xi−µki∥2 where ki = kµ(Xi)
60
+ and (b) decisiveness and fairness of predictions �
61
+ i H(σi)−H(¯σ)
62
+ where H(·) is entropy function, σi = σv(Xi) and ¯σ = avg{σi}.
63
+ The decisions kµ(X) in (a) are hard and σv(X) in (b) are soft
64
+ (distributions). The softness is visualized by transparency. The
65
+ optimal results in (a) and (b) are analyzed in Sec.1.1. The result in
66
+ (b) may require margin maximization term, see Fig.3 in Sec.2.1.
67
+ understood wrongly. Lapses of clarity call for simple illus-
68
+ trative tests, but we did not find any basic low-level exam-
69
+ ples of entropy clustering in prior work. We observe that
70
+ decisiveness and fairness are general criteria applicable to
71
+ any soft-max model, not necessarily deep. Thus, it should
72
+ be possible to use them for unsupervised clustering even
73
+ with a basic linear classifier using soft-max output. Our
74
+ Fig.1(b) shows decision regions for an optimal linear classi-
75
+ fier trained for 2D data without any supervision using only
76
+ the standard decisiveness & fairness loss. It is natural to
77
+ juxtapose such entropy-based linear clustering with the most
78
+ popular linear clustering method, K-means, see Fig.1(a).
79
+ arXiv:2301.11405v1 [cs.LG] 26 Jan 2023
80
+
81
+ kμ(X)= 0
82
+ kμ(X)=1
83
+ compactness
84
+ of clustersv(X)~ onehoto
85
+ 0v(X)
86
+ ~ onehot1
87
+ decisiveness & fairness
88
+ of predictionsRevisiting Discriminative Entropy Clustering and its relation to K-means
89
+ 0.0
90
+ 0.2
91
+ 0.4
92
+ 0.6
93
+ 0.8
94
+ corruption level
95
+ 15%
96
+ 25%
97
+ 35%
98
+ 45%
99
+ 55%
100
+ 65%
101
+ accuracy
102
+ forward CE: H(y,
103
+ )
104
+ forward CE: H(y,
105
+ )
106
+ reverse CE: H( , y)
107
+ Figure 2. Robustness to noisy labels:
108
+ reverse cross-entropy
109
+ H(σ, y) vs standard cross-entropy H(y, σ). These losses are used
110
+ to train VGG-4 network on fully-supervised STL10 data with cor-
111
+ rupted labels. The horizontal axis shows the level of corruption,
112
+ i.e. percentage η of training images where the correct ground
113
+ truth labels were replaced by a random label. We use soft target
114
+ distributions ˜y = η ∗ u + (1 − η) ∗ y representing the mixture
115
+ of one-hot distribution y for the observed corrupt label and the
116
+ uniform distribution u, as recommended in (M¨uller et al., 2019).
117
+ The vertical axis shows the test accuracy. Training with reverse
118
+ cross-entropy is robust to high levels of labeling errors.
119
+ Our paper provides both conceptual and algorithmic contri-
120
+ butions briefly summarized below. First, our simple illus-
121
+ trative example in Fig.1 works as a counterexample for the
122
+ main theoretical claim of a recent TPAMI paper (Jabi et al.,
123
+ 2021) wrongly stating the equivalence between the loss
124
+ functions for discriminative entropy clustering and variance
125
+ clustering, a.k.a. K-means. We point out specific technical
126
+ errors in their proof later in Section 1.2. Our paper also dis-
127
+ cusses the susceptibility of standard formulations of entropy
128
+ clustering to narrow decision margins and how to avoid
129
+ them. We also propose a new formulation of entropy-based
130
+ self-labeling loss for clustering. All standard self-labeling
131
+ methods replace the entropy H(σ) of soft-max predictions σ
132
+ (decisiveness) by cross-entropy H(y, σ) with pseudo-labels
133
+ y representing extra hidden variables. In contrast, we pro-
134
+ pose reverse cross-entropy H(σ, y) arguably demonstrating
135
+ improved robustness to labeling errors, e.g. Fig.2, which
136
+ are expected in estimated soft pseudo-labels y. Note that
137
+ the second position inside the cross-entropy is natural for
138
+ estimated distributions. At the same time, the place of the
139
+ first argument is natural for the network predictions σ since
140
+ cross-entropy H(σ, y) is an upper-bound approximation for
141
+ decisiveness H(σ). We also propose a stronger formula-
142
+ tion of the fairness constraint. Our new self-labeling loss
143
+ addresses limitations of the standard formulations and it is
144
+ amenable to an efficient EM solver derived in our paper.
145
+ The rest of this introductory section is organized as fol-
146
+ lows. First, Section 1.1 reviews the background and no-
147
+ tation for entropy-based clustering with soft-max models.
148
+ Then, Section 1.2 reviews the most closely related work
149
+ using self-labeling loss formulations of entropy clustering.
150
+ We conclude the introduction by summarizing our main
151
+ contributions and outlining the structure of the whole paper.
152
+ 1.1. Background on discriminative entropy clustering
153
+ The work of Bridle, Heading, and MacKay from 1991 (Bri-
154
+ dle et al., 1991) formulated mutual information (MI) loss
155
+ for unsupervised discriminative training of neural networks
156
+ using probability-type outputs, e.g. softmax σ : RK → ∆K
157
+ mapping K logits lk ∈ R to a point in the probability
158
+ simplex ∆K. Such output σ = (σ1, . . . , σK) is often in-
159
+ terpreted as a pseudo posterior1 over K classes, where
160
+ σk =
161
+ exp lk
162
+
163
+ i exp li is a scalar prediction for each class k.
164
+ The unsupervised loss proposed in (Bridle et al., 1991) trains
165
+ the model predictions to keep as much information about
166
+ the input as possible. They derived an estimate of MI as the
167
+ difference between the average entropy of the output and
168
+ the entropy of the average output
169
+ Lmi
170
+ :=
171
+ −MI(c, X)
172
+
173
+ H(σ) − H(σ)
174
+ (1)
175
+ where c is a random variable representing class prediction,
176
+ X represents the input, and the averaging is done over all
177
+ input samples {Xi}M
178
+ i=1, i.e. over M training examples.
179
+ The derivation in (Bridle et al., 1991) assumes that soft-
180
+ max represents the distribution Pr(c|X). However, since
181
+ softmax is not a true posterior, the right-hand side in (1)
182
+ can be seen only as a pseudo MI loss. In any case, (1)
183
+ has a clear discriminative interpretation that stands on its
184
+ own: H(σ) encourages “fair” predictions with a balanced
185
+ support of all categories across the whole training dataset,
186
+ while H(σ) encourages confident or “decisive” prediction
187
+ at each data point implying that decision boundaries are
188
+ away from the training examples (Grandvalet & Bengio,
189
+ 2004), see Fig.1(b). Our paper refers to unsupervised train-
190
+ ing of discriminative soft-max models using predictions’
191
+ entropies, e.g. see (1), as discriminative entropy clustering.
192
+ This should not be confused with generative entropy clus-
193
+ tering methods where the entropy is used as a measure of
194
+ compactness for clusters’ density functions2.
195
+ As mentioned earlier, discriminative clustering loss (1) can
196
+ be applied to deep or shallow models. For clarity, this paper
197
+ distinguishes parameters w of the representation layers of
198
+ the network computing features fw(X) ∈ RN for any input
199
+ X and the linear classifier parameters v of the output layer
200
+ computing K-logit vector v⊤f for any feature f ∈ RN.
201
+ The overall network model is defined as
202
+ σ(v⊤fw(X)).
203
+ (2)
204
+ 1The term pseudo emphasizes that discriminative training does
205
+ not lead to the true Bayesian posteriors, in general.
206
+ 2E.g., K-means minimizes clusters’ variances. Their logarithms
207
+ equal the cluster’s density entropies, assuming Gaussianity.
208
+
209
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
210
+ A special “shallow” case of the model in (2) is a basic linear
211
+ discriminator
212
+ σ(v⊤X)
213
+ (3)
214
+ directly operating on low-level input features f = X. Op-
215
+ timization of the loss (1) for the shallow model (3) is done
216
+ only over linear classifier parameters v, but the deeper net-
217
+ work model (2) is optimized over all network parameters
218
+ [v, w]. Typically, this is done via gradient descent or back-
219
+ propagation (Rumelhart et al., 1986; Bridle et al., 1991).
220
+ Our simple 2D example in Fig.1(b) illustrates “decisive-
221
+ ness” and “fairness” losses (1) in the context of a linear
222
+ classifier (3) and compares with the standard “compactness”
223
+ criterion optimized by K-means, see Fig.1(a). In this “shal-
224
+ low” setting both clustering methods are linear and have
225
+ similar parametric complexities, about K × N parameters.
226
+ K-means (a) finds balanced compact clusters of the least
227
+ squared deviations or variance. This can also be interpreted
228
+ “generatively”, see (Kearns et al., 1997), as MLE-based fit-
229
+ ting of two (isotropic) Gaussian densities, explaining the
230
+ failure for non-isotropic clusters in (a). To fix (a) “gener-
231
+ atively”, one should use non-isotropic Gaussian densities,
232
+ e.g. 2-mode GMM would produce soft clusters similar to
233
+ (b). However, this has costly parametric complexity - two
234
+ extra covariance matrices to estimate and quadratic decision
235
+ boundaries. In contrast, there is no estimation of complex
236
+ data density models in (b). Entropy-based loss (1) discrim-
237
+ inatively trains a simple linear classifier (3) to produce a
238
+ balanced (“fair”) decision boundary away from the data
239
+ points (“decisiveness”). Later, we show that the “decisive-
240
+ ness” may not be sufficient to avoid narrow decision margins
241
+ without an extra margin maximization term, see Fig.3.
242
+ In the context of deep models (2), the decision boundaries
243
+ between the clusters of training data points {Xi} can be ar-
244
+ bitrarily complex since the network learns high-dimensional
245
+ non-linear representation map or embedding fw(X). In this
246
+ case, loss (1) is optimized with respect to both represen-
247
+ tation w and classification v parameters. To avoid overly
248
+ complex clustering of the training data and to improve gen-
249
+ erality, it is common to use self-augmentation techniques
250
+ (Hu et al., 2017). For example, (Ji et al., 2019) maximize
251
+ the mutual information between class predictions for input
252
+ X and its augmentation counterpart X′ encouraging deep
253
+ features invariant to augmentation.
254
+ To reduce the complexity of the model, (Krause et al., 2010)
255
+ proposed to combine entropy-based loss (1) with regular-
256
+ ization of all network parameters interpreted as an isotropic
257
+ Gaussian prior on these weights
258
+ Lmi+decay
259
+ =
260
+ H(σ) −
261
+ H(σ)
262
+ + ∥[v, w]∥2
263
+ c=
264
+ H(σ) + KL(σ ∥ u) + ∥[v, w]∥2
265
+ (4)
266
+ where
267
+ c= represents equality up to an additive constant and
268
+ u is a uniform distribution over K classes. Of course, mini-
269
+ mizing the norm of network weights as above corresponds
270
+ to the weight decay - a common default for network training.
271
+ The second formulation of the loss (4) uses KL divergence
272
+ motivated in (Krause et al., 2010) by the possibility to gener-
273
+ alize “fairness” to balancing with respect to any given target
274
+ distribution different from the uniform u.
275
+ 1.2. Review of entropy-based self-labeling
276
+ Optimization of losses (1) or (4) during network training
277
+ is mostly done with standard gradient descent or backprop-
278
+ agation (Bridle et al., 1991; Krause et al., 2010; Hu et al.,
279
+ 2017). However, the difference between the two entropy
280
+ terms implies non-convexity, which makes such losses chal-
281
+ lenging for gradient descent. This motivates alternative
282
+ formulations and optimization approaches. For example,
283
+ it is common to extend the loss by incorporating auxiliary
284
+ or hidden variables y representing pseudo-labels for unla-
285
+ beled data points X, which are to be estimated jointly with
286
+ optimization of the network parameters (Ghasedi Dizaji
287
+ et al., 2017; Asano et al., 2020; Jabi et al., 2021). Typically,
288
+ such self-labeling approaches to unsupervised network train-
289
+ ing iterate optimization of the loss over pseudo-labels and
290
+ network parameters, similarly to Lloyd’s algorithm for K-
291
+ means or EM algorithm for Gaussian mixtures (Bishop,
292
+ 2006). While the network parameters are still optimized
293
+ via gradient descent, the pseudo-labels can be optimized via
294
+ more powerful algorithms.
295
+ For example, (Asano et al., 2020) formulate self-labeling
296
+ using the following constrained optimization problem with
297
+ discrete pseudo-labels y tied to predictions by cross entropy
298
+ function H(y, σ)
299
+ Lce
300
+ =
301
+ H(y, σ)
302
+ s.t.
303
+ y ∈ ∆K
304
+ 0,1
305
+ and
306
+ ¯y = u (5)
307
+ where ∆K
308
+ 0,1 are one-hot distributions, i.e. corners of the
309
+ probability simplex ∆K. Training of the network is done by
310
+ minimizing cross entropy H(y, σ), which is convex w.r.t. σ,
311
+ assuming fixed pseudo-labels y. Then, model predictions
312
+ get fixed and cross-entropy is minimized w.r.t variables y.
313
+ Note that cross-entropy H(y, σ) is linear with respect to y,
314
+ and its minimum over simplex ∆K is achieved by one-hot
315
+ distribution for a class label corresponding to arg max(σ)
316
+ at each training example. However, the balancing constraint
317
+ ¯y = u converts minimization of cross-entropy over all data
318
+ points into a non-trivial integer programming problem that
319
+ can be approximately solved via optimal transport (Cuturi,
320
+ 2013). The cross-entropy in (5) encourages the network
321
+ predictions σ to approximate the estimated one-hot target
322
+ distributions y, which implies the decisiveness.
323
+ Self-labeling methods for unsupervised clustering can also
324
+ use soft pseudo-labels y ∈ ∆K as target distributions inside
325
+
326
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
327
+ H(y, σ). In general, soft targets y are commonly used with
328
+ cross-entropy functions H(y, σ), e.g. in the context of noisy
329
+ labels (Tanaka et al., 2018; Song et al., 2022). Softened
330
+ targets y can also assist network calibration (Guo et al.,
331
+ 2017; M¨uller et al., 2019) and improve generalization by
332
+ reducing over-confidence (Pereyra et al., 2017). In the con-
333
+ text of unsupervised clustering, cross entropy H(y, σ) with
334
+ soft pseudo-labels y approximates the decisiveness since
335
+ it encourages σ ≈ y implying H(y, σ) ≈ H(y) ≈ H(σ)
336
+ where the latter is the decisiveness term in (1). Inspired
337
+ by (4), instead of the hard constraint ¯y = u used in (5),
338
+ self-labeling losses can represent the fairness using KL di-
339
+ vergence KL(¯y ∥ u), as in (Ghasedi Dizaji et al., 2017; Jabi
340
+ et al., 2021). In particular, (Jabi et al., 2021) formulates the
341
+ following entropy-based self-labeling loss
342
+ Lce+kl
343
+ =
344
+ H(y, σ)
345
+ + KL(¯y ∥ u)
346
+ (6)
347
+ encouraging decisiveness and fairness, as discussed. Simi-
348
+ larly to (5), the network parameters in loss (6) are trained
349
+ by the standard cross-entropy term. But, optimization over
350
+ relaxed pseudo-labels y ∈ ∆K is relatively easy due to the
351
+ convexity of KL divergence and linearity of cross-entropy
352
+ w.r.t. y. While there is no closed-form solution, the authors
353
+ offer an efficient approximate solver for y. Iterating steps
354
+ that estimate pseudo-labels y and optimize the model pa-
355
+ rameters resembles Lloyd’s algorithm for K-means. The
356
+ results in (Jabi et al., 2021) also establish a formal relation
357
+ between the loss (6) and the K-means objective.
358
+ Our work is closely related to self-labeling loss (6) and
359
+ the corresponding ADM algorithm proposed in (Jabi et al.,
360
+ 2021). Their inspiring approach is a good reference point
361
+ for our self-labeling loss proposal (10). It also helps to illu-
362
+ minate some problems with standard entropy-based losses
363
+ and their limited understanding.
364
+ In particular, we disagree with the main theoretical claim in
365
+ (Jabi et al., 2021) establishing a formal equivalence between
366
+ K-means and “regularized” entropy-based clustering with
367
+ soft-max models. In fact, our Figure 1 works as a simple
368
+ 2D counterexample to their claim3. Also, they extend the
369
+ entropy-based loss with the classifier regularization ∥v∥2,
370
+ but this extra quadratic term is mainly used as a technical
371
+ tool in their proof of algebraic similarity between their loss
372
+ and the standard K-means loss4. In contrast to related prior
373
+ work, we demonstrate that ∥v∥2 is needed in discriminative
374
+ entropy clustering for margin maximization.
375
+ 3The proof of Proposition 2 has a critical technical error - it
376
+ ignores normalization for soft-max prediction in their equation (5),
377
+ which is hidden via ∝ symbol. Such normalization is critical for
378
+ pseudo-posterior models.
379
+ 4Since they ignore normalization in the softmax prediction,
380
+ then ln σ in the cross-entropy H(y, σ) turns into a linear term w.r.t.
381
+ logits v⊤x. Adding regularization ∥v∥2 to such loss allows them
382
+ to create a quadratic form with respect to v that resembles squared
383
+ errors loss in K-means, which is quadratic w.r.t means µk).
384
+ 1.3. Summary of contributions
385
+ Our paper provides conceptual and algorithmic contribu-
386
+ tions. First of all, our paper disproves the main theoretical
387
+ claim (in the title) of a recent TPAMI paper (Jabi et al.,
388
+ 2021) wrongly stating the equivalence between the stan-
389
+ dard K-means loss and entropy-based clustering losses. Our
390
+ Figure 1 provides a simple counterexample to the claim,
391
+ but we also show specific technical errors in their proof.
392
+ Figure 1 helps to motivate entropy clustering with discrimi-
393
+ native soft-max models. This general methodology is unde-
394
+ servedly little-known to the broader ML community for two
395
+ reasons: (A) it was previously presented only in the context
396
+ of complex (non-linear, deep) softmax models obfuscating
397
+ the basics and (B) because there is confusion even among
398
+ the researchers who know about it. Besides clarifying ear-
399
+ lier claims about the relation to K-means, we also show that
400
+ entropy-based losses may lead to narrow decision margins,
401
+ which may contradict one common motivation for decisive-
402
+ ness (Grandvalet & Bengio, 2004). Unlike prior entropy
403
+ clustering work, we motivate classifier norm regularization
404
+ by demonstrating its importance for margin maximization.
405
+ We also discuss the limitations of the existing self-labeling
406
+ formulations of entropy clustering and propose a new loss,
407
+ as well as an efficient pseudo-labeling algorithm. In par-
408
+ ticular, we replace standard forward cross-entropy H(y, σ),
409
+ where y are soft pseudo-labels, by the reverse cross-entropy
410
+ H(σ, y) that is significantly more robust to errors in esti-
411
+ mated soft pseudo-labels, see Figures 2 and 4(b). Our for-
412
+ mulation of fairness is motivated by a zero-avoiding version
413
+ of KL divergence enforcing stronger fairness, see Figure
414
+ 4(a). We design a new EM algorithm with closed-form EM
415
+ steps. In part, our self-labeling formulation of entropy clus-
416
+ tering is motivated by its amenability to an efficient EM
417
+ solver. Our empirical results improve the state-of-the-art on
418
+ many standard benchmarks for deep clustering.
419
+ The rest of our paper is organized as follows. Section 2
420
+ motivates our new self-labeling loss for entropy clustering
421
+ and derives our EM algorithm. Section 3 compares our ap-
422
+ proach with the state-of-the-art entropy clustering methods.
423
+ Conclusions are provided in Section 4.
424
+ 2. Our entropy clustering approach
425
+ We are focused on entropy-based losses for clustering with
426
+ softmax models that typically enforce “decisiveness” and
427
+ “fairness”. First, In Section 2.1 we argue that common for-
428
+ mulations of such losses, e.g. (1) or (5), may produce narrow
429
+ classification margins. We show that some explicit margin
430
+ maximization constraints should be added, which motivates
431
+ the classifier norm regularization ∥v∥2 similarly to SVM
432
+ methods (Xu et al., 2004). Section 2.2 introduces our new
433
+ entropy-based self-labeling loss incorporating strong fair-
434
+
435
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
436
+ (a) γ = 0
437
+ (b) γ = 0.001
438
+ (c) γ = 0.01
439
+ Figure 3. Margin maximization term γ ∥v∥2 in our loss (7): low-
440
+ level clustering results for the softmax linear classifier model (3)
441
+ with N = 2 and different weights γ. The dots represent data points.
442
+ The optimal softmax clustering of the data and the decision regions
443
+ over the whole space are visualized by σ-weighted color trans-
444
+ parency, as in Fig.1(b). The “margin” is a weak-confidence “soft”
445
+ region around the linear decision boundary lacking color-saturation.
446
+ For small γ the classifier can “squeeze” a narrow-margin linear
447
+ decision boundary just between the data points, while maintaining
448
+ arbitrarily hard “decisiveness” on the data points themselves.
449
+ ness and reverse cross-entropy. Section 2.3 derives an effi-
450
+ cient EM algorithm for an important sub-problem - estima-
451
+ tion of pseudo-labels y.
452
+ 2.1. Margin maximization via norm regularization
453
+ The average entropy term in (1), a.k.a. “decisiveness”, is
454
+ recommended in (Grandvalet & Bengio, 2004) as a general
455
+ regularization term for semi-supervised problems. They
456
+ argue that it produces decision boundaries away from all
457
+ training examples, labeled or not. This seems to suggest
458
+ larger classification margins, which are good for general-
459
+ ization. However, the decisiveness may not automatically
460
+ imply large margins if the norm of classifier v in pseudo
461
+ posterior models (2, 3) is unrestricted, see Figure 3(a). Tech-
462
+ nically, this follows from the same arguments as in (Xu
463
+ et al., 2004) where regularization of the classifier norm is
464
+ formally related to the margin maximization in the context
465
+ of their SVM approach to clustering.
466
+ Interestingly, regularization of the norm for all network pa-
467
+ rameters [v, w] is motivated in (4) differently (Krause et al.,
468
+ 2010). But, since the classifier parameters v are included,
469
+ coincidentally, it also leads to margin maximization. On the
470
+ other hand, many MI-based methods (Bridle et al., 1991;
471
+ Ghasedi Dizaji et al., 2017; Asano et al., 2020) do not have
472
+ regularizer ∥v∥2 in their clustering loss, e.g. see (5). One
473
+ may argue that practical implementations of these meth-
474
+ ods implicitly benefit from the weight decay, which is om-
475
+ nipresent in network training. It is also possible that gradient
476
+ descent may implicitly restrict the classifier norm (Soudry
477
+ et al., 2018). In any case, since margin maximization is
478
+ important for clustering, ideally, it should not be left to
479
+ chance. Thus, the norm regularization term ∥v∥2 should be
480
+ explicitly present in any clustering loss for pseudo-posterior
481
+ models.
482
+ We extend MI loss (1) by combining it with the regulariza-
483
+ tion of the classifier norm ∥v∥ encouraging margin maxi-
484
+ mization, as shown in Figure 3
485
+ Lmi+mm
486
+ :=
487
+ H(σ) −
488
+ H(σ)
489
+ + γ ∥v∥2
490
+ c=
491
+ H(σ) +
492
+ KL(σ ∥ u) + γ ∥v∥2.
493
+ (7)
494
+ We note that (Jabi et al., 2021) also extend their entropy-
495
+ based loss (6) with the classifier regularization ∥v∥2, but
496
+ this extra term is used mainly as a technical tool in relating
497
+ their loss (6) to K-means, as detailed in Section 1.3. They
498
+ do not discuss its relation to margin maximization.
499
+ 2.2. Our self-labeling loss function
500
+ Below we motivate and put forward some new ideas for
501
+ entropy-based clustering losses. First, we observe that the
502
+ entropy H(¯σ) in (1) is a weak formulation of the fairness
503
+ constraint. Indeed, as clear from an equivalent formulation
504
+ in (7), it is enforced by the reverse KL divergence for the
505
+ average predictions ¯σ. It assigns a bounded penalty even
506
+ for highly unbalanced solutions where ¯σk = 0 for some
507
+ k, see the dashed red curve in Fig.4(a). Compare this with
508
+ the forward KL divergence KL(u ∥ σ), see the solid red
509
+ curve. We propose such zero-avoiding forward version of
510
+ KL divergence as a strong fairness loss
511
+ Lmi++
512
+ :=
513
+ H(σ) + λ KL(u ∥ σ) + γ ∥v∥2. (8)
514
+ We will derive our self-labeling loss directly from (8) using
515
+ standard splitting technique (Boyd & Vandenberghe, 2004)
516
+ to divide optimization of (8) into simpler sub-problems sep-
517
+ arating the “decisiveness” and “fairness” terms, as follows.
518
+ Introducing auxiliary splitting variables y ∈ ∆K, one for
519
+ each training example X, optimization of the loss (8) can
520
+ be equivalently written as
521
+ min
522
+ v,w
523
+ H(σ) + γ ∥v∥2
524
+ (decisiveness sub-problem)
525
+ min
526
+ y
527
+ KL(u ∥ y)
528
+ (fairness sub-problem)
529
+ s.t.
530
+ y = σ
531
+ (consistency constraint).
532
+ This constrained optimization problem can be formulated us-
533
+ ing a Lagrangian function enforcing the equality constraint
534
+ y = σ via the forward KL divergence for y (motivated
535
+ below)
536
+ Lour
537
+ :=
538
+ H(σ) + β KL(σ ∥ y) + λ KL(u ∥ ¯y) + γ ∥v∥2.
539
+ (9)
540
+ The Lagrangian is optimized with respect to both the net-
541
+ work parameters and latent variables y, but we treat the
542
+ Lagrange multiplier β as a fixed hyper-parameter. Thus,
543
+ the constraint y = σ may not be satisfied exactly and the
544
+ Lagrangian (9) works only an approximation of the loss (8).
545
+
546
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
547
+ (a) strong fairness KL(u∥¯σ)
548
+ (b) reverse cross-entropy H(σ, y)
549
+ Figure 4. “Forward” vs “reverse”: (a) KL-divergence and (b) cross-entropy. Assuming binary classification K = 2, we can represent all
550
+ possible probability distributions as points on the interval [0,1]. The solid curves in (a) illustrate our “strong” fairness constraint, i.e.
551
+ the forward KL-divergence KL(u∥¯σ) for the average prediction ¯σ. We show two examples of volumetric prior u1 = (0.9, 0.1) (blue
552
+ curve) and u2 = (0.5, 0.5) (red curve). For comparison, the dashed curves represent reverse KL-divergence KL(¯σ∥u) commonly used
553
+ for fairness in the prior art. The solid curves in (b) show our reverse cross-entropy H(σ, y) w.r.t the network prediction σ. The dashed
554
+ curves show the forward cross-entropy H(y, σ), which is standard in the prior art. The plots in (b) show examples for two fixed estimates
555
+ of pseudo-labels y1 = (0.9, 0.1) (blue curves) and y2 = (0.5, 0.5) (red curves). The boundedness of H(σ, y) represents robustness to
556
+ errors in y. For example, our loss H(σ, y) turns off the training (sets zero-gradients) when the estimated confidence is highly uncertain,
557
+ see y2 = (0.5, 0.5) (solid red). In contrast, the standard loss H(y, σ) trains the network to copy this uncertainty, e.g observe the optimum
558
+ σ for the dashed curves.
559
+ Also, one can justify hyper-parameter β = 1 empirically,
560
+ see Appendix I. Since H(σ) + KL(σ ∥ y) = H(σ ∥ y), we
561
+ get the following self-labeling loss formulation
562
+ Lour
563
+ β = 1
564
+ =
565
+ H(σ, y)
566
+ + λ KL(u ∥ ¯y) + γ ∥v∥2
567
+ (10)
568
+ where the reverse cross entropy H(σ, y) enforces both the
569
+ decisiveness and consistency y ≈ σ.
570
+ There are some notable differences between our loss (10)
571
+ and existing self-labeling losses. For example, consider the
572
+ loss (6) proposed in (Jabi et al., 2021). Our loss reverses the
573
+ order of both the KL divergence and the cross-entropy terms.
574
+ As explained earlier, our version of the KL divergence en-
575
+ forces stronger fairness, see Fig.4(a). The reversal of the
576
+ cross-entropy is motivated in two ways. First, it makes the
577
+ training of network predictions σ robust to errors in noisy
578
+ estimates y, see Figure 4(b), as the pseudo-labels y are not
579
+ guaranteed to be accurate. On the other hand, compared to
580
+ the standard cross-entropy, it enforces stronger consistency
581
+ of y with the predictions σ, which work as target distri-
582
+ butions for y. Thus, w.r.t. pseudo-labels y, our loss (10)
583
+ enforces stronger fairness and stronger consistency y ≈ σ.
584
+ The corresponding well-constrained optimization problem
585
+ for y allows an efficient EM solver derived in Section 2.3.
586
+ 2.3. Our EM algorithm for estimating pseudo-labels
587
+ To optimize (10) with respect to y, basic Newton’s meth-
588
+ ods (Kelley, 1995) can be applied. Although the overall
589
+ convergence rate of such second-order methods is fast, the
590
+ calculation or approximation of the inverse Hessian is com-
591
+ putationally costly as shown in Table 1. This motivates us
592
+ to derive the more efficient expectation-maximization (EM)
593
+ algorithm (Bishop, 2006) for optimizing y as below.
594
+ Here we present a new efficient algorithm for optimizing
595
+ our discriminative entropy-based loss (10) with respect to
596
+ the pseudo-labels y when the model predictions are fixed,
597
+ i.e. σ and v. Using the variational inference (Bishop,
598
+ 2006), we derive a new EM algorithm introducing a dif-
599
+ ferent type of latent variables, K distributions Sk ∈ ∆M
600
+ representing normalized support for each cluster over M
601
+ data points. We refer to each vector Sk as a normalized
602
+ cluster k. Note the difference with distributions represented
603
+ by pseudo-posteriors y ∈ ��K showing support for each
604
+ class at a given data point. Since we explicitly use indi-
605
+ vidual data points below, we will start to carefully index
606
+ them by i ∈ {1, . . . , M}. Thus, we will use yi ∈ ∆K and
607
+ σi ∈ ∆K. Individual components of distribution Sk ∈ ∆M
608
+ corresponding to data point i will be denoted by scalar Sk
609
+ i .
610
+ First, we expand our loss (10) introducing the latent vari-
611
+ ables Sk ∈ ∆M
612
+ Lour
613
+ c=
614
+ H(σ, y) + λ H(u, ¯y) + γ ∥v∥2
615
+ (11)
616
+ =
617
+ H(σ, y) − λ
618
+
619
+ k
620
+ uk ln
621
+
622
+ i
623
+ Sk
624
+ i
625
+ yk
626
+ i
627
+ Sk
628
+ i M + γ ∥v∥2
629
+
630
+ H(σ, y) − λ
631
+
632
+ k
633
+
634
+ i
635
+ ukSk
636
+ i ln
637
+ yk
638
+ i
639
+ Sk
640
+ i M + γ ∥v∥2
641
+ (12)
642
+ Due to the convexity of negative log, we apply Jensen’s
643
+ inequality to derive an upper bound, i.e. (12), to Lour. Such
644
+
645
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
646
+ a bound becomes tight when:
647
+ Estep :
648
+ Sk
649
+ i =
650
+ yk
651
+ i
652
+
653
+ j yk
654
+ j
655
+ (13)
656
+ Then, we fix Sk
657
+ i as (13) and solve the Lagrangian of (12)
658
+ with simplex constraint to update y as:
659
+ Mstep :
660
+ yk
661
+ i =
662
+ σk
663
+ i + λMukSk
664
+ i
665
+ 1 + λM �
666
+ c ucSc
667
+ i
668
+ (14)
669
+ We run these two steps until convergence with respect to
670
+ some predefined tolerance. Note that the minimum y is
671
+ guaranteed to be globally optimal since (11) is convex w.r.t.
672
+ y (Appendix. A). The empirical convergence rate is within
673
+ 15 steps on MNIST. The comparison of computation speed
674
+ on synthetic data is shown in Table 1. While the number
675
+ of iterations to convergence is roughly the same as New-
676
+ ton’s methods, our EM algorithm is much faster in terms
677
+ of running time and is extremely easy to implement using
678
+ the highly optimized built-in functions from the standard
679
+ PyTorch library that supports GPU.
680
+ number of iterations
681
+ running time in sec.
682
+ (to convergence)
683
+ (to convergence)
684
+ K2
685
+ K20
686
+ K200
687
+ K2
688
+ K20
689
+ K200
690
+ Newton
691
+ 3
692
+ 3
693
+ 4
694
+ 2.8e−2
695
+ 3.3e−2
696
+ 1.7e−1
697
+ EM
698
+ 2
699
+ 2
700
+ 2
701
+ 9.9e−4
702
+ 2.0e−3
703
+ 4.0e−3
704
+ Table 1. Comparison of our EM algorithm to Newton’s methods
705
+ (Kelley, 1995). K2, K20 and K200 stand for the number of classes.
706
+ Inspired by (Springenberg, 2015; Hu et al., 2017), we also
707
+ adapted our EM algorithm to allow for updating y within
708
+ each batch. In fact, the mini-batch approximation of (11) is
709
+ an upper bound. Considering the first two terms of (11), we
710
+ can use Jensen’s inequality to get:
711
+ H(σ, y) + λ H(u, ¯y)
712
+
713
+ EB[HB(σ, y) + λ H(u, ¯yB)]
714
+ (15)
715
+ where B is the batch randomly sampled from the whole
716
+ dataset. Now, we can apply our EM algorithm to update
717
+ y in each batch, which is even more efficient. Compared
718
+ to other methods (Ghasedi Dizaji et al., 2017; Asano et al.,
719
+ 2020; Jabi et al., 2021) which also use the auxiliary vari-
720
+ able y, we can efficiently update y on the fly while they
721
+ only update once or just a few times per epoch due to the
722
+ inefficiency to update y for the whole dataset per iteration.
723
+ Interestingly, we found that it is actually important to update
724
+ y on the fly, which makes convergence faster and improves
725
+ the performance significantly (Appendix. C). We use this
726
+ “batch version” EM throughout all the experiments. Our full
727
+ algorithm for the loss (10) is summarized in Appendix. B.
728
+ 3. Experimental results
729
+ Our experiments start from pure clustering on fixed features
730
+ to joint clustering with feature learning. We have also com-
731
+ pared different losses on weakly-supervised classification.
732
+ Note that our goal is comparing different losses together
733
+ with their own optimization algorithms, thus we keeping
734
+ our experimental setup as simple as possible to reduce the
735
+ distraction factors for analysis.
736
+ Dataset
737
+ For the clustering problem, we use four standard
738
+ benchmarks: MNIST (Lecun et al., 1998), CIFAR10/100
739
+ (Torralba et al., 2008) and STL10 (Coates et al., 2011). The
740
+ training and test data are the same. As for the weakly-
741
+ supervised setting, we conduct experiments on CIFAR10
742
+ and STL10. We split the data into training and test sets as
743
+ suggested by the instructions for the datasets.
744
+ Evaluation
745
+ As for the evaluation on clustering, we set the
746
+ number of clusters to the number of ground-truth categories
747
+ and we adopt the standard method (Kuhn, 1955) by finding
748
+ the best one-to-one mapping between clusters and labels.
749
+ We use the accuracy as the measure for both unsupervised
750
+ and weakly-supervised settings while the latter calculates
751
+ the accuracy on the test set.
752
+ 3.1. Clustering with fixed features
753
+ In this section, we test our loss (10) with a simple linear
754
+ classifier on MNIST (Lecun et al., 1998) by using the (fixed)
755
+ original features of the images. We compare it to K-means
756
+ and (4). The detailed experimental settings can be found
757
+ in Appendix. F. In Table 2, we report the mean accuracy
758
+ K-means
759
+ MI (Bridle et al., 1991; Krause et al., 2010)
760
+ Our
761
+ accuracy
762
+ 53.2% (Hu et al., 2017)
763
+ 60.2%(3.7)
764
+ 60.8%(1.1)
765
+ Table 2. Comparison of different losses on MNIST without learn-
766
+ ing features.
767
+ and standard deviation. Note that discriminative clustering
768
+ methods perform consistently much better than K-means
769
+ (≥ 7%) while our approach achieves a bit higher accuracy
770
+ but is more robust. Also, a low-level ablation study can be
771
+ found in Appendix. F.
772
+ STL10
773
+ CIFAR10
774
+ CIFAR100-20
775
+ MNIST
776
+ Kmeans
777
+ 85.20%(5.9)
778
+ 67.78%(4.6)
779
+ 42.99%(1.3)
780
+ 47.62%(2.1)
781
+ MI-GD (Bridle et al., 1991; Krause et al., 2010)
782
+ 89.56%(6.4)
783
+ 72.32%(5.8)
784
+ 43.59%(1.1)
785
+ 52.92%(3.0)
786
+ MI-ADM (Jabi et al., 2021)
787
+ 81.28%(7.2)
788
+ 56.07%(5.5)
789
+ 36.70%(1.1)
790
+ 47.15%(3.7)
791
+ SeLa (Asano et al., 2020)
792
+ 90.33%(4.8)
793
+ 63.31%(3.7)
794
+ 40.74%(1.1)
795
+ 52.38%(5.2)
796
+ Our
797
+ 92.2%(6.2)
798
+ 73.48%(6.2)
799
+ 43.8%(1.1)
800
+ 58.2%(3.1)
801
+ Table 3. Comparison of different methods on clustering with fixed
802
+ features extracted from Resnet-50. The numbers are the average
803
+ accuracy and the standard deviation over trials.
804
+ Besides using low-level features, we also compare our
805
+
806
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
807
+ method against the state-of-the-art methods using fixed deep
808
+ features generated by large models such as Resnet-50 (He
809
+ et al., 2016). We still use a one-layer linear classifier for
810
+ all loss functions except for Kmeans. The coefficients γ for
811
+ the margin maximization terms are set to 0.001, 0.02, 0.009,
812
+ and 0.02 for MNIST, CIFAR10, CIFAR100 and STL10 re-
813
+ spectively. As illustrated in Figure 3, γ is important for
814
+ the optimal decision boundary, especially when features are
815
+ fixed. If we jointly learn the representation and cluster the
816
+ data, we observed that the results are less sensitive to γ.
817
+ Note that this backbone network could be trained together
818
+ with the linear classifier even from the scratch. However, we
819
+ found that the clustering loss itself is not enough to generate
820
+ reasonable features for the backbone network. Thus, we
821
+ keep the backbone network fixed and only train the linear
822
+ classifier using different clustering loss functions.
823
+ 3.2. Joint clustering and representation learning
824
+ In this section, we train a deep network to jointly learn the
825
+ features and cluster the data on the four standard benchmark
826
+ datasets: STL10 (Coates et al., 2011), CIFAR10/CIFAR100
827
+ (Torralba et al., 2008) and MNIST (Lecun et al., 1998).
828
+ The only extra standard technique we add here is the self-
829
+ augmentation, following (Hu et al., 2017; Ji et al., 2019;
830
+ Asano et al., 2020). This technique is important for en-
831
+ forcing neural networks to learn augmentation-invariant fea-
832
+ tures, which are often semantically meaningful. While (Ji
833
+ et al., 2019) designed their loss directly based on such tech-
834
+ nique, our loss and (Krause et al., 2010; Asano et al., 2020;
835
+ Jabi et al., 2021) are more general for clustering without
836
+ any guarantee to generate semantic clusters. Thus, for fair
837
+ comparison and more reasonable results, we combine this
838
+ augmentation technique into network training. The exper-
839
+ imental settings and more detailed discussion are given in
840
+ Appendix. G. From Table 4, it can be seen that our approach
841
+ consistently achieves the best or the most competitive results
842
+ in terms of accuracy.
843
+ STL10
844
+ CIFAR10
845
+ CIFAR100-20
846
+ MNIST
847
+ MI-D⋆ (Hu et al., 2017)
848
+ 25.28%(0.5)
849
+ 21.4%(0.5)
850
+ 14.39%(0.7)
851
+ 92.90%(6.3)
852
+ IIC⋆ (Ji et al., 2019)
853
+ 24.12%(1.7)
854
+ 21.3%(1.4)
855
+ 12.58%(0.6)
856
+ 82.51%(2.3)
857
+ SeLa§ (Asano et al., 2020)
858
+ 23.99%(0.9)
859
+ 24.16%(1.5)
860
+ 15.34%(0.3)
861
+ 52.86%(1.9)
862
+ MI-ADM§ (Jabi et al., 2021)
863
+ 17.37%(0.9)
864
+ 17.27%(0.6)
865
+ 11.02%(0.5)
866
+ 17.75%(1.3)
867
+ Our⋆,§
868
+ 25.33%(1.4)
869
+ 24.16%(0.8)
870
+ 15.09%(0.5)
871
+ 93.58%(4.8)
872
+ Table 4. Quantitative results of accuracy for unsupervised cluster-
873
+ ing methods. We only use the 20 coarse categories for CIFAR100.
874
+ We reuse the code published by (Ji et al., 2019; Asano et al., 2020;
875
+ Hu et al., 2017) and implemented the optimization for loss of (Jabi
876
+ et al., 2021) according to the paper. ⋆: all variables are updated for
877
+ each batch. §: loss formula has pseudo-label.
878
+ Note that we only use a very small network architecture
879
+ (VGG4) here since we observed that more complex archi-
880
+ tectures require more additional techniques to obtain rea-
881
+ sonable results. For example, Ji et.al. (Ji et al., 2019) also
882
+ use auxiliary over-clustering, multiple heads, and more data
883
+ to obtain high numbers on STL10 with ResNet structure.
884
+ To emphasize on the effects of different loss functions, we
885
+ keep the experimental settings as simple as possible.
886
+ 3.3. Weakly-supervised classification
887
+ We also test different methods over different levels of (very)
888
+ weak supervision on STL10. In Table 5, we can see that
889
+ our approach still shows very competitive results, especially
890
+ with weaker supervision. More details are given in Ap-
891
+ pendix. H including another test on CIFAR 10.
892
+ 0.1
893
+ 0.05
894
+ 0.01
895
+ Only seeds
896
+ 40.27%
897
+ 36.26%
898
+ 26.1%
899
+ + MI-D (Hu et al., 2017)
900
+ 47.39%
901
+ 40.73%
902
+ 26.54%
903
+ + IIC (Ji et al., 2019)
904
+ 44.73%
905
+ 33.6%
906
+ 26.17%
907
+ + SeLa (Asano et al., 2020)
908
+ 44.84%
909
+ 36.4%
910
+ 25.08%
911
+ + MI-ADM (Jabi et al., 2021)
912
+ 45.83%
913
+ 40.41%
914
+ 25.79%
915
+ + Our
916
+ 47.20%
917
+ 41.13%
918
+ 26.76%
919
+ Table 5. Quantitative results for weakly-supervised classification
920
+ on STL10. 0.1, 0.05 and 0.01 correspond to different ratios of
921
+ labels used for supervision. “Only seeds” means that we only use
922
+ standard cross-entropy loss on labeled training data.
923
+ 4. Conclusions
924
+ Our paper proposed a new self-labeling algorithm for dis-
925
+ criminative entropy clustering, but we also clarify several
926
+ important conceptual properties of this general methodol-
927
+ ogy. For example, we disproved a theoretical claim in a
928
+ recent TPAMI paper stating the equivalence between vari-
929
+ ance clustering (K-means) and discriminative entropy-based
930
+ clustering. We also demonstrate that standard formulations
931
+ of entropy clustering losses may lead to narrow decision mar-
932
+ gins. Unlike prior work on discriminative entropy clustering,
933
+ we show that classifier norm regularization is important for
934
+ margin maximization.
935
+ We also discussed several limitations of the existing self-
936
+ labeling formulations of entropy clustering and propose
937
+ a new loss addressing such limitations. In particular, we
938
+ replace the standard (forward) cross-entropy by the reverse
939
+ cross-entropy that we show is significantly more robust to
940
+ errors in estimated soft pseudo-labels. Our loss also uses
941
+ a strong formulation of the fairness constraint motivated
942
+ by a zero-avoiding version of KL divergence. Moreover,
943
+ we designed an efficient EM algorithm minimizing our loss
944
+ w.r.t. pseudo-labels; it is significantly faster than standard
945
+ alternatives, e.g Newton’s method. Our empirical results
946
+ improved the state-of-the-art on many standard benchmarks
947
+ for deep clustering.
948
+
949
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
950
+ References
951
+ Asano, Y. M., Rupprecht, C., and Vedaldi, A. Self-labelling
952
+ via simultaneous clustering and representation learning.
953
+ In International Conference on Learning Representations,
954
+ 2020.
955
+ Bishop, C. M. Pattern Recognition and Machine Learning.
956
+ Springer, 2006.
957
+ Boyd, S. and Vandenberghe, L. Convex optimization. Cam-
958
+ bridge university press, 2004.
959
+ Bridle, J. S., Heading, A. J. R., and MacKay, D. J. C. Un-
960
+ supervised classifiers, mutual information and ’phantom
961
+ targets’. In NIPS, pp. 1096–1101, 1991.
962
+ Coates, A., Ng, A., and Lee, H. An analysis of single-
963
+ layer networks in unsupervised feature learning. In Pro-
964
+ ceedings of the fourteenth international conference on
965
+ artificial intelligence and statistics, pp. 215–223. JMLR
966
+ Workshop and Conference Proceedings, 2011.
967
+ Cuturi, M. Sinkhorn distances: Lightspeed computation
968
+ of optimal transport. Advances in neural information
969
+ processing systems, 26, 2013.
970
+ Ghasedi Dizaji, K., Herandi, A., Deng, C., Cai, W., and
971
+ Huang, H. Deep clustering via joint convolutional au-
972
+ toencoder embedding and relative entropy minimization.
973
+ In Proceedings of the IEEE international conference on
974
+ computer vision, pp. 5736–5745, 2017.
975
+ Grandvalet, Y. and Bengio, Y. Semi-supervised learning by
976
+ entropy minimization. Advances in neural information
977
+ processing systems, 17, 2004.
978
+ Guo, C., Pleiss, G., Sun, Y., and Weinberger, K. Q. On
979
+ calibration of modern neural networks. In International
980
+ conference on machine learning, pp. 1321–1330. PMLR,
981
+ 2017.
982
+ He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learn-
983
+ ing for image recognition. In Proceedings of the IEEE
984
+ conference on computer vision and pattern recognition,
985
+ pp. 770–778, 2016.
986
+ Hu, W., Miyato, T., Tokui, S., Matsumoto, E., and Sugiyama,
987
+ M. Learning discrete representations via information
988
+ maximizing self-augmented training. In International
989
+ conference on machine learning, pp. 1558–1567. PMLR,
990
+ 2017.
991
+ Jabi, M., Pedersoli, M., Mitiche, A., and Ayed, I. B. Deep
992
+ clustering: On the link between discriminative models
993
+ and k-means. IEEE Transactions on Pattern Analysis and
994
+ Machine Intelligence, 43(6):1887–1896, 2021.
995
+ Ji, X., Henriques, J. F., and Vedaldi, A. Invariant informa-
996
+ tion clustering for unsupervised image classification and
997
+ segmentation. In Proceedings of the IEEE/CVF Interna-
998
+ tional Conference on Computer Vision, pp. 9865–9874,
999
+ 2019.
1000
+ Kearns, M., Mansour, Y., and Ng, A. Y. An information-
1001
+ theoretic analysis of hard and soft assignment methods
1002
+ for clustering.
1003
+ In UAI ’97: Proceedings of the Thir-
1004
+ teenth Conference on Uncertainty in Artificial Intelli-
1005
+ gence, Brown University, Providence, Rhode Island, USA,
1006
+ August 1-3, 1997, pp. 282–293. Morgan Kaufmann, 1997.
1007
+ Kelley, C. T. Iterative methods for linear and nonlinear
1008
+ equations. SIAM, 1995.
1009
+ Kingma, D. P. and Ba, J. Adam: A method for stochastic
1010
+ optimization. In ICLR (Poster), 2015.
1011
+ Krause, A., Perona, P., and Gomes, R. Discriminative clus-
1012
+ tering by regularized information maximization.
1013
+ Ad-
1014
+ vances in neural information processing systems, 23,
1015
+ 2010.
1016
+ Kuhn, H. W. The hungarian method for the assignment
1017
+ problem. Naval research logistics quarterly, 2(1-2):83–
1018
+ 97, 1955.
1019
+ Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P. Gradient-
1020
+ based learning applied to document recognition. Proceed-
1021
+ ings of the IEEE, 86(11):2278–2324, 1998.
1022
+ M¨uller, R., Kornblith, S., and Hinton, G. E. When does
1023
+ label smoothing help? Advances in neural information
1024
+ processing systems, 32, 2019.
1025
+ Pereyra, G., Tucker, G., Chorowski, J., Kaiser, L., and
1026
+ Hinton, G. Regularizing neural networks by penalizing
1027
+ confident output distributions. 2017.
1028
+ Rumelhart, D. E., Hinton, G. E., and Williams, R. J. Learn-
1029
+ ing representations by back-propagating errors. Nature,
1030
+ 323(6088):533–536, 1986.
1031
+ Song, H., Kim, M., Park, D., Shin, Y., and Lee, J.-G. Learn-
1032
+ ing from noisy labels with deep neural networks: A sur-
1033
+ vey. IEEE Transactions on Neural Networks and Learn-
1034
+ ing Systems, 2022.
1035
+ Soudry, D., Hoffer, E., Nacson, M. S., Gunasekar, S., and
1036
+ Srebro, N. The implicit bias of gradient descent on sepa-
1037
+ rable data. The Journal of Machine Learning Research,
1038
+ 19(1):2822–2878, 2018.
1039
+ Springenberg, J. T. Unsupervised and semi-supervised learn-
1040
+ ing with categorical generative adversarial networks. In
1041
+ International Conference on Learning Representations,
1042
+ 2015.
1043
+
1044
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
1045
+ Tanaka, D., Ikami, D., Yamasaki, T., and Aizawa, K. Joint
1046
+ optimization framework for learning with noisy labels. In
1047
+ Proceedings of the IEEE conference on computer vision
1048
+ and pattern recognition, pp. 5552–5560, 2018.
1049
+ Torralba, A., Fergus, R., and Freeman, W. T. 80 million tiny
1050
+ images: A large data set for nonparametric object and
1051
+ scene recognition. IEEE transactions on pattern analysis
1052
+ and machine intelligence, 30(11):1958–1970, 2008.
1053
+ Xu, L., Neufeld, J., Larson, B., and Schuurmans, D. Maxi-
1054
+ mum margin clustering. In Saul, L., Weiss, Y., and Bottou,
1055
+ L. (eds.), Advances in Neural Information Processing Sys-
1056
+ tems, volume 17. MIT Press, 2004.
1057
+
1058
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
1059
+ A. Proof
1060
+ Lemma A.1. Given fixed σi ∈ ∆K where i ∈ {1, ..., M}
1061
+ and u ∈ ∆K, the objective
1062
+ E(y) = − β
1063
+ M
1064
+
1065
+ i
1066
+
1067
+ k
1068
+ σk
1069
+ i ln yk
1070
+ i − λ
1071
+
1072
+ k
1073
+ uk ln
1074
+
1075
+ i yk
1076
+ i
1077
+ M
1078
+ is convex for y, where yi ∈ ∆K.
1079
+ Proof. First, we rewrite E(y)
1080
+ E(y) = −
1081
+
1082
+ k
1083
+
1084
+ β
1085
+ M
1086
+
1087
+ i
1088
+ σk
1089
+ i ln yk
1090
+ i + λuk ln
1091
+
1092
+ i yk
1093
+ i
1094
+ M
1095
+
1096
+ := −
1097
+
1098
+ k
1099
+ fk(yk)
1100
+ (16)
1101
+ Next, we prove that fk : RM
1102
+ (0,1) → R is concave based on
1103
+ the definition of concavity(Boyd & Vandenberghe, 2004)
1104
+ for any k ∈ {1, ..., K}. Considering x = (1 − α)x1 + αx2
1105
+ where x1, x2 ∈ RM
1106
+ (0,1) and α ∈ [0, 1], we have
1107
+ fk(x) =
1108
+ β
1109
+ M
1110
+
1111
+ i
1112
+ σk
1113
+ i ln ((1 − α)x1i + αx2i)+
1114
+ λuk ln
1115
+
1116
+ i ((1 − α)x1i + αx2i)
1117
+ M
1118
+ ≥ β
1119
+ M
1120
+
1121
+ i
1122
+ (1 − α)σk
1123
+ i ln x1i + ασk
1124
+ i ln x2i
1125
+ + λuk
1126
+
1127
+ (1 − α) ln
1128
+
1129
+ i x1i
1130
+ M
1131
+ + α ln
1132
+
1133
+ i x2i
1134
+ M
1135
+
1136
+ = (1 − α)fk(x1) + αfk(x2)
1137
+ The inequality uses Jensen’s inequality. Now that fk is
1138
+ proved to be concave, −fk will be convex. Then E(y)
1139
+ can be easily proved to be convex using the definition of
1140
+ convexity with the similar steps above.
1141
+ B. Our Algorithm
1142
+ C. Loss Curve
1143
+ D. Network Architecture
1144
+ The network structure is VGG-style and adapted from (Ji
1145
+ et al., 2019).
1146
+ E. Dataset Summary
1147
+ Table 7 indicates the number of (training) data and the input
1148
+ size of each image for the unsupervised clustering. Training
1149
+ and test sets are the same.
1150
+ As for weakly-supervised classification on STL10, we use
1151
+ 5000 images for training and 8000 images for testing. We
1152
+ Algorithm 1 Optimization for our loss
1153
+ Input
1154
+ :network parameters [v, w] and dataset
1155
+ Output :network parameters [v∗, w∗]
1156
+ for each epoch do
1157
+ for each iteration do
1158
+ Initialize y by the network output at current stage as
1159
+ a warm start while not convergent do
1160
+ Sk
1161
+ i =
1162
+ yk
1163
+ i
1164
+
1165
+ j yk
1166
+ j
1167
+ yk
1168
+ i =
1169
+ σk
1170
+ i +λMukSk
1171
+ i
1172
+ 1+λM �
1173
+ c ucSc
1174
+ i
1175
+ end
1176
+ Update [w, v] using loss HB(σ, y) + γ ∥v∥2 via
1177
+ stochastic gradient descent
1178
+ end
1179
+ end
1180
+ Figure 5. Loss (10) curves for different update setting on y. This
1181
+ is generated with just a linear classifier on MNIST. We use the
1182
+ same initialization and run both for 50 epochs. The gray line has
1183
+ an accuracy of 52.35% while the yellow one achieves 63%.
1184
+ only keep a certain percentage of ground-truth labels for
1185
+ each class of training data. The accuracy is calculated on
1186
+ test set by comparing the hard-max of prediction to the
1187
+ ground-truth.
1188
+ F. Low-level Clustering
1189
+ As for the experiments on MNIST (Lecun et al., 1998), we
1190
+ transform the original image values linearly into [−1, 1] and
1191
+ use the flattened images as input features. Note that here
1192
+ we only use a linear classifier without training any features.
1193
+ We employ stochastic gradient descent with learning rate
1194
+ 0.07 to update v in (4) and (10). We use the same (random)
1195
+ intialization for both losses and run each 6 times up to 50
1196
+ epochs per run. We use 250 for batch size. We set γ = 0.01
1197
+ for both and use λ = 100 for (10) and λ = 1.3 for (4).
1198
+ We fix the hyperparameter values for (9) and (4)
1199
+ throughout the whole experimental sections.
1200
+ We also conducted an ablation study on toy examples as
1201
+ shown in Figure. 6. We use the normalized X-Y coordinates
1202
+ of the data points as the input. We can see that each part
1203
+ of our loss is necessary for obtaining a good result. Note
1204
+
1205
+ loss:
1206
+ update y on whole dataset once per epoch
1207
+ update y on batch data per iteration
1208
+ 232
1209
+ 231.5
1210
+ 231
1211
+ Iteration
1212
+ 0
1213
+ 2k
1214
+ 4k
1215
+ 6k
1216
+ 8k
1217
+ 10k
1218
+ 12kRevisiting Discriminative Entropy Clustering and its relation to K-means
1219
+ Grey(28x28x1)
1220
+ RGB(32x32x3)
1221
+ RGB(96x96x3)
1222
+ 1xConv(5x5,s=1,p=2)@64
1223
+ 1xConv(5x5,s=1,p=2)@32
1224
+ 1xConv(5x5,s=2,p=2)@128
1225
+ 1xMaxPool(2x2,s=2)
1226
+ 1xMaxPool(2x2,s=2)
1227
+ 1xMaxPool(2x2,s=2)
1228
+ 1xConv(5x5,s=1,p=2)@128
1229
+ 1xConv(5x5,s=1,p=2)@64
1230
+ 1xConv(5x5,s=2,p=2)@256
1231
+ 1xMaxPool(2x2,s=2)
1232
+ 1xMaxPool(2x2,s=2)
1233
+ 1xMaxPool(2x2,s=2)
1234
+ 1xConv(5x5,s=1,p=2)@256
1235
+ 1xConv(5x5,s=1,p=2)@128
1236
+ 1xConv(5x5,s=2,p=2)@512
1237
+ 1xMaxPool(2x2,s=2)
1238
+ 1xMaxPool(2x2,s=2)
1239
+ 1xMaxPool(2x2,s=2)
1240
+ 1xConv(5x5,s=1,p=2)@512
1241
+ 1xConv(5x5,s=1,p=2)@256
1242
+ 1xConv(5x5,s=2,p=2)@1024
1243
+ 1xLinear(512x3x3,K)
1244
+ 1xLinear(256x4x4,K)
1245
+ 1xLinear(1024x1x1,K)
1246
+ Table 6. Network architecture summary. s: stride; p: padding; K:
1247
+ number of clusters. The first column is used on MNIST (Lecun
1248
+ et al., 1998); the second one is used on CIFAR10/100 (Torralba
1249
+ et al., 2008); the third one is used on STL10 (Coates et al., 2011).
1250
+ Batch normalization is also applied after each Conv layer. ReLu is
1251
+ adopted for non-linear activation function.
1252
+ STL10
1253
+ CIFAR10
1254
+ CIFAR100-20
1255
+ MNIST
1256
+ 13000
1257
+ 60000
1258
+ 60000
1259
+ 70000
1260
+ 96x96x3
1261
+ 32x32x3
1262
+ 32x32x3
1263
+ 28x28x1
1264
+ Table 7. Dataset summary for unsupervised clustering.
1265
+ that, in Figure 6 (a), (c) of 3-label case, the clusters formed
1266
+ are the same, but the decision boundaries which implies the
1267
+ generalization are different. This emphasizes the importance
1268
+ of including L2 norm of v to enforce maximum margin for
1269
+ better generalization.
1270
+ 2 clusters
1271
+ 3 clusters
1272
+ (a) γ = 0
1273
+ (b) λ = 0
1274
+ (c) full setting
1275
+ Figure 6. “Shallow” ablation study on toy examples.
1276
+ G. Deep Clustering
1277
+ We add deep neural networks for learning features while
1278
+ doing the clustering simultaneously.
1279
+ We use four stan-
1280
+ dard benchmark datasets: STL10 (Coates et al., 2011), CI-
1281
+ FAR10/CIFAR100 (Torralba et al., 2008) and MNIST (Le-
1282
+ cun et al., 1998). As for the architectures, we followed (Ji
1283
+ et al., 2019) to use VGG11-like network structures whereas
1284
+ we use it for both gray-scale and RGB images with some
1285
+ adjustments as shown in Appendix. D.
1286
+ We achieved the self-augmentation by setting σi
1287
+ =
1288
+ Et[σ(v⊤fw(t(Xi))]. For each image, we generate two aug-
1289
+ mentations sampled from “horizontal flip”, “rotation” and
1290
+ “color distortion”.
1291
+ We use Adam (Kingma & Ba, 2015) with learning rate 1e−4
1292
+ for optimizing the network parameters. We set batch size to
1293
+ 250 for CIFAR10, CIFAR100 and MNIST, and we use 160
1294
+ for STL10. In Table 4, we report the mean accuracy and Std
1295
+ from 6 runs with different initializations while we use the
1296
+ same initialization for all methods in each run. We still use
1297
+ 50 epochs for each run and all methods reach convergence
1298
+ within 50 epochs.
1299
+ As for other methods in Table 4, MI-D has the most com-
1300
+ parable results to us, in part because our loss can be seen
1301
+ as an approximation to the MI and we both update all vari-
1302
+ ables per batch. SeLa achieves relatively better results on
1303
+ other three datasets than MNIST, because it enforces a hard
1304
+ constraint on “fairness” and MNIST is the only one out of
1305
+ four sets that is not exactly balanced. In real world, the data
1306
+ we collect is mostly not exactly balanced. This could be the
1307
+ reason why such method is better for the unsupervised rep-
1308
+ resentation learning where over-clustering can be employed
1309
+ and real clusters become less important. MI-ADM only up-
1310
+ dates the pseudo-labels once per epoch, thus easily leading
1311
+ the network towards a trap of local minimum created by the
1312
+ incorrect pseudo-labels through the forward cross-entropy
1313
+ loss as illustrated in Figure 4.
1314
+ H. Weakly-supervised Clustering
1315
+ We use the same experimental settings as that in unsuper-
1316
+ vised clustering except for two points: 1. We add cross-
1317
+ entropy loss on labelled data; 2. We separate the training
1318
+ data from test data while we use all the data for training and
1319
+ test in unsupervised clustering.
1320
+ While MI-ADM is the worst according to Table 4, it is
1321
+ improved significantly in weakly-supervised setting. This
1322
+ might be a sign that the advantage of more frequent update
1323
+ on variables in unsupervised clustering is waning since the
1324
+ seeds help the network keeping away from some bad local
1325
+ minima.
1326
+ Below is the result on CIFAR 10.
1327
+ 0.1
1328
+ 0.05
1329
+ 0.01
1330
+ Only seeds
1331
+ 58.77%
1332
+ 54.27%
1333
+ 39.01%
1334
+ + MI-D
1335
+ 65.54%
1336
+ 61.4%
1337
+ 46.97%
1338
+ + IIC
1339
+ 66.5%
1340
+ 61.17%
1341
+ 47.21%
1342
+ + SeLa
1343
+ 61.5%
1344
+ 58.35%
1345
+ 47.19%
1346
+ + MI-ADM
1347
+ 62.51%
1348
+ 57.05%
1349
+ 45.91%
1350
+ + Our
1351
+ 66.17%
1352
+ 61.59%
1353
+ 47.22%
1354
+
1355
+ Revisiting Discriminative Entropy Clustering and its relation to K-means
1356
+ I. Hyperparameter β
1357
+ Below is an empirical justification for setting hyper-
1358
+ parameter β = 1 in the loss (9). The first two terms in
1359
+ (9) can be written as (1 − β)H(σ) + βH(σ, y). If β > 1
1360
+ then the negative entropy pushes predictions σ away from
1361
+ one-hot solutions weakening the decisiveness. On the other
1362
+ hand, if β < 1 then the loss is non-convex w.r.t σ that may
1363
+ trap gradient descent in bad local minima, as illustrated by
1364
+ the plots for y = (0.9, 0.1) in Figure 7
1365
+ Figure 7. (1 − β)H(σ) + βH(σ, y)
1366
+
-tFIT4oBgHgl3EQf9Cv1/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
.gitattributes CHANGED
@@ -7217,3 +7217,48 @@ PdFQT4oBgHgl3EQfYjZ5/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -tex
7217
  F9AzT4oBgHgl3EQfHPuf/content/2301.01042v1.pdf filter=lfs diff=lfs merge=lfs -text
7218
  wNFJT4oBgHgl3EQffyxP/content/2301.11558v1.pdf filter=lfs diff=lfs merge=lfs -text
7219
  b9AyT4oBgHgl3EQfwfkG/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7217
  F9AzT4oBgHgl3EQfHPuf/content/2301.01042v1.pdf filter=lfs diff=lfs merge=lfs -text
7218
  wNFJT4oBgHgl3EQffyxP/content/2301.11558v1.pdf filter=lfs diff=lfs merge=lfs -text
7219
  b9AyT4oBgHgl3EQfwfkG/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7220
+ FNAzT4oBgHgl3EQfw_7Q/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7221
+ aNE2T4oBgHgl3EQfvgjw/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7222
+ StAzT4oBgHgl3EQfXfxy/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7223
+ RtE2T4oBgHgl3EQfBwaC/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7224
+ gtAzT4oBgHgl3EQf4P6f/content/2301.01842v1.pdf filter=lfs diff=lfs merge=lfs -text
7225
+ PtAyT4oBgHgl3EQfUvdy/content/2301.00131v1.pdf filter=lfs diff=lfs merge=lfs -text
7226
+ rNAzT4oBgHgl3EQfA_pb/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7227
+ gdFJT4oBgHgl3EQfUyyg/content/2301.11510v1.pdf filter=lfs diff=lfs merge=lfs -text
7228
+ ntE4T4oBgHgl3EQfug2d/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7229
+ S9E4T4oBgHgl3EQfmA05/content/2301.05164v1.pdf filter=lfs diff=lfs merge=lfs -text
7230
+ 1NAzT4oBgHgl3EQfRftD/content/2301.01216v1.pdf filter=lfs diff=lfs merge=lfs -text
7231
+ ntE4T4oBgHgl3EQfug2d/content/2301.05233v1.pdf filter=lfs diff=lfs merge=lfs -text
7232
+ mtFST4oBgHgl3EQfKjji/content/2301.13737v1.pdf filter=lfs diff=lfs merge=lfs -text
7233
+ RtAyT4oBgHgl3EQf7_pZ/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7234
+ FNAzT4oBgHgl3EQfw_7Q/content/2301.01732v1.pdf filter=lfs diff=lfs merge=lfs -text
7235
+ 1NAzT4oBgHgl3EQfRftD/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7236
+ fdE4T4oBgHgl3EQfqg0X/content/2301.05200v1.pdf filter=lfs diff=lfs merge=lfs -text
7237
+ aNE0T4oBgHgl3EQf4QKS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7238
+ 4dAyT4oBgHgl3EQf2Pkf/content/2301.00746v1.pdf filter=lfs diff=lfs merge=lfs -text
7239
+ F9AzT4oBgHgl3EQfHPuf/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7240
+ ftE1T4oBgHgl3EQfMAOS/content/2301.02984v1.pdf filter=lfs diff=lfs merge=lfs -text
7241
+ GNE4T4oBgHgl3EQfgA3B/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7242
+ pdAzT4oBgHgl3EQfOvsC/content/2301.01169v1.pdf filter=lfs diff=lfs merge=lfs -text
7243
+ m9E0T4oBgHgl3EQfZQCv/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7244
+ rNE1T4oBgHgl3EQfPwMX/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7245
+ ItE1T4oBgHgl3EQfYARl/content/2301.03133v1.pdf filter=lfs diff=lfs merge=lfs -text
7246
+ gdFJT4oBgHgl3EQfUyyg/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7247
+ 09FKT4oBgHgl3EQfOS13/content/2301.11758v1.pdf filter=lfs diff=lfs merge=lfs -text
7248
+ 3tAzT4oBgHgl3EQfD_po/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7249
+ 4tE4T4oBgHgl3EQf1A0X/content/2301.05286v1.pdf filter=lfs diff=lfs merge=lfs -text
7250
+ 3tAzT4oBgHgl3EQfD_po/content/2301.00985v1.pdf filter=lfs diff=lfs merge=lfs -text
7251
+ 0NFRT4oBgHgl3EQfkTex/content/2301.13595v1.pdf filter=lfs diff=lfs merge=lfs -text
7252
+ CdFAT4oBgHgl3EQftB4M/content/2301.08661v1.pdf filter=lfs diff=lfs merge=lfs -text
7253
+ SNFRT4oBgHgl3EQf8zg4/content/2301.13685v1.pdf filter=lfs diff=lfs merge=lfs -text
7254
+ SNFRT4oBgHgl3EQf8zg4/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7255
+ Q9FJT4oBgHgl3EQf3C2V/content/2301.11659v1.pdf filter=lfs diff=lfs merge=lfs -text
7256
+ ftE1T4oBgHgl3EQfMAOS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7257
+ mtFST4oBgHgl3EQfKjji/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7258
+ PtAyT4oBgHgl3EQfUvdy/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7259
+ TdAyT4oBgHgl3EQf8foB/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7260
+ utAzT4oBgHgl3EQfBvrw/content/2301.00949v1.pdf filter=lfs diff=lfs merge=lfs -text
7261
+ cNE2T4oBgHgl3EQfwgg2/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
7262
+ N9E3T4oBgHgl3EQfZQq9/content/2301.04496v1.pdf filter=lfs diff=lfs merge=lfs -text
7263
+ TdAyT4oBgHgl3EQf8foB/content/2301.00855v1.pdf filter=lfs diff=lfs merge=lfs -text
7264
+ StAzT4oBgHgl3EQfXfxy/content/2301.01319v1.pdf filter=lfs diff=lfs merge=lfs -text
09FKT4oBgHgl3EQfOS13/content/2301.11758v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0922c4e9a11d71c7af85dd41e93869c72f2d8513c711f7a6b6758074c578ce89
3
+ size 15995704
0NFRT4oBgHgl3EQfkTex/content/2301.13595v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa85e492c73d1e483dfcbb5dae56ed1ee273a71ab831d583a5b0c38d0b926e1a
3
+ size 944175
0NFRT4oBgHgl3EQfkTex/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f58c3c09f9795cf32bea1e55b5ce362870d0c709ece8eced9099602f7f31d767
3
+ size 64899
0tAyT4oBgHgl3EQfoPjL/content/tmp_files/2301.00505v1.pdf.txt ADDED
@@ -0,0 +1,419 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ PokAR: Facilitating Poker Play Through Augmented Reality
2
+ ADAM GAMBA and ANDRÉS MONROY-HERNÁNDEZ, Princeton University, USA
3
+ Fig. 1. Two players using the PokAR application.
4
+ We introduce PokAR, an augmented reality (AR) application to facilitate poker play. PokAR aims to alleviate three difficulties of
5
+ traditional poker by leveraging AR technology: (1) need to have physical poker chips, (2) complex rules of poker, (3) slow game pace
6
+ caused by laborious tasks. Despite the potential benefits of AR in poker, not much research has been done in the field. In fact, PokAR is
7
+ the first application to enable AR poker on a mobile device without requiring extra costly equipment. This has been done by creating
8
+ a Snapchat Lens 1 which can be used on most mobile devices. We evaluated this application by instructing 4 participant dyads to
9
+ use PokAR to engage in poker play and respond to survey questions about their experience. We found that most PokAR features
10
+ were positively received, AR did not significantly improve nor hinder socialization, PokAR slightly increased the game pace, and
11
+ participants had an overall enjoyable experience with the Lens. These findings led to three major conclusions: (1) AR has the potential
12
+ to augment and simplify traditional table games, (2) AR should not be used to replace traditional experiences, only augment them, (3)
13
+ Future work includes additional features like increased tactility and statistical annotations.
14
+ CCS Concepts: • Human-centered computing → Collaborative and social computing devices.
15
+ Additional Key Words and Phrases: connected lens, augmented reality, poker, co-located, interaction, socialization
16
+ ACM Reference Format:
17
+ Adam Gamba and Andrés Monroy-Hernández. 2023. PokAR: Facilitating Poker Play Through Augmented Reality. 1, 1 (January 2023),
18
+ 11 pages. https://doi.org/XXXXXXX.XXXXXXX
19
+ 1A Lens in Snapchat is an experience that utilizes augmented reality to transform the world around you [12].
20
+ Authors’ address: Adam Gamba, agamba@princeton.edu; Andrés Monroy-Hernández, andresmh@princeton.edu, Princeton University, Princeton, New
21
+ Jersey, USA, 08544.
22
+ Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not
23
+ made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components
24
+ of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to
25
+ redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org.
26
+ © 2023 Association for Computing Machinery.
27
+ Manuscript submitted to ACM
28
+ Manuscript submitted to ACM
29
+ 1
30
+ arXiv:2301.00505v1 [cs.HC] 2 Jan 2023
31
+
32
+ l:61:598
33
+ old2
34
+ Gamba and Monroy-Hernández
35
+ 1
36
+ INTRODUCTION
37
+ The goal of this project is to facilitate heads-up Texas hold’em poker play through augmented reality. Poker is
38
+ cumbersome to play in its current form, requiring players to have poker chips and knowledge of the complex rules to
39
+ play correctly. Without an experienced player to guide the game, new players often find it difficult to learn the rules
40
+ and play correctly [9]. Additionally, due to the burden of physical chips, it is difficult to play poker in many scenarios
41
+ (e.g., at the beach, while traveling, or camping). Finally, the game pace is often slowed due to poker’s complex rules and
42
+ the need for laborious tasks like counting chip stacks.
43
+ Augmented reality technology is well-equipped to solve these issues in three ways. Firstly, AR can eliminate the need
44
+ for physical poker chips by instead utilizing AR to render chips. Next, AR can help guide players through the complex
45
+ rules of poker by hinting at legal actions during gameplay. Finally, AR can help decrease the burden of laborious tasks
46
+ (like counting chips) and increase the game pace. PokAR helps alleviate these three issues, which we’ll discuss further
47
+ throughout this paper.
48
+ Poker is a popular game, with over 120 million players worldwide playing regularly online [7]. Texas hold’em is one
49
+ of the most popular poker variants. In this variant, players are dealt two private cards and five community cards, and
50
+ they battle to make the best hand or bluff opponents into folding. ’Heads-up’ poker is a term used to describe poker
51
+ played by just two players, head-to-head. In its current state, PokAR supports only heads-up Texas hold’em poker, but
52
+ with future work, it could be extended to more players and more variants. Throughout this paper, we will use the term
53
+ ’poker’ to refer to heads-up Texas hold’em poker.
54
+ Poker is a classic example of a social, co-located game, since poker, by design, emphasizes in-person, co-located
55
+ interaction. Players often look at each other and speak to each other during a poker game, either to gain information or
56
+ to socialize. Also, poker forces players to focus on the same enablers, or "physical objects that trigger and are the focus
57
+ of the AR experience" [1]. These enablers, like playing cards and poker chips, can help guide an AR experience and
58
+ engage players more closely than in games that do not have a similar shared focus. For the above reasons, we chose to
59
+ augment poker in this study.
60
+ PokAR is not intended to replace traditional poker, rather it helps people play when traditional poker would be
61
+ difficult or impractical. The goal of AR applications should be to disappear completely and seamlessly immerse the user
62
+ in a realistic experience that combines reality with augmentation [16]. This disappearance frees users to utilize these
63
+ applications more effortlessly, allowing them to focus on new goals, beyond the application itself.
64
+ 2
65
+ RELATED WORK
66
+ While we have established AR as a possible solution to the aforementioned issues with poker, very little work has
67
+ been done concerning AR poker. Additionally, while AR is not yet a heavily explored area, researchers have argued that
68
+ “A Poker-Assistance-Software is an ideal test area for an AR Application with real added value,” with possible areas to
69
+ add value including automation and statistical estimations [15].
70
+ Similar projects in the past have all relied on physical means to augment reality. For example, researchers used
71
+ overhead projectors to project all aspects of the poker game (e.g., cards, chips, etc.) onto a table [9]. Additionally,
72
+ researchers have used RFID playing cards to detect the dealt cards [9]. While this study succeeded in creating an AR
73
+ application to ease some of the same cumbersome aspects of poker tackled by PokAR (physical chips, complex rules,
74
+ slow game pace), it did so using a high-cost solution, which is impractical for most recreational use. Additionally, they
75
+ disregarded studying how this AR setup influenced social interactions in poker.
76
+ Manuscript submitted to ACM
77
+
78
+ PokAR: Facilitating Poker Play Through Augmented Reality
79
+ 3
80
+ Furthermore, people have utilized virtual reality (VR) in the past to create commercial poker video games. One
81
+ example is PokerVR by Meta, which uses “expressive avatars built for reading tells with growing customizations” [6].
82
+ While they may say this, VR poker applications still just employ static players’ avatars, which do not emphasize the
83
+ in-person, social nature of poker.
84
+ PokAR is the first project to enable AR poker without needing additional equipment other than a mobile device and
85
+ a regular deck of cards (like an overhead projector or a VR headset). This is a worthwhile problem because it is the
86
+ first application to enable AR poker at a low cost, since it only requires a few commonly-owned pieces of equipment
87
+ (mobile devices and playing cards). Additionally, utilizing AR over VR allows the gameplay to emphasize the social
88
+ aspects of co-location and increase socialization when compared to VR implementations.
89
+ Co-located gaming has been shown to lead to more effective and enjoyable gaming, as players can more easily
90
+ communicate and build social relationships [4]. One major reason for this is "out-of-the-game, game-related communi-
91
+ cation" [3]. By having the ability to converse about other topics while simultaneously being involved in a game with
92
+ another player, these players are given the opportunity to build a deeper connection.
93
+ Additionally, when comparing socialization in AR and VR, prior research tends to support increased socialization
94
+ in AR applications. AR games have the ability to "potentially enhance social communication and social interaction
95
+ between people" [10], whereas high-involvement in VR games could potentially isolate users socially and "negatively
96
+ affect their well-being" [5]. Thus, we chose to develop an AR application rather than a VR application to reap the social
97
+ benefits of shared, co-located experiences.
98
+ 3
99
+ POKAR SYSTEM
100
+ The PokAR Snapchat Lens allows users to play heads-up poker with another player on two mobile devices. AR
101
+ visual annotations include 3D models of poker chips which dynamically render with changing stack size, and 3D text
102
+ above the chip stacks denoting the size of each stack. 2D visual annotations include number of hands played, current
103
+ round, current dealer, previous action, amount to call, waiting message, and UI buttons with labels "Check," "Call," "Bet,"
104
+ "Raise," and "Fold." All AR and 2D annotations render dynamically with changing game state, stack amounts, and legal
105
+ actions. PokAR implements five main features to help achieve its motivating goals.
106
+ • “3D AR Chips” - PokAR renders 3D models of chips to eliminate the requirement of needing physical poker
107
+ chips.
108
+ • “UI Action Buttons” - 2D buttons rendered on the player’s screen allows them to select among and perform legal
109
+ actions at the current state of the game.
110
+ • “Game Messages” - Messages provide additional information to both players about the actions of players, bet
111
+ amounts, and more throughout the game.
112
+ • “Counting Stacks” - A live count of the number of chips in all chip stacks is rendered above the 3D models,
113
+ eliminating the hassle of counting chips manually.
114
+ • “Awarding Pots” - After a winner is determined (through folding or at showdown), the chips are automatically
115
+ awarded to the winner, eliminating the hassle of moving chips manually.
116
+ 3.1
117
+ Approach
118
+ To achieve our motivating goals, we developed a Snapchat Lens which could be used on any mobile device to
119
+ facilitate poker play through AR. Besides having physical playing cards, players will play the game of poker in AR
120
+ Manuscript submitted to ACM
121
+
122
+ 4
123
+ Gamba and Monroy-Hernández
124
+ Fig. 2. Side-by-side points of view of the same game of PokAR on two mobile devices.
125
+ by interacting with their augmented environment through a mobile device. We used the Snap Lens Studio IDE with
126
+ JavaScript for development, the Snapchat app for testing and deployment, and GitHub for version control. Code for this
127
+ project can be found at the link in Appendix D. Physical playing cards act as an enabler to ground the game in physical
128
+ reality. A demonstration of the completed application is shown in Fig 2. In the development of PokAR, we faced and
129
+ solved three major implementation subproblems. They will be discussed below.
130
+ 3.2
131
+ Subproblem 1: Modeling Poker in Code
132
+ The first implementation subproblem to solve was to figure out how to model a game of poker in code. By nature
133
+ of the rules of poker, the game is deterministic based on previous player actions within a betting round. Thus, the
134
+ game state can be modeled using a Deterministic Finite Automaton (DFA). The DFA for our application determines the
135
+ legal actions and/or termination state of the betting round, given previous actions within the betting round. At the
136
+ end of each betting round, one of two termination states is reached, dictating whether the hand has ended or players
137
+ will advance to the next betting round. Fig 3 and Fig 4 show a graphical representation of the DFAs we used in our
138
+ implementation. In both DFAs, the application begins with the Start state and terminates in either the endHand() or
139
+ advance() state. The double-headed arrow represents the possible cycle of betting, raising, reraising, etc., until one of the
140
+ players is eventually all-in. Player ’A’ is the one assigned ‘opponent’ at the start of a hand, and player ’B’ is the ‘dealer.’
