[Paper Review] Global fluctuations for 1D log-gas dynamics

Jérémie Unterberger|arXiv (Cornell University)|Jul 4, 2016

Stochastic processes and financial applications21 references3 citations

TL;DR

This paper establishes the Gaussian fluctuation limit for one-dimensional log-gas dynamics under the mean-field hydrodynamic scaling, proving that fluctuations around the macroscopic limit satisfy a linear stochastic PDE. The key result is a rigorous derivation of the fluctuation process as a solution to a non-Markovian SPDE driven by a transport operator and a bounded non-local jump component, extending prior results on Dyson Brownian motion to general β-ensembles with convex potentials.

ABSTRACT

We study in this article the hydrodynamic limit in the macroscopic regime of the coupled system of stochastic differential equations, \begin{equation} dλ_t^i=\frac{1}{\sqrt{N}} dW_t^i - V'(λ_t^i) dt+ \fracβ{2N} \sum_{j ot=i} \frac{dt}{λ^i_t-λ^j_t}, \qquad i=1,\ldots,N, \end{equation} with $β>1$, sometimes called generalized Dyson's Brownian motion, describing the dissipative dynamics of a log-gas of $N$ equal charges with equilibrium measure corresponding to a $β$-ensemble, with sufficiently regular convex potential $V$. The limit $N o\infty$ is known to satisfy a mean-field Mac-Kean-Vlasov equation. We prove that, for suitable initial conditions, fluctuations around the limit are Gaussian and satisfy an explicit PDE. The proof is very much indebted to the harmonic potential case treated in Israelsson \cite{Isr}. Our key argument consists in showing that the time-evolution generator may be written in the form of a transport operator on the upper half-plane, plus a bounded non-local operator interpreted in terms of a signed jump process. As an essential technical argument ensuring the convergence of the above scheme, we give an $N$-independent large-deviation type estimate for the probability that $\sup_{t\in[0,T]}\max_{i=1,\ldots,N} |λ_t^i|$ is large, based on a multi-scale argument and entropic bounds.

Motivation & Objective

• To extend the hydrodynamic fluctuation theory of Dyson Brownian motion to general β-ensembles with convex potentials.
• To establish the Gaussian nature of fluctuations around the macroscopic limit in the N→∞ regime.
• To derive an explicit stochastic PDE governing the fluctuation process, generalizing the harmonic potential case.
• To analyze the generator of the fluctuation process using Stieltjes transform techniques and non-local operators.
• To prove convergence of the fluctuation process to a solution of a linear SPDE with transport and bounded remainder terms.

Proposed method

• Uses the Stieltjes transform to analyze the empirical measure and its time evolution in the hydrodynamic limit.
• Expresses the generator of the SDE system as a transport operator on the upper half-plane plus a bounded non-local operator.
• Applies Plemelj formulas and Fourier analysis to relate boundary values of holomorphic functions to singular integrals.
• Employs the generalized transport operator framework to decompose the generator into physical and correctional components.
• Uses large deviation bounds on the support of the measure to control singularities and ensure regularity.
• Relies on the harmonic potential case (Israelsson [14]) as a foundational reference for the generalization.

Experimental results

Research questions

• RQ1 Can the fluctuation process for general β-ensembles with convex potentials be shown to converge to a Gaussian process?
• RQ2 How does the time-evolution generator of the log-gas dynamics decompose in the hydrodynamic limit?
• RQ3 What is the structure of the limiting SPDE governing fluctuations, and how does it differ from the harmonic case?
• RQ4 Can the non-local remainder term in the generator be bounded and interpreted as a signed jump process?
• RQ5 To what extent is the fluctuation PDE universal across different convex potentials?

Key findings

• The fluctuation process converges to a Gaussian process in the N→∞ limit, with finite-dimensional distributions satisfying a linear SPDE.
• The limiting SPDE is driven by a transport operator on the upper half-plane, plus a bounded non-local operator interpreted as a signed jump process.
• The generator of the fluctuation process decomposes into a transport part and a remainder term that is bounded in operator norm.
• The Stieltjes transform of the limiting measure satisfies bounds ensuring regularity and control of singular integrals.
• The fluctuation PDE is universal up to scaling and depends on the potential only through the support of the equilibrium measure.
• Large deviation bounds on the support of the measure ensure that the fluctuation process remains well-behaved even near the boundary.