Fluctuations for the Sherrington–Kirkpatrick spin glass model near the critical temperature

Partha S. Dey ⋆, Taegu Kang † · 2026

Read Fluctuations for the Sherrington–Kirkpatrick spin glass model near the critical temperature by Partha S. Dey ⋆, Taegu Kang † (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-05636.txt.

Introduction

Dey and Kang consider the Sherrington-Kirkpatrick spin glass with zero external field at inverse temperature β above zero. Here c_N equals N to the 1/3 power times one minus β_N squared. Assuming c_N stays bounded away from zero, they prove a log-partition variance formula. Variance equals negative one-half log of one minus β_N squared minus β_N squared over two, plus order c_N to the negative 3/2. Consequently, variance at one minus c times N to the negative 1/3 equals 1/6 log N plus O(1).

Catalog record

CREATORPartha S. Dey ⋆, Taegu Kang †
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52324
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.05636

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Fluctuations for the Sherrington–Kirkpatrick spin glass model near the critical temperature Partha S. Dey ⋆, Taegu Kang † Abstract . We consider the Sherrington–Kirkpatrick spin glass model with zero external field and at inverse temperature β > 0 \beta>0 . Let F N ​ ( β ) F_{N}(\beta) be the corresponding log-partition function. Under the assumption that c N := N 1 / 3 ​ ( 1 − β N 2 ) c_{N}:=N^{1/3}(1-\beta_{N}^{2}) is bounded away from 0 0 , we prove that Var ( F N ( β N ) ) = − 1 2 ln ( 1 − β N 2 ) − β N 2 / 2 + 𝒪 ( c N − 3 / 2 ) . \var(F_{N}(\beta_{N}))=-\frac{1}{2}\ln(1-\beta_{N}^{2})-{\beta_{N}^{2}}/{2}+\mathcal{O}(c_{N}^{-3/2}). As a consequence, we obtain Var ( F N ( 1 − c N − 1 / 3 ) ) = 1 6 ln N + O ( 1 ) \var(F_{N}(1-cN^{-1/3}))=\frac{1}{6}\ln N+O(1) for any fixed constant c ∈ ( 0 , ∞ ) c\in(0,\infty) . We also prove a Gaussian central limit theorem for the centered and scaled F N ​ ( β N ) F_{N}(\beta_{N}) . #1# 2020 Mathematics Subject Classification Primary: 60F05, 82D30. 1. Introduction We consider the Sherrington-Kirkpatrick (SK) spin glass model at inverse temperature \beta>0 and with zero external field. For each positive integer N , we define the configuration space \Sigma_{N}=\{-1,+1\}^{N} . For a configuration \boldsymbol{\sigma}=(\sigma_{i})_{i\in[N]}\in\Sigma_{N} , the Hamiltonian of the SK model is given by \displaystyle H_{N,\beta}(\boldsymbol{\sigma})=\frac{\beta}{\sqrt{N}}\sum_{1\leqslant i<j\leqslant N}g_{ij}\sigma_{i}\sigma_{j}, where g_{ij} are independent and identically distributed (i.i.d.) standard Gaussian random variables. We define the partition function, Gibbs measure, and free energy for the SK model, respectively, by Z_{N}(\beta)=\sum_{\boldsymbol{\sigma}\in\Sigma_{N}}\exp({H_{N,\beta}(\boldsymbol{\sigma})}) , G_{N,\beta}(\boldsymbol{\sigma})=\exp(H_{N,\beta}(\boldsymbol{\sigma}))/Z_{N}(\beta) , and F_{N}(\beta)=\ln Z_{N}(\beta) . The Sherrington-Kirkpatrick spin glass model has been a subject of intense study since its introduction, with the primary goal of characterizing its limiting free energy. A landmark achievement in this field was the rigorous proof of the Parisi formula, which yields the free-energy limit at any temperature. This was first established by Talagrand [ 8 ] for the standard SK model and later extended to mixed p -spin models by Panchenko [ 7 ] . While the Parisi formula describes the first-order limit of the free energy, the study of its