REM universality for linear random energy
Francesco Concetti, Simone Franchini · 2026
Read REM universality for linear random energy by Francesco Concetti, Simone Franchini (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-2604-06122.txt.
Introduction
Concetti and Franchini prove Random Energy Model universality for a linear random-energy model built from random Hamiltonians. When a huge set of configurations is sampled at random, the energy levels converge to a Poisson point process of exponential intensity. The result strengthens earlier local work by pinning down fluctuations of order one in the energy. It also improves a dilution-based result by allowing far more sampled configurations than before. The paper closes by deriving the long-run distribution of the Gibbs weight.
Catalog record
| CREATOR | Francesco Concetti, Simone Franchini |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87724 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.06122
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
REM universality for linear random energy Francesco Concetti, Simone Franchini Abstract . We consider a sequence of random Hamiltonians H n ( h , σ ) = ∑ i = 1 n h i ( σ i − m ) H_{n}(h,\sigma)=\sum^{n}_{i=1}h_{i}(\sigma_{i}-m) , and study the asymptotic ( n → ∞ n\to\infty ) distribution of the energy levels ( H n ( h , σ ) ) σ ∈ { − 1 , 1 } n \left(H_{n}(h,\sigma)\right)_{\sigma\in\{-1,1\}^{n}} , where h 1 , h 2 , ⋯ h_{1},h_{2},\cdots are i.i.d. random variables. We show that, when e O ( n ) e^{O(n)} configurations are sampled at random, the corresponding collection of energy levels converges in distribution to a Poisson point process with exponential intensity measure. This establishes the Random Energy Model (REM) universality for the present model. Our results strengthen earlier works on local REM universality by characterizing the distribution of O ( 1 ) − O(1)- order fluctuations of H n H_{n} . In addition, we improve upon the REM universality by dilution studied by Ben Arous, Gayrard, Kuptsov by allowing an exponentially large number e O ( n ) e^{O(n)} of sampled configurations, instead of e o ( n ) e^{o(\sqrt{n})} . Finally, we derive the asymptotic distribution of the Gibbs weight. MSC: 60G55,60F99, 82B44. 1. Introduction Let h:=(h_{i})_{i\in\mathbb{N}} be a sequence of independent and identically distributed (i.i.d.) real-valued random variables, and let \sigma:=(\sigma_{i})_{i\in\mathbb{N}} be a sequence of i.i.d. \{-1,1\}- valued random variables, independent of h . We denote by \mathbf{P}_{\sigma} the distribution of \sigma , characterized by (1.1) \mathbf{P}_{\sigma}(\sigma_{1}=1)=\frac{1+m}{2}, for some m\in(-1,1) . We denote by \mathbf{P}_{h} the joint distribution of h and denote by \mathbb{E}\left[\,\cdot\,\right] the expectation over \mathbf{P}_{h} . For any n\in\mathbb{N} , we define (1.2) H_{n}(h,\sigma):=\sum_{i=1}^{n}h_{i}(\sigma_{i}-m),\qquad\sigma\in\{-1,1\}^{\mathbb{N}}. The Hamiltonian H_{n} provides a simple example of a random Hamiltonian whose energy levels \bigl(H_{n}(h,\sigma)\bigr)_{\sigma\in\{-1,1\}^{\mathbb{N}}} are correlated random variables. Models of this type arise naturally in the statistical mechanics of disordered systems, notably in spin glasses, as well as in combinatorial optimization problems. In particular, H_{n} is closely related to the number partitioning problem [ 20 ] . It