141
+ 3.3
142
+ Subproblem 2: Rendering 3D Objects Stably
143
+ The second implementation subproblem was to figure out how to render 3D objects in the world and effectively
144
+ track them. Originally the implementation used Snap’s World Tracking [14] functionality, and then its Surface Tracking
145
+ [14] functionality, with neither proving to be too accurate. Chip stacks are small and users expect them to stay in
146
+ Manuscript submitted to ACM
147
+
148
+ 5:49
149
+ 5:49
150
+ 79
151
+ Hand #: 1
152
+ Hand #:1
153
+ Round: Preflop
154
+ Round:Preflop
155
+ Dealer:Opponent
156
+ Dealer: Me
157
+ Amount to Call: $1
158
+ $99
159
+ Opponent
160
+ $98
161
+ Pot:ss
162
+ Waiting for opponent..
163
+ 1
164
+ Z
165
+ 3
166
+ 4
167
+ 5
168
+ 6
169
+ 7
170
+ 8
171
+ 9
172
+ O
173
+
174
+ $
175
+ &
176
+ @
177
+ X
178
+ ABC
179
+ space
180
+ done
181
+ Send ChatPokAR: Facilitating Poker Play Through Augmented Reality
182
+ 5
183
+ Start
184
+ B Checks
185
+ B Bets
186
+ endhand()
187
+ A Checks
188
+ A Bets
189
+ A Folds
190
+ A Calls
191
+ advance()
192
+ B Folds
193
+ B Calls
194
+ Fig. 3. Pre-flop DFA (used before any community cards are dealt).
195
+ relatively the same position throughout a game. However, with World Tracking and Surface Tracking, chip stacks
196
+ would move throughout the room quite a bit if the mobile device’s camera was moved.
197
+ Then, we decided to use Marker Tracking [14], which uses a printed marker pattern to mark a position in the physical
198
+ world and allow the application to render objects relative to that position. Marker tracking proved to be very accurate
199
+ and stable for rendering 3D objects in AR. Chip stacks would no longer move throughout the room, as they were
200
+ grounded in a location in 3D space. Marker Tracking, however, is only a temporary solution while alternative tracking
201
+ solutions are improved with continued computer vision research.
202
+ 3.4
203
+ Subproblem 3: Connecting Multiple Players
204
+ The third implementation subproblem was to figure out how to connect two players within a single Snapchat Lens to
205
+ play together and share game data. To solve this problem, we designed an API to connect two different mobile devices
206
+ and allow them to send messages between each other, including updates on the game state and player actions. This API
207
+ is built on top of Snap’s Connected Lenses feature [11]. Thus, the two players will always observe the same data on
208
+ different devices in real time, unifying their gameplay experience.
209
+ Manuscript submitted to ACM
210
+
211
+ 6
212
+ Gamba and Monroy-Hernández
213
+ Start
214
+ A Calls
215
+ A Raises
216
+ endhand()
217
+ B Checks
218
+ B Raises
219
+ B Folds
220
+ B Calls
221
+ advance()
222
+ A Folds
223
+ A Calls
224
+ A Folds
225
+ Fig. 4. Post-flop DFA (used once community cards have started to been dealt).
226
+ 4
227
+ EVALUATION
228
+ We recruited 8 participants who were found through a poker club on campus and recruited by email. We asked
229
+ participants to respond to a pre-study survey to learn about their individual experience with poker and its rules. This
230
+ survey can be found in Appendix B. We randomly paired participants into 4 dyads to utilize PokAR to play heads-up
231
+ poker for 25 minutes. Then, we asked them to respond to a post-study survey about their experience with the application.
232
+ In this survey, we asked participants about the benefits and detriments of particular PokAR features, the effects of AR
233
+ on socialization, the effects of AR on game pace, and the overall experience with PokAR. This survey can be found in
234
+ Appendix C. The study protocol above is described in more detail in Appendix A.
235
+ 5
236
+ RESULTS
237
+ Although participants had varying levels of poker expertise, they all had a self-reported understanding of the rules.
238
+ Based on the results of the pre-study survey, we grouped the 8 participants into three groups of varying experience
239
+ levels for analysis: Highly Experienced (play poker multiple times a week, 𝑛 = 2), Moderately Experienced (play poker
240
+ weekly to monthly, 𝑛 = 3), and Slightly Experienced (play poker yearly or less, 𝑛 = 3).
241
+ Manuscript submitted to ACM
242
+
243
+ PokAR: Facilitating Poker Play Through Augmented Reality
244
+ 7
245
+ 5.1
246
+ Evaluation of Features
247
+ We asked participants to rate each of the five major PokAR features on a scale of 1 (detrimental) to 5 (beneficial) in
248
+ terms of its effectiveness compared to the corresponding object/action in real-life poker. Each feature earned an average
249
+ score > 3 (leaning beneficial) among all participants. Specifically, "3D AR Chips" scored a 4, "UI Action Buttons" scored
250
+ a 4.25, "Game Messages" scored a 3.875, "Counting Stacks" scored a 4.375, and "Awarding Pots" scored a 4.25.
251
+ Notably, the only features that scored < 3 (leaning detrimental) were “3D AR Chips,” “UI Action Buttons,” and
252
+ “Game Messages” for the Highly Experienced subgroup of participants. This could be explained by the fact that all
253
+ three of these features are intended to alleviate the requirement of knowing the complex rules of poker. However, in
254
+ the pre-study survey, all members of this subgroup answered that they play poker quite often and they confidently
255
+ understand all the rules, so these features likely just got in the way of their gameplay. The features of “Counting Stacks”
256
+ and “Awarding Pots” were, however, positively received by all three subgroups of participants.
257
+ 5.2
258
+ Evaluation of Socialization
259
+ The average response to the survey question: “How much did AR affect the in-person social aspects of the game of
260
+ poker?” was a 3.25 on a scale of 1 (negatively) to 5 (positively), meaning that AR neither significantly improved nor
261
+ impaired the in-person social aspects of poker. This is a beneficial result, as one of PokAR’s goals was to supplement the
262
+ game of poker. We did not implement social-related features intended to improve socialization, but this result supports
263
+ the claim that the AR features of PokAR did not impair socialization. In other words, players are utilizing PokAR as a
264
+ tool to enable poker play, which does not get in the way of the traditional social interactions at a poker table.
265
+ Additionally, multiple participants noted that AR did not heavily influence socialization. P1 stated that the experience
266
+ was “no different, we could still talk and converse,” and P5 stated that AR “Didn’t affect socialization because everything
267
+ was still in person.”
268
+ 5.3
269
+ Evaluation of Game Pace
270
+ The average response to the survey question: “How did augmented reality affect the game pace of poker?” was a
271
+ 3.75 on a scale of 1 (slowed the game) to 5 (sped up the game), meaning that AR slightly increased the game pace of
272
+ poker. In this study, the average game pace was 40.3 hands/hour (67.2 hands/hour for the Highly Experienced subgroup).
273
+ Comparatively, “A typical live poker game will deal 25-30 per hour,” assuming 9 players [2]. This section requires
274
+ additional study, including a control session of each participant group playing traditional poker to compare the game
275
+ pace with AR poker.
276
+ 5.4
277
+ Evaluation of Overall Experience
278
+ On average, participants rated their overall experience at a 4.5/5, overwhelmingly positive. P2 stated that “It’s quicker,
279
+ but annoying to hold the phone up.” P6 stated that “It was a different experience which took a little getting used to, but I
280
+ enjoyed it.” P4 stated that it “Felt cool to have the chips tracked for you. Definitely could see myself using it on a camping
281
+ trip or during traveling.” P5 stated that “It was cool, because sometimes I’d like to play poker but sometimes have no chips!”
282
+ 6
283
+ CONCLUSION
284
+ Through developing a complete AR application and studying how people utilize it, we have generated three main
285
+ conclusions.
286
+ Manuscript submitted to ACM
287
+
288
+ 8
289
+ Gamba and Monroy-Hernández
290
+ 6.1
291
+ AR Has the Potential to Augment and Simplify Traditional Table Games
292
+ After our work throughout this semester, we are confident that AR as a technology can and will be used in the future
293
+ to augment and simplify traditional table games, like poker. While the technology is currently in a primitive state, it is
294
+ continually evolving and progressing. Several participants noted that the gameplay of PokAR was clunky since they
295
+ had to constantly hold their mobile devices up to see the game. However, with improved AR technology, this annoyance
296
+ will begin to fade away. For example, AR glasses like Spectacles will eliminate the need to hold up a mobile device [13].
297
+ This study revealed promising results concerning the future potential of AR in games. For instance, the use of AR did
298
+ not hinder socialization, and participants had a positive overall experience. PokAR features meant to facilitate gameplay
299
+ were positively received by most players, and options to disable disruptive features would alleviate the rest.
300
+ 6.2
301
+ AR Should Not Be Used to Replace Traditional Experiences; It Should Be Used to Augment Them
302
+ This conclusion stems from the fact that AR technology has innate limitations compared to the physical world. For
303
+ instance, AR experiences are less tactile than physical world experiences. While software tricks exist to improve the
304
+ tactility of AR experiences (like hand tracking, which enables object manipulation), it will never feel quite like the
305
+ physical world. For example, several participants noted that PokAR was missing one important aspect of traditional
306
+ poker: chip shuffling. Chip shuffling is a common fidgeting technique among poker players in which they use one hand
307
+ to rearrange a stack of chips. Shuffling is almost unanimous among poker players and is commonly used to pass time
308
+ and cure boredom during long poker sessions. While AR could simulate chip shuffling, it could never reproduce the
309
+ experience perfectly.
310
+ Early intuition about this conclusion is one reason we decided to utilize physical playing cards in PokAR. If playing
311
+ cards were virtual, players would be playing an online poker game in which in-person social interactions were minimal.
312
+ Players wouldn’t even need to be co-located to play PokAR anymore. This would be a case of using AR to replace
313
+ a traditional game experience. Instead, we decided to use physical cards and virtual chips in PokAR to afford some
314
+ physical-world tactility to players and streamline some of the more annoying and time-consuming aspects of poker,
315
+ like counting chips.
316
+ As mentioned in the introduction, PokAR is not intended to replace traditional poker, only augment it. This is due to
317
+ the inherent limitations of AR technology. More broadly, AR should not be used to replace traditional experiences, only
318
+ augment them. Augmentations should be deliberately planned and carefully implemented to ensure that they do not
319
+ take over the spirit of the game. Go too far with augmentation, and you approach the virtual reality world and lose out
320
+ on social interaction.
321
+ 6.3
322
+ Future Work
323
+ Our work on PokAR has revealed possible directions for future study and enhancements to the application. Firstly,
324
+ due to time constraints, not all features of poker were able to be added. PokAR is currently limited to just two players.
325
+ This design choice was made to reduce the project’s complexity, but this limit should be increased to the accepted limit
326
+ of nine players to better simulate traditional poker. Additionally, the option to chop pots (split pots equally between
327
+ tied players) is currently not implemented. The option to run it multiple times (deal remaining cards multiple times in
328
+ an all-in situation and award the pot proportionally to winners) is also not yet implemented.
329
+ PokAR would benefit from increased tactility, which is why we believe that it is a necessary direction for future
330
+ work. Increased tactility could come in two forms, with the first being the ability to grab AR chips and manipulate them
331
+ Manuscript submitted to ACM
332
+
333
+ PokAR: Facilitating Poker Play Through Augmented Reality
334
+ 9
335
+ with your hands. This feature would let players bet more realistically (rather than just clicking a button) or could help
336
+ simulate chip shuffling (which was mentioned earlier as a lacking aspect). The other way to increase tactility would be
337
+ to utilize hand gestures, rather than UI buttons, to signal actions. For example, players could tap the table with a fist to
338
+ signal a ‘check,’ as in traditional poker. These features would further increase the immersion of PokAR.
339
+ Finally, AR could be utilized to provide helpful statistical annotations for players. Possible annotations could include
340
+ the probability of winning or the probability of making a certain hand. To implement this feature, one must first
341
+ implement a computer vision model to recognize and classify playing cards. This has been done in the past with high
342
+ accuracy (> 99%) [8]. This feature would further reduce the mental load on players and help them play and learn poker
343
+ more effectively.
344
+ Manuscript submitted to ACM
345
+
346
+ 10
347
+ Gamba and Monroy-Hernández
348
+ REFERENCES
349
+ [1] Ella Dagan, Ana Cárdenas Gasca, Ava Robinson, Anwar Noriega, Yu Jiang Tham, Rajan Vaish, and Andrés Monroy-Hernández. 2022. Project IRL:
350
+ Playful Co-Located Interactions with Mobile Augmented Reality. Proceedings of the ACM on Human-Computer Interaction 6, CSCW1 (March 2022),
351
+ 1–27. https://doi.org/10.1145/3512909 arXiv:2201.02558 [cs].
352
+ [2] Geoffrey Fisk. 2020. How Many Hands Are Played Per Hour in Live Poker Games? https://upswingpoker.com/hands-per-hour-live-poker-vs-online/
353
+ [3] Christothea Herodotou. 2010. Social Praxis Within and Around Online Gaming: The Case of World of Warcraft. In 2010 Third IEEE International
354
+ Conference on Digital Game and Intelligent Toy Enhanced Learning. 10–22.
355
+ [4] Christothea Herodotou, Niall Winters, and Maria Kambouri. 2015. An Iterative, Multidisciplinary Approach to Studying Digital Play Motivation:
356
+ The Model of Game Motivation. Games and Culture 10, 3 (May 2015), 249–268. https://doi.org/10.1177/1555412014557633
357
+ [5] Hyun-Woo Lee, Sanghoon Kim, and Jun-Phil Uhm. 2021. Social Virtual Reality (VR) Involvement Affects Depression When Social Connectedness
358
+ and Self-Esteem Are Low: A Moderated Mediation on Well-Being. Frontiers in Psychology 12 (2021). https://www.frontiersin.org/articles/10.3389/
359
+ fpsyg.2021.753019
360
+ [6] Meta. 2019. Poker VR - Multi Table Tournaments on Oculus Quest. https://www.oculus.com/experiences/quest/2257223740990488/
361
+ [7] Fast Offshore. 2021. Online poker sector overview for 2021: Stats, key drivers and more. https://fastoffshore.com/2021/10/online-poker-sector-
362
+ overview-2021/
363
+ [8] Arjun Rohlfing-Das. 2020. Image Classification for Playing Cards. https://medium.com/swlh/image-classification-for-playing-cards-26d660f3149e
364
+ [9] Hiroyuki Sakuma, Tetsuo Yamabe, and Tatsuo Nakajima. 2012. Enhancing Traditional Games with Augmented Reality Technologies. In 2012 9th
365
+ International Conference on Ubiquitous Intelligence and Computing and 9th International Conference on Autonomic and Trusted Computing. 822–825.
366
+ https://doi.org/10.1109/UIC-ATC.2012.95
367
+ [10] Nina Savela, Atte Oksanen, Markus Kaakinen, Marius Noreikis, and Yu Xiao. 2020. Does Augmented Reality Affect Sociability, Entertainment, and
368
+ Learning? A Field Experiment. Applied Sciences 10, 4 (Jan. 2020), 1392. https://doi.org/10.3390/app10041392 Number: 4 Publisher: Multidisciplinary
369
+ Digital Publishing Institute.
370
+ [11] Snap. 2022. Connected Lenses Overview | Docs. https://docs.snap.com/lens-studio/references/guides/lens-features/connected-lenses/connected-
371
+ lenses-overview
372
+ [12] Snap. 2022. How do I use Lenses on Snapchat? https://support.snapchat.com/en-US/a/face-world-lenses
373
+ [13] Snap. 2022. Spectacles by Snap Inc. • The Next Generation of Spectacles. https://www.spectacles.com/
374
+ [14] Snap. 2022. Tracking Modes | Docs. https://docs.snap.com/lens-studio/references/guides/lens-features/tracking/world/tracking-modes
375
+ [15] Christoph Thul. 2013. PokerTool - Entwicklung und Implementierung einer AR-Android-Anwendung für Wahrscheinlichkeitsberechnungen bei
376
+ Texas Holdem Poker. (Sept. 2013). https://kola.opus.hbz-nrw.de/opus45-kola/frontdoor/index/index/docId/769
377
+ [16] Mark Weiser. 1991. The Computer for the 21st Century. (1991).
378
+ Manuscript submitted to ACM
379
+
380
+ PokAR: Facilitating Poker Play Through Augmented Reality
381
+ 11
382
+ A
383
+ STUDY PROTOCOL
384
+ Below is the process we asked participants to follow during the study:
385
+ (1) Participants were found through a poker club on campus and recruited by email.
386
+ (2) Participants were asked to respond to the pre-study survey.
387
+ (3) The participants were randomly paired into dyads for heads-up poker play.
388
+ (4) During the study:
389
+ (a) Participants were asked to download Snapchat (if necessary) and scan a code to gain access to the PokAR
390
+ Snapchat Lens.
391
+ (b) Participants were asked to sign consent forms.
392
+ (c) Participants were asked to use PokAR to play heads-up poker (without using real-world money) for 25 minutes.
393
+ (d) We took notes on comments, reactions, game pace, frustrations, etc. We took photos and videos throughout.
394
+ We also answered questions about the application when asked, but we avoided guiding the players.
395
+ (5) After play, participants were asked to respond to the post-study survey.
396
+ B
397
+ PRE-STUDY SURVEY
398
+ Below are the questions asked during the pre-study survey.
399
+ • How well do you know the rules of Heads-Up Texas Hold’em Poker?
400
+ • How often do you play poker?
401
+ C
402
+ POST-STUDY SURVEY
403
+ Below are the questions asked during the post-study survey.
404
+ • Please rate each of the following PokAR features in terms of its effectiveness when compared to the corresponding
405
+ object/action in real-life Texas hold’em poker?
406
+ – 3D AR Chips
407
+ – UI Action Buttons
408
+ – Game Messages
409
+ – Counting Stacks
410
+ – Awarding Pots
411
+ • How much did AR affect the in-person social aspects of the game of poker?
412
+ • How did augmented reality affect the game pace of poker (# of hands played / unit time)? Ignore the first few
413
+ hands in which you were learning the application.
414
+ • Overall, how would you describe your experience with PokAR?
415
+ D
416
+ CODE REPOSITORY
417
+ The code for this project can be found at the following GitHub repository: https://github.com/adamgamba/PokAR.
418
+ Manuscript submitted to ACM
419
+
0tAyT4oBgHgl3EQfoPjL/content/tmp_files/load_file.txt ADDED
@@ -0,0 +1,343 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf,len=342
2
+ page_content='PokAR: Facilitating Poker Play Through Augmented Reality ADAM GAMBA and ANDRÉS MONROY-HERNÁNDEZ, Princeton University, USA Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
3
+ page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
4
+ page_content=' Two players using the PokAR application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
5
+ page_content=' We introduce PokAR, an augmented reality (AR) application to facilitate poker play.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
6
+ page_content=' PokAR aims to alleviate three difficulties of traditional poker by leveraging AR technology: (1) need to have physical poker chips, (2) complex rules of poker, (3) slow game pace caused by laborious tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
7
+ page_content=' Despite the potential benefits of AR in poker, not much research has been done in the field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
8
+ page_content=' In fact, PokAR is the first application to enable AR poker on a mobile device without requiring extra costly equipment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
9
+ page_content=' This has been done by creating a Snapchat Lens 1 which can be used on most mobile devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
10
+ page_content=' We evaluated this application by instructing 4 participant dyads to use PokAR to engage in poker play and respond to survey questions about their experience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
11
+ page_content=' We found that most PokAR features were positively received, AR did not significantly improve nor hinder socialization, PokAR slightly increased the game pace, and participants had an overall enjoyable experience with the Lens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
12
+ page_content=' These findings led to three major conclusions: (1) AR has the potential to augment and simplify traditional table games, (2) AR should not be used to replace traditional experiences, only augment them, (3) Future work includes additional features like increased tactility and statistical annotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
13
+ page_content=' CCS Concepts: • Human-centered computing → Collaborative and social computing devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
14
+ page_content=' Additional Key Words and Phrases: connected lens, augmented reality, poker, co-located, interaction, socialization ACM Reference Format: Adam Gamba and Andrés Monroy-Hernández.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
15
+ page_content=' 2023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
16
+ page_content=' PokAR: Facilitating Poker Play Through Augmented Reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
17
+ page_content=' 1, 1 (January 2023), 11 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
18
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
19
+ page_content='org/XXXXXXX.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
20
+ page_content='XXXXXXX 1A Lens in Snapchat is an experience that utilizes augmented reality to transform the world around you [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
21
+ page_content=' Authors’ address: Adam Gamba, agamba@princeton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
22
+ page_content='edu;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
23
+ page_content=' Andrés Monroy-Hernández, andresmh@princeton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
24
+ page_content='edu, Princeton University, Princeton, New Jersey, USA, 08544.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
25
+ page_content=' Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
26
+ page_content=' Copyrights for components of this work owned by others than ACM must be honored.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
27
+ page_content=' Abstracting with credit is permitted.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
28
+ page_content=' To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
29
+ page_content=' Request permissions from permissions@acm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
30
+ page_content='org.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
31
+ page_content=' © 2023 Association for Computing Machinery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
32
+ page_content=' Manuscript submitted to ACM Manuscript submitted to ACM 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
33
+ page_content='00505v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
34
+ page_content='HC] 2 Jan 2023 l:61:598 old2 Gamba and Monroy-Hernández 1 INTRODUCTION The goal of this project is to facilitate heads-up Texas hold’em poker play through augmented reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
35
+ page_content=' Poker is cumbersome to play in its current form, requiring players to have poker chips and knowledge of the complex rules to play correctly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
36
+ page_content=' Without an experienced player to guide the game, new players often find it difficult to learn the rules and play correctly [9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
37
+ page_content=' Additionally, due to the burden of physical chips, it is difficult to play poker in many scenarios (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
38
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
39
+ page_content=', at the beach, while traveling, or camping).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
40
+ page_content=' Finally, the game pace is often slowed due to poker’s complex rules and the need for laborious tasks like counting chip stacks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
41
+ page_content=' Augmented reality technology is well-equipped to solve these issues in three ways.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
42
+ page_content=' Firstly, AR can eliminate the need for physical poker chips by instead utilizing AR to render chips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
43
+ page_content=' Next, AR can help guide players through the complex rules of poker by hinting at legal actions during gameplay.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
44
+ page_content=' Finally, AR can help decrease the burden of laborious tasks (like counting chips) and increase the game pace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
45
+ page_content=' PokAR helps alleviate these three issues, which we’ll discuss further throughout this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
46
+ page_content=' Poker is a popular game, with over 120 million players worldwide playing regularly online [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
47
+ page_content=' Texas hold’em is one of the most popular poker variants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
48
+ page_content=' In this variant, players are dealt two private cards and five community cards, and they battle to make the best hand or bluff opponents into folding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
49
+ page_content=' ’Heads-up’ poker is a term used to describe poker played by just two players, head-to-head.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
50
+ page_content=' In its current state, PokAR supports only heads-up Texas hold’em poker, but with future work, it could be extended to more players and more variants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
51
+ page_content=' Throughout this paper, we will use the term ’poker’ to refer to heads-up Texas hold’em poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
52
+ page_content=' Poker is a classic example of a social, co-located game, since poker, by design, emphasizes in-person, co-located interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
53
+ page_content=' Players often look at each other and speak to each other during a poker game, either to gain information or to socialize.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
54
+ page_content=' Also, poker forces players to focus on the same enablers, or "physical objects that trigger and are the focus of the AR experience" [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
55
+ page_content=' These enablers, like playing cards and poker chips, can help guide an AR experience and engage players more closely than in games that do not have a similar shared focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
56
+ page_content=' For the above reasons, we chose to augment poker in this study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
57
+ page_content=' PokAR is not intended to replace traditional poker, rather it helps people play when traditional poker would be difficult or impractical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
58
+ page_content=' The goal of AR applications should be to disappear completely and seamlessly immerse the user in a realistic experience that combines reality with augmentation [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
59
+ page_content=' This disappearance frees users to utilize these applications more effortlessly, allowing them to focus on new goals, beyond the application itself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
60
+ page_content=' 2 RELATED WORK While we have established AR as a possible solution to the aforementioned issues with poker, very little work has been done concerning AR poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
61
+ page_content=' Additionally, while AR is not yet a heavily explored area, researchers have argued that “A Poker-Assistance-Software is an ideal test area for an AR Application with real added value,” with possible areas to add value including automation and statistical estimations [15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
62
+ page_content=' Similar projects in the past have all relied on physical means to augment reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
63
+ page_content=' For example, researchers used overhead projectors to project all aspects of the poker game (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
64
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
65
+ page_content=', cards, chips, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
66
+ page_content=') onto a table [9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
67
+ page_content=' Additionally, researchers have used RFID playing cards to detect the dealt cards [9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
68
+ page_content=' While this study succeeded in creating an AR application to ease some of the same cumbersome aspects of poker tackled by PokAR (physical chips, complex rules, slow game pace), it did so using a high-cost solution, which is impractical for most recreational use.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
69
+ page_content=' Additionally, they disregarded studying how this AR setup influenced social interactions in poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
70
+ page_content=' Manuscript submitted to ACM PokAR: Facilitating Poker Play Through Augmented Reality 3 Furthermore, people have utilized virtual reality (VR) in the past to create commercial poker video games.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
71
+ page_content=' One example is PokerVR by Meta, which uses “expressive avatars built for reading tells with growing customizations” [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
72
+ page_content=' While they may say this, VR poker applications still just employ static players’ avatars, which do not emphasize the in-person, social nature of poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
73
+ page_content=' PokAR is the first project to enable AR poker without needing additional equipment other than a mobile device and a regular deck of cards (like an overhead projector or a VR headset).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
74
+ page_content=' This is a worthwhile problem because it is the first application to enable AR poker at a low cost, since it only requires a few commonly-owned pieces of equipment (mobile devices and playing cards).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
75
+ page_content=' Additionally, utilizing AR over VR allows the gameplay to emphasize the social aspects of co-location and increase socialization when compared to VR implementations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
76
+ page_content=' Co-located gaming has been shown to lead to more effective and enjoyable gaming, as players can more easily communicate and build social relationships [4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
77
+ page_content=' One major reason for this is "out-of-the-game, game-related communi- cation" [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
78
+ page_content=' By having the ability to converse about other topics while simultaneously being involved in a game with another player, these players are given the opportunity to build a deeper connection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
79
+ page_content=' Additionally, when comparing socialization in AR and VR, prior research tends to support increased socialization in AR applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
80
+ page_content=' AR games have the ability to "potentially enhance social communication and social interaction between people" [10], whereas high-involvement in VR games could potentially isolate users socially and "negatively affect their well-being" [5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
81
+ page_content=' Thus, we chose to develop an AR application rather than a VR application to reap the social benefits of shared, co-located experiences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
82
+ page_content=' 3 POKAR SYSTEM The PokAR Snapchat Lens allows users to play heads-up poker with another player on two mobile devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
83
+ page_content=' AR visual annotations include 3D models of poker chips which dynamically render with changing stack size, and 3D text above the chip stacks denoting the size of each stack.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
84
+ page_content=' 2D visual annotations include number of hands played, current round, current dealer, previous action, amount to call, waiting message, and UI buttons with labels "Check," "Call," "Bet," "Raise," and "Fold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
85
+ page_content='" All AR and 2D annotations render dynamically with changing game state, stack amounts, and legal actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
86
+ page_content=' PokAR implements five main features to help achieve its motivating goals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
87
+ page_content=' “3D AR Chips” - PokAR renders 3D models of chips to eliminate the requirement of needing physical poker chips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
88
+ page_content=' “UI Action Buttons” - 2D buttons rendered on the player’s screen allows them to select among and perform legal actions at the current state of the game.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
89
+ page_content=' “Game Messages” - Messages provide additional information to both players about the actions of players, bet amounts, and more throughout the game.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
90
+ page_content=' “Counting Stacks” - A live count of the number of chips in all chip stacks is rendered above the 3D models, eliminating the hassle of counting chips manually.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
91
+ page_content=' “Awarding Pots” - After a winner is determined (through folding or at showdown), the chips are automatically awarded to the winner, eliminating the hassle of moving chips manually.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
92
+ page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
93
+ page_content='1 Approach To achieve our motivating goals, we developed a Snapchat Lens which could be used on any mobile device to facilitate poker play through AR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
94
+ page_content=' Besides having physical playing cards, players will play the game of poker in AR Manuscript submitted to ACM 4 Gamba and Monroy-Hernández Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
95
+ page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
96
+ page_content=' Side-by-side points of view of the same game of PokAR on two mobile devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
97
+ page_content=' by interacting with their augmented environment through a mobile device.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
98
+ page_content=' We used the Snap Lens Studio IDE with JavaScript for development, the Snapchat app for testing and deployment, and GitHub for version control.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
99
+ page_content=' Code for this project can be found at the link in Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
100
+ page_content=' Physical playing cards act as an enabler to ground the game in physical reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
101
+ page_content=' A demonstration of the completed application is shown in Fig 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
102
+ page_content=' In the development of PokAR, we faced and solved three major implementation subproblems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
103
+ page_content=' They will be discussed below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
104
+ page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
105
+ page_content='2 Subproblem 1: Modeling Poker in Code The first implementation subproblem to solve was to figure out how to model a game of poker in code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
106
+ page_content=' By nature of the rules of poker, the game is deterministic based on previous player actions within a betting round.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
107
+ page_content=' Thus, the game state can be modeled using a Deterministic Finite Automaton (DFA).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
108
+ page_content=' The DFA for our application determines the legal actions and/or termination state of the betting round, given previous actions within the betting round.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
109
+ page_content=' At the end of each betting round, one of two termination states is reached, dictating whether the hand has ended or players will advance to the next betting round.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
110
+ page_content=' Fig 3 and Fig 4 show a graphical representation of the DFAs we used in our implementation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
111
+ page_content=' In both DFAs, the application begins with the Start state and terminates in either the endHand() or advance() state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
112
+ page_content=' The double-headed arrow represents the possible cycle of betting, raising, reraising, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
113
+ page_content=', until one of the players is eventually all-in.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
114
+ page_content=' Player ’A’ is the one assigned ‘opponent’ at the start of a hand, and player ’B’ is the ‘dealer.’ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
115
+ page_content='3 Subproblem 2: Rendering 3D Objects Stably The second implementation subproblem was to figure out how to render 3D objects in the world and effectively track them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
116
+ page_content=' Originally the implementation used Snap’s World Tracking [14] functionality, and then its Surface Tracking [14] functionality, with neither proving to be too accurate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
117
+ page_content=' Chip stacks are small and users expect them to stay in Manuscript submitted to ACM 5:49 5:49 79 Hand #: 1 Hand #:1 Round: Preflop Round:Preflop Dealer:Opponent Dealer: Me Amount to Call: $1 $99 Opponent $98 Pot:ss Waiting for opponent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
118
+ page_content='. 1 Z 3 4 5 6 7 8 9 O ) $ & @ X ABC space done Send ChatPokAR: Facilitating Poker Play Through Augmented Reality 5 Start B Checks B Bets endhand() A Checks A Bets A Folds A Calls advance() B Folds B Calls Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
119
+ page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
120
+ page_content=' Pre-flop DFA (used before any community cards are dealt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
121
+ page_content=' relatively the same position throughout a game.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
122
+ page_content=' However, with World Tracking and Surface Tracking, chip stacks would move throughout the room quite a bit if the mobile device’s camera was moved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
123
+ page_content=' Then, we decided to use Marker Tracking [14], which uses a printed marker pattern to mark a position in the physical world and allow the application to render objects relative to that position.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
124
+ page_content=' Marker tracking proved to be very accurate and stable for rendering 3D objects in AR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
125
+ page_content=' Chip stacks would no longer move throughout the room, as they were grounded in a location in 3D space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
126
+ page_content=' Marker Tracking, however, is only a temporary solution while alternative tracking solutions are improved with continued computer vision research.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
127
+ page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
128
+ page_content='4 Subproblem 3: Connecting Multiple Players The third implementation subproblem was to figure out how to connect two players within a single Snapchat Lens to play together and share game data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
129
+ page_content=' To solve this problem, we designed an API to connect two different mobile devices and allow them to send messages between each other, including updates on the game state and player actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
130
+ page_content=' This API is built on top of Snap’s Connected Lenses feature [11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
131
+ page_content=' Thus, the two players will always observe the same data on different devices in real time, unifying their gameplay experience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
132
+ page_content=' Manuscript submitted to ACM 6 Gamba and Monroy-Hernández Start A Calls A Raises endhand() B Checks B Raises B Folds B Calls advance() A Folds A Calls A Folds Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
133
+ page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
134
+ page_content=' Post-flop DFA (used once community cards have started to been dealt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
135
+ page_content=' 4 EVALUATION We recruited 8 participants who were found through a poker club on campus and recruited by email.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
136
+ page_content=' We asked participants to respond to a pre-study survey to learn about their individual experience with poker and its rules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
137
+ page_content=' This survey can be found in Appendix B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
138
+ page_content=' We randomly paired participants into 4 dyads to utilize PokAR to play heads-up poker for 25 minutes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
139
+ page_content=' Then, we asked them to respond to a post-study survey about their experience with the application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
140
+ page_content=' In this survey, we asked participants about the benefits and detriments of particular PokAR features, the effects of AR on socialization, the effects of AR on game pace, and the overall experience with PokAR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
141
+ page_content=' This survey can be found in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
142
+ page_content=' The study protocol above is described in more detail in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
143
+ page_content=' 5 RESULTS Although participants had varying levels of poker expertise, they all had a self-reported understanding of the rules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
144
+ page_content=' Based on the results of the pre-study survey, we grouped the 8 participants into three groups of varying experience levels for analysis: Highly Experienced (play poker multiple times a week, 𝑛 = 2), Moderately Experienced (play poker weekly to monthly, 𝑛 = 3), and Slightly Experienced (play poker yearly or less, 𝑛 = 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
145
+ page_content=' Manuscript submitted to ACM PokAR: Facilitating Poker Play Through Augmented Reality 7 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
146
+ page_content='1 Evaluation of Features We asked participants to rate each of the five major PokAR features on a scale of 1 (detrimental) to 5 (beneficial) in terms of its effectiveness compared to the corresponding object/action in real-life poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
147
+ page_content=' Each feature earned an average score > 3 (leaning beneficial) among all participants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
148
+ page_content=' Specifically, "3D AR Chips" scored a 4, "UI Action Buttons" scored a 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
149
+ page_content='25, "Game Messages" scored a 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
150
+ page_content='875, "Counting Stacks" scored a 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
151
+ page_content='375, and "Awarding Pots" scored a 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
152
+ page_content='25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
153
+ page_content=' Notably, the only features that scored < 3 (leaning detrimental) were “3D AR Chips,” “UI Action Buttons,” and “Game Messages” for the Highly Experienced subgroup of participants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
154
+ page_content=' This could be explained by the fact that all three of these features are intended to alleviate the requirement of knowing the complex rules of poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
155
+ page_content=' However, in the pre-study survey, all members of this subgroup answered that they play poker quite often and they confidently understand all the rules, so these features likely just got in the way of their gameplay.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
156
+ page_content=' The features of “Counting Stacks” and “Awarding Pots” were, however, positively received by all three subgroups of participants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
157
+ page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
158
+ page_content='2 Evaluation of Socialization The average response to the survey question: “How much did AR affect the in-person social aspects of the game of poker?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
159
+ page_content=' was a 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
160
+ page_content='25 on a scale of 1 (negatively) to 5 (positively), meaning that AR neither significantly improved nor impaired the in-person social aspects of poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
161
+ page_content=' This is a beneficial result, as one of PokAR’s goals was to supplement the game of poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
162
+ page_content=' We did not implement social-related features intended to improve socialization, but this result supports the claim that the AR features of PokAR did not impair socialization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
163
+ page_content=' In other words, players are utilizing PokAR as a tool to enable poker play, which does not get in the way of the traditional social interactions at a poker table.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
164
+ page_content=' Additionally, multiple participants noted that AR did not heavily influence socialization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
165
+ page_content=' P1 stated that the experience was “no different, we could still talk and converse,” and P5 stated that AR “Didn’t affect socialization because everything was still in person.” 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
166
+ page_content='3 Evaluation of Game Pace The average response to the survey question: “How did augmented reality affect the game pace of poker?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
167
+ page_content=' was a 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
168
+ page_content='75 on a scale of 1 (slowed the game) to 5 (sped up the game), meaning that AR slightly increased the game pace of poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
169
+ page_content=' In this study, the average game pace was 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
170
+ page_content='3 hands/hour (67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
171
+ page_content='2 hands/hour for the Highly Experienced subgroup).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
172
+ page_content=' Comparatively, “A typical live poker game will deal 25-30 per hour,” assuming 9 players [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
173
+ page_content=' This section requires additional study, including a control session of each participant group playing traditional poker to compare the game pace with AR poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
174
+ page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
175
+ page_content='4 Evaluation of Overall Experience On average, participants rated their overall experience at a 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
176
+ page_content='5/5, overwhelmingly positive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
177
+ page_content=' P2 stated that “It’s quicker, but annoying to hold the phone up.” P6 stated that “It was a different experience which took a little getting used to, but I enjoyed it.” P4 stated that it “Felt cool to have the chips tracked for you.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
178
+ page_content=' Definitely could see myself using it on a camping trip or during traveling.” P5 stated that “It was cool, because sometimes I’d like to play poker but sometimes have no chips!”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
179
+ page_content=' 6 CONCLUSION Through developing a complete AR application and studying how people utilize it, we have generated three main conclusions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
180
+ page_content=' Manuscript submitted to ACM 8 Gamba and Monroy-Hernández 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
181
+ page_content='1 AR Has the Potential to Augment and Simplify Traditional Table Games After our work throughout this semester, we are confident that AR as a technology can and will be used in the future to augment and simplify traditional table games, like poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
182
+ page_content=' While the technology is currently in a primitive state, it is continually evolving and progressing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
183
+ page_content=' Several participants noted that the gameplay of PokAR was clunky since they had to constantly hold their mobile devices up to see the game.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
184
+ page_content=' However, with improved AR technology, this annoyance will begin to fade away.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
185
+ page_content=' For example, AR glasses like Spectacles will eliminate the need to hold up a mobile device [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
186
+ page_content=' This study revealed promising results concerning the future potential of AR in games.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
187
+ page_content=' For instance, the use of AR did not hinder socialization, and participants had a positive overall experience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
188
+ page_content=' PokAR features meant to facilitate gameplay were positively received by most players, and options to disable disruptive features would alleviate the rest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
189
+ page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
190
+ page_content='2 AR Should Not Be Used to Replace Traditional Experiences;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
191
+ page_content=' It Should Be Used to Augment Them This conclusion stems from the fact that AR technology has innate limitations compared to the physical world.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
192
+ page_content=' For instance, AR experiences are less tactile than physical world experiences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
193
+ page_content=' While software tricks exist to improve the tactility of AR experiences (like hand tracking, which enables object manipulation), it will never feel quite like the physical world.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
194
+ page_content=' For example, several participants noted that PokAR was missing one important aspect of traditional poker: chip shuffling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
195
+ page_content=' Chip shuffling is a common fidgeting technique among poker players in which they use one hand to rearrange a stack of chips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
196
+ page_content=' Shuffling is almost unanimous among poker players and is commonly used to pass time and cure boredom during long poker sessions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
197
+ page_content=' While AR could simulate chip shuffling, it could never reproduce the experience perfectly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
198
+ page_content=' Early intuition about this conclusion is one reason we decided to utilize physical playing cards in PokAR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
199
+ page_content=' If playing cards were virtual, players would be playing an online poker game in which in-person social interactions were minimal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
200
+ page_content=' Players wouldn’t even need to be co-located to play PokAR anymore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
201
+ page_content=' This would be a case of using AR to replace a traditional game experience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
202
+ page_content=' Instead, we decided to use physical cards and virtual chips in PokAR to afford some physical-world tactility to players and streamline some of the more annoying and time-consuming aspects of poker, like counting chips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
203
+ page_content=' As mentioned in the introduction, PokAR is not intended to replace traditional poker, only augment it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
204
+ page_content=' This is due to the inherent limitations of AR technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
205
+ page_content=' More broadly, AR should not be used to replace traditional experiences, only augment them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
206
+ page_content=' Augmentations should be deliberately planned and carefully implemented to ensure that they do not take over the spirit of the game.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
207
+ page_content=' Go too far with augmentation, and you approach the virtual reality world and lose out on social interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
208
+ page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
209
+ page_content='3 Future Work Our work on PokAR has revealed possible directions for future study and enhancements to the application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
210
+ page_content=' Firstly, due to time constraints, not all features of poker were able to be added.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
211
+ page_content=' PokAR is currently limited to just two players.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
212
+ page_content=' This design choice was made to reduce the project’s complexity, but this limit should be increased to the accepted limit of nine players to better simulate traditional poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
213
+ page_content=' Additionally, the option to chop pots (split pots equally between tied players) is currently not implemented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
214
+ page_content=' The option to run it multiple times (deal remaining cards multiple times in an all-in situation and award the pot proportionally to winners) is also not yet implemented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
215
+ page_content=' PokAR would benefit from increased tactility, which is why we believe that it is a necessary direction for future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
216
+ page_content=' Increased tactility could come in two forms, with the first being the ability to grab AR chips and manipulate them Manuscript submitted to ACM PokAR: Facilitating Poker Play Through Augmented Reality 9 with your hands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
217
+ page_content=' This feature would let players bet more realistically (rather than just clicking a button) or could help simulate chip shuffling (which was mentioned earlier as a lacking aspect).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
218
+ page_content=' The other way to increase tactility would be to utilize hand gestures, rather than UI buttons, to signal actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
219
+ page_content=' For example, players could tap the table with a fist to signal a ‘check,’ as in traditional poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
220
+ page_content=' These features would further increase the immersion of PokAR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
221
+ page_content=' Finally, AR could be utilized to provide helpful statistical annotations for players.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
222
+ page_content=' Possible annotations could include the probability of winning or the probability of making a certain hand.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
223
+ page_content=' To implement this feature, one must first implement a computer vision model to recognize and classify playing cards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
224
+ page_content=' This has been done in the past with high accuracy (> 99%) [8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
225
+ page_content=' This feature would further reduce the mental load on players and help them play and learn poker more effectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
226
+ page_content=' Manuscript submitted to ACM 10 Gamba and Monroy-Hernández REFERENCES [1] Ella Dagan, Ana Cárdenas Gasca, Ava Robinson, Anwar Noriega, Yu Jiang Tham, Rajan Vaish, and Andrés Monroy-Hernández.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
227
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
228
+ page_content=' Project IRL: Playful Co-Located Interactions with Mobile Augmented Reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
229
+ page_content=' Proceedings of the ACM on Human-Computer Interaction 6, CSCW1 (March 2022), 1–27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
230
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
231
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
232
+ page_content='1145/3512909 arXiv:2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
233
+ page_content='02558 [cs].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
234
+ page_content=' [2] Geoffrey Fisk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
235
+ page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
236
+ page_content=' How Many Hands Are Played Per Hour in Live Poker Games?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
237
+ page_content=' https://upswingpoker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
238
+ page_content='com/hands-per-hour-live-poker-vs-online/ [3] Christothea Herodotou.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
239
+ page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
240
+ page_content=' Social Praxis Within and Around Online Gaming: The Case of World of Warcraft.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
241
+ page_content=' In 2010 Third IEEE International Conference on Digital Game and Intelligent Toy Enhanced Learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
242
+ page_content=' 10–22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
243
+ page_content=' [4] Christothea Herodotou, Niall Winters, and Maria Kambouri.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
244
+ page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
245
+ page_content=' An Iterative, Multidisciplinary Approach to Studying Digital Play Motivation: The Model of Game Motivation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
246
+ page_content=' Games and Culture 10, 3 (May 2015), 249–268.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
247
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
248
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
249
+ page_content='1177/1555412014557633 [5] Hyun-Woo Lee, Sanghoon Kim, and Jun-Phil Uhm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
250
+ page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
251
+ page_content=' Social Virtual Reality (VR) Involvement Affects Depression When Social Connectedness and Self-Esteem Are Low: A Moderated Mediation on Well-Being.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
252
+ page_content=' Frontiers in Psychology 12 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
253
+ page_content=' https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
254
+ page_content='frontiersin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
255
+ page_content='org/articles/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
256
+ page_content='3389/ fpsyg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
257
+ page_content='2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
258
+ page_content='753019 [6] Meta.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
259
+ page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
260
+ page_content=' Poker VR - Multi Table Tournaments on Oculus Quest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
261
+ page_content=' https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
262
+ page_content='oculus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
263
+ page_content='com/experiences/quest/2257223740990488/ [7] Fast Offshore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
264
+ page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
265
+ page_content=' Online poker sector overview for 2021: Stats, key drivers and more.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
266
+ page_content=' https://fastoffshore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
267
+ page_content='com/2021/10/online-poker-sector- overview-2021/ [8] Arjun Rohlfing-Das.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
268
+ page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
269
+ page_content=' Image Classification for Playing Cards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
270
+ page_content=' https://medium.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
271
+ page_content='com/swlh/image-classification-for-playing-cards-26d660f3149e [9] Hiroyuki Sakuma, Tetsuo Yamabe, and Tatsuo Nakajima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
272
+ page_content=' 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
273
+ page_content=' Enhancing Traditional Games with Augmented Reality Technologies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
274
+ page_content=' In 2012 9th International Conference on Ubiquitous Intelligence and Computing and 9th International Conference on Autonomic and Trusted Computing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
275
+ page_content=' 822–825.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
276
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
277
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
278
+ page_content='1109/UIC-ATC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
279
+ page_content='2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
280
+ page_content='95 [10] Nina Savela, Atte Oksanen, Markus Kaakinen, Marius Noreikis, and Yu Xiao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
281
+ page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
282
+ page_content=' Does Augmented Reality Affect Sociability, Entertainment, and Learning?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
283
+ page_content=' A Field Experiment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
284
+ page_content=' Applied Sciences 10, 4 (Jan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
285
+ page_content=' 2020), 1392.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
286
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
287
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
288
+ page_content='3390/app10041392 Number: 4 Publisher: Multidisciplinary Digital Publishing Institute.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
289
+ page_content=' [11] Snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
290
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
291
+ page_content=' Connected Lenses Overview | Docs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
292
+ page_content=' https://docs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
293
+ page_content='snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
294
+ page_content='com/lens-studio/references/guides/lens-features/connected-lenses/connected- lenses-overview [12] Snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
295
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
296
+ page_content=' How do I use Lenses on Snapchat?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
297
+ page_content=' https://support.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
298
+ page_content='snapchat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
299
+ page_content='com/en-US/a/face-world-lenses [13] Snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
300
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
301
+ page_content=' Spectacles by Snap Inc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
302
+ page_content=' • The Next Generation of Spectacles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
303
+ page_content=' https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
304
+ page_content='spectacles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
305
+ page_content='com/ [14] Snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
306
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
307
+ page_content=' Tracking Modes | Docs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
308
+ page_content=' https://docs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
309
+ page_content='snap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
310
+ page_content='com/lens-studio/references/guides/lens-features/tracking/world/tracking-modes [15] Christoph Thul.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
311
+ page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
312
+ page_content=' PokerTool - Entwicklung und Implementierung einer AR-Android-Anwendung für Wahrscheinlichkeitsberechnungen bei Texas Holdem Poker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
313
+ page_content=' (Sept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
314
+ page_content=' 2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
315
+ page_content=' https://kola.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
316
+ page_content='opus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
317
+ page_content='hbz-nrw.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
318
+ page_content='de/opus45-kola/frontdoor/index/index/docId/769 [16] Mark Weiser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
319
+ page_content=' 1991.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
320
+ page_content=' The Computer for the 21st Century.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
321
+ page_content=' (1991).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
322
+ page_content=' Manuscript submitted to ACM PokAR: Facilitating Poker Play Through Augmented Reality 11 A STUDY PROTOCOL Below is the process we asked participants to follow during the study: (1) Participants were found through a poker club on campus and recruited by email.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
323
+ page_content=' (2) Participants were asked to respond to the pre-study survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
324
+ page_content=' (3) The participants were randomly paired into dyads for heads-up poker play.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
325
+ page_content=' (4) During the study: (a) Participants were asked to download Snapchat (if necessary) and scan a code to gain access to the PokAR Snapchat Lens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
326
+ page_content=' (b) Participants were asked to sign consent forms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
327
+ page_content=' (c) Participants were asked to use PokAR to play heads-up poker (without using real-world money) for 25 minutes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
328
+ page_content=' (d) We took notes on comments, reactions, game pace, frustrations, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
329
+ page_content=' We took photos and videos throughout.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
330
+ page_content=' We also answered questions about the application when asked, but we avoided guiding the players.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
331
+ page_content=' (5) After play, participants were asked to respond to the post-study survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
332
+ page_content=' B PRE-STUDY SURVEY Below are the questions asked during the pre-study survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
333
+ page_content=' How well do you know the rules of Heads-Up Texas Hold’em Poker?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
334
+ page_content=' How often do you play poker?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
335
+ page_content=' C POST-STUDY SURVEY Below are the questions asked during the post-study survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
336
+ page_content=' Please rate each of the following PokAR features in terms of its effectiveness when compared to the corresponding object/action in real-life Texas hold’em poker?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
337
+ page_content=' – 3D AR Chips – UI Action Buttons – Game Messages – Counting Stacks – Awarding Pots How much did AR affect the in-person social aspects of the game of poker?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
338
+ page_content=' How did augmented reality affect the game pace of poker (# of hands played / unit time)?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
339
+ page_content=' Ignore the first few hands in which you were learning the application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
340
+ page_content=' Overall, how would you describe your experience with PokAR?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
341
+ page_content=' D CODE REPOSITORY The code for this project can be found at the following GitHub repository: https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
342
+ page_content='com/adamgamba/PokAR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
343
+ page_content=' Manuscript submitted to ACM' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/0tAyT4oBgHgl3EQfoPjL/content/2301.00505v1.pdf'}
19E2T4oBgHgl3EQfiwfE/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76d609a55f97acd9c9112ef5a22af45c952d34d5e3baf8dd1600ce6b99c44230
3
+ size 272976
1NAyT4oBgHgl3EQfbff_/content/tmp_files/2301.00266v1.pdf.txt ADDED
The diff for this file is too large to render. See raw diff
 
1NAyT4oBgHgl3EQfbff_/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
1NAzT4oBgHgl3EQfRftD/content/2301.01216v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13fa9d2c3c91c839887c8e3eb1f0f55d0e85cd3ba0479c06328e635e64eab2be
3
+ size 1057369
1NAzT4oBgHgl3EQfRftD/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06f2b8f59ebcaaa5408ff7e846c2cf6eef89c5dd1e8f737236fe82122409d38e
3
+ size 3538989
1NAzT4oBgHgl3EQfRftD/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fec5845204ca36d84ab523b9c316c2a1f0c39ba37c2ab5b7e437c0348f064f0
3
+ size 128032
3tAzT4oBgHgl3EQfD_po/content/2301.00985v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0d48e1ba1e49c96139213477fd76d427107e6d0645625dc4f591b087034303a
3
+ size 9894094
3tAzT4oBgHgl3EQfD_po/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93e1dd07c64152f5e77f8e193147f459aa883e3f69dc7545ebca7191ab50efd9
3
+ size 6422573
49E0T4oBgHgl3EQfegBh/content/tmp_files/2301.02391v1.pdf.txt ADDED
@@ -0,0 +1,1561 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ arXiv:2301.02391v1 [math.NT] 6 Jan 2023
2
+ On effective irrationality exponents of cubic irrationals
3
+ Dzmitry Badziahin
4
+ January 9, 2023
5
+ Abstract
6
+ We provide an upper bound on the efficient irrationality exponents of cubic algebraics
7
+ x with the minimal polynomial x3 − tx2 − a. In particular, we show that it becomes
8
+ non-trivial, i.e. better than the classical bound of Liouville in the case |t| > 19.71a4/3.
9
+ Moreover, under the condition |t| > 86.58a4/3, we provide an explicit lower bound on the
10
+ expression ||qx|| for all large q ∈ Z. These results are based on the recently discovered
11
+ continued fractions of cubic irrationals [1] and improve the currently best-known bounds
12
+ of Wakabayashi.
13
+ Keywords: cubic irrationals, continued fractions, irrationality exponent, effective irrationality exponent
14
+ Math Subject Classification 2010: 11J68, 11J70, 11J82
15
+ 1
16
+ Introduction
17
+ The irrationality exponent λ(x) of an irrational real number x is defined as the supremum of
18
+ real numbers λ such that the inequality
19
+ ����x − p
20
+ q
21
+ ���� < q−λ
22
+ (1)
23
+ has infinitely many rational solutions p/q. It follows from the classical Dirichlet theorem that
24
+ for all x ∈ R \ Q, λ(x) ⩾ 2. On the other hand, Khintchine’s theorem implies that almost
25
+ all x ∈ R with respect to the Lebesgue measure satisfy λ(x) = 2. In the first half of the
26
+ XX century, there was a big interest in estimating the irrationality exponent of real algebraic
27
+ numbers. It culminated in 1955 with the work of Roth [8], who established the best possible
28
+ result, i.e. that for any algebraic x ∈ R\Q, λ(x) = 2. Unfortunately, that result is ineffective,
29
+ i.e. for λ > 2 it does not allow us to find all rational p/q that satisfy (1). Therefore, for
30
+ example, it can not be used to solve the Thue equations
31
+ F(p, q) = c
32
+ in integer p, q, where F ∈ Z[x, y] is a homogeneous polynomial of degree d ⩾ 3 and c is
33
+ some integer parameter. Since then, many mathematicians were working on effective results
34
+ regarding the irratoinality exponents of algebraic numbers.
35
+ Given x ∈ R \ Q, by the effective irrationality exponent of x we define a positive real
36
+ number λeff(x) such that for all λ > λeff(x) there exists an effectively computable Q > 0
37
+ such that all rational solutions of the inequality (1) in reduced form satisfy q ⩽ Q.
38
+ All known upper bounds on λeff(x) for algebraic real x are much weaker than in Roth’s
39
+ theorem. First of all, the classical theorem of Liouville states that λeff(x) ⩽ d where d is
40
+ the degree of x. Therefore any non-trivial bound on λeff(x) should be strictly smaller than
41
+ 1
42
+
43
+ d. One of the notable improvements of Liouville’s bound is based on Feldman’s refinement of
44
+ the theory of linear forms in logarithms [6]. Its advantage is that it gives λeff(x) < d for all
45
+ algebraic numbers. However, the difference between λeff(x) and d is usually extremely small.
46
+ For state-of-the-art results regarding this approach, we refer to the book of Bugeaud [5]. For
47
+ other notable achievements on this problem, we refer to [3, 4] and the references therein.
48
+ In this paper, we focus on the case of cubic irrationals.
49
+ Multiplying by some integer
50
+ number and shifting by another rational number, we can always guarantee that the minimal
51
+ polynomial of the resulting cubic x is x3 + px + q for some p, q ∈ Z. Notice also that such a
52
+ transformation does not change the (effective) irrationality exponent of x. The first general
53
+ result about λeff(x) for these specific values x was achieved by Bombieri, van der Poorten
54
+ and Vaaler [4] in 1996. They showed that under the conditions |p| > e1000 and |p| ⩾ q2, one
55
+ has
56
+ λeff(x) ⩽ 2 log(|p|3)|
57
+ log(|p|3/q2) +
58
+ 14
59
+ (log(|p|3/q2))1/3 .
60
+ Later, Wakabayashi [9] improved that bound and showed that λeff(x) ⩽ λw(p, q). It becomes
61
+ non-trivial (i.e. smaller than 3) under the condition
62
+ |p| > 22/334|q|8/3
63
+
64
+ 1 +
65
+ 1
66
+ 390|q|3
67
+ �2/3
68
+ (2)
69
+ and for large p and q it asymptotically behaves like λw(p, q)
70
+
71
+ 2 + (4 log |q| +
72
+ 2 log 108)/(3 log |p|).
73
+ In this paper we investigate what estimates on λeff(x) can be achieved with help of the
74
+ convergents of the recently discovered continued fractions [1] of cubic irrationals with the
75
+ minimal polynomial x3 − tx2 − a ∈ Z[x]. It was shown there that, as soon as |t|3 > 12a > 0,
76
+ the real root of this equation with the largest absolute value admits the continued fraction
77
+ x = K
78
+
79
+ t
80
+ 3(12k + 1)(3k + 1)α
81
+ 3(12k + 5)(3k + 2)α
82
+ 3(12k + 7)(6k + 5)α
83
+ 3(12k − 1)(6k + 1)α
84
+ (2i + 1)t2
85
+ (2i + 1)t
86
+ 2(2i + 1)t2
87
+ (2i + 1)t
88
+
89
+ .
90
+ Here i is the index of the corresponding partial quotient and k =
91
+ � i
92
+ 4
93
+
94
+ . Notice that the change
95
+ of variables y = q
96
+ x transforms cubic x from [4, 9] to numbers in this paper with t = −p and
97
+ a = −q2.
98
+ The resulting upper bounds on λeff(x) (see Theorems 1 and 2) depend on prime factori-
99
+ sations of a and t but in any case they are better than those in [9]. Theorem 1 states that
100
+ under the condition |t|3 > 12|a| the largest real root of the above cubic equation satisfies
101
+ ||qx|| ⩾ τ(t, a)q−λ(t,a)(log(8|t1|q))λ(t,a)−1/2,
102
+ for all q > Q0(t, a),
103
+ where all values for τ(t, a), λ(t, a) and q0(t, a) are explicitly provided. This inequality always
104
+ gives a non-trivial upper bound on λeff(x) under the condition |t| > 86.58a4/3. It translates
105
+ to the condition |p| > 86.58|q|8/3 which is better than (2), where even without the term in
106
+ brackets we have |p| > 128.57|q|8/3. Moreover, the parameter a does not have to be of the
107
+ form −q2, therefore many cubics from this paper do not plainly transfer to those in [4, 9].
108
+ On top of that, Theorem 2 provides an even better upper bound on λeff(x) but does not
109
+ give an explicit lower bound on ||qx|| as above. That bound always becomes non-trivial as
110
+ soon as |t| > 19.71a4/3. These bounds between t and a are obtained in Section 3.
111
+ 2
112
+ Preliminaries and main results
113
+ Consider the real root of the equation x3 − tx2 − a = 0 that has the largest absolute value
114
+ among other roots of the same equation. Notice that a ̸= 0 because otherwise x is not cubic.
115
+ Also, by replacing x with −x if needed, we can guarantee that a > 0.
116
+ 2
117
+
118
+ Next, by the standard CF transformations (see [1, Lemma 1]), we can cancel some
119
+ common divisors of t and 3a from the continued fraction of x. Let g1 := gcd(t2, 3a) and
120
+ g2 := gcd(t, 3a/g1). For convenience, we denote t2 = g1t2, t = g2t1 and 3a = g1g2a∗. We di-
121
+ vide the following partial quotients of x by g1: β1, a1, β2; β3, a3, β4; . . . , β2k−1, a2k−1, β2k,. . . .
122
+ After that we divide the following partial quotients by g2:
123
+ β0, a0, β1; β2, a2, β3; . . . ;
124
+ β2k, a2k, β2k+1,. . . The resulting continued fraction has the same limit x as the initial one.
125
+ To make β0 integer, we consider the number x/g2 instead of x. Its continued fraction is then
126
+ K
127
+
128
+ t1
129
+ (12k + 1)(3k + 1)a∗
130
+ (12k + 5)(3k + 2)a∗
131
+ (12k + 7)(6k + 5)a∗
132
+ (12k − 1)(6k + 1)a∗
133
+ (2i + 1)t2
134
+ (2i + 1)t1
135
+ 2(2i + 1)t2
136
+ (2i + 1)t1
137
+ · · ·
138
+
139
+ (3)
140
+ We define the following notions:
141
+ c6 = c6(t1, t2, a∗) :=
142
+
143
+
144
+
145
+
146
+
147
+
148
+
149
+
150
+
151
+ √64t1t2 + 270a∗
152
+ 27/4ec1
153
+ if t > 0
154
+
155
+ 64|t1t2| − 54a∗
156
+ 27/4ec1
157
+ if t < 0,
158
+ (4)
159
+ c7 = c7(t1, t2, a∗) :=
160
+
161
+
162
+
163
+
164
+
165
+
166
+
167
+
168
+
169
+
170
+
171
+ 27/4ec1(16t1t2 + 9a∗)
172
+ 9a∗√64t1t2 + 270a∗
173
+ if t > 0
174
+ 223/4ec1(|t1t2| − 3a∗)
175
+ 9a∗�
176
+ 64|t1t2| − 54a∗
177
+ if t < 0,
178
+ (5)
179
+ where c1 = c1(a∗) is defined in (10). Next,
180
+ τ = τ(t1, t2, a∗) := g2 · (log c7)1/2
181
+ 8c8
182
+ 6(8|t1|)
183
+ log c6
184
+ log c7
185
+ ;
186
+ q0 = q0(t1, t2, a∗) :=
187
+ c4
188
+ 7
189
+ 8|t1|.
190
+ (6)
191
+ The main result of this paper is
192
+ Theorem 1 For all integer q ⩾ q0 one has
193
+ ||qx|| > τq−λ log(8|t1|q)−λ−1/2
194
+ where λ = log c6
195
+ log c7. In particular, λeff(x) ⩽ λ.
196
+ As shown in Section 5, the constant c1 can be replaced by a bigger constant c2 defined
197
+ in (11). But in that case, writing such an explicit inequality as in Theorem 1 is much harder
198
+ (but theoretically possible). We do not provide it in the exact form here but only state the
199
+ following result. Let c∗
200
+ 6 and c∗
201
+ 7 be defined in the same way as c6 and c7 but with the constant
202
+ c2 instead of c1.
203
+ Theorem 2 The effective irrationality exponent of x satisfies
204
+ λeff(x) ⩽ λ∗ := log c∗
205
+ 6
206
+ log c∗
207
+ 7
208
+ .
209
+ 3
210
+ Analysis of the results
211
+ Theorem 1 provides a nontrivial lower bound for ||qx|| as soon as log c6
212
+ log c7 is strictly less that 2
213
+ or in other words, c6 < c2
214
+ 7. In view of (4) and (5), for t > 0 this is equivalent to
215
+ √64t1t2 + 270a∗
216
+ 27/4ec1
217
+ < 27/2e2c2
218
+ 1(16t1t2 + 9a∗)2
219
+ 81a∗2(64t1t2 + 270a∗)
220
+ 3
221
+
222
+ ⇐⇒
223
+ a∗2(64t1t2 + 270a∗)3/2 < 221/4e3c3
224
+ 1
225
+ 81
226
+ (16t1t2 + 9a∗)2.
227
+ Define the parameter u such that 16t1t2 = ua∗4 − 9a∗. Also for convenience define τ :=
228
+ 221/4e3c3
229
+ 1
230
+ 81
231
+ . Then the last inequality rewrites
232
+ a∗2(4ua∗4 + 234a∗)3/2 < τu2a∗8
233
+ ⇐⇒
234
+ 4ua∗3 + 234 < τ 2/3u4/3a∗3.
235
+ Notice that for u ⩾
236
+
237
+ 4
238
+ τ 2/3 +
239
+ 234
240
+ 64a∗3 τ 4/3�3
241
+ one has
242
+ τ 2/3u4/3a∗3 ⩾
243
+
244
+ 4 + 234
245
+ 64a∗3 τ 2
246
+
247
+ · ua∗3 ⩾ 4ua∗3 + 234τ 2
248
+ 64a∗3 · 43
249
+ τ 2 a∗3 = 4ua∗3 + 234
250
+ and the condition c6 < c2
251
+ 7 is satisfied.
252
+ Recall that t1t2 = t3/(g1g2) and a∗ = 3a/(g1g2).
253
+ Therefore, 16t1t2 > ua∗4 − 9a∗ is
254
+ equivalent to 16t3 >
255
+ 81
256
+ (g1g2)3 ua4 − 27a.
257
+ From the definition (10) we see that c1 and hence τ depends on the prime factorisation
258
+ of a∗. If 3 ∤ a∗ then we always get c1 > 0.0924 which in turn implies τ > 0.00744 and
259
+ � 4
260
+ τ 2/3 + 234
261
+ 64a∗3 τ 4/3
262
+ �3
263
+ ⩽ 104.973.
264
+ On the other hand, we have g1g2 ⩾ 3.
265
+ Finally, we get that for t > 0 the non-trivial bound on λeff(x) is always achieved if
266
+ t > 104.97
267
+ 3
268
+ ·
269
+ � 81
270
+ 16
271
+ �1/3 a4/3 ≈ 60.08a4/3, however for many pairs a and t it is satisfied under
272
+ essentially weaker conditions.
273
+ In the case 3 | a∗ we get c1 > 0.13329, thus τ > 0.0223 and
274
+ � 4
275
+ τ 2/3 + 234
276
+ 64a∗3 τ 4/3
277
+ �3
278
+ ⩽ 50.423.
279
+ Then the non-trivial bound is achieved if t > 50.42 ·
280
+ � 81
281
+ 16
282
+ �1/3 a4/3 ≈ 86.57a4/3.
283
+ The case t < 0 is dealt analogously. The condition c6 < c2
284
+ 7 is equivalent to
285
+
286
+ 64|t1t2| − 54a∗
287
+ 27/4ec1
288
+ < 27/2e2c2
289
+ 1(16|t1t2| − 48a∗)2
290
+ 81a∗2(64|t1t2| − 54a∗)
291
+ ⇐⇒
292
+ a∗2(64|t1t2| − 54a∗)3/2 < τ(16|t1t2| − 48a∗)2.
293
+ Define u such that 16|t1t2| = ua∗4 + 48a∗. Then the last inequality rewrites
294
+ 4ua∗3 + 138 < τ 2/3u4/3a∗3.
295
+ One can check that the last inequality is satisfied for u ⩾
296
+
297
+ 4
298
+ τ 2/3 + 138τ 4/3
299
+ 64a∗3
300
+ �3
301
+ . As in the case of
302
+ positive t, the right hand side is always smaller than 104.933 in the case 3 ∤ a∗ and is smaller
303
+ than 50.423 in the case 3 | a∗. Therefore for 3 ∤ a∗ the condition c6 < c2
304
+ 7 is always satisfied
305
+ if |t|3 > 104.933·81
306
+ 33·16
307
+ a4 + 9a which follows from |t| > 60.06a4/3. For 3 | a∗, similar computations
308
+ give us |t|3 > 50.423·81
309
+ 16
310
+ a4 + 9a which follows from |t| > 86.58a4/3.
311
+ One can repeat the same analysis as above for Theorem 2. In that case, the constant
312
+ c1 in the computations should be replaced by c2. We have that if 3 ∤ a∗ it always satisfies
313
+ c2 ⩾ 0.1939, which in turn implies τ ⩾ 0.0688 and then
314
+ � 4
315
+ τ 2/3 + 234
316
+ 64a∗3 τ 4/3
317
+ �3
318
+ ⩽ 23.933.
319
+ 4
320
+
321
+ If 3 | a∗, we have c2 ⩾ 0.2797, τ ⩾ 0.206 and
322
+ � 4
323
+ τ 2/3 + 234
324
+ 64a∗3 τ 4/3
325
+ �3
326
+ ⩽ 11.473.
327
+ Finally, for the case 3 ∤ a∗, the condition c6 < c2
328
+ 7 is always satisfied if |t|3 > 23.933·81
329
+ 33·16
330
+ a4 +9a
331
+ which follows from |t| > 13.72a4/3. For the case 3 | a∗ similar calculations give |t| > 19.71a4/3.
332
+ 4
333
+ Nice, convenient and perfect continued fractions
334
+ Definition 1 Let x be a continued fraction given by
335
+ x = K
336
+ � β0
337
+ β1
338
+ β2
339
+ a0
340
+ a1
341
+ a2 · · ·
342
+
343
+ ;
344
+ βi, ai ∈ Z, ∀i ∈ Z⩾0.
345
+ For given positive integers k, r with −1 ⩽ r ⩽ k we define
346
+ γk,r :=
347
+ k+r
348
+
349
+ i=k−r
350
+ 2|(i−k+r)
351
+ βi,
352
+ γk,−1 := 1.
353
+ We say that x is d-nice at index k where d, k ∈ N if d | ak and for all positive integer r ⩽ k
354
+ one has ak−rβk+rγk,r−2 ≡ −ak+rγk,r−1 (mod d). We call x (d, r)-perfect at index k, where
355
+ 0 ⩽ r ⩽ k if it is d-nice at index k and βk−r ≡ βk+r+1 ≡ 0 (mod d).
356
+ We say that x is eventually d-nice at index k from position k0 if the same conditions
357
+ as above are satisfied for all 0 ⩽ r ⩽ k − k0. In this paper the value k0 will often be a fixed
358
+ absolute constant. If there is no confusion about its value we will omit it in the text.
359
+ Let pn/qn be the n’th convergent of x. Define the following matrices
360
+ Sn :=
361
+
362
+ pn
363
+ qn
364
+ pn−1
365
+ qn−1
366
+
367
+ ;
368
+ Cn :=
369
+ � an
370
+ βn
371
+ 1
372
+ 0
373
+
374
+ .
375
+ From the theory of continued fractions we know that Sn = CnCn−1 · · · C0. To make this
376
+ product shorter, we use the usual notation but in the descending order: Sn = �0
377
+ i=n Ci.
378
+ Lemma 1 Let the continued fraction x be eventually d-nice at index k from the position k0.
379
+ Then for all 0 ⩽ r ⩽ k − k0 one has
380
+ k−r
381
+
382
+ i=k+r
383
+ Ci ≡
384
+
385
+ 0
386
+ γk,r
387
+ γk,r−1
388
+ 0
389
+
390
+ (mod d).
391
+ Moreover, if x is (d, r)-perfect at index k then �k−r
392
+ i=k+r+1 Ci ≡ 0 (mod d).
393
+ Proof.
394
+ We prove by induction on r. For r = 0 the statement is straightforward. Suppose that
395
+ the statement is true for r and verify it for r + 1.
396
+ k−r−1
397
+
398
+ i=k+r+1
399
+ Ci
400
+
401
+ � ak+r+1
402
+ βk+r+1
403
+ 1
404
+ 0
405
+ � �
406
+ 0
407
+ γk,r
408
+ γk,r−1
409
+ 0
410
+ � � ak−r−1
411
+ βk−r−1
412
+ 1
413
+ 0
414
+
415
+
416
+ � ak−r−1βk+r+1γk,r−1 + ak+r+1γk,r
417
+ βk−r−1βk+r+1γk,r−1
418
+ γk,r
419
+ 0
420
+
421
+ (mod d).
422
+ 5
423
+
424
+ By the conditions of d-nice CF at index k, the last matrix is congruent to
425
+
426
+ 0
427
+ γk,r+1
428
+ γk,r
429
+ 0
430
+
431
+ .
432
+ If x is (d, r)-perfect at index k then γk,r ≡ 0 (mod d) and we have
433
+ k−r
434
+
435
+ i=k+r+1
436
+ Ci ≡
437
+ � ak+r+1
438
+ 0
439
+ 1
440
+ 0
441
+ � �
442
+ 0
443
+ γk,r
444
+ γk,r−1
445
+ 0
446
+
447
+ ≡ 0 (mod d).
448
+
449
+ Another two properties of d-nice continued fractions that easily follow from the definition
450
+ are
451
+ • Let d1, d2 be two coprime positive integer numbers. If a continued fraction is eventually
452
+ d1-nice and eventually d2-nice at the same index k for the same position k0 then it is
453
+ eventually d1d2-nice at index k.
454
+ • If a continued fraction is eventually d-nice at index k then it is also eventually e-nice
455
+ at the same index from the same position for all positive integer divisors e of d.
456
+ Definition 2 We say that the continued fraction x is (eventually) d-convenient at index k if
457
+ there exists a sequence (cr)0⩽r⩽⌊k/2⌋ of residues modulo m such that for all positive integers
458
+ r ⩽ k (resp. r ⩽ k − k0) one has
459
+ • βk+r+1 ≡ c⌊ r
460
+ 2 ⌋βk−r (mod d);
461
+ • if r is odd then ak+r ≡ −c⌊ r
462
+ 2⌋ak−r (mod d);
463
+ • if r is even then ak+r ≡ −ak−r (mod d).
464
+ Lemma 2 Let d > 2. Then any eventually d-convenient continued fraction at index k is
465
+ also eventually d-nice. For d = 2, any d-convenient continued fraction at index k such that
466
+ ak ≡ 0 (mod d) is also d-nice.
467
+ Proof. First of all, for d > 2 and r = 0 the condition ak+r ≡ −ak−r (mod d) implies that
468
+ ak ≡ 0 (mod d), which is the first condition of d-nice CF.
469
+ Secondly, one can check that the first condition of d-convenient CF implies that for odd
470
+ r, γk,r ≡ c⌊ r
471
+ 2 ⌋γk,r−1βk−r ≡ γk,r−1βk+r+1 (mod d). Then we get
472
+ ak−r−1βk+r+1γk,r−1 ≡ ak−r−1γk,r ≡ −ak+r+1γk,r (mod d)
473
+ and the second condition of d-nice CF is verified.
474
+ Thirdly, for even r we get cr/2γk,r ≡ cr/2γk,r−1βk−r ≡ γk,r−1βk+r+1 (mod d) and therefore
475
+ ak−r−1βk+r+1γk,r−1 ≡ cr/2ak−r−1γk,r ≡ −ak+r+1γk,r (mod d).
476
+ Again, the second condition of d-nice CF is satisfied.
477
+
478
+ 5
479
+ Divisibility of entries of Sn
480
+ Lemma 3 Let k ∈ N, k ⩾ 2 and d be any integer divisor of 2k+1. The continued fraction (3)
481
+ is eventually d-convenient at index k from the position 2. Additionally, the same statement
482
+ is true for k ≡ 3 (mod 4) and d = 2.
483
+ 6
484
+
485
+ In the further discussion we will always deal with eventually d-convenient or d-nice contin-
486
+ ued fractions from the position 2. Therefore, to make the description shorter, we will omit the
487
+ words ‘eventually’ and ‘from the position 2’ and call the continued fraction (3) d-convenient
488
+ or d-nice.
489
+ Proof. We will check the conditions of d-convenient continued fraction separately for
490
+ each of the cases, depending on k modulo 4.
491
+ Case k = 4m + 1. Then m ≡ − 3
492
+ 8 (mod d) and we use (3) to compute
493
+ ak+4r ≡ 8rt2,
494
+ βk+4r = a∗(12m + 1 + 12r)(3m + 1 + 3r) ≡ a∗
495
+ 16(24r − 7)(24r − 1) (mod d);
496
+ ak+4r+1 ≡ (8r + 2)t1,
497
+ βk+4r+1 ≡ a∗
498
+ 16(24r + 1)(24r + 7) (mod d);
499
+ ak+4r+2 ≡ 2(8r + 4)t2,
500
+ βk+4r+2 ≡ a∗
501
+ 8 (24r + 5)(24r + 11) (mod d);
502
+ ak+4r−1 ≡ (8r − 2)t1,
503
+ βk+4r−1 ≡ a∗
504
+ 8 (24r − 11)(24r − 5) (mod d).
505
+ The conditions of d-convenient continued fraction at index k can now be easily checked where
506
+ cr is the constant 1 sequence.
507
+ We proceed the same way in all other cases.
508
+ Case k = 4m + 2. Then m ≡ − 5
509
+ 8 (mod d) and
510
+ ak+4r ≡ 8rt1,
511
+ βk+4r ≡ a∗
512
+ 16(24r − 5)(24r + 1) (mod d);
513
+ ak+4r+1 ≡ 2(8r + 2)t2,
514
+ βk+4r+1 ≡ a∗
515
+ 8 (24r − 1)(24r + 5) (mod d);
516
+ ak+4r+2 ≡ (8r + 4)t1,
517
+ βk+4r+2 ≡ a∗
518
+ 8 (24r + 7)(24r + 13) (mod d);
519
+ ak+4r−1 ≡ (8r − 2)t2,
520
+ βk+4r−1 ≡ a∗
521
+ 16(24r − 13)(24r − 7) (mod d).
522
+ One can easily check that for s ≡ 0, 1 (mod 4), βk+s+1 ≡ 2βk−s (mod d) and for s ≡ 2, 3
523
+ (mod 4), βk+s+1 ≡ 2−1βk−s (mod d). Also, for s ≡ 1 (mod 4), ak+s ≡ 2ak−s (mod d) and
524
+ for s ≡ 3 (mod 4), ak+s ≡ 2−1ak−s (mod d). Hence, the conditions of d-convenient CF are
525
+ verified, where the sequence cs is periodic with the period 2, 2−1.
526
+ Case k = 4m + 3, d ̸= 2. Then m ≡ − 7
527
+ 8 (mod d) and
528
+ ak+4r ≡ 16rt2,
529
+ βk+4r ≡ a∗
530
+ 8 (24r − 7)(24r − 1) (mod d);
531
+ ak+4r+1 ≡ (8r + 2)t1,
532
+ βk+4r+1 ≡ a∗
533
+ 8 (24r + 1)(24r + 7) (mod d);
534
+ ak+4r+2 ≡ (8r + 4)t2,
535
+ βk+4r+2 ≡ a∗
536
+ 16(24r + 5)(24r + 11) (mod d);
537
+ ak+4r−1 ≡ (8r − 2)t1,
538
+ βk+4r−1 ≡ a∗
539
+ 16(24r − 11)(24r − 5) (mod d).
540
+ One can then check the conditions of d-convenient CF at index k for the constant 1 sequence
541
+ cr.
542
+ Case k = 4m + 3, d = 2.
543
+ in this case one can easily see that ak+4r ≡ 0 (mod 2),
544
+ ak+4r+1 ≡ ak+4r+3 ≡ t1 (mod 2), ak+4r+2 ≡ t2 (mod 2); βk+4r ≡ βk+4r+1 ≡ a∗ (mod 2) and
545
+ 7
546
+
547
+ βk+4r+2 ≡ βk+4r−1 (mod 2). Therefore, the CF is 2-convenient at index k with the constant
548
+ 1 sequence cr.
549
+ Case k = 4m. Then m ≡ − 1
550
+ 8 (mod d) and
551
+ ak+4r ≡ 8rt1,
552
+ βk+4r ≡ a∗
553
+ 8 (24r − 5)(24r + 1) (mod d);
554
+ ak+4r+1 ≡ (8r + 2)t2,
555
+ βk+4r+1 ≡ a∗
556
+ 16(24r − 1)(24r + 5) (mod d);
557
+ ak+4r+2 ≡ (8r + 4)t1,
558
+ βk+4r+2 ≡ a∗
559
+ 16(24r + 7)(24r + 13) (mod d);
560
+ ak+4r−1 ≡ 2(8r − 2)t2,
561
+ βk+4r−1 ≡ a∗
562
+ 8 (24r − 13)(24r − 7) (mod d).
563
+ One can then check the conditions of d-convenient CF with the periodic sequence cr with the
564
+ period 2−1, 2.
565
+
566
+ Lemmata 2 and 3 show that x is d-nice at each index k ⩾ 2 for appropriately chosen d.
567
+ As the next step, we show that for almost every prime p, it is also (p, t)-perfect at infinitely
568
+ many carefully chosen indices k and t. This fact will allow us to show that all the entries of
569
+ CkCk−1 · · · C1 are multiples of some big power of p.
570
+ First of all, let’s consider the case p > 2 and p | a∗. Let s ∈ N be such that ps||a∗.
571
+ Consider q = pl for some 1 ⩽ l ⩽ s.
572
+ If we write q = 2m + 1 then we have q | αk for
573
+ k = m + rq = (2r+1)q−1
574
+ 2
575
+ where r ∈ Z⩾0. One can easily see that for any such value of k, x is
576
+ (q, 0)-perfect at index k. In view of Lemma 1, we can then split the product Sk into
577
+
578
+ 2k+q−1
579
+ 2q
580
+
581
+ groups such that all entries of the resulting product matrix in each group are multiples of q.
