Poisson Statistics for Coulomb Gases at Intermediate Temperature Regimes

David Padilla-Garza, Luke Peilen, Eric Thoma · 2025

Read Poisson Statistics for Coulomb Gases at Intermediate Temperature Regimes by David Padilla-Garza, Luke Peilen, Eric Thoma (2025) 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-2507-08198.txt.

Introduction

David Padilla-Garza, Luke Peilen, and Eric Thoma study the fine pattern of a Coulomb gas in a 2025 math paper. A Coulomb gas is a set of points that repel each other under a known law. The team looks at middle temperature ranges, between hot and cold extremes.

They show the fine point pattern settles toward a Poisson process, where points fall in a uniform random way. Their proof extends past results to wider ranges of heat and size. A sharp account of correlation functions drives the work and may interest others on its own. The result clarifies how order fades in such gases.

Catalog record

CREATORDavid Padilla-Garza, Luke Peilen, Eric Thoma
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE66 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-08198

Browse the full library · About the Tower of Babel

Full text

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

Poisson Statistics for Coulomb Gases at Intermediate Temperature Regimes David Padilla-Garza, Luke Peilen, Eric Thoma Abstract . We consider the microscopic statistics of a Coulomb gas in ℝ 2 \mathbb{R}^{2} at intermediate temperatures. In particular, we show that the microscopic point process associated to the Coulomb gas converges to a homogeneous Poisson point process at intermediate temperature regimes β ​ N → ∞ \beta N\rightarrow\infty and β ​ N ​ log ⁡ N → 0 \beta\sqrt{N}\log N\rightarrow 0 , extending previous results. Our approach relies on a novel quantitative asymptotic description of correlation functions, which is of its own interest. 1. Introduction We are interested in the Coulomb gas in \mathbb{R}^{\mathsf{d}} , an interacting particle system whose equilibrium statistics are governed by the Gibbs measure at inverse temperature \beta (1.1) \mathbb{P}_{N,\beta}(dX_{N})=\frac{1}{\mathsf{Z}_{N,\beta}}e^{-\beta\mathcal{H}_{N}(X_{N})}\penalty\ dX_{N} where X_{N}=(x_{1},\ldots,x_{N})\subset(\mathbb{R}^{\mathsf{d}})^{N} denotes the particle configuration and \mathcal{H}_{N} is an interaction Hamiltonian given by (1.2) \mathcal{H}_{N}(X_{N})=\frac{1}{2}\sum_{i\neq j}\mathsf{g}(x_{i}-x_{j})+N\sum_{i=1}^{N}V(x_{i}). V:\mathbb{R}^{\mathsf{d}}\rightarrow\mathbb{R} is an external confinement potential, assumed to grow rapidly enough at infinity so that the particles remain asymptotically confined to a compact subset of \mathbb{R}^{\mathsf{d}} with overwhelming probability. The kernel \mathsf{g} we consider is the Coulomb kernel (1.3) \mathsf{g}(x)=\begin{cases}-\log|x|&\text{if }\mathsf{d}=2,\\ |x|^{2-\mathsf{d}}&\text{if }\mathsf{d}\geq 3.\end{cases} There is also interest in studying more general Riesz gases, where the interaction power is allowed to vary between 0 and \mathsf{d} : (1.4) \mathsf{g}(x):=\mathsf{g}_{\mathsf{s}}(x)=\begin{cases}|x|^{-s}&\text{if }0<\mathsf{s}<\mathsf{d},\\ -\log|x|&\text{if }\mathsf{s}=0.