fluctuations focuses on the second-order behavior. In the high-temperature regime \beta<1 , the first rigorous proof of the central limit theorem (CLT) for the free energy was established by Aizenman, Lebowitz, and Ruelle [ 1 ] , who showed that for the entire regime \beta<1 , the fluctuations of the free energy are Gaussian with O(1) variance. This foundational result was later re-established by Comets and Neveu [ 5 ] using an alternative martingale-based approach. In contrast, the low-temperature regime \beta>1 exhibits much larger fluctuations. Chatterjee [ 3 ] provided a general upper bound in this region, proving that the variance of the free energy is at most of the order N/\ln N . As the temperature approaches the critical value \beta_{c}=1 , the fluctuations of the free energy are predicted to diverge. In the physics literature, it is widely believed that at \beta=1 , the variance behaves as \frac{1}{6}\ln N+O(1) . However, establishing this scaling rigorously remains a formidable challenge. A significant step forward was made by Chen and Lam [ 4 ] , who established that the variance at the critical temperature is at most of the order (\ln N)^{2} . For the spherical SK model, which is often more tractable due to its connections with random matrix theory, Baik and Lee [ 2 ] provided a complete characterization of the free energy fluctuations, identifying a transition to the Tracy–Widom distribution at low temperatures. Recently, Landon [ 6 ] characterized the near-critical fluctuations for the spherical model within a critical window of size \sqrt{\ln N}\cdot N^{-1/3} . A central quantity in spin glass theory is the overlap between two configurations \boldsymbol{\sigma}^{1},\boldsymbol{\sigma}^{2} chosen independently from the Gibbs measure G_{N,\beta} , defined as \displaystyle R(\boldsymbol{\sigma}^{1},\boldsymbol{\sigma}^{2}):=\frac{1}{N}\sum_{i\leqslant N}\sigma_{i}^{1}\sigma_{i}^{2} In the standard SK model with Ising spins, the analysis is significantly more involved due to the discrete nature of the configuration space \Sigma_{N}=\{-1,1\}^{N} . The control of fluctuations in this case relies heavily on the concentration of the overlap R(\boldsymbol{\sigma}^{1},\boldsymbol{\sigma}^{2}) , for which Talagrand [ 10 ] established crucial bounds that become increasingly difficult to control as \beta\uparrow 1 . Our work builds upon the variance representation via Gaussian interpolation introduced by Chatterjee [ 3 ] and employs Stein’s method to provide a refined CLT that remains valid as \beta_{N} approaches 1 at the scaling rate N^{-1/3} . 1.1. Main Results Our main objective is to characterize the fluctuation of the free energy near, but above the critical temperature \beta_{c}=1 . Theorem 1.1 . Assume that the inverse temperature sequence (\beta_{N})_{N\geqslant 1} satisfies \lim_{N\to\infty}\beta_{N}=1 and (1) \displaystyle\lim_{N\to\infty}N^{1/3}(1-\beta_{N}^{2})=c\in(0,\infty)\cup\{\infty\}. Define the function \displaystyle\nu(\beta):=-\frac{1}{2}\ln(1-\beta^{2})-\frac{\beta^{2}}{2} for \beta<1 . Then, the variance satisfies (2) \displaystyle\left|\var(F_{N}(\beta_{N}))-\nu(\beta_{N})\right|\leqslant L\left(N^{1/3}(1-\beta_{N}^{2})\right)^{-3/2}. Moreover, we have the following distributional convergence for the free energy (3) \displaystyle\dtv\left(\frac{F_{N}(\beta_{N})-\E F_{N}(\beta_{N})}{\sqrt{\nu(\beta_{N})}},g\right)\leqslant\frac{L}{\sqrt{\nu(\beta_{N})}}\left(N^{1/3}(1-\beta_{N}^{2})\right)^{-3/4}. Here, L is a finite