582
+ Finally, we combine this information for ql for all 1 ⩽ l ⩽ s and derive that all entries of the
583
+ product �2
584
+ i=k Ci are divisible by
585
+ p
586
+ s�
587
+ i=1
588
+
589
+ 2k+pi−1
590
+ 2pi
591
+
592
+ .
593
+ Next, consider the case p = 2 and p | a∗. We have p | ak for all k ≡ 3 (mod 4) and one can
594
+ easily see that for all such k, x is (p, 0) perfect at index k. Then the analogous application
595
+ of Lemma 1 as in the previous case implies that all entries of �2
596
+ i=k Ci are divisible by 2⌊k/4⌋.
597
+ For the case p = 2, p ∤ a∗ the result is slightly weaker. From (3) one can verify that
598
+ β8m+2 ≡ β8m+5 ≡ 0 (mod 2) for all m ∈ Z⩾0 and therefore x is (2, 1)-perfect at indices
599
+ 8m + 3. Then Lemma 1 then implies that all entries of �2
600
+ i=k Ci are divisible by 2⌊(k+3)/8⌋.
601
+ Finally, in the next lemma we consider the remaining case of p ∈ N that do not divide a∗.
602
+ Lemma 4 Let p ∈ N be such that gcd(p, 6) = 1. Then for all k ∈ Z all the entries of the
603
+ product of matrices �2
604
+ i=k Ci are divisible by
605
+ p
606
+
607
+ 3k+p−2
608
+ 3p
609
+
610
+ .
611
+ Proof. We prove by routinely considering all the cases, depending on p modulo 12.
612
+ Case p = 12m + 1. Then with help of (3) one can verify that for all r ∈ Z⩾0,
613
+ 0
614
+ ≡ β4(m+rp)+1 ≡ β4(2m+rp) ≡ β4(4m+rp)+1 ≡ β4(5m+rp)+2 ≡ β4(7m+rp)+3 ≡ β4(8m+rp)+2
615
+ ≡ β4(10m+rp)+3 ≡ β4(11m+rp)+4 (mod p)
616
+ 8
617
+
618
+ and 0 ≡ a6m+rp (mod p). In view of Lemma 3, we then derive that x is (p, 2m − 1)-perfect
619
+ at indices k = 6m + 2rp for all r ∈ Z⩾0 and is (p, 2m)-perfect at indices k = 6m + (2r + 1)p.
620
+ Lemma 1 then implies that all the entries of the following products of matrices are divisible
621
+ by p:
622
+ 4m+2rp+1
623
+
624
+ i=8m+2rp
625
+ Ci,
626
+ 16m+4rp+1
627
+
628
+ i=20m+2rp+2
629
+ Ci.
630
+ Finally, one can easily check that for k = (n + 1)p − p−1
631
+ 3 , the product �2
632
+ i=k Ci contains n + 1
633
+ blocks of the above form. Therefore all its entries are divisible by pn+1.
634
+ The other cases are done analogously.
635
+ Case p = 12m + 5. One verifies that for all r ∈ Z⩾0,
636
+ 0
637
+ ≡ β4(m+rp)+2 ≡ β4(2m+rp)+3 ≡ β4(4m+rp)+6 ≡ β4(5m+rp)+9 ≡ β4(7m+rp)+12 ≡ β4(8m+rp)+13
638
+ ≡ β4(10m+rp)+16 ≡ β4(11m+rp)+19 (mod p)
639
+ and 0 ≡ a6m+rp+2 (mod p).
640
+ Then Lemma 3 implies that x is (p, 2m)-perfect at indices
641
+ k = 6m + 2rp for all r ∈ Z and is (p, 2m − 1)-perfect at indices k = 6m + (2r + 1)p. Lemma 1
642
+ then implies that all the entries of the following products are divisible by p:
643
+ 4m+2rp+2
644
+
645
+ i=8m+2rp+3
646
+ Ci,
647
+ 16m+4rp+6
648
+
649
+ i=20m+2rp+9
650
+ Ci.
651
+ For k ⩾ (n + 1)p − p−2
652
+ 3
653
+ one can easily check that the product �2
654
+ i=k Ci contains n + 1 blocks
655
+ of the above form. Therefore all its entries are divisible by pn+1.
656
+ Case p = 12m + 7. Then for all r ∈ Z⩾0,
657
+ 0
658
+ ≡ β4(m+rp)+3 ≡ β4(2m+rp)+4 ≡ β4(4m+rp)+9 ≡ β4(5m+rp)+12 ≡ β4(7m+rp)+17 ≡ β4(8m+rp)+18
659
+ ≡ β4(10m+rp)+23 ≡ β4(11m+rp)+26 (mod p)
660
+ and 0 ≡ a6m+rp+3 (mod p). Lemmata 3 and 1 imply that all the entries of the following
661
+ products are divisible by p:
662
+ 4m+2rp+3
663
+
664
+ i=8m+2rp+4
665
+ Ci,
666
+ 16m+4rp+9
667
+
668
+ i=20m+2rp+12
669
+ Ci.
670
+ For k ⩾ (n + 1)p − p−1
671
+ 3
672
+ one can easily check that the product �2
673
+ i=k Ci contains n + 1 blocks
674
+ of the above form. Therefore all its entries are divisible by pn+1.
675
+ Case p = 12m + 11. Then for all r ∈ Z⩾0,
676
+ 0
677
+ ≡ β4(m+rp)+4 ≡ β4(2m+rp)+7 ≡ β4(4m+rp)+14 ≡ β4(5m+rp)+19 ≡ β4(7m+rp)+26 ≡ β4(8m+rp)+29
678
+ ≡ β4(10m+rp)+36 ≡ β4(11m+rp)+41 (mod p)
679
+ and 0 ≡ a6m+rp+5 (mod p). Lemmata 3 and 1 imply that all the entries of the following
680
+ products are divisible by p:
681
+ 4m+2rp+4
682
+
683
+ i=8m+2rp+7
684
+ Ci,
685
+ 16m+4rp+14
686
+
687
+ i=20m+2rp+19
688
+ Ci.
689
+ For k ⩾ (n + 1)p − p−2
690
+ 3
691
+ one can easily check that the product �2
692
+ i=k Ci contains n + 1 blocks
693
+ of the above form. Therefore all its entries are divisible by pn+1.
694
+ 9
695
+
696
+ In all four cases we have that for k ⩾ (n+1)p− p−2
697
+ 3
698
+ all the entries of �2
699
+ i=k Ci are divisible
700
+ by pn+1. Writing it in terms of k we get that this power of p is
701
+
702
+ k + p−2
703
+ 3
704
+ p
705
+
706
+ =
707
+ �3k + p − 2
708
+ 3p
709
+
710
+ .
711
+
712
+ We combine all the divisibility properties of �1
713
+ i=n Ci together and get the following
714
+ Proposition 1 Let the prime factorisation of a∗ be a∗ = 2σ0pσ1
715
+ 1 pσ2
716
+ 2 · · · pσd
717
+ d
718
+ where σ0 can
719
+ be equal to zero while the other powers σi are strictly positive. Define P1 := {p1, . . . , pd},
720
+ P2 := P \ (P1 ∪ {2, 3}). If 2 | a∗ then
721
+ gcd(pn, qn) ⩾ 2
722
+
723
+ n
724
+ 4
725
+ � �
726
+ pi∈P1
727
+ p
728
+ σi
729
+
730
+ j=1
731
+
732
+ 2n+pj −1
733
+ 2pj
734
+
735
+ i
736
+ ·
737
+
738
+ p∈P2
739
+ p
740
+
741
+ 3n+p−2
742
+ 3p
743
+
744
+ .
745
+ (7)
746
+ If 2 ∤ a∗ then
747
+ gcd(pn, qn) ⩾ 2
748
+
749
+ n+3
750
+ 8
751
+ � �
752
+ pi∈P1
753
+ p
754
+ σi
755
+
756
+ j=1
757
+
758
+ 2n+pj −1
759
+ 2pj
760
+
761
+ i
762
+ ·
763
+
764
+ p∈P2
765
+ p
766
+
767
+ 3n+p−2
768
+ 3p
769
+
770
+ .
771
+ (8)
772
+ We now provide shorter lower bounds for (7) and (8) and then provide slightly better
773
+ ones that, after some efforts, can still be made effective for large enough n. Observe that
774
+ � 2n+pj−1
775
+ 2pj
776
+
777
+
778
+ � n
779
+ pj
780
+
781
+ and
782
+ � 3n+p−2
783
+ 3p
784
+
785
+
786
+ � n
787
+ p
788
+
789
+ . For convenience, if 3 ∤ a∗ we still add 3 to the set P1
790
+ by setting pd+1 := 3, σd+1 := 0. Then
791
+ 2
792
+ �∞
793
+ i=1
794
+ n
795
+ 2i ·
796
+
797
+ pj∈P1
798
+ p
799
+ �σj
800
+ i=1
801
+
802
+ n
803
+ pi
804
+ j
805
+
806
+ j
807
+ ·
808
+
809
+ p∈P2
810
+ p
811
+
812
+ n
813
+ p
814
+
815
+ ·
816
+
817
+ pj∈P1
818
+ p
819
+ �∞
820
+ i=σj+1
821
+ n
822
+ pi
823
+ j
824
+ j
825
+ ·
826
+
827
+ p∈P2
828
+ p
829
+ �∞
830
+ i=2
831
+ n
832
+ pi ⩾ n! ⩾
833
+
834
+ 2πn
835
+ �n
836
+ e
837
+ �n
838
+ .
839
+ The last inequality infers that
840
+ gcd(pn, qn) ⩾
841
+
842
+ 2πn(c1n)n
843
+ (9)
844
+ where c1 = c1(a∗) is defined as
845
+ c1 =
846
+
847
+
848
+
849
+
850
+
851
+
852
+
853
+
854
+
855
+ 2−3/4 exp
856
+
857
+ −1 − �
858
+ pj∈P1
859
+ ln pj
860
+ p
861
+ σj
862
+ j (pj−1) − �
863
+ p∈P2
864
+ ln p
865
+ p(p−1)
866
+
867
+ if 2 | a∗;
868
+ 2−7/8 exp
869
+
870
+ −1 − �
871
+ pj∈P1
872
+ ln pj
873
+ p
874
+ σj
875
+ j (pj−1) − �
876
+ p∈P2
877
+ ln p
878
+ p(p−1)
879
+
880
+ if 2 ∤ a∗.
881
+ (10)
882
+ c1 reaches its minimal value in the case P1 = {3} with σ1 = 0. Then c1 ≈ 0.0924. However, if
883
+ a∗ = 3 then c1 ≈ 0.1333. In general, more squares of small prime numbers divide a∗, bigger
884
+ is the value of c1.
885
+ We can provide a better asymptotic lower estimate on gcd(pn, qn) for large enough n. The
886
+ exact condition on n can be effectively computed, however the computations will not be nice.
887
+ Consider a prime p ∈ P2. The term p
888
+
889
+ 3n+p−2
890
+ 3p
891
+
892
+ has an extra power of p compared to p⌊n/p⌋ if
893
+ for some integer k,
894
+ n
895
+ p < k ⩽ 3n + p − 2
896
+ 3p
897
+ ⇐⇒
898
+ n
899
+ k < p ⩽ 3n − 2
900
+ 3k − 1.
901
+ 10
902
+
903
+ We also have �
904
+ p∈P1 p ≍ 1 where the implied constants only depend on a∗ but not on n.
905
+ Define the set
906
+ K :=
907
+ n�
908
+ k=1
909
+ �n
910
+ k , 3n − 2
911
+ 3k − 1
912
+
913
+ Then gcd(pn, qn) ⩾ T ·
914
+
915
+ 2πn(c1n)n where
916
+ T ≍
917
+
918
+ p∈P∩K
919
+ p = exp
920
+
921
+  �
922
+ p∈K∩P
923
+ ln p
924
+
925
+  = exp
926
+ � n
927
+
928
+ k=1
929
+
930
+ θ
931
+ �3n − 2
932
+ 3k − 1
933
+
934
+ − θ
935
+ �n
936
+ k
937
+ ���
938
+ ,
939
+ where θ(x) is the first Chebyshev function. It is well known (see [7] for example) that for
940
+ large enough x, |θ(x) − x| <
941
+ x
942
+ 2 ln x. Therefore for y > x one has θ(y) − θ(x) ⩾ y − x −
943
+ y
944
+ ln y.
945
+ This implies
946
+
947
+ ln n
948
+
949
+ k=1
950
+
951
+ θ
952
+ �3n − 2
953
+ 3k − 1
954
+
955
+ − θ
956
+ �n
957
+ k
958
+ ��
959
+
960
+
961
+ ln n
962
+
963
+ k=1
964
+ n − 2k
965
+ k(3k − 1) − O
966
+
967
+ n
968
+
969
+ ln n
970
+
971
+ = n
972
+
973
+ ln n
974
+
975
+ k=1
976
+ 1
977
+ k(3k − 1) − O
978
+
979
+ n
980
+
981
+ ln n
982
+
983
+ .
984
+ For any ε > 0 and for large enough n, the last expression can be made bigger that (τ − ε)n
985
+ where τ := �∞
986
+ k=1
987
+ 1
988
+ k(3k−1) ≈ 0.74102. Therefore T ≫ e(τ−ε)n = γn(1−ε) where γ = eτ. Finally,
989
+ we get
990
+ gcd(pn, qn) ≫ ((c2 − δ)n)n,
991
+ where c2 = c1 · γ,
992
+ (11)
993
+ δ can be made arbitrarily small and the implied constant in the inequality only depends on
994
+ a∗ and δ but not on n. For the case a∗ = 1, when the constant c1 is minimal possible, we get
995
+ c2 ≈ 0.1939. Respectively, for a∗ = 6, c2 ≈ 0.2797.
996
+ 6
997
+ Lower and upper bounds on the denominators qn.
998
+ In this section we will get upper and lower bounds of the denominators qn, compared to qn−1.
999
+ Since the recurrent formulae between qn, qn−1 and qn−2 depend on n modulo 4, it makes sense
1000
+ to compare q4k and q4k+4.
1001
+ We adapt some notation from [1]. Denote
1002
+ T4k :=
1003
+
1004
+ p4k
1005
+ q4k
1006
+ p4k−4
1007
+ q4k−4
1008
+
1009
+ Then [1, (69) and (70)] one has
1010
+ T4k+4 =
1011
+ � ak11
1012
+ ak12
1013
+ 1
1014
+ 0
1015
+
1016
+ S4k
1017
+ (12)
1018
+ where ak11 and ak12 are the corresponding indices of C4k+4C4k+3C4k+2C4k+1. In view of (3),
1019
+ one computes
1020
+ ak11 = 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)(t1t2)2
1021
+ +6(8k + 5)(8k + 7)(36k2 + 55k + 16)a∗t1t2
1022
+ +(12k + 5)(12k + 11)(3k + 2)(6k + 7)a∗2
1023
+ ;
1024
+ (13)
1025
+ ak12 = 2(12k + 1)(3k + 1)(8k + 7)a∗t1((8k + 5)(8k + 9)t1t2 + 2(36k2 + 63k + 25)a∗).
1026
+ To make the notation shorter, we write ak12 = 2(12k + 1)(3k + 1)(8k + 7)a∗t1p(k) where p(k)
1027
+ is a polynomial of k with parameters t1t2 and a∗.
1028
+ Then an easy adaptation of the proof of [1, Lemma 16] gives
1029
+ 11
1030
+
1031
+ Lemma 5 Let a∗ ∈ N and t1, t2 ∈ Z satisfy 12a∗ ⩽ |t1t2|. Then q4k+4 and q4k satisfy the
1032
+ relation
1033
+ |q4k+4| > (8k + 3)(8k + 5)(8k + 7)(8k + 9)(t1t2 + 2a∗)2|q4k|.
1034
+ (14)
1035
+ Now we will provide an opposite inequality between the denominators q4k+4 and q4k.
1036
+ Three consecutive denominators of this form are related by the equation [1, (72)]:
1037
+ q4k+4 = ak11q4k + (dq4k−4 − bk21q4k)ak12
1038
+ bk22
1039
+ ,
1040
+ (15)
1041
+ where bk21/d and bk22/d are the corresponding entries of C−1
1042
+ 4k−2C−1
1043
+ 4k−1C−1
1044
+ 4k , i.e.
1045
+ d = −(12k − 7)(12k − 5)(12k − 1)(3k − 1)(6k − 1)(6k + 1)a∗3,
1046
+ (16)
1047
+ bk21 = −(12k − 5)(6k − 1)a∗ − 2(8k − 3)(8k − 1)t1t2,
1048
+ bk22 = 2(8k − 1)t1((8k − 3)(8k + 1)t1t2 + 2(36k2 − 9k − 2)a∗) =: 2(8k − 1)t1p(k − 1).
1049
+ Lemma 6 Let a∗, t1, t2 be the same as in Lemma 5. Then q4k+4 and q4k satisfy the following
1050
+ relations:
1051
+ |q4k+4| ⩽ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)
1052
+
1053
+ t1t2 + 135
1054
+ 32 a∗
1055
+ �2
1056
+ |q4k|,
1057
+ if t1t2 > 0,
1058
+ (17)
1059
+ |q4k+4| ⩽ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)
1060
+
1061
+ t1t2 + 27
1062
+ 32a∗
1063
+ �2
1064
+ |q4k|,
1065
+ if t1t2 < 0.
1066
+ (18)
1067
+ Proof. First, we estimate the terms in (15). Since |t1t2| ⩾ 12a∗, we get for all k ⩾ 1
1068
+ that
1069
+ 1
1070
+ 12(12k − 5)(6k − 1)(12a∗) < (8k − 3)(8k − 1)(12a∗) ⩽ (8k − 3)(8k − 1)|t1t2|. Therefore
1071
+ |bk21| ⩽ 3(8k − 3)(8k − 1)|t1t2|.
1072
+ (19)
1073
+ Next, by Lemma 5 we have
1074
+ |dq4k−4| ⩽ (12k − 7)(12k − 5)(12k − 1)(3k − 1)(6k − 1)(6k + 1)a∗3
1075
+ (8k − 5)(8k − 3)(8k − 1)(8k + 1)(t1t2 + 2a∗)2
1076
+ q4k.
1077
+ Since |t1t2 + 2a∗| ⩾ 10a∗ and 12a∗ ⩽ |t1t2|, one can verify that
1078
+ |dq4k−4| < (8k − 3)(8k − 1)|t1t2q4k|.
1079
+ (20)
1080
+ Next, we have (8k + 5)(8k + 9)|t1t2| > 12(36k2 + 63k + 25)a∗, therefore we always have
1081
+ |p(k)|
1082
+ (8k + 5)(8k + 9)|t1t2| ∈
1083
+ � �
1084
+ 1, 7
1085
+ 6
1086
+
1087
+ if t1t2 ⩾ 0;
1088
+ � 5
1089
+ 6, 1
1090
+
1091
+ if t1t2 < 0.
1092
+ The last inequality in turn implies that for k ⩾ 1 the ratio ak12/bk22 is always positive and
1093
+ satisfies
1094
+ ak12
1095
+ bk12
1096
+ ⩽ 6(12k + 1)(3k + 1)(8k + 7)a∗
1097
+ 8k − 1
1098
+ .
1099
+ (21)
1100
+ Assume that t1t2 ⩾ 0. In that case, the last inequality together with (19) and (20) imply
1101
+ that
1102
+ ����(dq4k−4 − bk21q4k)ak12
1103
+ bk22
1104
+ ���� ⩽ 24(8k − 3)(8k + 7)(12k + 1)(3k + 1)a∗t1t2q4k.
1105
+ 12
1106
+
1107
+ One can check that for all k ⩾ 1,
1108
+ 6(8k + 5)(8k + 7)(36k2 + 55k + 16) + 24(8k − 3)(8k + 7)(12k + 1)(3k + 1)
1109
+ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)
1110
+ < 135
1111
+ 16
1112
+ (22)
1113
+ and
1114
+ 81
1115
+ 256 < (12k + 5)(12k + 11)(3k + 2)(6k + 7)
1116
+ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)
1117
+ ⩽ 23
1118
+ 66 < 1.
1119
+ (23)
1120
+ These bounds together with the formula (13) and equation (15) imply the inequality (17)
1121
+ for k ⩾ 1.
1122
+ Finally, this bound can be easily verified for k = 0 from the equation q4 =
1123
+ a011q0 + a012q−1 and q−1 = 0.
1124
+ Consider the case t1 < 0. One can check that for all k ⩾ 1,
1125
+ 321
1126
+ 187 ⩾ 6(8k + 5)(8k + 7)(36k2 + 55k + 16)
1127
+ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9) ⩾ 27
1128
+ 16.
1129
+ (24)
1130
+ This together with the condition |t1t2| > 12a∗2 imply that ak11 > 0 and q4k and q4k+4 share
1131
+ the same sign for all k ∈ N. Next, since (12k − 5)(6k − 1)a∗ < (8k − 3)(8k − 1)|t1t2|, we
1132
+ have that bk21 > 0 and then in view of (20) and ak12
1133
+ bk22 > 0, the term (dq4k−4 − bk21q4k)ak12
1134
+ bk22 has
1135
+ the opposite sign compared to ak11q4k. That all implies that |q4k+4| ⩽ |ak11q4k|. Finally, the
1136
+ inequalities (23) together with (24) establish the bound (18).
1137
+
1138
+ Lemma 6 immediately implies that for t1t2 > 0,
1139
+ |q4k| ⩽ 2k
1140
+
1141
+ t1t2 + 135
1142
+ 32 a∗
1143
+ �2k
1144
+ (8k + 1)!! ⩽ 16k
1145
+
1146
+ 8 · 21/4e−1
1147
+
1148
+ t1t2 + 135
1149
+ 32 a∗
1150
+ �4k
1151
+ k4k.
1152
+ The case of t1t2 < 0 can be dealt with in a similar way. Finally, we get the estimate
1153
+ |q4k| ⩽ 16kc4k
1154
+ 3 k4k,
1155
+ (25)
1156
+ where
1157
+ c3 = c3(t1, t2, a∗) =
1158
+
1159
+
1160
+
1161
+
1162
+
1163
+ 8 · 21/4e−1
1164
+
1165
+ t1t2 + 135
1166
+ 32 a∗
1167
+ if t1t2 > 0
1168
+ 8 · 21/4e−1
1169
+
1170
+ |t1t2| − 27
1171
+ 32a∗
1172
+ if t1t2 < 0.
1173
+ Lemma 7 Under the same conditions on a∗, t1, t2 as in the previous lemma, one has
1174
+ |q4k+4| ⩾ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)
1175
+
1176
+ t1t2 + 9
1177
+ 16a∗
1178
+ �2
1179
+ |q4k|,
1180
+ if t1t2 > 0,
1181
+ (26)
1182
+ |q4k+4| ⩾ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9) (t1t2 + 3a∗)2 |q4k|,
1183
+ if t1t2 < 0.
1184
+ (27)
1185
+ Proof. If t1t2 > 0 we have q4k+4 > ak11q4k. Then the lower bound in (23) together with
1186
+ the lower bound in (24) imply the bound (26).
1187
+ Now assume that t1t2 < 0. Then, as we have shown in the proof of Lemma 6, bk21 > 0 and
1188
+ dq4k−4 and bk21q4k have the opposite signs. This together with ak12
1189
+ bk22 > 0, the inequality (21)
1190
+ and 0 < bk21 ⩽ 2(8k − 3)(8k − 1)|t1t2| in turn imply that
1191
+ |q4k+4| ⩾ |ak11q4k| − bk21ak12
1192
+ bk22
1193
+ |q4k| ⩾ (ak11 + 12(8k − 3)(12k + 1)(3k + 1)(8k + 7)a∗t1t2)|q4k|
1194
+ 13
1195
+
1196
+ We need to show that the expression
1197
+ ak11+12(8k−3)(12k+1)(3k+1)(8k+7)a∗t1t2−2(8k+3)(8k+5)(8k+7)(8k+9) (t1t2 + 3a∗)2
1198
+ is always positive. Notice that after substituting (13) into it and expanding the brackets, the
1199
+ term with (t1t2)2 disappears. The term for a∗t1t2 then equals to
1200
+ −6(8k + 7)(160k3 + 1532k2 + 1063k + 196)a∗t1t2
1201
+ and the term for a∗2 is
1202
+ −(71136k4 + 212976k3 + 227970k2 + 102633k + 16240)
1203
+ (we made these computations with Wolfram Mathematika). Finally, one can check that in
1204
+ the case |t1t2| > 12a∗, the absolute value of the first term is always bigger than that of the
1205
+ second term and therefore the whole expression is positive.
1206
+ Remark.
1207
+ By performing neater computations, one can make the coefficient 3 in
1208
+ (t1t2 + 3a∗)2 slightly smaller. However we decide not to further complicate already tedious
1209
+ calculations.
1210
+
1211
+ Analogously to (25), one can find shorter lower bounds for |q4k|. With help of the known
1212
+ inequality (8k + 1)!! ⩾ 8k(8k/e)2k, Lemma 7 infers
1213
+ |q4k| ⩾ 8kc4k
1214
+ 4 k4k,
1215
+ (28)
1216
+ where
1217
+ c4 = c4(t1, t2, a∗) =
1218
+
1219
+
1220
+
1221
+ 8 · 21/4e−1
1222
+
1223
+ t1t2 + 9
1224
+ 16a∗
1225
+ if t1t2 > 0
1226
+ 8 · 21/4e−1�
1227
+ |t1t2| − 3a∗
1228
+ if t1t2 < 0.
1229
+ 7
1230
+ Distance between x and the convergents
1231
+ From [1, Lemma 17] we know that, under the condition 12a∗ ⩽ |t1t2|, one has
1232
+ ����x − p4k
1233
+ q4k
1234
+ ���� < 2
1235
+ ����
1236
+ p4k
1237
+ q4k
1238
+ − p4k+4
1239
+ q4k+4
1240
+ ���� .
1241
+ In order to estimate the right hand side, we use the matrix equation [1, (72)]:
1242
+ Tk+1 =
1243
+ � ak11 − ak12
1244
+ bk21
1245
+ bk22
1246
+ dak12
1247
+ bk22
1248
+ 1
1249
+ 0
1250
+
1251
+ Tk.
1252
+ Notice that the values of d in fact depends on k (see the formula (16)). to emphasize this
1253
+ dependence, in this section we write d(k) for it. Then the above equation gives the following
1254
+ formula:
1255
+ ����
1256
+ p4k
1257
+ q4k
1258
+ − p4k+4
1259
+ q4k+4
1260
+ ���� =
1261
+ ���
1262
+ �k
1263
+ i=1
1264
+ d(i)ai12
1265
+ bi22
1266
+ ��� · |p0q4 − q0p4|
1267
+ q4kq4k+4
1268
+ .
1269
+ We first compute its product term:
1270
+ �����
1271
+ k
1272
+
1273
+ i=1
1274
+ d(i)ai12
1275
+ bi22
1276
+ ����� =
1277
+ k
1278
+
1279
+ i=1
1280
+ 2(12i − 7)(12i − 5)(12i − 1)(3i − 1)(6i − 1)(6i + 1)(12i + 1)(3i + 1)(8i + 7)a∗4|t1p(i)|
1281
+ 2(8i − 1)|t1p(i − 1)|
1282
+ 14
1283
+
1284
+ = (8k + 7)|p(k)|
1285
+ 7|p(0)|
1286
+ · (3k + 1)(6k + 1)(12k + 1)a∗4k(12k)!
1287
+ 26k34k(4k)!
1288
+
1289
+
1290
+ 3(3k + 1)(6k + 1)(12k + 1)(8k + 7)|p(k)|
1291
+ 7|p(0)|
1292
+ ·
1293
+ �122k2a∗
1294
+ 2
1295
+
1296
+ 2e2
1297
+ �4k
1298
+ .
1299
+ Next, from (12) for k = 0 we get that |p0q4 − p4q0| = 14a∗t1|p(0)|. Finally, we unite all these
1300
+ bounds together with the lower bounds (26), (27) and (28) for |q4k| to get
1301
+ ����x − p4k
1302
+ q4k
1303
+ ���� ⩽
1304
+ 2
1305
+
1306
+ 3(3k + 1)(6k + 1)(12k + 1)(8k + 7)|t1a∗p(k)|
1307
+ 2(8k + 3)(8k + 5)(8k + 7)(8k + 9)(|t1t2| − 3a∗)2 · 64k2 ·
1308
+
1309
+ 122a∗
1310
+ 2
1311
+
1312
+ 2e2c2
1313
+ 4
1314
+ �4k
1315
+ To simplify the right hand side, notice that (3k+1)(6k+1)(12k+1)
1316
+ (8k+3)(8k+5)(8k+9) < 27
1317
+ 64. Next, since |t1t2| ⩾
1318
+ 12a∗, one has |p(k)| = |(8k + 5)(8k + 9)t1t2 + 2(36k2 + 63k + 25)a∗| ⩽ 2(8k + 5)(8k + 9)|t1t2|
1319
+ which for all k ⩾ 1 is smaller than 442k2|t1t2|. Finally, (|t1t2| − 3a∗)2 ⩾
1320
+ 9
1321
+ 16(t1t2)2. Collecting
1322
+ all of these inequalities together gives,
1323
+ ����x − p4k
1324
+ q4k
1325
+ ���� ⩽
1326
+
1327
+ 3 · 27 · 442|t2
1328
+ 1t2a∗|
1329
+ 64 · (9/16) · 64(t1t2)2 ·
1330
+
1331
+ 72a∗
1332
+
1333
+ 2e2c2
1334
+ 4
1335
+ �4k
1336
+ ⩽ |t1|c4k
1337
+ 5 ,
1338
+ (29)
1339
+ where
1340
+ c5 =
1341
+
1342
+
1343
+
1344
+ 9a∗
1345
+ (16t1t2+9a∗)
1346
+ if t1t2 > 0
1347
+ 9a∗
1348
+ 16(|t1t2|−3a∗)
1349
+ if t1t2 < 0.
1350
+ 8
1351
+ Estimating the irrationality exponent
1352
+ In this section we establish Theorems refth1 and 2. Consider p∗
1353
+ k := p4k/ gcd(p4k, q4k) and
1354
+ q∗
1355
+ k := q4k/ gcd(p4k, q4k). Definitely, they are both integers and (25) together with (9) imply
1356
+ |q∗
1357
+ k| ⩽ 4
1358
+
1359
+ 2k
1360
+ π
1361
+ � c3
1362
+ 4c1
1363
+ �4k
1364
+ =: 4
1365
+
1366
+ 2k
1367
+ π · c4k
1368
+ 6 .
1369
+ (30)
1370
+ For arbitrary δ > 0 and large enough k, one can use the inequality (11) to get
1371
+ |q∗
1372
+ k| ≪ 16k
1373
+
1374
+ c3
1375
+ 4(c2 − δ)
1376
+ �4k
1377
+
1378
+ � c3
1379
+ 4c2
1380
+ + δ1
1381
+ �4k
1382
+ =: (c∗
1383
+ 6 + δ1)4k
1384
+ (31)
1385
+ where δ1 > 0 can be made arbitrarily close to zero for large enough k. Denote the upper
1386
+ bound for b∗
1387
+ k by Q(k, t, a).
1388
+ Next, we combine the last two inequalities with (29) and get
1389
+ ||q∗
1390
+ kx|| ⩽ |t1|c4k
1391
+ 5 · 4
1392
+
1393
+ 2k
1394
+ π
1395
+ � c3
1396
+ 4c1
1397
+ �4k
1398
+ ⩽ 4|t1|
1399
+
1400
+ k
1401
+ �c3c5
1402
+ 4c1
1403
+ �4k
1404
+ =: 4|t1|
1405
+
1406
+ kc−4k
1407
+ 7
1408
+ (32)
1409
+ or
1410
+ ||q∗
1411
+ kx|| ≪
1412
+ � 4c2
1413
+ c3c5
1414
+ − δ2
1415
+ �−4k
1416
+ =: (c∗
1417
+ 7 − δ2)−4k
1418
+ (33)
1419
+ where δ2 can be made arbitrarily small and k is large enough, depending on δ2. Denote the
1420
+ upper bound of ||q∗
1421
+ kx|| by R(k, t, a).
1422
+ 15
1423
+
1424
+ Consider an arbitrary q ⩾
1425
+ 1
1426
+ 2R(1,t,a) = q0. We now impose the condition c7 > e1/4. In
1427
+ this case, by examining the derivative of
1428
+
1429
+ kc−4k
1430
+ 7
1431
+ , one can check that it strictly decreases for
1432
+ k ⩾ 1. Therefore, there exists a unique k ⩾ 2 such that R(k, t, a) <
1433
+ 1
1434
+ 2q ⩽ R(k − 1, t, a). Let
1435
+ p ∈ Z be such that ||qx|| = |qx − p|. Since two vectors (p∗
1436
+ k, q∗
1437
+ k) and (p∗
1438
+ k+1, q∗
1439
+ k+1) are linearly
1440
+ independent, at least one of them must be linearly independent with (p, q). Suppose that is
1441
+ (p∗
1442
+ k, q∗
1443
+ k). Then we estimate the absolute value of the following determinant:
1444
+ 1 ⩽
1445
+ ����
1446
+ q
1447
+ q∗
1448
+ k
1449
+ p
1450
+ p∗
1451
+ k
1452
+ ���� ⩽
1453
+ ����
1454
+ q
1455
+ q∗
1456
+ k
1457
+ p − qx
1458
+ p∗
1459
+ k − q∗
1460
+ kx
1461
+ ���� ⩽ qR(k, t, a) + ||qx||Q(k, t, a).
1462
+ Since qR(k, t, a) < 1
1463
+ 2, we must have ||qx|| ⩾ (2Q(k, t, a))−1. Analogously, if (p, q) is linearly
1464
+ independent with (p∗
1465
+ k+1, q∗
1466
+ k+1), we have ||qx|| ⩾ (2Q(k + 1, t, a))−1. The latter lower bound
1467
+ is weaker. Now, we need to rewrite the right hand side of the inequality in terms of q rather
1468
+ than k.
1469
+ Since
1470
+ 1
1471
+ 2q ⩽ R(k − 1, t, a), we have that
1472
+ c4(k−1)
1473
+ 7
1474
+ 8|t1|
1475
+
1476
+ k − 1 ⩽ q
1477
+ =⇒
1478
+ k − 1 ⩽ log(8|t1|q) + log log(8|t1|q)
1479
+ 4 log c7
1480
+ .
1481
+ The last implication can be justified by standard techniques on working with logarithms,
1482
+ see [1, (41)].
1483
+ Finally, substitute the last lower bound for k in ||qx|| ⩾ (2Q(k + 1, t, a))−1 and get
1484
+ ||qx|| ⩾
1485
+ √π
1486
+ 8
1487
+
1488
+ 2(k + 1)c4(k+1)
1489
+ 6
1490
+
1491
+ 2√π(log c7)1/2
1492
+ 8
1493
+
1494
+ 6c8
1495
+ 6(log(8|t1|q) + log log(8|t1|q))1/2 · (8|t1|q)
1496
+ log c6
1497
+ log c7 (log(8|t1|q))
1498
+ log c6
1499
+ log c7
1500
+ .
1501
+
1502
+ (log c7)1/2
1503
+ 8c8
1504
+ 6(8|t1|)
1505
+ log c6
1506
+ log c7
1507
+ · q− log c6
1508
+ log c7 (log(8|t1|q))− log c6
1509
+ log c7 − 1
1510
+ 2 = τ
1511
+ g2
1512
+ q−λ(log(8|t1|q))−λ− 1
1513
+ 2 .
1514
+ To finish the proof of Theorem 1, we recall, that for convenience, we in fact worked with the
1515
+ number x/g2 rather than x, i.e. the inequality above is for ||qx/g2||. Hence one needs to
1516
+ multiply both sides by g2.
1517
+ Regarding theorem 2, we use inequalities (31) and (33). in this case the computations are
1518
+ much easier and we get for any δ3 > 0 and large enough integer q that
1519
+ ||qx|| ⩾ q
1520
+
1521
+ log c∗
1522
+ 6
1523
+ log c∗
1524
+ 7 −δ3
1525
+ or in other words λeff(x) ⩽ log c∗
1526
+ 6
1527
+ log c∗
1528
+ 7 . That completes the proof of Theorem 2.
1529
+ References
1530
+ [1] D.
1531
+ Badziahin.
1532
+ Continued
1533
+ fractions
1534
+ of
1535
+ cubic
1536
+ Laurent
1537
+ series.
1538
+ Preprint.
1539
+ https://arxiv.org/pdf/2211.08663.
1540
+ [2] A. Baker. Rational approximations to certain algebraic numbers. Proc. LMS 14 (1964),
1541
+ No 3, 385–398.
1542
+ [3] M. A. Bennett. Effective Measures of Irrationality for Certain Algebraic Numbers. J.
1543
+ AustMS 62 (1997), 329–344.
1544
+ 16
1545
+
1546
+ [4] E. Bombieri, A.J. van der Poorten, J. D. Vaaler. Effective measures of irrationality for
1547
+ cubic extensions of number fields. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) 23 (1996)
1548
+ No 2, 211–248.
1549
+ [5] Y. Bugeaud. Linear forms in logarithms and applications. European Mathematical So-
1550
+ ciety (2018).
1551
+ [6] N. I. Feldman. Improved estimate for a linear form of the logarithms of algebraic num-
1552
+ bers. Mat. Sb. 77 (1968), 256–270 (in Russian). English translation in Math. USSR. Sb.
1553
+ 6 (1968), 393–406.
1554
+ [7] J. B. Rosser, L. Schoenfeld. Approximate formulas for some functions of prime numbers.
1555
+ Illinois J. Math., 6 (1962), 64–94.
1556
+ [8] K. F. Roth. Rational approximations to algebraic numbers. Mathematika, 2 (1955), 337–
1557
+ 360.
1558
+ [9] I. Wakabayashi, Cubic Thue inequalities with negative discriminant. J. Number Theory,
1559
+ 97 (2002), No 2, 225–251.
1560
+ 17
1561
+
49E0T4oBgHgl3EQfegBh/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
49E1T4oBgHgl3EQfAwK8/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db277b23780e24d4f5bbfb4fd04f89af5cffc6f046879da999c6fc3be7851244
3
+ size 163014
4dAyT4oBgHgl3EQf2Pkf/content/2301.00746v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f71658d7463733f8e8cdac4b92c1eb67254d27a5d01492f2a908f25f241a4949
3
+ size 2524795
4dAyT4oBgHgl3EQf2Pkf/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0dd1108031e1e9a5888a389b9a2bedcc2bab3dd1ae10c66764b2709c0800774
3
+ size 139621
4tE4T4oBgHgl3EQf1A0X/content/2301.05286v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e3782100b5db017cd6fdd8bdae6657fd788573f37bb51eb21cb03e8083357ca
3
+ size 13836874
6NA0T4oBgHgl3EQfN_-e/content/tmp_files/2301.02155v1.pdf.txt ADDED
@@ -0,0 +1,1206 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ A Shannon-Theoretic Approach to the Storage-Retrieval Tradeoff in
2
+ PIR Systems
3
+ Chao Tian, Hua Sun, and Jun Chen
4
+ Abstract
5
+ We consider the storage-retrieval rate tradeoff in private information retrieval (PIR) systems using a
6
+ Shannon-theoretic approach. Our focus is mostly on the canonical two-message two-database case, for
7
+ which a coding scheme based on random codebook generation and the binning technique is proposed.
8
+ This coding scheme reveals a hidden connection between PIR and the classic multiple description source
9
+ coding problem. We first show that when the retrieval rate is kept optimal, the proposed non-linear
10
+ scheme can achieve better performance over any linear scheme. Moreover, a non-trivial storage-retrieval
11
+ rate tradeoff can be achieved beyond space-sharing between this extreme point and the other optimal
12
+ extreme point, achieved by the retrieve-everything strategy. We further show that with a method akin
13
+ to the expurgation technique, one can extract a zero-error PIR code from the random code. Outer
14
+ bounds are also studied and compared to establish the superiority of the non-linear codes over linear
15
+ codes.
16
+ 1
17
+ Introduction
18
+ Private information retrieval (PIR) addresses the situation of storing K messages of L-bits each in N
19
+ databases, with the requirement that the identity of any requested message must be kept private from any
20
+ one (or any small subset) of the databases. The early works were largely computer science theoretic [1],
21
+ where L = 1, and the main question is the scaling law of the retrieval rate in terms of (K, N).
22
+ The storage overhead in PIR systems has been studied in the coding and information theory com-
23
+ munity, from several perspectives using mainly two problem formulations. Shah et al. [2] considered the
24
+ problem when N is allowed to vary with L and K, and obtained some conclusive results. In a similar
25
+ vein, for L = 1, Fazeli et al. [3] proposed a technique to convert any linear PIR code to a new one with
26
+ low storage overhead by increasing N. Other notable results along this line can be found in [4–9].
27
+ An information theoretic formulation of the PIR problem was considered in [10], where L is allowed to
28
+ increase, while (N, K) are kept fixed. Important properties on the tradeoff between the storage rate and
29
+ retrieval rate were identified in [10], and a linear code construction was proposed. In this formulation,
30
+ even without any storage overhead constraint, characterizing the minimum retrieval rate in the PIR
31
+ systems is nontrivial, and this capacity problem was settled in [11]. Tajeddine et al. [12] considered the
32
+ capacity problem when the message is coded across the databases with a maximum-distance separable
33
+ (MDS) code, which was later solved by Banawan and Ulukus [13]. Capacity-achieving code designs with
34
+ optimal message sizes were given in [14,15]. Systems where servers can collude were considered in [16].
35
+ There have been various extensions and generalizations, and the recent survey article [17] provides a
36
+ comprehensive overview on efforts following this information theoretic formulation.
37
+ In many existing works, the storage component and the PIR component are largely designed sepa-
38
+ rately, usually by placing certain structural constraints on one of them; e.g., the MDS coding requirement
39
+ for the storage component [13], or the storage is uncoded [18]; moreover, the code constructions are al-
40
+ most all linear. The few exceptions we are aware of are [19–21]. In this work, we consider the information
41
+ theoretic formulation of the PIR problem, without placing any additional structural constraints on the
42
+ two components, and explicitly investigate the storage-retrieval tradeoff region. We mostly focus on the
43
+ 1
44
+ arXiv:2301.02155v1 [cs.IT] 5 Jan 2023
45
+
46
+ case N = K = 2 here since it provides the most important intuition; we refer to this as the (2, 2) PIR
47
+ system. Our approach naturally allows the joint design of the two components using either linear or
48
+ non-linear schemes.
49
+ The work in [19] is of significant relevance to our work, where the storage overhead was considered
50
+ in both single-round and multi-round PIR systems, when the retrieval rate must be optimal. Although
51
+ multi-round PIR has the same capacity as single-round PIR, it was shown that at the minimum retrieval
52
+ rate, a multi-round, ϵ-error, non-linear code can indeed break the storage performance barrier of an
53
+ optimal single-round, zero error, linear code. The question whether all the three differences are essential
54
+ to overcome this barrier was left as an open question.
55
+ In this work, we show that a non-linear code is able to achieve better performance than the optimal
56
+ linear code in the single-round zero-error (2, 2) PIR system, over a range of the storage rates. This is
57
+ accomplished by providing a Shannon-theoretic coding scheme based on random codebook generation
58
+ and the binning technique. The proposed scheme at the minimum retrieval rate is conceptually simpler,
59
+ and we present it as an explicit example. The general inner bound is then provided, and we show an
60
+ improved tradeoff can be achieved beyond space-sharing between the minimum retrieval rate code and
61
+ the other optimal extreme point. By leveraging a method akin to the expurgation technique, we further
62
+ show that one can extract a zero-error deterministic PIR code from the random ϵ-error PIR code. Outer
63
+ bounds are also studied for both general codes and linear codes, which allow us to establish conclusively
64
+ the superiority of non-linear codes over linear codes. Our work essentially answers the open question
65
+ in [19], and shows that in fact only non-linearity is essential in breaking the aforementioned barrier.
66
+ A preliminary version of this work was presented first in part in [22]. In this updated article, we provide
67
+ a more general random coding scheme, which reveals a hidden connection to the multiple description
68
+ source coding problem [23]. Intuitively, we can view the retrieved message as certain partial reconstruction
69
+ of the full set of messages, instead of a complete reconstruction of a single message. Therefore, the answers
70
+ from the servers can be viewed as descriptions of the full set of messages, which are either stored directly at
71
+ the servers or formed at the time of request, and the techniques seen in multiple description coding become
72
+ natural in the PIR setting. Since the publication of the preliminary version [22], several subsequent efforts
73
+ have been made in studying the storage-retrieval tradeoff in the PIR setting, which provided stronger and
74
+ more general information theoretic outer bounds and several new linear code constructions [20, 21, 24].
75
+ However, the Shannon-theoretic random coding scheme given in [22] remains the best performance for
76
+ the (2, 2) case, which motivate us to provide the general coding scheme in this work and to make the
77
+ connection to multiple description source coding more explicit. It is our hope that this connection may
78
+ bring existing coding techniques for the multiple description problem to the study of the PIR problem.
79
+ 2
80
+ Preliminaries
81
+ The problem we consider is essentially the same as that in [11], with the additional consideration on
82
+ the storage overhead constraint at the databases. We provide a formal problem definition in the more
83
+ traditional Shannon-theoretic language, to facilitate subsequent treatment. Some relevant results on this
84
+ problem are also reviewed briefly in this section.
85
+ 2.1
86
+ Problem Definition
87
+ There are two independent messages, denoted as W1 and W2, in this system, each of which is generated
88
+ uniformly at random in the finite field FL
89
+ 2 , i.e., each message is an L-bit sequence. There are two databases
90
+ to store the messages, which are produced by two encoding functions operating on (W1, W2)
91
+ φn : FL
92
+ 2 × FL
93
+ 2 → Fαn
94
+ 2 ,
95
+ n = 1, 2,
96
+ where αn is the number of storage symbols at database-n, n = 1, 2, which is a deterministic function of
97
+ L, i.e., we are using fixed length codes for storage. We write S1 = φ1(W1, W2) and S2 = φ2(W1, W2).
98
+ 2
99
+
100
+ When a user requests message-k, it generates two queries (Q[k]
101
+ 1 , Q[k]
102
+ 2 ) to be sent to the two databases,
103
+ randomly in the alphabet Q × Q. Note the joint distribution satisfies the condition
104
+ PW1,W2,Q[k]
105
+ 1 ,Q[k]
106
+ 2 = PW1,W2PQ[k]
107
+ 1 ,Q[k]
108
+ 2 ,
109
+ k = 1, 2,
110
+ (1)
111
+ i.e.,
112
+ the messages and the queries are independent. The marginal distributions PW1,W2 and PQ[k]
113
+ 1 ,Q[k]
114
+ 2 ,
115
+ k = 1, 2, thus fully specify the randomness in the system.
116
+ After receiving the queries, the databases produce the answers to the query via a set of deterministic
117
+ functions
118
+ ϕ(q)
119
+ n
120
+ : Fαn
121
+ 2
122
+ → Fβ(q)
123
+ n
124
+ 2
125
+ ,
126
+ q ∈ Q, n = 1, 2.
127
+ (2)
128
+ We also write the answers A[k]
129
+ n = ϕ(Q[k]
130
+ n )
131
+ n
132
+ (Sn), n = 1, 2. The user, with the retrieved information, wishes
133
+ to reproduce the desired message through a set of decoding functions
134
+ ψ(k,q1,q2) : Fβ(q1)
135
+ 1
136
+ 2
137
+ × Fβ(q2)
138
+ 2
139
+ 2
140
+ → FL
141
+ 2 .
142
+ (3)
143
+ The outputs of the functions ˆWk = ψ(k,Q[k]
144
+ 1 ,Q[k]
145
+ 2 )(A[k]
146
+ 1 , A[k]
147
+ 2 ) are essentially the retrieved messages. We
148
+ require the system to retrieve the message correctly (zero-error), i.e., ˆWk = Wk for k = 1, 2.
149
+ Alternatively, we can require the system to have a small error probability. Denote the average prob-
150
+ ability of coding error of a PIR code as
151
+ Pe = 0.5
152
+
153
+ k=1,2
154
+ PW1,W2,Q[k]
155
+ 1 ,Q[k]
156
+ 2 (Wk ̸= ˆWk).
157
+ (4)
158
+ An (L, α1, α2, β1, β2) ϵ-error PIR code is defined similar as a (zero-error) PIR code, except that the
159
+ correctness condition is replaced by the condition that the probability of error Pe ≤ ϵ.
160
+ Finally, the privacy constraint stipulates that the identical distribution condition must be satisfied:
161
+ PQ[1]
162
+ n ,A[1]
163
+ n ,Sn = PQ[2]
164
+ n ,A[2]
165
+ n ,Sn,
166
+ n = 1, 2.
167
+ (5)
168
+ Note that one obvious consequence is that PQ[1]
169
+ n = PQ[2]
170
+ n ≜ PQn, for n = 1, 2.
171
+ We refer to the code, which is specified by two probability distributions PQ[k]
172
+ 1 ,Q[k]
173
+ 2 , k = 1, 2, and a
174
+ valid set of coding functions {φn, ϕ(q)
175
+ n , ψk,q1,q2} that satisfy both the correctness and privacy constraints,
176
+ as an (L, α1, α2, β1, β2) PIR code, where βn = EQn[β(Qn)
177
+ n
178
+ ], for n = 1, 2.
179
+ Definition 1. A normalized storage-retrieval rate pair (¯α, ¯β) is achievable, if for any ϵ > 0 and sufficiently
180
+ large L, there exists an (L, α1, α2, β1, β2) PIR code, such that
181
+ L(¯α + ϵ) ≥ 1
182
+ 2(α1 + α2), L(¯β + ϵ) ≥ 1
183
+ 2 (β1 + β2) .
184
+ (6)
185
+ The collection of the achievable normalized storage-retrieval rate pair (¯α, ¯β) is the achievable storage-
186
+ retrieval rate region, denoted as R.
187
+ Unless explicitly stated, the rate region R is used for the zero-error PIR setting. In the definition
188
+ above, we have used the average rates (¯α, ¯β) across the databases instead of the individual rate vectors
189
+ 1
190
+ n(α1, α2, EQ1[β(Q1)
191
+ 1
192
+ ], EQ2[β(Q2)
193
+ 2
194
+ ]). This can be justified using the following lemma.
195
+ Lemma 1. If an (L, α1, α2, β1, β2) PIR code exists, then a (2L, α, α, β, β) PIR code exists, where
196
+ α = α1 + α2,
197
+ β = β1 + β2.
198
+ (7)
199
+ This lemma can essentially be proved by a space-sharing argument, the details of which can be found
200
+ in [19]. The following lemma is also immediate using a conventional space-sharing argument.
201
+ Lemma 2. The region R is convex.
202
+ 3
203
+
204
+ database-1
205
+ database-2
206
+ Figure 1: A possible coding structure.
207
+ 2.2
208
+ Some Relevant Known Results
209
+ The capacity of a general PIR system with K messages and N databases is identified in [11] as
210
+ C = 1 − 1/N
211
+ 1 − 1/N K ,
212
+ (8)
213
+ which in our definition corresponds to the case when ¯β is minimized, and the proposed linear code achieves
214
+ (¯α, ¯β) = (K, (1 − 1/N K)/(N − 1)). The capacity of MDS-code PIR systems was established in [13]. In
215
+ the context of storage-retrieval tradeoff, this result can be viewed as providing the achievable tradeoff
216
+ pairs
217
+ (¯α, ¯β) =
218
+
219
+ t, 1 − tK/N K
220
+ N − t
221
+
222
+ , t = 1, 2, . . . , N.
223
+ (9)
224
+ However when specialized to the (2, 2) PIR problem, this does not provide any improvement over the
225
+ space-sharing strategy between the trivial code of retrieval-everything and the code in [11]. By specializing
226
+ the code in [11], it was shown in [19] that for the (2, 2) PIR problem, at the minimal retrieval value
227
+ ¯β = 0.75, the storage rate ¯αl = 1.5 is achievable using a single-round, zero-error linear code, and in fact,
228
+ it is the optimal storage rate that any single-round, zero-error linear code can achieve.
229
+ One of the key observations in [19] is that a special coding structure appears to be the main difficulty
230
+ in the (2, 2) PIR setting, which is illustrated in Fig. 1. Here message W1 can be recovered from either
231
+ (X1, Y1) or (X2, Y2), and message W2 can be recovered from either (X1, Y2) or (X2, Y1); (X1, X2) is
232
+ essentially S1 and is stored at database-1, and (Y1, Y2) is essentially S2 and is stored at database-2.
233
+ It is clear that we can use the following strategy to satisfy the privacy constraint: when message W1
234
+ is requested, with probability 1/2, the user queries for either (X1, Y1) or (X2, Y2); for message 2, with
235
+ probability 1/2, the user queries for either (X1, Y2) or (X2, Y1). More precisely, the following probability
236
+ distribution PQ[1]
237
+ 1 ,Q[1]
238
+ 2
239
+ and PQ[2]
240
+ 1 ,Q[2]
241
+ 2
242
+ can be used:
243
+ PQ[1]
244
+ 1 ,Q[1]
245
+ 2 =
246
+
247
+ 0.5
248
+ (Q[1]
249
+ 1 , Q[1]
250
+ 2 ) = (11)
251
+ 0.5
252
+ (Q[1]
253
+ 1 , Q[1]
254
+ 2 ) = (22)
255
+ ,
256
+ (10)
257
+ and
258
+ PQ[2]
259
+ 1 ,Q[2]
260
+ 2 =
261
+
262
+ 0.5
263
+ (Q[2]
264
+ 1 , Q[2]
265
+ 2 ) = (12)
266
+ 0.5
267
+ (Q[2]
268
+ 1 , Q[2]
269
+ 2 ) = (21)
270
+ .
271
+ (11)
272
+ 4
273
+
274
+ 2.3
275
+ Multiple Description Source Coding
276
+ The multiple description source coding problem [23] considers compressing a memoryless source S into
277
+ a total of M descriptions, i.e., M compressed bit sequences, such that the combinations of any subset
278
+ of these descriptions can be used to reconstruct the source S to guarantee certain quality requirements.
279
+ The motivation of this problem is mainly to address the case when packets can be dropped randomly on
280
+ a communication network.
281
+ Denote the coding rate for each description as Ri, i = 1, 2, . . . , M. A coding scheme was proposed
282
+ in [25], which leads to the following rate region.
283
+ Let U1, U2, . . . , UM be M random variables jointly
284
+ distributed with S, then the following rates (R1, R2, . . . , RM) and distortions (DA, A ⊆ {1, 2, . . . , M})
285
+ are achievable:
286
+
287
+ i∈A
288
+ Ri ≥
289
+
290
+ i∈A
291
+ H(Ui) − H({Ui, i ∈ A}|S),
292
+ A ⊆ {1, 2, . . . , M},
293
+ (12)
294
+ DA ≥ E[d(S, fA(Ui, i ∈ A))],
295
+ A ⊆ {1, 2, . . . , M}.
296
+ (13)
297
+ Here fA is a reconstruction mapping from the random variables {Ui, i ∈ A} to the reconstruction domain,
298
+ d(·, ·) is a distortion metric that is used to measure the distortion, and DA is the distortion achievable using
299
+ the descriptions in the set A. Roughly speaking, the coding scheme requires generating approximately
300
+ 2nRi length-n codewords in an i.i.d. manner using the marginal distribution Ui for each i = 1, 2, . . . , M,
301
+ and the rate constraints insure that when n is sufficiently large, with overwhelming probability there is a
302
+ tuple of M codewords (un
303
+ 1, un
304
+ 2, . . . , un
305
+ M), one in each codebook constructed earlier, that are jointly typical
306
+ with the source vector Sn. In this coding scheme, the descriptions are simply the codeword indices of
307
+ these codewords in these codebooks. For a given joint distribution (S, U1, U2, . . . , UM), we refer to the
308
+ rate region in (12) as the MD rate region RMD(S, U1, U2, . . . , UM), and the corresponding random code
309
+ construction the MD codebooks associated with (S, U1, U2, . . . , UM).
310
+ The binning technique [26] can be applied in the multiple description problem to provide further per-
311
+ formance improvements, particularly when not all the combinations of the descriptions are required
312
+ to satisfy certain performance constraints, but only a subset of them are; this technique has pre-
313
+ viously been used in [27] and [28] for this purpose.
314
+ Assume that only the subsets of descriptions
315
+ A1, A2, . . . , AT ⊆ {1, 2, . . . , M} have distortion requirements associated with the reconstructions using
316
+ these descriptions, which are denoted as DAi, i = 1, 2, . . . , T. Consider the MD codebooks associated with
317
+ (S, U1, U2, . . . , UM) at rates (R′
318
+ 1, R′
319
+ 2, . . . , R′
320
+ M) ∈ RMD(S, U1, U2, . . . , UM), then assign the codewords in
321
+ the i-th codebook uniformly at random into 2nRi bins with 0 ≤ Ri ≤ R′
322
+ i. The coding rates and distortions
323
+ that satisfy the following constraints simultaneously for all Ai, i = 1, 2, . . . , T are achievable:
324
+
325
+ j∈J
326
+ (R′
327
+ j − Rj) ≤
328
+
329
+ j∈J
330
+ H(Uj) − H
331
+
332
+ {Uj, j ∈ J }
333
+ ����
334
+
335
+ Uj′, j′ ∈ Ai \ J
336
+ ��
337
+ ,
338
+ ∀J ⊆ Ai,
339
+ (14)
340
+ DAi ≥ E[d(S, fAi(Uj, j ∈ Ai))].
341
+ (15)
342
+ We denote the collection of such rate vectors (R1, R2, . . . , RM, R′
343
+ 1, R′
344
+ 2, . . . , R′
345
+ M) as R∗
346
+ MD((S, U1, U2, . . . , UM), ({Uj, j ∈
347
+ Ai}, i = 1, 2, . . . , T)), and refer to the corresponding codebooks as the MD∗ codebooks associated with
348
+ the random variables (S, U1, U2, . . . , UM) and the reconstruction sets (A1, A2, . . . , AT ).
349
+ 3
350
+ A Special Case: Slepian-Wolf Coding for Minimum Retrieval Rate
351
+ In this section, we consider the minimum-retrieval-rate case, and show that non-linear and Shannon-
352
+ theoretic codes are beneficial. We will be rather cavalier here and ignore some details, in the hope of
353
+ better conveyance of the intuition. In particular, we ignore the asymptotic-zero probability of error that
354
+ is usually associated with a random coding argument, but this will be addressed more carefully in Section
355
+ 4.
356
+ 5
357
+
358
+ Let us rewrite the L-bit messages as
359
+ Wk = (Vk[1], . . . , Vk[L]) ≜ V L
360
+ k ,
361
+ k = 1, 2.
362
+ (16)
363
+ The messages can be viewed as being produced from a discrete memoryless source PV1,V2 = PV1 · PV2,
364
+ where V1 and V2 are independent uniform-distributed Bernoulli random variables.
365
+ Consider the following auxiliary random variables
366
+ X1 ≜ V1 ∧ V2,
367
+ X2 ≜ (¬V1) ∧ (¬V2),
368
+ Y1 ≜ V1 ∧ (¬V2),
369
+ Y2 ≜ (¬V1) ∧ V2,
370
+ (17)
371
+ where ¬ is the binary negation, and ∧ is the binary “and” operation. This particular distribution satisfies
372
+ the coding structure depicted in Fig. 1, with (V1, V2) taking the role of (W1, W2), and the relation is
373
+ non-linear. The same distribution was used in [19] to construct a multiround PIR code. This non-linear
374
+ mapping appears to allow the resultant code to be more efficient than linear codes.
375
+ We wish to store (XL
376
+ 1 , XL
377
+ 2 ) at the first database in a lossless manner, however, store only certain
378
+ necessary information regarding Y L
379
+ 1 and Y L
380
+ 2 to facilitate the recovery of W1 or W2. For this purpose, we
381
+ will encode the message as follows:
382
+ • At database-1, compress and store (XL
383
+ 1 , XL
384
+ 2 ) losslessly;
385
+ • At database-2, encode Y L
386
+ 1 using a Slepian-Wolf code (or more precisely Sgarro’s code with uncer-
387
+ tainty side information [29]), with either XL
388
+ 1 or XL
389
+ 2 at the decoder, whose resulting code index is
390
+ denoted as CY1; encode Y L
391
+ 2 in the same manner, independent of Y L
392
+ 1 , whose code index is denoted
393
+ as CY2.
394
+ It is clear that for database-1, we need roughly ¯α1 = H(X1, X2). At database-2, in order to guarantee
395
+ successful decoding of the Slepian-Wolf code, we can chose roughly
396
+ ¯α2 = max(H(Y1|X1), H(Y1|X2)) + max(H(Y2|X1), H(Y2|X2))
397
+ = 2H(Y1|X1),
398
+ (18)
399
+ where the second equality is due to the symmetry in the probability distribution. Thus we find that this
400
+ code achieves
401
+ ¯αnl = 0.5[H(X1, X2) + 2H(Y1|X1)]
402
+ = 0.75 + 0.75H(1/3, 2/3)
403
+ = 0.25 + 0.75 log2 3 ≈ 1.4387.
404
+ (19)
405
+ The retrieval strategy is immediate from the coding structure in Fig. 1, with (V L
406
+ 1 , V L
407
+ 2 , XL
408
+ 1 , XL
409
+ 2 , CY1, CY2)
410
+ serving the roles of (W1, W2, X1, X2, Y1, Y2), and thus indeed the privacy constraint is satisfied. The re-
411
+ trieval rates are roughly as follows
412
+ ¯β(1)
413
+ 1
414
+ = ¯β(2)
415
+ 1
416
+ = H(X1) = H(X2),
417
+ (20)
418
+ ¯β(1)
419
+ 2
420
+ = ¯β(2)
421
+ 2
422
+ = H(Y1|X1),
423
+ (21)
424
+ implying
425
+ ¯β = 0.5[H(X1) + H(Y1|X1)] = 0.5H(Y1, X1) = 0.75.
426
+ Thus at the optimal retrieval rate ¯β = 0.75, we have
427
+ ¯αl = 1.5 vs. ¯αnl ≈ 1.4387,
428
+ (22)
429
+ and clearly the proposed non-linear Shannon-theoretic code is able to perform better than the optimal
430
+ linear code. We note that it was shown in [19] by using a multround approach, the storage rate ¯α can be
431
+ further reduced, however this issue is beyond the scope of this work. In the rest of the paper, we build
432
+ on the intuition in this special case to generalize and strengthen the coding scheme.
433
+ 6
434
+
435
+ 4
436
+ Main Result
437
+ 4.1
438
+ A General Inner Bound
439
+ We first present a general inner bound to the storage-retrieval tradeoff region. Let (V1, V2) be independent
440
+ random variables uniformly distributed on Ft
441
+ 2 × Ft
442
+ 2. Define the region R(t)
443
+ in to be the collection of (¯α, ¯β)
444
+ pairs for which there exist random variables (X0, X1, X2, Y1, Y2) jointly distributed with (V1, V2) such
445
+ that:
446
+ 1. There exist deterministic functions f1,1, f1,2, f2,1, and f2,2 such that
447
+ V1 = f1,1(X0, X1, Y1) = f2,2(X0, X2, Y2),
448
+ V2 = f1,2(X0, X1, Y2) = f2,1(X0, X2, Y1);
449
+ (23)
450
+ 2. There exist non-negative coding rates
451
+ (β(0)
452
+ 1 , β(1)
453
+ 1 , β(2)
454
+ 1 , β(1)
455
+ 2 , β(2)
456
+ 2 , γ(0)
457
+ 1 , γ(1)
458
+ 1 , γ(2)
459
+ 1 , γ(1)
460
+ 2 , γ(2)
461
+ 2 )
462
+ ∈ R∗
463
+ MD (((V1, V2), X0, X1, X2, Y1, Y2), ({X0, X1, Y1}, {X0, X1, Y2}, {X0, X2, Y1}, {X0, X2, Y2})) ;
464
+ (24)
465
+ 3. There exist non-negative storage rates (α(0)
466
+ 1 , α(1)
467
+ 1 , α(2)
468
+ 1 , α(1)
469
+ 2 , α(2)
470
+ 2 ) such that
471
+ α(0)
472
+ 1
473
+ ≤ β(0)
474
+ 1 , α(1)
475
+ 1
476
+ ≤ β(1)
477
+ 1 , α(2)
478
+ 1
479
+ ≤ β(2)
480
+ 1 , α(1)
481
+ 2
482
+ ≤ β(1)
483
+ 2 , α(2)
484
+ 2
485
+ ≤ β(2)
486
+ 2 ,
487
+ (25)
488
+ and if
489
+ γ(0)
490
+ 1
491
+ − β(0)
492
+ 1
493
+ + γ(1)
494
+ 1
495
+ − β(1)
496
+ 1
497
+ + γ(2)
498
+ 1
499
+ − β(2)
500
+ 1
501
+ < H(X1) + H(X2) + H(X3) − H(X0, X1, X2),
502
+ (26)
503
+ choose
504
+ (α(0)
505
+ 1 , α(1)
506
+ 1 , α(2)
507
+ 1 , γ(0)
508
+ 1 , γ(1)
509
+ 1 , γ(2)
510
+ 1 ) ∈ R∗
511
+ MD (((V1, V2), X0, X1, X2), ({X0, X1, X2})) ;
512
+ (27)
513
+ otherwise, choose (α(0)
514
+ 1 , α(1)
515
+ 1 , α(2)
516
+ 1 ) = (β(0)
517
+ 1 , β(1)
518
+ 1 , β(2)
519
+ 1 ). Similarly, if
520
+ γ(1)
521
+ 2
522
+ − β(1)
523
+ 2
524
+ + γ(2)
525
+ 2
526
+ − β(2)
527
+ 2
528
+ < I(Y1; Y2),
529
+ (28)
530
+ choose
531
+ (α(1)
532
+ 2 , α(2)
533
+ 2 , γ(1)
534
+ 2 , γ(2)
535
+ 2 ) ∈ R∗
536
+ MD (((V1, V2), Y1, Y2), ({Y1, Y2})) ,
537
+ (29)
538
+ otherwise (α(1)
539
+ 2 , α(2)
540
+ 2 ) = (β(1)
541
+ 1 , β(2)
542
+ 1 );
543
+ 4. The normalized average retrieval and storage rates
544
+ 2t¯α ≥ α(0)
545
+ 1
546
+ + α(1)
547
+ 1
548
+ + α(2)
549
+ 1
550
+ + α(1)
551
+ 2
552
+ + α(2)
553
+ 2 ,
554
+ (30)
555
+ 4t¯β ≥ 2β(0)
556
+ 1
557
+ + β(1)
558
+ 1
559
+ + β(2)
560
+ 1
561
+ + β(1)
562
+ 2
563
+ + β(2)
564
+ 2 .
565
+ (31)
566
+ Then we have the following theorem.
567
+ Theorem 1. R(t)
568
+ in ⊆ R.
569
+ We can in fact potentially enlarge the achievable region by taking ∪∞
570
+ t=1R(t)
571
+ in . However, unless R(t+1)
572
+ in
573
+
574
+ R(t)
575
+ in for all t ≥ 1, the region ∪∞
576
+ t=1R(t)
577
+ in is even more difficult to characterize. Nevertheless, for each fixed
578
+ t, we can identify inner bounds by specifying a feasible set of random variables X0, X1, X2, Y1, Y2.
579
+ Instead of directly establishing this theorem, we shall prove the following theorem which establishes
580
+ the existence of a PIR code with diminishing error probability, and then use an expurgation technique
581
+ to extract a zero-error PIR code.
582
+ 7
583
+
584
+ Theorem 2. Consider any (¯α, ¯β) ∈ R(t)
585
+ in .
586
+ For any ϵ > 0 and sufficiently large L, there exists an
587
+ (L, L(¯α + ϵ), L(¯α + ϵ), L(¯β + ϵ), L(¯β + ϵ)) ϵ-error PIR code with the query distribution given in (10) and
588
+ (11).
589
+ The key observation to establish this theorem is that there are five descriptions in this setting, however,
590
+ the retrieval and storage place different constraints on different combination of descriptions, and some
591
+ descriptions can in fact be stored, recompressed, and then retrieved. Such compression and recompression
592
+ may lead to storage savings. The description based on X0 can be viewed as some common information
593
+ to X1 and X2, which allows us to tradeoff the storage and retrieval rates.
594
+ Proof of Theorem 2. Codebook generation: Codebooks are built using the MD codebooks based on the
595
+ distribution ((V1, V2), X0, X1, X2, Y1, Y2).
596
+ Storage codes:
597
+ The bin indices of the codebooks are stored in the two servers: those of X0, X1, and X2
598
+ are stored at server-1 at rates α(0)
599
+ 1 , α(1)
600
+ 1 , and α(2)
601
+ 1 , respectively; those of Y1 and Y2 are stored at server-2
602
+ at rates α(1)
603
+ 2
604
+ and α(2)
605
+ 2 . Note that at such rates, the codewords for X0, X1, and X2 can be recovered jointly
606
+ with overwhelming probability, while those for Y1 and Y2 can also be recovered jointly with overwhelming
607
+ probability.
608
+ Retrieval codes:
609
+ A different set of bin indices of the codebooks are retrieved during the retrieval process,
610
+ again based on the MD∗ codebooks: those of X0, X1, and X2 are retrieved at server-1 at rates β(0)
611
+ 1 , β(1)
612
+ 1 ,
613
+ and β(2)
614
+ 1 , respectively; those of Y1 and Y2 are retrieved at server-2 at rates β(1)
615
+ 2
616
+ and β(2)
617
+ 2 . Note that at such
618
+ rates, the codewords of X0, X1, and Y1 can be jointly recovered such that using the three corresponding
619
+ codewords, the required V1 source vector can be recovered with overwhelming probability. Similarly, the
620
+ three retrieval patterns of (X0, X1, Y2) → V2, (X0, X2, Y1) → V2, and (X0, X2, Y2) → V2 will succeed with
621
+ overwhelming probabilities.
622
+ Storage and retrieval rates:
623
+ The rates can be computed straightforwardly, after normalization by the
624
+ parameter t.
625
+ Next we use it to prove Theorem 1.
626
+ Proof of Theorem 1. Given an ϵ > 0, according to Proposition 2, we can find an (L, L(¯α + ϵ), L(¯α +
627
+ ϵ), L(¯β + ϵ), L(¯β + ϵ)) ϵ-error PIR code for some sufficient large L. The probability of error of this code
628
+ can be rewritten as
629
+ Pe = 0.5
630
+
631
+ k=1,2
632
+
633
+ (w1,w2)
634
+ 2−2LPQ[k]
635
+ 1 ,Q[k]
636
+ 2 |(w1,w2)(wk ̸= ˆWk).
637
+ For a fixed (w1, w2) pair, denote the event that there exists a (q1, q2) ∈ {(11), (22)}, i.e., when (Q[1]
638
+ 1 , Q[1]
639
+ 2 ) =
640
+ (q1, q2), such that ˆw1 ̸= w1 as E(1)
641
+ w1,w2, and there exists a (q1, q2) ∈ {(12), (21)} such that ˆw2 ̸= w2 as
642
+ E(2)
643
+ w1,w2. Since (Q[k]
644
+ 1 , Q[k]
645
+ 2 ) is independent of (W1, W2), if P(E(k)
646
+ w1,w2) ̸= 0, we must have P(E(k)
647
+ w1,w2) ≥ 0.5.
648
+ It follows that
649
+ Pe ≥ 0.25
650
+
651
+ (w1,w2)
652
+ 2−2L1(E[1]
653
+ w1,w2 ∪ E[2]
654
+ w1,w2),
655
+ (32)
656
+ where (·) is the indicator function. This implies that for any ϵ ≤ 0.125, there are at most 22L−1 pairs of
657
+ (w1, w2) that will induce any coding error. We can use any 22L−2 of the remaining 22L−1 pairs of L-bit
658
+ sequence pairs to instead store a pair of (L − 1)-bit messages, through an arbitrary but fixed one-to-one
659
+ mapping. This new code has a factor of 1 + 1/(L − 1) increase in the normalized coding rates, which is
660
+ negligible when L is large. Thus a zero-error PIR code has been found with the same normalized rates
661
+ as the ϵ-error code asymptotically, and this completes the proof.
662
+ 8
663
+
664
+ 4.2
665
+ Outer bounds
666
+ We next turn our attention to the outer bounds for R, summarized in the following theorem.
667
+ Theorem 3. Any (¯α, ¯β) ∈ R must satisfy
668
+ ¯β ≥ 0.75,
669
+ ¯α + ¯β ≥ 2,
670
+ 3¯α + 8¯β ≥ 10.
671
+ (33)
672
+ Moreover, if (¯α, ¯β) ∈ R can be achieved by a linear code, it must satisfy
673
+ ¯α + 6¯β ≥ 6.
674
+ (34)
675
+ The inequality ¯β ≥ 0.75 follows from [11], while the two other bounds in (33) were proved in [24].
676
+ Therefore we only need to prove (34).
677
+ Proof of Theorem 3. Following [19], we make the following simplifying assumptions that have no loss of
678
+ generality. Define Q = {Q[1]
679
+ 1 , Q[2]
680
+ 1 , Q[1]
681
+ 2 , Q[2]
682
+ 2 }.
683
+ 1. Q[1]
684
+ 1 = Q[2]
685
+ 1
686
+ ⇒ A[1]
687
+ 1 = A[2]
688
+ 1 ,
689
+ (35)
690
+ 2. H(A[1]
691
+ 1 |Q) = H(A[1]
692
+ 2 |Q) = H(A[2]
693
+ 2 |Q),
694
+ H(S1) = H(S2)
695
+ (36)
696
+ ⇒ H(A[1]
697
+ 1 |Q) ≤ β ≤ (¯β + ϵ)L,
698
+ H(S2) ≤ α ≤ (¯α + ϵ)L.
699
+ (37)
700
+ Assumption 1 states that the query to the first database is the same regardless of the desired message
701
+ index. This is justified by the privacy condition that the query to one database is independent of the
702
+ desired message index.
703
+ Assumption 2 states that the scheme is symmetric after the symmetrization
704
+ operation in Lemma 1 (the proof is referred to Theorem 3 in [19]). (37) follows from the fact that to
705
+ describe S2, A[1]
706
+ 1 , the number of bits needed can not be less than the entropy value, and Definition 1.
707
+ In the following, we use (c) to refer to the correctness condition, (i) to refer to the constraint that
708
+ queries are independent of the messages, (a) to refer to the constraint that answers are deterministic
709
+ functions of the storage variables and corresponding queries, and (p) to refer to the privacy condition.
710
+ From A[1]
711
+ 1 , A[1]
712
+ 2 , Q, we can decode W1.
713
+ H(A[1]
714
+ 1 , A[1]
715
+ 2 |W1, Q)
716
+ =
717
+ H(A[1]
718
+ 1 , A[1]
719
+ 2 , W1|Q) − H(W1|Q)
720
+ (38)
721
+ (c)(i)
722
+ =
723
+ H(A[1]
724
+ 1 , A[1]
725
+ 2 |Q) − L
726
+ (39)
727
+ (36)
728
+
729
+ 2H(A[1]
730
+ 1 |Q) − L.
731
+ (40)
732
+ Next, consider Ingleton’s inequality.
733
+ I(A[1]
734
+ 2 ; A[2]
735
+ 2 |Q)
736
+
737
+ I(A[1]
738
+ 2 ; A[2]
739
+ 2 |W1, Q) + I(A[1]
740
+ 2 ; A[2]
741
+ 2 |W2, Q)
742
+ (41)
743
+ =
744
+ 2I(A[1]
745
+ 2 ; A[2]
746
+ 2 |W1, Q)
747
+ (42)
748
+ =
749
+ 2
750
+
751
+ H(A[1]
752
+ 2 |W1, Q) + H(A[2]
753
+ 2 |W1, Q) − H(A[1]
754
+ 2 , A[2]
755
+ 2 |W1, Q)
756
+
757
+ (43)
758
+ (p)
759
+ =
760
+ 2
761
+
762
+ 2H(A[1]
763
+ 2 |W1, Q) − H(A[1]
764
+ 2 , A[2]
765
+ 2 |W1, Q)
766
+
767
+ (44)
768
+
769
+ 2
770
+
771
+ 2H(A[1]
772
+ 2 |W1, Q) + H(A[1]
773
+ 1 , A[1]
774
+ 2 |W1, Q)
775
+ − H(A[1]
776
+ 1 , A[1]
777
+ 2 , A[2]
778
+ 2 |W1, Q) − H(A[1]
779
+ 2 |W1, Q)
780
+
781
+ (45)
782
+ (c)(35)
783
+ =
784
+ 2
785
+
786
+ H(A[1]
787
+ 2 |W1, Q) + H(A[1]
788
+ 1 , A[1]
789
+ 2 |W1, Q)
790
+ − H(A[1]
791
+ 1 , A[1]
792
+ 2 , A[2]
793
+ 2 , W2|W1, Q)
794
+
795
+ (46)
796
+ 9
797
+
798
+ (i)
799
+
800
+ 2
801
+
802
+ 2H(A[1]
803
+ 1 , A[1]
804
+ 2 |W1, Q) − H(W2)
805
+
806
+ (47)
807
+ (40)
808
+
809
+ 2
810
+
811
+ 2(2H(A[1]
812
+ 1 |Q) − L) − L
813
+
814
+ (48)
815
+ where (42) follows from the observation that the second term can be bounded using the same method as
816
+ that bounds the first term by switching the message index. A more detailed derivation of (44) appears
817
+ in (79) of [19]. (45) is due to sub-modularity of entropy.
818
+ Note that
819
+ I(A[1]
820
+ 2 ; A[2]
821
+ 2 |Q)
822
+ =
823
+ H(A[1]
824
+ 2 |Q) + H(A[2]
825
+ 2 |Q) − H(A[1]
826
+ 2 , A[2]
827
+ 2 |Q)
828
+ (49)
829
+ (36)
830
+
831
+ 2H(A[1]
832
+ 1 |Q) − (¯α + ϵ)L
833
+ (50)
834
+ where in (50), and the second term is bounded as follows :
835
+ H(A[1]
836
+ 2 , A[2]
837
+ 2 |Q) ≤ H(A[1]
838
+ 2 , A[2]
839
+ 2 , S2|Q)
840
+ (a)
841
+ = H(S2|Q)
842
+ (37)
843
+ ≤ (¯α + ϵ)L.
844
+ (51)
845
+ Combining (48) and (50), we have
846
+ 2H(A[1]
847
+ 1 |Q)/L − (¯α + ϵ) ≥ 2(4H(A[1]
848
+ 1 |Q)/L − 3)
849
+
850
+ ¯α + ϵ + 6H(A[1]
851
+ 1 |Q)/L ≥ 6
852
+ (52)
853
+ (37)
854
+
855
+ ¯α + 6¯β ≥ 6.
856
+ (53)
857
+ The proof is complete.