\end{cases} The reason for taking \mathsf{s}<\mathsf{d} is to preserve local integrability of the kernel at the origin. In the Coulomb case \mathsf{g} satisfies -\Delta\mathsf{g}=\mathsf{c}_{\mathsf{d}}\delta_{0}, unlocking access to a trove of elliptic techniques. For the Riesz gases, one has a similar relationship with a fractional operator, namely (-\Delta)^{\frac{\mathsf{d}-\mathsf{s}}{2}}\mathsf{g}=\mathsf{c}_{\mathsf{d},\mathsf{s}}\delta_{0} in the sense of distributions. The superCoulombic Riesz gas, for \mathsf{d}-2<\mathsf{s}<\mathsf{d} is then particularly tractable. The fractional Laplacian with power 0<\alpha<1 has an interpretation as a (possibly degenerate) elliptic operator in \mathbb{R}^{\mathsf{d}+1} via the Caffarelli-Silvestre extension procedure (see [ CS07 ] ) When \mathsf{s}<\mathsf{d}-2 (the subCoulombic Riesz gas) the fractional operator is extremely nonlocal and the analysis of the gas is rather different. In this article, we will focus on the Coulomb gas, where the superharmonicity of the kernel \mathsf{g} plays a key role. We will work mainly in dimension \mathsf{d}=2 . We will also focus on the temperature regime \frac{1}{N}\ll\beta\ll 1 , which we call the intermediate temperature regime. 1.1. Splitting and the Thermal Equilibrium Measure From [ Fro35 ] (see also [ ST97 ] ), if one assumes that the potential V is everywhere finite, lower semicontinuous, and grows faster than the logarithm at infinity, i.e. (1.5) \lim_{|x|\rightarrow\infty}V(x)+\mathsf{g}(x)=+\infty then the mean-field energy (1.6) \mathcal{E}_{V}(\mu):=\mathcal{E}(\mu)+\int V(x)\penalty\ \mu(dx):=\frac{1}{2}\iint\mathsf{g}(x-y)\penalty\ \mu(dx)\mu(dy)+\int V(x)\penalty\ \mu(dx) has a compactly supported minimizer among probability measures known as the equilibrium measure , which we denote \mu_{V} . The equilibrium measure satisfies the Euler-Lagrange equation (1.7) \begin{cases}\mathsf{h}^{\mu_{V}}+V&\geq c_{V}\quad\text{ in }\mathbb{R}^{d}\\ \mathsf{h}^{\mu_{V}}+V&=c_{V}\quad\text{ in }\mathrm{supp}\left(\mu_{V}\right)\end{cases} for some constant c_{V} , where \mathsf{h}^{\nu} is the Coulomb potential (1.8) \mathsf{h}^{\nu}(x):=\int\mathsf{g}(x-y)\penalty\ \nu(dy). The equilibrium measure is a good approximation of the empirical measure in the sense that \mathsf{emp}_{N}:=\frac{1}{N}\sum_{i=1}^{N}\delta_{x_{i}}\rightarrow\mu_{V} in a large deviations sense so long as V is lower semicontinuous and \beta\gg\frac{1}{N} (see for instance [ GZ19 ] ). At temperatures \beta\simeq\frac{1}{N} the gas undergoes thermalization, and this equilibrium measure is no longer an adequate approximation. Entropic effects appear at leading order, and particle locations are no longer strongly confined to a compact set; instead, one needs to consider a thermal equilibrium measure \mu_{\theta} minimizing (1.9) \mathcal{E}_{\theta}(\mu):=\mathcal{E}_{V}(\mu)+\frac{1}{\theta}\int\mu\log\mu with \theta:=\beta N . This minimizer exists (see [ Lam21b ] ) and satisfies the Euler-Lagrange equation (1.10) \mathsf{h}^{\mu_{\theta}}+V+\frac{1}{\theta}\log\mu_{\theta}=c_{\theta} for some constant c_{\theta} . Furthermore, one has a law of large numbers (see [ GZ19 ] ) for the empirical measures \mathsf{emp}_{N}\rightarrow\mu_{\theta} in a large deviations sense when \beta N\to\theta as N\to\infty . Indeed, the authors in [ AS21 ] argue that \mu_{\theta} is a more precise approximation of \mu_{V} at all temperature regimes. Concentration of measure is studied in [ PG23 ] and quantiative approximation of \mu_{V} by \mu_{\theta} at temperatures \beta\gg\frac{1}{N} is established in [ AS22 ] via connections between the equilibrium problem and the classical obstacle problem (see [ Ser24 ] and [ Caf98 ] ). Given these deterministic first-order asymptotics, it is natural to “split” off the deterministic first-order limit when considering finer statistics of the gas. In particular, one has the so-called “splitting formula” (see [ Ser24 , Lemma 5.2] , introduced in [ SS12 ] ) (1.11) \mathcal{H}_{N}(X_{N})=N^{2}\mathcal{E}_{\theta}(\mu_{\theta})-\frac{N}{\theta}\sum_{i=1}^{N}\log\mu_{\theta}(x_{i})+N^{2}\mathsf{F}_{N}(X_{N},\mu_{\theta}), where \mathsf{F}_{N} is a next-order energy given by (1.12) \mathsf{F}_{N}(X_{N},\mu)=\frac{1}{2}\iint_{\Delta^{c}}\mathsf{g}(x-y)\left(\frac{1}{N}\sum_{i=1}^{N}\delta_{x_{i}}-\mu\right)^{\otimes 2}(dx,dy), which can be thought of as the energy of a “jellium” consisting of positive point charges and a neutralizing, negatively charged background measure \mu . The set \Delta=\{x=y\}\subset(\mathbb{R}^{\mathsf{d}})^{2} above denotes the diagonal, which is removed to neglect the formally infinite self-energy of a point charge. Using ( 1.11 ), one can equivalently write the Gibbs measure ( 1.1 ) as (1.13) \mathbb{P}_{N,\beta}(dX_{N})=\frac{1}{\mathsf{K}_{N,\beta}}e^{-\beta N^{2}\mathsf{F}_{N}(X_{N},\mu_{\theta})}\mu_{\theta}^{\otimes N}(dX_{N}) where \theta=\beta N , and \mathsf{K}_{N,\beta} is the next-order partition function (1.14) \mathsf{K}_{N,\beta}(X_{N}):=\int e^{-\beta N^{2}\mathsf{F}_{N}(X_{N},\mu_{\theta})}\mu_{\theta}^{\otimes N}(dX_{N}). We take ( 1.13 ) as our preferred representation of the Coulomb gas in the remainder of this paper. The next-order partition function is conveniently bounded below, which we state here for reference. Proposition 1.1 . ( [ PG23 , Prop. 5.10] ) Let \mathsf{K}_{N,\beta} be as in ( 1.14 ). Then \mathsf{K}_{N,\beta}\geq 1 . 1.2. Main results Our main goal in this paper is to better understand the microscopic behavior of the gas. A useful means of examining this behavior is by way of the microscopic point process near a point \overline{z}\in\mathbb{R}^{\mathsf{d}} , which is defined by (1.15) Q_{\overline{z},N}=\sum_{i=1}^{N}\delta_{N^{1/\mathsf{d}}(x_{i}-\overline{z})}. Effectively, Q_{\overline{z},N} centers the configuration around \overline{z} and then blows up the distance between particles so that the typical interparticle distance (previously \simeq N^{-1/\mathsf{d}} ) is now order one. We will study the convergence as N\to\infty of Q_{\overline{z},N} in the weak topology on point processes, Definition 3.1 , generated by integration against continuous, compactly supported functions on \mathbb{R}^{\mathsf{d}} . In general, the process ( 1.15 ) is hard to study due to the long-range nature of the Coulomb energy. The authors in [ LS15 ] studied the push-forward of Q_{\overline{z},N} under an averaging operator and identified a critical temperature scale, \beta\simeq N^{\frac{2}{\mathsf{d}}-1} (\beta\simeq N^{-\frac{\mathsf{s}}{\mathsf{d}}} for Riesz interactions with kernels \mathsf{g}_{\mathsf{s}}:=|x|^{-\mathsf{s}} ) at which the averaged point process converges as N\rightarrow\infty to a point process that minimizes a free energy with competing energy and entropic terms. For the \mathsf{d}=1 log-gas this process is the celebrated \mathrm{Sine}_{\beta} point process, whose convergence has been well-studied (cf. [ VV09 ] , [ KS09 ] ). For the Coulomb gas in \mathsf{d}=2 , this corresponds to the Ginibre point process at \beta=2 , but is otherwise not identified. At regimes \beta\gg N^{\frac{2}{\mathsf{d}}-1} one would correspondingly expect the energy term to dominate and see particles concentrate around energy minimizers, and as \beta\downarrow 0 faster than N^{\frac{2}{\mathsf{d}}-1} one would expect to see entropic behavior dominate. This corresponds to a kind of complete independence of the particles, and convergence to a Poisson point process (see \lx@sectionsign 3 for a more thorough discussion of this process). Indeed, this regime was studied in [ Leb16 ] , where the author established convergence of this averaged point process to a Poisson point process. In the high temperature regime \beta\simeq\frac{1}{N} where the strength of the interaction

Read the complete text