constant depending on the sequence (\beta_{N})_{N\geqslant 1} , g\sim\N(0,1) and \dtv is the total variation distance. Corollary 1.2 . In particular, when \lim_{N\to\infty}N^{1/3}(1-\beta_{N}^{2})=c\in(0,\infty) , we have \var(F_{N}(\beta_{N}))=\frac{1}{6}\log N+\mathcal{O}(1) as N\to\infty . Remark 1 . In the physics literature, it is conjectured that the variance of the free energy at the critical temperature \beta_{c}=1 satisfies \displaystyle\var(F_{N}(\beta_{c}))=\frac{1}{6}\ln N+\mathcal{O}(1). To extend the methods developed in this paper to the critical case, one would require precise estimates of the overlap moments at \beta_{c} , similar to Lemma 1.4 . For instance, it is conjectured in [ 10 , Conjecture 11.7.5] that \displaystyle\lim_{N\to\infty}N^{2/3}\langle R^{2}(\boldsymbol{\sigma}^{1},\boldsymbol{\sigma}^{2})\rangle=a for some constant a>0 . Alternatively, one could attempt to compare the free energy at the scaling \beta_{N}=1-cN^{-1/3} with that at \beta_{c}=1 using a martingale approach. Define the process \displaystyle M_{N}(t):=\sum_{\boldsymbol{\sigma}\in\Sigma_{N}}\exp\left(\frac{1}{\sqrt{N}}\sum_{1\leqslant i<j\leqslant N}B_{ij}(t)\sigma_{i}\sigma_{j}-\frac{(N-1)t}{4}\right) where ((B_{ij}(t)))_{i<j} are independent standard Brownian motions. By Itô’s formula, we have \displaystyle\ln M_{N}(\beta_{c}^{2})-\ln M_{N}(\beta_{N}^{2})=\int_{\beta_{N}^{2}}^{\beta_{c}^{2}}\frac{1}{M_{N}(t)}dM_{N}(t)-\frac{1}{2}\int_{\beta_{N}^{2}}^{\beta_{c}^{2}}\frac{1}{M_{N}^{2}(t)}d\langle M_{N}\rangle_{t}. If one can show that the term \displaystyle\E\left[\int_{\beta_{N}^{2}}^{\beta_{c}^{2}}\frac{1}{M_{N}^{2}(t)}d\langle M_{N}\rangle_{t}\right]=\int_{\beta_{N}}^{\beta_{c}}\beta(N\E\langle R^{2}(\boldsymbol{\sigma}^{1},\boldsymbol{\sigma}^{2})\rangle-1)d\beta is uniformly bounded in N , a CLT at the critical temperature could potentially be established. The aforementioned conjecture on the overlap scaling would provide a path to such a proof. Remark 2 . Our proof approach remains valid in both the high-temperature regime and the near-critical window. Specifically, we extend the classical results of [ 1 ] and [ 5 ] to sequences of inverse temperatures (\beta_{N}) that approach the critical value \beta_{c}=1 at the rate of N^{-1/3} . By employing a sharper analysis, the variance bound can be improved further. Specifically, for any \epsilon>0 , it can be shown that \displaystyle\left|\var(F_{N}(\beta_{N}))-\nu(\beta_{N})\right|\leqslant L\left(N^{1/3}(1-\beta_{N}^{2})\right)^{-2+\epsilon}. For any real-valued function f on \Sigma_{N}^{n} , we define its Gibbs average under the Gibbs measure G_{N,\beta} as \displaystyle\langle f\rangle=\sum_{\boldsymbol{\sigma}^{1},\ldots,\boldsymbol{\sigma}^{n}\in\Sigma_{N}}f(\boldsymbol{\sigma}^{1},\ldots,\boldsymbol{\sigma}^{n})G_{N,\beta}(\boldsymbol{\sigma}^{1})\cdots G_{N,\beta}(\boldsymbol{\sigma}^{n}). The main technical tools used in our proof are the Gaussian interpolation method and Stein’s method for normal approximation. In the following two sections, we briefly discuss these now-classical tools. Throughout the rest of the article, we write H_{N} , Z_{N},G_{N} , and F_{N} instead of H_{N,\beta} , Z_{N}(\beta),G_{N,\beta} , and F_{N}(\beta) , unless we need to explicitly indicate the dependence on \beta . Without loss of generality, we will also assume that N^{1/3}(1-\beta_{N}^{2})\geqslant\varepsilon_{0} for all N\geqslant 1 and some \varepsilon_{0}>0 . 1.2. Coupled Hamiltonians

Read the complete text