has long been conjectured that, for a broad class of random Hamiltonians, the properly rescaled energy levels converge in distribution to a Poisson point process (PPP). [ 20 , 4 , 3 ] . Consequently, the asymptotic statistics of the energy levels coincide with those of Derrida’s Random Energy Model (REM): a spin-glass model in which the energy levels are independent by construction [ 14 ] . This conjecture is commonly referred to as REM universality . In the original REM, the energies are Gaussian random variables. It was later shown that the convergence of rescaled energy levels to a PPP holds for a broader class of models with independent energies drawn from more general distributions [ 8 ] . In a series of works, Borgs, Chayes, Mertens, and Nair for the partitioning [ 6 , 7 ] , and Bovier and Kourkova for more general spin glass Hamiltonians [ 9 ] , proved that the fluctuations of the energy levels converges to a PPP, when observed in a small window of the spectrum whose width shrinks exponentially fast with the system size n . They called this local property of the energy spectrum local REM universality . A complementary perspective was later introduced by Ben Arous, Gayrard, and Kuptsov [ 5 ] , who established REM universality by dilution . Specifically, they proved that REM universality persists for energy levels arising from random subsets of configurations whose cardinality is sub-exponential ( \sim e^{o(\sqrt{n})} ). The present work substantially extends these results. For the Hamiltonian H_{n} , we establish REM universality for energy fluctuations of order 1 and for families of configurations whose cardinality grows exponentially with the system size ( \sim e^{\alpha n} ). In particular, this proves REM universality for an extensive portion of the energy levels. Recently, REM universality has attracted renewed interest in physics literature, particularly in connection with advances in mean-field spin glass theory, where new methods that exploit REM-like behavior are proposed [ 15 , 16 , 17 ] . The results of this manuscript are based on the following assumption. Assumption 1.1 . The distribution of h_{1} has an absolutely continuous part, there exists \varepsilon>0 and a constant p_{1}>0 such that \mathbf{P}_{h}(|h_{1}|<t)\leq p_{1}t for all t\in[0,\varepsilon) , and there exists an interval [c,d] such that the density of h_{1} on [c,d] is bounded from below by a constant p_{2}>0 . Moreover, the first, second, and third moments exist with (1.3) \mathbb{E}[h_{1}]=\psi_{1},\quad\mathbb{E}[h_{1}^{2}]=\psi_{2},\quad\mathbb{E}[|h_{1}|]=\psi_{3},\quad\mathbb{E}[|h_{1}|^{3}]=\psi_{4}. Given \lambda>0 , we define the locally finite measure \mathbf{D}_{\lambda} on \mathbb{R} by (1.4) \mathbf{D}_{\lambda}(\mathfrak{U}):=\int_{\mathfrak{U}}e^{-\lambda x}dx, for any Borel set \mathfrak{U} in the Borel \sigma- algebra \mathcal{B}(\mathbb{R}) . The first result of the manuscript is the following. Define the functions (1.5) G(\lambda):=\mathbb{E}\left[\log((1+m)e^{\lambda(1-m)h_{1}}+(1-m)e^{-\lambda(1+m)h_{1}})\right]-\log(2),\qquad G^{*}(a):=\sup_{\lambda\in\mathbb{R}}\left(\lambda a-G(\lambda)\right), and denote by G^{\prime} the first derivative of G . We also define the quantities (1.6) \varsigma:=-m\psi_{1}+\psi_{3},\qquad\gamma:=-\mathbb{E}[\log(1+\sign(h_{1})m)]+\log(2),\quad\Gamma_{n}(h):=-\sum^{n}_{i=1}\log\left(1+\sign(h_{i})m\right)+n\log(2) Our