858
+ 4.3
859
+ Specialization of the Inner Bound
860
+ The inner bound given in Theorem 1 is general but more involved, and we can specialize it in multiple
861
+ ways in order to simplify it. One particularly interesting approach is as follows. Define the region ˜R(t)
862
+ in
863
+ to be the collection of (¯α, ¯β) pairs such that there exists random variables (X0, X1, X2, Y1, Y2) jointly
864
+ distributed with (V1, V2) such that
865
+ 1. The distribution factorizes as follows
866
+ PV1,V2,X0,X1,X2,Y1,Y2 = PV1,V2PX0|V1,V2PX1|V1,V2PX2|V1,V2PY1|V1,V2PY2|V1,V2;
867
+ 2. There exist deterministic functions f1,1, f1,2, f2,1, and f2,2 such that
868
+ V1 = f1,1(X0, X1, Y1) = f2,2(X0, X2, Y2),
869
+ (54)
870
+ V2 = f1,2(X0, X1, Y2) = f2,1(X0, X2, Y1);
871
+ (55)
872
+ 3. A set of rates
873
+ γ(0)
874
+ 1
875
+ = I(V1, V2; X0), γ(1)
876
+ 1
877
+ = I(V1, V2; X1), γ(2)
878
+ 1
879
+ = I(V1, V2; X2),
880
+ (56)
881
+ γ(1)
882
+ 2
883
+ = I(V1, V2; Y1), γ(2)
884
+ 2
885
+ = I(V1, V2; Y2),
886
+ (57)
887
+ β(0)
888
+ 1
889
+ = γ(0)
890
+ 1 , β(1)
891
+ 1
892
+ = I(V1, V2; X1|X0), β(2)
893
+ 1
894
+ = I(V1, V2; X2|X0),
895
+ (58)
896
+ β(1)
897
+ 2
898
+ = max(I(V1, V2; Y1|X0, X1), I(V1, V2; Y1|X0, X2)),
899
+ (59)
900
+ β(2)
901
+ 2
902
+ = max(I(V1, V2; Y2|X0, X1), I(V1, V2; Y2|X0, X2)),
903
+ (60)
904
+ and (α(0)
905
+ 1
906
+ = γ(0)
907
+ 1 , α(1)
908
+ 1 , α(2)
909
+ 1 , α(1)
910
+ 2 , α(2)
911
+ 2 ) as defined in item 3 for the general region R(t);
912
+ 10
913
+
914
+ 1
915
+ 1.05
916
+ 1.1
917
+ 1.15
918
+ 1.2
919
+ 1.25
920
+ 1.3
921
+ 1.35
922
+ 1.4
923
+ 1.45
924
+ 1.5
925
+ 0.75
926
+ 0.8
927
+ 0.85
928
+ 0.9
929
+ 0.95
930
+ 1
931
+ inner bound via a non-linear scheme
932
+ non-linear scheme: time-sharing
933
+ inner bound via a linear scheme
934
+ an outer bound on linear schemes
935
+ information theoretic outer bound
936
+ Figure 2: Illustration of inner bounds and outer bounds.
937
+ 4. The normalized average retrieval and storage rates
938
+ 2t¯α ≥ α(0)
939
+ 1
940
+ + α(1)
941
+ 1
942
+ + α(2)
943
+ 1
944
+ + α(1)
945
+ 2
946
+ + α(2)
947
+ 2 ,
948
+ (61)
949
+ 4t¯β ≥ 2β(0)
950
+ 1
951
+ + β(1)
952
+ 1
953
+ + β(2)
954
+ 1
955
+ + β(1)
956
+ 2
957
+ + β(2)
958
+ 2 .
959
+ (62)
960
+ Then we have the following corollary.
961
+ Corollary 1. ˜R(t)
962
+ in ⊆ R.
963
+ This inner bound is illustrated together with the outer bounds in Fig. 2.
964
+ Proof. The main difference from Theorem 1 is in the special dependence structure of (X0, X1, X2, Y1, Y2)
965
+ jointly distributed with (V1, V2), i.e., the Markov structure. We verify that the rate assignments satisfy
966
+ all the constraints in Theorem 1. Due to the special dependence structure of (X0, X1, X2, Y1, Y2) jointly
967
+ distributed with (V1, V2), it is straightforward to verify that
968
+ (γ(0)
969
+ 1 , γ(1)
970
+ 1 , γ(2)
971
+ 1 , γ(1)
972
+ 2 , γ(2)
973
+ 2 ) ∈ RMD((V1, V2), X0, X1, X2, Y1, Y2).
974
+ We next verify (24) holds with the choice given above. Due to the symmetry in the structure, we only need
975
+ to confirm one subset of random variables, i.e., {X0, X1, Y1}, and the three other subsets {X0, X1, Y2},
976
+ {X0, X2, Y1}, and {X0, X2, Y2} follow similarly. There are a total of 7 conditions in the form of (14)
977
+ associated with this subset {X0, X1, Y1}. Notice that
978
+ γ(0)
979
+ 1
980
+ − β(0)
981
+ 1
982
+ = 0, γ(1)
983
+ 1
984
+ − β(1)
985
+ 1
986
+ = I(X1; X0), γ(2)
987
+ 2
988
+ − β(2)
989
+ 2
990
+ ≤ I(Y1; X0, X1),
991
+ which in fact confirm three of the seven conditions when J is a singleton. Next when J has two elements,
992
+ we verify that
993
+ γ(0)
994
+ 1
995
+ − β(0)
996
+ 1
997
+ + γ(1)
998
+ 1
999
+ − β(1)
1000
+ 1
1001
+ = I(X1; X0) = H(X0) + H(X1) − H(X0, X1)
1002
+ ≤ H(X0) + H(X1) − H(X0, X1|Y1),
1003
+ (63)
1004
+ γ(0)
1005
+ 1
1006
+ − β(0)
1007
+ 1
1008
+ + γ(1)
1009
+ 2
1010
+ − β(1)
1011
+ 2
1012
+ ≤ I(Y1; X0, X1) = H(Y1) + H(X0, X1) − H(X0, X1, Y1)
1013
+ ≤ H(Y1) + H(X0) + H(X1) − H(X0, X1, Y1)
1014
+ 11
1015
+
1016
+ Table 1: Conditional distribution PX0|W1,W2 used in Corollary 2.
1017
+ (w1, w2) x0 = (00) x0 = (01) x0 = (10) x0 = (11)
1018
+ (00)
1019
+ 1/2
1020
+ 1/2
1021
+ (10)
1022
+ (1 − p)/2
1023
+ p
1024
+ (1 − p)/2
1025
+ (01)
1026
+ (1 − p)/2
1027
+ p
1028
+ (1 − p)/2
1029
+ (11)
1030
+ 1/2
1031
+ 1/2
1032
+ = H(X0) + H(Y1) − H(X0, Y1|X1),
1033
+ (64)
1034
+ γ(1)
1035
+ 1
1036
+ − β(1)
1037
+ 1
1038
+ + γ(1)
1039
+ 2
1040
+ − β(1)
1041
+ 2
1042
+ ≤ I(X1; X0) + I(Y1; X0, X1) = H(X1) + H(Y1) − H(X1, Y1|X0).
1043
+ (65)
1044
+ Finally when J has all the three elements, we have
1045
+ γ(0)
1046
+ 1
1047
+ − β(0)
1048
+ 1
1049
+ + γ(1)
1050
+ 1
1051
+ − β(1)
1052
+ 1
1053
+ + γ(1)
1054
+ 2
1055
+ − β(1)
1056
+ 2
1057
+ = I(X0; X1) + I(V1, V2; X1) − max(I(V1, V2; Y1|X0, X1), I(V1, V2; Y1|X0, X2))
1058
+ (66)
1059
+ ≤ I(X0; X1) + I(V1, V2; X1) − I(V1, V2; Y1|X0, X1)
1060
+ (67)
1061
+ = H(X0) + H(X1) + H(Y1) − H(X0, X1, Y1).
1062
+ (68)
1063
+ Thus (24) is indeed true with the assignments (56)-(60). This in fact completes the proof.
1064
+ We can use any explicit distribution (X0, X1, X2, Y1, Y2) to obtain an explicit inner bound to ˜R(t)
1065
+ in , and
1066
+ the next corollary provides one such non-trivial bound. For convenience, we write the entropy function
1067
+ of a probability mass (p1, . . . , pt) as H(p1, . . . , pt).
1068
+ Corollary 2. The following (¯α, ¯β) ∈ R for any p ∈ [0, 1]:
1069
+ ¯α =9
1070
+ 4 − H(1
1071
+ 4, 3
1072
+ 4) + 1
1073
+ 4H(1 − p
1074
+ 2
1075
+ , 1 − p
1076
+ 2
1077
+ , p
1078
+ 2, p
1079
+ 2)
1080
+ + 1
1081
+ 2H(2 − p
1082
+ 4
1083
+ , 2 − p
1084
+ 4
1085
+ , p
1086
+ 2) − 3
1087
+ 4H(3 − 2p
1088
+ 6
1089
+ , 3 − 2p
1090
+ 6
1091
+ , p
1092
+ 3, p
1093
+ 3),
1094
+ ¯β =5
1095
+ 8 + 1
1096
+ 4H(2 − p
1097
+ 4
1098
+ , 2 − p
1099
+ 4
1100
+ , p
1101
+ 2) − 1
1102
+ 8H(1 − p
1103
+ 2
1104
+ , 1 − p
1105
+ 2
1106
+ , p).
1107
+ Proof. These tradeoff pairs are obtained by applying Corollary 1, and choosing t = 1 and setting
1108
+ (X1, X2, Y1, Y2) as given in (17), and letting X0 be defined as in Table 1. Note that the joint distri-
1109
+ bution indeed satisfies the required Markov structure, and in this case α(1)
1110
+ 2
1111
+ = β(1)
1112
+ 2
1113
+ and α(2)
1114
+ 2
1115
+ = β(2)
1116
+ 2 .
1117
+ 5
1118
+ Conclusion
1119
+ We consider the problem of private information retrieval using a Shannon-theoretic approach. A new
1120
+ coding scheme based on random coding and binning is proposed, which reveals a hidden connection to the
1121
+ multiple description problem. It is shown that for the (2, 2) PIR setting, this non-linear coding scheme is
1122
+ able to provide the best known tradeoff between retrieval rate and storage rate, which is strictly better
1123
+ than that achievable using linear codes. We further investigate the relation between zero-error PIR codes
1124
+ and ϵ-error PIR codes in this setting, and shows that they do not causes any essential difference in this
1125
+ problem setting. We hope that the hidden connection to multiple description coding can provide a new
1126
+ revenue to design more efficient PIR codes.
1127
+ 12
1128
+
1129
+ References
1130
+ [1] B. Chor, O. Goldreich, E. Kushilevitz, and M. Sudan, “Private information retrieval,” in Foundations
1131
+ of Computer Science, 1995. Proceedings., 36th Annual Symposium on, Oct. 1995, pp. 41–50.
1132
+ [2] N. Shah, K. Rashmi, and K. Ramchandran, “One extra bit of download ensures perfectly private
1133
+ information retrieval,” in Proceedings of 2014 IEEE International Symposium on Information Theory
1134
+ (ISIT), Jun.-Jul. 2014, pp. 856–860.
1135
+ [3] A. Fazeli, A. Vardy, and E. Yaakobi, “Codes for distributed PIR with low storage overhead,” in
1136
+ 2015 Proceedings of IEEE International Symposium on Information Theory (ISIT), Jun. 2015, pp.
1137
+ 2852–2856.
1138
+ [4] S. Rao and A. Vardy,
1139
+ “Lower bound on the redundancy of PIR codes,”
1140
+ arXiv preprint
1141
+ arXiv:1605.01869, 2016.
1142
+ [5] S. R. Blackburn and T. Etzion, “Pir array codes with optimal virtual server rate,” IEEE Transactions
1143
+ on Information Theory, vol. 65, no. 10, pp. 6136–6145, 2019.
1144
+ [6] S. R. Blackburn, T. Etzion, and M. B. Paterson, “Pir schemes with small download complexity and
1145
+ low storage requirements,” IEEE Transactions on Information Theory, vol. 66, no. 1, pp. 557–571,
1146
+ 2019.
1147
+ [7] Y. Zhang, X. Wang, H. Wei, and G. Ge, “On private information retrieval array codes,” IEEE
1148
+ Transactions on Information Theory, vol. 65, no. 9, pp. 5565–5573, 2019.
1149
+ [8] M. Vajha, V. Ramkumar, and P. V. Kumar, “Binary, shortened projective reed muller codes for
1150
+ coded private information retrieval,” in 2017 IEEE International Symposium on Information Theory
1151
+ (ISIT), 2017, pp. 2648–2652.
1152
+ [9] H. Asi and E. Yaakobi, “Nearly optimal constructions of pir and batch codes,” IEEE Transactions
1153
+ on Information Theory, vol. 65, no. 2, pp. 947–964, 2018.
1154
+ [10] T. H. Chan, S.-W. Ho, and H. Yamamoto, “Private information retrieval for coded storage,” in
1155
+ Proceedings of 2015 IEEE International Symposium on Information Theory (ISIT), Jun. 2015, pp.
1156
+ 2842–2846.
1157
+ [11] H. Sun and S. A. Jafar, “The capacity of private information retrieval,” IEEE Transactions on
1158
+ Information Theory, vol. 63, no. 7, pp. 4075–4088, Jul. 2017.
1159
+ [12] R. Tajeddine, O. W. Gnilke, and S. El Rouayheb, “Private information retrieval from MDS coded
1160
+ data in distributed storage systems,” IEEE Transactions on Information Theory, vol. 64, no. 11, pp.
1161
+ 7081 – 7093, 2018.
1162
+ [13] K. Banawan and S. Ulukus, “The capacity of private information retrieval from coded databases,”
1163
+ IEEE Transactions on Information Theory, vol. 64, no. 3, pp. 1945–1956, Mar. 2018.
1164
+ [14] C. Tian, H. Sun, and J. Chen, “Capacity-achieving private information retrieval codes with optimal
1165
+ message size and upload cost,” IEEE Transactions on Information Theory, vol. 65, no. 11, pp.
1166
+ 7613–7627, Nov. 2019.
1167
+ [15] R. Zhou, C. Tian, H. Sun, and T. Liu, “Capacity-achieving private information retrieval codes from
1168
+ mds-coded databases with minimum message size,” IEEE Transactions on Information Theory,
1169
+ vol. 66, no. 8, pp. 4904–4916, 2020.
1170
+ 13
1171
+
1172
+ [16] H. Sun and S. A. Jafar, “The capacity of robust private information retrieval with colluding
1173
+ databases,” IEEE Transactions on Information Theory, vol. 64, no. 4, pp. 2361–2370, 2018.
1174
+ [17] S. Ulukus, S. Avestimehr, M. Gastpar, S. Jafar, R. Tandon, and C. Tian, “Private retrieval, com-
1175
+ puting and learning: Recent progress and future challenges,” IEEE Journal on Selected Areas in
1176
+ Communications, 2022.
1177
+ [18] M. A. Attia, D. Kumar, and R. Tandon, “The capacity of private information retrieval from uncoded
1178
+ storage constrained databases,” IEEE Transactions on Information Theory, vol. 66, no. 11, pp. 6617–
1179
+ 6634, 2020.
1180
+ [19] H. Sun and S. A. Jafar, “Multiround private information retrieval: Capacity and storage overhead,”
1181
+ IEEE Transactions on Information Theory, vol. 64, no. 8, pp. 5743–5754, 2018.
1182
+ [20] H. Sun and C. Tian, “Breaking the MDS-PIR capacity barrier via joint storage coding,” Information,
1183
+ vol. 10, no. 9, p. 265, 2019.
1184
+ [21] T. Guo, R. Zhou, and C. Tian, “New results on the storage-retrieval tradeoff in private information
1185
+ retrieval systems,” IEEE Journal on Selected Areas in Information Theory, vol. 2, no. 1, pp. 403–414,
1186
+ 2021.
1187
+ [22] C. Tian, H. Sun, and J. Chen, “A shannon-theoretic approach to the storage-retrieval tradeoff in pir
1188
+ systems,” in 2018 IEEE International Symposium on Information Theory (ISIT), 2018, pp. 1904–
1189
+ 1908.
1190
+ [23] A. Gamal and T. Cover, “Achievable rates for multiple descriptions,” IEEE Transactions on Infor-
1191
+ mation Theory, vol. 28, no. 6, pp. 851–857, 1982.
1192
+ [24] C. Tian, “On the storage cost of private information retrieval,” IEEE Transactions on Information
1193
+ Theory, vol. 66, no. 12, pp. 7539–7549, 2020.
1194
+ [25] R. Venkataramani, G. Kramer, and V. K. Goyal, “Multiple description coding with many channels,”
1195
+ IEEE Transactions on Information Theory, vol. 49, no. 9, pp. 2106–2114, 2003.
1196
+ [26] A. Wyner and J. Ziv, “The rate-distortion function for source coding with side information at the
1197
+ decoder,” IEEE Transactions on information Theory, vol. 22, no. 1, pp. 1–10, 1976.
1198
+ [27] S. S. Pradhan, R. Puri, and K. Ramchandran, “n-channel symmetric multiple descriptions-part i:
1199
+ (n, k) source-channel erasure codes,” IEEE Transactions on Information Theory, vol. 50, no. 1, pp.
1200
+ 47–61, 2004.
1201
+ [28] C. Tian and J. Chen, “New coding schemes for the symmetric k-description problem,” IEEE Trans-
1202
+ actions on Information Theory, vol. 56, no. 10, pp. 5344–5365, 2010.
1203
+ [29] A. Sgarro, “Source coding with side information at several decoders,” IEEE Transactions on Infor-
1204
+ mation Theory, vol. 23, no. 2, pp. 179–182, 1977.
1205
+ 14
1206
+
6NA0T4oBgHgl3EQfN_-e/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
7NE2T4oBgHgl3EQf7giQ/content/tmp_files/2301.04210v1.pdf.txt ADDED
@@ -0,0 +1,1080 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Next nearest neighbour coupling with spinor polariton condensates
2
+ Dmitriy Dovzhenko,1, ∗ Denis Aristov,1 Lucy Pickup,1 Helgi Sigurdsson,1, 2 and Pavlos Lagoudakis3, 1
3
+ 1School of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, UK
4
+ 2Science Institute, University of Iceland, Dunhagi 3, IS-107, Reykjavik, Iceland
5
+ 3Hybrid Photonics Laboratory, Skolkovo Institute of Science and Technology,
6
+ Territory of Innovation Center Skolkovo, Bolshoy Boulevard 30, building 1, 121205 Moscow, Russia
7
+ (Dated: January 12, 2023)
8
+ We report on experimental observation of next-nearest-neighbour coupling between ballistically
9
+ expanding spinor exciton-polariton condensates in a planar semiconductor microcavity. All-optical
10
+ control over the coupling strength between neighbouring condensates is demonstrated through
11
+ distance-periodic pseudospin screening of their ballistic particle outflow due to the inherent splitting
12
+ of the planar cavity transverse-electric (TE) and transverse-magnetic (TM) modes. By screening
13
+ the nearest-neighbour coupling we overcome the conventional spatial coupling hierarchy between
14
+ condensates. This offers a promising route towards creating unconventional non-planar many-body
15
+ Hamiltonians using networks of ballistically expanding spinor exciton-polariton condensates.
16
+ Strongly correlated quantum many-body systems have
17
+ attracted a lot of interest as a promising tool to engi-
18
+ neer and explore phases of matter in extreme settings
19
+ [1–3] and to simulate complex Hamiltonians [4, 5]. Such
20
+ systems include ultracold atomic ensembles [4], trapped
21
+ ions [6, 7], nuclear and electronic spins [8, 9], supercon-
22
+ ducting circuits [10, 11], and nonlinear photonic systems
23
+ [12]. Of interest, recent milestone achievements in pro-
24
+ grammable connectivity in condensed matter using cold
25
+ atomic gases [13] now permit construction of intriguing
26
+ networks of coupled elements. However, in general, many
27
+ lab systems are by their physical nature unable to form
28
+ unconventional graph topologies.
29
+ In the past decade,
30
+ driven-dissipative Bose-Einstein condensates of exciton-
31
+ polaritons (from here on, polaritons) in planar microcavi-
32
+ ties have substantially advanced in optical reprogramma-
33
+ bility [14–21]. There, each condensate is driven by a fo-
34
+ cused non-resonant optical excitation beam forming a lo-
35
+ calized macroscopically coherent wavefunction [22]. The
36
+ coupling strength between neighbouring condensates is
37
+ roughly given by their mutual overlap with an expo-
38
+ nential fall-off as a function of separation distance [23–
39
+ 25]. This means that nearest-neighbour (NN) coupling
40
+ dominates over next-nearest-neighbour (NNN) coupling
41
+ making polariton networks inherently planar in a graph
42
+ topology sense.
43
+ Overcoming this spatial coupling hi-
44
+ erarchy can offer opportunities to observe spontaneous
45
+ ordering and emergent polariton effects in non-planar
46
+ graph topologies [26–31].
47
+ However, this is extremely
48
+ challenging, requiring very fine control over the two-
49
+ dimensional polariton potential landscape with limita-
50
+ tions of its own [32].
51
+ In this Letter, we demonstrate that spin-orbit coupled
52
+ (SOC) exciton-polariton condensates can overcome this
53
+ challenge. Polaritons are quasiparticles exhibiting inter-
54
+ mixed properties of excitons and photons, which appear
55
+ when light and matter are brought to the strong coupling
56
+ regime [33]. As a consequence, the photon polarisation is
57
+ explicitly connected to the polariton pseudospin (or just
58
+ "spin" for short) with ˆσz = ±1 spin-projections along
59
+ the cavity growth axis representing σ± circularly polar-
60
+ ized light. Their two-component integer spin structure
61
+ has led to deep exploration into nonequilibrium spinor
62
+ quantum fluids [34].
63
+ Polaritons mostly decay through
64
+ photons leaking out of the cavity containing all the in-
65
+ formation on the condensate such as energy, momentum,
66
+ density, and spin. This salient feature allows direct, yet
67
+ non-destructive, measurement of the condensate spin dis-
68
+ tribution using polarization resolved photoluminescence
69
+ (PL) imaging.
70
+ Both the polariton condensate and the incoherent pho-
71
+ toexcited background of excitons sustaining it adopt the
72
+ circular polarisation of the nonresonant excitation [35,
73
+ 36] due to the optical orientation effect of excitons [37, 38]
74
+ and spin-preserving stimulated scattering of excitons into
75
+ the condensate [39]. This permits excitation of a con-
76
+ densate of a well defined macroscopic Sz ∼ ⟨ˆσz⟩ spin
77
+ projection [40–43]. Subsequently, the inherent TE-TM
78
+ splitting of the microcavity [44] will start rotating the
79
+ spin of any condensate polaritons which obtain finite
80
+ wavevector and flow away from the pump spot [45, 46].
81
+ This is also referred to as the optical spin Hall ef-
82
+ fect [47, 48].
83
+ Namely, the splitting between TE and
84
+ TM polarized cavity photon modes acts as a direction-
85
+ ally dependent in-plane effective magnetic field [47, 49]
86
+ (i.e., effective SOC [50]) causing the spins of outflowing
87
+ condensate polaritons to start precessing [see Fig. 1(a)
88
+ and Fig. 1(b)]. The strength of this effective SOC scales
89
+ quadratically with the polariton momentum, ∝ k2 and
90
+ can even be electrically tuned [51, 52]. This makes so-
91
+ called ballistic condensates ideal for enhanced SOC ef-
92
+ fects [45, 46] due to their extremely high kinetic en-
93
+ ergies obtained through repulsive Coulomb interactions
94
+ with the locally pump-induced exciton reservoir. More-
95
+ over, because of their long-range coherent particle out-
96
+ flow, ballistic condensates can couple over macroscopic
97
+ distances much greater than their respective full-width-
98
+ at-half-maximum [24] while also preserving their spin in-
99
+ arXiv:2301.04210v1 [cond-mat.mes-hall] 10 Jan 2023
100
+
101
+ 2
102
+ formation [43, 45, 46].
103
+ Recently, it was theoretically predicted that ballistic
104
+ condensates could invert their neighbour coupling hier-
105
+ archy, making NNN stronger than NN, through a spin-
106
+ screening effect made possible by the effective SOC stem-
107
+ ming from TE-TM splitting [53]. Here, we provide exper-
108
+ imental evidence of these recent predictions. We present
109
+ a study of a spinor polariton dyad (two coupled con-
110
+ densates) and a triad [three coupled condensates, see
111
+ schematic Fig. 1(c)] wherein each condensate ballistically
112
+ emits a coherent pseudospin current which rapidly pre-
113
+ cesses as it propagates [45, 46]. We demonstrate control
114
+ over the coupling strength between neighbouring conden-
115
+ sates by changing the spatial distance between them (de-
116
+ noted d) relative to the spatial precession period of the
117
+ condensate pseudospin (denoted ξ).
118
+ We briefly explain the idea of spin-screened polariton
119
+ coupling. The three peaks in Fig. 1(c) represent the con-
120
+ densate centers excited by three co-localized Gaussian
121
+ pump spots of equal intensity. The red-blue colour map
122
+ shows the precession of the polariton pseudospin as it
123
+ radially propagates in-plane away from each condensate
124
+ center, with red representing Sz = +1 (spin-up polari-
125
+ tons) and blue representing Sz = −1 (spin-down polari-
126
+ tons). The height of the peaks represents the intensity of
127
+ the condensate emission. The distance between the con-
128
+ densate centers relative to the spatial oscillations of the
129
+ pseudospin modifies the coupling between them. In the
130
+ non-screened state [Fig. 1(c)] NN condensates are excited
131
+ at a distance equal to integer number of periods of pseu-
132
+ dospin oscillations, d = nξ where n = 1, 2, 3, . . . . This
133
+ means that propagating condensate polaritons arrive at
134
+ NNs with unchanged spin projection. On the contrary,
135
+ in the screened state [Fig. 1(d)] NNs are separated by
136
+ d = (n−1/2)ξ and polaritons arrive at their NNs with op-
137
+ posite spin-projection which reduces the condensate cou-
138
+ pling, while coupling between NNNs is still maintained.
139
+ The microcavity used in this study consists of a 5λ/2
140
+ AlGaAs cavity surrounded by two distributed Bragg mir-
141
+ rors (DBR) of 35 and 32 pairs of ALGaAs/AlAs for the
142
+ bottom and top DBR correspondingly with the 12 GaAs
143
+ QWs separated into four sets of three QWs placed at the
144
+ antinodes of electric field within the cavity. The cavity
145
+ quality factor is around Q ∼ 16000 with the correspond-
146
+ ing polariton lifetime τp ≈ 5 ps and Rabi splitting of 9
147
+ meV. The measured TE-TM splitting is ≈ 0.2 meV at
148
+ k = 3 µm−1 in-plane wavevector. See section S1 in the
149
+ Supplemental Material [54] for further experimental de-
150
+ tails.
151
+ The normalized Stokes parameters of the cavity emis-
152
+ sion are written,
153
+ Sx,y,z(r) = IH,D,σ+(r) − IV,A,σ−(r)
154
+ IH,D,σ+(r) + IV,A,σ−(r),
155
+ (1)
156
+ where
157
+ r
158
+ =
159
+ (x, y)
160
+ is
161
+ the
162
+ in-plane
163
+ coordinate
164
+ and
165
+ IH(V ),D(A),σ+(σ−)(r)
166
+ corresponds
167
+ to
168
+ horizon-
169
+ Figure 1.
170
+ (a) Schematic of the effective SOC magnetic field
171
+ distribution (dark olive arrows) from the TE-TM splitting
172
+ on a momentum-space circle. (b) Schematic of the Poincaré
173
+ sphere showing example pseudospin precession as polaritons
174
+ propagate (blue and red arrows). Schematic representing two
175
+ pump geometries where the distance between the central and
176
+ edge pump spots equals to (c) one full period of pseudospin
177
+ oscillation (NN is stronger than NNN) and (d) half oscilla-
178
+ tion period (NN is weaker than NNN). The height of the peaks
179
+ represents the intensity of the condensate emission, and the
180
+ red, white, and blue colour map shows the precession of the
181
+ polariton pseudospin propagating in the cavity plane, with
182
+ red representing Sz = +1 (spin-up polaritons) and blue repre-
183
+ senting Sz = −1 (spin-down polaritons). Red and blue arrows
184
+ show the pseudospin precession of the polaritons propagating
185
+ from the edge condensates along the triad axis
186
+ tally(vertically), diagonally(antidiagonally), and right-
187
+ circularly(left-circularly)
188
+ polarized
189
+ (RCP
190
+ and
191
+ LCP
192
+ for short) PL, respectively.
193
+ Formally, the Stokes pa-
194
+ rameters relate to the condensate pseudospin through
195
+ S = ⟨Ψ†|ˆσ|Ψ⟩/⟨Ψ†|Ψ⟩ where Ψ = (ψ+, ψ−)T is the
196
+ condensate spinor order parameter and ˆσ is the Pauli
197
+
198
+ (b)
199
+ a
200
+ D
201
+ H
202
+ (c) NN > NNN
203
+ Microcavity
204
+ (d) NN < NNN
205
+ Microcavity3
206
+ matrix-vector. The Sx(r) and Sy(r) components repre-
207
+ sent the degree of linear and diagonal polarisation but
208
+ are not important in this study (also due to the pre-
209
+ dominant circular polarisation of the condensates used
210
+ here).
211
+ Experimental measurements were reproduced
212
+ using a generalised two-dimensional Gross-Pitaevskii
213
+ equation (2DGPE) (see section S2 in the Supplemental
214
+ Material [54]).
215
+ In Fig. 2 we present results for two polariton con-
216
+ densates separated by d ≈ ξ/2. Data for a single iso-
217
+ lated condensate gives a Sz period around ξ ≈ 90 µm
218
+ (see section S1 in the Supplemental Material [54]). Fig-
219
+ ures 2(a) and 2(b) show the measured and simulated spa-
220
+ tial distribution of the Sz component with spatial pseu-
221
+ dospin oscillations clearly visible due to the SOC rotat-
222
+ ing the spin of the outflowing polaritons. Note that un-
223
+ avoidable dephasing of polaritons in experiment results
224
+ in lowered Sz values compared to simulations as indi-
225
+ cated on the colorbars. Smaller ripple-like modulations
226
+ are also visible due to the standing wave interference be-
227
+ tween the two phase-locked condensates as reported be-
228
+ fore [24, 43, 53].
229
+ These ripples are characterized by a
230
+ small-scale period λ = 2π/⟨kc⟩ ≈ 3 µm, where ⟨kc⟩ is the
231
+ average outflow momentum of polaritons from their con-
232
+ densates. In contrast, the large-scale Sz period is given
233
+ by ξ = 2π/∆k ≫ λ where ℏ∆k/√2εc = |√mTE−√mTM|
234
+ and εc ≈ 3 meV is the condensate energy (measured from
235
+ k = 0 at the dispersion) and mTE,TM are the effective
236
+ masses of TE and TM polarized polaritons [44].
237
+ The spin screening effect can be observed as periodic
238
+ extrema in the integrated PL intensity, which represents
239
+ the condensate occupation, as a function of separation
240
+ distance d in Fig. 2(c). At the maxima the coupling is
241
+ unscreened and NN coupling is strong. At the minima
242
+ the coupling is screened and NN coupling is weak. Black
243
+ dots and black solid curve denote experimental measure-
244
+ ments and calculations, respectively. In the absence of
245
+ SOC one would observe monotonically decreasing emis-
246
+ sion intensity with only short variations (order of λ)
247
+ corresponding to in-phase and anti-phase flip-flop tran-
248
+ sitions between the synchronized condensates [24]. In-
249
+ stead, we observe strong non-monotonic behaviour with
250
+ clearly visible maxima around 67 and 154 µm, and min-
251
+ ima around 56 and 135 µm.
252
+ Notice that the distance
253
+ between the two maxima and the two minima correlates
254
+ with the measured ξ ≈ 90 µm period of Sz oscillations.
255
+ The discrepancy between the absolute locations of the
256
+ minima and maxima with the predicted critical distances
257
+ for screened (ξ/2, 3ξ/2) and unscreened (ξ, 2ξ) coupling,
258
+ respectively, can be understood as follows. Firstly, when
259
+ two condensates are coupled their energy is redshifted
260
+ on average [24] leading to smaller εc and thus larger ξ in
261
+ the coupled system. Second, the finite width of the pump
262
+ spots modulates the phase of polaritons and causes a shift
263
+ in the Sz period. Third, the cavity here has higher levels
264
+ of disorder than strain-compensated cavities [55] which
265
+ Figure 2.
266
+ Two polariton condensates. (a) Experimentally
267
+ measured and (b) simulated numerically real space Sz compo-
268
+ nent of the Stokes vector of the polariton condensates emis-
269
+ sion. In panel (a) black circles show the position of pump
270
+ spots. (c) Total integrated emission intensity dependence on
271
+ the separation distance between two condensates pump spots.
272
+ In panel (c) black dots shows the experimentally measured
273
+ values with red region representing the error of the total inten-
274
+ sity value. Black curve shows the same dependence calculated
275
+ numerically
276
+ can affect the spatial coupling. That’s why the relative
277
+ distances between the extrema are more meaningful than
278
+ their absolute locations. This interpretation is verified in
279
+ 2DGPE modeling which accurately reproduces the loca-
280
+ tions of the extrema. Note that the slight discrepancy
281
+ between modeling and experiment in Fig. 2(c) between
282
+ 70 and 120 µm can be attributed to the large parame-
283
+ ter space of the 2DGPE making quantitative matching
284
+ somewhat challenging.
285
+ In order to demonstrate the NNN coupling using the
286
+ all-optical spin screening effect we investigated the sys-
287
+ tem containing a chain of three condensates similar to the
288
+ system depicted schematically in Fig. 1. As in the pre-
289
+ vious experiment with two condensates, all condensates
290
+ were excited using tightly focused RCP laser pump spots
291
+ of equal intensity above threshold. Figures 3(a) and 3(b)
292
+ show the measured and simulated spatial distribution of
293
+ the three condensate Sz component with NN distance of
294
+ d ≈ ξ/2. As in the previous case of two condensates, the
295
+ system forms a joint macroscopic coherent state result-
296
+ ing in an oscillating Sz pattern elongated along the hor-
297
+ izontal axis with three RCP condensate circles of equal
298
+
299
+ X (um)
300
+ -100
301
+ 0
302
+ 100
303
+ -100
304
+ 0
305
+ 100
306
+ (b)
307
+ a
308
+ 100
309
+ (μm)
310
+ 0
311
+ -100
312
+ 0.6
313
+ -0.6
314
+ (c)
315
+ 1.0
316
+ 0.8
317
+ 0.6
318
+ 0.4
319
+ 40
320
+ 80
321
+ 120
322
+ 160
323
+ Pump spot separation distance (um)4
324
+ degree of polarisation in the centre. Amazingly, the in-
325
+ tensity of the central condensate was suppressed relative
326
+ to the outer ones, evidencing reduced NN coupling due
327
+ to the spin screening effect, see in Fig. 3(c) measured
328
+ (red diamonds) and simulated (black solid curve) inten-
329
+ sity distribution along the triad axis.
330
+ To unambiguously demonstrate the spin screening ef-
331
+ fect in the triad, we measured (dots) and simulated (solid
332
+ curve) the dependence of the central condensate intensity
333
+ as a function of NN separation distance with results pre-
334
+ sented in Fig. 3(d). Both experiment and calculations
335
+ show a clear dip around d = 52 µm ≈ ξ/2, corresponding
336
+ to spin-screened NN coupling, followed by a small peak
337
+ around d = 80 µm ≈ ξ where the NN coupling is re-
338
+ stored. The observed suppression of the central conden-
339
+ sate intensity provides strong evidence of spin-screened
340
+ NN coupling mediated by the spin coherence of the sys-
341
+ tem.
342
+ Moreover, the experimentally measured pump power
343
+ dependence for each separation distance and the ex-
344
+ tracted polariton condensation threshold values are
345
+ shown in Fig. 3(e) (red circles). The horizontal dashed
346
+ line is the threshold value of the isolated condensate. In
347
+ the absence of the TE-TM splitting monotonic increase of
348
+ the threshold value converging to the isolated condensate
349
+ threshold is expected with the increase of the separation
350
+ distance between the condensates. In our system we ob-
351
+ serve maximum threshold at the separation distance of
352
+ 52 µm, which precisely corresponded to the minimum of
353
+ the central condensate intensity in Figs. 3(c) and 3(d). It
354
+ confirms that the NN condensate interaction is effectively
355
+ screened at this separation distance due to the TE-TM
356
+ splitting. Around a separation distance close to the full
357
+ period of Sz oscillation (d ≈ ξ) a decrease in the thresh-
358
+ old power was observed, as expected with NN coupling
359
+ restored. A simple linear coupled oscillator model [solid
360
+ curve in Fig. 3(e)] is able to explain the behaviour of
361
+ the threshold power (see section S3 in the Supplemental
362
+ Material [54]).
363
+ In summary, we have experimentally demonstrated
364
+ that next-nearest-neighbours coupling can be made
365
+ stronger than nearest-neighbour coupling in ballistically
366
+ expanding spinor exciton-polariton condensates which
367
+ was recently proposed in Ref. [53]. This unconventional
368
+ near-inversion of the spatial coupling hierarchy between
369
+ condensates stems from the combination of TE-TM split-
370
+ ting and the ballistic polariton flow from each conden-
371
+ sate.
372
+ Outflowing polaritons experience effective spin-
373
+ orbit coupling which rotates their spin state as they prop-
374
+ agate from one condensate to the next. Depending on
375
+ distance, the overlap (coupling) between the condensates
376
+ can become spin-screened depending on the polariton
377
+ spin projection upon arrival at its neighbour. We believe
378
+ that the demonstrated alteration of the conventional con-
379
+ densate coupling hierarchy could pave the way towards
380
+ all-optical simulation of many-body ballistic systems be-
381
+ Figure 3.
382
+ Three polariton condensates. (a) Experimentally
383
+ measured and (b) simulated real space Sz component of the
384
+ Stokes vector of the polariton condensates emission. In panel
385
+ (a) black circles show the position of pump spots. (c) Mea-
386
+ sured experimentally (red diamonds) and calculated numeri-
387
+ cally (solid black curve) real space intensity distribution along
388
+ the triad axis. (d) Dependence of the central condensate PL
389
+ intensity on the separation distance between the condensates
390
+ pump spots measured experimentally (black dots) and calcu-
391
+ lated numerically (solid black curve); red region represents the
392
+ error of the total intensity value. The dashed curves are guides
393
+ to the eye. (e) The system threshold power dependence on
394
+ the separation distance between the condensates pump spots
395
+ measured experimentally (red circles) and calculated numer-
396
+ ically (solid black curve); red bars represent the error. Grey
397
+ dashed line in panel (e) shows the threshold power for single
398
+ isolated condensate.
399
+ longing to non-planar graph topologies using networks of
400
+ spinor polariton condensates.
401
+ The authors acknowledge the support of the European
402
+ Union’s Horizon 2020 program, through a FET Open re-
403
+ search and innovation action under the grant agreements
404
+ No.
405
+ 899141 (PoLLoC) and no.
406
+ 964770 (TopoLight).
407
+ H.S. acknowledges the Icelandic Research Fund (Rannis),
408
+ Grant No. 217631-051.
409
+
410
+ X (μm)
411
+ -100
412
+ 0
413
+ 100
414
+ -100
415
+ 0
416
+ 100
417
+ (a)
418
+ 100
419
+ 100
420
+ Y
421
+ (um)
422
+ (μm)
423
+ 0
424
+ 0
425
+ Y
426
+ -100
427
+ -100
428
+ -0.7
429
+ 0.7
430
+ SZ
431
+ (c)
432
+ 1.0
433
+ Int
434
+ 1.0
435
+ tegrated
436
+ 0.8
437
+ (a.u.)
438
+ 0.8
439
+ intensity
440
+ 0.6
441
+ Intensity
442
+ 0.6
443
+ 0.4
444
+ 0.4 @
445
+ 0.2
446
+ .u.
447
+ 20
448
+ 40
449
+ 60
450
+ 80
451
+ ¥100
452
+ -100
453
+ 0
454
+ 100
455
+ (un) X
456
+ Pump spot separation (μm)
457
+ (e)
458
+ .0
459
+ 0.9
460
+ 0.8
461
+ 0
462
+ 20
463
+ 40
464
+ 60
465
+ 80
466
+ 100
467
+ Pump spot separation (um)5
468
+ ∗ DovzhenkoDS@gmail.com
469
+ [1] F. Alet, A. M. Walczak, and M. P. Fisher, Exotic quan-
470
+ tum phases and phase transitions in correlated matter,
471
+ Phys. A Stat. Mech. its Appl. 369, 122 (2006).
472
+ [2] M. Greiner, O. Mandel, T. Esslinger, T. W. Hänsch, and
473
+ I. Bloch, Quantum phase transition from a superfluid to
474
+ a Mott insulator in a gas of ultracold atoms, Nature 415,
475
+ 39 (2002).
476
+ [3] L. Fallani, J. E. Lye, V. Guarrera, C. Fort, and M. In-
477
+ guscio, Ultracold Atoms in a Disordered Crystal of Light:
478
+ Towards a Bose Glass, Phys. Rev. Lett. 98, 130404
479
+ (2007).
480
+ [4] I. Bloch, J. Dalibard, and S. Nascimbène, Quantum sim-
481
+ ulations with ultracold quantum gases, Nat. Phys. 8, 267
482
+ (2012).
483
+ [5] I. M. Georgescu, S. Ashhab, and F. Nori, Quantum sim-
484
+ ulation, Rev. Mod. Phys. 86, 153 (2014).
485
+ [6] D. Hanneke, J. P. Home, J. D. Jost, J. M. Amini,
486
+ D. Leibfried, and D. J. Wineland, Realization of a pro-
487
+ grammable two-qubit quantum processor, Nat. Phys. 6,
488
+ 13 (2010).
489
+ [7] M. Johanning, A. F. Varón, and C. Wunderlich, Quan-
490
+ tum simulations with cold trapped ions, J. Phys. B At.
491
+ Mol. Opt. Phys. 42, 154009 (2009).
492
+ [8] X. Peng, J. Zhang, J. Du, and D. Suter, Quantum Simu-
493
+ lation of a System with Competing Two- and Three-Body
494
+ Interactions, Phys. Rev. Lett. 103, 140501 (2009).
495
+ [9] R. Hanson and D. D. Awschalom, Coherent manipula-
496
+ tion of single spins in semiconductors, Nature 453, 1043
497
+ (2008).
498
+ [10] J. Q. You and F. Nori, Atomic physics and quantum
499
+ optics using superconducting circuits, Nature 474, 589
500
+ (2011).
501
+ [11] M. P. Harrigan, K. J. Sung, M. Neeley, K. J. Satzinger,
502
+ F. Arute, K. Arya, J. Atalaya, J. C. Bardin, R. Barends,
503
+ S. Boixo, et al., Quantum approximate optimization of
504
+ non-planar graph problems on a planar superconducting
505
+ processor, Nat. Phys. 17, 332 (2021).
506
+ [12] D. E. Chang, V. Vuletić, and M. D. Lukin, Quantum
507
+ nonlinear optics — photon by photon, Nat. Photonics 8,
508
+ 685 (2014).
509
+ [13] A. Periwal, E. S. Cooper, P. Kunkel, J. F. Wienand,
510
+ E. J. Davis, and M. Schleier-Smith, Programmable in-
511
+ teractions and emergent geometry in an array of atom
512
+ clouds, Nature 600, 630 (2021).
513
+ [14] S. Alyatkin, J. D. Töpfer, A. Askitopoulos, H. Sigurds-
514
+ son, and P. G. Lagoudakis, Optical Control of Couplings
515
+ in Polariton Condensate Lattices, Phys. Rev. Lett. 124,
516
+ 207402 (2020).
517
+ [15] S. Alyatkin, H. Sigurdsson, A. Askitopoulos, J. D.
518
+ Töpfer, and P. G. Lagoudakis, Quantum fluids of light
519
+ in all-optical scatterer lattices, Nat. Commun. 12, 5571
520
+ (2021).
521
+ [16] A. Kavokin,
522
+ T. C. H. Liew,
523
+ C. Schneider,
524
+ P. G.
525
+ Lagoudakis, S. Klembt, and S. Hoefling, Polariton con-
526
+ densates for classical and quantum computing, Nat. Rev.
527
+ Phys. 4, 435 (2022).
528
+ [17] Y. Zhang, X. Zhang, B. Tang, C. Tian, C. Xu, H. Dong,
529
+ and W. Zhou, Realization of an all-optically controlled
530
+ dynamic superlattice for exciton–polaritons, Nanoscale
531
+ 10, 14082 (2018).
532
+ [18] L. Pickup, H. Sigurdsson, J. Ruostekoski, and P. G.
533
+ Lagoudakis, Synthetic band-structure engineering in po-
534
+ lariton crystals with non-Hermitian topological phases,
535
+ Nat. Commun. 11, 4431 (2020).
536
+ [19] M. Pieczarka, M. Pieczarka, M. Pieczarka, E. Estrecho,
537
+ S. Ghosh, M. Wurdack, M. Steger, M. Steger, D. W.
538
+ Snoke, K. West, L. N. Pfeiffer, T. C. H. Liew, A. G. Tr-
539
+ uscott, E. A. Ostrovskaya, and E. A. Ostrovskaya, Topo-
540
+ logical phase transition in an all-optical exciton-polariton
541
+ lattice, Optica 8, 1084 (2021).
542
+ [20] J. D. Töpfer, I. Chatzopoulos, H. Sigurdsson, T. Cook-
543
+ son, Y. G. Rubo, and P. G. Lagoudakis, Engineering spa-
544
+ tial coherence in lattices of polariton condensates, Optica
545
+ 8, 106 (2021).
546
+ [21] N. G. Berloff, M. Silva, K. Kalinin, A. Askitopoulos, J. D.
547
+ Töpfer, P. Cilibrizzi, W. Langbein, and P. G. Lagoudakis,
548
+ Realizing the classical XY Hamiltonian in polariton sim-
549
+ ulators, Nat. Mater. 16, 1120 (2017).
550
+ [22] J. Kasprzak, M. Richard, S. Kundermann, A. Baas,
551
+ P. Jeambrun, J. M. J. Keeling, F. M. Marchetti, M. H.
552
+ Szymańska, R. André, J. L. Staehli, V. Savona, P. B. Lit-
553
+ tlewood, B. Deveaud, and L. S. Dang, Bose–Einstein con-
554
+ densation of exciton polaritons, Nature 443, 409 (2006).
555
+ [23] H. Ohadi, R. L. Gregory, T. Freegarde, Y. G. Rubo, A. V.
556
+ Kavokin, N. G. Berloff, and P. G. Lagoudakis, Nontrivial
557
+ Phase Coupling in Polariton Multiplets, Phys. Rev. X 6,
558
+ 031032 (2016).
559
+ [24] J. D. Töpfer, H. Sigurdsson, L. Pickup, and P. G.
560
+ Lagoudakis, Time-delay polaritonics, Commun. Phys. 3,
561
+ 2 (2020).
562
+ [25] A. S. Abdalla, B. Zou, and Y. Zhang, Optical joseph-
563
+ son oscillation achieved by two coupled exciton-polariton
564
+ condensates, Opt. Express 28, 9136 (2020).
565
+ [26] C. Domb and R. Potts, Order-disorder statistics IV. A
566
+ two-dimensional model with first and second interactions,
567
+ Proc. R. Soc. Lond. A: Math. Phys. Sci. 210, 125 (1951).
568
+ [27] W. Selke, The ANNNI model — Theoretical analysis and
569
+ experimental application, Phys. Rep. 170, 213 (1988).
570
+ [28] M. Wolf and K. D. Schotte, Ising model with compet-
571
+ ing next-nearest-neighbour interactions on the Kagome
572
+ lattice, J. Phys. A. Math. Gen. 21, 2195 (1988).
573
+ [29] A. J. Ramírez-Pastor, F. Nieto, and E. E. Vogel, Ising lat-
574
+ tices with ±J second-nearest-neighbor interactions, Phys.
575
+ Rev. B 55, 14323 (1997).
576
+ [30] K. P. Kalinin, P. G. Lagoudakis, and N. G. Berloff, Exotic
577
+ states of matter with polariton chains, Phys. Rev. B 97,
578
+ 161101 (2018).
579
+ [31] K. P. Kalinin and N. G. Berloff, Computational complex-
580
+ ity continuum within Ising formulation of NP problems,
581
+ Commun. Phys. 5, 20 (2022).
582
+ [32] C. Schneider, K. Winkler, M. D. Fraser, M. Kamp, Y. Ya-
583
+ mamoto, E. A. Ostrovskaya, and S. Höfling, Exciton-
584
+ polariton trapping and potential landscape engineering,
585
+ Rep. Prog. Phys. 80, 016503 (2016).
586
+ [33] H. Deng, H. Haug, and Y. Yamamoto, Exciton-polariton
587
+ bose-einstein condensation, Rev. Mod. Phys. 82, 1489
588
+ (2010).
589
+ [34] H. Yang and N. Y. Kim, Microcavity exciton-polariton
590
+ quantum spin fluids, Adv. Quantum Technol. 5, 2100137
591
+ (2022).
592
+ [35] D. Read, T. C. H. Liew, Y. G. Rubo, and A. V. Kavokin,
593
+ Stochastic polarization formation in exciton-polariton
594
+ Bose-Einstein condensates, Phys. Rev. B 80, 195309
595
+ (2009).
596
+
597
+ 6
598
+ [36] Y. del Valle-Inclan Redondo, H. Sigurdsson, H. Ohadi,
599
+ I. A. Shelykh, Y. G. Rubo, Z. Hatzopoulos, P. G. Sav-
600
+ vidis, and J. J. Baumberg, Observation of inversion, hys-
601
+ teresis, and collapse of spin in optically trapped polariton
602
+ condensates, Phys. Rev. B 99, 165311 (2019).
603
+ [37] F. Meier and B. Zakharchenya, Optical orientation
604
+ (North Holland, 1984) p. 523.
605
+ [38] S. Pfalz, R. Winkler, T. Nowitzki, D. Reuter, A. D.
606
+ Wieck, D. Hägele, and M. Oestreich, Optical orientation
607
+ of electron spins in gaas quantum wells, Phys. Rev. B 71,
608
+ 165305 (2005).
609
+ [39] P. Renucci, T. Amand, X. Marie, P. Senellart, J. Bloch,
610
+ B. Sermage, and K. V. Kavokin, Microcavity polariton
611
+ spin quantum beats without a magnetic field: A man-
612
+ ifestation of coulomb exchange in dense and polarized
613
+ polariton systems, Phys. Rev. B 72, 075317 (2005).
614
+ [40] A. Askitopoulos, A. V. Nalitov, E. S. Sedov, L. Pickup,
615
+ E. D. Cherotchenko, Z. Hatzopoulos, P. G. Savvidis,
616
+ A. V. Kavokin, and P. G. Lagoudakis, All-optical quan-
617
+ tum fluid spin beam splitter, Phys. Rev. B 97, 235303
618
+ (2018).
619
+ [41] I. Gnusov, H. Sigurdsson, S. Baryshev, T. Ermatov,
620
+ A. Askitopoulos, and P. G. Lagoudakis, Optical orienta-
621
+ tion, polarization pinning, and depolarization dynamics
622
+ in optically confined polariton condensates, Phys. Rev. B
623
+ 102, 125419 (2020).
624
+ [42] I. Gnusov, H. Sigurdsson, J. Töpfer, S. Baryshev, S. Aly-
625
+ atkin, and P. Lagoudakis, All-Optical Linear-Polarization
626
+ Engineering in Single and Coupled Exciton-Polariton
627
+ Condensates, Phys. Rev. Appl. 16, 034014 (2021).
628
+ [43] L. Pickup, J. D. Töpfer, H. Sigurdsson, and P. G.
629
+ Lagoudakis, Polariton spin jets through optical control,
630
+ Phys. Rev. B 103, 155302 (2021).
631
+ [44] G. Panzarini, L. C. Andreani, A. Armitage, D. Baxter,
632
+ M. S. Skolnick, V. N. Astratov, J. S. Roberts, A. V. Ka-
633
+ vokin, M. R. Vladimirova, and M. A. Kaliteevski, Cavity-
634
+ polariton dispersion and polarization splitting in single
635
+ and coupled semiconductor microcavities, Phys. Solid
636
+ State 41, 1223 (1999).
637
+ [45] E. Kammann, T. C. H. Liew, H. Ohadi, P. Cilibrizzi,
638
+ P. Tsotsis, Z. Hatzopoulos, P. G. Savvidis, A. V. Kavokin,
639
+ and P. G. Lagoudakis, Nonlinear Optical Spin Hall Ef-
640
+ fect and Long-Range Spin Transport in Polariton Lasers,
641
+ Phys. Rev. Lett. 109, 036404 (2012).
642
+ [46] C. Antón, S. Morina, T. Gao, P. S. Eldridge, T. C. H.
643
+ Liew, M. D. Martín, Z. Hatzopoulos, P. G. Savvidis, I. A.
644
+ Shelykh, and L. Viña, Optical control of spin textures in
645
+ quasi-one-dimensional polariton condensates, Phys. Rev.
646
+ B 91, 075305 (2015).
647
+ [47] A. Kavokin, G. Malpuech, and M. Glazov, Optical Spin
648
+ Hall Effect, Phys. Rev. Lett. 95, 136601 (2005).
649
+ [48] C. Leyder, M. Romanelli, J. P. Karr, E. Giacobino,
650
+ T.
651
+ C.
652
+ H.
653
+ Liew,
654
+ M.
655
+ M.
656
+ Glazov,
657
+ A.
658
+ V.
659
+ Kavokin,
660
+ G. Malpuech, and A. Bramati, Observation of the op-
661
+ tical spin hall effect, Nat. Phys. 3, 628 (2007).
662
+ [49] M. Maragkou, C. E. Richards, T. Ostatnický, A. J. D.
663
+ Grundy, J. Zajac, M. Hugues, W. Langbein, and P. G.
664
+ Lagoudakis, Optical analogue of the spin Hall effect in a
665
+ photonic cavity, Opt. Lett. 36, 1095 (2011).
666
+ [50] K. Y. Bliokh, F. J. Rodríguez-Fortuño, F. Nori, and A. V.
667
+ Zayats, Spin–orbit interactions of light, Nat. Photonics 9,
668
+ 796 (2015).
669
+ [51] K. Lekenta, M. Król, R. Mirek, K. Łempicka, D. Stephan,
670
+ R. Mazur, P. Morawiak, P. Kula, W. Piecek, P. G.
671
+ Lagoudakis, B. Piętka, and J. Szczytko, Tunable opti-
672
+ cal spin hall effect in a liquid crystal microcavity, Light
673
+ Sci. Appl. 7, 74 (2018).
674
+ [52] K. Łempicka-Mirek, M. Król, H. Sigurdsson, A. Win-
675
+ cukiewicz,
676
+ P. Morawiak,
677
+ R. Mazur,
678
+ M. Muszyński,
679
+ W. Piecek, P. Kula, T. Stefaniuk, M. Kamińska, L. De
680
+ Marco, P. G. Lagoudakis, D. Ballarini, D. Sanvitto,
681
+ J. Szczytko, and B. Pi¸etka, Electrically tunable Berry
682
+ curvature and strong light-matter coupling in liquid crys-
683
+ tal microcavities with 2D perovskite, Sci. Adv. 8, 1
684
+ (2022).
685
+ [53] D. Aristov, H. Sigurdsson, and P. G. Lagoudakis, Screen-
686
+ ing nearest-neighbor interactions in networks of exciton-
687
+ polariton condensates through spin-orbit coupling, Phys.
688
+ Rev. B 105, 155306 (2022).
689
+ [54] See Supplemental Material for additional information.
690
+ [55] P. Cilibrizzi,
691
+ A. Askitopoulos,
692
+ M. Silva,
693
+ F. Basti-
694
+ man,
695
+ E. Clarke,
696
+ J. M. Zajac,
697
+ W. Langbein, and
698
+ P. G. Lagoudakis, Polariton condensation in a strain-
699
+ compensated planar microcavity with ingaas quantum
700
+ wells, Appl. Phys. Lett. 105, 191118 (2014).
701
+
702
+ Supplemental Material: Next nearest neighbour coupling with spinor polariton
703
+ condensates
704
+ Dmitriy Dovzhenko,1, ∗ Denis Aristov,1 Lucy Pickup,1 Helgi Sigurdsson,1, 2 and Pavlos Lagoudakis1, 3
705
+ 1School of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, UK
706
+ 2Science Institute, University of Iceland, Dunhagi 3, IS-107, Reykjavik, Iceland
707
+ 3Hybrid Photonics Laboratory, Skolkovo Institute of Science and Technology,
708
+ Territory of Innovation Center Skolkovo, Bolshoy Boulevard 30, building 1, 121205 Moscow, Russia
709
+ (Dated: January 12, 2023)
710
+ S1.
711
+ EXPERIMENTAL DETAILS, TE-TM SPLITTING AND SINGLE ISOLATED CONDENSATE REAL
712
+ SPACE Sz COMPONENT OF THE STOKES VECTOR
713
+ In this supplemental section we present experimental details, experimentally measured TE-TM splitting and real
714
+ space distribution of Sz component of Stokes vector of the emission from a single isolated polariton condensate.
715
+ All measurements were performed at 6 K using a continuous flow cold finger cryostat. We used a right circularly
716
+ (σ+) polarized non-resonant continuous wave laser excitation tuned to the first Bragg minimum of the microcavity
717
+ reflection spectra at 754 nm. To reduce sample heating we used an acousto-optic modulator driven by rectangular
718
+ voltage pulse train at 10 kHz repetition rate with 5% duty cycle. A spatial light modulator was used to structure the
719
+ pump spatial profile into one, two, or three Gaussian spots focused on the sample using a microscope objective lens
720
+ with 0.4 numerical aperture. In order to obtain simultaneously real space, k-space, and spectrally resolved k-space
721
+ images of the time-averaged PL two separate CCD cameras and a 0.75 m monochromator with 1200g/mm diffraction
722
+ grating equipped with the CCD camera were used. A quarter wave plate and a Wollaston prism were introduced
723
+ in the optical path for real space imaging to simultaneously measure the right-circularly polarized and left-circularly
724
+ polarized components of the PL with the same CCD camera.
725
+ In Fig. S1(a) we show the Sz spatial oscillations due to the spin-orbit coupling (SOC) rotating the pseudospin of
726
+ the polaritons propagating away from the condensate excited using single Gaussian spot. The oscillation period ξ was
727
+ measured to be around 90 µm with the oscillations amplitude of ±0.6.
728
+ In order to experimentally estimate the value of TE-TM splitting we measured dispersion of the lower polariton
729
+ branch in a linear regime (i.e., below condensation threshold) along the in-plane k∥ momentum axis [see Fig. S1(b)].
730
+ Splitting of the dispersion is clearly observed at the higher values of in-plane k vector with the higher energy branch
731
+ corresponding to the emission from the vertically polarized polaritons. The energy splitting between horizontally
732
+ and vertically polarized polaritons possesses parabolic dependence on the in-plane momentum and ≈ 0.2 meV at
733
+ k = 3 µm−1 in-plane wavevector.
734
+ S2.
735
+ TWO DIMENSIONAL SPINOR POLARITON MODEL
736
+ The experimental observations are reproduced by numerically solving a generalized Gross-Pitaevskii equation (S1)
737
+ for macroscopic spinor polariton wavefunction Ψ(r, t) = (ψ+, ψ−)T coupled to an active exciton reservoir with density
738
+ nA(r, t) = (nA+, nA−)T rate equation [1],
739
+ i∂ψ±
740
+ ∂t
741
+ =
742
+
743
+ − ℏ∇2
744
+ 2m + i
745
+ 2
746
+
747
+ RnA± − γ
748
+
749
+ + α1|ψ±|2 + α2|ψ∓|2 + U±(r) + V (r)
750
+
751
+ ψ± + ∆LT
752
+ � ∂
753
+ ∂x
754
+ ∓ i ∂
755
+ ∂y
756
+ �2
757
+ ψ∓,
758
+ (S1)
759
+ U± = G1
760
+
761
+ nA± + nI±
762
+
763
+ + G2
764
+
765
+ nA∓ + nI∓
766
+
767
+ ,
768
+ (S2)
769
+ ∂nA±
770
+ ∂t
771
+ = −
772
+
773
+ ΓA + Γs + R|ψ±|2�
774
+ nA± + WnI± + ΓsnA∓.
775
+ (S3)
776
+ Here, ± represents the spin of polaritons and excitons along the cavity growth axis, m is the polariton effective
777
+ mass in parabolic dispersion approximation, γ is the polariton decay rate, G1 = 2g|χ|2 and α1 = g|χ|4 are the same
778
+ spin polariton-reservoir and polariton-polariton interaction strengths, respectively, g is the exciton-exciton Coulomb
779
+ ∗ DovzhenkoDS@gmail.com
780
+ arXiv:2301.04210v1 [cond-mat.mes-hall] 10 Jan 2023
781
+
782
+ 2
783
+ Figure S1.
784
+ (a) Experimentally measured real space Sz component of the Stokes vector of the single isolated polariton
785
+ condensate emission. (b) Energy and in-plane wavevector resolved normalized PL intensity from the lower polariton branch
786
+ below the threshold
787
+ interaction strength, |χ|2 is the excitonic Hopfield fraction of the polariton, and ∆LT represents the strength of the
788
+ TE-TM splitting. Opposite spin interactions, usually much weaker, were chosen to be G2 = −0.2G1 and α2 = −0.2α1
789
+ for completeness but we note that our results to not qualitatively depend on these terms. R is the scattering rate of
790
+ reservoir excitons into the condensate, ΓA is the active reservoir decay rate, and Γs represents exciton spin relaxation
791
+ rate [2].
792
+ A so-called inactive reservoir of excitons nI,± also contributes to the blueshift of polaritons as depicted in Eq. (S2).
793
+ This reservoir corresponds to high-momentum excitons which do not scatter into the condensate but instead drive the
794
+ active low-momentum excitons (S3). In continuous wave experiments the inactive reservoir density can be written
795
+ WnI,+ =
796
+ P0(r)
797
+ W + 2Γs
798
+ (W cos2 (θ) + Γs),
799
+ WnI,− =
800
+ P0(r)
801
+ W + 2Γs
802
+ (W sin2 (θ) + Γs),
803
+ (S4)
804
+ where P0 is the total power density of the incident coherent light with degree of circular polarization expressed as S3 =
805
+ P0[cos2 (θ)−sin2 (θ)] = P0 cos (2θ). Since our experiment is performed with fully right hand circularly polarized light,
806
+ we set θ = 0 from here on. The phenomenological parameter W quantifies conversion rate between same-spin inactive
807
+ and active exciton reservoirs. The pump profile is written as a superposition of Gaussians P0(r) = p0
808
+
809
+ n e−|r−rn|2/2w2.
810
+ To represent tight focusing of excitation beams we used Gaussians with 2 µm full-width-at-half-maximum.
811
+ Lastly, given the disorder present at the large spatial scales of the experiment we include a random potential
812
+ landscape in our simulation given by V (r) generated as a random Gaussian-correlated potential [3]. The simulation
813
+ parameters are based on previous GaAs microcavity experiments [4, 5]: m = 5×10−5 of free electron mass; γ−1 = 5.5
814
+ ps; |χ|2 = 0.4; ℏg = 0.5 µeV µm2; R = 3.2g; W = ΓA = γ; Γs = γ/4; ∆LT = 0.036 ps−1 µm2. The disorder potential
815
+ was generated with 1.5 µm correlation length and 0.06 meV root mean squared amplitude.
816
+ We note that in order to compensate for additional background noise in experiment (i.e., additional light coming
817
+ from spontaneous emission of bottleneck excitons) we applied a global shift to the integrated densities of the condensate
818
+ |ψ±|2 by approximately 10 percent in order to match the experimental values in Figure 3(d) in the main text. This
819
+ difference between modeling and experiment is more evident in Figure 3(c) where the experimentally measured
820
+ photoluminescence (PL) intensity is more spread out than simulated condensate densities. This can also come from
821
+ the finite diffusion of excitons which we have neglected here for simplicity.
822
+ Nevertheless, the calculated relative
823
+
824
+ X (μm)
825
+ k, (μm-l)
826
+ -200
827
+ 200
828
+ 0
829
+ -3-2
830
+ -1
831
+ 0
832
+ 1
833
+ 2
834
+ [(b)
835
+ (a)
836
+ 1.543
837
+ 200
838
+ 1.542
839
+ (un)
840
+ 0
841
+ (eV)
842
+ 1.540
843
+ 1.539
844
+ -200
845
+ 1.538
846
+ 0.6
847
+ Sz
848
+ -0.6
849
+ PL intensity
850
+ 03
851
+ amplitude of the PL at the pump positions follows the experimental results quite precisely, which, therefore, justifies
852
+ the use of current model and provides a clear quantitative evidence of the spin-screening happening in the system.
853
+ S3.
854
+ THEORY OF THE THRESHOLD BEHAVIOUR IN A SPIN SCREENED CONDENSATE TRIAD
855
+ The behaviour of the pump threshold from experiment in the triad configuration can be reproduced by scrutinizing
856
+ the eigenenergies of an appropriate linear operator which couples the three condensates together. In other words, we
857
+ neglect polariton nonlinearities so close to the threshold. The threshold is reached when a single eigenvalue belonging
858
+ to the three coupled condensates crosses from the lower- to the upper-half of the complex plane.
859
+ We will start by defining the state vector of the system,
860
+ |Ψ⟩ = (ψ1,+, ψ1,−, ψ2,+, ψ2,−, ψ3,+, ψ3,−)T.
861
+ (S5)
862
+ Here, the index n ∈ {1, 2, 3} denotes the left, middle, and right condensate, respectively. The spectrum of the coupled
863
+ system in the linear regime (i.e., close to threshold |ψn,±|2 ≃ 0) can be described with the following non-Hermitian
864
+ operator separated into three parts for clarity,
865
+ ˆH =
866
+
867
+
868
+
869
+
870
+
871
+
872
+
873
+ ω+
874
+ 0
875
+ 0
876
+ 0
877
+ 0
878
+ 0
879
+ 0
880
+ ω−
881
+ 0
882
+ 0
883
+ 0
884
+ 0
885
+ 0
886
+ 0
887
+ ω+
888
+ 0
889
+ 0
890
+ 0
891
+ 0
892
+ 0
893
+ 0
894
+ ω−
895
+ 0
896
+ 0
897
+ 0
898
+ 0
899
+ 0
900
+ 0
901
+ ω+
902
+ 0
903
+ 0
904
+ 0
905
+ 0
906
+ 0
907
+ 0
908
+ ω−
909
+
910
+
911
+
912
+
913
+
914
+
915
+
916
+ +
917
+
918
+
919
+
920
+
921
+
922
+
923
+
924
+ 0
925
+ 0
926
+ J+ δJ
927
+ 0
928
+ 0
929
+ 0
930
+ 0
931
+ δJ J−
932
+ 0
933
+ 0
934
+ J+ δJ
935
+ 0
936
+ 0
937
+ J+ δJ
938
+ δJ J−
939
+ 0
940
+ 0
941
+ δJ J−
942
+ 0
943
+ 0
944
+ J+ δJ
945
+ 0
946
+ 0
947
+ 0
948
+ 0
949
+ δJ J−
950
+ 0
951
+ 0
952
+
953
+
954
+
955
+
956
+
957
+
958
+
959
+ +
960
+
961
+
962
+
963
+
964
+
965
+
966
+
967
+ 0
968
+ 0
969
+ 0 0 K+ δK
970
+ 0
971
+ 0
972
+ 0 0 δK K−
973
+ 0
974
+ 0
975
+ 0 0
976
+ 0
977
+ 0
978
+ 0
979
+ 0
980
+ 0 0
981
+ 0
982
+ 0
983
+ K+ δK 0 0
984
+ 0
985
+ 0
986
+ δK K− 0 0
987
+ 0
988
+ 0
989
+
990
+
991
+
992
+
993
+
994
+
995
+
996
+ (S6)
997
+ The first matrix describes the complex self-energy of each oscillator (condensate) composed of the local pump blueshift
998
+ (G) and gain (R), and cavity losses (γ). This contribution from the pump can be parametrized in terms of the reservoir
999
+ spin populations,
1000
+ ω± =
1001
+
1002
+ G1 + iR
1003
+ 2
1004
+
1005
+ (NA,± + NI,±) − iγ
1006
+ 2 .
1007
+ (S7)
1008
+ where
1009
+
1010
+ nA(I),± dr = NA(I),± [i.e., spatially integrating (S3) and (S4)]. Here, we will neglect opposite spin interaction
1011
+ G2 for simplicity.
1012
+ Each condensate is coupled ballistically with its nearest neighbours with coupling strength J± and next-nearest
1013
+ neighbours with strength K± determined by the overlap between different condensates over their respective pump
1014
+ spots. Approximating the tightly focused pump spots as delta functions, we can write the coupling between the
1015
+ ballistic condensates as [4],
1016
+ J± = cos2 (ξd + Φ)
1017
+
1018
+ G1 + iR
1019
+ 2
1020
+
1021
+ (NA,± + NI,±)H(1)
1022
+ 0 (kd + φ),
1023
+ (S8)
1024
+ K± = sin2 (2ξd + Φ)
1025
+
1026
+ G1 + iR
1027
+ 2
1028
+
1029
+ (NA,± + NI,±)H(1)
1030
+ 0 (2kd + φ)
1031
+ (S9)
1032
+ The square cosine (sine) modulations in the coupling stem from a pseudospin screening effect coming from the strong
1033
+ influence of TE-TM splitting on the ballistic condensates [6] as explained in the main manuscript. Here, ξ denotes
1034
+ the period of the pseudospin precession for a single condensate in experiment. H(1)
1035
+ 0 (kd) is the zeroth order Hankel
1036
+ function of the first kind. The coupling depends on the product kd where d is the separation distance between two
1037
+ pump spots and k is the complex wavevector of the polaritons with mass m propagating outside the pump spot,
1038
+ k ≈ kc + i Γm
1039
+ 2ℏkc
1040
+ .
1041
+ (S10)
1042
+ Here, kc is the average real wavevector of the outflowing polaritons. The finite size of the Gaussian pump spots intro-
1043
+ duces some lag into the pseudospin precession because outflowing polaritons need to gradually build up momentum
1044
+ as they leave the pump spot. This is captured in the fitting parameter Φ. For the same reason, an overall phase-lag
1045
+ fitting parameter φ is also needed in the coupling term between the condensates.
1046
+
1047
+ 4
1048
+ The presence of TE-TM splitting also introduces coupling between opposite spin components denoted δJ and δK
1049
+ written in a similar fashion,
1050
+ δJ = δ cos2 (ξd + Φ)
1051
+
1052
+ G1 + iR
1053
+ 2
1054
+
1055
+ (NA,+ + NI,+ + NA,− + NI,−)H(1)
1056
+ 0 (kd + φ),
1057
+ (S11)
1058
+ δK = δ sin2 (2ξd + Φ)
1059
+
1060
+ G1 + iR
1061
+ 2
1062
+
1063
+ (NA,+ + NI,+ + NA,− + NI,−)H(1)
1064
+ 0 (2kd + φ)
1065
+ (S12)
1066
+ Here, δ < 1 is a fitting parameter describing the amount of opposite spin coupling. Diagonalizing ˆH for increasing
1067
+ pump power P0 we identify the threshold as the point in which a single eigenenergy crosses from the lower-half to
1068
+ the upper-half of the complex plane. The results are plotted in Fig. 3(e) in the main text (solid curve) alongside the
1069
+ experimental data, normalized in units of threshold power for the single isolated condensates Pthr,iso.
1070
+ [1] H. Deng, H. Haug, and Y. Yamamoto, Exciton-polariton bose-einstein condensation, Rev. Mod. Phys. 82, 1489 (2010).
1071
+ [2] M. Z. Maialle, E. A. de Andrada e Silva, and L. J. Sham, Exciton spin dynamics in quantum wells, Phys. Rev. B 47, 15776
1072
+ (1993).
1073
+ [3] V. Savona, Effect of interface disorder on quantum well excitons and microcavity polaritons, J. Phys. Condens. Matter 19,
1074
+ 295208 (2007).
1075
+ [4] J. D. T¨opfer, H. Sigurdsson, L. Pickup, and P. G. Lagoudakis, Time-delay polaritonics, Commun. Phys. 3, 2 (2020).
1076
+ [5] L. Pickup, J. D. T¨opfer, H. Sigurdsson, and P. G. Lagoudakis, Polariton spin jets through optical control, Phys. Rev. B
1077
+ 103, 155302 (2021).
1078
+ [6] D. Aristov, H. Sigurdsson, and P. G. Lagoudakis, Screening nearest-neighbor interactions in networks of exciton-polariton
1079
+ condensates through spin-orbit coupling, Phys. Rev. B 105, 155306 (2022).
1080
+
7NE2T4oBgHgl3EQf7giQ/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
7tFAT4oBgHgl3EQfoR3-/content/tmp_files/2301.08634v1.pdf.txt ADDED
@@ -0,0 +1,2775 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ICARUS at the Fermilab Short-Baseline Neutrino Program
2
+ - Initial Operation
3
+ P. Abratenko𝑎 A. Aduszkiewicz𝑏 F. Akbar𝑐 M. Artero Pons𝑑 J. Asaadi𝑒 M. Aslin 𝑓 ,1
4
+ M. Babicz𝑔,2 W.F. Badgett 𝑓 L.F. Bagby 𝑓 B. Baibussinov𝑑 B. Beheraℎ V. Bellini𝑖
5
+ O. Beltramello𝑔 R. Benocci 𝑗 J. Bergerℎ S. Berkman 𝑓 S. Bertolucci𝑘 R. Bertoni𝑗
6
+ M. Betancourt 𝑓 M. Bettini𝑑 S. Biagi𝑙 K. Biery 𝑓 O. Bitter 𝑓 ,3 M. Bonesini𝑗 T. Booneℎ
7
+ B. Bottino𝑚 A. Braggiotti𝑑,4 D. Brailsford5 J. Bremer𝑔 S.J. Brice 𝑓 V. Brio𝑖 C. Brizzolari𝑗
8
+ J. Brown 𝑓 H.S. Budd𝑐 F. Calaon𝑑 A. Campani𝑚 D. Carberℎ M. Carneiro𝑛 I. Caro Terrazasℎ
9
+ H. Carranza𝑒 D. Casazza𝑚 L. Castellani𝑑 A. Castro𝑜 S. Centro𝑑 G. Cerati 𝑓 M. Chalifour𝑔
10
+ P. Chambouvet𝑔 A. Chatterjee𝑝 D. Cherdack𝑏 S. Cherubini𝑙 N. Chithirasreemadam𝑞
11
+ M. Cicerchia𝑑 V. Cicero𝑘 T. Coan𝑟 A. Cocco𝑠 M.R. Convery𝑡 S. Copello𝑢 E. Cristaldo6
12
+ A.A. Dange𝑒 I. de Icaza Astiz7 A. De Roeck𝑔 S. Di Domizio𝑚 L. Di Noto𝑚 C. Di Stefano𝑙
13
+ D. Di Ferdinando𝑘 M. Diwan𝑛 S. Dolan𝑔 L. Domine𝑡 S. Donati𝑞 R. Doubnik 𝑓 F. Drielsma𝑡
14
+ J. Dyerℎ S. Dytman𝑣 C. Fabre𝑔 F. Fabris𝑑 A. Falcone𝑗 C. Farnese𝑑 A. Fava 𝑓 H. Ferguson 𝑓
15
+ A. Ferrari𝑤 F. Ferraro𝑚 N. Gallice𝑤 F.G. Garcia𝑡 M. Geynisman 𝑓 M. Giarin𝑑 D. Gibin𝑑
16
+ S.G. Gigli𝑢 A. Gioiosa𝑞 W. Gu𝑛 M. Guerzoni𝑘 A. Guglielmi𝑑 G. Gurung𝑒 S. Hahn 𝑓 K. Hardin 𝑓
17
+ H. Hausner 𝑓 A. Heggestuenℎ C. Hilgenbergℎ,8 M. Hoganℎ B. Howard 𝑓 R. Howell𝑐
18
+ J. Hrivnak𝑔 M. Iliescu𝑘,9 I. Ingratta𝑘 C. James 𝑓 W. Jang𝑒 M. Jung𝑥,10 Y.-J. Jwa𝑡 L. Kashurℎ
19
+ W. Ketchum 𝑓 J.S. Kim𝑐 D.-H. Koh𝑡 U. Kose𝑔,11 J. Larkin𝑛 G. Laurenti𝑘 G. Lukhanin 𝑓
20
+ S. Marchini𝑑 C.M. Marshall𝑐 S. Martynenko𝑛 N. Mauri𝑘 A. Mazzacane 𝑓 K.S. McFarland𝑐
21
+ D.P. Méndez𝑛 A. Menegolli𝑢,12 G. Meng𝑑 O.G. Miranda𝑜 D. Mladenov𝑔 A. Moganℎ N. Moggi𝑘
22
+ E. Montagna𝑘 C. Montanari 𝑓 ,13 A. Montanari𝑘 M. Mooneyℎ G. Moreno-Granados𝑜 J. Muellerℎ
23
+ D. Naples𝑣 M. Nebot-Guinot14 M. Nessi𝑔 T. Nichols 𝑓 M. Nicoletto𝑑 B. Norris 𝑓 S. Palestini𝑔
24
+ M. Pallavicini𝑚 V. Paolone𝑣 R. Papaleo𝑙 L. Pasqualini𝑘 L. Patrizii𝑘 R. Peghin𝑑 G. Petrillo𝑡
25
+ C. Petta𝑖 V. Pia𝑘 F. Pietropaolo𝑔,15 J. Poirot𝑔 F. Poppi𝑘 M. Pozzato𝑘 M.C. Prata𝑢 A. Prosser 𝑓
26
+ G. Putnam𝑤 X. Qian𝑛 G. Rampazzo𝑑 A. Rappoldi𝑢 G.L. Raselli𝑢 R. Rechenmacher 𝑓
27
+ F. Resnati𝑔 A.M. Ricci𝑞 G. Riccobene𝑙 L. Rice𝑣 E. Richards𝑣 A. Rigamonti𝑔 M. Rosenberg𝑎
28
+ 1Now at University of Wisconsin, Madison, USA
29
+ 2Also at INP-Polish Acad. Sci, Krakow,Poland. Now at University of Zurich, Switzerland
30
+ 3Now at Northwestern University, USA
31
+ 4Also at Istituto di Neuroscienze, CNR, Padova, Italy
32
+ 5SBND Collaboration, Lancaster University, UK
33
+ 6SBND Collaboration, Universidad Nacional de Asuncion, San Lorenzo, Paraguay
34
+ 7SBND Collaboration, University of Sussex, UK
35
+ 8Now at University of Minnesota, USA
36
+ 9Now at INFN-LNF
37
+ 10SBND Collaboration
38
+ 11Now at ETH Zurich, Switzerland
39
+ 12Corresponding author.
40
+ 13on leave of absence from INFN Pavia, Italy
41
+ 14SBND Collaboration, University of Edinburgh, UK
42
+ 15On leave of absence from INFN Padova, Italy
43
+ arXiv:2301.08634v1 [hep-ex] 20 Jan 2023
44
+
45
+ M. Rossella𝑢 C. Rubbia𝑦 P. Sala𝑤 P. Sapienza𝑙 G. Savage 𝑓 A. Scaramelli𝑢 A. Scarpelli𝑛
46
+ D. Schmitz𝑥 A. Schukraft 𝑓 F. Sergiampietri𝑔,16 G. Sirri𝑘 J.S. Smedley𝑐 A.K. Soha 𝑓
47
+ M. Spanu 𝑗 L. Stanco𝑑 J. Stewart𝑛 N.B. Suarez𝑣 C. Sutera𝑖 H.A. Tanaka𝑡 M. Tenti𝑘 K. Terao𝑡
48
+ F. Terranova 𝑗 V. Togo𝑘 D. Torretta 𝑓 M. Torti 𝑗 F. Tortorici𝑖 N. Tosi𝑘 Y.-T. Tsai𝑡 S. Tufanli𝑔
49
+ M. Turcato𝑑 T. Usher𝑡 F. Varanini𝑑 S. Ventura𝑑 F. Vercellati𝑢 M. Vicenzi𝑚 C. Vignoli𝑧
50
+ B. Viren𝑛 D. Warnerℎ Z. Williams𝑒 R.J. Wilsonℎ P. Wilson 𝑓 J. Wolfs𝑐 T. Wongjirad𝑎 A. Wood𝑏
51
+ E. Worcester𝑛 M. Worcester𝑛 M. Wospakrik 𝑓 H. Yu𝑛 J. Yu𝑒 A. Zani𝑤 P.G. Zatti𝑑 J. Zennamo 𝑓
52
+ J.C. Zettlemoyer 𝑓 C. Zhang𝑛 S. Zucchelli𝑘 and M. Zuckerbrot 𝑓
53
+ 𝑎Tufts University, Medford, MA 02155, USA
54
+ 𝑏University of Houston, Houston, TX 77204, USA
55
+ 𝑐University of Rochester, Rochester, NY 14627, USA
56
+ 𝑑INFN Sezione di Padova and University of Padova, Padova, Italy
57
+ 𝑒University of Texas at Arlington, Arlington, TX 76019, USA
58
+ 𝑓 Fermi National Accelerator Laboratory, Batavia, IL 60510, USA
59
+ 𝑔CERN, European Organization for Nuclear Research 1211 Genève 23, Switzerland, CERN
60
+ ℎColorado State University, Fort Collins, CO 80523, USA
61
+ 𝑖INFN Sezione di Catania and University of Catania, Catania, Italy
62
+ 𝑗INFN Sezione di Milano Bicocca and University of Milano Bicocca, Milano, Italy
63
+ 𝑘INFN Sezione di Bologna and University of Bologna, Bologna, Italy
64
+ 𝑙INFN LNS, Catania, Italy
65
+ 𝑚INFN Sezione di Genova and University of Genova, Genova, Italy
66
+ 𝑛Brookhaven National Laboratory, Upton, NY 11973, USA
67
+ 𝑜Centro de Investigacion y de Estudios Avanzados del IPN (Cinvestav), Mexico City
68
+ 𝑝Physical Research Laboratory, Ahmedabad, India
69
+ 𝑞INFN Sezione di Pisa, Pisa, Italy
70
+ 𝑟Southern Methodist University, Dallas, TX 75275, USA
71
+ 𝑠INFN Sezione di Napoli, Napoli, Italy
72
+ 𝑡SLAC National Acceleratory Laboratory, Menlo Park, CA 94025, USA
73
+ 𝑢INFN Sezione di Pavia and University of Pavia, Pavia, Italy
74
+ 𝑣University of Pittsburgh, Pittsburgh, PA 15260, USA
75
+ 𝑤INFN Sezione di Milano, Milano, Italy
76
+ 𝑥University of Chicago, Chicago, IL 60637, USA
77
+ 𝑦INFN GSSI, L’Aquila, Italy
78
+ 𝑧INFN LNGS, Assergi, Italy
79
+ E-mail: alessandro.menegolli@unipv.it
80
+ 16Now at IPSI-INAF Torino, Italy
81
+
82
+ Abstract: The ICARUS collaboration employed the 760-ton T600 detector in a successful three-
83
+ year physics run at the underground LNGS laboratory studying neutrino oscillations with the
84
+ CERN Neutrino to Gran Sasso beam (CNGS) and searching for atmospheric neutrino interactions.
