Universality of Wigner random matrices: a Survey of Recent Results
Laszlo Erdos
TL;DR
This survey demonstrates that the local spectral statistics of large Wigner-type matrices are universal, matching those of classical Gaussian ensembles in the bulk and at the edge under minimal assumptions. The authors develop a three-step program: (1) establish a strong local semicircle law to control eigenvalue densities and eigenvectors at microscopic scales, (2) prove bulk universality for Gaussian-divisible ensembles via a local relaxation flow tied to Dyson Brownian motion, and (3) remove the Gaussian component using Green function comparison and four-moment matching. The combination of precise resolvent estimates, flow-based ergodicity, and moment-matching yields universality for general Wigner matrices with subexponential tails, and edge universality follows from refined local laws. The work provides a unifying framework that explains universality as a manifestation of rapid local equilibration in the Dyson Brownian motion and offers robust tools for extending universality to broader matrix ensembles and models.
Abstract
We study the universality of spectral statistics of large random matrices. We consider $N\times N$ symmetric, hermitian or quaternion self-dual random matrices with independent, identically distributed entries (Wigner matrices) where the probability distribution for each matrix element is given by a measure $ν$ with a subexponential decay. Our main result is that the correlation functions of the local eigenvalue statistics in the bulk of the spectrum coincide with those of the Gaussian Orthogonal Ensemble (GOE), the Gaussian Unitary Ensemble (GUE) and the Gaussian Symplectic Ensemble (GSE), respectively, in the limit $N\to \infty$. Our approach is based on the study of the Dyson Brownian motion via a related new dynamics, the local relaxation flow. As a main input, we establish that the density of eigenvalues converges to the Wigner semicircle law and this holds even down to the smallest possible scale, and, moreover, we show that eigenvectors are fully delocalized. These results hold even without the condition that the matrix elements are identically distributed, only independence is used. In fact, we give strong estimates on the matrix elements of the Green function as well that imply that the local statistics of any two ensembles in the bulk are identical if the first four moments of the matrix elements match. Universality at the spectral edges requires matching only two moments. We also prove a Wegner type estimate and that the eigenvalues repel each other on arbitrarily small scales.