first result establishes a universal asymptotic behavior of the distribution of H_{n} , conditionally on h . Theorem 1.2 . Assume that Assumption 1.1 holds. Given a deterministic number c\in(0,\gamma) , there exists a unique \tilde{a}\in(0,\varsigma) and \tilde{\lambda}\in\mathbb{R}_{>0} (depending on c ) such that (1.7) G^{*}(\tilde{a})=c,\quad G^{\prime}(\tilde{\lambda})=\tilde{a}. Moreover, for any random sequence (C_{n})_{n\in\mathbb{N}} such that (1.8) C_{n}(h)\in(0,\Gamma_{n}(h)),\quad\mathbf{P}_{h}-\text{e.a.s.},\qquad\textup{and}\qquad\lim_{n\to\infty}\tfrac{1}{n}C_{n}=c,\quad\mathbf{P}_{h}-\text{a.s.}, there exists a h- measurable random sequence (A_{n})_{n\in\mathbb{N}} such that the sequence of measure kernels \left(\mathbf{K}_{n}\right)_{n\in\mathbb{N}} , defined by (1.9) \mathbf{K}_{n}(h,\mathfrak{U}):=e^{C_{n}(h)}\mathbf{P}_{\sigma}(\{\sigma:\,H_{n}(h,\sigma)-A_{n}(h)\in\mathfrak{U}\}),\quad h\in\mathbb{R}^{\mathbb{N}},\,\mathfrak{U}\in\mathcal{B}(\mathbb{R}), converges vaguely \mathbf{P}_{h}-\text{almost surely} to the deterministic measure \mathbf{D}_{\tilde{\lambda}} . Remark . The main novelty of Theorem 1.2 , with respect to [ 5 ] , is that the centering sequence (A_{n})_{n\in\mathbb{N}} is allowed to depend on the environment h . This random centering is needed to obtain the convergence of the kernels \mathbf{K}_{n}(h,\cdot) . Given n\in\mathbb{N} and \sigma\in\{-1,1\}^{\mathbb{N}} , let \sigma_{[n]} denote its projection onto \{-1,1\}^{n} , namely (1.10) \sigma_{[n]}:=(\sigma_{1},\ldots,\sigma_{n}). For \tau\in\{-1,1\}^{n} , we define the associated n -dimensional cylinder by (1.11) [\tau]:=\{\sigma\in\{-1,1\}^{\mathbb{N}}:\sigma_{[n]}=\tau\}. For fixed h\in\mathbb{R}^{\mathbb{N}} , the map \sigma\mapsto H_{n}(h,\sigma) depends only on \sigma_{[n]} and is therefore constant on each n -dimensional cylinder. Fix \rho\in(0,1) . For each fixed n\in\mathbb{N} , let \mathbf{Q}^{(n\rho)}_{\sigma} denote the finite measure on the \sigma -algebra on \{-1,1\}^{\mathbb{N}} generated by the n -dimensional cylinders, defined by (1.12) \mathbf{Q}^{(n\rho)}_{\sigma}([\tau]):=e^{n\rho(\log 2-\log(1+|m|))}\mathbf{P}_{\sigma}([\tau]),\qquad\tau\in\{-1,1\}^{n}. Throughout the paper, we use the symbol \sigma to denote an entire infinite configuration in \{-1,1\}^{\mathbb{N}} and work with measures defined on this space. Let \Omega_{n}\subset\{-1,1\}^{\mathbb{N}} be a set containing exactly one representative from each n -dimensional cylinder. Equivalently, \Omega_{n} consists of configurations for which the coordinates \sigma_{n+1},\sigma_{n+2},\dots are fixed, so that only the first n spins vary. Whenever only the first n coordinates are relevant–such as in the definition of the point process below–summation is taken over configurations in \Omega_{n} . Let (U_{\sigma})_{\sigma\in\{-1,1\}^{\mathbb{N}}} be a family of independent random variables, uniformly distributed on [0,1] , and independent of both h and \sigma . We now state the main theorem of the paper, establishing REM universality for H_{n} . We say that a random variable is h- measurable if it is measurable with respect to the \sigma- algebra generated by h , Theorem 1.3 (REM universality) . Given \rho\in(0,1) , and let \tilde{a} and \tilde{\lambda} satisfy (1.13) G^{*}(\tilde{a})=\rho\left(\log(2)-\log(1+|m|)\,\right),\qquad G^{\prime}(\tilde{\lambda})=\tilde{a}.