85
+ ICARUS performed a sensitive search for LSND-like anomalous 𝜈𝑒 appearance in the CNGS
86
+ beam, which contributed to the constraints on the allowed parameters to a narrow region around
87
+ 1 eV2, where all the experimental results can be coherently accommodated at 90% C.L.. After a
88
+ significant overhaul at CERN, the T600 detector has been installed at Fermilab. In 2020, cryogenic
89
+ commissioning began with detector cool down, liquid argon filling and recirculation. ICARUS has
90
+ started operations and successfully completed its commissioning phase, collecting the first neutrino
91
+ events from the Booster Neutrino Beam (BNB) and the Neutrinos at the Main Injector (NuMI)
92
+ beam off-axis, which were used to test the ICARUS event selection, reconstruction and analysis
93
+ algorithms. The first goal of the ICARUS data taking will then be a study to either confirm or refute
94
+ the claim by Neutrino-4 short baseline reactor experiment both in the 𝜈𝜇 channel with the BNB and
95
+ in the 𝜈𝑒 with NuMI. ICARUS will also address other fundamental studies such as neutrino cross
96
+ sections with the NuMI beam and a number of Beyond Standard Model searches. After the first
97
+ year of operations, ICARUS will commence its search for evidence of a sterile neutrino jointly with
98
+ the Short Baseline Near Detector, within the Short-Baseline Neutrino program.
99
+ Keywords: Large detector systems for particle and astro-particle physics, Liquid Argon, Time
100
+ Projection Chambers (TPC)
101
+
102
+ Contents
103
+ 1
104
+ Introduction
105
+ 2
106
+ 2
107
+ The ICARUS-T600 detector
108
+ 3
109
+ 3
110
+ The overhaul of ICARUS-T600
111
+ 4
112
+ 3.1
113
+ The TPC electronics
114
+ 4
115
+ 3.2
116
+ The scintillation light detection system
117
+ 5
118
+ 4
119
+ The Cosmic Ray Tagger
120
+ 6
121
+ 5
122
+ First operations at FNAL
123
+ 7
124
+ 5.1
125
+ Cryogenic plant installation
126
+ 7
127
+ 5.2
128
+ TPC electronics installation
129
+ 9
130
+ 5.3
131
+ PMT system installation
132
+ 9
133
+ 5.4
134
+ Cosmic Ray Tagger installation
135
+ 10
136
+ 6
137
+ ICARUS-T600 commissioning
138
+ 10
139
+ 6.1
140
+ TPC commissioning
141
+ 12
142
+ 6.2
143
+ PMT commissioning
144
+ 16
145
+ 6.3
146
+ CRT commissioning
147
+ 17
148
+ 6.4
149
+ Triggering on the BNB and NuMI neutrinos
150
+ 18
151
+ 6.5
152
+ DAQ implementation
153
+ 20
154
+ 6.6
155
+ First operations with the BNB and NuMI
156
+ 21
157
+ 7
158
+ Observation and reconstruction of neutrino events
159
+ 21
160
+ 7.1
161
+ Wire signal reconstruction
162
+ 21
163
+ 7.2
164
+ PMT signal reconstruction
165
+ 22
166
+ 7.3
167
+ CRT reconstruction
168
+ 23
169
+ 7.4
170
+ Event display study
171
+ 25
172
+ 7.5
173
+ Event reconstruction
174
+ 25
175
+ – 1 –
176
+
177
+ 1
178
+ Introduction
179
+ The Liquid Argon Time Projection Chamber
180
+ (LAr-TPC) is a continuously sensitive and self
181
+ triggering detector that can provide excellent
182
+ 3D imaging and calorimetric reconstruction of
183
+ any ionizing event. First proposed by C. Rub-
184
+ bia in 1977 [1], this detection technique allows
185
+ a detailed study of neutrino interactions, span-
186
+ ning a wide energy spectrum (from a few keV
187
+ to several hundreds of GeV), as demonstrated
188
+ by the first large scale experiment performed by
189
+ the ICARUS Collaboration at the LNGS under-
190
+ ground laboratory.
191
+ Several experiments, in particular the Liq-
192
+ uid Scintillator Neutrino Detector (LSND) [2]
193
+ and MiniBooNE [3], have reported anomalous
194
+ signals that may imply the presence of additional
195
+ (mass-squared difference Δ𝑚2
196
+ 𝑛𝑒𝑤 ∼ 1 eV2) flavor
197
+ oscillations at small distances pointing toward
198
+ the possible existence of nonstandard heavier
199
+ sterile neutrino(s). A sensitive search for a possi-
200
+ ble 𝜈𝑒 excess related to the LSND anomaly in the
201
+ CNGS 𝜈𝜇 beam has already been performed us-
202
+ ing the neutrino events collected in the ICARUS-
203
+ T600 detector during the Gran Sasso run. A total
204
+ of 2,650 CNGS neutrino interactions, identified
205
+ in 7.9·1019 POT (Protons On Target) exposure,
206
+ have been studied to identify the 𝜈𝑒 interactions.
207
+ Globally, 7 electron-like events have been ob-
208
+ served to be compared to 8.5±1.1 expected from
209
+ the intrinsic beam contamination and standard
210
+ 3-flavor oscillations. This study constrained the
211
+ LSND signal to a narrow parameter region at
212
+ sin22𝜃 ∼ 0.005, Δ𝑚2 < 1 eV2, which requires
213
+ further investigation [4].
214
+ The primary goal of the Short-Baseline
215
+ Neutrino (SBN) program at Fermilab is to fur-
216
+ ther investigate the possibility of sterile neutri-
217
+ nos in the O(1 eV) mass range and provide the
218
+ required clarification of the LSND anomaly. It
219
+ is based on three LAr-TPC detectors (ICARUS-
220
+ T600, with 476 tons active mass, MicroBooNE
221
+ with 89 tons active mass and SBND with 112
222
+ tons active mass) exposed at shallow depth to
223
+ the ∼ 0.8 GeV Booster Neutrino Beam (BNB) at
224
+ different distances from the target (600 m, 470
225
+ m and 110 m respectively) [5, 6].
226
+ The detection technique used will provide
227
+ an unambiguous identification of neutrino in-
228
+ teractions, measurement of their energy and a
229
+ strong mitigation of possible sources of back-
230
+ ground. Performing this study with almost iden-
231
+ tical detectors at various distances from the neu-
232
+ trino source allows identification of any variation
233
+ of the spectra, which is a clear signature of neu-
234
+ trino oscillations.
235
+ In particular, SBN will allow for a very sen-
236
+ sitive search for 𝜈𝜇 → 𝜈𝑒 appearance signals,
237
+ covering the LSND 99% C.L. allowed region at
238
+ ∼ 5𝜎 C.L. [5, 6]. The high correlations between
239
+ the event samples of the three LAr-TPC’s and the
240
+ huge event statistics at the near detector will also
241
+ allow for a simultaneous sensitive search in the
242
+ 𝜈𝜇 disappearance channel.
243
+ During data taking at Fermilab, the 760-
244
+ ton T600 detector is also exposed to the off-axis
245
+ neutrinos from the Neutrinos at the Main Injec-
246
+ tor (NuMI) beam, where most of events are in
247
+ the 0 – 3 GeV energy range, with an enriched
248
+ component of electron neutrinos (few %). The
249
+ analysis of these events will provide useful infor-
250
+ mation related to detection efficiencies and neu-
251
+ trino cross-sections at energies relevant to the
252
+ future long baseline experiment with the multi-
253
+ kiloton DUNE LAr-TPC detector.
254
+ In addition to the LSND anomaly, ICARUS
255
+ will test the oscillation signal reported by the
256
+ Neutrino-4 experiment [7] both in the 𝜈𝜇 and
257
+ 𝜈𝑒 channels with the BNB and NuMI beams,
258
+ respectively.
259
+ This paper is organized as follows: in Sec-
260
+ tion 2 the ICARUS-T600 detector is described
261
+ with a particular emphasis on its achievements
262
+ during three years data taking at the INFN LNGS
263
+ underground laboratories in Italy; in Section 3,
264
+ – 2 –
265
+
266
+ the ICARUS-T600 overhauling activities, most
267
+ of which were carried out at CERN in the Neu-
268
+ trino Platform framework [8], are shown; the
269
+ new Cosmic Ray Tagger (CRT) detector, used
270
+ to mitigate the cosmic ray background due to
271
+ operating ICARUS at shallow depth, is detailed
272
+ in Section 4. In Section 5, the first operations
273
+ of ICARUS at Fermilab, in particular the instal-
274
+ lation of the cryogenic plant, TPC electronics,
275
+ scintillation light detection system and CRT are
276
+ described. A successful commissioning phase
277
+ followed soon after as described in Section 6.
278
+ Finally, the procedure for the selection, recon-
279
+ struction, and analysis of the first collected BNB
280
+ and NuMI off-axis neutrino events is introduced
281
+ in Section 7.
282
+ 2
283
+ The ICARUS-T600 detector
284
+ The ICARUS-T600, with a total active mass of
285
+ 476 ton, is the first large-scale operating LAr-
286
+ TPC detector [9]: it consists of two large and
287
+ identical adjacent modules with internal dimen-
288
+ sions 3.6 × 3.9 × 19.6 m3, filled with a total of
289
+ 760 tons of ultra-pure liquid argon. Each mod-
290
+ ule houses two LAr-TPCs separated by a com-
291
+ mon cathode with a maximum drift distance of
292
+ 1.5 m, equivalent to ∼ 1 ms drift time for the
293
+ nominal 500 V/cm electric drift field. The cath-
294
+ ode is built up by an array of nine panels made of
295
+ punched stainless-steel, allowing for a 58% op-
296
+ tical transparency between the two drift regions.
297
+ The anode is made of three parallel wire planes
298
+ positioned 3 mm apart, where the stainless-steel
299
+ 100 µm wires are oriented on each plane at a
300
+ different angle with respect to the horizontal di-
301
+ rection: 0◦ (Induction 1), +60◦ (Induction 2)
302
+ and -60◦ (Collection).
303
+ In total, 53,248 wires
304
+ with a 3 mm pitch and length up to 9 m are in-
305
+ stalled in the detector. By appropriate voltage
306
+ biasing, the first two planes (Induction 1 and In-
307
+ duction 2) provide a nondestructive charge mea-
308
+ surement, whereas the ionization charge is fully
309
+ collected by the last Collection plane. Photo-
310
+ Multiplier Tubes (PMTs) are located behind the
311
+ wire planes to collect the scintillation light pro-
312
+ duced by charged particles in LAr and used for
313
+ the trigger of the detector.
314
+ In 2013, ICARUS concluded a very suc-
315
+ cessful 3-year long run in the Gran Sasso under-
316
+ ground laboratory [10], demonstrating the feasi-
317
+ bility of the LAr-TPC technology at the kiloton
318
+ scale in a deep underground environment and
319
+ paving the way to the construction of the next
320
+ generation of experiments dedicated to study
321
+ neutrino oscillation physics such as DUNE. Dur-
322
+ ing the data taking, the liquid argon was kept
323
+ at an exceptionally high purity level (< 50 ppt
324
+ of O2 equivalent contaminants) reaching in 2013
325
+ a 16 ms lifetime corresponding to 20 ppt O2
326
+ equivalent LAr contamination [11], demonstrat-
327
+ ing the possibility to build larger LAr-TPC de-
328
+ tectors with drift distances up to 5 m.
329
+ The detector has been exposed to the CNGS
330
+ neutrino beam and to cosmic rays, recording
331
+ events that demonstrate high-level performance
332
+ and the physical potential of this detection tech-
333
+ nique: the detector showed a remarkable 𝑒/𝛾
334
+ separation and particle identification exploiting
335
+ the measurement of 𝑑𝐸/𝑑𝑥 versus range [12].
336
+ The momentum of escaping muons has been
337
+ measured by studying the multiple Coulomb
338
+ scattering with ∼ 15% average resolution in the
339
+ 0.4 – 4 GeV/c energy range, which is relevant for
340
+ the next generation neutrino experiments [13].
341
+ Events related to cosmic rays have been
342
+ studied to identify atmospheric neutrino interac-
343
+ tions: 6 𝜈𝜇CC and 8 𝜈𝑒CC events in a 0.43 kton·y
344
+ exposure have been identified and reconstructed,
345
+ demonstrating that the automatic search for the
346
+ 𝜈𝑒CC in the sub-GeV range of interest for the
347
+ future short and long baseline neutrino experi-
348
+ ments is feasible [14].
349
+ – 3 –
350
+
351
+ 3
352
+ The overhaul of ICARUS-T600
353
+ The ICARUS-T600 detector at Fermilab takes
354
+ data at shallow depth, shielded by a ∼ 3-meter
355
+ concrete overburden: neutrino interactions must
356
+ be recognized among the ∼ 11 cosmic muons
357
+ that are expected to cross the detector randomly
358
+ in the 1 ms drift time during each triggered event.
359
+ High-energy photons produced by cosmic rays
360
+ can become a serious background source for the
361
+ 𝜈𝑒 search since the electrons produced via Comp-
362
+ ton scattering and pair production can mimic
363
+ 𝜈𝑒CC events.
364
+ In order to prepare the detector for SBN data
365
+ taking, the T600 underwent an intensive overhaul
366
+ at CERN in the Neutrino Platform framework
367
+ (WA104/NP01 project) before being shipped to
368
+ the USA in 2017, introducing several technology
369
+ developments while maintaining the achieved
370
+ performance at Gran Sasso.
371
+ The refurbishing
372
+ mainly consisted of: the realization of new cold
373
+ vessels (Fig. 1) with purely passive insulation; an
374
+ update of the cryogenics and of the LAr purifi-
375
+ cation equipment; flattening of the TPC cathode
376
+ (the punched hole stainless-steel panels under-
377
+ went a thermal treatment improving the planarity
378
+ to a few mm); the implementation of new, higher
379
+ performance TPC read-out electronics; the up-
380
+ grade of the LAr light detection system.
381
+ 3.1
382
+ The TPC electronics
383
+ The electronics used at LNGS was based on
384
+ flange modularity, each flange serving 576 TPC
385
+ wire-channels.
386
+ The analogue front-end was a
387
+ Radeka type amplifier, using a custom BiCMOS
388
+ chip to integrate the cascode stage with two dif-
389
+ ferent filtering, one for Collection and Induc-
390
+ tion 1, another for Induction 2 with the aim
391
+ to produce in all the cases a unipolar signal.
392
+ This solution, however, showed strong limita-
393
+ tions in the Induction 2 signals in the case of
394
+ dense showers. Analog signals were converted
395
+ to digital via multiplexers by 10-bit ADCs with
396
+ Figure 1. One of the two new ICARUS cryostats
397
+ during its assembly at a CERN workshop.
398
+ sampling rate of 400 ns. The analogue circuits
399
+ were housed in a custom crate, connected to the
400
+ flange by flat cables, with 18 boards (32 chan-
401
+ nels per board). Analogue boards had a digital
402
+ link to corresponding digital modules hosted in
403
+ VME crates that contained memory buffers and
404
+ performed lossless data compression and data
405
+ transmission through a VME bus. Both crates
406
+ were housed in a rack next to the flange.
407
+ One of the largest tasks of the overhauling
408
+ program was the design of new electronics for
409
+ the 53,248 wire-channels that would be compat-
410
+ ible with higher data rates foreseen at shallow
411
+ depth operation at FNAL. The new electronics
412
+ adopts the same modularity and architecture but
413
+ takes advantage of newer technology that allows
414
+ for integrating both the analogue and the digital
415
+ electronics on the same board on a custom crate
416
+ mounted onto the flange [15].
417
+ New packaging for the BiCMOS custom
418
+ cascode allowed the design of a small piggyback
419
+ module with 8 amplifiers and to house 8 of these
420
+ modules on a single board serving 64 channels,
421
+ see Fig. 2 (top-left). The digital part is also com-
422
+ pletely contained in the same board. Moreover,
423
+ all the amplifiers now have the same filtering,
424
+ preserving the bipolar structure of Induction 2
425
+ signals without distortion. Each amplifier is fol-
426
+ – 4 –
427
+
428
+ lowed by a serial 12-bit ADC avoiding the cum-
429
+ bersome signal multiplexing.
430
+ The digital part
431
+ is based essentially on a large powerful FPGA
432
+ allowing the possibility to use different signal
433
+ treatments if required from running experience.
434
+ The VME standard was abandoned in favor of
435
+ a serial optical link, allowing for gigabit band-
436
+ width data transmission compatible with shallow
437
+ depth data rates.
438
+ Figure 2. A2795 custom board housing 64 amplifiers
439
+ (far end), AD converter, digital control, and optical
440
+ link (top-left). An assembled feed-through with nine
441
+ DBBs and the biasing cables (top-right).
442
+ A mini-
443
+ crate populated by the nine A2795 boards installed
444
+ on a feed-through flange (bottom).
445
+ TPC wire signals are fed into the front-
446
+ end amplifiers by means of Decoupling Biasing
447
+ Boards (DBBs). The DBB has two functions:
448
+ biasing of each wire and conveying, with block-
449
+ ing capacitors, the signals to the amplifiers. The
450
+ DBBs work in argon gas and can withstand up
451
+ to 400 V input biasing. The flange CF250 is re-
452
+ alized with a G10 multi-layer solid PCB, about
453
+ 6 mm thick with three internal layers of copper
454
+ to guarantee the required stiffness. SMD exter-
455
+ nal connectors provide receptacles for the A2795
456
+ boards, while another set of SMD connectors in
457
+ correspondence (inner side) provide receptacles
458
+ for DBBs, see Fig. 2 (top-right). Finally, nine
459
+ electronic A2795 boards are hosted by a mini-
460
+ crate which is installed on a feed-through CF250
461
+ flange, see Fig. 2 (bottom).
462
+ 3.2
463
+ The scintillation light detection system
464
+ A new light detection system that is sensitive to
465
+ the photons produced by the LAr scintillation is
466
+ a fundamental feature for the T600 operation at
467
+ shallow depth (contributing to the rejection of the
468
+ cosmic background). The light detection system
469
+ complements the 3D track reconstruction, unam-
470
+ biguously providing the absolute timing for each
471
+ track and identifying the interactions occurring
472
+ in the BNB and NuMI spill gates.
473
+ The ICARUS-T600 light detection system
474
+ consists of 360 8" Hamamatsu R5912-MOD
475
+ PMTs deployed behind the 4 wire chambers,
476
+ 90 PMTs per TPC [16, 17], see Fig. 3. Since
477
+ the PMT glass is not transparent to the 128 nm
478
+ wavelength scintillation light produced in liquid
479
+ argon, each unit is provided with a ≈ 200 µg/cm2
480
+ coating of Tetra-Phenyl Butadiene (TPB), to con-
481
+ vert the VUV photons to visible light [18].
482
+ All PMTs are mounted onto the wire cham-
483
+ ber mechanical frames using a supporting sys-
484
+ tem, that allows the PMT to be positioned about
485
+ 5 mm behind the Collection planes wires.
486
+ A
487
+ stainless steel grid cage is mounted around each
488
+ PMT to mitigate the induction of fake signals
489
+ on the nearby wire planes by the relatively large
490
+ PMT signals.
491
+ The light detection setup, realized by INFN,
492
+ is complemented by a laser calibration system
493
+ allowing for gain equalization, timing and moni-
494
+ – 5 –
495
+
496
+ Figure 3. The new ICARUS PMTs mounted behind
497
+ the wires of one TPC.
498
+ toring of all the PMTs. Laser pulses (𝜆 = 405 nm,
499
+ FWHM = 60 ps), generated by a laser diode head
500
+ (Hamamatsu PLP10), are sent to each PMT win-
501
+ dow by means of a light distribution system based
502
+ on optical fibers, light splitters and an optical
503
+ switch [19].
504
+ 4
505
+ The Cosmic Ray Tagger
506
+ ICARUS-T600 based at FNAL faces more chal-
507
+ lenging experimental conditions than at LNGS:
508
+ due to its shallow depth operation, identifica-
509
+ tion of neutrino interactions among 11 kHz of
510
+ cosmic rays is required. A ∼ 3-meter concrete
511
+ overburden was designed to almost completely
512
+ remove the contribution from charged hadrons
513
+ and high energy photons [20]. However, ∼ 11
514
+ muon tracks occur per triggered event in the 1 ms
515
+ TPC drift readout; photons associated with the
516
+ muons represent a serious background for identi-
517
+ fying 𝜈𝑒 candidates since electrons produced via
518
+ Compton scattering/pair production can mimic a
519
+ genuine 𝜈𝑒CC event.
520
+ Rejecting the cosmic background, i.e. re-
521
+ constructing the triggering event, requires to
522
+ know precisely the timing of each track in the
523
+ TPC image. Operating at FNAL, ICARUS ex-
524
+ ploits an improved light detection system with
525
+ high granularity and 𝑂(1 ns) time resolution, and
526
+ an external ∼ 4𝜋 high coverage Cosmic Ray Tag-
527
+ ger (CRT). The primary function of the CRT is to
528
+ tag muons passing through or near the cryostats.
529
+ Timestamps associated to a particle tagged
530
+ by the CRT are compared with timestamps from
531
+ PMT signals, both with a few nanosecond res-
532
+ olution, allow the determination of whether an
533
+ interaction in the TPC originated from an outside
534
+ cosmic ray or from an internal interaction. The
535
+ ICARUS CRT consists of a top, side and bottom
536
+ subsystem.
537
+ The ICARUS Top CRT system is divided in
538
+ 123 detector modules covering a surface of about
539
+ 426 m2: 84 horizontal and 39 vertical modules
540
+ along the perimeter of the cryostat top surface.
541
+ Its design is such that more than 80% of the cos-
542
+ mic muon flux is intercepted by the Top CRT.
543
+ Each module is a 1.86 × 1.86 m2 aluminum
544
+ box containing two orthogonal layers of eight
545
+ scintillator bars for position reconstruction. The
546
+ bars, coated with white paint, are 23 cm wide,
547
+ 184 cm long and have different thickness de-
548
+ pending on the layer: 1 cm and 1.5 cm for the
549
+ top layer and the bottom layer, respectively. In
550
+ each scintillator, the light is collected by two
551
+ wave-length shifting (WLS) fibers Kuraray Y-
552
+ 11(200) then read out from one end by a Silicon
553
+ Photo-Multiplier (SiPM), Hamamatsu S13360-
554
+ 1350C model. The 32 SiPM signals of one mod-
555
+ ule are routed via 50 Ω micro-coaxial cables to
556
+ a patch panel connected to the CAEN DT5702
557
+ Front End Board (FEB) which provides a bias
558
+ voltage adjustable for each channel. The FEB
559
+ triggers on the coincidence between two SiPM
560
+ signals of the same bar and provides a coinci-
561
+ dence logic between the two scintillator layers
562
+ in the module. In Fig. 4, a picture of a vertical
563
+ Top CRT module installed in the detector hall is
564
+ shown. The Top CRT was a brand new detector
565
+ – 6 –
566
+
567
+ designed and built by INFN and CERN before
568
+ shipping to Fermilab in summer 2021.
569
+ Front End
570
+ Board
571
+ Aluminum Box
572
+ containing Top
573
+ CRT module
574
+ Figure 4. Picture of a vertical TOP CRT module
575
+ installed in the detector hall.
576
+ The ICARUS Side CRT makes use of scin-
577
+ tillator modules formerly used by the MINOS ex-
578
+ periment. Each module is composed of twenty
579
+ adjacent strips of 800 × 4 × 1 cm3 Polystyrene
580
+ (1.0% PPO, 0.03% POPOP) scintillator.
581
+ The
582
+ full Side CRT system consists of 2,710 readout
583
+ channels across 93 FEBs, with 136 full and 81
584
+ cut modules in total.
585
+ The scintillator is con-
586
+ tained in a metal sheath and each strip has an
587
+ embedded WLS fiber running down the mid-
588
+ dle. These fibers are collected into “snouts” at
589
+ the ends of the modules, onto which the opti-
590
+ cal readout, consisting of an array of ten Hama-
591
+ matsu S14160-3050HS SiPMs, is mounted onto
592
+ a snout. Each SiPM reads out two fibers and cor-
593
+ responds to a single electronic readout channel
594
+ on CAEN A1702 Front-End Boards (FEBs). A
595
+ full MINOS module has two snouts, one on each
596
+ end. The ICARUS Side CRT System is double
597
+ layered, with an inner and outer layer of MINOS
598
+ modules to apply coincidence logic between the
599
+ two layers. To account for geometric constraints,
600
+ some MINOS modules were cut and sealed on
601
+ the cut end with mylar and tape to only have a
602
+ single snout for readout. The South Side CRT
603
+ wall consists of an inner and outer layer of cut
604
+ modules oriented orthogonally in an X-Y con-
605
+ figuration, with the added benefit of improved
606
+ position reconstruction on the southern side of
607
+ the TPCs, upstream along the BNB beam. The
608
+ East and West walls utilize full length MINOS
609
+ modules mounted horizontally, while the North
610
+ Wall use cut modules mounted horizontally.
611
+ The Bottom CRT consists of 14 modules di-
612
+ vided into two daisy chains of 7 modules each,
613
+ positioned underneath the warm vessel in a north
614
+ and south section.
615
+ These modules are refur-
616
+ bished veto modules from the Double Chooz re-
617
+ actor neutrino experiment. Each module consists
618
+ of 64 Polystyrene scintillator strips, running in
619
+ parallel and divided into two layers of 32 strips
620
+ offset 2.5 cm from each other. Scintillation light
621
+ is collected in a WLS optical fiber and read out at
622
+ one end of each strip by an Hamamatsu H7546B
623
+ M64 multi-anode PMT, while the other end is
624
+ mirrored to maximize light collection.
625
+ 5
626
+ First operations at FNAL
627
+ Following the overhauling activities at CERN,
628
+ ICARUS-T600 was shipped to Fermilab in July
629
+ 2017 and the two cryostats hosting the TPCs were
630
+ finally deployed in their shallow depth position
631
+ in August 2018. Work began soon after to install
632
+ and test all main subsystems before the cryogenic
633
+ commissioning, see Fig. 5.
634
+ 5.1
635
+ Cryogenic plant installation
636
+ The ICARUS cryogenic plant was designed,
637
+ built and installed at Fermilab by a collabora-
638
+ tion of three international institutions, CERN,
639
+ INFN and Fermilab to support operations of the
640
+ ICARUS LAr-TPC. For the installation at Fer-
641
+ milab, the entire ICARUS-T600 cryogenic and
642
+ purification system was rebuilt anew. The new
643
+ design followed closely the original implementa-
644
+ tion at the LNGS with one important exception:
645
+ – 7 –
646
+
647
+ Figure 5. Deployment of the ICARUS cryostats inside the pit of the SBN Far Detector experimental hall at
648
+ Fermilab in August 2018 (left). Installation of TPC, PMT and laser feed-through flanges in December 2018
649
+ (center). Status of the ICARUS detector at the beginning of data taking for commissioning (right).
650
+ at Fermilab, the LN2 boiloff is vented to the
651
+ atmosphere (open loop cooling circuit), while
652
+ at LNGS the LN2 boiloff was re-condensed by
653
+ means of a set of cryocoolers (closed loop cool-
654
+ ing circuit). The main components of the cryo-
655
+ genic and purification system are the following:
656
+ • Main LAr containers (2× cold vessels):
657
+ 273 m3 each, containing the TPC detectors
658
+ and the LAr scintillation light system.
659
+ • Cold shields: set of heat exchangers filled
660
+ with LN2, completely surrounding the
661
+ main LAr containers and designed to pre-
662
+ vent heat, coming from the thermal insu-
663
+ lation, to reach the LAr volumes.
664
+ • Thermal insulation:
665
+ polyurethane foam
666
+ panels, ∼ 600 mm thick, surrounding the
667
+ cold shields.
668
+ • Warm vessel: provides enclosure and me-
669
+ chanical support for the thermal insula-
670
+ tion.
671
+ • LN2 cooling circuits: piping, circulation
672
+ pumps, regulating valves, phase separa-
673
+ tors, etc., providing LN2 supply to heat
674
+ exchangers serving the cold shields and
675
+ the purifying units.
676
+ • Argon gas recirculation units (4×, two per
677
+ cold vessel): set of units that re-condense
678
+ and purify the argon flowing from the gas
679
+ phase on top of the main LAr containers.
680
+ • Liquid argon recirculation units (2×, one
681
+ per cold vessel): provide forced circula-
682
+ tion, with a cryogenic pump, of argon
683
+ coming from the cold vessels through a set
684
+ of purifiers before injecting it back into the
685
+ cold vessel.
686
+ • Cryogenic control system:
687
+ to provide
688
+ automation, data display, recording and
689
+ alarming.
690
+ • LN2 and LAr storage dewars and relative
691
+ transfer lines.
692
+ • A dedicated purification unit used for the
693
+ filling of the cold vessels, equipped with a
694
+ regeneration system and a set of gas ana-
695
+ lyzers.
696
+ The ICARUS cryogenic plant at the SBN Far
697
+ Detector Hall at Fermilab was fully designed,
698
+ delivered, and installed by July 2019, with the
699
+ commissioning phase started by January 2020.
700
+ The equipment included the ICARUS cryogenic
701
+ plant is schematically divided into the external
702
+ components supplied by Fermilab, the proximity
703
+ components supplied by Demaco Holland B.V.
704
+ under contract with CERN and components in-
705
+ ternal to the cryostats supplied by INFN. Fig. 6
706
+ shows the ICARUS plant physical layout.
707
+ – 8 –
708
+
709
+ lcarusTritFigure 6. ICARUS cryogenic plant physical layout.
710
+ 5.2
711
+ TPC electronics installation
712
+ Each mini-crate, housing nine A2795 boards,
713
+ was mounted onto the flange on top of the chim-
714
+ ney that contains flat cables connecting wires of
715
+ the chambers to DBBs and powered by a linear
716
+ power supply next to the chimney, see Fig. 7.
717
+ Each set of nine A2795 in a single crate are read
718
+ out through two fibers that implement a CAEN
719
+ proprietary protocol named CONET (Chain-able
720
+ Optical NETwork). The two sets of fibers are
721
+ read through an A3818 PCI Express board in-
722
+ stalled in dedicated PCs.
723
+ The full TPC electronics (96 mini-crates) is
724
+ synchronized by a serial link (one cable), named
725
+ TTLink, which sends clock, trigger, and com-
726
+ mands. The TTLinks are distributed to all mini-
727
+ crates by four fan-out modules with the same
728
+ cable lengths to guarantee equal time delay. The
729
+ TPC electronics system is fully installed and op-
730
+ erational.
731
+ 5.3
732
+ PMT system installation
733
+ Electrical connections between PMTs and elec-
734
+ tronics, located in a building alcove adjacent to
735
+ the detector, were realized by means of 360 sig-
736
+ nal cables and 360 high voltage cables.
737
+ Sig-
738
+ nal cables are RG316/U, 7 m of which are de-
739
+ Figure 7. Two Low Voltage Power Supply (LVPS)
740
+ modules powering the two adjacent mini-crates pop-
741
+ ulated with nine A2795 boards, serving 576 wires
742
+ each.
743
+ ployed inside the detector and 37 m outside, the
744
+ two parts connected by means of BNC-BNC
745
+ feedthrough flanges.
746
+ High voltage cables are
747
+ 7-m long HTC-50-1-1 deployed inside the de-
748
+ tector and 37 m RG58/U outside; the two parts
749
+ connected by means of SHV-SHV feedthrough
750
+ flanges. Power supply voltages are generated and
751
+ distributed by 8 CAEN A7030 boards, each with
752
+ 48 channels that can provide 3 kV, housed in two
753
+ CAEN SY4527 crates.
754
+ The PMT electronics are designed to allow
755
+ continuous read-out, digitization and indepen-
756
+ – 9 –
757
+
758
+ Proximity cryogenics on detector top:
759
+ LN2 shields valve boxes
760
+ LAR
761
+ GAr re-condensers valve boxes
762
+ Transfer lines and gas collection piping
763
+ Fill Filter
764
+ External cryogenics:
765
+ LAr and LN2 dewars
766
+ Transfer and vent lines
767
+ Proximity cryogenics in pit and mezzanine:
768
+ Regeneration skids for filter media
769
+ LAr pumps valve boxes
770
+ Gas analyzers
771
+ LAr filters valve boxes
772
+ Process controls system
773
+ LN2 Phase separator and pumps valve boxes
774
+ ← 23 m
775
+ Safety controls systemWE05/06
776
+ TRIPP-LITE
777
+ CINENdent waveform recording of signals coming from
778
+ the 360 PMTs. This operation is performed by
779
+ 24 CAEN V1730B digitizers installed in 8 VME
780
+ crates (3 digitizers per crate). Each module con-
781
+ sists of a 16-channel 14-bit 500-MSa/s FLASH
782
+ ADC with 2 Vpp input dynamic range. In each
783
+ board 15 channels are used for the acquisition of
784
+ PMT pulses, while one channel is used for the
785
+ acquisition of ancillary signals such as the beam
786
+ gates and the trigger pulses.
787
+ For each channel, an internal trigger-request
788
+ logic signal is generated every time the sam-
789
+ pled PMT pulse passes through a programmable
790
+ threshold. For each couple of adjacent channels,
791
+ trigger-requests are logically combined (OR,
792
+ AND, Ch0, Ch1) and the result is presented in a
793
+ low-voltage differential signaling (LVDS) logic
794
+ output with settable duration. For triggering pur-
795
+ poses, an OR logic between neighboring PMTs
796
+ is adopted. A total of 192 LVDS lines (8 lines
797
+ per digitizer) are connected to the ICARUS trig-
798
+ ger system for exploiting the scintillation light
799
+ information for trigger purposes.
800
+ The PMT electronics are complemented by
801
+ a common 62.5 MHz clock distribution system,
802
+ an external trigger network, an external time-
803
+ stamp reset network, and 24 optical link inter-
804
+ faces based on the CAEN CONET2 protocol.
805
+ 5.4
806
+ Cosmic Ray Tagger installation
807
+ The Side CRT system was installed over the pe-
808
+ riod from November 2019 to April 2021 (Fig. 8
809
+ left). Following its shipping in summer 2021, the
810
+ installation of Top CRT modules was carried out
811
+ and completed in December 2021 (Fig. 8 right).
812
+ All Top and Side CRT modules were tested be-
813
+ fore and after their installation to check for elec-
814
+ tronic functionality of the channels. Data trans-
815
+ mission to the servers is performed via ethernet
816
+ cables connecting the modules in daisy chain.
817
+ The distribution of a Pulse Per Second (PPS)
818
+ signal (see Sec. 6.4) for absolute time reference
819
+ and trigger signal to the FEBs was performed
820
+ with lemo cables. A voltage of 5.5 V to be pro-
821
+ vided to the FEBs is distributed via power lines
822
+ assembled at FNAL during the installation. All
823
+ the information on modules to cables connec-
824
+ tions, SiPM bias voltages, module positions, etc.
825
+ are stored in a Fermilab SQL database.
826
+ The last ICARUS installation activity was
827
+ the deployment of the 2.85-meter concrete over-
828
+ burden above the Top CRT. The overburden is
829
+ composed of three layers of concrete blocks,
830
+ each approximately 1-meter tall, giving a total
831
+ mass of 5 million pounds. The installation of the
832
+ last concrete block was completed June 7, 2022,
833
+ marking the beginning of ICARUS data taking
834
+ for physics with both BNB and NuMI beams.
835
+ 6
836
+ ICARUS-T600 commissioning
837
+ After the placement of the two ICARUS modules
838
+ in the pit in August 2018, all the feed-through
839
+ flanges for the TPC and PMT signals and for
840
+ the injection of the laser flashes used to calibrate
841
+ the PMTs were installed in December 2018. The
842
+ gain and the dark rate for all 360 PMTs were mea-
843
+ sured as a function of the applied voltage at room
844
+ temperature. All the new TPC readout electron-
845
+ ics in the 96 mini-crates and the low noise power
846
+ supplies were installed and verified. In particular
847
+ the full readout chain has been tested by injecting
848
+ test pulses in wires at the far end of the cham-
849
+ ber and reading out the signals with the A2795
850
+ boards on the other end to check the full system
851
+ for noise monitoring purposes.
852
+ In parallel all the cryogenic equipment were
853
+ installed, welded and the complete system has
854
+ been tested at 350 mbar over-pressure. The cold
855
+ vessels were then successfully brought to vac-
856
+ uum, with a 10−5 mbar residual pressure.
857
+ The
858
+ cryogenic
859
+ commissioning
860
+ of
861
+ the
862
+ ICARUS-T600 detector started on February 13,
863
+ 2020 by breaking the vacuum in the two main
864
+ cold vessels with ultra-purified argon gas. Cool
865
+ down started on February 14 by injecting liq-
866
+ – 10 –
867
+
868
+ Figure 8.
869
+ Left: picture of the Side CRT. Right: Top CRT horizontal modules whose installation was
870
+ completed in December 2021.
871
+ uid nitrogen in the cold shields. It took about
872
+ four days to bring the temperature on the wire
873
+ chamber below 100 K. The cooling process was
874
+ continuous and the maximum temperature gra-
875
+ dient on the wire chambers was about 35 K. On
876
+ February 19, the gas recirculation units were put
877
+ into operation to purify the argon gas before the
878
+ start of the liquid filling.
879
+ The continuous filling with ultra-purified
880
+ liquid argon started on February 24. The filling
881
+ was interrupted at around 50% to regenerate the
882
+ filling filter. The filling was stopped again when
883
+ the liquid reached the −6 cm LAr level probes
884
+ (6 cm below the nominal level) to perform the
885
+ final pressure test of the two cold vessels. After
886
+ the test, the gas recirculation units were put into
887
+ operation.
888
+ The filling was completed on April 19, see
889
+ Fig. 9. On April 21 the liquid recirculation was
890
+ started. The recirculation rates were 1.85 m3/h
891
+ in the West module and 2.25 m3/h in the East
892
+ module.
893
+ The cryogenic stabilization phase was com-
894
+ pleted around the end of May 2020. Pressures
895
+ and temperatures in the two modules were stable
896
+ and no cold spots were observed on the exter-
897
+ nal surface of the Warm Vessel.
898
+ At the start
899
+ of the cryogenic commissioning, all activities in
900
+ the detector building not related to cryogenics
901
+ were suspended and the building was put in a
902
+ high safety condition, with strong limitations to
903
+ the presence of people onsite. At the end of the
904
+ liquid argon filling, after the final pressure test,
905
+ the standard safety conditions were restored and
906
+ regular activities on top of the detector could be
907
+ restarted to complete the installation and test of
908
+ all sub-detectors.
909
+ During the cryogenic commissioning, there
910
+ were several activities both related to monitoring
911
+ the status of the detectors (wire chambers, wires
912
+ readout electronics, PMTs, CRT) and to develop-
913
+ ments for the following detector commissioning
914
+ phases. Noise data have been continuously taken
915
+ of wire readout electronics, PMTs and CRT. Ef-
916
+ fects on the noise from the activation of the cryo-
917
+ genic plant have been continuously monitored.
918
+ Functionality and stress tests of the DAQ were
919
+ conducted with several useful results.
920
+ The detector activation took place on Au-
921
+ gust 27, 2020 when the TPC wire planes and the
922
+ cathode high voltage (HV) were taken to nomi-
923
+ nal voltages. HV has remained stable at −75 kV.
924
+ Significant currents were found only on a few
925
+ wire bias and were addressed. All PMTs were
926
+ switched on and calibrated with the laser system.
927
+ Cosmic-ray interaction events were initially
928
+ collected with a random 5 Hz trigger and data
929
+ analyzed for calibration purposes (i.e. electron
930
+ – 11 –
931
+
932
+ DUMMY
933
+ Genie29Figure 9. Trend of the liquid argon level inside the two ICARUS cryostats during the filling phase.
934
+ lifetime, space charge, drift velocity measure-
935
+ ments).
936
+ Dedicated runs were also carried out
937
+ for specific commissioning tasks, such as inves-
938
+ tigation of TPC noise, PMT calibration with the
939
+ laser system, DAQ upgrades/longevity tests, etc.
940
+ One of the first measurements carried out
941
+ was the free electron lifetime 𝜏𝑒𝑙𝑒. This parame-
942
+ ter is fundamental for the monitoring of the liquid
943
+ argon condition in the TPCs and to obtain the
944
+ precise measurement of the energy deposition
945
+ from the ionization charge signal in the collected
946
+ events. The LAr purity is continuously moni-
947
+ tored by measuring the charge attenuation along
948
+ the drift path of the electron ionization signals
949
+ generated by cosmic ray tracks crossing the de-
950
+ tector. A fast procedure has been setup starting
951
+ from the method developed and used during the
952
+ Gran Sasso run [11]; it has been applied to the
953
+ recorded data since the detector activation.
954
+ The 𝜏𝑒𝑙𝑒 measurement is based on a simpli-
955
+ fied identification of the wire signals in the Col-
956
+ lection plane and of the anode to cathode cross-
957
+ ing muon tracks that have no indication of asso-
958
+ ciated 𝛿-rays or electromagnetic showers along
959
+ the track. It is used to provide a fast, real time,
960
+ measurement within 5-10% precision dominated
961
+ mostly by effects related to space charge and to
962
+ the electron diffusion, see Fig. 10.
963
+ The steady state values of 𝜏𝑒𝑙𝑒, exceeding
964
+ 3 ms in both cryostats, are high enough to al-
965
+ low for efficient detection and reconstruction of
966
+ ionizing events inside the active volume.
967
+ 6.1
968
+ TPC commissioning
969
+ After the TPC wires were biased and the cath-
970
+ ode HV was raised to nominal operating condi-
971
+ tions, the TPC commissioning began. With the
972
+ liquid argon at a sufficient level of purity, cos-
973
+ mogenic activity in the detector can be used to
974
+ study the detector response to ionization signals
975
+ in the TPC. To characterize the performance of
976
+ the ICARUS TPC, a variety of measurements
977
+ were taken between August 2020 and May 2022
978
+ as summarized below.
979
+ Noise levels in the TPC can be measured us-
980
+ ing the RMS of waveforms from the TPC read-
981
+ out, with an equivalent noise charge (ENC) of
982
+ roughly 550 e−/ADC [21]. Measured TPC noise
983
+ levels at ICARUS are shown in Fig. 11, both
984
+ before and after the filtering of coherent noise,
985
+ which was performed across sets of 64 channels
986
+ associated with the same front-end electronics
987
+ board.
988
+ – 12 –
989
+
990
+ 4
991
+ Start Gas and Liquid re-circulation
992
+ 3,5
993
+ 3
994
+ 2, 5
995
+ EAST Module
996
+ 2
997
+ WESTModule
998
+ 1,5
999
+ 1
1000
+ 0,5
1001
+ 0
1002
+ 0,00
1003
+ 200,00
1004
+ 400,00
1005
+ 600,00
1006
+ 800,00
1007
+ 1000,00
1008
+ 1200,00
1009
+ [400,00
1010
+ 1600,00
1011
+ Feb 19 : Filling Start
1012
+ Elapsed Time (hours)
1013
+ Apr 19 : Filling CompleteFigure 10. Trend of the drift electron lifetime in the two ICARUS cryostats during the commissioning phase.
1014
+ The sharp decreases of the lifetime are due to programmed interventions on the LAr recirculation pumps or
1015
+ on the cryogenic system. The lifetime is quickly recovered after the end of the interventions.
1016
+ Waveforms containing ionization signals are
1017
+ identified by simply applying a threshold and re-
1018
+ moving from consideration to ensure there is no
1019
+ bias to the noise measurements. The measure-
1020
+ ments were repeated with the cathode HV off
1021
+ and consistent results were obtained, validating
1022
+ the ionization signal identification methodology
1023
+ and indicating that a negligible amount of TPC
1024
+ noise is caused by interference from the cathode
1025
+ HV system.
1026
+ The noise levels after coherent noise filter-
1027
+ ing shown in Fig. 11 are consistent with previous
1028
+ noise measurements of the TPC electronics in a
1029
+ test setup [21].
1030
+ Fast Fourier transforms (FFTs) of the same
1031
+ noise waveforms used in the results shown in
1032
+ Fig. 11 are calculated for each of the three
1033
+ wire planes and averaged across the entire de-
1034
+ tector;
1035
+ these results are shown in Fig. 12.
1036
+ FFTs are shown both before and after coherent
1037
+ noise removal, showing the expected approxi-
1038
+ mate Rayleigh distribution of the intrinsic noise
1039
+ spectrum [22] on all three planes after coherent
1040
+ noise is removed. This provides strong evidence
1041
+ of extrinsic noise being almost completely re-
1042
+ moved from the TPC waveform data by the noise
1043
+ filtering algorithm.
1044
+ The Induction 2 plane and Collection plane
1045
+ spectra show a similar normalization, which is
1046
+ expected given the same length of the wires of
1047
+ these two planes. The Induction 1 plane spec-
1048
+ trum has instead a larger normalization given the
1049
+ longer wires and thus a higher capacitance, in-
1050
+ creasing the intrinsic noise levels. Further work
1051
+ is being carried out to understand the source of
1052
+ the coherent noise.
1053
+ In runs with sufficiently high electron life-
1054
+ time (most runs after the very beginning of
1055
+ commissioning in 2020),
1056
+ ionization signals
1057
+ from anode-cathode-crossing cosmic muons are
1058
+ used to evaluate the peak signal-to-noise ratio
1059
+ (PSNR) for minimum-ionizing particles (MIPs)
1060
+ in the TPC. Anode-cathode-crossing cosmic
1061
+ muon tracks traverse the full drift length of the
1062
+ detector and therefore allow for knowledge of the
1063
+ drift coordinate of each ionization signal along
1064
+ the track. Fig. 13 shows the PSNR of ioniza-
1065
+ tion signals for each plane using a large sample
1066
+ of cosmic muons in ICARUS data with coherent
1067
+ noise removed.
1068
+ In this study, the peak signal
1069
+ (numerator in the ratio) is defined as the maxi-
1070
+ mum signal ADC value minus the baseline ADC
1071
+ value for the unipolar signals of the Collection
1072
+ plane and the absolute value of the maximum sig-
1073
+ – 13 –
1074
+
1075
+ Lifetime [ms]
1076
+ West
1077
+ L
1078
+ East
1079
+
1080
+ 30/Sep
1081
+ 31/Dec
1082
+ 01/Apr
1083
+ 01/Jul
1084
+ 01/Oct
1085
+ 31/Dec
1086
+ 01/Apr
1087
+ 2020
1088
+ 2020
1089
+ 2021
1090
+ 2021
1091
+ 2021
1092
+ 2021
1093
+ 2022
1094
+ DateFigure 11. TPC noise levels at ICARUS before and after filtering of coherent noise, as measured by waveform
1095
+ RMS in ADC counts (with ENC of roughly 550 e−/ADC [21]). Results are shown separately for the Induction
1096
+ 1 plane (left), Induction 2 plane (center), and Collection plane (right). Mean values of the shown distributions
1097
+ are presented at the bottom of each figure.
1098
+ 0.0
1099
+ 0.2
1100
+ 0.4
1101
+ 0.6
1102
+ 0.8
1103
+ 1.0
1104
+ 1.2
1105
+ Induction 1
1106
+ Raw Spectra
1107
+ Noise-Filtered Spectra
1108
+ 0.0
1109
+ 0.2
1110
+ 0.4
1111
+ 0.6
1112
+ 0.8
1113
+ 1.0
1114
+ 1.2
1115
+ Induction 2
1116
+ 0
1117
+ 100
1118
+ 200
1119
+ 300
1120
+ 400
1121
+ 500
1122
+ 0.0
1123
+ 0.2
1124
+ 0.4
1125
+ 0.6
1126
+ 0.8
1127
+ 1.0
1128
+ 1.2
1129
+ Collection
1130
+ Frequency [kHz]
1131
+ Power [ADC2/kHz]
1132
+ Figure 12. Fast Fourier transforms (FFTs) of noise
1133
+ waveform data collected by the ICARUS TPCs, be-
1134
+ fore and after filtering of coherent noise. Results are
1135
+ shown separately for the Induction 1 plane (top), In-
1136
+ duction 2 plane (middle), and Collection plane (bot-
1137
+ tom).
1138
+ nal ADC value minus the minimum signal ADC
1139
+ value for the bipolar signals of the two induc-
1140
+ tion planes. The noise level (denominator in the
1141
+ ratio) is the RMS of signal-removed waveforms
1142
+ from the same TPC channel in units of ADCs,
1143
+ as shown in Fig. 11. Cosmic muon tracks used
1144
+ in the PSNR measurement are required to be
1145
+ oriented at an angle of 20 degrees or less with
1146
+ respect to the anode plane, and have a "3D pitch"
1147
+ (track segment length corresponding to the ion-
1148
+ ization signal from a single wire) of 4 mm or less
1149
+ for the wire plane of interest. These selection
1150
+ criteria probe the phase space most relevant for
1151
+ beam neutrinos interacting in the detector, which
1152
+ have interaction products that travel mainly in the
1153
+ forward direction. Furthermore, only parts of the
1154
+ track within 2 cm to 10 cm of the anode are used
1155
+ in order to minimize impact from charge attenua-
1156
+ tion due to impurities in the liquid argon. Fig. 13
1157
+ illustrates the performance of the TPC.
1158
+ The detector enables robust identification of
1159
+ ionization signals embedded within electronics
1160
+ noise background, with more than 99% of the
1161
+ MIP ionization signals having a PSNR greater
1162
+ than four.
1163
+ Anode-cathode-crossing
1164
+ cosmic
1165
+ muon
1166
+ tracks are also used to make a measurement
1167
+ of ionization drift velocity in the detector.
1168
+ The distance between the anode and cathode,
1169
+ 148.2 cm, is divided by the maximum ionization
1170
+ drift time, or the difference in time between
1171
+ the first and last ionization signals associated
1172
+ with the cosmic muon tracks.
1173
+ The latter
1174
+ measurement should yield the time it takes for
1175
+ ionization to drift from the cathode (one end of
1176
+ – 14 –
1177
+
1178
+ Average Noise by Plane
1179
+ Full Noise
1180
+ Noise-Filtered
1181
+ Induction 2
1182
+ Collection
1183
+ Induction 1
1184
+ 2.5
1185
+ 2.0
1186
+ Units
1187
+ 1.5
1188
+ Arbitrary l
1189
+ 1.0
1190
+ 0.5
1191
+ 0.0
1192
+ 0.0
1193
+ 2.5
1194
+ 5.0
1195
+ 7.5
1196
+ 10.0
1197
+ 0.0
1198
+ 2.5
1199
+ 5.0
1200
+ 7.5
1201
+ 10.0
1202
+ 0.0
1203
+ 2.5
1204
+ 5.0
1205
+ 7.5
1206
+ 10.0
1207
+ RMS[ADC]
1208
+ RMS [ADC]
1209
+ RMS [ADC]
1210
+ μ: 3.80 ADC
1211
+ 6.02 ADC
1212
+ μ: 2.51 ADC
1213
+ 3.65 ADC
1214
+ μ: 2.53 ADC
1215
+ 3.47 ADC0
1216
+ 5
1217
+ 10
1218
+ 15
1219
+ 20
1220
+ 25
1221
+ 30
1222
+ 35
1223
+ 40
1224
+ Peak Signal-to-Noise Ratio (Noise-Filtered)
1225
+ 0.00
1226
+ 0.05
1227
+ 0.10
1228
+ 0.15
1229
+ 0.20
1230
+ 0.25
1231
+ 0.30
1232
+ 0.35
1233
+ 0.40
1234
+ 0.45
1235
+ Arbitrary Units
1236
+ Induction 1 (Peak: 6.74)
1237
+ Induction 2 (Peak: 8.64)
1238
+ Collection (Peak: 9.05)
1239
+ Figure 13. Peak signal-to-noise ratio (PSNR) of ion-
1240
+ ization signals for each of the three TPC wire planes
1241
+ using cosmic muons in ICARUS data. Coherent noise
1242
+ is removed from the TPC waveforms prior to iden-
1243
+ tification and measurement of the ionization signal
1244
+ amplitude. See text for details on the cosmic muon
1245
+ data selection.
1246
+ the track) to the anode (other end of the track),
1247
+ so the ratio should provide the drift velocity of
1248
+ the ionization electrons in liquid argon at the
1249
+ nominal drift electric field of roughly 500 V/cm
1250
+ and temperature of roughly 87.5 K.
1251
+ A correction is made to account for a small
1252
+ bias in precisely reconstructing the drift times as-
1253
+ sociated with the track end points, derived from
1254
+ Monte Carlo simulation. A Crystal Ball func-
1255
+ tion1 is then fit to the maximum ionization drift
1256
+ time distribution associated with cosmic muon
1257
+ tracks in each TPC volume (two per cryostat),
1258
+ with the peak value of each fit used in the ion-
1259
+ ization drift velocity calculation. The results of
1260
+ the ionization drift velocity measurements in the
1261
+ west cryostat are shown in Fig. 14. The results
1262
+ of the measurements, roughly 0.1572 cm/µs for
1263
+ both TPC volumes in the west cryostat, agree
1264
+ with the predicted value of 0.1576 cm/µs to
1265
+ within 0.3% [23, 24].
1266
+ 1The Crystal Ball function, named after the Crystal
1267
+ Ball Collaboration, is a probability density function com-
1268
+ monly used to model various lossy processes in high-energy
1269
+ physics. It consists of a Gaussian core portion and a power-
1270
+ law low-end tail, below a certain threshold.
1271
+ 880
1272
+ 900
1273
+ 920
1274
+ 940
1275
+ 960
1276
+ 980
1277
+ 1000
1278
+ Maximum Ionization Drift Time [ s]
1279
+ 0
1280
+ 5000
1281
+ 10000
1282
+ 15000
1283
+ Tracks
1284
+ TPC E Drift V:
1285
+ 0.1572 cm/ s
1286
+ TPC W Drift V:
1287
+ 0.1572 cm/ s
1288
+ Ionization Drift Velocity: West Cryostat
1289
+ TPC E Fit
1290
+ TPC E Data
1291
+ TPC W Fit
1292
+ TPC W Data
1293
+ Figure 14.
1294
+ Results of the ionization drift veloc-
1295
+ ity measurement using ICARUS cosmic muon data.
1296
+ Shown are Crystal Ball fits to the maximum ioniza-
1297
+ tion drift time distributions associated with anode-
1298
+ cathode-crossing cosmic muons in the two TPCs in
1299
+ the west cryostat.
1300
+ Electric field distortions in near-surface
1301
+ LAr-TPCs can arise due to the accumulation
1302
+ of space charge, i.e.
1303
+ slow-moving positively-
1304
+ charged argon ions originating from cosmic
1305
+ muon ionization within the detector [25]. These
1306
+ argon ions, which drift slowly toward the cath-
1307
+ ode at a drift velocity of several millimeters per
1308
+ second at a drift electric field of 500 V/cm [24],
1309
+ linger around long enough to create substantial
1310
+ electric field distortions that pull ionization elec-
1311
+ trons toward the middle of the TPC volume as
1312
+ they drift toward the anode. These electric field
1313
+ distortions lead to biases in reconstructing the
1314
+ point of origin of ionization within the detector,
1315
+ a secondary effect referred to as "spatial distor-
1316
+ tions" in LAr-TPC detectors; collectively, these
1317
+ two related distortions are referred to as space
1318
+ charge effects (SCE).
1319
+ Using
1320
+ anode-cathode-crossing
1321
+ cosmic
1322
+ muon tracks, the magnitude of SCE in the
1323
+ ICARUS detector is estimated by utilizing
1324
+ methodology developed to measure SCE in
1325
+ previous near-surface running of the ICARUS
1326
+ detector [26]. The results of measurements in
1327
+ the two TPC volumes of the west cryostat are
1328
+ shown in Fig. 15, where they are compared
1329
+ – 15 –
1330
+
1331
+ to a calculation of SCE [24] used in ICARUS
1332
+ Monte Carlo simulations prior to measuring
1333
+ the magnitude of SCE in ICARUS data.
1334
+ The
1335
+ magnitude of SCE is observed to be very similar
1336
+ in the two TPC volumes, though underestimated
1337
+ by roughly 30% in simulation.
1338
+ 0
1339
+ 20
1340
+ 40
1341
+ 60
1342
+ 80
1343
+ 100
1344
+ 120
1345
+ 140
1346
+ Drift Coordinate X [cm]
1347
+ 0
1348
+ 0.1
1349
+ 0.2
1350
+ 0.3
1351
+ 0.4
1352
+ 0.5
1353
+ 0.6
1354
+ 0.7
1355
+ 0.8
1356
+ X [cm]
1357
+
1358
+ Drift Direction Spatial Offset
1359
+ WE TPC Data
1360
+ WW TPC Data
1361
+ Calculation for Simulation
1362
+ X vs. X
1363
+
1364
+ Data SCE Comparison:
1365
+ Figure 15. Measured spatial offsets in the drift di-
1366
+ rection as a function of ionization drift distance for
1367
+ the two TPCs in the west cryostat, evaluated us-
1368
+ ing anode-cathode-crossing cosmic muon tracks in
1369
+ ICARUS data. The results are compared with pre-
1370
+ dictions of spatial distortions from a calculation of
1371
+ space charge effects (SCE) presently used in ICARUS
1372
+ Monte Carlo simulations (to be updated with data-
1373
+ driven SCE measurement).
1374
+ The energy scale of MIPs can be probed
1375
+ with cosmic muons that stop in the ICARUS de-
1376
+ tector, as done in similar calibrations performed
1377
+ at other LAr-TPC neutrino experiments [27].
1378
+ The known profile of muon energy loss per unit
1379
+ length (𝑑𝐸/𝑑𝑥) in liquid argon as a function of
1380
+ kinetic energy [28] can be used to predict the
1381
+ value of 𝑑𝐸/𝑑𝑥 versus residual range, the dis-
1382
+ tance from the end of a stopped muon track
1383
+ in reconstructed TPC data.
1384
+ After accounting
1385
+ for prompt electron-ion recombination [29] and
1386
+ charge attenuation during ionization drift due to
1387
+ electro-negative impurities in the detector, one
1388
+ can compare the most-probable value (MPV) of
1389
+ 𝑑𝐸/𝑑𝑥 versus residual range from a sample of
1390
+ stopping muons in ICARUS data (evaluated by
1391
+ fitting the data with a Landau distribution con-
1392
+ volved with a Gaussian, performed in bins of
1393
+ residual range) to the MPV 𝑑𝐸/𝑑𝑥 curve ex-
1394
+ pected from theory.
1395
+ The result of the Collection plane energy
1396
+ scale calibration for the east TPC of the west
1397
+ cryostat is shown in Fig. 16 (left). Good agree-
1398
+ ment between calibrated data and predictions
1399
+ from theory is found for all values of stopping
1400
+ muon residual range after this calibration has
1401
+ been performed, with sub-percent agreement for
1402
+ values of 𝑑𝐸/𝑑𝑥 < 4 MeV/cm; similar levels of
1403
+ agreement are observed for the other three TPCs
1404
+ as well. Additionally, the energy scale calibra-
1405
+ tion is further scrutinized by comparing two dif-
1406
+ ferent methods of stopping muon kinetic energy
1407
+ reconstruction: one by calorimetry (summing up
1408
+ charge associated with energy deposition along
1409
+ the track), 𝐸calo, and another by range (convert-
1410
+ ing distance from end of stopping muon track
1411
+ to kinetic energy by use of a look-up table [28]),
1412
+ 𝐸range. The result of this cross-check is presented
1413
+ in Fig. 16 (right), showing little bias between the
1414
+ two methods for stopping muons in ICARUS cos-
1415
+ mic muon data after the energy scale calibration
1416
+ is applied.
1417
+ Future measurements will include
1418
+ protons from ICARUS data, allowing for probing
1419
+ of the energy scale of highly-ionizing particles
1420
+ in the detector.
1421
+ 6.2
1422
+ PMT commissioning
1423
+ The whole light detection system was tested at
1424
+ Fermilab before the cooling of the detector, once
1425
+ the dark condition inside the cryostats was guar-
1426
+ anteed. A total of 357 (out of 360) PMTs were
1427
+ found to be working with performances consis-
1428
+ tent with the tests performed at CERN [16]. The
1429
+ same number of working PMTs were found af-
1430
+ ter the filling of the detector with liquid argon,
1431
+ demonstrating the ability of this PMT model to
1432
+ withstand low temperatures.
1433
+ A PMT signal, recorded by the light detec-
1434
+ tion system electronics, is shown in Fig. 17. A
1435
+ gain calibration/equalization campaign was car-
1436
+ – 16 –
1437
+
1438
+ 0
1439
+ 20
1440
+ 40
1441
+ 60
1442
+ 80
1443
+ 100
1444
+ Residual Range [cm]
1445
+ 1
1446
+ 2
1447
+ 3
1448
+ 4
1449
+ 5
1450
+ 6
1451
+ Calibrated dE/dx [MeV/cm]
1452
+ Predicted MPV dE/dx
1453
+ 0.4
1454
+ 0.2
1455
+ 0.0
1456
+ 0.2
1457
+ 0.4
1458
+ (Ecalo
1459
+ Erange) / Erange
1460
+ 0
1461
+ 5000
1462
+ 10000
1463
+ 15000
1464
+ 20000
1465
+ 25000
1466
+ 30000
1467
+ Tracks
1468
+ 1: 4.7%
1469
+ 1: 0.3%
1470
+ 2: 14.5%
1471
+ 2: 0.8%
1472
+ Figure 16. Calibrated Collection plane 𝑑𝐸/𝑑𝑥 as a function of residual range for a selection of stopping muons
1473
+ in ICARUS cosmic muon data, including a comparison to the most-probable value (MPV) of 𝑑𝐸/𝑑𝑥 from
1474
+ stopping muons predicted from theory [28] (left); comparison of cosmic muon kinetic energy reconstruction
1475
+ by calorimetry, 𝐸calo, and by range, 𝐸range, showing little bias between the two methods for stopping muons
1476
+ in ICARUS cosmic muon data after the energy scale calibration is applied (right).
1477
+ ried out during the PMT commissioning. At first,
1478
+ external fast laser pulses focused on each PMT
1479
+ window by means of dedicated optical fibers
1480
+ were used to obtain a coarse gain curve for each
1481
+ PMT as a function of the applied voltage around
1482
+ the expected values. Laser pulses were also used
1483
+ to characterize, to within 1 ns precision, the delay
1484
+ response of each PMT channel, which can dif-
1485
+ fer due to different PMT and cable transit times.
1486
+ Voltages were set to values corresponding to a
1487
+ gain of 5 · 106, resulting in an equalization within
1488
+ 16%, as a first approximation.
1489
+ Fine tuning was carried out to improve the
1490
+ gain equalization by means of an automatic pro-
1491
+ cedure.
1492
+ To this purpose the response of each
1493
+ PMT to background single photons (≈ 250 kHz)
1494
+ was measured, and the voltages were adjusted
1495
+ according to the gain curves. This procedure led
1496
+ to a final equalization with a spread less than 1%,
1497
+ as shown in Fig. 18.
1498
+ 6.3
1499
+ CRT commissioning
1500
+ The side and top CRT modules were tested before
1501
+ the installation at ICARUS using a test stand. Af-
1502
+ ter the installation of all CRT modules, the cos-
1503
+ mic rate over time was obtained. The event rates
1504
+ Figure 17. PMT signal as recorded by the light de-
1505
+ tection system electronics.
1506
+ for each wall of the side CRT as a function of time
1507
+ are constant, as shown in Fig. 19. The higher
1508
+ rates on north wall (black) are due to the proxim-
1509
+ ity with the cryogenic pumps, with these mod-
1510
+ ules experiencing higher electrical noise rates in
1511
+ addition to cosmic rates on the surface. In addi-
1512
+ tion, the rates from the west north and east north
1513
+ walls are slightly higher from being closer to the
1514
+ cryogenics. Following work to characterize and
1515
+ mitigate the noise, electrical chokes (inductors)
1516
+ were installed along all Side CRT FEB power
1517
+ cables to reduce noise rates.
1518
+ Top CRT cosmic event rates before and after
1519
+ the installation of concrete overburden are shown
1520
+ – 17 –
1521
+
1522
+ 15000
1523
+ Amplitude (ADC Counts
1524
+ 14900
1525
+ 14800
1526
+ 14700
1527
+ 14600
1528
+ 14500
1529
+ 0
1530
+ 2
1531
+ 4
1532
+ 6
1533
+ 8
1534
+ 10
1535
+ Time (us)Figure 18. Gain distribution for 354 PMTs after the
1536
+ fine tuning equalization. The automatic procedure
1537
+ was not applied on 6 PMTs (not present in the plot)
1538
+ that were manually calibrated.
1539
+ in Fig. 20 for horizontal (left) and vertical (right)
1540
+ modules.
1541
+ Before the installation of the over-
1542
+ burden the mean rate was ∼ 610 Hz and 260 Hz
1543
+ for horizontal and vertical modules, respectively.
1544
+ After the installation of the overburden the rates
1545
+ reduced to 330 Hz and 180 Hz for horizontal and
1546
+ vertical modules, respectively. Except for varia-
1547
+ tion due to concrete blocks placement above the
1548
+ detector, the rates are stable on a time scale of
1549
+ months.
1550
+ 12/23/20
1551
+ 12/30/20
1552
+ 01/06/21
1553
+ Date
1554
+ 0
1555
+ 2
1556
+ 4
1557
+ 6
1558
+ 8
1559
+ 10
1560
+ 12
1561
+ 14
1562
+ 16
1563
+ 18
1564
+ 20
1565
+ Rate [kHz]
1566
+ North Wall
1567
+ West North Wall
1568
+ West Central Wall
1569
+ West South Wall
1570
+ East North Wall
1571
+ East Central Wall
1572
+ East South Wall
1573
+ Figure 19. Side CRT cosmic event rates as a function
1574
+ of time. The black points corresponds to the rates
1575
+ from the north side CRT wall, the pink and blue
1576
+ points corresponds to East and West north walls, and
1577
+ the remaining walls are at 1 kHz rate.
1578
+ 6.4
1579
+ Triggering on the BNB and NuMI neu-
1580
+ trinos
1581
+ The initial ICARUS trigger system exploits the
1582
+ coincidence of the BNB and NuMI beams spills,
1583
+ 1.6 µs and 9.6 µs respectively, with the prompt
1584
+ scintillation light detected by the PMT system in-
1585
+ stalled behind the wire planes of each TPC [30].
1586
+ The generation of the beam spill gates is
1587
+ based on receiving the “Early Warning” (EW)
1588
+ signals for BNB and NuMI beams, 35 and 730 ms
1589
+ in advance of protons on target, respectively.
1590
+ LVDS signals from the PMT digitizers, in terms
1591
+ of the OR signal of adjacent PMTs, are pro-
1592
+ cessed by programmable FPGA logic boards to
1593
+ implement trigger logic for the activation of the
1594
+ ICARUS read-out. Additional trigger signals are
1595
+ generated for calibration purposes in correspon-
1596
+ dence with a subset of the beam spills without
1597
+ any requirement on the scintillation light (Min-
1598
+ Bias trigger) and outside of the beam spills to
1599
+ detect cosmic ray interactions (Off-Beam trig-
1600
+ ger).
1601
+ To synchronize all detector subsystems’
1602
+ read-outs with the proton beam spill extraction
1603
+ at the level of few nanosecond accuracy, a White
1604
+ Rabbit (WR) network [31] has been deployed for
1605
+ distributing the beam extraction signals. An ab-
1606
+ solute GPS timing signal, in the form of PPS, is
1607
+ used as a reference for generating phase locked
1608
+ digitization clocks (62.5 MHz for the PMT and
1609
+ 2.5 MHz for the TPC) and for time-stamping
1610
+ the beam gates and trigger signals. In addition,
1611
+ the signals of Resistive Wall Monitor detectors
1612
+ (RWM) at 2 GHz sampling frequency are also
1613
+ recorded to precisely measure the timing and
1614
+ the bunched structure of protons on target, see
1615
+ Fig. 21.
1616
+ In the presence of a global trigger signal,
1617
+ 1.5 ms and 30 µs acquisition windows are acti-
1618
+ vated for the TPC and PMT signal recording,
1619
+ respectively. In addition, PMT waveforms are
1620
+ collected inside a 2 ms time window around the
1621
+ – 18 –
1622
+
1623
+ PMTs
1624
+ 90
1625
+ Entries
1626
+ 354
1627
+ Constant
1628
+ 94.57
1629
+ #
1630
+ 80
1631
+ Mean
1632
+ 0.4967
1633
+ 70
1634
+ Sigma
1635
+ 0.003574
1636
+ 60
1637
+ 50
1638
+ 40
1639
+ 30
1640
+ 20
1641
+ 10
1642
+ 0.45
1643
+ 0.46
1644
+ 0.47
1645
+ 0.48
1646
+ 0.49
1647
+ 0.5
1648
+ 0.510.520.530.540.55
1649
+ Gain [10' electrons]02/26/22
1650
+ 03/28/22
1651
+ 04/27/22
1652
+ 05/27/22
1653
+ Date
1654
+ 0.25
1655
+ 0.3
1656
+ 0.35
1657
+ 0.4
1658
+ 0.45
1659
+ 0.5
1660
+ 0.55
1661
+ 0.6
1662
+ 0.65
1663
+ 0.7
1664
+ Rate [kHz]
1665
+ FEB 172
1666
+ FEB 114
1667
+ FEB 100
1668
+ FEB 150
1669
+ FEB 238
1670
+ FEB 234
1671
+ FEB 238
1672
+ FEB 170
1673
+ FEB 101
1674
+ FEB 142
1675
+ FEB 6
1676
+ FEB 232
1677
+ FEB 237
1678
+ FEB 239
1679
+ 02/26/22
1680
+ 03/28/22
1681
+ 04/27/22
1682
+ 05/27/22
1683
+ Date
1684
+ 0
1685
+ 0.05
1686
+ 0.1
1687
+ 0.15
1688
+ 0.2
1689
+ 0.25
1690
+ 0.3
1691
+ Rate [kHz]
1692
+ FEB 81
1693
+ FEB 119
1694
+ FEB 87
1695
+ FEB 92
1696
+ FEB 180
1697
+ FEB 97
1698
+ FEB 174
1699
+ FEB 189
1700
+ FEB 190
1701
+ Figure 20. Cosmic ray rates as a function of time for a set of Top CRT horizontal (left) and vertical (right)
1702
+ modules. Numbers in the legend indicate the module’s Front End Board and the black dot lines indicate the
1703
+ beginning and the end of 3 m overburden installation over the displayed modules: the rates reduced from
1704
+ ∼ 610 (260) Hz before to 330 (180) Hz after the installation of the overburden for the horizontal (vertical)
1705
+ modules.
1706
+ Figure 21.
1707
+ Layout of the trigger system.
1708
+ SPEXI
1709
+ board: synchronizes the whole ICARUS detector,
1710
+ generates clocks and readout signals, handles beam
1711
+ extraction messages; 7820 FPGA boards: generate a
1712
+ Global Trigger in coincidence with beam extraction
1713
+ (Early Warning) on the basis of selected PMT sig-
1714
+ nal majorities to recognize an event interaction in the
1715
+ LAr, to start the PMT activity recording; RT Con-
1716
+ troller implements all the features for communication
1717
+ with DAQ.
1718
+ beam spill to record all cosmic muons crossing
1719
+ the ICARUS TPCs during the electron drift time.
1720
+ The timing of the beam spills was first ap-
1721
+ proximately determined by measuring with an
1722
+ oscilloscope the difference between the EW sig-
1723
+ nals arrival time and the actual proton extraction
1724
+ signal by RWM counters at the target. Then neu-
1725
+ trino interactions were identified and associated
1726
+ with the muons of the beam spill in excess to
1727
+ cosmic rays that were clearly identified inside
1728
+ the time profile of the scintillation light signals
1729
+ (flashes) by requiring at least 5 fired PMT pairs
1730
+ in the left and right TPC (Fig. 22).
1731
+ Due to the energy range of BNB and NuMI
1732
+ neutrino beams, neutrino interactions are ex-
1733
+ pected to be contained in an ∼ 4 m section of
1734
+ ICARUS along the beam direction, suggesting
1735
+ the implementation of a trigger logic based on
1736
+ the recognition of fired PMTs inside a limited
1737
+ TPC region. The logic for processing the PMT
1738
+ LVDS signals has been initially determined with
1739
+ Monte Carlo calculations, and then it has been
1740
+ refined by analyzing a sample of events collected
1741
+ with a beam spill signal only (Min-Bias trigger),
1742
+ i.e. without any requirement on the scintillation
1743
+ light. The 18-m long TPC walls have been sub-
1744
+ divided in 3 consecutive longitudinal slices of
1745
+ 6-m length including 30 PMTs each. In each of
1746
+ opposite facing slices a majority of 5 LVDS sig-
1747
+ – 19 –
1748
+
1749
+ WhiteRabbitnetwork
1750
+ PMT
1751
+ PMT
1752
+ CPU W RT
1753
+ TRIG
1754
+ GLOBAL
1755
+ TRIG
1756
+ controller
1757
+ TRIG
1758
+ SPEXI
1759
+ EAST
1760
+ WEST
1761
+ PXle8135
1762
+ 7820R
1763
+ 7820R
1764
+ 7820R
1765
+ TPC
1766
+ TPC
1767
+ A2795's
1768
+ A2795's
1769
+ PMT
1770
+ PMT
1771
+ TPC WIRES
1772
+ TPC WIRES
1773
+ V1730B's
1774
+ V1730B's
1775
+ T300 E
1776
+ T300 E
1777
+ PMT DAQ
1778
+ TPC DAQ
1779
+ PMT's
1780
+ PMT's
1781
+ T300 E
1782
+ T300 W
1783
+ ENABLEGATE(2mS
1784
+ Trigger
1785
+ BEAMGATE[1.6uS,9.8ys]
1786
+ laptop
1787
+ GlobalTriggerOutput cryostat1-2
1788
+ Central
1789
+ PMT Trigger cryostat 1-2
1790
+ DAQ
1791
+ ALLBus LinesFigure 22. Time distribution of the recorded PMT light flashes (≥ 5 fired PMT pairs in the left and right
1792
+ TPCs within 150 ns): the beam event excess is observed for BNB (left) and NuMI beam (right). The 1.6 µs
1793
+ and 9.6 µs spills duration of the beams are well recognized.
1794
+ nals, with 8 photo-electron (phe) discrimination
1795
+ threshold and an OR of two adjacent PMTs, has
1796
+ been required to produce a PMT trigger primi-
1797
+ tive signal. The same logic with a majority of
1798
+ 10 LVDS PMT signals is applied to generate a
1799
+ PMT trigger primitive in time period prior to and
1800
+ after a beam spill. This trigger provides collec-
1801
+ tion of data sampling the 15 kHz of cosmic rays
1802
+ crossing the detector during the drift time.
1803
+ With trigger gates of duration 4 ms and 14
1804
+ ms for BNB and NuMI, respectively, a trigger
1805
+ rate of ∼ 0.7 Hz has been obtained (0.3 and 0.15
1806
+ Hz from the BNB and NuMI components, re-
1807
+ spectively, and 0.25 Hz for the Off-Beam). This
1808
+ is in a manageable data read-out bandwidth with
1809
+ good operational stability. The trigger efficiency
1810
+ for neutrino interactions is under study with data;
1811
+ expectations based on the Monte Carlo simula-
1812
+ tions indicate a > 90% efficiency for neutrino CC
1813
+ interactions with >100 MeV energy deposition.
1814
+ 6.5
1815
+ DAQ implementation
1816
+ The ICARUS data acquisition (DAQ) system uti-
1817
+ lizes the general artdaq data acquisition software
1818
+ development toolkit [32], providing customiz-
1819
+ able applications for reading data from detector
1820
+ elements (BoardReaders), and configurable ap-
1821
+ plications for performing event-building, data-
1822
+ logging, and data-dispatch to downstream online
1823
+ data quality monitoring processes.
1824
+ Customized BoardReaders acquire data
1825
+ fragments from the TPC, PMT, and CRT read-
1826
+ out electronics, and from the trigger and White
1827
+ Rabbit timing systems.
1828
+ They then assign ap-
1829
+ propriate event counters and timestamps to each
1830
+ fragment and then queue that data for transfer
1831
+ to a configurable number of EventBuilder appli-
1832
+ cations. For each triggered event, the ICARUS
1833
+ trigger BoardReader sends its data fragment to
1834
+ an EventBuilder, triggering a request for data
1835
+ from all other configured BoardReaders in the
1836
+ DAQ system. Events are written using the art
1837
+ event-processing framework [33]. Data are writ-
1838
+ ten on separate file streams using simple filters
1839
+ on trigger type. Each event in ICARUS, after
1840
+ lossless data compression, is approximately 160
1841
+ MB, with the majority of data corresponding to
1842
+ the TPCs. The DAQ system is capable of stably
1843
+ supporting trigger rates in excess of 5 Hz, though
1844
+ typical operational trigger rates are of roughly 1
1845
+ Hz or below.
1846
+ The BoardReader for the trigger system
1847
+ sends a single fragment containing the trigger
1848
+ and beam-gate timing, the type of beam gate,
1849
+ a global trigger counter, and a counter for the
1850
+ number of beam gates of each type in that DAQ
1851
+ – 20 –
1852
+
1853
+ BnB
1854
+ NuM
1855
+ 4
1856
+ Background
1857
+ .50 Ms
1858
+ 2140
1859
+ .
1860
+ 425
1861
+ Beam gate:1.6 μs
1862
+ 400
1863
+ Data
1864
+ 375
1865
+ lashes
1866
+ lashes
1867
+ 100
1868
+ 329
1869
+ Background
1870
+ DL
1871
+ Beamgate:9.5μs
1872
+ 275
1873
+ Data
1874
+ Pmt flash start tinmne [us]
1875
+ PMT flash start tirme Lus]run.
1876
+ The global trigger counter and time are
1877
+ used for collection of data from other subsys-
1878
+ tems; the latter derives from the common White
1879
+ Rabbit timing system, and is checked for validity
1880
+ against the network protocol time of the trigger
1881
+ BoardReader server. The number of beam gates
1882
+ of each type in the run is used offline for proper
1883
+ accounting of the total number of POT and de-
1884
+ tector exposure within a run.
