On the Spontaneous Breaking of U(N) symmetry in invariant Matrix Models

TL;DR

This paper demonstrates that certain invariant matrix models can spontaneously break their rotational symmetry, leading to eigenvector localization and deviations from universal eigenvalue statistics, thus connecting symmetry breaking with localization phenomena.

Contribution

It reveals a novel spontaneous symmetry breaking mechanism in invariant matrix models, linking eigenvector localization with deviations from universal eigenvalue distributions.

Findings

Matrix models can spontaneously break U(N) symmetry.

Eigenvector localization occurs alongside eigenvalue distribution deviations.

The work introduces a new perspective on symmetry breaking and localization in matrix models.

Abstract

Matrix Models are the most effective way to describe strongly interacting systems with many degrees of freedom. They have proven successful in describing very different settings, from nuclei spectra to conduction in mesoscopic systems, from holographic models to various aspects of mathematical physics. This success reflects the existence of a large universality class for all these systems, signaled by the Wigner-Dyson statistics for the matrix eigenvalues. These models are defined in a base invariant way and this rotational symmetry has traditionally been read to imply that they describe extended system. In this work we show that certain matrix models, which show deviations from the Wigner-Dyson distribution, can spontaneously break their rotational ($U(N)$) invariance and localize their eigenvectors on a portion of the Hilbert space. This conclusion establishes once more a direct connection between the eigenvalue and eigenvector distributions. Recognizing this loss of ergodicity discloses the power of non-perturbative techniques available for matrix models to the study of localization problems and introduces a novel spontaneous symmetry breaking mechanism. Moreover, it brings forth the overlooked role of eigenvectors in the study of matrix models and allows for the consideration of new types of observable.