1885
+ In order to handle large data volumes stored
1886
+ on tape, the Fermilab based SAM (Serial Ac-
1887
+ cess to Metadata) system is exploited. For this
1888
+ purpose, a set of metadata is associated to each
1889
+ data file using Python scripts. The metadata al-
1890
+ low users to create large data sets for the analysis
1891
+ by requiring matching with data’s relevant infor-
1892
+ mation such as run number, data type (raw or
1893
+ reconstructed), run configuration, date, etc.
1894
+ 6.6
1895
+ First operations with the BNB and
1896
+ NuMI
1897
+ The ICARUS-T600 detector was first fully op-
1898
+ erational in June 2021 before the summer shut-
1899
+ down. It restarted data collection when beam
1900
+ returned November 5, 2021. Figure 23 shows
1901
+ the amounts of POT delivered by the accelerator
1902
+ and collected by the detector during its commis-
1903
+ sioning phase, concluded in June 2022, for a
1904
+ total of 296 · 1018 and 503 · 1018 POT collected
1905
+ for BNB and NuMI, respectively. Beam utiliza-
1906
+ tion - defined as the amount of POT collected
1907
+ divided by the delivered - of 89% for BNB and
1908
+ 88% for NuMI. In Fig. 23, daily variations of the
1909
+ beam utilization are also visible: periods with
1910
+ low utilization (less than 60%) correspond to
1911
+ days where the data acquisition was suspended
1912
+ in order to proceed with detector commission-
1913
+ ing activities.
1914
+ Apart from this, the utilization
1915
+ is an average over 91% per day for both beams,
1916
+ which corresponds to a downtime of less than
1917
+ two hours per day. The most frequent causes of
1918
+ operation downtime are data acquisition issues
1919
+ and less commonly hardware problems.
1920
+ The
1921
+ detector and data collection status are continu-
1922
+ ously supervised with fully-remote shifts staffed
1923
+ by collaborators and with the support of on-call
1924
+ experts for each of the main detector subsystems.
1925
+ 7
1926
+ Observation and reconstruction of
1927
+ neutrino events
1928
+ The data collected by the detector are processed
1929
+ by offline software to obtain information neces-
1930
+ sary for reconstruction and analysis of events.
1931
+ The procedure to reconstruct the TPC wire and
1932
+ PMT signals is briefly described in the following
1933
+ Sec. 7.1, 7.2 and 7.3.
1934
+ The detector behavior was first investigated
1935
+ by a visual selection of neutrino interactions in
1936
+ the active liquid argon, as described in Sec. 7.4.
1937
+ These sample were an important component of
1938
+ the development and validation of an automated
1939
+ event selection scheme.
1940
+ 7.1
1941
+ Wire signal reconstruction
1942
+ The ICARUS wire signal processing chain fol-
1943
+ lows a logic similar to other LAr-TPC experi-
1944
+ ments, based on the deconvolution of the wire
1945
+ signal waveform. This procedure, explained in
1946
+ more detail in [34], has the goal to recover the
1947
+ original time structure of the current of drift
1948
+ electrons generating the signal on each wire, up-
1949
+ stream of the distortions produced by the electric
1950
+ field in the wire region and the shaping by the
1951
+ front-end electronics.
1952
+ Mathematically, this is
1953
+ obtained by inverting the response functions de-
1954
+ scribing both the electric field and the electronics
1955
+ effects; the resulting deconvolved signal shape is
1956
+ approximately Gaussian for all wire planes.
1957
+ After the removal of the coherent noise (de-
1958
+ scribed in 6.1), the deconvolution is performed
1959
+ on each wire waveform.
1960
+ Segments of wave-
1961
+ forms corresponding to physical signals (hits) are
1962
+ searched for in the deconvolved waveform with a
1963
+ threshold-based hit finding algorithm. Each hit
1964
+ – 21 –
1965
+
1966
+ 06-01
1967
+ 2021
1968
+ 11-19
1969
+ 2021
1970
+ 12-29
1971
+ 2021
1972
+ 02-14
1973
+ 2022
1974
+ 03-27
1975
+ 2022
1976
+ 05-06
1977
+ 2022
1978
+ 20
1979
+ 40
1980
+ 60
1981
+ 80
1982
+ 100
1983
+ Beam utilization [%]
1984
+ 06-01
1985
+ 2021
1986
+ 11-19
1987
+ 2021
1988
+ 12-29
1989
+ 2021
1990
+ 02-14
1991
+ 2022
1992
+ 03-27
1993
+ 2022
1994
+ 05-06
1995
+ 2022
1996
+ 20
1997
+ 40
1998
+ 60
1999
+ 80
2000
+ 100
2001
+ Beam utilization [%]
2002
+ 0
2003
+ 50
2004
+ 100
2005
+ 150
2006
+ 200
2007
+ 250
2008
+ 300
2009
+ POT (1018)
2010
+ BNB
2011
+ Delivered: 334.2 1018 POT
2012
+ Collected: 296.1 1018 POT
2013
+ 0
2014
+ 100
2015
+ 200
2016
+ 300
2017
+ 400
2018
+ 500
2019
+ POT (1018)
2020
+ NuMI
2021
+ Delivered: 573.6 1018 POT
2022
+ Collected: 503.1 1018 POT
2023
+ Figure 23. Cumulative sum of POT delivered by the accelerator and collected by the detector and daily beam
2024
+ utilization coefficient as a function of the operation time for BNB (NuMI) on the left (right). The dotted
2025
+ black line marks the separation between the two operation periods of the detector: the full month of June
2026
+ 2021 and between November 5, 2021 and June 1, 2022 (the long break between the two periods is hidden in
2027
+ the plot).
2028
+ is then fit with a Gaussian, whose area is propor-
2029
+ tional to the number of drift electrons generating
2030
+ the signal.
2031
+ Globally, the efficiency for identifying a
2032
+ wire signal and associating it with the corre-
2033
+ sponding track that generated is exceeding 90%
2034
+ for all three wire planes when the 3D track seg-
2035
+ ment length contributing to each hit (pitch) is
2036
+ larger than 3.4 mm (Fig. 24).
2037
+ Figure 24.
2038
+ Hit efficiency as a function of wire
2039
+ "pitch": blue, red and green points correspond to
2040
+ Induction 1, Induction 2 and Collection wires respec-
2041
+ tively. Measurement made by means of a sample of
2042
+ cosmic muon tracks crossing the cathode.
2043
+ 7.2
2044
+ PMT signal reconstruction
2045
+ The reconstruction of the scintillation light as-
2046
+ sociated with the event of interest is based on
2047
+ the recorded PMTs signals in the event, sam-
2048
+ pled at 500 MHz.
2049
+ For any event triggered in
2050
+ coincidence with the beam spill, all 360 PMTs
2051
+ digitized signals are recorded in 30 µs long time
2052
+ intervals. In addition, for cosmic rays crossing
2053
+ the detector in ±1 ms around the beam gate and
2054
+ identified by the trigger logic, all 180 PMTs be-
2055
+ longing to the ICARUS module containing the
2056
+ event are recorded in 10 µs long time intervals.
2057
+ A threshold-based algorithm is applied to
2058
+ each recorded signal, to identify fired PMTs and
2059
+ to reconstruct the characteristics of the detected
2060
+ light to be used in the event analysis. Whenever
2061
+ a PMT signal exceeds the baseline by 0.5 phe,
2062
+ a new OpHit object is created, characterized by
2063
+ a start time, a time interval for the signal to re-
2064
+ turn back to baseline, a maximal amplitude, and
2065
+ an integral of the signal over the baseline. As
2066
+ a second stage all OpHits in coincidence within
2067
+ 100 ns are clustered together into an OpFlash
2068
+ object. The Opflash is then expanded to include
2069
+ also OpHits within 1 µs after the first OpHit time.
2070
+ Nominally, an OpFlash should correspond to the
2071
+ – 22 –
2072
+
2073
+ Efficiency
2074
+ 0.9
2075
+ 0.8
2076
+ 0.7
2077
+ efficiency profile
2078
+ 0.6
2079
+ 0.5
2080
+ 0.4
2081
+ 0.3
2082
+ 0.4
2083
+ 0.5
2084
+ 0.6
2085
+ 0.7
2086
+ pitch [cm]
2087
+ 0.8total detected light associated to each interac-
2088
+ tion, either due to cosmic rays or to a neutrino
2089
+ interaction. The distribution of the PMT signals
2090
+ in an OpFlash (time, amplitudes, integrals and
2091
+ geometrical positions) is clearly determined by
2092
+ the associated interaction in the TPC (Fig. 25).
2093
+ Figure 25. The PMTs associated with a cosmic ray
2094
+ muon crossing the cathode.
2095
+ Initially, a very simple association between
2096
+ the event in the TPC and the corresponding de-
2097
+ tected light that is based on the comparison of
2098
+ the track and the light barycentre along the lon-
2099
+ gitudinal z axis (zTPC, zPMT) has been adopted.
2100
+ A correlation within few tens of centimeters
2101
+ was observed for the TPC and light barycen-
2102
+ tre (Δz = zTPC − zPMT) for both cosmic muons
2103
+ crossing the cathode (Fig. 26) and for a sample
2104
+ of BNB neutrino interactions (Fig. 27) selected
2105
+ by visual scanning.
2106
+ By requiring |Δz| < 100 cm it is possible
2107
+ to restrict the analysis of the event to a detector
2108
+ slide that is approximately 5% of the total active
2109
+ LAr, with a corresponding reduction of randomly
2110
+ overlapping cosmic rays.
2111
+ 7.3
2112
+ CRT reconstruction
2113
+ The CRT hit reconstruction algorithm was vali-
2114
+ dated during the commissioning phase [35]. The
2115
+ first step in the reconstruction chain is to con-
2116
+ struct CRT hits defined as points in space and
2117
+ time corresponding to a muon track crossing the
2118
+ CRT volume. CRT data coming from Front End
2119
+ Board (FEB) read-outs in a given event are or-
2120
+ dered in time and grouped by CRT region. Due
2121
+ Figure 26. Distribution of Δz = zTPC − zPMT for a
2122
+ sample of cosmic ray muons crossing the cathode.
2123
+ Figure 27. Distribution of Δz = zTPC − zPMT for a
2124
+ sample BNB 𝜈 interactions identified by visual scan-
2125
+ ning.
2126
+ to the differences in design of the side and top
2127
+ CRT systems, the Side and Top CRT Hits have
2128
+ to be handled differently.
2129
+ The coincidence logic in the Side CRTs is
2130
+ performed offline in the reconstruction stage due
2131
+ to the inner and outer CRT modules being con-
2132
+ nected to FEBs in adjacent layers, whereas each
2133
+ top CRT module is a self-contained coincidence
2134
+ unit. In order to identify a coincident grouping of
2135
+ CRT data objects, a software-based coincidence
2136
+ gate is performed (the hardware-based coinci-
2137
+ dence gate width is 150 ns and this value is the
2138
+ minimum for the software gate). The reason for
2139
+ not making the coincidence window too large
2140
+ is to avoid introducing fake coincidences from
2141
+ – 23 –
2142
+
2143
+ PMTs (behindthe
2144
+ Fired
2145
+ wires)
2146
+ PMTs
2147
+ Central cathode
2148
+ PMTs(behind thewires
2149
+ 50
2150
+ z axis24000
2151
+ Cosmic
2152
+ Entries
2153
+ 22000
2154
+ 282361
2155
+ Mean
2156
+ 0.7743
2157
+ 20000
2158
+ muons
2159
+ RMS
2160
+ 51.16
2161
+ 18000
2162
+ 16000
2163
+ 14000
2164
+ 12000
2165
+ 10000
2166
+ 8000
2167
+ 6000
2168
+ 4000
2169
+ 2000
2170
+ 0
2171
+ 400
2172
+ -200
2173
+ 0
2174
+ 200
2175
+ 400#events
2176
+ BNB vuCC
2177
+ 12
2178
+ candidates
2179
+ 10
2180
+ 8
2181
+ rms=41 cm
2182
+ 6
2183
+ -50
2184
+ 0
2185
+ 50
2186
+ 100
2187
+ 150
2188
+ 200
2189
+ cmlow energy events. Studies are underway to es-
2190
+ tablish a gate width that optimizes the tagging
2191
+ efficiency while avoiding introducing fake coin-
2192
+ cidences with low energy events if the gate is too
2193
+ wide.
2194
+ After the creation of coincident groupings
2195
+ of CRT data, the spatial information is extracted
2196
+ to reconstruct the position of the crossing track.
2197
+ The channel with the largest amplitude is the
2198
+ channel that generated the FEB trigger signal.
2199
+ The channel position is identified and extracted
2200
+ from the geometry based on the global coordi-
2201
+ nates of the ICARUS building. The hit position
2202
+ is taken as the mean strip position where a track
2203
+ crosses multiple strips in each layer.
2204
+ When the charge amplitude exceeds the dis-
2205
+ criminator threshold, a CRT hit is acquired by
2206
+ the front-end electronics recording the values of
2207
+ two different time counters. The first counter,
2208
+ T0, is reset every second by means of the PPS
2209
+ signal (see Sec. 5.4) and it provides the global
2210
+ timing of the recorded hit. The second counter,
2211
+ T1, is reset by the event trigger signal and is used
2212
+ to determine the hit relative timing with respect
2213
+ to the event trigger. Each CRT hit timestamp is
2214
+ corrected to account for cable delays and light
2215
+ propagation in the scintillator and in the WLS
2216
+ fiber.
2217
+ The Top CRT hit is defined by the FEB inter-
2218
+ nal triggering logic (see Sec. 4) where a signal
2219
+ threshold of 1.5 phe is applied to each chan-
2220
+ nel. The position within a module is determined
2221
+ by selecting the four channels with the largest
2222
+ amplitude and projected in the global detector
2223
+ coordinates.
2224
+ The CRT timing system has been cross-
2225
+ calibrated with the PMT signals, using the com-
2226
+ mon trigger pulse recorded by the CRT and PMT
2227
+ systems. A preliminary evaluation of the Time-
2228
+ Of-Flight (TOF) of cosmic muons has been per-
2229
+ formed by selecting particles entering the detec-
2230
+ tor from the Top CRT and generating a flash in
2231
+ the active argon volume. The preliminary distri-
2232
+ Figure 28. Time difference between matched CRT
2233
+ hits and PMT flashes. The plot refers to Top CRT
2234
+ data in time with the BNB spill.
2235
+ bution of the time differences between Top CRT
2236
+ hits and PMT signals is shown in Fig. 28: the
2237
+ measured average TOF of 24±9 ns is in agree-
2238
+ ment with the expected ∼ 26 ns evaluated from
2239
+ the distance between the Top CRT plane and the
2240
+ first PMT row.
2241
+ 10
2242
+
2243
+ 8
2244
+
2245
+ 6
2246
+
2247
+ 4
2248
+
2249
+ 2
2250
+
2251
+ 0
2252
+ 2
2253
+ 4
2254
+ 6
2255
+ 8
2256
+ 10
2257
+ s)
2258
+ µ
2259
+ CRT Hit T0 - gate start time (
2260
+ 0
2261
+ 100
2262
+ 200
2263
+ 300
2264
+ 400
2265
+ 500
2266
+ 600
2267
+ 700
2268
+ 800
2269
+ 900
2270
+ Number of CRT Hits
2271
+ BNB, Side, South
2272
+ s
2273
+ µ
2274
+ bin size = 0.2
2275
+ Figure 29. CRT hit time relative to the neutrino gate
2276
+ start time in the south wall (side CRT) for the BNB
2277
+ beam.
2278
+ Figure 29 shows the CRT hit time relative
2279
+ to the neutrino gate start time in the south side
2280
+ CRT wall for the BNB neutrino beam. Using
2281
+ 11 days of commissioning data, a clear peak can
2282
+ be observed, showing activity in the 4 µs trigger
2283
+ coincidence window. Additional activity due to
2284
+ the beam appears inside the smaller BNB gate
2285
+ – 24 –
2286
+
2287
+ Bins
2288
+ 1400
2289
+ Events/100 I
2290
+ Fit parameters:
2291
+ 1200
2292
+ mean= -24 ns
2293
+ sigma= 9 ns
2294
+ 1000
2295
+ 800
2296
+ 600
2297
+ 400
2298
+ 200
2299
+ -80
2300
+ -60
2301
+ -40
2302
+ -20
2303
+ 20
2304
+ 40
2305
+ 60
2306
+ 0
2307
+ 80
2308
+ 100
2309
+ CRT Hit timestamp - PMT Flash [ns](1.6 µs within the 4 µs window), the rest of the
2310
+ activity outside the 1.6 µs window is due to cos-
2311
+ mic ray triggering.
2312
+ 7.4
2313
+ Event display study
2314
+ As a first check of the general behavior of the de-
2315
+ tector, a visual study campaign was performed
2316
+ to select and identify neutrino interactions in the
2317
+ active liquid argon using a graphical event dis-
2318
+ play.
2319
+ As a first step, all the events recorded in the
2320
+ BNB and NuMI beam for some runs were studied
2321
+ selecting the tracks in the cryostat where the trig-
2322
+ ger signal has been produced. An interaction was
2323
+ classified as a neutrino candidate if a clear vertex
2324
+ with more than one track was visually identified:
2325
+ electron neutrino CC candidate events require
2326
+ the presence of a clear electromagnetic shower
2327
+ connected to the primary vertex, while the muon
2328
+ neutrino CC events are selected by requiring the
2329
+ presence of a long track (at least 0.5 m) from
2330
+ the primary vertex. In addition, only events with
2331
+ the primary vertex at least 5 cm from top/bottom
2332
+ TPC sides, 50 cm from the upstream/downstream
2333
+ TPC wall, and 5 cm from the anode position have
2334
+ been initially selected.
2335
+ An example of a 𝜈𝜇CC candidate is shown in
2336
+ Fig. 30, with an estimated total deposited energy
2337
+ of ∼ 1.1 GeV. The CC muon candidate is 3.8 m
2338
+ long, while the highly ionizing track from the pri-
2339
+ mary vertex is identified as a 20 cm long proton.
2340
+ The full wire signal calibration is in the finaliza-
2341
+ tion stage, but by a very preliminary wire signal
2342
+ conversion to estimate the deposited energy, it is
2343
+ possible to reconstruct the dE/dx associated to
2344
+ the individual hits of the muon candidate in the
2345
+ same event, distributed as expected for a MIP
2346
+ particle particle, as shown in Fig. 31.
2347
+ Visual scanning also permitted identifica-
2348
+ tion of 𝜈𝑒CC candidates in the NuMI beam: a
2349
+ remarkable example is shown in Fig. 32 for an
2350
+ event of ∼ 600 MeV deposited energy.
2351
+ 7.5
2352
+ Event reconstruction
2353
+ For a given cryostat, hits identified and passing
2354
+ a multi-plane matching algorithm are passed as
2355
+ input to Pandora [36]: a pattern reconstruction
2356
+ code that performs a 3D reconstruction of the
2357
+ full image recorded in the collected event, in-
2358
+ cluding the identification of interaction vertices
2359
+ and of tracks and showers inside the TPC. These
2360
+ are organized into a hierarchical structure (called
2361
+ a slice) of particles generated starting from a pri-
2362
+ mary interaction vertex or particle.
2363
+ The analysis uses information reconstructed
2364
+ in Pandora to tag and reject “clear cosmic” slices
2365
+ by identifying straight tracks crossing the full ac-
2366
+ tive liquid argon volume or that are clearly out
2367
+ of time with respect to the beam gate. In Monte
2368
+ Carlo studies, selection criteria require that the
2369
+ reconstructed vertex is in the fiducial volume and
2370
+ that PMT timing signals and the reconstructed
2371
+ angle of the muon track are inconsistent with
2372
+ that of a cosmic ray.
2373
+ These requirements re-
2374
+ ject 99.7% of cosmic rays, while accepting more
2375
+ than 82% of true 𝜈𝜇CC events in the fiducial vol-
2376
+ ume. Requiring that a particle identified as a pro-
2377
+ ton be reconstructed in the event further reduces
2378
+ background from cosmic rays. After all criteria
2379
+ are applied, 0.8% of a selected 𝜈𝜇CC contained
2380
+ sample is made up of background from cosmic
2381
+ rays, with 0.6% coming from intime cosmic rays
2382
+ and 0.2% coming from out-of-time cosmic rays.
2383
+ Further tagging and rejection of cosmic rays out
2384
+ of time with respect to the beam spill is possi-
2385
+ ble with the CRT detector, which can provide a
2386
+ few nanosecond absolute time measurement for
2387
+ the TPC tracks when they are unambiguously
2388
+ matched to signals on the CRT. This TPC track-
2389
+ CRT hit matching algorithm is still being tuned
2390
+ and validated with cosmic ray data collected off-
2391
+ beam, but is expected to facilitate improved ef-
2392
+ ficiency and allow further optimization of the
2393
+ cosmic rejection criteria.
2394
+ Pandora and a set of algorithms to iden-
2395
+ – 25 –
2396
+
2397
+ Figure 30. A visually selected 𝜈𝜇CC candidate from the BNB beam.
2398
+ Figure 31. Distribution of the measured dE/dx of the
2399
+ muon candidate in the event shown in Fig. 30. dE/dx
2400
+ is reconstructed on each wire applying a preliminary
2401
+ calibration constant.
2402
+ tify, measure and reconstruct tracks and show-
2403
+ ers can be exploited for the event reconstruction
2404
+ and analysis. These reconstruction tools repre-
2405
+ sent a legacy from past efforts and made avail-
2406
+ able within the LArSoft framework [37], com-
2407
+ plemented by new efforts carried out within the
2408
+ joint SBN effort for a common near and far detec-
2409
+ tor analysis. This set of algorithms is applied to
2410
+ Figure 32. A visually selected 𝜈𝑒CC candidate from
2411
+ the NuMI beam .
2412
+ tracks and showers from any slice in the event to
2413
+ perform particle identification and estimate the
2414
+ momentum from range, calorimetry and multiple
2415
+ Coulomb Scattering.
2416
+ A dedicated visual study of events was per-
2417
+ formed to select ∼ 600 𝜈𝜇CC interactions from
2418
+ BNB in the active liquid argon. These events
2419
+ have been used for validation of the Pandora
2420
+ – 26 –
2421
+
2422
+ Collection plane
2423
+ p
2424
+ Primary
2425
+ vertex
2426
+ Beam direction
2427
+ Cathode90
2428
+ 80
2429
+ Mean
2430
+ 2.235
2431
+ 70
2432
+ RMS
2433
+ 1.211
2434
+ 60
2435
+ 50
2436
+ 40
2437
+ 30
2438
+ 20
2439
+ 10
2440
+ 0
2441
+ 9
2442
+ 1
2443
+ 2
2444
+ 3
2445
+ 4
2446
+ 5
2447
+ 6
2448
+ 8
2449
+ dE/dx[MeV/cm]NuMI veCC
2450
+ candidate
2451
+ Track 1
2452
+ Track 2
2453
+ e-shower
2454
+ (~600MeV)
2455
+ COLL
2456
+ 1 m
2457
+ Wiresreconstruction. In order to reduce the manual
2458
+ effort, events to be visually studied are first se-
2459
+ lected by requiring, offline, the absence of signals
2460
+ in the CRT in coincidence with the trigger. In ad-
2461
+ dition, full 3D reconstruction was performed for
2462
+ the events and only reconstructed tracks longer
2463
+ than 30 cm, fully contained in the detector, and
2464
+ whose barycenter was in agreement within 1 m
2465
+ with the barycenter of the light signal generating
2466
+ the trigger, have been visually studied. For this
2467
+ sample, the neutrino interaction vertex was iden-
2468
+ tified and measured in 3D coordinates as well
2469
+ as the final point associated with the muon can-
2470
+ didate track.
2471
+ Out of the full selected sample,
2472
+ 476 neutrino events present in the analysis files
2473
+ showed a reasonable match with a reconstructed
2474
+ object based on vertex location and were adopted
2475
+ as a benchmark for the validation of the recon-
2476
+ struction tools.
2477
+ As an example, in ∼ 90% of
2478
+ these events the reconstruction reasonably iden-
2479
+ tifies the neutrino interaction vertex along the
2480
+ beam direction, meaning the difference between
2481
+ the two estimates is within 3 cm, as shown in
2482
+ Fig. 33.
2483
+ Comparison of the visual study to auto-
2484
+ mated reconstruction, along with studies of
2485
+ Monte Carlo simulation, will enable further un-
2486
+ derstanding of where to focus efforts and im-
2487
+ provements in the automatic reconstruction. For
2488
+ example, in some cases inefficiencies in a wire
2489
+ plane for a given event reconstruction leading to
2490
+ loss of hits may impact some 3D steps and lead to
2491
+ a track broken into one or more smaller pieces;
2492
+ or algorithms may lead to improper clustering
2493
+ or determination of particle types, etc. Further
2494
+ tuning of the reconstruction is progressing, as
2495
+ well as the complete calibration of the detec-
2496
+ tor. However the first results are quite promis-
2497
+ ing, demonstrating that the basic tools for the
2498
+ event reconstruction and the event selection are
2499
+ operational and allow an initial identification and
2500
+ measurement of neutrino interactions.
2501
+ 10
2502
+
2503
+ 8
2504
+
2505
+ 6
2506
+
2507
+ 4
2508
+
2509
+ 2
2510
+
2511
+ 0
2512
+ 2
2513
+ 4
2514
+ 6
2515
+ 8
2516
+ 10
2517
+ (cm)
2518
+ vertex Z
2519
+
2520
+ 0
2521
+ 20
2522
+ 40
2523
+ 60
2524
+ 80
2525
+ 100
2526
+ 120
2527
+ Slices
2528
+ (scan-reco)
2529
+
2530
+ Figure 33. Difference Δ𝑍 between the automatic and
2531
+ manual measured longitudinal (beam) coordinate of
2532
+ the neutrino interaction vertex for a sample of 476
2533
+ 𝜈𝜇CC candidates from the BNB beam.
2534
+ Conclusions
2535
+ After the successful three-year physics run at
2536
+ the underground LNGS laboratories studying
2537
+ neutrino oscillations with the CERN Neutrino
2538
+ to Gran Sasso beam, the ICARUS T600 LAr-
2539
+ TPC detector underwent a significant overhaul
2540
+ at CERN and was then installed at Fermilab.
2541
+ Detector activation began in 2020 with the cryo-
2542
+ genic commissioning and, despite serious chal-
2543
+ lenges and delays caused by prolonged restric-
2544
+ tions related to the COVID-19 pandemic, it
2545
+ started operations in 2021 and successfully com-
2546
+ pleted its commissioning phase in 2022. It col-
2547
+ lected neutrino events from both the Booster
2548
+ Neutrino Beam (BNB) and the Main Injector
2549
+ (NuMI) beam off-axis.
2550
+ Data taking started in
2551
+ June 2021 with the beam data acquisition, with
2552
+ the detector commissioning activities being con-
2553
+ ducted in parallel. An event sample correspond-
2554
+ ing to ∼ 3 · 1020 and 5 · 1020 POT of the Booster
2555
+ and NuMI beam respectively has been collected
2556
+ with an efficiency exceeding 91% during the
2557
+ normal operations.
2558
+ This data set was used to
2559
+ study the single detector subsystems calibration
2560
+ and to test the ICARUS event selection and re-
2561
+ construction procedure and analysis algorithms.
2562
+ – 27 –
2563
+
2564
+ ICARUS has already started the first year of reg-
2565
+ ular data taking devoted to a sensitive study of the
2566
+ claim by Neutrino-4 short baseline reactor exper-
2567
+ iment both in the 𝜈𝜇 channel with the BNB and in
2568
+ the 𝜈𝑒 channel with NuMI. ICARUS will also ad-
2569
+ dress other fundamental studies such as neutrino
2570
+ cross sections with the NuMI beam and a number
2571
+ of Beyond Standard Model searches. The search
2572
+ for evidence of a sterile neutrino jointly with the
2573
+ Short-Baseline Near Detector, within the Short-
2574
+ Baseline Neutrino program, will follow.
2575
+ Acknowledgements
2576
+ This document was prepared by the ICARUS
2577
+ Collaboration using the resources of the Fermi
2578
+ National Accelerator Laboratory (Fermilab), a
2579
+ U.S. Department of Energy, Office of Science,
2580
+ HEP User Facility.
2581
+ Fermilab is managed by
2582
+ Fermi Research Alliance, LLC (FRA), acting
2583
+ under Contract No.
2584
+ DE-AC02-07CH11359.
2585
+ This work was supported by the US Depart-
2586
+ ment of Energy, INFN, EU Horizon 2020
2587
+ Research and Innovation Program under the
2588
+ Marie Sklodowska-Curie Grant Agreement No.
2589
+ 734303, 822185, 858199, and 101003460 and
2590
+ Horizon Europe Program research and innova-
2591
+ tion programme under the Marie Sklodowska-
2592
+ Curie Grant Agreement No. 101081478. Part
2593
+ of the work resulted from the implementation of
2594
+ the research Project No. 2019/33/N/ST2/02874
2595
+ funded by the National Science Centre, Poland.
2596
+ The ICARUS Collaboration would like to thank
2597
+ the MINOS Collaboration for having provided
2598
+ the side CRT panels as well as Double Chooz
2599
+ (University of Chicago) for the bottom CRT pan-
2600
+ els. We also acknowledge the contribution of
2601
+ many SBND colleagues, in particular for the de-
2602
+ velopment of a number of simulation, recon-
2603
+ struction and analysis tools which are shared
2604
+ within the SBN program. Finally, our experi-
2605
+ ment could not have been carried out without
2606
+ the major support of CERN in the detector over-
2607
+ hauling within the Neutrino Platform framework
2608
+ and of Fermilab in the detector installation and
2609
+ commissioning, and in providing the BNB and
2610
+ NuMI beams.
2611
+ References
2612
+ [1] C. Rubbia. The Liquid Argon Time Projection
2613
+ Chamber: A New Concept for Neutrino
2614
+ Detectors. CERN-EP, 77-08, 1977.
2615
+ [2] A.A. Aguilar-Arevalo et al.
2616
+ (LSND Collaboration). Evidence for Neutrino
2617
+ Oscillations from the Observation of Electron
2618
+ Anti-neutrinos in a Muon Anti-Neutrino
2619
+ Beam. Phys. Rev., D64:112007, 2001.
2620
+ [3] A.A. Aguilar-Arevalo et al.
2621
+ (MiniBooNE Collaboration). Updated
2622
+ MiniBooNE neutrino oscillation results with
2623
+ increased data and new background studies.
2624
+ Phys. Rev., D103:052002, 2021.
2625
+ [4] M. Antonello et al. (ICARUS Collaboration).
2626
+ Search for anomalies in the 𝜈𝑒 appearance
2627
+ from a 𝜈𝜇 beam. Eur. Phys. J., C73:2599,
2628
+ 2013.
2629
+ [5] R. Acciarri et al. (SBND MicroBooNE
2630
+ ICARUS Collaborations). A Proposal for a
2631
+ Three Detector Short-Baseline Neutrino
2632
+ Oscillation Program in the Fermilab Booster
2633
+ Neutrino Beam. arXiv:1503.01520, 2015.
2634
+ [6] O. Palamara P.A.N. Machado and D.W.
2635
+ Schmitz. The Short-Baseline Neutrino
2636
+ Program at Fermilab. Annual Review of
2637
+ Nuclear and Particle Science, 69:367–387,
2638
+ 2019.
2639
+ [7] A.P. Serebrov et al.
2640
+ (Neutrino-4 Collaboration). First Observation
2641
+ of the Oscillation Effect in the Neutrino-4
2642
+ Experiment on the Search for the Sterile
2643
+ Neutrino. JETP lett., 109:213–221, 2019.
2644
+ [8] G.L. Raselli (on behalf of the
2645
+ ICARUS Collaboration). The upgrading of the
2646
+ ICARUS T600 detector. POS
2647
+ (EPS-HEP2017), page 515, 2017.
2648
+ [9] S. Amerio et al. (ICARUS Collaboration).
2649
+ Design, construction and tests of the ICARUS
2650
+ – 28 –
2651
+
2652
+ T600 detector. Nucl. Instr. Meth.,
2653
+ A526:329–410, 2004.
2654
+ [10] C. Rubbia et al. (ICARUS Collaboration).
2655
+ Underground operation of the ICARUS T600
2656
+ LAr-TPC: first results. JINST, 6 P07011, 2011.
2657
+ [11] M. Antonello et al. (ICARUS Collaboration).
2658
+ Experimental observation of an extremely
2659
+ high electron lifetime with the ICARUS-T600
2660
+ LAr-TPC. JINST, 9 P12006, 2014.
2661
+ [12] M. Antonello et al. (ICARUS Collaboration).
2662
+ Precise 3D track reconstruction algorithm for
2663
+ the ICARUS T600 liquid argon time
2664
+ projection chamber detector. Advances in
2665
+ High Energy Physics, 2013:260820, 2013.
2666
+ [13] M. Antonello et al. (ICARUS Collaboration).
2667
+ Muon momentum measurement in
2668
+ ICARUS-T600 LAr-TPC via multiple
2669
+ scattering in few-GeV range. JINST, 12
2670
+ P04010, 2017.
2671
+ [14] C. Farnese (on behalf of the
2672
+ ICARUS Collaboration). Atmospheric
2673
+ Neutrino Search in the ICARUS T600
2674
+ Detector. Universe, 5(1), 2019.
2675
+ [15] L. Bagby et al. (ICARUS Collaboration).
2676
+ Overhaul and installation of the
2677
+ ICARUS-T600 liquid argon TPC electronics
2678
+ for the FNAL Short Baseline Neutrino
2679
+ program. JINST, 16 P01037, 2021.
2680
+ [16] M. Babicz et al. (ICARUS Collaboration).
2681
+ Test and characterization of 400 Hamamatsu
2682
+ R5912-MOD photomultiplier tubes for the
2683
+ ICARUS T600 detector. JINST, 13 P10030,
2684
+ 2018.
2685
+ [17] B. Ali-Mohammadzadeh et al.
2686
+ (ICARUS Collaboration). Design and
2687
+ implementation of the new scintillation light
2688
+ detection system of ICARUS T600. JINST, 15
2689
+ T10007, 2020.
2690
+ [18] M. Bonesini et al. An innovative technique for
2691
+ TPB deposition on convex window
2692
+ photomultiplier tubes. JINST, 13 P12020,
2693
+ 2018.
2694
+ [19] M. Bonesini et al. (on behalf of the
2695
+ ICARUS Collaboration). The laser diode
2696
+ calibration system of the Icarus T600 detector
2697
+ at FNAL. JINST, 15 C05042, 2020.
2698
+ [20] B. Behera (on behalf of the
2699
+ ICARUS Collaboration). Cosmogenic
2700
+ background suppression at the ICARUS using
2701
+ a concrete overburden. J. Phys. Conf. Ser.,
2702
+ 2156(1):012181, 2021.
2703
+ [21] L. Bagby et al. (ICARUS Collaboration).
2704
+ Overhaul and installation of the
2705
+ ICARUS-T600 liquid argon TPC electronics
2706
+ for the FNAL Short Baseline Neutrino
2707
+ program. JINST, 16 P01037, 2021.
2708
+ [22] R. Acciarri et al.
2709
+ (MicroBooNE Collaboration). Noise
2710
+ Characterization and Filtering in the
2711
+ MicroBooNE Liquid Argon TPC. JINST, 12
2712
+ P08003, 2017.
2713
+ [23] W. Walkowiak. Drift velocity of free electrons
2714
+ in liquid argon. Nucl. Instrum. Methods Phys.
2715
+ Res. A, 449:288–294, 2000.
2716
+ [24] P. Abratenko et al.
2717
+ (MicroBooNE Collaboration). Measurement
2718
+ of space charge effects in the MicroBooNE
2719
+ LArTPC using cosmic muons. JINST, 15
2720
+ P12037, 2020.
2721
+ [25] M. Mooney. The MicroBooNE Experiment
2722
+ and the Impact of Space Charge Effects.
2723
+ arXiv:1511.01563, 2015.
2724
+ [26] M. Antonello et al. (ICARUS Collaboration).
2725
+ Study of space charge in the ICARUS T600
2726
+ detector. JINST, 15:P07001, 2020.
2727
+ [27] C. Adams et al. (MicroBooNE Collaboration).
2728
+ Calibration of the charge and energy loss per
2729
+ unit length of the MicroBooNE liquid argon
2730
+ time projection chamber using muons and
2731
+ protons. JINST, 15 P03022, 2020.
2732
+ [28] P.A. Zyla et al. (Particle Data Group). The
2733
+ Review of Particle Physics. Prog. Theor. Exp.
2734
+ Phys., 2020 083C01, 2020.
2735
+ [29] R. Acciarri et al. (ArgoNeuT Collaboration).
2736
+ A study of electron recombination using
2737
+ highly ionizing particles in the ArgoNeuT
2738
+ Liquid Argon TPC. JINST, 8 P08005, 2013.
2739
+ [30] C. Farnese et al. (ICARUS Collaboration).
2740
+ – 29 –
2741
+
2742
+ Implementation of the trigger system of the
2743
+ ICARUS-T600 detector at Fermilab. Nucl.
2744
+ Instr. Meth., A1045:167498, 2023.
2745
+ [31] J. Serrano et al. The White Rabbit Project.
2746
+ Proceedings of the 12𝑡ℎ International
2747
+ Conference On Accelerator And Large
2748
+ Experimental Physics Control Systems, Kobe,
2749
+ Japan, pages 93–95, 2009.
2750
+ [32] K. Biery et al. artdaq: An Event-Building,
2751
+ Filtering, and Processing Framework. IEEE
2752
+ Trans. Nucl. Sci., 60:3764–3771, 2013.
2753
+ [33] C. Green et al. The Art Framework. J. Phys.
2754
+ Conf. Ser., 396:022020, 2012.
2755
+ [34] C. Adams (on behalf of the
2756
+ MicroBooNE Collaboration). Ionization
2757
+ electron signal processing in single phase
2758
+ LArTPCs. Part I. Algorithm Description and
2759
+ quantitative evaluation with MicroBooNE
2760
+ simulation. JINST, 13 P07006, 2018.
2761
+ [35] B. Behera (on behalf of the
2762
+ ICARUS Collaboration). First Data from the
2763
+ Commissioned ICARUS Side Cosmic Ray
2764
+ Tagger. PoS, NuFact2021:201, 2022.
2765
+ [36] R. Acciarri et al.
2766
+ (MicroBooNE Collaboration). The Pandora
2767
+ multi-algorithm approach to automated
2768
+ pattern recognition of cosmic-ray muon and
2769
+ neutrino events in the MicroBooNE detector.
2770
+ arXiv:1708.03135v1, 2017.
2771
+ [37] R. Pordes and E. Snider. The Liquid Argon
2772
+ Software Toolkit (LArSoft): Goals, Status and
2773
+ Plan. PoS, ICHEP2016:182, 2016.
2774
+ – 30 –
2775
+
7tFAT4oBgHgl3EQfoR3-/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
9tAzT4oBgHgl3EQfFPrO/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96abb337afa24021a10d7f46c3d6765c185c6fd390e2bbe7e8c1e9ebd317bdbf
3
+ size 69395
A9E1T4oBgHgl3EQfVgQy/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:069100fda825e90317d14206d5b6b2465032937913db7dc2a27d9fd4a1fbedf5
3
+ size 80359
AdAzT4oBgHgl3EQfTPx3/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5f070a60eb981db0b2162811d3db544378c105c920804eb99e0967b059daa75
3
+ size 37696
CdFAT4oBgHgl3EQftB4M/content/2301.08661v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34ae70177938c493c88ebf5d7bde587ccd5028e28b50131bbff92922ca6865a8
3
+ size 14522646
DtAyT4oBgHgl3EQfSPej/content/tmp_files/2301.00083v1.pdf.txt ADDED
@@ -0,0 +1,1482 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ arXiv:2301.00083v1 [math.PR] 31 Dec 2022
2
+ Weak semiconvexity estimates for Schr¨odinger potentials and
3
+ logarithmic Sobolev inequality for Schr¨odinger bridges
4
+ Giovanni Conforti ∗,
5
+ January 3, 2023
6
+ Contents
7
+ 1
8
+ Introduction and statement of the main results
9
+ 2
10
+ 2
11
+ Invariant sets of weakly convex functions for the HJB flow
12
+ 7
13
+ 3
14
+ Second order bounds for Schr¨odinger potentials
15
+ 10
16
+ 3.1
17
+ Weak semiconvexity of ψ implies weak semiconcavity of ϕ . . . . . . . . . . . . . . .
18
+ 10
19
+ 3.2
20
+ Weak semiconcavity of ϕ implies weak semiconvexity of ψ . . . . . . . . . . . . . . .
21
+ 11
22
+ 3.3
23
+ Proof of Theorem 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
+ 14
25
+ 4
26
+ Logarithmic Sobolev inequality for Schr¨odinger bridges
27
+ 15
28
+ 5
29
+ Appendix
30
+ 18
31
+ ∗CMAP, CNRS, Ecole polytechnique, Institut Polytechnique de Paris, 91120 Palaiseau, France E-mail address:
32
+ giovanni.conforti@polytechnique.edu. Research supported by the ANR project ANR-20-CE40-0014.
33
+ 1
34
+
35
+ Abstract
36
+ We investigate the quadratic Schr¨odinger bridge problem, a.k.a. Entropic Optimal Trans-
37
+ port problem, and obtain weak semiconvexity and semiconcavity bounds on Schr¨odinger poten-
38
+ tials under mild assumptions on the marginals that are substantially weaker than log-concavity.
39
+ We deduce from these estimates that Schr¨odinger bridges satisfy a logarithmic Sobolev inequal-
40
+ ity on the product space. Our proof strategy is based on a second order analysis of coupling
41
+ by reflection on the characteristics of the Hamilton-Jacobi-Bellman equation that reveals the
42
+ existence of new classes of invariant functions for the corresponding flow.
43
+ Mathematics Subject Classification (2020)
44
+ 49Q22,49L12,35G50,60J60,39B62
45
+ 1
46
+ Introduction and statement of the main results
47
+ The Schr¨odinger problem [36] (SP) is a statistical mechanics problem that consists in finding the
48
+ most likely evolution of a cloud of independent Brownian particles conditionally to observations.
49
+ Also known as Entopic Optimal Transport (EOT) problem and formulated with the help of large
50
+ deviations theory as a constrained entropy minimization problem, it stands nowadays at the cross
51
+ of several research lines ranging from functional inequalities [13, 25], statistical machine learning
52
+ [15, 35], control engineering [9, 10], and numerics for PDEs [5, 4]. Given two probability distributions
53
+ µ, ν on Rd, the corresponding (quadratic) Schr¨odinger problem is
54
+ inf
55
+ π∈Π(µ,ν) H(π|R0T ),
56
+ (1)
57
+ where Π(µ, ν) represents the set of couplings of µ and ν and H(π|R0T ) is the relative entropy of a
58
+ coupling π computed against the joint law R0T at times 0 and T of a Brownian motion with initial
59
+ law µ. It is well known that under mild conditions on the marginals, the optimal coupling ˆπ, called
60
+ (static) Schr¨odinger bridge, is unique and admits the representation
61
+ ˆπ(dx dy) = exp(−ϕ(x) − ψ(y)) exp
62
+
63
+ − |x − y|2
64
+ 2T
65
+
66
+ dxdy
67
+ (2)
68
+ where ϕ, ψ are two functions, known as Schr¨odinger potentials [31] that can be regarded as prox-
69
+ ies for the Brenier potentials of optimal transport, that are recovered in the short-time (T → 0)
70
+ limit [34, 12]. In this article we seek for convexity and concavity estimates for Schr¨odinger po-
71
+ tentials. Such estimates have been recently established in [11] and [24] working under a set of
72
+ assumptions that implies in particular log-concavity of at least one of the two marginals. Such
73
+ assumption is crucial therein as it allows to profit from classical functional inequalities such as
74
+ Pr´ekopa-Leindler inequality and Brascamp-Lieb inequality. In particular, the estimates obtained
75
+ in the above-mentioned works yield alternative proofs of Caffarelli’s contraction Theorem [8] in the
76
+ short-time limit. The purpose of this work is twofold: in first place we show at Theorem 1.2 that,
77
+ for any fixed T > 0 it is possible to leverage the probabilistic interpretation of (1) to establish lower
78
+ and upper bounds on the functions
79
+ ⟨∇ϕ(x) − ∇ϕ(y), x − y⟩
80
+ and
81
+ ⟨∇ψ(x) − ∇ψ(y), x − y⟩
82
+ that are valid for all x, y ∈ Rd and do not require strict log concavity of the marginals to hold, but
83
+ still allow to recover the results of [11] as a special case. The second main contribution is to apply
84
+ 2
85
+
86
+ these bounds to prove that static Schr¨odinger bridges satisfy the logarithmic Sobolev inequality
87
+ (LSI for short) at Theorem 1.3. In our main results we shall quantify the weak semiconvexity of a
88
+ potential U : Rd −→ R appealing to the function κU, defined as follows:
89
+ κU : (0, +∞) −→ R,
90
+ κU(r) = inf{|x − y|−2⟨∇U(x) − ∇U(y), x − y⟩ : |x − y| = r}.
91
+ (3)
92
+ κU(r) may be regarded as an averaged or integrated convexity lower bound for U for points that
93
+ are at distance r. This function is often encountered in applications of the coupling method to the
94
+ study of the long time behavior of Fokker-Planck equations [22, 32]. Obviously κU ≥ 0 is equivalent
95
+ to the convexity of U, but working with non-uniform lower bounds on κU allows to design efficient
96
+ generalizations of the classical notion of convexity. A commonly encountered sufficient condition
97
+ on κU ensuring the exponential trend to equilibrium of the Fokker-Planck equation
98
+ ∂tµt − 1
99
+ 2∆µt − ∇ ·
100
+
101
+ ∇U µt
102
+
103
+ = 0
104
+ is the following
105
+ κU(r) ≥
106
+
107
+ α,
108
+ if r > R,
109
+ α − L′,
110
+ if r ≤ R,
111
+ (4)
112
+ for some α, L′, R > 0. In this work, we refer to assumptions of the form (4) and variants thereof
113
+ as to weak convexity assumptions and our main result require an assumption of this kind, namely
114
+ (6) below, that is shown to be no more demanding than (4) (see Proposition 5.1), and is expressed
115
+ through a rescaled version of the hyperbolic tangent function. These functions play a special role in
116
+ this work since, as we show at Theorem 2.1, they define a weak convexity property that propagates
117
+ backward along the flow of the Hamilton-Jacobi-Bellman (HJB) equation
118
+ ∂tϕt + 1
119
+ 2∆ϕt − 1
120
+ 2|∇ϕt|2 = 0.
121
+ Such invariance property represents the main innovation in our proof strategy: the propagation
122
+ of the classical notion of convexity along the HJB equation used in [24] as well as the Brascamp-Lieb
123
+ inequality employed in [11] are both consequences of the Pr´ekopa-Leindler inequality, see [7]. In
124
+ the framework considered here, such powerful tool becomes ineffective due to the possible lack of
125
+ log-concavity in both marginals. To overcome this obstacle we develop a probabilistic approach
126
+ based on a second order analysis of coupling by reflection on the solutions of the SDE
127
+ dXt = −∇ϕt(Xt)dt + dBt,
128
+ also known as characteristics of the HJB equation, that enables to establish the above mentioned
129
+ propagation of weak convexity (Theorem 2.1). This property is a key ingredient the proof of the
130
+ semiconvexity bounds of Theorem 1.2. Static Schr¨odinger bridges are not log-concave probability
131
+ measures in general, not even in the case when both marginals are strongly log-concave. For this
132
+ reason, one cannot infer LSI directly from Theorem 1.2 and the Bakry-´Emery criterion. However,
133
+ reintroducing a dynamical viewpoint and representing Schr¨odinger bridges as Doob h-transforms
134
+ of Brownian motion [21] reveals all the effectiveness of Theorem 1.2 that gives at once gradient
135
+ estimates and local (or conditional, or heat kernel) logarithmic Sobolev inequalities and gradient
136
+ estimates for the h-transform semigroup. By carefully mixing the local inequalities with the help
137
+ of the gradient estimates, we finally establish at Theorem 1.3 LSI for ˆπ, that is our second main
138
+ 3
139
+
140
+ contribution. It is worth noticing that in the T → +∞ asymptotic regime, our approach to LSI
141
+ can be related to the techniques recently developed in [33] to construct Lipschitz transports be-
142
+ tween the Gaussian distribution and probability measures that are approximately log-concave in
143
+ a suitable sense. Because of the intrinsic probabilistic nature of our proof strategy, our ability to
144
+ compensate for the lack of log-concavity in the marginals depends on the size of the regularization
145
+ parameter T , and indeed vanishes as T → 0. Thus, our main results do not yield any sensible
146
+ convexity/concavity estimate on Brenier potentials that improves on Caffarelli’s Theorem. On the
147
+ other hand, the semiconvexity bounds of 1.2 find applications beyond LSI, that we shall address in
148
+ future works. For example, following classical arguments put forward in [20], they can be shown
149
+ to imply transport-entropy (a.k.a. Talagrand) inequalities on path space for dynamic Schr¨odiner
150
+ bridges. Moreover, building on the results of [13], they shall imply new semiconvexity estimates
151
+ for the Fisher information along entropic interpolations. It is also natural to conjecture that these
152
+ bounds will provide with new stability estimates for Schr¨odinger bridges under marginal perturba-
153
+ tions, thus addressing a question that has recently drawn quite some attention, see [19, 12, 23, 26, 3]
154
+ for example. Finally, we point out that Hessian bounds for potentials can play a relevant role in
155
+ providing theoretical guarantees for learning algorithms that make use of dynamic Schr¨odinger
156
+ bridges and conditional processes. In this framework, leveraging Doob’s h-transform theory and
157
+ time reversal arguments, they directly translate into various kinds of quantitative stability estimates
158
+ for the diffusion processes used for sampling, see e.g. [18, 17, 37].
159
+ Organization
160
+ The document is organized as follows. In remainder of the first section we state
161
+ and comment our main hypothesis and results. In Section 2 we study invariant sets for the HJB
162
+ flow. Sections 3 and 4 are devoted to the proof our two main results, Theorem 1.2 and Theorem
163
+ 1.3. Technical results and background material are collected in the Appendix section.
164
+ Assumption 1.1. We assume µ, ν admit a positive density against the Lebesgue measure which
165
+ can be written in the form exp(−U µ) and exp(−U ν) respectively. U µ, U ν are of class C2(Rd).
166
+ (H1) µ has finite second moment and finite relative entropy against the Lebsegue measure. More-
167
+ over, there exists βµ > 0 such that
168
+ ⟨v, ���2U µ(x)v⟩ ≤ βµ|v|2
169
+ ∀x, v ∈ Rd.
170
+ (5)
171
+ One of the following holds
172
+ (H2) There exist αν, L > 0 such that
173
+ κUν(r) ≥ αν − r−1fL(r)
174
+ ∀r > 0,
175
+ (6)
176
+ where the function fL is defined for any L > 0 by:
177
+ fL : [0, +∞] −→ [0, +∞],
178
+ fL(r) = (2L)1/2 tanh
179
+ �1
180
+ 2(2L)1/2r
181
+
182
+ .
183
+ (H2′) There exist αν, L′ > 0 such that
184
+ κUν(r) ≥
185
+
186
+ αν,
187
+ if r > R,
188
+ αν − L′,
189
+ if r ≤ R.
190
+ In this case, we set
191
+ L = inf{¯L : R−1f¯L(R) ≥ L′}.
192
+ (7)
193
+ 4
194
+
195
+ Clearly, imposing (6) is less restrictive than asking that ν is strongly log-concave.
196
+ Remark 1.1. We show that (H2′) implies (H2) at Proposition 5.1.
197
+ Remark 1.2. The requirement that the density of ν is strictly positive everywhere could be dropped
198
+ at the price of additional technicalities. For µ, such requirement is a consequence of (5).
199
+ The Schr¨odinger system
200
+ Let (Pt)t≥0 the semigroup generated by a d-dimensional Brownian
201
+ motion. For given marginals, µ, ν and T > 0 the Schr¨odinger system, whose unknowns ϕ, ψ we
202
+ shall refer to as Schr¨odinger potentials, is given by
203
+
204
+ ϕ(x) = U µ(x) + log PT exp(−ψ)(x),
205
+ x ∈ Rd,
206
+ ψ(y) = U ν(y) + log PT exp(−ϕ)(y),
207
+ y ∈ Rd.
208
+ (8)
209
+ Under Assumption 1.1, it is known that the Schr¨odinger system admits a solution, and that if ( ¯ϕ, ¯ψ)
210
+ is another solution, then there exists c ∈ R such that (ϕ, ψ) = ( ¯ϕ + c, ¯ψ − c), see [34, sec. 2][31] and
211
+ references therein.
212
+ Weak semiconvexity and semiconcavity bounds for Schr¨odinger potentials
213
+ In the rest
214
+ of the article, given a scalar function U, any pointwise lower bound on κU implying in particular
215
+ that
216
+ lim inf
217
+ r→+∞ κU(r) > −∞
218
+ shall be called a weak semiconvexity bound for U. Next, in analogy with (3) we introduce for a
219
+ differentiable U : Rd −→ R the function ℓU as follows:
220
+ ℓU : (0, +∞) −→ R,
221
+ ℓU(r) = sup{|x − y|−2⟨∇U(x) − ∇U(y), x − y⟩ : |x − y| = r},
222
+ and call a weak semiconcavity bound for U any pointwise upper bound for ℓU implying in particular
223
+ that
224
+ lim sup
225
+ r→+∞ ℓU(r) < +∞.
226
+ Our first main result is about weak semiconvexity and weak semiconcavity bounds for Schr¨odinger
227
+ potentials.
228
+ Theorem 1.2. Let Assumption 1.1 hold and (ϕ, ψ) be solutions of the Schr¨odinger system. Then
229
+ ϕ, ψ are twice differentiable and for all r > 0 we have
230
+ κψ(r) ≥ αψ − r−1fL(r),
231
+ (9)
232
+ ℓϕ(r) ≤ βµ −
233
+ α
234
+ (1 + T α) + r−1fL(r)
235
+ (1 + T α)2 ,
236
+ (10)
237
+ where αψ > αν − 1/T can be taken to be the smallest solution of the fixed point equation
238
+ α = αν − 1
239
+ T + G(α, 2)
240
+ 2T 2
241
+ ,
242
+ α ∈ (αν − 1/T, +∞)
243
+ (11)
244
+ 5
245
+
246
+ with
247
+ G(α, u) = inf{s ≥ 0 : F(α, s) ≥ u},
248
+ u > 0
249
+ (12)
250
+ and
251
+ F(α, s) = βµs +
252
+ s
253
+ T (1 + T α) + s1/2fL(s1/2)
254
+ (1 + T α)2 ,
255
+ s > 0.
256
+ Remark 1.3. It is proven at Lemma 3.2 that F(α, ·) is increasing on (0, +∞) for all α > −1/T .
257
+ G(α, ·) is therefore its inverse.
258
+ Remark 1.4. We conjecture that αψ can be taken to be the largest solution of the fixed point
259
+ equation (11).
260
+ To prove so, it would suffice to show that Sinkhorn’s iterates (see [34, Sec 6])
261
+ converge to solutions of the Schr¨odinger system under Assumption 1.1 for a large set of initial
262
+ conditions. We could not find such result in the existing literature.
263
+ Remark 1.5. It is possible to check that if (H2) holds with L = 0, Theorem 1.2 recovers the
264
+ conclusion of [11, Theorem 4],after a change of variable. To be more precise, the potentials (ϕε, ψε)
265
+ considered there are related to the couple (ϕ, ψ) appearing in (8) by choosing ε = T and setting
266
+ ϕε = ε
267
+
268
+ ϕ − U µ + | · |2
269
+
270
+
271
+ ,
272
+ ψε = ε
273
+
274
+ ψ − U ν + | · |2
275
+
276
+
277
+ .
278
+ Remark 1.6. The rescaled potential T ϕ converges to the Brenier potential in the small noise limit
279
+ [34]. As explained in the introduction, one cannot deduce from Theorem 1.2 an improvement over
280
+ Caffarelli’s Theorem [8] by letting T → 0 in Theorem 1.2.
281
+ Our second main result is that the static Schr¨odinger bridge ˆπ satisfies LSI with an explicit
282
+ constant. We recall here that a probability measure ρ on Rd satisfies LSI with constant C if and
283
+ only if for all positive differentiable function f
284
+ Entρ(f) ≤ C
285
+ 2
286
+ � |∇f|2
287
+ f
288
+ dρ,
289
+ where
290
+ Entρ(f) =
291
+
292
+ f log fdρ −
293
+
294
+ fdρ log
295
+ � �
296
+ fdρ
297
+
298
+ .
299
+ Theorem 1.3. Let Assumption 1.1 hold and assume furthermore that µ satisfies LSI with constant
300
+ Cµ. Then the static Schr¨odinger bridge ˆπ satisfies LSI with constant
301
+ max
302
+
303
+ Cµ, CµC0,T +
304
+ � T
305
+ 0
306
+ Ct,T dt
307
+
308
+ ,
309
+ where for all t ≤ T
310
+ Ct,T := exp
311
+
312
+
313
+ � T
314
+ t
315
+ αψ
316
+ s ds
317
+
318
+ ,
319
+ αψ
320
+ t :=
321
+ αψ
322
+ 1 + (T − t)αψ
323
+
324
+ L
325
+ (1 + (T − t)αψ)2 ,
326
+ and αψ is as in Theorem 1.2.
327
+ It is well known that LSI has a number of remarkable consequences including, but certainly not
328
+ limited to, spectral gaps and concentration of measure inequalities for Lipschitz observables.
329
+ Remark 1.7. By taking µ to be a Gaussian distribution, we obtain as a corollary of Theorem 1.3
330
+ that any probability ν fulfilling (6) satisfies a logarithmic Sobolev inequality. To the best of our
331
+ knowledge, this is a new result. It is worth noticing (6) does not imply that ν is a bounded or
332
+ Lipschitz perturbation of a log-concave distribution: therefore the results of [28][1] do not apply in
333
+ this context.
334
+ 6
335
+
336
+ 2
337
+ Invariant sets of weakly convex functions for the HJB flow
338
+ We introduce the notation
339
+ U T,g
340
+ t
341
+ (x) := − log PT −t exp(−g)(x) = − log
342
+
343
+ 1
344
+ (2π(T − t))d/2
345
+
346
+ exp
347
+
348
+ − |y − x|2
349
+ 2(T − t) − g(y)
350
+
351
+ dy
352
+
353
+ . (13)
354
+ With this notation at hand, (8) rewrites as follows:
355
+
356
+ ϕ = U µ − U T,ψ
357
+ 0
358
+ ,
359
+ ψ = U ν − U T,ϕ
360
+ 0
361
+ .
362
+ (14)
363
+ It is well known that under mild conditions on g, the map [0, T ] × Rd ∋ (t, x) �→ U T,g
364
+ t
365
+ (x) is a
366
+ classical solution of th HJB equation
367
+
368
+ ∂tϕt(x) + 1
369
+ 2∆ϕt(x) − 1
370
+ 2|∇ϕt|2(x) = 0,
371
+ ϕT (x) = g(x).
372
+ (15)
373
+ In the next theorem, we construct for any L > 0 a set of weakly convex functions FL that is shown
374
+ to be invariant for the HJB flow. In the proof, and in the rest of the paper we shall denote by [·, ·]
375
+ the quadratic covariation of two Itˆo processes.
376
+ Theorem 2.1. Fix L > 0 and define
377
+ FL = {g ∈ C1(Rd) : κg(r) ≥ −r−1fL(r)
378
+ ∀r > 0}.
379
+ Then for all 0 ≤ t ≤ T < +∞ we have
380
+ g ∈ FL ⇒ U T,g
381
+ t
382
+ ∈ FL.
383
+ (16)
384
+ In the proof of the Theorem we profit from the fact that fL solves the ODE
385
+ ff ′(r) + 2f ′′(r) = 0
386
+ ∀r > 0,
387
+ f(0) = 0, f ′(0) = L.
388
+ (17)
389
+ To verify the above, it suffices to compute
390
+ f ′
391
+ L(r) =
392
+ L
393
+ cosh2( 1
394
+ 2(2L)1/2r),
395
+ f ′′
396
+ L(r) = 2−1/2L3/2 sinh( 1
397
+ 2(2L)1/2r)
398
+ cosh3( 1
399
+ 2(2L)1/2r)
400
+ Moreover, we recall here some useful properties of fL:
401
+ fL(r) > 0, f ′
402
+ L(r) > 0, f ′′
403
+ L(r) < 0, fL(r) ≥ rf ′
404
+ L(r)
405
+ ∀r > 0.
406
+ (18)
407
+ We are now in position to prove Theorem 2.1. As anticipated above, the proof relies on the analysis
408
+ of coupling by reflection along the characteristics of the HJB equation. In the recent article [14, Thm
409
+ 1.3] Hessian bounds for HJB equations originating from stochastic control problems are obtained
410
+ by means of coupling techniques. These are two-sided bounds that require an a priori knowledge
411
+ of global Lipschitz bounds on solutions of the HJB equation to hold.
412
+ The one-sided estimates
413
+ of Theorem 2.1 do not require any Lipschitz property of solutions and their proof require finer
414
+ arguments than those used in [14].
415
+ 7
416
+
417
+ Proof. We first assume w.l.o.g. that t = 0 and work under the additional assumption that
418
+ g ∈ C3(Rd),
419
+ sup
420
+ x∈Rd |∇2g|(x) < +∞.
421
+ (19)
422
+ Combining the above with g ∈ FL, we can justify differentiation under the integral sign in (13) and
423
+ establish that
424
+ [0, T ] × Rd ∋ (t, x) �→ U T,g
425
+ t
426
+ (x)
427
+ is a classical solution of (15) such that
428
+ [0, T ] × Rd ∋ (t, x) �→ ∇U T,g
429
+ t
430
+ (x)
431
+ is continuously differentiable in t as well as twice continuously differentiable and uniformly Lipschitz
432
+ in x. Under these regularity assumptions, for given x, ˆx ∈ Rd, coupling by reflection of two diffusions
433
+ started at x and ˆx respectively and whose drift field is −∇U T,g
434
+ t
435
+ is well defined, see [22]. That is
436
+ to say, there exist a stochastic process (Xt, ˆXt)0≤t≤T with (X0, ˆX0) = (x, ˆx) and two Brownian
437
+ motions (Bt, ˆBt)0≤t≤T all defined on the same probability space and such that
438
+
439
+ dXt = −∇U T,g
440
+ t
441
+ (Xt)dt + dBt,
442
+ for 0 ≤ t ≤ T ,
443
+ d ˆXt = −∇U T,g
444
+ t
445
+ ( ˆXt)dt + d ˆBt,
446
+ for 0 ≤ t ≤ τ, Xt = ˆXt for t > τ,
447
+ where
448
+ et = r−1
449
+ t
450
+ (Xt − ˆXt),
451
+ rt = |Xt − ˆXt|,
452
+ d ˆBt = dBt − 2et⟨et, dBt⟩
453
+ and
454
+ τ = inf{t ∈ [0, T ] : Xt = ˆXt} ∧ T.
455
+ We now define
456
+ U : [0, T ] × Rd × Rd −→ R,
457
+ Ut(x, ˆx) =
458
+
459
+ |x − ˆx|−1⟨∇U T,g
460
+ t
461
+ (x) − ∇U T,g
462
+ t
463
+ (ˆx), x − ˆx⟩,
464
+ if x ̸= ˆx,
465
+ 0
466
+ if x = ˆx,
467
+ and proceed to prove that (U(Xt, ˆXt))0≤t≤T is a supermartingale. To this aim, we first deduce from
468
+ (15) and Itˆo’s formula that
469
+ d∇U T,g
470
+ t
471
+ (Xt) = dMt,
472
+ d∇U T,g
473
+ t
474
+ ( ˆXt) = d ˆ
475
+ Mt
476
+ (20)
477
+ where M·, ˆ
478
+ M· are square integrable martingales. Indeed we find from Itˆo’s formula
479
+ d∇U T,g
480
+ t
481
+ (Xt) =
482
+
483
+ ∂t∇U T,g
484
+ t
485
+ (Xt) − ∇2U T,g
486
+ t
487
+ ∇U T,g
488
+ t
489
+ (Xt) + 1
490
+ 2∆∇U T,g
491
+ t
492
+ (Xt)
493
+
494
+ dt + ∇2U T,g
495
+ t
496
+ (Xt) · dBt
497
+ (15)
498
+ = ∇2U T,g
499
+ t
500
+ (Xt) · dBt,
501
+ and a completely analogous argument shows that ∇U T,g
502
+ t
503
+ ( ˆ
504
+ Xt) is a square integrable martingale. We
505
+ shall also prove separately at Lemma 2.1 that
506
+ det = −r−1
507
+ t proje⊥
508
+ t (∇U T,g
509
+ t
510
+ (Xt) − ∇U T,g
511
+ t
512
+ ( ˆXt))dt
513
+ ∀t < τ,
514
+ (21)
515
+ 8
516
+
517
+ where proje⊥
518
+ t denotes the orthogonal projection on the orthogonal complement of the linear subspace
519
+ generated by et. Combining together (20) and(21) we find that dUt(Xt, ˆXt) = 0 for t ≥ τ, whereas
520
+ for t < τ
521
+ dUt(Xt, ˆXt) = ⟨∇U T,g
522
+ t
523
+ (Xt) − ∇U T,g
524
+ t
525
+ ( ˆXt), det⟩
526
+ + ⟨et, d(∇U T,g
527
+ t
528
+ (Xt) − ∇U T,g
529
+ t
530
+ ( ˆXt))⟩ + d[(∇U T,g
531
+ ·
532
+ (X·) − ∇U T,g
533
+ ·
534
+ ( ˆX·)), e·]t
535
+ (20)+(21)
536
+ =
537
+ −r−1
538
+ t
539
+ |proje⊥
540
+ t (∇U T,g
541
+ t
542
+ (Xt) − ∇U T,g
543
+ t
544
+ ( ˆXt))|2dt + d ˜
545
+ Mt.
546
+ proving that (U(Xt, ˆXt))0≤t≤T is a supermartingale. In the above, ˜
547
+ M· denotes a square integrable
548
+ martingale and to obtain the last equality we used that the quadratic variation term vanishes
549
+ because of (21). Next, arguing exactly as in [22, Eq. 60] (see also (25) below for more details) on
550
+ the basis of Itˆo’s formula and invoking (17) we get
551
+ dfL(rt) = [−f ′
552
+ L(rt)Ut(Xt, ˆXt) + 2f ′′
553
+ L(rt)]dt + dNt
554
+ (17)
555
+ = −f ′
556
+ L(rt)[Ut(Xt, ˆXt) + fL(rt)]dt + dNt,
557
+ where N· is a square integrable martingale. It then follows that
558
+ d
559
+
560
+ Ut(Xt, ˆXt) + fL(rt)
561
+
562
+ ≤ −f ′
563
+ L(rt)
564
+
565
+ Ut(Xt, ˆXt) + fL(rt)
566
+
567
+ dt + dNt + d ˜
568
+ Mt.
569
+ (22)
570
+ from which we deduce that the process
571
+ Γt = exp
572
+ � � t
573
+ 0
574
+ f ′
575
+ L(rs)ds
576
+ ��
577
+ Ut(Xt, ˆXt) + fL(rt)
578
+
579
+ is a supermartingale and in particular is decreasing on average. This gives
580
+ |x − ˆx|−1⟨∇U T,g
581
+ 0
582
+ (x) − ∇U T,g
583
+ 0
584
+ (ˆx), x − ˆx⟩ + fL(|x − ˆx|) = E[Γ0]
585
+ ≥ E[ΓT ] ≥ E
586
+
587
+ exp(
588
+ � T
589
+ 0
590
+ f ′
591
+ L(rs)ds)
592
+
593
+ |XT − ˆXT |κg(|XT − ˆXT |) + fL(|XT − ˆXT |)
594
+ ��
595
+ ≥ 0,
596
+ where the last inequality follows from g ∈ FL.
597
+ We have thus completed the proof under the
598
+ additional assumption (19). In order to remove it, consider any g ∈ FL. Then there exist (gn) ⊆ FL
599
+ such that (19) holds for any of the gn, gn → g pointwise and gn is uniformly bounded below. From
600
+ this, one can prove that ∇U gn,T
601
+ 0
602
+ → ∇U g,T
603
+ 0
604
+ pointwise by differentiating (13) under the integral sign.
605
+ Using this result in combination with the fact that (16) holds for any gn allows to reach the desired
606
+ conclusion.
607
+ Lemma 2.1. Under the same assumptions and notations of Theorem 2.1 we have
608
+ det = −r−1
609
+ t proje⊥
610
+ t (∇U T,g
611
+ t
612
+ (Xt) − ∇U T,g
613
+ t
614
+ ( ˆXt))dt
615
+ ∀t < τ.
616
+ Proof. Recall that if θ : Rd → R is the map z �→ |z|, then we have
617
+ ∇θ(z) = z
618
+ |z|,
619
+ ∇2θ(z) = I
620
+ |z| − zz⊤
621
+ |z|3 ,
622
+ z ̸= 0.
623
+ (23)
624
+ 9
625
+
626
+ The proof consist of several applications of Itˆo’s formula. We first observe that for t < τ
627
+ d(Xt − ˆXt) = −(∇U T,g
628
+ t
629
+ (Xt) − ∇U T,g
630
+ t
631
+ ( ˆXt))dt + 2etdWt,
632
+ with
633
+ dWt = ⟨et, dBt⟩.
634
+ (24)
635
+ Note that by L´evy characterization, (Wt)0≤t≤T is a Brownian motion. Thus, invoking (23) (or
636
+ refferring directly to [22, Eq. 60] we obtain
637
+ drt = −⟨∇U T,g
638
+ t
639
+ (Xt) − ∇U T,g
640
+ t
641
+ ( ˆXt), et⟩dt + 2dWt,
642
+ (25)
643
+ whence
644
+ dr−1
645
+ t
646
+ = −r−2
647
+ t drt + r−3
648
+ t
649
+ d[r]t
650
+ =
651
+
652
+ r−2
653
+ t
654
+ ⟨∇U T,g
655
+ t
656
+ (Xt) − ∇U T,g
657
+ t
658
+ ( ˆXt), et⟩ + 4r−3
659
+ t
660
+
661
+ dt − 2r−2
662
+ t dWt.
663
+ (26)
664
+ Combining (24) with (26) we find that for t < τ
665
+ det = d
666
+
667
+ r−1
668
+ t
669
+ (Xt − ˆXt))
670
+ = r−1
671
+ t
672
+ d(Xt − ˆXt) + (Xt − ˆXt)d(r−1
673
+ t
674
+ ) + d[X· − ˆX·, r−1
675
+ ·
676
+ ]t
677
+ = −r−1
678
+ t
679
+ (∇U T,g
680
+ t
681
+ (Xt) − ∇U T,g
682
+ t
683
+ ( ˆXt))dt + 2r−1
684
+ t
685
+ etdWt
686
+ +
687
+
688
+ r−2
689
+ t ⟨∇U T,g
690
+ t
691
+ (Xt) − ∇U T,g
692
+ t
693
+ ( ˆ
694
+ Xt), et⟩ + 4r−3
695
+ t
696
+
697
+ (Xt − ˆXt)dt
698
+ − 2r−2
699
+ t (Xt − ˆXt)dWt − 4r−2
700
+ t etdt
701
+ = −r−1
702
+ t
703
+
704
+ ∇U T,g
705
+ t
706
+ (Xt) − ∇U T,g
707
+ t
708
+ ( ˆXt) − ⟨∇U T,g
709
+ t
710
+ (Xt) − ∇U T,g
711
+ t
712
+ ( ˆXt), et⟩et
713
+
714
+ dt
715
+ = −(r−1
716
+ t
717
+ )proje⊥
718
+ t (∇U T,g
719
+ t
720
+ (Xt) − ∇U T,g
721
+ t
722
+ ( ˆXt))dt.
723
+ 3
724
+ Second order bounds for Schr¨odinger potentials
725
+ From now on Assumption 1.1 is in force, even if we do not specify it. Moreover, since we show at
726
+ Proposition 5.1 in the appendix that (H2′) implies (H2), we shall always assume that (H2) holds
727
+ in the sequel. The next two subsections are devoted to establish the key estimates needed in the
728
+ proof of Theorem 1.2, that is carried out immediately afterwards.
729
+ 3.1
730
+ Weak semiconvexity of ψ implies weak semiconcavity of ϕ
731
+ Lemma 3.1. Assume that α > −1/T exists such that
732
+ κψ(r) ≥ α − r−1fL(r)
733
+ ∀r > 0.
734
+ Then we have
735
+ ℓϕ(r) ≤ βµ −
736
+ α
737
+ 1 + T α + r−1fL(r)
738
+ (1 + T α)2 = r−2F(α, r2) − 1
739
+ T
740
+ ∀r > 0.
741
+ 10
742
+
743
+ Proof. We define
744
+ ˆψ(·) = ψ(·) − α
745
+ 2 | · |2.
746
+ and note by assumption ˆψ ∈ FL. We claim that
747
+ U T,ψ
748
+ 0
749
+ (x) =
750
+ α
751
+ 2(1 + T α)|x|2 + U T/(1+T α), ˆψ
752
+ 0
753
+ ((1 + T α)−1x) + C,
754
+ (27)
755
+ where C is some constant independent of x. Indeed we have
756
+ U T,ψ
757
+ 0
758
+ (x) − d
759
+ 2 log(2πT ) = − log
760
+
761
+ exp
762
+
763
+ − |y − x|2
764
+ 2T
765
+ − α
766
+ 2 |y|2 − ˆψ(y)
767
+
768
+ dy
769
+ = − log
770
+
771
+ exp
772
+
773
+
774
+ α|x|2
775
+ 2(1 + T α) − 1 + T α
776
+ 2T
777
+ |y − (1 + T α)−1x|2 − ˆψ(y)
778
+
779
+ dy
780
+ =
781
+ α|x|2
782
+ 2(1 + T α) + U T/(1+T α), ˆ
783
+ ψ
784
+ 0
785
+ ((1 + T α)−1x) − d
786
+ 2 log(2πT/(1 + T α))
787
+ Since ˆψ ∈ FL, we can invoke Theorem 2.1 to obtain
788
+ κUT,ψ
789
+ 0
790
+ (r) ≥
791
+ α
792
+ 1 + T α − r−1fL(r)
793
+ (1 + T α)2
794
+ ∀r > 0.
795
+ (28)
796
+ The desired conclusion is then obtained from (14) and Assumption 1.1.
797
+ 3.2
798
+ Weak semiconcavity of ϕ implies weak semiconvexity of ψ
799
+ We begin by recording some useful properties of the functions F(·, ·) and G(·, ·).
800
+ Lemma 3.2. Let T, βµ > 0, L ≥ 0 be given.
801
+ (i) For any α > −1/T the function
802
+ s �→ F(α, s)
803
+ is concave and increasing [0, +∞).
804
+ (ii) α �→ G(α, 2) is positive and non decreasing over (− 1
805
+ T , +∞).
806
+ (iii) The fixed point equation (11) admits at least one solution on (αν − 1/T, +∞) and αν − 1/T
807
+ is not an accumulation point for the set of solutions.
808
+ Proof. We begin with the proof of (i). To this aim, we observe that fL is increasing on [0, +∞)
809
+ and therefore so is s �→ s1/2fL(s1/2). Therefore
810
+ d
811
+ dsF(α, s) ≥ βµ +
812
+ 1
813
+ T (1 + T α) > 0,
814
+ where we used α > −1/T in the last inequality. To prove concavity, we observe that
815
+ d2
816
+ du2
817
+
818
+ u1/2fL(u1/2)
819
+ ����
820
+ u=s = s−1/2
821
+ 4
822
+ f
823
+ ′′
824
+ L(s1/2) + s−3/2
825
+ 4
826
+ (f
827
+
828
+ L(s1/2)s1/2 − fL(s1/2))
829
+ (18)
830
+ < 0.
831
+ 11
832
+
833
+ Thus s �→ s1/2fL(s1/2) is concave and so is F(α, ·). We now move on to the proof of (ii) by first
834
+ showing that G(·, 2) is positive and then showing that it is increasing. If this was not the case then
835
+ G(α, 2) = 0 for some α > −1/T and therefore there exists a sequence (sn)n≥0 such that sn → 0
836
+ and F(α, sn) ≥ 2. But this is impossible since lims↓0 F(α, sn) = 0. Next, we observe that F(α, s)
837
+ is increasing in s from item (i) and decreasing in α for α ∈ (−1/T, +∞). For this reason, for any
838
+ u and α′ ≥ α we have
839
+ {s : F(α′, s) ≥ u} ⊆ {s : F(α, s) ≥ u}
840
+ and therefore
841
+ G(α′, u) ≥ G(α, u).
842
+ To prove (iii), we introduce
843
+ h : [αν − 1
844
+ T , +∞) −→ R,
845
+ h(α) := α −
846
+
847
+ αν − 1
848
+ T + G(α, 2)
849
+ 2T 2
850
+
851
+ Note that that h is continuous on its domain since G(·, 2) is so. Therefore, to reach the conclusion
852
+ it suffices to show that
853
+ h(αν − 1
854
+ T ) < 0,
855
+ lim
856
+ α→+∞ h(α) = +∞.
857
+ (29)
858
+ The first inequality is a direct consequence of G(αν − 1/T ) > 0, that we have already proven. The
859
+ second inequality is proven if we can show that
860
+ lim sup
861
+ α→+∞ G(α, 2) ≤
862
+ 1
863
+ 2βµ
864
+ .
865
+ (30)
866
+ To see that this relation holds, observe that, using fL(r) ≥ 0 we obtain that for any α > −1/T
867
+ F(α, s) ≥ βµs
868
+ ∀s > 0.
869
+ But then we obtain directly from (12) that
870
+ G(α, 2) ≤
871
+ 1
872
+ 2βµ
873
+ ,
874
+ thus proving (30).
875
+ We shall now introduce the modified potential ¯ψ as follows
876
+ ¯ψ(y) = T
877
+
878
+ ψ(y) − U ν(y) + |y|2
879
+ 2T
880
+
881
+ ,
882
+ (31)
883
+ It has been proven at [11, Lemma 1] that the Hessian of ¯ψ relates to the covariance matrix of the
884
+ conditional distributions of the static Schr¨odinger bridge ˆπ. That is to say,
885
+ ∇2 ¯ψ(y) = 1
886
+ T CovX∼ˆπy(X)
887
+ (32)
888
+ where ˆπy is (a version of) the conditional distribution of ˆπ that, in view of (8) has the following
889
+ form:
890
+ ˆπy(dx) = exp(−V ˆπy(x)))dx
891
+
892
+ exp(−V ˆπy(¯x))d¯x,
893
+ V ˆπy(x) := ϕ(x) + |x|2
894
+ 2T − xy
895
+ T .
896
+ (33)
897
+ 12
898
+
899
+ We shall give an independent proof of (32) under additional regularity assumptions at Proposition
900
+ 5.2 in the Appendix for the readers’ convenience. A consequence of (32) is that ¯ψ is convex and we
901
+ obtain from (31) that
902
+ κψ(r) ≥ αν − 1
903
+ T − r−1fL(r)
904
+ ∀r > 0.
905
+ (34)
906
+ This is a first crude weak semiconvexity bound on ψ upon which Theorem 1.2 improves by means
907
+ of a recursive argument. We show in the forthcoming Lemma how to deduce weak semiconvexity
908
+ of ψ from weak semiconcavity of ϕ. In the L = 0 setting, this step is carried out in [11] invoking
909
+ the Cramer-Rao inequality, whose application is not justified in the present more general setup.
910
+ Lemma 3.3. Assume that α > −1/T exists such that
911
+ ℓϕ(r) ≤ − 1
912
+ T + r−2F(α, r2)
913
+ ∀r > 0.
914
+ (35)
915
+ Then
916
+ κψ(r) ≥ αν − 1
917
+ T + G(α, 2)
918
+ 2T 2
919
+ − r−1fL(r)
920
+ ∀r > 0.
921
+ Proof. Recalling the definition of V ˆπy given at (33) we observe that the standing assumptions imply
922
+ ℓV ˆπy (r) ≤ r−2F(α, r2)
923
+ ∀r > 0.
924
+ (36)
925
+ In view of (32), we now proceed to bound VarX∼ˆπy(X1) from below for a given y, where we adopted
926
+ the notational convention X = (X1, . . . , Xd) for the components of random vectors. We first observe
927
+ that
928
+ VarX∼ˆπy(X1) ≥ EX∼ˆπy[VarX∼ˆπy(X1|X2, . . . , Xd)].
929
+ (37)
930
+ Moreover, upon setting for any z = (z2, . . . , zd)
931
+ V ˆπy,z(·) := V ˆπy(·, z),
932
+ ˆπy,z(dx) =
933
+ exp(−V ˆπy,z(x))dx
934
+
935
+ exp(−V ˆπy,z(¯x))d¯x
936
+ we have
937
+ VarX∼ˆπy(X1|X2 = z2, . . . , Xd = zd) = 1
938
+ 2
939
+
940
+ |x − ˆx|2ˆπy,z(dx)ˆπy,z(dˆx)
941
+ With this notation at hand, we find that, uniformly in z ∈ Rd−1,
942
+ 1 = 1
943
+ 2
944
+
945
+ (∂xV ˆπy,z(x) − ∂xV ˆπy,z(ˆx))(x − ˆx)ˆπy,z(dx)ˆπy,z(dˆx)
946
+ = 1
947
+ 2
948
+
949
+ ⟨∇V ˆπy(x, z) − ∇V ˆπy(ˆx, z), (x, z) − (ˆx, z)⟩ˆπy,z(dx)ˆπy,z(dˆx)
950
+ (36)
951
+ ≤ 1
952
+ 2
953
+
954
+ F(α, |x − ˆx|2)ˆπy,z(dx)ˆπy,z(dˆx)
955
+ ≤ 1
956
+ 2F(α, 2VarX∼ˆπy(X1|X2 = z2, . . . , Xd = zd))
957
+ where to establish the last inequality we used Lemma 3.2(i) and Jensen’s inequality. Since α >
958
+ −1/T , invoking again Lemma 3.2(i) we have that s �→ F(α, s) is non decreasing. But then, we get
959
+ from (37) and the last bound that
960
+ VarX∼ˆπy(X1) ≥ 1
961
+ 2G(α, 2),
962
+ ∀y ∈ Rd.
963
+ 13
964
+
965
+ Next, we observe that, because of the fact that if ϕ(·) satisfies (35) then so does ϕ(O·) for any
966
+ orthonormal matrix O, repeating the argument above yields
967
+ VarX∼ˆπy(⟨v, X⟩) ≥ 1
968
+ 2G(α, 2),
969
+ ∀y, v ∈ Rd s.t. |v| = 1.
970
+ In light of (32), this implies
971
+ ⟨v, ∇2 ¯ψ(y)v⟩ ≥ G(α, 2)
972
+ 2T
973
+ |v|2
974
+ ∀v, y ∈ Rd.
975
+ But then, since
976
+ ψ(y) = U ν(y) − |y|2
977
+ 2T +
978
+ ¯ψ(y)
979
+ T
980
+ we immediately obtain
981
+ κψ(r) ≥ αν − 1
982
+ T + G(α, 2)
983
+ 2T 2
984
+ − r−1fL(r)∀r > 0.
985
+ 3.3
986
+ Proof of Theorem 1.2
987
+ The proof is obtained combining the results of the former two sections through a fixed point argu-
988
+ ment.
989
+ Proof of Theorem 1.2. We define a sequence (αn)n≥0 via
990
+ α0 = αν − 1
991
+ T ,
992
+ αn = αν − 1
993
+ T + G(αn−1, 2)
994
+ 2T 2
995
+ ,
996
+ n ≥ 1.
997
+ Using Lemma 3.2(ii) and an induction argument, we obtain that α1 ≥ α0 and (αn)n≥0 is a non
998
+ decreasing sequence. If we denote by α∗ the limit, then by continuity of G(·, 2), we know that
999
+ α∗ > αν − 1/T and α∗ satisfies the fixed point equation (11). To conclude the proof, we show by
1000
+ induction that
1001
+ κψ(r) ≥ αn − r−1fL(r)
1002
+ ∀n ≥ 1.
1003
+ (38)
1004
+ The case n = 0 is (34). For the inductive step, suppose (38) holds for a given n. Then Lemma 3.1
1005
+ gives that
1006
+ ℓϕ(r) ≤ r−2F(αn, r2) − 1
1007
+ T
1008
+ ∀r > 0.
1009
+ But then, an application of Lemma 3.1 proves that for all r > 0 we have
1010
+ κψ(r) ≥ αν − 1
1011
+ T + G(αn, 2)
1012
+ 2T 2
1013
+ − r−1fL(r) = αn+1 − r−1fL(r).
1014
+ The proof of (9) is now finished. To conclude, we observe that (10) follows directly from (9) and
1015
+ Lemma 3.3.
1016
+ 14
1017
+
1018
+ 4
1019
+ Logarithmic Sobolev inequality for Schr¨odinger bridges
1020
+ This section is devoted to the proof of Theorem 1.3 and is structured as follows: we first recall
1021
+ known facts about logaithmic Sobolev inequalities and gradient estimates for diffusion semigroups
1022
+ whose proofs can be found e.g. in [2] and eventually prove at Lemma 4.1 a sufficient condition
1023
+ for the two-times distribution of a diffusion process to satisfy LSI. Though such a result may not
1024
+ appear surprising, we could not find it in this form in the existing literature. We then proceed to
1025
+ elucidate the connection between Schr¨odinger bridges and Doob h-transforms at Lemma 4.2, and
1026
+ then finally prove Theorem 1.3.
1027
+ Local LSIs and gradient estimates
1028
+ Let [0, T ] × Rd ∋ (t, x) �→ Ut(x) be continuous in the time
1029
+ variable and twice differentiable in the space variable with
1030
+ ⟨v, ∇2Ut(x)v⟩ ≥ αt|v|2
1031
+ ∀x, v ∈ Rd, t ∈ [0, T ]
1032
+ for some function αt uniformly bounded below. We consider the time-inhomogeneous semigroup
1033
+ (Ps,t)0≤s≤t≤T generated by the diffusion process whose generator at time t acts on smooth functions
1034
+ with bounded support as follows
1035
+ f �→ 1
1036
+ 2∆f − ⟨∇Ut, ∇f⟩.
1037
+ We now recall some basic fact about gradient estimates and local LSIs for the semigroup (Ps,t)0≤s≤t≤T .
1038
+ For time-homogeneous semigroups these facts are well known and can be found e.g. in [2]: the
1039
+ adaptation to the time-inhomogeneous setting is straightforward. The first result we shall need
1040
+ afterwards is the gradient estimate (see [2, Thm. 3.3.18])
1041
+ |∇Pt,T f|(x) ≤ Ct,T Pt,T (|∇f|)(x),
1042
+ Ct,T = exp
1043
+
1044
+
1045
+ � T
1046
+ t
1047
+ αsds
1048
+
1049
+ ,
1050
+ (39)
1051
+ that holds for all (t, x) ∈ [0, T ] × Rd and any continuously differentiable f. Moreover, the local
1052
+ logarithmic Sobolev inequalities (see [2, Thm. 5.5.2])
1053
+ (P0,T f log f)(x) − (P0,T f)(x) log(P0,T f)(x) ≤
1054
+ ˜C0,T
1055
+ 2
1056
+ P0,T (|∇f|2/f)(x),
1057
+ ˜C0,T =
1058
+ � T
1059
+ 0
1060
+ Ct,T dt
1061
+ (40)
1062
+ hold for all x ∈ Rd and all positive continuously differentiable f. In the next Lemma we show how
1063
+ obtain LSI for the joint law at times 0 and T of a diffusion process with initial distribution µ and
1064
+ drift −∇Ut, that is to say for the coupling π defined by
1065
+
1066
+ Rd×Rd f(x, y)π(dxdy) =
1067
+
1068
+ Rd P0,T f(x, ·)(x)µ(dx)
1069
+ ∀f > 0.
1070
+ (41)
1071
+ Lemma 4.1. Assume that µ satisfies LSI with constant Cµ and let π be as in (41). Then π satisfies
1072
+ LSI with constant
1073
+ max{Cµ, CµC0,T + ˜C0,T }.
1074
+ To proof is carried out ”mixing” carefully with the help of the gradient estimate the local
1075
+ (conditional) LSIs (40). Similar arguments and ideas can be found e.g. in [6, 27].
1076
+ 15
1077
+
1078
+ Proof. We recall the decomposition of the entropy formula (see [30, Thm. 2.4])
1079
+ Entπ(f) = Entµ(f0) +
1080
+
1081
+ Rd Entπx(f x)f0(x)µ(dx),
1082
+ where we adopted the following conventions
1083
+ f0(x) = (P0,T f(x, ·))(x),
1084
+ f x(y) = f(x, y)/f0(x),
1085
+
1086
+ g(y)πx(dy) =
1087
+
1088
+ P0,T g
1089
+
1090
+ (x) ∀g > 0.
1091
+ We know from (40) that uniformly in x, f we have
1092
+ Entπx(f x) = P0,T
1093
+
1094
+ f x log f x�
1095
+ (x) −
1096
+
1097
+ P0,T f x log P0,T f x�
1098
+ (x) ≤
1099
+ ˜C0,T
1100
+ 2f0(x)
1101
+
1102
+ |∇yf(x, y)|2/f(x, y)πx(dy).
1103
+ This gives
1104
+
1105
+ Entπx(f x)f0(x)µ(dx) ≤
1106
+ ˜C0,T
1107
+ 2
1108
+ � |∇yf(x, y)|2
1109
+ f(x, y)
1110
+ π(dxdy).
1111
+ (42)
1112
+ Next, we use LSI for µ to obtain
1113
+ Entµ(f0) ≤ Cµ
1114
+ 2
1115
+
1116
+ |∇xf0(x)|2/f0(x)µ(dx)
1117
+ = Cµ
1118
+ 2
1119
+
1120
+ |P0,T (∇xf(x, ·))(x)|2(P0,T f(x, ·))−1(x) µ(dx)
1121
+ + Cµ
1122
+ 2
1123
+
1124
+ |∇zP0,T (f(x, ·))(z)|2���
1125
+ z=x(P0,T f(x, ·))−1(x) µ(dx)
1126
+ (43)
1127
+ For the first summand on the rhs of (43), we can argue on the basis of Jensen’s inequality applied
1128
+ to the convex function a, b �→ a2/b to obtain
1129
+
1130
+ 2
1131
+
1132
+ |P0,T (∇xf(x, ·))(x)|2(P0,T f(x, ·))−1(x)µ(dx)
1133
+ ≤ Cµ
1134
+ 2
1135
+
1136
+ P0,T
1137
+
1138
+ |∇xf(x, ·)|2/f(x, ·)
1139
+
1140
+ (x)µ(dx)
1141
+ = Cµ
1142
+ 2
1143
+
1144
+ |∇xf(x, y)|2/f(x, y)π(dxdy).
1145
+ (44)
1146
+ For the second summand on the rhs of (43), we first invoke the gradient estimate (39) and eventually
1147
+ apply again Jensen’s inequality to obtain
1148
+
1149
+ 2
1150
+
1151
+ |∇zP0,T (f(x, ·))(z)|2���
1152
+ z=x(P0,T f(x, ·))−1(x)µ(dx)
1153
+ ≤ CµC0,T
1154
+ 2
1155
+
1156
+ (P0,T (|∇yf(x, ·)|)(x))2(P0,T f(x, ·))−1(x)µ(dx)
1157
+ ≤ CµC0,T
1158
+ 2
1159
+
1160
+ P0,T
1161
+
1162
+ |∇yf(x, ·)|2/f(x, ·)
1163
+
1164
+ (x)µ(dx)
1165
+ = CµC0,T
1166
+ 2
1167
+
1168
+ |∇yf(x, y)|2/f(x, y) π(dxdy).
1169
+ (45)
1170
+ Gathering (42)-(44)-(45) we obtain the desired result.
1171
+ 16
1172
+
1173
+ In the next lemma, we represent the static Schr¨odinger bridge (2) through a diffusion process.
1174
+ It is a classical result saying that Schr¨odinger bridges are indeed Doob’s h-transforms, see e.g. [31,
1175
+ Sec. 4][16].
1176
+ Lemma 4.2. Let Assumption 1.1 hold and ˆπ be the static Schr¨odinger bridge (2). Then ˆπ has the
1177
+ form (41), where (Ps,t)0≤s≤t≤T is the time-inhomogeneous semigroup associated with the generator
1178
+ acting on smooth test functions as follows
1179
+ f �→ 1
1180
+ 2∆f − ⟨∇U T,ψ
1181
+ t
1182
+ , ∇f⟩,
1183
+ t ∈ [0, T ],
1184
+ (46)
1185
+ where U T,ψ
1186
+ t
1187
+ has been defined at (13).
1188
+ Proof. Let ψ be the Schr¨odinger potential issued from (8) and denote by Q the law on C([0, T ]; Rd)
1189
+ of a solution (Xt)t∈[0,T ] to the stochastic differential equation
1190
+ dXt = −∇U T,ψ
1191
+ t
1192
+ (Xt)dt + dBt,
1193
+ X0 ∼ µ.
1194
+ Note that, because of Theorem 1.2, existence of strong solutions and pathwise uniqueness hold for
1195
+ the above equation. Next, we denote by P the Wiener measure with initial distribution µ. By
1196
+ Girsanov’s Theorem, see [29] for a version that applies in the current setting, we know that
1197
+ dQ
1198
+ dP (ω) = exp
1199
+
1200
+
1201
+ � T
1202
+ 0
1203
+ ∇U T,ψ
1204
+ t
1205
+ (ωt)dωt − 1
1206
+ 2
1207
+ � T
1208
+ 0
1209
+ |∇U T,ψ
1210
+ t
1211
+ (ωt)|2dt
1212
+
1213
+ P − a.s.,
1214
+ where we denote by ω the typical element of the canonical space C([0, T ]; Rd). Using Itˆo formula
1215
+ we rewrite the above as
1216
+ dQ
1217
+ dP (ω) = exp
1218
+
1219
+ U T,ψ
1220
+ 0
1221
+ (ω0) − U T,ψ
1222
+ T
1223
+ (ωT ) +
1224
+ � T
1225
+ 0
1226
+
1227
+ ∂tU T,ψ
1228
+ t
1229
+ + 1
1230
+ 2∆U T,ψ
1231
+ t
1232
+ − 1
1233
+ 2|∇U T,ψ
1234
+ t
1235
+ |2�
1236
+ (ωt)dt
1237
+
1238
+ = exp(U µ(ω0) − ϕ(ω0) − ψ(ωT ))
1239
+ where we used the Schr¨odinger system (8) and the HJB equation (15) to obtain the last expression.
1240
+ Indeed because of Theorem 1.2 one can deduce that [0, T ] × Rd ∋ (t, x) �→ U T,ψ
1241
+ t
1242
+ (x), is a classical
1243
+ solution of (15) by differentiating under the integral sign in (13). From this, we deduce that
1244
+ dQ0T
1245
+ dP0T
1246
+ (x, y) = exp
1247
+
1248
+ U µ(x) − ϕ(x) − ψ(y)
1249
+
1250
+ P0T − a.s.,
1251
+ where Q0T (resp. P0T ) denotes the joint distribution of Q (resp. P) at times 0 and T . Since
1252
+ dP0T (dxdy) = (2πT )−d/2 exp(−U µ(x)) exp
1253
+
1254
+ − |y − x|2
1255
+ 2T
1256
+
1257
+ dxdy,
1258
+ we conclude that
1259
+ dQ0T (dxdy) = (2πT )−d/2 exp
1260
+
1261
+ − ϕ(x) − ψ(y) − |y − x|2
1262
+ 2T
1263
+
1264
+ dxdy.
1265
+ But then Q0T = ˆπ, where ˆπ is defined at (2). To conclude, we observe that Q0T has the desired
1266
+ form (41) where (Ps,t)0≤s≤t is indeed the semigroup generated by (46)
1267
+ 17
1268
+
1269
+ Proof of Theorem 1.3. We know by Lemma 4.2 that ˆπ has the form (41) for the inhomogeneous
1270
+ semigroup generated by (46). We now set
1271
+ αψ
1272
+ t =
1273
+ inf
1274
+ x,v∈Rd,|v|=1⟨v, ∇2U ψ,T
1275
+ t
1276
+ (x), v⟩
1277
+ and proceed to estimate αψ
1278
+ t from below. To do so, we observe that U ψ,T
1279
+ t
1280
+ = U ψ,T −t
1281
+ 0
1282
+ and argue
1283
+ exactly as we did to establish (28) to obtain that
1284
+ κUT,ψ
1285
+ t
1286
+ (r) ≥
1287
+ αψ
1288
+ 1 + (T − t)αψ −
1289
+ r−1fL(r)
1290
+ (1 + (T − t)αψ)2
1291
+ ∀r > 0.
1292
+ From here, using the concavity of fL and f ′
1293
+ L(0) = L we obtain
1294
+ αψ
1295
+ t ≥
1296
+ αψ
1297
+ 1 + (T − t)αψ −
1298
+ L
1299
+ (1 + (T − t)αψ)2 .
1300
+ At this point, the conclusion follows from Lemma 4.1
1301
+ 5
1302
+ Appendix
1303
+ Proposition 5.1. Assume that U satisfies (4) for some α > 0, L′, R ≥ 0. Then
1304
+ κU(r) ≥ α − r−1fL(r)
1305
+ ∀r > 0.
1306
+ with L given by (7).
1307
+ Proof. If r > R the claim is a simple consequence of fL(r) ≥ 0. If r ≤ R, using (18) to get that
1308
+ r′ �→ r′−1fL(r′) is non increasing on (0, +∞), we obtain
1309
+ r−1fL(r) ≥ R−1fL(R) = L′,
1310
+ from which the conclusion follows.
1311
+ Proposition 5.2. Let Assumption 1.1 hold and assume furthermore that there exist ε, γ′ > 0 such
1312
+ that
1313
+
1314
+ exp(γ′|x|1+ε)µ(dx) < +∞.
1315
+ (47)
1316
+ Moreover, let ¯ψ be as in (31). Then ¯ψ is twice differentiable and we have
1317
+ ∇2 ¯ψ(y) = 1
1318
+ T CovX∼ˆπy(X)
1319
+ ∀y ∈ Rd,
1320
+ where ˆπy is given by (33).
1321
+ Proof. From (8) we obtain that
1322
+ ¯ψ(y) + d
1323
+ 2 log(π) = T log
1324
+
1325
+ Rd exp
1326
+
1327
+ − ϕ(x) − |x|2
1328
+ 2T + ⟨x, y⟩
1329
+ T
1330
+
1331
+ dx.
1332
+ (48)
1333
+ 18
1334
+
1335
+ From Assumption 1.1, (8) and (47) it follows that
1336
+
1337
+ Rd×Rd exp
1338
+
1339
+ γ′|x|1+ε − ϕ(x) − ψ(y) − |x − y|2
1340
+ 2T
1341
+
1342
+ dx dy < +∞,
1343
+ whence the existence of some y′ such that
1344
+
1345
+ Rd×Rd exp
1346
+
1347
+ γ′|x|1+ε − ϕ(x) − |x|2
1348
+ 2T + ⟨x, y′⟩
1349
+ T
1350
+
1351
+ dx < +∞.
1352
+ From this, we easily obtain that for all γ < γ′
1353
+
1354
+ Rd exp
1355
+
1356
+ γ|x|1+ε − ϕ(x) − |x|2
1357
+ 2T + ⟨x, y⟩
1358
+ T
1359
+
1360
+ dx < +∞
1361
+ ∀y ∈ Rd.
1362
+ (49)
1363
+ Thanks to (49) we can apply the dominated convergence theorem and differentiate under the integral
1364
+ sign in (15) to obtain that ¯ψ is differentiable and
1365
+ ∇ ¯ψ(y) =
1366
+
1367
+ x exp(−ϕ(x) − |x|2
1368
+ 2T + ⟨x,y⟩
1369
+ T
1370
+ )dx
1371
+
1372
+ exp(−ϕ(¯x) − |¯x|2
1373
+ 2T + ⟨¯x,y⟩
1374
+ T
1375
+ )d¯x
1376
+ (33)
1377
+ = EX∼ˆπy[X]
1378
+ Using once again (49) to differentiate under the integral sign in (48) we conclude that ¯ψ is twice
1379
+ differentiable and that (32) holds.
1380
+ References
1381
+ [1] Shigeki Aida, Takao Masuda, and Ichiro Shigekawa.
1382
+ Logarithmic sobolev inequalities and
1383
+ exponential integrability. Journal of Functional Analysis, 126(1):83–101, 1994.
1384
+ [2] Dominique Bakry, Ivan Gentil, and Michel Ledoux. Analysis and geometry of Markov diffusion
1385
+ operators, volume 348. Springer Science & Business Media, 2013.
1386
+ [3] Erhan Bayraktar, Stephan Eckstein, and Xin Zhang.
1387
+ Stability and sample complexity of
1388
+ divergence regularized optimal transport. arXiv preprint arXiv:2212.00367, 2022.
1389
+ [4] Jean-David Benamou.
1390
+ Optimal transportation, modelling and numerical simulation.
1391
+ Acta
1392
+ Numerica, 30:249–325, 2021.
1393
+ [5] Jean-David Benamou, Guillaume Carlier, Marco Cuturi, Luca Nenna, and Gabriel Peyr´e. Iter-
1394
+ ative Bregman projections for regularized transportation problems. SIAM Journal on Scientific
1395
+ Computing, 37(2):A1111–A1138, 2015.
1396
+ [6] Th Bodineau and B Helffer. The log-sobolev inequality for unbounded spin systems. Journal
1397
+ of functional analysis, 166(1):168–178, 1999.
1398
+ [7] H.J. Brascamp and E.H. Lieb. On extensions of the Brunn-Minkowski and Pr´ekopa-leindler
1399
+ theorems, including inequalities for log concave functions, and with an application to the
1400
+ diffusion equation. Journal of Functional Analysis, 22(4):366–389, 1976.
1401
+ 19
1402
+
1403
+ [8] Luis A Caffarelli. Monotonicity properties of optimal transportation and the FKG and related
1404
+ inequalities. Communications in Mathematical Physics, 214(3):547–563, 2000.
1405
+ [9] Y. Chen, T. Georgiou, and M. Pavon.
1406
+ On the relation between optimal transport and
1407
+ Schr¨odinger bridges: A stochastic control viewpoint. preprint arXiv:1412.4430, 2014.
1408
+ [10] Yongxin Chen, Tryphon T Georgiou, and Michele Pavon. Stochastic Control Liaisons: Richard
1409
+ Sinkhorn Meets Gaspard Monge on a Schrodinger Bridge. SIAM Review, 63(2):249–313, 2021.
1410
+ [11] Sinho Chewi and Aram-Alexandre Pooladian. An entropic generalization of caffarelli’s con-
1411
+ traction theorem via covariance inequalities. arXiv preprint arXiv:2203.04954, 2022.
1412
+ [12] Alberto Chiarini, Giovanni Conforti, Giacomo Greco, and Luca Tamanini. Gradient estimates
1413
+ for the Schr¨odinger potentials: convergence to the Brenier map and quantitative stability.
1414
+ arXiv preprint arXiv:2207.14262, 2022.
1415
+ [13] Giovanni Conforti. A second order equation for Schr¨odinger bridges with applications to the
1416
+ hot gas experiment and entropic transportation cost. Probability Theory and Related Fields,
1417
+ 174(1-2):1–47, 2019.
1418
+ [14] Giovanni Conforti. Coupling by reflection for controlled diffusion processes: turnpike property
1419
+ and large time behavior of Hamilton Jacobi Bellman equations. Annals of Applied Probability
1420
+ (to appear), 2022.
1421
+ [15] Marco Cuturi. Sinkhorn distances: Lightspeed computation of optimal transport. In Advances
1422
+ in Neural Information Processing Systems, pages 2292–2300, 2013.
1423
+ [16] P. Dai Pra and M. Pavon. On the Markov processes of Schr¨odinger, the Feynman-Kac formula
1424
+ and stochastic control. In Realization and Modelling in System Theory, volume 3, pages 497–
1425
+ 504. Springer, 1990.
1426
+ [17] Valentin De Bortoli, Arnaud Doucet, Jeremy Heng, and James Thornton. Simulating diffusion
1427
+ bridges with score matching. arXiv preprint arXiv:2111.07243, 2021.
1428
+ [18] Valentin De Bortoli, James Thornton, Jeremy Heng, and Arnaud Doucet. Diffusion schr¨odinger
1429
+ bridge with applications to score-based generative modeling. Advances in Neural Information
1430
+ Processing Systems, 34, 2021.
1431
+ [19] George Deligiannidis, Valentin De Bortoli, and Arnaud Doucet. Quantitative uniform stability
1432
+ of the iterative proportional fitting procedure. arXiv preprint arXiv:2108.08129, 2021.
1433
+ [20] Hacene Djellout, Arnaud Guillin, and Liming Wu. Transportation cost-information inequalities
1434
+ and applications to random dynamical systems and diffusions.
1435
+ The Annals of Probability,
1436
+ 32(3B):2702–2732, 2004.
1437
+ [21] Joseph Doob. Conditional Brownian motion and the boundary limits of harmonic functions.
1438
+ Bulletin de la Soci´et´e Math´ematique de France, 85:431–458, 1957.
1439
+ [22] Andreas Eberle. Reflection couplings and contraction rates for diffusions. Probability Theory
1440
+ and Related Fields, 166(3-4):851–886, 2016.
1441
+ 20
1442
+
1443
+ [23] Stephan Eckstein and Marcel Nutz. Quantitative stability of regularized optimal transport
1444
+ and convergence of sinkhorn’s algorithm. arXiv preprint arXiv:2110.06798, 2021.
1445
+ [24] Max Fathi, Nathael Gozlan, and Maxime Prodhomme. A proof of the caffarelli contraction
1446
+ theorem via entropic regularization. Calculus of Variations and Partial Differential Equations,
1447
+ 59(96), 2020.
1448
+ [25] Ivan Gentil, Christian L´eonard, and Luigia Ripani.
1449
+ Dynamical aspects of the generalized
1450
+ schr¨odinger problem via otto calculus–a heuristic point of view. Revista Matem´atica Iberoamer-
1451
+ icana, 36(4):1071–1112, 2020.
1452
+ [26] Promit Ghosal and Marcel Nutz. On the convergence rate of sinkhorn’s algorithm. arXiv
1453
+ preprint 2212.06000, 2022.
1454
+ [27] Natalie Grunewald, Felix Otto, C´edric Villani, and Maria G Westdickenberg.
1455
+ A two-scale
1456
+ approach to logarithmic sobolev inequalities and the hydrodynamic limit. In Annales de l’IHP
1457
+ Probabilit´es et statistiques, volume 45, pages 302–351, 2009.
1458
+ [28] Richard Holley and Daniel W Stroock. Logarithmic sobolev inequalities and stochastic ising
1459
+ models. 1986.
1460
+ [29] Christian L´eonard. Girsanov theory under a finite entropy condition. In C. Donati-Martin, A.
1461
+ Lejay, and A. Rouault, editors, S´eminaire de Probabilit´es XLIV, volume 2046 of Lecture Notes
1462
+ in Mathematics, pages 429–465. Springer, 2012.
1463
+ [30] Christian L´eonard. Some properties of path measures. In S´eminaire de Probabilit´es XLVI,
1464
+ pages 207–230. Springer, 2014.
1465
+ [31] Christian L´eonard. A survey of the Schr¨odinger problem and some of its connections with
1466
+ optimal transport. Discrete and Continuous Dynamical Systems, 34(4):1533–1574, 2014.
1467
+ [32] Torgny Lindvall and L Cris G Rogers. Coupling of multidimensional diffusions by reflection.
1468
+ The Annals of Probability, pages 860–872, 1986.
1469
+ [33] Dan Mikulincer and Yair Shenfeld. On the Lipschitz properties of transportation along heat
1470
+ flows. arXiv preprint arXiv:2201.01382, 2022.
1471
+ [34] Marcel Nutz and Johannes Wiesel. Entropic optimal transport: Convergence of potentials.
1472
+ Probability Theory and Related Fields, 184(1):401–424, 2022.
1473
+ [35] Gabriel Peyr´e and Marco Cuturi. Computational optimal transport. Foundations and Trends
1474
+ in Machine Learning, 11(5-6):355–607, 2019.
1475
+ [36] Erwin Schr¨odinger.
1476
+ ¨Uber die Umkehrung der Naturgesetze. Sitzungsberichte Preuss. Akad.
1477
+ Wiss. Berlin. Phys. Math., 144:144–153, 1931.
1478
+ [37] Yuyang Shi, Valentin De Bortoli, George Deligiannidis, and Arnaud Doucet.
1479
+ Conditional
1480
+ simulation using diffusion schr¨odinger bridges. arXiv preprint arXiv:2202.13460, 2022.
1481
+ 21
1482
+
DtAyT4oBgHgl3EQfSPej/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
ENE1T4oBgHgl3EQfEQP0/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0304c4e22e91c4882af1b1b0a979b109104530de11aa39e6ff23e791bc9a040
3
+ size 238394
EdE3T4oBgHgl3EQfVQpt/content/tmp_files/2301.04458v1.pdf.txt ADDED
@@ -0,0 +1,1536 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Femtosecond laser-induced sub-wavelength plasma inside dielectrics: III.
2
+ Terahertz radiation emission
3
+ Kazem Ardaneh,1, 2, a) Ken-Ichi Nishikawa,3 Remo Giust,1 Benoit Morel,1 Pierre-Jean Charpin,1 Arnaud
4
+ Couairon,4 Guy Bonnaud,5 and Francois Courvoisier1, b)
5
+ 1)FEMTO-ST Institute, Univ. Bourgogne Franche-Comt´e, CNRS, 15B avenue des Montboucons, 25030,
6
+ Besan¸con Cedex, France
7
+ 2)Sorbonne University, Pierre and Marie Curie Campus, 4 place Jussieu, 75252, Paris Cedex 5,
8
+ France
9
+ 3)Department of Physics, Chemistry and Mathematics, V. Murry Chambers Bld., Alabama A&M University,
10
+ Huntsville, AL 35810, USA
11
+ 4)CPHT, CNRS, Ecole Polytechnique, Institut Polytechnique de Paris, Route de Saclay, F-91128 Palaiseau,
12
+ France
13
+ 5)CEA, Centre de Paris-Saclay, DRF, Univ.
14
+ Paris-Saclay, 91191 Gif-sur-Yvette,
15
+ France
16
+ (Dated: 12 January 2023)
17
+ Electromagnetic radiation within the terahertz (THz) frequency range is of great interest for applications
18
+ in remote sensing and time-domain spectroscopy.
19
+ The laser-induced plasmas are promising mediums for
20
+ generating THz radiation.
21
+ It has been recently reported that focusing femtosecond Bessel pulses inside
22
+ dielectrics induces a high aspect ratio over-critical plasmas. Here we show that the intense resonantly driven
23
+ electrostatic fields at the so-called critical surface lead to THz radiation emission. Through three-dimensional
24
+ particle-in-cell simulation and analytical derivation, we have investigated the emission of THz radiation. We
25
+ show that the THz radiation is associated with a hot population of electrons trapped in ambipolar electric
26
+ fields of the double layers.
27
+ I.
28
+ INTRODUCTION
29
+ Terahertz (THz) radiation, typically referred to as the
30
+ frequency band 100 GHz − 10 THz, in the infrared and
31
+ microwaves ranges, has been attracting ongoing interest
32
+ because of its broad applications ranging from biomedical
33
+ imaging, security or packaged goods inspection, to time-
34
+ domain spectroscopy.1–4 Using femtosecond laser pulses,
35
+ the techniques for generating THz radiation are gen-
36
+ erally classified as either optical rectification,5–11 tran-
37
+ sient current sources,2,12–23 or a combination of these two
38
+ mechanisms.24
39
+ Optical rectification can induce THz radiation in
40
+ non-centrosymmetric crystals, e.g., ZnTe and GaAs, in
41
+ which the fundamental frequency of an infrared fem-
42
+ tosecond laser pulse is down-converted to the THz fre-
43
+ quency via the second-order susceptibility where the po-
44
+ larization reads P (ωTHz) = χ(2)E(ω + ωTHz)E∗(ω).5,7
45
+ The frequency of the rectified pulse envelope is in
46
+ the range of 3-10 THz.7 In the centrosymmetric me-
47
+ dia, e.g., gases, two-color illumination can be used
48
+ to mix the fundamental frequency with the second-
49
+ harmonic in a four-wave mixing process as P (ωTHz) =
50
+ χ(3)E(2ω −ωTHz) E∗(ω)E∗(ω).8,11 Correspondingly, the
51
+ THz component might match with the fundamental one
52
+ in an inverse process and leads to second-harmonic gen-
53
+ eration, which is a method for THz detection. A THz
54
+ emission with an estimated field strength ∼ 400 kV/cm
55
+ a)Electronic mail: kazem.arrdaneh@gmail.com
56
+ b)Electronic mail: francois.courvoisier@femto-st.fr
57
+ has been reported in which the presence of plasma was
58
+ essential for the high-efficiency process.10
59
+ Laser-induced plasmas are attractive for THz radia-
60
+ tion generation because of their ability to sustain ex-
61
+ tremely intense electromagnetic fields.13,14,25–27 In this
62
+ context, femtosecond laser-induced breakdown of gases
63
+ is investigated widely,28,29 mostly using the two-color
64
+ approach.8 The peak of the THz field however satu-
65
+ rates for laser intensities higher than 1015 W/cm2 be-
66
+ cause of the strong THz absorption in the long (∼7 mm)
67
+ air plasma.17 Plasma generation in laser-solid interac-
68
+ tions offer an alternative: experiments of ultrashort laser
69
+ pulses-solid interaction have shown a monotonous in-
70
+ crease in THz radiation with the incident laser intensity
71
+ up to 1019 W/cm2.18,19,21
72
+ Illumination of solid targets by intense ultrashort laser
73
+ beams results in the generation of hot electron currents
74
+ that are the source of the secondary electromagnetic ra-
75
+ diation ranging from x-rays30–33 to THz radiation.19 Un-
76
+ der p−polarized laser illumination of a short-scale inho-
77
+ mogeneous plasma, for moderate laser intensities (bellow
78
+ 1014 W/cm2), resonance absorption is the main mecha-
79
+ nism for hot electron generation.34–36
80
+ In previous papers, we have reported that over-critical
81
+ plasmas, i.e. with density above the reflection density for
82
+ the incident laser wavelength, are generated by focusing
83
+ Bessel beams with moderate intensities on the order of
84
+ 1014 W/cm2 inside sapphire.37–39 A Bessel beam is a so-
85
+ lution of the wave equation in which the wave amplitude
86
+ is defined by the Bessel function of the first kind.40 Im-
87
+ portantly, the axial intensity profile of the Bessel beam is
88
+ propagation invariant. Therefore, all segments of the di-
89
+ electric along the Bessel zone will receive simultaneously
90
+ arXiv:2301.04458v1 [physics.plasm-ph] 11 Jan 2023
91
+
92
+ 2
93
+ FIG. 1. The energy spectrum of the radiation emitted by a population of hot electrons: (a) the spatially averaged, (b) angular
94
+ distribution, and (c) spatial distribution averaged in the frequency range of 1-30 rad/ps.
95
+ The hot electrons are randomly
96
+ selected.
97
+ the same amount of energy which results in a high aspect
98
+ ratio plasma rod.
99
+ In the first article of this series (Ardaneh et al.,38 Pa-
100
+ per I hereafter), we confirmed that the resonance of the
101
+ plasma waves can explain the experimental diagnostics
102
+ of total absorption, and far-field intensity pattern. We
103
+ reported electron acceleration up to several keV while
104
+ surfing the plasma waves. In the outward propagation of
105
+ hot electrons, electrostatic ambipolar fields form at the
106
+ plasma surface due to the different inertia of the electron
107
+ and ion. Moreover, in the second article of this series
108
+ (Ardaneh et al.,39 Paper II hereafter), we reported the
109
+ second-harmonic generation by a second-order current
110
+ of hot electrons near the critical surface. The electron
111
+ currents form by the resonance absorption and radiation
112
+ force of the incident laser wave.
113
+ In the current work as the third in this series, we estab-
114
+ lish a link between resonance absorption-driven currents
115
+ and THz radiation. This is based on calculating the co-
116
+ herent radiation spectrum of the hot electrons for the
117
+ performed Particle-In-Cell (PIC) simulation. The sim-
118
+ ulation consists of electron-ion plasma initially induced
119
+ by multi-photon and collisional ionizations. The dipole
120
+ moments are induced due to the radiation force of the
121
+ resonance fields. For a laser field with frequency ω0, this
122
+ force induces a second-harmonic component at 2ω0 and
123
+ a low-frequency component by separating the light elec-
124
+ trons from the heavy ions. We have developed an ana-
125
+ lytical model for THz generation in laser-plasma inter-
126
+ actions to explain the underlying physics, in particular,
127
+ how the dipole moment is created in the plasma and the
128
+ characteristics of the radiation spectrum.
129
+ We organized the paper as follows. In Sec. II, we re-
130
+ call the setup of the PIC simulation as discussed in Paper
131
+ I38, we detail the radiation diagnostic, and the results of
132
+ the simulation. Then, in Sec. III, we derive an analyti-
133
+ cal solution for the current source of THz radiation, and
134
+ the radiated electromagnetic fields with their frequency
135
+ spectrums.
136
+ TABLE I. Simulation setup.
137
+ Parameter
138
+ Value
139
+ Simulation volume
140
+ 15 × 15 × 30 µm3
141
+ Grid resolution
142
+ ∆x:yk0
143
+ r = 0.04a, ∆zk0
144
+ z = 0.1b
145
+ FDTD order
146
+ Second-order
147
+ BC for fieldsc
148
+ Perfectly matched layers
149
+ BC for particles
150
+ Outflow
151
+ Pulse energy (Ep)
152
+ 1.2 µJ
153
+ Pulse frequency (ω0)
154
+ 2.35 rad/fs
155
+ Pulse cone angle (θ)
156
+ 25◦
157
+ Pulse temporal profile
158
+ exp[−(t − tc)2/T 2]
159
+ Central time (tc)
160
+ 130 fs
161
+ Pulse FWHM =
162
+
163
+ 2 ln 2T
164
+ 100 fs
165
+ Pulse spatial profile
166
+ exp(−r2/w2
167
+ 0)
168
+ Pulse spatial waist (w0)
169
+ 10 µm
170
+ Maximum density (nmax)
171
+ 5 nc
172
+ Density profile (axial)
173
+ tanh(zµm)
174
+ Mass ratio (mi/me)
175
+ 102 × 1836d
176
+ Plasma distribution [f(ve:i)] Maxwellian
177
+ Plasma temperature (Te:i)
178
+ 1 eV
179
+ Particles per cell per species 32
180
+ Particle shape function
181
+ triangle
182
+ Time step
183
+ ∆tω0 = 0.07
184
+ Simulation time
185
+ 320 fs
186
+ a k0
187
+ r = k0 sin θ.
188
+ b k0
189
+ z = k0 cos θ.
190
+ c BC: Boundary condition.
191
+ d 102 is the sapphire molar mass.
192
+ II.
193
+ PIC SIMULATION
194
+ We performed self-consistent PIC simulation using
195
+ the three-dimensional massively parallel electromag-
196
+ netic code EPOCH41.
197
+ In our simulation, we used
198
+ the plasma parameters that could reproduce our ex-
199
+ perimental measurements (far-field, near-field, absorp-
200
+
201
+ dW(WTHz)/dQ[a.u. ]
202
+ dW(WTHz)/dΩ[a. u. ]
203
+ 0.25
204
+ 0.50
205
+ 0.75
206
+ 1.00
207
+ 0.25
208
+ 0.50
209
+ 0.75
210
+ 1.00
211
+ 180
212
+ 1
213
+ 100↓
214
+ (a)
215
+ (b)
216
+ (c)
217
+ u.
218
+ 135
219
+ esin Φ
220
+ 10-1.
221
+ deg]
222
+ dW/dw [
223
+ 90
224
+ 0
225
+ sin(
226
+ 10-2
227
+ 45 -
228
+ 10-3
229
+ 0
230
+ -1
231
+ -2
232
+ 0
233
+ 90
234
+ 180
235
+ 270
236
+ 360
237
+ -1
238
+ 0
239
+ 2
240
+ 0
241
+ [0m]3
242
+ Φ[deg]
243
+ sin Ocos Φ3
244
+ tion) as reported in Paper I,38 and II.39 The simula-
245
+ tion setup is summarized in Table I. The plasma is
246
+ fully ionized and composed of electrons and ions with
247
+ equal densities (to preserve electric neutrality) given
248
+ by n = nmax exp(−x2/w2
249
+ x) exp(−y2/w2
250
+ y) tanh(zµm) with
251
+ FWHMx =
252
+
253
+ 2 ln 2wx = 250 nm, and FWHMy
254
+ =
255
+
256
+ 2 ln 2wy = 600 nm.
257
+ There are initially 32 particles
258
+ per cell per species leading to the total number of parti-
259
+ cles in the simulation ∼ 109. The collisions are modeled
260
+ through a binary model as presented in Refs.41,42.
261
+ We
262
+ injected
263
+ from
264
+ the
265
+ zmin
266
+ boundary
267
+ a
268
+ linearly
269
+ x−polarized Gaussian pulse propagating along the pos-
270
+ itive z−direction. We applied a phase to the Gaussian
271
+ beam to create a Bessel-Gauss beam.43 The peak inten-
272
+ sity in the Bessel zone is 6 × 1014 W/cm2 in absence of
273
+ plasma. The time step is limited by the Courant con-
274
+ dition.
275
+ The minimum frequency in the simulation is
276
+ 1.5 rad/ps which is well below the peak frequency of the
277
+ THz spectrum at 30 rad/ps.
278
+ One of the primary advantages of PIC codes is the pos-
279
+ sibility to access full information about the particles. We
280
+ have developed a radiation diagnostic that utilizes the
281
+ position and momentum of particles over time and cal-
282
+ culates the radiated fields and energy. For this purpose,
283
+ let us consider a particle at position r (t) at time t. At the
284
+ same time, we observe the radiated electromagnetic fields
285
+ from the particle at position x. Due to the finite velocity
286
+ of light, we observe the particle at an earlier position r (t′)
287
+ where it was at the retarded time t′ = t−R (t′) /c, where
288
+ R (t′) = |x − r (t′)| is the distance from the charged par-
289
+ ticle (at the retarded time t′) to the observer. The mag-
290
+ netic and electric fields produced from a moving point
291
+ charge can be calculated directly from their scalar and
292
+ vector potentials known as the Li´enard–Wiechert poten-
293
+ tials. The electric field reads:44
294
+ E(x, t) =
295
+ Velocity field
296
+
297
+ ��
298
+
299
+ e
300
+
301
+ n − β
302
+ γ2(1 − β · n)3R2
303
+
304
+ ret
305
+ +
306
+ Acceleration field
307
+
308
+ ��
309
+
310
+ e
311
+ c
312
+
313
+ n × {(n − β) × ˙β}
314
+ (1 − β · n)3R
315
+
316
+ ret
317
+ (1)
318
+ where n = R (t′) / |R (t′)| is a unit vector pointing
319
+ from the particle retarded position to the observer, β =
320
+ v/c the particle instantaneous velocity,
321
+ ˙β = dβ/dt is
322
+ the acceleration divided by c, γ is the Lorentz fac-
323
+ tor. The spatial spectra are obtained by the choice of
324
+ n
325
+
326
+ n2
327
+ x + n2
328
+ y + n2
329
+ z = 1
330
+
331
+ . The field in Eq. (1) divides itself
332
+ into ”velocity fields,” which are independent of acceler-
333
+ ation, and ”acceleration fields,” which depend linearly
334
+ on ˙β. The velocity field is a static field decreasing as
335
+ R−2 while the acceleration field is a radiation field, be-
336
+ ing transverse to the radius vector and falling off as R−1.
337
+ The total energy W radiated per unit solid angle dΩ per
338
+ unit frequency dω from the accelerated charged particle
339
+ reads:44
340
+ d2W
341
+ dωdΩ = e2ω2
342
+ 4π2c
343
+ ����
344
+ � ∞
345
+ −∞
346
+ dt′ˆn × (ˆn × β)ejω(t′+R(t′)/c)
347
+ ����
348
+ 2
349
+ (2)
350
+ In our simulations, we collected the THz radiation
351
+ emissions from the hot electrons as follows.
352
+ We have
353
+ tracked 105 electrons in the simulations and recorded the
354
+ information of these electrons. We calculated the energy
355
+ spectrum of the radiation emitted by 100 randomly se-
356
+ lected electrons according to Eq. (2). The result is shown
357
+ in Fig.
358
+ 1(a).
359
+ We see two peaks around ω = 0, with
360
+ a width of typically 50 rad/ps. We note that, because
361
+ of computing memory limitations, the time resolution of
362
+ the particle positions is insufficient to capture the second
363
+ harmonic emission.
364
+ The angular and spatial distribu-
365
+ tions of the THz emission are obtained by averaging the
366
+ energy spectrum in the frequency range of 1-30 rad/ps
367
+ [Figs. 1(b) and 1(c)].
368
+ As one expects, the energy spectrum has sharp max-
369
+ ima at the laser frequency ω0 due to strong electron ac-
370
+ celeration in resonantly driven plasma waves at the crit-
371
+ ical surfaces.
372
+ The spectrum also has maxima at ω ≈
373
+ 30 rad/ps. The angular distribution of this THz radia-
374
+ tion in Fig. 1(b) shows maxima around (θ, φ) ≈ (0, π/2)
375
+ and (0, 3π/2), perpendicular to the electron acceleration
376
+ which is mainly in the x−direction.
377
+ The small tilt in
378
+ Fig. 1(c) is due to the asymmetric distribution of the
379
+ randomly selected electrons in xy−space over the inte-
380
+ grated time (a similar deviation occurred for another set
381
+ of 100 electrons).
382
+ We select some representative electrons to calculate
383
+ the radiated fields in a spatial window |x| ⩽ 45 µm and
384
+ |y| ⩽ 45 µm at z = 0. Using a time window, we also ex-
385
+ amined in which part of the trajectory, the electron emits
386
+ electromagnetic radiation in the THz frequency range
387
+ [Figs. 2(a)]. For each time window, we calculated the
388
+ intensity distribution I(ωTHz, x, y) by performing a dis-
389
+ crete Fourier transform on each component of the electric
390
+ field, Ex:y:z(t, x, y) and averaging in the frequency range
391
+ of 1-30 rad/ps [Figs. 2(b)-2(f)].
392
+ Figure 2(a) shows the time evolution of the Ex com-
393
+ ponent, parallel to the incident laser polarization over-
394
+ plotted with the trajectory of a representative electron.
395
+ One can see the resonance plasma waves induced at the
396
+ critical surfaces (x = ±0.2 µm), in the time between 70-
397
+ 200 fs (See Fig. 4 in Paper I38 for more details). Near
398
+ the peak of the laser field, the intense ambipolar fields
399
+ propagating with the sound speed are generated at the
400
+ surface of the plasma (dashed lines). The ambipolar field
401
+ sign is positive for x > 0 and negative for x < 0. The ra-
402
+ diation force (See Appendix A) due to the intense, local-
403
+ ized resonance field ejects electrons from the resonance
404
+ region. The electrons are ejected from the critical sur-
405
+ faces in the positive x−direction where x > 0 and nega-
406
+ tive x−direction where x < 0 as shown in Fig. 6 Paper I.
407
+ Therefore, the electrons ejected with energy less than the
408
+ potential barrier of the ambipolar field will be reflected by
409
+
410
+ 4
411
+ FIG. 2. THz radiation from an electron trapped in the ambipolar electric fields of double layers. Shown are: (a) x−component
412
+ of the electric field over-plotted by the trajectory of a representative electron, electron emission for a time window between:
413
+ (b) 20-150 fs [shown by blue � symbols in panel (a), and blue line in panel (b)], and 182-312 fs [shown by red � symbols in
414
+ panel (a) and red line in panel (b)], (c-f) cE2
415
+ x/8π, cE2
416
+ y/8π, cE2
417
+ z /8π and the total intensity of the THz radiation emitted by
418
+ the electron for the time window between 182-224 fs. The dashed lines in panel (a) show the expansion of the plasma at the
419
+ sound velocity. The color in the electron trajectory reflects its energy based on the color bar of the panel (b).
420
+ −eE force. Consequently, these electrons will be trapped
421
+ between the ambipolar electric fields on either side of the
422
+ plasma. An ejected electron oscillates between the am-
423
+ bipolar fields with a period that increases with time due
424
+ to the energy exchange between the electrons and ions.
425
+ We monitored the electron radiation using a time win-
426
+ dow of 130 fs. The electron emission between 20-150 fs
427
+ [shown by blue � symbols in panel (a)], is sharply
428
+ peaked at the laser frequency ω0 [blue line in panel (b)].
429
+ This emission is due to the electron acceleration while
430
+ surfing the resonantly driven plasma waves (See Paper
431
+ I38 for more details). During the time interval 182-312 fs
432
+ [shown by red � symbols in panel (a)], the electron is
433
+ trapped between two ambipolar fields and emits the THz
434
+ radiation with a peak frequency at ω = 24 rad/ps [red
435
+ line in panel (b)]. Figures 2(c-f) show respectively the
436
+ intensity of the electric field components computed using
437
+ Eq. (1), c(E2
438
+ x, E2
439
+ y, E2
440
+ z )/8π, and the total intensity radi-
441
+ ated by the electron during the time between 182-312 fs.
442
+ The THz radiation is mainly polarized in the x−direction
443
+ because Ex component is dominant in the radiated field.
444
+ This polarization is the same as the incident pulse and
445
+ second-harmonic detailed in Paper II39. In Sec. III, we
446
+ will show that the emission pattern corresponds with an
447
+ electron current in the x−direction.
448
+ III.
449
+ ELECTRON THZ EMISSION IN AMBIPOLAR
450
+ ELECTRIC FIELDS
451
+ The starting point in understanding the mechanism
452
+ responsible for THz radiation is the identification of its
453
+ current sources. We have seen earlier, in Figs. 2, that
454
+ the electrons emit THz radiation while they are trapped
455
+ in the ambipolar electric fields of
456
+ plasma double lay-
457
+ ers. Taking the strength of the resonance electric field
458
+ of about 50 GV/m integrated over its width of 70 nm
459
+ (See Paper I38), one arrives at a potential of about a few
460
+ keV which corresponds to the temperature of the hottest
461
+ electrons in the simulation. The hot electrons propagate
462
+ outside of the plasma and consequently, the separation of
463
+ charges forms an electric double layer where an ambipo-
464
+
465
+ Ex[GV/m]
466
+ Energy[keV]
467
+ -50
468
+ 0
469
+ 50 0
470
+ 2
471
+ 4
472
+ (a)
473
+ (b)
474
+ -1.0
475
+ 0.8'
476
+ [μm]
477
+ 0.6
478
+ 0
479
+ xy
480
+ X
481
+ 0.4
482
+ 0.2
483
+ -1
484
+ 0.0
485
+ 50
486
+ 100
487
+ 150
488
+ 200
489
+ 250
490
+ 300
491
+ -1
492
+ 0
493
+ 0
494
+ 1
495
+ 2
496
+ t[fs]
497
+ [m]m
498
+ 0.0
499
+ 2.0
500
+ 4.00.0
501
+ 0.2
502
+ 0.50.0
503
+ 0.2
504
+ 0.50.0
505
+ 2.0
506
+ 4.0
507
+ (c)
508
+ (d)
509
+ (e)
510
+ (f)
511
+ 30
512
+ [un]
513
+ 0o
514
+ 0
515
+ y
516
+ -30
517
+ cE2/8r [W/m?]
518
+ cE/8π [W/m²]
519
+ cE3/8π [W/m2]
520
+ I [W/m?]
521
+ 0
522
+ -30
523
+ -30
524
+ 0
525
+ -30
526
+ 0
527
+ 0
528
+ 30
529
+ 30
530
+ 30
531
+ -30
532
+ 30
533
+ x[μm]
534
+ x[μm]
535
+ x[μm]
536
+ x[μm]5
537
+ lar electric field is present. An analytical solution for this
538
+ field is possible by using the two-fluid plasma equations
539
+ for continuity and momentum (See Appendix B).
540
+ Here for simplicity,
541
+ we considered a s−polarized
542
+ monochromatic laser wave as E = Es(r) cos(ω0t) where
543
+ Es(r) includes the spatial dependence.
544
+ The ambipo-
545
+ lar electric field Ea is described by an inhomogeneous
546
+ second-order differential equation for a classical, damped,
547
+ driven harmonic oscillator given by [See Eq. (B7) in Ap-
548
+ pendix B]:
549
+ ∂2
550
+ t Ea + 2Γ∂tEa + Ω2
551
+ pEa = Ω2
552
+ p [E0 + E2 cos(2ω0t)]
553
+ (3)
554
+ where Γ = νei/2 (1 + Zme/mi), νei is the electron-ion
555
+ collision frequency, Ω2
556
+ p
557
+ = ω2
558
+ pe (1 + Zme/mi), ωpe
559
+ =
560
+ (4πnee2/me)1/2 is the electron plasma frequency (in cgs
561
+ units), and
562
+ E0 =4πe
563
+ Ω2p
564
+
565
+ ∂x
566
+
567
+ Z Pi
568
+ mi
569
+ − Pe
570
+ me
571
+ + Zniv2
572
+ i − nev2
573
+ e
574
+ ��
575
+
576
+ 4πe
577
+ meΩ2p
578
+ ω2
579
+ pe
580
+ ω2
581
+ 0
582
+ ∂x
583
+ �E2
584
+
585
+
586
+ (4a)
587
+ E2 = −
588
+ 4πe
589
+ meΩ2p
590
+ ω2
591
+ pe
592
+ ω2
593
+ 0
594
+ ∂x
595
+ �E2
596
+
597
+
598
+ (4b)
599
+ with the standard notation (t, x, v, P, ms, ns, Z) for the
600
+ time, space, velocity, pressure, mass and density of a
601
+ particle of species s, and ion charge respectively.
602
+ The
603
+ ⟨⟩ denotes an average over a laser cycle. The coupling
604
+ to the laser was included in the momentum equation via
605
+ the radiation force density fRF = (ϵ − 1)/8π∇E2 where
606
+ ϵ is the plasma permittivity (See Appendix A). One can
607
+ find an equation similar to Eq. (3) in Refs.2,16,20,24 but
608
+ with a different right-hand side (different current sources
609
+ of the THz radiation).
610
+ The solution of Eq. (3) under the initial conditions of
611
+ (Ea, ∂tEa) = (0, 0) reads (See for example Ref.45):
612
+ Ea(t) =
613
+ Terahertz oscillation
614
+
615
+ ��
616
+
617
+ E0
618
+
619
+ 1 − exp (−Γt)
620
+
621
+ cos (ϖt) + Γ
622
+ ϖ sin (ϖt)
623
+ ��
624
+ +
625
+ Second-harmonic oscillation
626
+
627
+ ��
628
+
629
+ Ω2
630
+ pE2
631
+
632
+ Ω2
633
+ p − 4ω2
634
+ 0
635
+
636
+ cos(2ω0t) + 4ω0Γ sin(2ω0t)
637
+
638
+ Ω2p − 4ω2
639
+ 0
640
+ �2 + 16Γ2ω2
641
+ 0
642
+ (5)
643
+ where ϖ2 = Ω2
644
+ p − Γ2. This solution includes two com-
645
+ ponents. The first component oscillates with a frequency
646
+ close to the plasma frequency ωpe when ωpe ≫ νei. This
647
+ oscillation, however, decays exponentially at a rate close
648
+ to the collision frequency. This component is established
649
+ by the spatial gradients of the pressure difference between
650
+ the light electrons and the heavy ions as represented in
651
+ Eq. (4a). This part induces the dipole moment in the
652
+ plasma by separating the electrons from the ions. After
653
+ a time t ≫ 1/νei, neglecting the electron and ion veloc-
654
+ ities and assuming Te ≫ Ti (See Paper I38), a nearly
655
+ constant electric field remains eEa ≈ −1/nedPe/dx =
656
+ −γTed ln ne/dx, considering an adiabatic equation of
657
+ state with the adiabatic index γ. Therefore, the ambipo-
658
+ lar field oscillations are driven by the electron density
659
+ gradient. The work function of the electrons that moved
660
+ from the plasma interior (density n1) to the exterior (den-
661
+ sity n2) then reads −e∆φ = γTe ln(n1/n2) ≈ 4 keV.
662
+ The second part in Eq.
663
+ 5 arises where gradients of
664
+ the laser intensity induce a second harmonic longitu-
665
+ dinal field oscillation.
666
+ This term has a resonance at
667
+ 2ω = Ωp ≈ ωpe (four times the critical density) for
668
+ the evanescent part of the wave causing a very steep in-
669
+ crease of the oscillation amplitude. This resonance for
670
+ s−polarized lasers is different from the Denisov reso-
671
+ nance absorption occurring under oblique incidence of
672
+ p−polarized lasers.46,47
673
+ An example of the electric field given by Eq. (5) is
674
+ shown in Fig. 3(a) for ϖ = 30 rad/ps (corresponds to an
675
+ edge plasma density of n2/nc = 10−4), and Γ = 3 rad/ps
676
+ where we have supposed that the electron collision fre-
677
+ quency is small compared to the plasma frequency. The
678
+ wave shows 2-3 oscillations and damps out within a time
679
+ scale of ∼ 2 ps.
680
+ The Fourier spectrum of the elec-
681
+ tric current associated with the quasi-static electric field,
682
+ 4πJa(ω) = jωEa(ω)/(1 + jνei/ω), is shown in Fig. 3(b).
683
+ As one can see, it has a maximum at ω ≈ ϖ = 30 rad/ps.
684
+ To compare with the numerical results of the radi-
685
+ ated emission in Fig. 1(b), we derive the angular dis-
686
+ tribution of the radiated energy from a current source.
687
+ For a number of accelerated charges, the integrand
688
+ in Eq.
689
+ (2) involves the replacement eβejωR(t′)/c →
690
+ �N
691
+ m=1 emβmejωRm(t′)/c.
692
+ In the limit of a continu-
693
+ ous distribution of charge, the summation becomes an
694
+ integral over the current density as eβejωR(t′)/c
695
+
696
+ 1/c
697
+
698
+ d3r′ J(r′, t′)ejωR(t′)/c.
699
+ Hence, the radiation en-
700
+ ergy per solid angle per frequency of the current source
701
+ reads:44
702
+ d2W
703
+ dωdΩ =
704
+ ω2
705
+ 4π2c3
706
+ ����
707
+
708
+ dt′
709
+
710
+ d3r′ ˆn × [ˆn × J(r′, t′)] ejω[t′+R(t′)/c]
711
+ ����
712
+ 2
713
+ (6)
714
+ We consider an emission length of L for the plasma
715
+ rod oriented parallel to the z−direction, and a cur-
716
+ rent of electron in the x−direction as J(r′, t′)
717
+ =
718
+ ˆxJa(t′)δ(x′)δ(y′) exp (jkz′).
719
+ For a coordinate system
720
+ with the spherical angle θ = cos−1 (z/r) and the azimuth
721
+ angle φ = tan−1 (y/x) defining the direction of observa-
722
+ tion n, Eq. (6) reduces to:
723
+
724
+ 6
725
+ FIG. 3. Temporal profile of the THz component of the ambipolar electric field from Eq. 5, panel (a), the frequency spectrum
726
+ of the electromagnetic radiation, panel (b), angular distributions of the radiated energy from Eq. 7(b), panel (c), and electric
727
+ field components, panels (d)-(f).
728
+ d2W
729
+ dωdΩ =|Ja(ω)|2
730
+ π2c
731
+ f(φ, θ)
732
+ (7a)
733
+ f(φ, θ) =sin2 θ sin2 φ + cos2 θ
734
+ (1 − cos θ)2
735
+ sin2
736
+
737
+ Lk sin2 θ
738
+ 2
739
+
740
+ (7b)
741
+ where k is the emission wave-vector.
742
+ The angular distribution given in Eq. (7b) is shown in
743
+ Fig. 3(c) for an emission length of L = 10 µm. It fits well
744
+ with the distribution obtained from the PIC simulation
745
+ in Fig. 1(b) where the emission is beamed in the posi-
746
+ tive z−direction and has two maxima in the y−direction,
747
+ perpendicular to the current source. The vector poten-
748
+ tial for this current source is in the x−direction and at
749
+ far-field, it reads:44
750
+ A(x, ω) =1
751
+ c
752
+ ejkR
753
+ R
754
+
755
+ d3r′J (r′, ω) e−jkn·r′
756
+ =ˆx2Ja(ω)
757
+ c
758
+ ejkR
759
+ kR
760
+ sin
761
+
762
+ Lk sin2 θ
763
+ 2
764
+
765
+ 1 − cos θ
766
+ (8)
767
+ One can derive the scalar potential Φ using the Lorenz
768
+ gauge and then the components of the electric field as
769
+ follows:
770
+ ∇ · A = − 1
771
+ c
772
+ ∂Φ
773
+ ∂t
774
+ (9a)
775
+ E = − 1
776
+ c
777
+ ∂A
778
+ ∂t − ∇Φ
779
+ (9b)
780
+ The components of the radiated electric field calculated
781
+ using Eq. (9b) are shown in Fig. 3(d)-(f). In agreement
782
+ with the results of the PIC simulation [Figs. 2(c-d)], the
783
+ radiated emission is polarized in the x−direction. More-
784
+ over, the angular distributions of the electric field compo-
785
+ nents agree with the results of the PIC simulation [Figs.
786
+ 2(c)-(e)].
787
+ The quadrupole pattern in Fig.
788
+ 2(d) is not
789
+ symmetric like Fig. 3(e). The asymmetry is because the
790
+ trajectory of the electron is not symmetric as the electron
791
+ spends more time in the x > 0 region relative to x < 0
792
+ [Fig. 2(a)].
793
+ This analytical model allows us to explain the THz ra-
794
+ diation emission in PIC simulations. (i) We show that
795
+ the current source of the THz emission originates from
796
+ the electrons which are trapped between the double lay-
797
+ ers. (ii) The current source is parallel to the incident laser
798
+ polarization, and consequently, the THz radiation is po-
799
+ larized like the incident laser polarization. (iii) The THz
800
+ radiation shows a much higher signal along the angles
801
+ corresponding to the forward direction (0◦ < θ ≤ 90◦)
802
+ than for the backward direction (90◦ < θ ≤ 180◦). This
803
+
804
+ 180
805
+ 1.0
806
+ (a)
807
+ (b)
808
+ (c)
809
+ 1.0
810
+ 1.0
811
+ 135
812
+ 0.8
813
+ 0.8
814
+ 9
815
+ a
816
+ 0.6
817
+ 0.6
818
+ [de
819
+ 90-
820
+ -0.5
821
+ f(Φ,
822
+ Ea(t) [
823
+ 0.4
824
+ 45 -
825
+ 0.2 -
826
+ 0.2
827
+ 0.0 -
828
+ 0.0
829
+ 0
830
+ 0.0
831
+ 0.0
832
+ 0.5
833
+ 1.0
834
+ 1.5
835
+ 2.0
836
+ 0
837
+ 25
838
+ 50
839
+ 75
840
+ 100
841
+ 0
842
+ 90
843
+ 180
844
+ 270
845
+ 360
846
+ t[ps]
847
+ w [rad/ps]
848
+ Φ[deg]
849
+ 5
850
+ -1
851
+ 5
852
+ 5
853
+ -1
854
+ (d)
855
+ (e)
856
+ (f)
857
+ -2
858
+ F-2
859
+ -2
860
+ 8
861
+ 0
862
+ 0
863
+ 0
864
+ y
865
+ -3
866
+ -3
867
+ -3
868
+ log]Ey]?[a. u. ]
869
+ loglEx]²[a. u. ]
870
+ loglEz]2[a. u. ]
871
+ -5
872
+ -5.
873
+ -4
874
+ -5
875
+ .4
876
+ -5
877
+ 0
878
+ 5
879
+ -5
880
+ 0
881
+ 5
882
+ 0
883
+ 5
884
+ x[入]
885
+ x[入]
886
+ x[入]7
887
+ FIG. 4.
888
+ The frequency spectrum of the THz wave [Eq. (5)]
889
+ is shown for different plasma frequencies, panel (a), different
890
+ electron-ion collision frequencies, panel (b), different pulse du-
891
+ rations, panel (c), and for different pulse central wavelengths,
892
+ panel (d).
893
+ is due to the coherence of the phases of the dipole mo-
894
+ ments induced along the plasma rod [Eq. (7)].
895
+ It would be of interest to see what parameters affect the
896
+ THz radiation. The THz radiation forms due to the os-
897
+ cillating dipole moments in the plasma. Hence, the peak
898
+ frequency of the THz spectrum is determined by plasma
899
+ frequency (ϖ2 = Ω2
900
+ p−Γ2) as shown in Fig. 4(a). A higher
901
+ plasma frequency will cause a greater radiation force and
902
+ increases the energy of THz radiation. The electron-ion
903
+ collision frequency Γ is one of the important factors af-
904
+ fecting the THz spectrum. Increasing the collision time
905
+ slows down the thermal equilibrium between the elec-
906
+ trons and ions and leads to a longer-lasting ambipolar
907
+ electric field and a broader THz spectrum as shown in
908
+ Fig. 4(b).
909
+ The THz wave amplitude is proportional to the radia-
910
+ tion force driven E0 field as expressed in Eq. (5). To have
911
+ an estimate of this amplitude, let us suppose a pulse in-
912
+ tensity given by I = I0/√π exp
913
+
914
+ −r2/w2
915
+ 0
916
+
917
+ exp
918
+
919
+ −t2/T 2�
920
+ ,
921
+ where T is the duration of the pulse, and w0 the waist
922
+ of the pulse.
923
+ Under the equilibrium between radia-
924
+ tion and space charge forces, the THz wave amplitude
925
+ reads E0 = 4E2
926
+ p/(ecω2
927
+ 0T 3), where the pulse energy is
928
+ Ep = I0πw2
929
+ 0T. Hence, reducing the pulse duration while
930
+ keeping the pulse energy constant strongly increases the
931
+ radiated THz energy [Fig. 4(c)]. This is due to the higher
932
+ peak intensity of the incident field. Moreover, increasing
933
+ the laser wavelength enforces the exerted radiation force
934
+ on electrons during a laser cycle [Eq. (A2)]. It leads to
935
+ a stronger net electron current and amplification of the
936
+ THz radiation [Fig. 4(d)].
937
+ IV.
938
+ DISCUSSION
939
+ In this work, we have extended our study of femtosec-
940
+ ond Bessel beam-induced plasmas inside the dielectrics.
941
+ A single-shot Bessel beam can generate a high aspect ra-
942
+ tio over-critical plasma inside the dielectric.37 The gen-
943
+ erated plasma offers a promising medium for the THz
944
+ radiation due to the current hot electrons driven by the
945
+ resonance absorption.
946
+ Based on an analytical approach, we derived the cur-
947
+ rent source, the electric field components, and the angu-
948
+ lar distribution of THz radiation. The analytical deriva-
949
+ tion reproduces the main characteristics of the THz ra-
950
+ diation calculated using the radiation diagnostic of PIC
951
+ simulation. Under the linear mode conversion, the radia-
952
+ tion force of the resonantly driven plasma waves kicks the
953
+ electrons from the critical surfaces. Due to the different
954
+ mobility of the plasma species, charge separations known
955
+ as double layers, and consequently, ambipolar electric
956
+ fields form at the plasma surfaces. Most of the ejected
957
+ electrons from the critical surfaces trap in the potentials
958
+ of the ambipolar electric fields at plasma edges.
959
+ The
960
+ trapped electrons oscillate with a period of around 130 fs
961
+ and radiate in the THz frequency domain.
962
+ Although in this work we have examined the over-
963
+ critical plasma, the presented study is valid for the sub-
964
+ critical plasma, because the radiation force of an intense
965
+ laser (≳ 1018 W/cm2) can also induce the quasi-static
966
+ fields and the associated THz radiation.
967
+ The second-
968
+ harmonic part of the ambipolar field offers an experi-
969
+ mental diagnostic for the detection of THz radiation. Its
970
+ pattern at the far-field (a central spot) differs from the
971
+ one generated at the critical surfaces (two lobes parallel
972
+ with the incident laser polarization discussed in Paper
973
+ II39).
974
+ To estimate power radiated within the THz range,
975
+ equating the radiation force with the force due to the
976
+ space charge field generated from electron-ion separa-
977
+ tion gives an acceleration a = eE0/m = 4E2
978
+ p/(mcω2
979
+ 0T 3),
980
+ and the power using the Larmor formula44
981
+ P
982
+ =
983
+ 2e2a2/3c3, is P
984
+ = 32e2E2
985
+ p/(3m2c5ω4
986
+ 0T 6), or PW
987
+
988
+ 108 �
989
+ EµJ
990
+ p (λµm
991
+ 0 )2/(T fs)3�2.
992
+ Assuming a microjoule laser pulse with a duration of
993
+ 100 fs at 800 nm wavelength, the laser to THz efficiency
994
+ is predicted to be about ∼ 10−8.
995
+ This value appears
996
+ to be very small compared with the THz efficiency of
997
+ ∼ 10−3 − 10−6 for femtosecond pulses with the energy of
998
+ ∼ 10−3 − 1 Joules.19,23,48 Unlike our works, the absorp-
999
+ tion process for interactions of 10−3−1 Joules class lasers
1000
+ with solids relies on the Brunel mechanism49,50 and THz
1001
+ radiation generation is due to the surface currents18,48
1002
+ or highly relativistic particles passing through the differ-
1003
+ ent dielectrics, the so-called transition radiation.23 The
1004
+ THz energy might be improved using the Bessel beams.
1005
+
1006
+ 1.2
1007
+ 1.2
1008
+ (a)
1009
+ Ωp = 10rad/ps
1010
+ (b)
1011
+ 『= 3rad/ps
1012
+ 1.0-
1013
+ Qp = 20 rad/ps
1014
+ 1.0
1015
+ -- 「= 6rad/ps
1016
+ -- Ωp = 30rad/ps
1017
+ -- 「= 12rad/ps
1018
+ 0.8
1019
+ 0.8
1020
+ "n'el(m)
1021
+ 0.6-
1022
+ 0.6
1023
+ 0.4-
1024
+ 0.4 -
1025
+ 0.2
1026
+ 0.2
1027
+ 0.0
1028
+ 0.0
1029
+ 0
1030
+ 25
1031
+ 50
1032
+ 75
1033
+ 100
1034
+ 25
1035
+ 50
1036
+ 75
1037
+ 0
1038
+ 100
1039
+ w[rad/ps]
1040
+ w[rad/ps]
1041
+ 1.2
1042
+ 2.5
1043
+ (c)
1044
+ (d)
1045
+ T=50fs
1046
+ 入o = 0.8 μm
1047
+
1048
+ 1.0
1049
+ -- T= 75 fs
1050
+ --- 入o = 1.0 μm
1051
+ 2.0
1052
+ --- T= 100fs
1053
+ - - - 入o = 1.2 μm
1054
+ 0.8-
1055
+ 1.5
1056
+ 0.6-
1057
+ 1.0
1058
+ 0.4 -
1059
+ 0.5
1060
+ 0.2 -
1061
+ 0.0
1062
+ 0.0
1063
+ 25
1064
+ 50
1065
+ 75
1066
+ 100
1067
+ 0
1068
+ 25
1069
+ 50
1070
+ 75
1071
+ 0
1072
+ 100
1073
+ w[rad/ps]
1074
+ w[rad/ps]8
1075
+ The long plasma (recently, we reached cm-scale over-
1076
+ critical plasmas inside dielectrics)51 created by Bessel
1077
+ beams yields a longer double layer at the plasma surface.
1078
+ A longer double layer traps the ejected hot electrons from
1079
+ the critical surface on a longer distance, for a longer time
1080
+ which results in a longer THz pulse.
1081
+ Assuming that individual electrons radiate incoher-
1082
+ ently, we might estimate the THz intensity and conver-
1083
+ sion efficiency in the presented PIC simulation. As re-
1084
+ ported in Paper I38, the high-energy electrons represent
1085
+ around 4% of the electrons in the simulation (∼ 109).
1086
+ Considering the THz intensity for a single electron of
1087
+ 0.04 W/cm2 [Fig.
1088
+ 2(f)], the radiated THz intensity
1089
+ amounts to about 106 W/cm2, corresponding to a con-
1090
+ version efficiency of 10−8, in agreement with the above
1091
+ predicted efficiency.
1092
+ We require a picosecond timescale to observe the com-
1093
+ plete process of the THz wave generation (for example,
1094
+ a THz wave at 0.5 THz corresponds to a timescale of 2
1095
+ ps). However, the numerical heating appearing for high-
1096
+ density plasmas in several picoseconds imposes a limita-
1097
+ tion on the maximum duration of the PIC simulations.41
1098
+ For this reason, we did not run our simulations beyond
1099
+ 320 fs. The THz radiation corresponds to a frequency
1100
+ around 30 rad/s.
1101
+ Another challenge is calculating the
1102
+ radiation integral using the whole set of electrons in the
1103
+ PIC simulation. This requires the implementation of the
1104
+ radiation integral in the MPI-based parallel PIC codes,
1105
+ as done in Refs.52,53
1106
+ ACKNOWLEDGMENTS
1107
+ We thank the EPOCH support team for help https:
1108
+ //cfsa-pmw.warwick.ac.uk.
1109
+ The authors acknowl-
1110
+ edge the financial supports of:
1111
+ European Research
1112
+ Council (ERC) 682032-PULSAR, Region Bourgogne-
1113
+ Franche-Comte and Agence Nationale de la Recherche
1114
+ (EQUIPEX+ SMARTLIGHT platform ANR-21-ESRE-
1115
+ 0040), Labex ACTION ANR-11-LABX-0001-01, I-SITE
1116
+ BFC project (contract ANR-15-IDEX-0003), and the
1117
+ EIPHI Graduate School ANR-17-EURE-0002.
1118
+ This
1119
+ work was granted access to the PRACE HPC resources
1120
+ MARCONI-KNL, MARCONI-M100, and GALILEO at
1121
+ CINECA, Casalecchio di Reno, Italy, under the Project
1122
+ ”PULSARPIC” (PRA19 4980), PRACE HPC resource
1123
+ Joliot-Curie Rome at TGCC, CEA, France under the
1124
+ Project ”PULSARPIC” (RA5614), HPC resource Joliot-
1125
+ Curie Rome/SKL/KNL at TGCC, CEA, France un-
1126
+ der the projects A0070511001 and A0090511001, and
1127
+ M´esocentre de Calcul de Franche-Comt´e.
1128
+ Appendix A: Radiation force density
1129
+ The radiation force per unit volume, force density fRF,
1130
+ on the free electrons can be written:54
1131
+ fRF = ϵ − 1
1132
+ 8π ∇E2 + ϵ − 1
1133
+ 4πc ∂t(E × B)
1134
+ (A1)
1135
+ Usually, the average values of the force density dur-
1136
+ ing one period of the laser wave are considered.
1137
+ This
1138
+ is because the time envelope of the laser wave is much
1139
+ slower in comparison with the frequency of the laser
1140
+ wave.
1141
+ Hence, one can neglect the time average of the
1142
+ Poynting term, the last term in Eq. (A1). Let us con-
1143
+ sider the monochromatic solutions of the wave equation
1144
+ E = Es(r) cos(ω0t) where Es(r) includes the field’s spa-
1145
+ tial dependence. The radiation force density then reads
1146
+ fRF = − ω2
1147
+ pe
1148
+ 8πω2
1149
+ 0
1150
+ cos2(ω0t)∇E2
1151
+ s
1152
+ = −
1153
+ ω2
1154
+ pe
1155
+ 16πω2
1156
+ 0
1157
+ ∇E2
1158
+ s −
1159
+ ω2
1160
+ pe
1161
+ 16πω2
1162
+ 0
1163
+ cos(2ω0t)∇E2
1164
+ s
1165
+ = − ω2
1166
+ pe
1167
+ ω2
1168
+ 0
1169
+
1170
+ �E2
1171
+
1172
+
1173
+ − ω2
1174
+ pe
1175
+ ω2
1176
+ 0
1177
+ cos(2ω0t)∇
1178
+ �E2
1179
+
1180
+
1181
+ (A2)
1182
+ We have used
1183
+
1184
+ E2�
1185
+ = E2
1186
+ s /2 in Eq. (A2).
1187
+ Appendix B: Ambipolar electric field of double layer
1188
+ Following Refs.,46,47,55 we use the two-fluid plasma
1189
+ equations for continuity and momentum to derive an an-
1190
+ alytical solution for the ambipolar electric field of the
1191
+ double layer. The continuity equations read
1192
+ ∂t (neme) + ∂x (nemeve) =0
1193
+ (B1a)
1194
+ ∂t (nimi) + ∂x (nimivi) =0
1195
+ (B1b)
1196
+ where indexes e and i refer to electrons and ions, re-
1197
+ spectively. The equations for conservation of momentum
1198
+ read:
1199
+ ∂t (nemeve) = − ∂x
1200
+
1201
+ nemev2
1202
+ e
1203
+
1204
+ − ∂xPe − eneEa
1205
+ − nemeνei (ve − vi) + fRF
1206
+ (B2a)
1207
+ ∂t (nimivi) = − ∂x
1208
+
1209
+ nimiv2
1210
+ i
1211
+
1212
+ − ∂xPi + eniZEa
1213
+ + nemeνei (ve − vi)
1214
+ (B2b)
1215
+ In Eq. (B2a), the radiation force density is given by Eq.
1216
+ (A2). We have neglected the radiation force on the ions
1217
+ Zme/mifRF in Eq. (B2b).
1218
+ The Gauss law for the electric field Ea reads:
1219
+ ∂xEa = −4πe (ne − Zni)
1220
+ (B3)
1221
+ Taking the time derivative of the Gauss law, using the
1222
+ equations of continuity in Eqs. (B1), and spatial integra-
1223
+ tion gives:
1224
+ ∂tEa = 4πe (neve − Znivi)
1225
+ (B4)
1226
+
1227
+ 9
1228
+ The second derivative in time results in:
1229
+ ∂2
1230
+ t Ea = 4πe [∂t (neve) − Z∂t (nivi)]
1231
+ (B5)
1232
+ Substituting from the equations of momentum in Eqs.
1233
+ (B2) results in
1234
+ 1
1235
+ 4πe∂2
1236
+ t Ea = −∂x
1237
+
1238
+ nev2
1239
+ e
1240
+
1241
+ − 1
1242
+ me
1243
+ ∂xPx − eneEa
1244
+ me
1245
+ +νeine (vi − ve) + fRF
1246
+ me
1247
+ +Z∂x
1248
+
1249
+ niv2
1250
+ i
1251
+
1252
+ + Z
1253
+ mi
1254
+ ∂xPi − Z2eniEa
1255
+ mi
1256
+ +Zνeine (vi − ve) me
1257
+ mi
1258
+ (B6)
1259
+ The rearrangements of the terms result in the following
1260
+ differential equation that described a damped oscillator
1261
+ subjected to an external force (inhomogeneous second-
1262
+ order differential equation).
1263
+ ∂2
1264
+ t Ea + 2Γ∂tEa + Ω2
1265
+ pEa = Ω2
1266
+ p [E0 + E2 cos(2ω0t)]
1267
+ (B7)
1268
+ where
1269
+ Γ =νei
1270
+ 2
1271
+
1272
+ 1 + Zme
1273
+ mi
1274
+
1275
+ (B8a)
1276
+ Ω2
1277
+ p =ω2
1278
+ pe
1279
+
1280
+ 1 + Zme
1281
+ mi
1282
+
1283
+ (B8b)
1284
+ E0 =4πe
1285
+ Ω2p
1286
+
1287
+ ∂x
1288
+
1289
+ Z Pi
1290
+ mi
1291
+ − Pe
1292
+ me
1293
+ + Zniv2
1294
+ i − nev2
1295
+ e
1296
+ ��
1297
+
1298
+ 4πe
1299
+ meΩ2p
1300
+ ω2
1301
+ pe
1302
+ ω2
1303
+ 0
1304
+ ∂x
1305
+ �E2
1306
+
1307
+
1308
+ (B8c)
1309
+ E2 = −
1310
+ 4πe
1311
+ meΩ2p
1312
+ ω2
1313
+ pe
1314
+ ω2
1315
+ 0
1316
+ ∂x
1317
+ �E2
1318
+
1319
+
1320
+ (B8d)
1321
+ REFERENCES
1322
+ 1D. Grischkowsky, S. Keiding, M. van Exter, and C. Fattinger,
1323
+ “Far-infrared time-domain spectroscopy with terahertz beams of
1324
+ dielectrics and semiconductors,” J. Opt. Soc. Am. B 7, 2006–2015
1325
+ (1990).
1326
+ 2C. D’Amico, A. Houard, S. Akturk, Y. Liu, J. Le Bloas,
1327
+ M. Franco, B. Prade, A. Couairon, V. T. Tikhonchuk, and
1328
+ A. Mysyrowicz, “Forward thz radiation emission by femtosecond
1329
+ filamentation in gases: theory and experiment,” New Journal of
1330
+ Physics 10, 013015 (2008).
1331
+ 3B. Ferguson and X.-C. Zhang, “Materials for terahertz science
1332
+ and technology,” Nature Materials 1, 26–33 (2002).
1333
+ 4M. Tonouchi, “Cutting-edge terahertz technology,” Nature Pho-
1334
+ tonics 1, 97–105 (2007).
1335
+ 5M. Bass, P. A. Franken, J. F. Ward, and G. Weinreich, “Optical
1336
+ rectification,” Phys. Rev. Lett. 9, 446–448 (1962).
1337
+ 6L. Xu, X. Zhang, and D. H. Auston, “Terahertz beam generation
1338
+ by femtosecond optical pulses in electro-optic materials,” Applied
1339
+ Physics Letters 61, 1784–1786 (1992).
1340
+ 7A. Nahata, A. S. Weling, and T. F. Heinz, “A wideband coher-
1341
+ ent terahertz spectroscopy system using optical rectification and
1342
+ electro-optic sampling,” Applied Physics Letters 69, 2321–2323
1343
+ (1996).
1344
+ 8D. J. Cook and R. M. Hochstrasser, “Intense terahertz pulses by
1345
+ four-wave rectification in air,” Opt. Lett. 25, 1210–1212 (2000).
1346
+ 9M. Kress, T. L¨offler, S. Eden, M. Thomson, and H. G. Roskos,
1347
+ “Terahertz-pulse generation by photoionization of air with laser
1348
+ pulses composed of both fundamental and second-harmonic
1349
+ waves,” Opt. Lett. 29, 1120–1122 (2004).
1350
+ 10T. Bartel, P. Gaal, K. Reimann, M. Woerner, and T. Elsaesser,
1351
+ “Generation of single-cycle thz transients with high electric-field
1352
+ amplitudes,” Opt. Lett. 30, 2805–2807 (2005).
1353
+ 11X. Xie, J. Dai, and X.-C. Zhang, “Coherent control of thz wave
1354
+ generation in ambient air,” Phys. Rev. Lett. 96, 075005 (2006).
1355
+ 12D. H. Auston, K. P. Cheung, and P. R. Smith, “Picosecond pho-
1356
+ toconducting hertzian dipoles,” Applied Physics Letters 45, 284–
1357
+ 286 (1984).
1358
+ 13H. Hamster, A. Sullivan, S. Gordon, W. White, and R. W. Fal-
1359
+ cone, “Subpicosecond, electromagnetic pulses from intense laser-
1360
+ plasma interaction,” Phys. Rev. Lett. 71, 2725–2728 (1993).
1361
+ 14H. Hamster, A. Sullivan, S. Gordon, and R. W. Falcone, “Short-
1362
+ pulse terahertz radiation from high-intensity-laser-produced plas-
1363
+ mas,” Phys. Rev. E 49, 671–677 (1994).
1364
+ 15C.-C. Cheng, E. M. Wright, and J. V. Moloney, “Generation of
1365
+ electromagnetic pulses from plasma channels induced by fem-
1366
+ tosecond light strings,” Phys. Rev. Lett. 87, 213001 (2001).
1367
+ 16P. Sprangle, J. R. Pe˜nano, B. Hafizi, and C. A. Kapetanakos,
1368
+ “Ultrashort laser pulses and electromagnetic pulse generation in
1369
+ air and on dielectric surfaces,” Phys. Rev. E 69, 066415 (2004).
1370
+ 17K.-Y. Kim, J. H. Glownia, A. J. Taylor, and G. Rodriguez, “Ter-
1371
+ ahertz emission from ultrafast ionizing air in symmetry-broken
1372
+ laser fields,” Optics express 15, 4577–4584 (2007).
1373
+ 18Y. Li, C. Li, M. Zhou, W. Wang, F. Du, W. Ding, X. Lin, F. Liu,
1374
+ Z. Sheng, X. Peng, et al., “Strong terahertz radiation from rel-
1375
+ ativistic laser interaction with solid density plasmas,” Applied
1376
+ Physics Letters 100, 254101 (2012).
1377
+ 19C. Li, Y.-Q. Cui, M.-L. Zhou, F. Du, Y.-T. Li, W.-M. Wang, L.-
1378
+ M. Chen, Z.-M. Sheng, J.-L. Ma, X. Lu, et al., “Role of resonance
1379
+ absorption in terahertz radiation generation from solid targets,”
1380
+ Optics Express 22, 11797–11803 (2014).
1381
+ 20S. Mitryukovskiy, Coherent secondary radiation from femtosec-
1382
+ ond laser filaments, Ph.D. thesis, ´Ecole Polytechnique (2014).
1383
+ 21F. Buccheri and X.-C. Zhang, “Terahertz emission from laser-
1384
+ induced microplasma in ambient air,” Optica 2, 366–369 (2015).
1385
+ 22C. Miao, J. P. Palastro, and T. M. Antonsen, “Laser pulse driven
1386
+ terahertz generation via resonant transition radiation in inhomo-
1387
+ geneous plasmas,” Physics of Plasmas 23, 063103 (2016).
1388
+ 23J. D´echard, X. Davoine, L. Gremillet, and L. Berg´e, “Terahertz
1389
+ emission from submicron solid targets irradiated by ultraintense
1390
+ femtosecond laser pulses,” Physics of Plasmas 27, 093105 (2020).
1391
+ 24V.
1392
+ A.
1393
+ Andreeva,
1394
+ O.
1395
+ G.
1396
+ Kosareva,
1397
+ N.
1398
+ A.
1399
+ Panov,
1400
+ D.
1401
+ E.
1402
+ Shipilo, P. M. Solyankin, M. N. Esaulkov, P. Gonz´alez de
1403
+ Alaiza Mart´ınez, A. P. Shkurinov, V. A. Makarov, L. Berg´e, and
1404
+ S. L. Chin, “Ultrabroad terahertz spectrum generation from an
1405
+ air-based filament plasma,” Phys. Rev. Lett. 116, 063902 (2016).
1406
+ 25J. Yoshii, C. H. Lai, T. Katsouleas, C. Joshi, and W. B. Mori,
1407
+ “Radiation from cerenkov wakes in a magnetized plasma,” Phys.
1408
+ Rev. Lett. 79, 4194–4197 (1997).
1409
+ 26W. P. Leemans, C. G. R. Geddes, J. Faure, C. T´oth, J. van
1410
+ Tilborg, C. B. Schroeder, E. Esarey, G. Fubiani, D. Auerbach,
1411
+ B. Marcelis, M. A. Carnahan, R. A. Kaindl, J. Byrd, and M. C.
1412
+ Martin, “Observation of terahertz emission from a laser-plasma
1413
+ accelerated electron bunch crossing a plasma-vacuum boundary,”
1414
+ Phys. Rev. Lett. 91, 074802 (2003).
1415
+ 27Z.-M. Sheng, K. Mima, J. Zhang, and H. Sanuki, “Emission of
1416
+ electromagnetic pulses from laser wakefields through linear mode
1417
+ conversion,” Phys. Rev. Lett. 94, 095003 (2005).
1418
+ 28S. Tzortzakis, G. M´echain, G. Patalano, Y.-B. Andr´e, B. Prade,
1419
+ M.
1420
+ Franco,
1421
+ A.
1422
+ Mysyrowicz,
1423
+ J.-M.
1424
+ Munier,
1425
+ M.
1426
+ Gheudin,
1427
+
1428
+ 10
1429
+ G. Beaudin, et al., “Coherent subterahertz radiation from fem-
1430
+ tosecond infrared filaments in air,” Optics Letters 27, 1944–1946
1431
+ (2002).
1432
+ 29T. L¨offler, M. Kress, M. Thomson, and H. Roskos, “Efficient
1433
+ terahertz pulse generation in laser-induced gas plasmas,” Acta
1434
+ Phys. Pol. A 107, 99 (2005).
1435
+ 30U. Teubner, J. Bergmann, B. van Wonterghem, F. P. Sch¨afer,
1436
+ and R. Sauerbrey, “Angle-dependent x-ray emission and reso-
1437
+ nance absorption in a laser-produced plasma generated by a high
1438
+ intensity ultrashort pulse,” Phys. Rev. Lett. 70, 794–797 (1993).
1439
+ 31R. Sauerbrey, J. Fure, S. P. Le Blanc, B. van Wonterghem,
1440
+ U. Teubner, and F. P. Sch¨afer, “Reflectivity of laser-produced
1441
+ plasmas generated by a high intensity ultrashort pulse*,” Physics
1442
+ of Plasmas 1, 1635–1642 (1994).
1443
+ 32M. M. Murnane, H. C. Kapteyn, and R. W. Falcone, “High-
1444
+ density plasmas produced by ultrafast laser pulses,” Phys. Rev.
1445
+ Lett. 62, 155–158 (1989).
1446
+ 33E. Brambrink, H. G. Wei, B. Barbrel, P. Audebert, A. Benuzzi-
1447
+ Mounaix, T. Boehly, T. Endo, C. D. Gregory, T. Kimura, R. Ko-
1448
+ dama, N. Ozaki, H.-S. Park, and M. Koenig, “Direct density
1449
+ measurement of shock-compressed iron using hard x rays gener-
1450
+ ated by a short laser pulse,” Phys. Rev. E 80, 056407 (2009).
1451
+ 34W. Kruer, The Physics Of Laser Plasma Interactions (Avalon
1452
+ Publishing, 1988).
1453
+ 35S. Eliezer, The Interaction of High-Power Lasers with Plasmas,
1454
+ Series in Plasma Physics (CRC Press, 2002).
1455
+ 36P. Gibbon, Short Pulse Laser Interactions with Matter: An In-
1456
+ troduction (Imperial College Press, 2005).
1457
+ 37K. Ardaneh, R. Meyer, M. Hassan, R. Giust, C. Xie, B. Morel,
1458
+ I. Ouadghiri-Idrissi, L. Furfaro, L. Froehly, A. Couairon, G. Bon-
1459
+ naud, and F. Courvoisier, “High energy density plasma mediated
1460
+ by collisionless resonance absorption inside dielectrics,” (2021),
1461
+ arXiv:2109.00803 [physics.plasm-ph].
1462
+ 38K. Ardaneh, R. Meyer, M. Hassan, R. Giust, B. Morel, A. Coua-
1463
+ iron, G. Bonnaud, and F. Courvoisier, “Femtosecond laser-
1464
+ induced sub-wavelength plasma inside dielectrics:
1465
+ I. field en-
1466
+ hancement,” Physics of Plasmas 29, 072715 (2022).
1467
+ 39K. Ardaneh, M. Hassan, B. Morel, R. Meyer, R. Giust, A. Coua-
1468
+ iron, G. Bonnaud, and F. Courvoisier, “Femtosecond laser-
1469
+ induced sub-wavelength plasma inside dielectrics. ii. second-
1470
+ harmonic generation,” Physics of Plasmas 29, 072716 (2022).
1471
+ 40J. Durnin, J. J. Miceli, and J. H. Eberly, “Diffraction-free
1472
+ beams,” Phys. Rev. Lett. 58, 1499–1501 (1987).
1473
+ 41T. D. Arber, K. Bennett, C. S. Brady, A. Lawrence-Douglas,
1474
+ M. G. Ramsay,
1475
+ N. J. Sircombe,
1476
+ P. Gillies,
1477
+ R. G. Evans,
1478
+ H. Schmitz, A. R. Bell, and C. P. Ridgers, “Contemporary
1479
+ particle-in-cell approach to laser-plasma modelling,” Plasma
1480
+ Physics and Controlled Fusion 57, 113001 (2015).
1481
+ 42Y. Sentoku and A. Kemp, “Numerical methods for particle sim-
1482
+ ulations at extreme densities and temperatures: Weighted parti-
1483
+ cles, relativistic collisions and reduced currents,” Journal of Com-
1484
+ putational Physics 227, 6846 – 6861 (2008).
1485
+ 43K. Ardaneh, R. Giust, B. Morel, and F. Courvoisier, “Genera-
1486
+ tion of a bessel beam in fdtd using a cylindrical antenna,” Opt.
1487
+ Express 28, 2895–2908 (2020).
1488
+ 44J. Jackson, Classical Electrodynamics (Wiley, 1998).
1489
+ 45W. R. Inc., “Mathematica, Version 13.1,” Champaign, IL, 2022.
1490
+ 46M. Goldsworthy, F. Green, P. Lalousis, R. Stening, S. Eliezer,
1491
+ and H. Hora, “Hydrodynamic analysis of the high electrc fields
1492
+ and double layers in expanding inhomogeneous plasmas,” IEEE
1493
+ transactions on plasma science 14, 823–837 (1986).
1494
+ 47H. Hora, Plasmas at High Temperature and Density:
1495
+ Appli-
1496
+ cations and Implications of Laser-Plasma Interaction, Lecture
1497
+ Notes in Physics Monographs (Springer Berlin Heidelberg, 2008).
1498
+ 48Y. T. Li, X. H. Yuan, M. H. Xu, Z. Y. Zheng, Z. M. Sheng,
1499
+ M. Chen, Y. Y. Ma, W. X. Liang, Q. Z. Yu, Y. Zhang, F. Liu,
1500
+ Z. H. Wang, Z. Y. Wei, W. Zhao, Z. Jin, and J. Zhang, “Observa-
1501
+ tion of a fast electron beam emitted along the surface of a target
1502
+ irradiated by intense femtosecond laser pulses,” Phys. Rev. Lett.
1503
+ 96, 165003 (2006).
1504
+ 49L.
1505
+ Chopineau,
1506
+ A.
1507
+ Leblanc,
1508
+ G.
1509
+ Blaclard,
1510
+ A.
1511
+ Denoeud,
1512
+ M. Th´evenet, J.-L. Vay, G. Bonnaud, P. Martin, H. Vincenti,
1513
+ and F. Qu´er´e, “Identification of coupling mechanisms between ul-
1514
+ traintense laser light and dense plasmas,” Phys. Rev. X 9, 011050
1515
+ (2019).
1516
+ 50J. F. Ong, P. Ghenuche, and K. A. Tanaka, “Electron transport
1517
+ in a nanowire irradiated by an intense laser pulse,” Phys. Rev.
1518
+ Research 3, 033262 (2021).
1519
+ 51R. Meyer, L. Froehly, R. Giust, J. D. Hoyo, L. Furfaro, C. Bil-
1520
+ let, and F. Courvoisier, “Extremely high-aspect-ratio ultrafast
1521
+ Bessel beam generation and stealth dicing of multi-millimeter
1522
+ thick glass,” Applied Physics Letters 114, 201105 (2019).
1523
+ 52J. T. Frederiksen, T. Haugbølle, M. V. Medvedev, and ˚A. Nord-
1524
+ lund, “Radiation spectral synthesis of relativistic filamentation,”
1525
+ The Astrophysical Journal Letters 722, L114 (2010).
1526
+ 53K.-I. Nishikawa, J. Niemiec, M. Medvedev, B. Zhang, P. Hardee,
1527
+ A. Nordlund, J. Frederiksen, Y. Mizuno, H. Sol, M. Pohl, et al.,
1528
+ “Radiation from relativistic shocks in turbulent magnetic fields,”
1529
+ Advances in Space Research 47, 1434–1440 (2011).
1530
+ 54L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous
1531
+ Media (Pergamon, New York, 1984).
1532
+ 55P. Lalousis and H. Hora, “First direct electron and ion fluid com-
1533
+ putation of high electrostatic fields in dense inhomogeneous plas-
1534
+ mas with subsequent nonlinear laser interaction,” Laser and Par-
1535
+ ticle Beams 1, 283–304 (1983).
1536
+
EdE3T4oBgHgl3EQfVQpt/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
F9AzT4oBgHgl3EQfHPuf/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a98d6c21ad615d8794030f850843aea710283892eb6b712ba9961e161aedcaf7
3
+ size 5046317
FNAzT4oBgHgl3EQfw_7Q/content/2301.01732v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f56c9b54c4e66791e0f22c948ec28826da80b8ebfc8c88c9f74a6c3f89f6e604
3
+ size 1391697
FNAzT4oBgHgl3EQfw_7Q/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11d3570687d2a0d0ad015cf3ea850fac36a70a6c34dbcae9cb0e5b912b9f9bbd
3
+ size 3145773
FNE2T4oBgHgl3EQf-Anw/content/tmp_files/2301.04235v1.pdf.txt ADDED
The diff for this file is too large to render. See raw diff
 
FNE2T4oBgHgl3EQf-Anw/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
GNE4T4oBgHgl3EQfgA3B/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbb77c7839bc762bc16c959fdcf648d744e7c84018d0dd3dfb432506c7aa1f21
3
+ size 21626925
INAzT4oBgHgl3EQfU_y9/content/tmp_files/2301.01277v1.pdf.txt ADDED
@@ -0,0 +1,1185 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Artificial Intelligence in Wholesale and Retail
2
+ AE
3
+
4
+ Vol. 23 • No. 56 • February 2021
5
+ 155
6
+ CONSUMER ACCEPTANCE OF THE USE OF ARTIFICIAL INTELLIGENCE
7
+ IN ONLINE SHOPPING: EVIDENCE FROM HUNGARY
8
+
9
+ Szabolcs Nagy1* and Noémi Hajdú2
10
+ 1)2) University of Miskolc, Miskolc, Hungary
11
+
12
+
13
+
14
+ Please cite this article as:
15
+ Nagy, S. and Hadjú, N., 2021. Consumer Acceptance
16
+ of the Use of Artificial Intelligence in Online
17
+ Shopping:
18
+ Evidence
19
+ From
20
+ Hungary.
21
+ Amfiteatru
22
+ Economic, 23(56), pp.155-173.
23
+
24
+ DOI: 10.24818/EA/2021/56/155
25
+
26
+ Article History
27
+ Received: 30 September 2020
28
+ Revised: 7 November 2020
29
+ Accepted: 26 December 2020
30
+
31
+
32
+ Abstract
33
+ The rapid development of technology has drastically changed the way consumers do their
34
+ shopping. The volume of global online commerce has significantly been increasing partly
35
+ due to the recent COVID-19 crisis that has accelerated the expansion of e-commerce.
36
+ A growing number of webshops integrate Artificial Intelligence (AI), state-of-the-art
37
+ technology into their stores to improve customer experience, satisfaction and loyalty.
38
+ However, little research has been done to verify the process of how consumers adopt and
39
+ use AI-powered webshops. Using the technology acceptance model (TAM) as a theoretical
40
+ background, this study addresses the question of trust and consumer acceptance of
41
+ Artificial Intelligence in online retail. An online survey in Hungary was conducted to build
42
+ a database of 439 respondents for this study. To analyse data, structural equation modelling
43
+ (SEM) was used. After the respecification of the initial theoretical model, a nested model,
44
+ which was also based on TAM, was developed and tested. The widely used TAM was
45
+ found to be a suitable theoretical model for investigating consumer acceptance of the use of
46
+ Artificial Intelligence in online shopping. Trust was found to be one of the key factors
47
+ influencing consumer attitudes towards Artificial Intelligence. Perceived usefulness as the
48
+ other key factor in attitudes and behavioural intention was found to be more important than
49
+ the perceived ease of use. These findings offer valuable implications for webshop owners to
50
+ increase customer acceptance.
51
+
52
+ Keywords: consumer acceptance, artificial intelligence, online shopping, AI-powered
53
+ webshops, technology acceptance model, trust, perceived usefulness, perceived ease of use,
54
+ attitudes, behavioural intention, Hungary
55
+
56
+ JEL Classification: L81, M31, O30
57
+
58
+
59
+
60
+ * Corresponding author, Szabolcs Nagy – e-mail: nagy.szabolcs@uni-miskolc.hu
61
+
62
+ AE
63
+ Consumer Acceptance of the Use of Artificial Intelligence
64
+ in Online Shopping: Evidence From Hungary
65
+
66
+ 156
67
+ Amfiteatru Economic
68
+ Introduction
69
+ The rapid development of digital technology has changed online shopping (Daley, 2018). In
70
+ recent years, the use of Artificial Intelligence (AI) in online commerce has been increased since
71
+ AI is an excellent tool to meet rapidly changing consumer demand and to increase sales
72
+ efficiency. The global spending by retailers on AI services is expected to quadruple and reach
73
+ $12 billion by 2023, and over 325000 retailers will adopt AI technology (Maynard, 2019).
74
+ Smidt and Power (2020) claimed that online product research has significantly increased
75
+ over the past years. USA's largest online retailer, Amazon, is the exemplary case of how
76
+ to effectively integrate AI into online retail. Besides the rich assortment, fast delivery
77
+ and competitive prices, a more localised shopping journey can be created. Thus Amazon
78
+ can use location-specific pricing and send destination-specific messages to its
79
+ customers, who will pay in their local currency (Barmada, 2020).
80
+ Novel marketing techniques supported by new technologies, including the use of AI systems
81
+ spark the proliferation of new marketing methods to effectively reach target consumers and to
82
+ offer enhanced consumer experiences (Pusztahelyi, 2020). Pursuant to Asling (2017), the use
83
+ of AI in online shopping makes customer-centric search and a new level of personalisation
84
+ possible resulting in a more efficient sales process. Information technology (IT) has changed
85
+ the nature of company-customer relationships (Rust and Huang, 2014). However, any
86
+ technology-driven transformation is based on trust (Pricewaterhouse Coopers, 2018).
87
+ Online retailers need more in-depth insight into how consumers perceive and accept the use of
88
+ AI in webshops and how much they trust them. They also need to know how to use AI most
89
+ effectively to increase online spending and online purchase frequency since the importance of
90
+ time and cost efficiency in shopping has recently become more and more critical. In this
91
+ regard, online shopping means a convenient way for customers to buy the desired products.
92
+ So far, only a few researchers have addressed the question of trust and consumer
93
+ acceptance of AI in online retail. Based on the technology acceptance model (TAM), this
94
+ study aims to fill this research gap and proposes an integrated theoretical framework of
95
+ consumers' acceptance of AI-powered webshops. Further objectives of this paper are to
96
+ investigate the relationships between the elements of TAM; to analyse the effects of trust,
97
+ perceived usefulness and perceived ease of use on attitudes and behavioural intention.
98
+ After reviewing the use of AI in online shopping, this paper discusses the role of trust in
99
+ online shopping and presents the technology acceptance model. The next section deals with
100
+ the research methodology, including the research questions, hypotheses and the sample. In
101
+ the results and discussion section, the validity and reliability of the model, as well as the
102
+ model fit are presented. Hypothesis testing, detailed analysis of the relationships between
103
+ the elements of the nested model, and comparison of the results with the previous research
104
+ findings are also discussed here before the conclusions sections.
105
+
106
+ 1. Literature review
107
+ According to IBM's U.S. Retail Index, the COVID-19 has speeded up the change from
108
+ traditional shopping to online purchasing by circa five years (Haller, Lee and Cheung,
109
+ 2020). Due to the pandemic situation, there is an increased demand for AI in the retail
110
+ industry (Meticulous Market Research, 2020).
111
+
112
+ Artificial Intelligence in Wholesale and Retail
113
+ AE
114
+
115
+ Vol. 23 • No. 56 • February 2021
116
+ 157
117
+ 1.1. The use of AI in online shopping
118
+ AI systems are a set of software and hardware that can be used to continuously assess and
119
+ analyse data to characterise environmental factors and to determine decisions and actions
120
+ (European Commission, 2018). Prior research mainly focused on the advantages of the use
121
+ of AI in online settings and failed to address how consumers accept AI in online retail.
122
+ According to utility theory, this new technology helps consumers to find and choose the
123
+ best product alternatives, while decreases the search cost and search time (Pricewaterhouse
124
+ Coopers, 2018), thus increasing utility (Stigler, 1961; Bakos, 1977; Stigler and Becker,
125
+ 1977; André, et al. 2017; Lynch and Ariely, 2000). AI filters the information for each target
126
+ customer and provides what exactly is needed (Paschen, Wilson and Ferreira, 2020). AI
127
+ supports automating business processes, gains insight through data analysis, and engages
128
+ with customers and employees (Davenport and Ronanki, 2018).
129
+ Artificial intelligence is widely used to increase the efficiency of marketing (Kwong, Jiang, and
130
+ Luo, 2016) and retail (Weber and Schütte, 2019) and to automate marketing (Dumitriu and
131
+ Popescu, 2020). AI-powered online stores provide their customers with automated assistance
132
+ during the consumer journey (Yoo, Lee and Park, 2010; Pantano and Pizzi, 2020). It is a great
133
+ advantage, especially for the elder people, who are averse to technical innovations.
134
+ Consumers' online information search and product selection habits can be better understood by
135
+ AI to offer a more personalised shopping route (Rust and Huang, 2014). It is a great
136
+ opportunity for online shops to analyse the profile of existing and potential customers and
137
+ thereby suggest tailor-made marketing offerings for them (Onete, Constantinescu and Filip,
138
+ 2008). AI also makes the contact with both the customers and the employees continuous and
139
+ interactive. Frequently asked questions (FAQs) regarding the products, product-use and
140
+ ordering process can be automated by a chatbot. New sales models use automated algorithms
141
+ to recommend unique, personalised marketing offerings, thus increasing customer satisfaction
142
+ and engagement. To sum up the advantages, AI systems operate automatically and analyse big
143
+ data in real-time to interpret and shape consumer behavioural patterns to offer products and
144
+ services in a personalised way, thus enhancing the shopping experience.
145
+ However, AI systems also have some disadvantages. They work most effectively with big data;
146
+ therefore, the implementation of AI systems requires huge investments (Roetzer, 2017).
147
+ 1.2. The role of trust in online shopping
148
+ Trust is of great importance in online commerce. According to Kim, Ferrin and Rao (2008),
149
+ consumer confidence has a positive effect on a consumer's intention to buy. The higher the
150
+ consumer trust in an online shop is, the more likely the consumer will be to go through the
151
+ buying process. Trust is especially crucial when the customer perceives a financial risk.
152
+ Thatcher et al. (2013) identified two types of trust: general and specific trust. General trust
153
+ concerns the e-commerce environment, consumer beliefs about and attitudes towards it.
154
+ Specific trust is related to the shopping experience in a specific virtual store. Confidence can be
155
+ enhanced through interactive communication between the retailer and the buyer by using
156
+ appropriate product descriptions and images to reduce the perceived risk. As stated in Cătoiu et
157
+ al. (2014) there is a strong negative correlation between perceived risks and trust. According to
158
+ Reichheld and Schefter (2000, p. 107), “price does not rule the Web; trust does”.
159
+
160
+ AE
161
+ Consumer Acceptance of the Use of Artificial Intelligence
162
+ in Online Shopping: Evidence From Hungary
163
+
164
+ 158
165
+ Amfiteatru Economic
166
+ Aranyossy and Magisztrák (2016) found that a higher level of e-commerce trust was
167
+ associated with more frequent online shopping. However, when shopping online, customers
168
+ do not necessarily notice that a website uses AI tools (Daley, 2018).
169
+ All things considered, AI marks a new era in online sales. However, continuous
170
+ technological development such as the use of AI-powered websites divides society, as there
171
+ are those who accept novelty while others reject it.
172
+
173
+ 1.3. Technology Acceptance Model (TAM)
174
+ Consumers' adaptation to new technologies can be explained by several models. Dhagarra,
175
+ Goswami and Kumar (2020) summarised them as follows: (1) Theory of Reasoned Action
176
+ (TRA) by Fishbein and Ajzen (1975); (2) Theory of Planned Behaviour (TPB) by Ajzen
177
+ (1985); (3) Technology Acceptance Model (TAM) by Davis (1986); (4) Innovation
178
+ Diffusion Theory (IDT) by Rajagopal (2002); (5) Technology Readiness Index (TRI) by
179
+ Parasuraman, (2000); and (6) Unified Theory of Acceptance and Use of Technology
180
+ (UTAUT) by Venkatesh, et al. (2003).
181
+ Technology acceptance model (TAM), an extension of (TRA), is one of the most widely-
182
+ used theoretical models (Venkatesh, 2000) to explain why an IT user accepts or rejects
183
+ information technology and to predict IT user behaviour (Legris, Ingham, and Collerette,
184
+ 2003). The original TAM contains six elements: external variables, perceived usefulness,
185
+ perceived ease of use, attitude, behavioural intention to use and actual use. According to
186
+ TAM, external variables have a direct influence on perceived usefulness (PU) and
187
+ perceived ease of use (PEU), i.e. the two cognitive belief components. Perceived ease of
188
+ use directly influences PU and attitude, whereas perceived usefulness has a direct impact on
189
+ attitude and behavioural intention to use, which affects actual use (Figure no. 1).
190
+
191
+ Figure no. 1. The original technology acceptance model (TAM)
192
+ Source: Davis, 1986.
193
+ Ha and Stoel (2008) examined the factors affecting customer acceptance of online shopping
194
+ and found that perceived ease of use, perceived trust and perceived shopping enjoyment
195
+ had the greatest impact on customer acceptance. Ease of use, trust and shopping enjoyment
196
+ had a significant impact on perceived usefulness; trust, shopping enjoyment, and usefulness
197
+
198
+ Perceived
199
+ Usefulness
200
+ External
201
+ Attitude
202
+ Behavioral
203
+ Variables
204
+ Towards
205
+ Intentionto
206
+ Actual Use
207
+ Use
208
+ Use
209
+ Perceived
210
+ Easeof UseArtificial Intelligence in Wholesale and Retail
211
+ AE
212
+
213
+ Vol. 23 • No. 56 • February 2021
214
+ 159
215
+ had a significant effect on attitude towards online shopping. They also found that attitude
216
+ and perceived usefulness had an influential role in consumer intention to purchase online.
217
+ According to Vijayasarathy (2004), there is a positive association between consumer attitude
218
+ towards online shopping and the beliefs concerning usefulness, compatibility, security and ease
219
+ of use. Also, the intention to purchase online is strongly influenced by consumer beliefs about
220
+ online shopping, self-efficacy and attitude. Surprisingly, no positive relationship between
221
+ purchasing intention and consumer beliefs about the usefulness of online shopping was
222
+ reported (Vijayasarathy, 2004). Gefen, Karahanna and Straub (2003) found that perceived
223
+ usefulness and perceived ease of use influence consumer repurchase intention.
224
+ It must be noted that Schepman and Rodway (2020) expressed some criticisms about the
225
+ applicability of TAM to measure attitudes towards AI. According to them, it is the online
226
+ retailers that can decide to integrate AI into webshops, and consumers have no choice but to
227
+ use it when shopping online in such stores. Therefore, traditional technology acceptance
228
+ models might not be ideal to measure attitudes towards AI. However, we are convinced that
229
+ consumers still have the free will to decide whether to use new technology, i.e. to shop
230
+ online in an AI-powered webshop, or not.
231
+
232
+ 2. Methodology and research questions
233
+ 2.1. Methodology
234
+ The constructs and the measurement instruments presented in Table no. 1 were developed
235
+ based on the literature review, and according to the Technology Acceptance Model.
236
+ Variables with asterisk and in italics were adapted from Park (2009), the others were
237
+ adapted from Hu and O'Brien (2016). However, each variable was modified by the authors
238
+ to make it possible to measure the perceived role of AI in online shopping.
239
+ For data collection, a questionnaire made up of 26 questions (variables) was used (Table
240
+ no. 1). Additionally, six demographics variables - gender, education, age, occupation, place
241
+ of residence and internet subscription - were also included in the survey. All measurement
242
+ instruments were listed in Table no. 1 but the demographics variables were measured on a
243
+ seven-point Likert-scale ranging from strongly disagree (1) strongly agree (7).
244
+ In the very first section of the questionnaire, respondents were provided with a detailed
245
+ explanation of AI-powered webshops and shopping apps, which are online stores where
246
+ shopping is supported by artificial intelligence. AI-powered webshops present personalised
247
+ product/service offerings based on previous search patterns and purchases that we made
248
+ before, and automatically display products that AI chooses for us. Also, AI offers similar
249
+ products to those that were originally viewed but were not available in the right size
250
+ (product recommendation based on visual similarity). Another typical sign of an AI-
251
+ powered webshop is that when the customer is leaving the web store, AI warns about the
252
+ products left in the cart, to complete the purchase. AI-powered webshops often use
253
+ chatbots, i.e. a virtual assistant is available if the customer has any questions, and visual
254
+ (image-based) search is also possible: after uploading a product picture, AI recommends
255
+ the most similar ones to that. Virtual changing rooms, voice recognition and automatic
256
+ search completion are also available in AI-powered webshops such as Amazon, e-Bay,
257
+ Alibaba, AliExpress, GearBest, eMAG.hu, PCland.hu, Ecipo, Bonprix, Answear, Reserved,
258
+ Fashiondays, Fashionup, Spartoo, Orsay, to mention just a few.
259
+
260
+ AE
261
+ Consumer Acceptance of the Use of Artificial Intelligence
262
+ in Online Shopping: Evidence From Hungary
263
+
264
+ 160
265
+ Amfiteatru Economic
266
+ Table no. 1. Constructs and measurement instruments
267
+ Construct
268
+ Definition
269
+ Measurement Instruments
270
+ Perceived
271
+ Usefulness
272
+ (PU)
273
+ The degree to which a
274
+ consumer believes that AI
275
+ used in online shopping
276
+ would make his or her
277
+ purchases more effective.
278
+ PU1. The use of AI in retail (shopping ads and
279
+ webshops) allows me to find the best deals.
280
+ PU2. The use of AI in retail enhances my
281
+ effectiveness in purchasing.
282
+ PU3. The use of AI in retail is useful to me.
283
+ PU4 The use of AI in retail saves time for me. *
284
+ Perceived
285
+ Ease of Use
286
+ (PEU)
287
+ The degree to which a
288
+ consumer believes that
289
+ using AI in webshops will
290
+ be free of effort.
291
+ PEU1. AI-powered shopping apps and webshops
292
+ are easy to use.
293
+ PEU2. Shopping does not require a lot of my
294
+ mental efforts if supported by AI (alternatives are
295
+ offered by AI).
296
+ PEU3. Shopping is not so complicated if AI offers
297
+ products to me.
298
+ PEU4 Learning how to use AI-powered shopping
299
+ apps and webshops is easy for me. *
300
+ PEU5 It is easy to become skilful at using AI-
301
+ powered shopping apps and webshops*
302
+ Experience
303
+ (EXP)
304
+ The consumers'
305
+ knowledge about and the
306
+ experience with
307
+ purchasing in an AI-
308
+ powered webshop.
309
+ EXP1. I'm experienced in online shopping.
310
+ EXP2. I have already used AI-powered applications
311
+ (chatbots, etc.)
312
+ Trust
313
+ (TRUST)
314
+ The subjective probability
315
+ with which people believe
316
+ that AI works for their
317
+ best interest.
318
+ T1. I am convinced that AI in retail is used to
319
+ provide customers with the best offerings.
320
+ T2. I trust in apps and webshops that use AI.
321
+ Subjective
322
+ Norm
323
+ (SN)
324
+ The degree to which a
325
+ consumer perceives that
326
+ most people who are
327
+ important to him or her
328
+ think he or she should or
329
+ should not make
330
+ purchases in AI-powered
331
+ webshops.
332
+ SN1. People who influence my behaviour would
333
+ prefer me to use AI-powered shopping apps and
334
+ webshops.
335
+ SN2. I like using AI-powered webshops and
336
+ shopping apps based on the similarity of my values
337
+ and the social values underlying its use. *
338
+ Task
339
+ Relevance
340
+ (TR)
341
+ The degree to which a
342
+ consumer believes that
343
+ AI-powered webshops are
344
+ applicable to his or her
345
+ shopping task.
346
+ TR1 I think AI can be used effectively in webshops
347
+ and shopping apps.
348
+ Compen-
349
+ sation
350
+ (COMP)
351
+ The degree to which a
352
+ consumer believes that he
353
+ or she has the ability to
354
+ make purchases in AI-
355
+ powered webshops.
356
+ I would prefer AI-powered shopping apps and
357
+ webshops…
358
+ C1. if there was no one around to visit physical
359
+ shops/shopping malls with.
360
+ C2. if I had less time.
361
+ C3. if I had a built-in help facility for assistance
362
+ when needed.
363
+ Perceived
364
+ Quality
365
+ PQ
366
+ The degree of how good a
367
+ consumer perceives the
368
+ quality of a product in AI-
369
+ powered webshops.
370
+ PQ1 AI finds/offers better products for me than I
371
+ could.
372
+
373
+ Artificial Intelligence in Wholesale and Retail
374
+ AE
375
+
376
+ Vol. 23 • No. 56 • February 2021
377
+ 161
378
+ Construct
379
+ Definition
380
+ Measurement Instruments
381
+ Perceived
382
+ Enjoyment
383
+ PE
384
+ The extent to which
385
+ shopping in AI-powered
386
+ webshops is perceived
387
+ to be enjoyable.
388
+ PE1 Shopping is more fun, enjoyable when AI
389
+ helps me to find the best-suited products.
390
+ Attitude
391
+ ATT
392
+ The consumer's attitude
393
+ towards shopping in AI-
394
+ powered webshops.
395
+ ATT1 Shopping in a webshop/shopping app that is
396
+ powered by AI is a good idea
397
+ ATT2 Shopping in a webshop/shopping app that is
398
+ powered by AI is a wise idea
399
+ ATT3 I am positive towards webshop/shopping app
400
+ that is powered by AI
401
+ Behavioural
402
+ Intention
403
+ BI
404
+ A consumer's behavioural
405
+ intention to do the
406
+ shopping in AI-powered
407
+ webshops.
408
+ BI1 I intend to visit webshops and to use shopping
409
+ apps that are powered by AI more frequently.
410
+ BI2 I'm willing to spend more on products offered
411
+ by webshops and apps powered by AI
412
+ Sources: Adapted from Hu and O'Brien, 2016; *Park, 2009.
413
+ An online survey in Google Form was conducted to collect data in July and August 2020 in
414
+ Hungary. Because of the Theory Acceptance Model, previous online shopping experience
415
+ with AI-powered webshops was the one and only eligibility criterion for respondents to
416
+ participate in this study. Convenience sampling method was used to reach the maximum
417
+ number of respondents. Data was migrated from Google Form to MS Excel, SPSS 24 and
418
+ AMOS, and was checked for coding accuracy. The database was complete and contained
419
+ no missing data. Descriptive statistical analyses were done in SPSS. AMOS was employed
420
+ to test the hypotheses and the theoretical model by structural equation modelling (SEM).
421
+
422
+ 2.2. Research questions and hypotheses
423
+ Based on the literature review, this study aims to address the following research questions
424
+ respectively:
425
+  R1: Can the technology acceptance model (TAM) be used for investigating consumer
426
+ acceptance of the use of artificial intelligence in online shopping?
427
+  R2: If so, what are the key factors influencing behavioural intention to visit AI-
428
+ powered webshops and apps?
429
+ Based on the Technology Acceptance Model, an initial theoretical model was developed
430
+ (Figure no. 2). The arrows that link constructs (latent variables such as COMP, EXP,
431
+ TRUST, SN, PEU, PU, ATT, BI) represent hypothesised causal relationships (hypotheses)
432
+ in the direction of arrows. One of the objectives of this study is to test those hypotheses.
433
+ Error terms for all observed indicators are indicated by e1 to e35, respectively.
434
+
435
+
436
+ AE
437
+ Consumer Acceptance of the Use of Artificial Intelligence
438
+ in Online Shopping: Evidence From Hungary
439
+
440
+ 162
441
+ Amfiteatru Economic
442
+
443
+ Figure no. 2. The initial theoretical model
444
+
445
+ 2.3. The sample
446
+ A sample size of 200 is an appropriate minimum for SEM in AMOS (Marsh, Balla, and
447
+ MacDonald, 1988), and a minimum of 10-20 subjects per parameter estimates in the model
448
+ are optimal (Schumacker and Lomax, 2010). Therefore, the ideal sample size is between
449
+ 380 and 760, considering the number of parameter estimates (38) in the initial model. The
450
+ actual sample size of 439 respondents fits into this category.
451
+ Of the sample of 439 respondents, 62.2% were female, 37.8% male. Their average age was
452
+ 32.2 years. 60,8% of the respondents had tertiary education, 38% had secondary education,
453
+ and 1.2% had primary education. Most respondents resided in county seats (47.6%); the
454
+ rest lived in other towns/cities (24.1%), villages (17.8%) and the capital (10.5%). Most
455
+ respondents were subscribed to both mobile and wired internet services (84.3%), while
456
+ 7.7% had only mobile internet, and 7.1% had only wired internet services. Only 0.9% of
457
+ respondents had not got any subscription to internet services (wired or mobile). There is no
458
+ data available on the distribution of the e-shoppers in Hungary, therefore, it is impossible to
459
+ tell if this sample reflects the characteristics of the e-shoppers’ population in Hungary.
460
+
461
+ 3. Results and discussion
462
+ The initial model (Figure no. 2), which proved to be too complex and did not fit the current
463
+ data (CMIN/DF=7.72; p=.00; GFI=,693; CFI=.723; RMSEA=.124; HOELTER 0.5= 65),
464
+ was absolutely rejected. Therefore, it was not appropriate to interpret any individual
465
+
466
+ C2
467
+ COMP
468
+ PEU1PEU2PEU3PEU4PEU5
469
+ PEU
470
+ e24
471
+ EXP
472
+ EXP
473
+ EXP
474
+ ATT1ATT2ATT3
475
+ BI2
476
+ TRUST
477
+ SN
478
+ SN
479
+ PU
480
+ P2
481
+ P3
482
+ PU4Artificial Intelligence in Wholesale and Retail
483
+ AE
484
+
485
+ Vol. 23 • No. 56 • February 2021
486
+ 163
487
+ parameter estimates, and further model modifications were required to obtain a better-
488
+ fitting model. Respecification of the initial model led to a nested model that fitted well and
489
+ is discussed further. During the respecification, the alternative model approach was used
490
+ (Malkanthie, 2015). To test the model, the same data set was used. Several modified
491
+ models were developed, and out of the theoretically justifiable models, the model with the
492
+ best data fit was selected (Figure no. 3) as suggested by Mueller and Hancock (2008).
493
+ The respecification process was started with testing the measurement model by a series of
494
+ Principal Component Analysis (PCA). Variables with factor loadings under 0.7 were
495
+ deleted. A rule of thumb in confirmatory factor analysis suggests that variables with factor
496
+ loadings under |0.7| must be dropped (Malkanthie, 2015). As a result, only one external
497
+ variable, which is related to trust (T2), remained in the model (Table no. 2). Perceived
498
+ Usefulness (PU) was measured by three variables (PU1, PU2 and PU3), whereas Perceived
499
+ Ease of Use was made up of two variables (PEU2 and PEU3), and Behavioural Intention
500
+ became unidimensional (B1). The attitude was composed of three variables (ATT1, ATT2
501
+ and ATT3). The nested model, which is theoretically consistent with the research goals,
502
+ contains eight hypotheses:
503
+  H1: Attitude has a positive effect on behavioural intention.
504
+  H2: Perceived usefulness positively affects behavioural intention.
505
+  H3: Perceived usefulness has a positive effect on attitude.
506
+  H4: Perceived ease of use positively influences attitude.
507
+  H5: Perceived ease of use positively influences perceived usefulness.
508
+  H6: Perceived ease of use has a positive impact on trust.
509
+  H7: Trust has a positive effect on perceived usefulness.
510
+  H8: Trust positively influences attitude.
511
+
512
+ Figure no. 3. The nested model
513
+
514
+
515
+ PU1PU2PU3
516
+ PU
517
+ H7
518
+ H3
519
+ H2
520
+ er
521
+ H8
522
+ H1
523
+ T2
524
+ ATT
525
+ BI1
526
+ H5
527
+ H4
528
+ H6
529
+ ATT3
530
+ TATT2
531
+ [ATT1
532
+ PEU
533
+ PEU2PEU3AE
534
+ Consumer Acceptance of the Use of Artificial Intelligence
535
+ in Online Shopping: Evidence From Hungary
536
+
537
+ 164
538
+ Amfiteatru Economic
539
+ 3.1. Validity
540
+ To investigate the extent to which a set of items reflect the theoretical latent-construct they
541
+ are designed to measure, both convergent and discriminant validity were checked.
542
+ Convergent validity suggests that the variables of a factor that are theoretically related are
543
+ expected to correlate highly. According to the Fornell-Larcker criterion for convergent
544
+ validity, the Average Variance Extracted (AVE) should be greater than 0.5. According to
545
+ the Hair, et al. (1998) criteria, AVE should be greater than 0.5, standardised factor loading
546
+ of all items should be above 0.5, and composite reliability should be above 0.7.
547
+ In the nested measurement model, each factor loading was above .84 (Table no. 2).
548
+ Table no. 2. Summary of means, standard deviations, normality,
549
+ validity and reliability measures
550
+ Cons-
551
+ truct Measurement Instrument Mean STD Z Skew
552
+ Z
553
+ Kurt
554
+ Loa-
555
+ ding
556
+ α
557
+ AVE
558
+ CR
559
+ Perceived
560
+ Usefulness
561
+ PU1. The use of AI in
562
+ retail (shopping ads and
563
+ webshops) allows me to
564
+ find the best deals.
565
+ 4.68
566
+ 1.53
567
+ -4.06
568
+ -1.40
569
+ 0.85
570
+ 0.91
571
+ 0.76
572
+ 0.91
573
+ PU2. The use of AI in
574
+ retail enhances my
575
+ effectiveness in
576
+ purchasing.
577
+ 4.67
578
+ 1.63
579
+ -4.56
580
+ -1.87
581
+ 0.89
582
+ PU3. The use of AI in
583
+ retail is useful to me.
584
+ 4.73
585
+ 1.69
586
+ -4.16
587
+ -2.70
588
+ 0.89
589
+ Perceived Ease
590
+ of Use
591
+ PEU2. Shopping does not
592
+ require a lot of my mental
593
+ efforts if supported by AI
594
+ (alternatives are offered by
595
+ AI).
596
+ 5.15
597
+ 1.62
598
+ -6.38
599
+ -0.59
600
+ 0.90
601
+ 0.88
602
+ 0.81
603
+ 0.9
604
+ PEU3. Shopping is not so
605
+ complicated if AI offers
606
+ products to me.
607
+ 5.06
608
+ 1.64
609
+ -6.44
610
+ -0.68
611
+ 0.90
612
+ Trust
613
+ T2. I trust in apps and
614
+ webshops that use AI.
615
+ 4.11
616
+ 1.62
617
+ -2.00
618
+ -2.90
619
+ 1.00
620
+ 1
621
+ n.a.
622
+ n.a.
623
+ Attitude
624
+ ATT1 Shopping in a
625
+ webshop/shopping app that
626
+ is powered by AI is a good
627
+ idea
628
+ 5.02
629
+ 1.63
630
+ -4.99
631
+ -1.58
632
+ 0.90
633
+ 0.9
634
+ 0.79
635
+ 0.92
636
+ ATT2 Shopping in a
637
+ webshop/shopping app that
638
+ is powered by AI is a wise
639
+ idea
640
+ 4.23
641
+ 1.62
642
+ -1.60
643
+ -2.39
644
+ 0.86
645
+ ATT3 I am positive
646
+ towards webshop/shopping
647
+ app that is powered by AI
648
+ 4.72
649
+ 1.70
650
+ -4.11
651
+ -1.87
652
+ 0.90
653
+
654
+ Artificial Intelligence in Wholesale and Retail
655
+ AE
656
+
657
+ Vol. 23 • No. 56 • February 2021
658
+ 165
659
+ Cons-
660
+ truct Measurement Instrument Mean STD Z Skew
661
+ Z
662
+ Kurt
663
+ Loa-
664
+ ding
665
+ α
666
+ AVE
667
+ CR
668
+ Behavioural
669
+ Intention
670
+ BI1 I intend to visit
671
+ webshops and use
672
+ shopping apps that are
673
+ powered by AI more
674
+ frequently.
675
+ 3.35
676
+ 1.78
677
+ 2.13
678
+ -3.93
679
+ 1.0
680
+ 1
681
+ n.a.
682
+ n.a.
683
+ Notes: STD=Standard Deviation, Z Skew=Z score for skewness, Z Kurt=Z score for
684
+ Kurtosis, α=Cronbach's alpha, AVE=Average Variance Extracted, CR=Composite
685
+ Reliability, N=439.
686
+
687
+ Moreover, all AVE scores were also well above the threshold level (AVE (ATT)=0.79;
688
+ AVE (PU)=.76 and AVE (PEU)=0.81), and all CR scores exceeded 0.7 (CR (PU)=.91; CR
689
+ (PEU)=0.90 and CR (ATT)=0.92). Therefore, the model meets both the Fornell-Larcker
690
+ (1981) criterion and the Hair et al. (1998) criteria for convergent validity, so the internal
691
+ consistency of the model is acceptable.
692
+ To assess discriminant validity, i.e. the extent to which a construct is truly distinct to other
693
+ constructs, AVEs were compared with squared inter-construct correlations (SIC). AVE
694
+ scores higher than SIC scores indicate that discriminant validity is acceptable (ATT
695
+ AVE=0.79, SIC1=0.61 and SIC2=0.32; PU AVE=0.76, SIC1=0.40 and SIC2=0.61; PEU
696
+ AVE=0.81, SIC1=0.40 and SIC2=0.32). Discriminant validity was also confirmed by
697
+ investigating correlations among the constructs. Since there were no correlations above .85,
698
+ which is a threshold limit of poor discriminant validity in structural equation modelling
699
+ (David, 1998), results also confirmed adequate discriminant validity (PEU*T2=0.52;
700
+ PEU*PU=0.64;
701
+ PEU*ATT=0.57;
702
+ PEU*BI1=0.43;
703
+ T2*PU=0.73;
704
+ T2*ATT=0.74;
705
+ T2*BI1=0.53; PU*ATT=0.78; PU*BI1=0.64; ATT*BI1=0.66).
706
+ 3.2. Reliability
707
+ To test the accuracy and consistency of the nested model, three reliability tests were used:
708
+ Cronbach's alpha (α), the Average Variance Extracted index (AVE) and Composite
709
+ Reliability (CR). The threshold value for an acceptable Cronbach's alpha is .70 (Cronbach,
710
+ 1951). The measurement model is acceptable if all estimates are significant and above 0.5
711
+ or 0.7 ideally; AVEs for all constructs are above 0.5 (Forner and Larcker, 1981); and
712
+ finally, CRs for all constructs are above 0.7 (Malkanthie, 2015). Table no. 2 shows that the
713
+ calculated Cronbach's alphas of all constructs were at least .87 or higher, and the AVE
714
+ scores were also higher than 0.76, as well as the CRs were above 0.9; therefore, the
715
+ reliability of the measurement model is optimal.
716
+ 3.3. Model fit
717
+ Absolute- and relative model fits were tested. Each absolute measure was significant and
718
+ indicated a good fit. Although Chi-square statistics are sensitive to large sample size and
719
+ assume a multivariate normal distribution (Kelloway, 1998), even those measures were
720
+ acceptable. However, other model fit indexes are better to consider as criteria. Therefore,
721
+ the goodness-of-fit index (GFI), the adjusted goodness-of-fit index (AGFI), the root mean
722
+ squared error of approximation (RMSEA) and the standardised root mean squared residual
723
+
724
+ AE
725
+ Consumer Acceptance of the Use of Artificial Intelligence
726
+ in Online Shopping: Evidence From Hungary
727
+
728
+ 166
729
+ Amfiteatru Economic
730
+ (SRMR) were also examined. All of them indicated a good absolute model fit. (Absolute
731
+ measures: Chi square=34.154 (DF=29); Probability level=0.23; CMIN/DF=1.18;
732
+ GFI=0.98; AGFI=0.96; RMSEA=0.02; SRMR=0.04). As far as the relative model fit is
733
+ concerned, TLI or NNFI, GFI, AGFI, NFI, IFI, CFI and Critical N (CN or HOELTER)
734
+ were calculated. All but CN range from zero to one. Values exceeding .9 show an
735
+ acceptable fit, above .95 a good fit (Bentler and Bonnet, 1980). CN (HOELTER), which
736
+ favours large samples over small ones (Bollen, 1990), is an improved method for
737
+ investigating model fit (Hoelter, 1983). CN should be above 200 to indicate a good model
738
+ fit. (Relative measures: TLI/NNFI=0.98; GFI=0.98; AGFI=0.96; NFI=0.93; IFI=0.99;
739
+ CFI=0.99 and HOELTER (CN)=546). The results of the absolute and relative model fit test
740
+ confirmed that the structural model is acceptable and suitable for the analysis and
741
+ interpretation of the parameter estimates. Therefore, it can be concluded that the technology
742
+ acceptance model is suitable for investigating consumer acceptance of the use of artificial
743
+ intelligence in online shopping, which is the answer to the first research question (R1).
744
+ 3.4. Hypothesis testing and estimates
745
+ Because of the non-normality of the variables in the nested model, the asymptotically
746
+ distribution-free (ADF) method was used to estimate parameters in AMOS. ADF calculates
747
+ the asymptotically unbiased estimates of the chi-square goodness-of-fit test, the parameter
748
+ estimates, and the standard errors. The limitation of ADF is that it needs a large sample size
749
+ (Bian, 2012), which criterion was met in this study (N=439). Skewness and Kurtosis z-
750
+ values of the variables were out of the range of the normal distribution that is -2 and +2
751
+ (George and Mallery, 2010). Moreover, the p values of the variables were significant
752
+ (p=.000) in the Shapiro-Wilk and Kolmogorov-Smirnov tests, which also confirmed non-
753
+ normality.
754
+ To address the second research question (R2) and to determine the key factors influencing
755
+ behavioural intention to use AI-powered webshops and apps, hypotheses were tested in the
756
+ structural model (Table no. 3).
757
+ Table no. 3. Direct, indirect, total effects and hypothesis testing
758
+ Hypothesis
759
+ Relationship
760
+ P
761
+ St. direct eff.
762
+ St. indirect eff.
763
+ St. total eff.
764
+ Result
765
+ H1
766
+ BI1 ← ATT
767
+ ***
768
+ 0.41
769
+ 0.00
770
+ 0.41
771
+ accepted
772
+ H2
773
+ BI1 ← PU
774
+ ***
775
+ 0.32
776
+ 0.19
777
+ 0.51
778
+ accepted
779
+ H3
780
+ ATT ← PU
781
+ ***
782
+ 0.48
783
+ 0.00
784
+ 0.48
785
+ accepted
786
+ H4
787
+ ATT←PEU
788
+ 0.1
789
+ 0.09
790
+ 0.48
791
+ 0.57
792
+ rejected
793
+ H5
794
+ PU ← PEU
795
+ ***
796
+ 0.35
797
+ 0.28
798
+ 0.64
799
+ accepted
800
+ H6
801
+ T2 ← PEU
802
+ ***
803
+ 0.52
804
+ 0.00
805
+ 0.52
806
+ accepted
807
+ H7
808
+ PU ← T2
809
+ ***
810
+ 0.55
811
+ 0.00
812
+ 0.55
813
+ accepted
814
+ H8
815
+ ATT ← T2
816
+ ***
817
+ 0.35
818
+ 0.26
819
+ 0.61
820
+ accepted
821
+ The arrows linking constructs represent hypotheses in the direction of arrows in the nested
822
+ model (Figure no. 3 and Figure no. 4). Asterisks signal statistically significant relations
823
+ between constructs. Gamma estimates were calculated from exogenous construct to
824
+ endogenous construct, and beta estimates between two endogenous constructs. Figure no. 4
825
+ shows the standardised estimates, loadings and residuals regarding the relationships
826
+
827
+ Artificial Intelligence in Wholesale and Retail
828
+ AE
829
+
830
+ Vol. 23 • No. 56 • February 2021
831
+ 167
832
+ between constructs and observed indicators. A hypothesis was accepted if the presence of a
833
+ statistically significant relationship in the predicted direction was confirmed.
834
+ As Table no. 3 shows, all hypotheses were accepted except for H4. So, the present findings,
835
+ except for the relationship between perceived ease of use and attitude, are consistent with
836
+ the Technology Acceptance Model proposed by Davis (1986). Surprisingly, perceived ease
837
+ of use (PEU) was found to have no direct, significant effect on attitude (ATT), which is not
838
+ in agreement with the original TAM (H4 rejected). This discrepancy could be attributed to
839
+ the fact that shopping is not too complicated in AI-powered webshops, and it does not
840
+ require too much mental effort. However, this slightly unexpected result coincides with the
841
+ findings of a previous research by Ha and Stoel (2008), who examined the effect of PEU on
842
+ attitude towards online shopping.
843
+ In this study, with H5 and H6 accepted, perceived ease of use (PEU) was found to have a
844
+ significant, direct, positive impact on both the perceived usefulness (PU) and trust (T2). It
845
+ suggests that the easier it is for a consumer to use an AI-powered webshop, the higher level
846
+ of customer trust and perceived usefulness can be expected. Consumers trust in AI-powered
847
+ shopping apps and stores that are easy to use, and consider those that are too complicated
848
+ less useful. Similar results were obtained by Ha and Stoel (2008), who focused on
849
+ consumers' acceptance of e-shopping. Gefen, Karahanna and Straub (2003) also found that
850
+ perceived ease of use positively affected the perceived usefulness of a B2C website and the
851
+ trust in an e-vendor.
852
+
853
+ Figure no. 4. Parameter estimates of the nested model
854
+ Trust in AI-powered webshops has a central role in forming attitudes and perceived
855
+ usefulness. Similar to what Gefen, Karahanna and Straub (2003), and Ha and Stoel (2008)
856
+ found, trust directly influenced perceived usefulness (H7 accepted). Moreover, trust also
857
+ impacted attitude (H8 accepted), in line with the research findings of Ha and Stoel (2008).
858
+
859
+ ,72
860
+ ,78
861
+ ,78
862
+ PU1
863
+ PU2
864
+ PU3
865
+ +
866
+ ,85
867
+ 488
868
+ 89
869
+ 62
870
+ PU
871
+ 32
872
+ ,55
873
+ 48
874
+ 68
875
+ ,47
876
+ ,35
877
+ ATT
878
+ ,41
879
+ T2
880
+ BI1
881
+ ,35
882
+ ,90
883
+ 52
884
+ ,86
885
+ ,90
886
+ ,09
887
+ 81
888
+ ,74
889
+ ,81
890
+ 00
891
+ ATT3
892
+ ATT2
893
+ ATT1
894
+ PEU
895
+ e11
896
+ 90
897
+ ,90
898
+ ,81
899
+ ,82
900
+ PEU2
901
+ PEU3
902
+ eg
903
+ e10AE
904
+ Consumer Acceptance of the Use of Artificial Intelligence
905
+ in Online Shopping: Evidence From Hungary
906
+
907
+ 168
908
+ Amfiteatru Economic
909
+ The strongest direct effect was found between trust and perceived usefulness (H7
910
+ accepted). It suggests that the more we trust in Artificial Intelligence during the online
911
+ shopping journey, the more likely it is that we consider AI-powered apps and webshops
912
+ useful. Besides, a higher level of trust forms a more positive attitude towards shopping in
913
+ such webshops. Perceived usefulness has a central role in this model as it (PU) significantly
914
+ impacted attitude (H3 accepted) and behavioural intention (H2 accepted). The more useful
915
+ we find the use of artificial intelligence in online shopping believing that it allows us to
916
+ grab the best deals, the more likely we are to consider it a wise decision to do the shopping
917
+ in AI-powered webshops and apps more frequently. Not surprisingly, attitude towards AI-
918
+ powered webshops and apps was found to have a strong, significant, positive direct impact
919
+ on behavioural intention (H1 accepted). It suggests that forming consumers’ attitude plays a
920
+ vital role in increasing the traffic of AI-powered webshops and apps (Figure no. 4).
921
+ Although there was no significant direct relationship between perceived ease of use and
922
+ attitude, the indirect effect of PEU on attitude (PEU->ATT=0.48) was quite strong, similar
923
+ to its indirect impact on behavioural intention (PEU->BI1=0.43). Also, trust was found to
924
+ indirectly influence behavioural intention (T2->BI1=0.42). It suggests that if shopping
925
+ requires much mental effort and seems to be complicated in AI-powered webshops and
926
+ apps, consumers tend to form stronger negative attitudes towards them and also tend to
927
+ trust them less, which will result in weaker consumer intention to visit such webshops.
928
+ In the nested model perceived usefulness had the highest total effect on behavioural
929
+ intention. Therefore, AI-powered webshops and apps are advised to increase the level of
930
+ perceived usefulness to succeed by enabling customers to maximise purchase effectiveness
931
+ to grab the best deals, i.e. the ideal product with the highest utility.
932
+
933
+ Conclusions
934
+ This research extends our knowledge of consumer acceptance of the use of artificial
935
+ intelligence in online shopping in many aspects. The widely used technology acceptance
936
+ model (TAM) was proved to be suitable for investigating consumer acceptance of the use
937
+ of artificial intelligence in online shopping.
938
+ As expected, it was confirmed in the nested model that the key factors influencing
939
+ consumer’ behavioural intention to use AI-powered webshops and apps are trust, perceived
940
+ usefulness, perceived ease of use and attitudes. In contrast to the original TAM (Davis,
941
+ 1986), the direct relationship between perceived ease of use and attitudes was insignificant.
942
+ Nevertheless, it does not mean that user-friendliness of a webshop is not crucial as
943
+ perceived ease of use indirectly affects attitude and the behavioural intention. Instead, user-
944
+ friendliness and flawless operation of an artificial intelligence-powered website are the
945
+ prerequisites for market success.
946
+ Building trust has a central role in consumer acceptance of the use of artificial intelligence
947
+ in online shopping. If consumers do not trust in an AI-powered webshop/app, they tend to
948
+ consider it less useful and form a negative attitude towards it, which will result in less
949
+ online traffic. Also, AI must provide online consumers with tailor-made offerings to grab
950
+ the best deals, i.e. products with the highest value; and it is expected to shorten the product
951
+ search time to enhance shopping effectiveness. Not surprisingly, the favourable attitude
952
+
953
+ Artificial Intelligence in Wholesale and Retail
954
+ AE
955
+
956
+ Vol. 23 • No. 56 • February 2021
957
+ 169
958
+ towards AI-powered webshops leads to more frequent online traffic in such electronic
959
+ stores.
960
+ Considering the strong positive impact of the recent COVID-19 crisis on e-commerce, the
961
+ use of artificial intelligence in online shopping is expected to expand further. According to
962
+ Bloomberg (2020) the pandemic lockdowns have a dual effect on consumer behaviour on
963
+ the development of AI. Nowadays, it is more important than ever to create a personalised
964
+ customer journey, to meet customers' demand and to provide a greater online shopping
965
+ experience. In these efforts, artificial intelligence can be a very effective tool, which was
966
+ confirmed by the research findings of this paper.
967
+ This study has several practical applications. It is useful for webshop owners and online
968
+ marketing managers to understand how consumers adapt to the new technology, i.e. the use
969
+ of artificial intelligence in online shopping. It is also beneficial to academics and
970
+ researchers who are interested in the adaptation of the Technology Acceptance Model in
971
+ online shopping. Those who are interested in the role of trust in consumer choices in the
972
+ online environment will also benefit from this study.
973
+ As far as the future research directions are concerned, it would be advisable to repeat this
974
+ study in a multi-cultural context. It might also be useful to test the model of the Technology
975
+ Readiness Index proposed by Parasuraman (2000) and to compare the results presented
976
+ here with the new findings.
977
+
978
+ Acknowledgements
979
+ “The described article/presentation/study was carried out as part of the EFOP-3.6.1-16-
980
+ 2016-00011 “Younger and Renewing University – Innovative Knowledge City –
981
+ institutional development of the University of Miskolc aiming at intelligent specialisation”
982
+ project implemented in the framework of the Szechenyi 2020 program. The realization of
983
+ this project is supported by the European Union, co-financed by the European Social
984
+ Fund.”
985
+
986
+ References
987
+ Ajzen, I., 1985. The Theory of Planned Behavior. Organisational Behavior and Human
988
+ Decision Processes, 50, pp.179-211.
989
+ André, Q., Carmon, Z., Wertenbroch, K., Crum, A., Frank, D., Goldstein, W., Huber, J.,
990
+ Boven, L., Weber, B. and Yang, H., 2017. Consumer Choice and Autonomy in the Age
991
+ of Artificial Intelligence and Big Data. Customer needs and solutions, 5(1-2), pp.28-37.
992
+ Aranyossy, M. and Magisztrák, B., 2016. A vásárlói bizalom hatása az e-kereskedelmi
993
+ vásárlási hajlandóságra. Marketing & Menedzsment, 3-4, pp.73-87.
994
+ Asling, D., 2017. 19 Powerful Ways To Use Artificial Intelligence In eCommerce. [online]
995
+ Available
996
+ at:
997
+ <https://blog.linnworks.com/artificial-intelligence-in-ecommerce>
998
+ [Accessed 27 August 2020]
999
+ Bakos, J.Y., 1997. Reducing buyer search costs: implications for electronic marketplaces.
1000
+ Management Science, 43(12), pp.1676-1692.
1001
+
1002
+ AE
1003
+ Consumer Acceptance of the Use of Artificial Intelligence
1004
+ in Online Shopping: Evidence From Hungary
1005
+
1006
+ 170
1007
+ Amfiteatru Economic
1008
+ Barmada, N., 2020. Grow your business in the Nordics when Amazon Sweden launches this
1009
+ year. [online] Available at: <https://blog.linnworks.com/grow-your-business-across-
1010
+ borders-when-amazon-sweden-launches-this-year> [Accessed 27 August 2020].
1011
+ Bentler, P.M. and Bonnet, D.G., 1980. Significance tests and goodness-of-fit in the analysis
1012
+ of covariance structure. Psychological Bulletin, 88(3), pp.588-606.
1013
+ Bian, H., 2012. Structual Equation Modeling Using Amos. New York: Routledge.
1014
+ Bloomberg, 2020. Coronavirus will finally give artificial intelligence its moment. [online]
1015
+ Available
1016
+ at:
1017
+ <https://economictimes.indiatimes.com/small-biz/startups/features/
1018
+ coronavirus-will-finally-give-artificial-intelligence-its-
1019
+ moment/articleshow/76477021.cms?from=mdr> [Accessed 27 November 2020].
1020
+ Bollen, K.A., 1990. Overall fit in covariance structure models: two types of sample size
1021
+ effects. Psychological Bulletin, 107(2), pp.256-259.
1022
+ Cătoiu, I., Orzan, M., Macovei, O.I. and Iconaru, C., 2014. Modelling Users’ trust in online
1023
+ social networks. Amfiteatru Economic, 16(35), pp.289-302.
1024
+ Cronbach, L.J., 1951. Coefficient alpha and the internal structure of tests. Psychometrika
1025
+ 16, pp.297-334. https://doi.org/10.1007/BF02310555.
1026
+ Daley, S. 2018. 19 examples of artificial intelligence shaking up business as usual [online]
1027
+ Available
1028
+ at:
1029
+ <https://builtin.com/artificial-intelligence/examples-ai-in-industry>
1030
+ [Accessed 17 August 2020]
1031
+ Davenport, T.H. and Ronanki, R., 2018. Artificial Intelligence for the Real World. Harvard
1032
+ Business Review, January-February 2018., pp.108-116.
1033
+ Davis, F.D., 1986. A Technology Acceptance Model for empirical testing new end-user
1034
+ information systems: Theory and results, Doctoral Dissertation, MIT. [online]
1035
+ <https://dspace.mit.edu/handle/1721.1/15192> [Accessed 17 August 2020].
1036
+ Dhagarra, D., Goswami, M. and Kumar, G., 2020. Impact of Trust and Privacy Concerns
1037
+ on Technology Acceptance in Healthcare: An Indian Perspective. International Journal
1038
+ of Medical Informatics, 141, pp.104164. doi:10.1016/j.ijmedinf.2020.104164.
1039
+ Dumitriu, D. and Popescu, M.A.M., 2020. Artificial Intelligence Solutions for Digital
1040
+ Marketing. Procedia Manufacturing, 46, pp.630-636.
1041
+ Gefen, D., Karahanna, E. and Straub, D.W., 2003. Trust and TAM in Online Shopping: An
1042
+ Integrated Model. MIS Quarterly, 1, pp.51-90.
1043
+ European Commission, 2018. A definition of AI: Main Capabilities and Disciplines.
1044
+ Definition developed for the purpose of the AI HLEG's deliverables. [online] Available
1045
+ at: <https://ec.europa.eu/digital-single-market/en/news/definition-artificial-intelligence-
1046
+ main-capabilities-and-scientific-disciplines> [Accessed 07 August 2020].
1047
+ Fishbein, M. and Ajzen, I., 1975. Belief, attitude, intention and behavior: An introduction
1048
+ to theory and research. Massachusetts: Addison-Wesley.
1049
+ Fornell, C. and Larcker, D., 1981. Evaluating Structural Equation Models with
1050
+ Unobservable Variables and Measurement Error. Journal of Marketing Research, 18(1),
1051
+ pp.39-50. doi:10.2307/3151312.
1052
+ George, D. and Mallery, M., 2010. SPSS for Windows Step by Step: A Simple Guide and
1053
+ Reference, 17.0 update (10a ed.) Boston: Pearson.
1054
+
1055
+ Artificial Intelligence in Wholesale and Retail
1056
+ AE
1057
+
1058
+ Vol. 23 • No. 56 • February 2021
1059
+ 171
1060
+ Ha, S. and Stoel, L. 2008. Consumer e-shopping acceptance: Antecedents in a technology
1061
+ acceptance model. Journal of Business Research, 62, pp.565-571.
1062
+ Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W.C. 1998. Multivariate data analysis:
1063
+ (5th ed.). Upper Saddle River, NJ: Prentice Hill.
1064
+ Haller, K., Lee, J. and Cheung, J., 2020. Meet the 2020 consumers driving change. Why
1065
+ brands must deliver on omnipresence, agility, and sustainability. IBM Institute for
1066
+ Business Value. [online] Available at: <https://www.ibm.com/thought-leadership/
1067
+ institute-business-value/report/consumer-2020#> [Accessed 27 November 2020].
1068
+ Hoelter, J.W., 1983. The analysis of covariance structures: Goodness-of-fit indices.
1069
+ Sociological Methods and Research, 11, pp.325-344.
1070
+ Hu, K. and O'Brien, S., 2016. Applying TAM (Technology Acceptance Model) to testing MT
1071
+ acceptance.
1072
+ [online]
1073
+ Available
1074
+ at:
1075
+ <https://ec.europa.eu/info/sites/info/files/
1076
+ tef2016_kehu_en.pdf > [Accessed 07 July 2020].
1077
+ Kim, D., Ferrin, D.L. and Rao, H.R. 2008. A trust-based consumer decision-making model
1078
+ in electronic commerce: The role of trust, perceived risk, and their antecedents.
1079
+ Decision Support Systems, 44(2), pp.544-564.
1080
+ Kelloway, E.K., 1998. Using LISREL for structural equation modeling: A researcher's
1081
+ guide. Thousand Oaks, CA: Sage.
1082
+ Kwong, C.K., Jiang, H. and Luo, X., 2016. AI-based methodology of integrating affective
1083
+ design, engineering, and marketing for defining design specifications of new products.
1084
+ Engineering Applications of Artificial Intelligence, 47, pp.49-60.
1085
+ Legris, P., Ingham, J. and Collerette, P., 2003. Why do people use information technology?
1086
+ A critical review of the technology acceptance model. Information & Management
1087
+ 40(3), pp.191-204.
1088
+ Lynch, J.G. and Ariely, D., 2000. Wine Online: Search Costs Affect Competition on Price,
1089
+ Quality, and Distribution. Marketing Science, 19(1), pp.83-103.
1090
+ Malkanthie, A., 2015. Structural Equation Modeling with AMOS. Lap Lambert Academic
1091
+ Publishing, Germany.
1092
+ Marsh, H.W, Balla, J.R. and MacDonald, R.P., 1988. Goodness-of-fit indexes in
1093
+ confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 88,
1094
+ pp.245-258.
1095
+ Maynard, N., 2019. Juniper research. How AI can revive retail? [online] Available at:
1096
+ <https://www.juniperresearch.com/document-library/white-papers/how-ai-can-revive-
1097
+ retail> [Accessed 07 August 2020]
1098
+ Meticulous Market Research, 2020. Artificial Intelligence (AI) in Retail Market Worth
1099
+ $19.9 billion by 2027- Exclusive Report Covering Pre and Post COVID-19 Market
1100
+ Analysis. [online] Available at: <https://www.prnewswire.com/news-releases/artificial-
1101
+ intelligence-ai-in-retail-market-worth-19-9-billion-by-2027--exclusive-report-covering-
1102
+ pre-and-post-covid-19-market-analysis-by-meticulous-research-301098029.html>
1103
+ [Accessed 27 November 2020]
1104
+ Mueller, R. and Hancock, G., 2008. Best practices in structural equation modeling. In:
1105
+ J. Osborne, ed. 2008. Best practices in quantitative methods. Thousand Oaks, CA: Sage
1106
+ Publications, Inc. pp. 488-508.
1107
+
1108
+ AE
1109
+ Consumer Acceptance of the Use of Artificial Intelligence
1110
+ in Online Shopping: Evidence From Hungary
1111
+
1112
+ 172
1113
+ Amfiteatru Economic
1114
+ Onete, B., Constantinescu, M. and Filip, A., 2008. Internet buying behavior. Case study:
1115
+ research of AES students' behavior regarding online shopping. Amfiteatru Economic,
1116
+ November, pp.18-24.
1117
+ Pantano, E. and Pizzi, G., 2020. Forecasting artificial intelligence on online customer
1118
+ assistance: evidence from chatbot patents analysis. Journal of Retailing and Consumer
1119
+ Services, 55, pp.102096.
1120
+ Parasuraman, A., 2000. Technology Readiness Index (TRI): A Multiple item scale to
1121
+ measure readiness to embrace new technologies. Journal of Service Research, 2(4),
1122
+ pp.307-320.
1123
+ Park, S.Y., 2009. An Analysis of the Technology Acceptance Model in Understanding
1124
+ University Students' Behavioral Intention to Use e-Learning. Educational Technology &
1125
+ Society, 12(3), pp.150-162.
1126
+ Paschen, J., Wilson, M. and Ferreira, J., 2020. Collaborative intelligence: How human and
1127
+ artificial intelligence create value along the B2B sales funnel. Business Horizons, 63(3),
1128
+ pp.403-414.
1129
+ Pricewaterhouse Coopers, 2018. Künstliche Intelligenz als Innovationsbeschleuniger in
1130
+ Unternehmen – Zuversicht und Vertrauen in Künstliche Intelligenz. [online] Available
1131
+ at: <https://www.pwc.de/de/digitale-transformation/ki-als-innovationsbeschleuniger-in-
1132
+ unternehmen-whitepaper.pdf > [Accessed 07 August 2020].
1133
+ Pusztahelyi, R., 2020. Emotional AI and its challenges in the viewpoint of online
1134
+ marketing. Curentul Juridic, 23(2), pp.13-31.
1135
+ Rajagopal, P., 2002. An innovation-diffusion view of implementation of enterprise resource
1136
+ planning (ERP) systems and development of a research model. Information &
1137
+ Management, 40(2), pp.87-114.
1138
+ Reichheld, F.F. and Schefter, P., 2000. E-loyalty your secret weapon on the web. Harvard
1139
+ Business Review, 78(4), pp.105-113.
1140
+ Roetzer, P., 2017. 6 Limitations of Marketing Artificial Intelligence, According to Experts.
1141
+ [online]
1142
+ Available
1143
+ at:
1144
+ <https://www.marketingaiinstitute.com/blog/limitations-of-
1145
+ marketing-artificial-intelligence> [Accessed 25 September 2020].
1146
+ Rust, R.T. and Huang, M.H., 2014. The Service Revolution and the Transformation of
1147
+ Marketing Science. Marketing Science, 33(2), pp.206-221.
1148
+ Schepman, A. and Rodway, P., 2020. Initial validation of the general attitudes towards
1149
+ Artificial Intelligence Scale. Computers in Human Behavior Reports, 1, pp.
1150
+ 100014. DOI: 10.1016/j.chbr.2020.100014.
1151
+ Schumacker, R.E. and Lomax, R.G., 2010. A Beginner's Guide to Structural Equation
1152
+ Modeling. New York: Routledge. https://doi.org/10.4324/9780203851319.
1153
+ Shankar, V., 2018. How Artificial Intelligence (AI) Is Reshaping Retailing. Journal of
1154
+ Retailing, 94(4), pp.343-348.
1155
+ Smidt, F. and Power, B. 2020. 8 ways consumers across Europe adapted their shopping
1156
+ behaviour this year. [online] Available at: <https://www.thinkwithgoogle.com/intl/en-
1157
+ cee/insights-trends/industry-perspectives/consumers-adapted-shopping-behaviour-
1158
+ covid/> [Accessed 27 August 2020].
1159
+ Stigler, G.J., 1961. The Economics of Information. Journal of Political Economy, 69(3),
1160
+ pp.213-225.
1161
+
1162
+ Artificial Intelligence in Wholesale and Retail
1163
+ AE
1164
+
1165
+ Vol. 23 • No. 56 • February 2021
1166
+ 173
1167
+ Stiegler, G.J. and Becker, G.S., 1977. De Gustibus Non Est Disputandum. The American
1168
+ Review. 67(2), pp.76-90.
1169
+ Thatcher, J.B., Carter, M., Li, X. and Rong, G., 2013. A Classification and Investigation of
1170
+ Trustees in B-to-C e-Commerce: General vs. Specific Trust. Communications of the
1171
+ Association for Information Systems. 32(4). 10.17705/1CAIS.03204.
1172
+ Venkatesh, V., 2000. Determinants of perceived ease of use: integrating control, intrinsic
1173
+ motivation, and emotion into the technology acceptance model. Information Systems
1174
+ Research, 11(4), pp.342-365.
1175
+ Venkatesh, V., Morris, M.G., Davis, G.B. and Davis, F.D., 2003. User Acceptance of
1176
+ Information Technology: Toward a Unified View, MIS Quarterly, 27(3), pp.425-478.
1177
+ Vijayasarathy, L.R., 2004. Predicting consumer intentions to use online shopping: the case
1178
+ for an augmented technology acceptance model. Information & Management, 41,
1179
+ pp.747-762.
1180
+ Weber, F. and Schütte, R., 2019. A Domain-Oriented Analysis of the Impact of Machine
1181
+ Learning-The Case of Retailing. Big Data Cognition Computation, 3(1), pp.1-14.
1182
+ Yoo, W.-S., Lee, Y. and Park, J.K., 2010. The role of interactivity in e-tailing: Creating
1183
+ value and increasing satisfaction. Journal of Retailing and Consumer Services, 17(2),
1184
+ pp.89-96.
1185
+
INAzT4oBgHgl3EQfU_y9/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
IdE0T4oBgHgl3EQfRwCH/content/tmp_files/2301.02212v1.pdf.txt ADDED
The diff for this file is too large to render. See raw diff
 
IdE0T4oBgHgl3EQfRwCH/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
ItE1T4oBgHgl3EQfYARl/content/2301.03133v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02282ebd2236f2ed503ec486852004a4fcf433f92aad0b8d4825332f0a22bc33
3
+ size 816807
ItE1T4oBgHgl3EQfYARl/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08b024494970378e88b06a125be0d7bcabcc46b94bb5bf82c129248530f9be4f
3
+ size 71318
N9E3T4oBgHgl3EQfZQq9/content/2301.04496v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc7248e9ba36522a6eeca0600ed6d6b400f1284cfaa9e202ec1e42e29757364e
3
+ size 596717