Kac-Rice inspired approach to non-Hermitian random matrices

TL;DR

This paper introduces a novel method based on the Kac-Rice formula to analyze the joint distribution of eigenvalues and eigenvectors in non-Hermitian random matrices, with applications to various ensembles and perturbations.

Contribution

The paper develops a new approach for analyzing eigenvalue-eigenvector joint distributions in non-Hermitian matrices, extending the Kac-Rice method to this context.

Findings

Derived joint probability densities for specific matrix ensembles.

Identified a new 'weak non-reality' scaling regime near the real axis.

Analyzed eigenvalue and eigenvector distributions under rank-one perturbations.

Abstract

We suggest a method of analyzing the joint probability density (JPD) ${\cal P}_N(z,{\bf v})$ of an eigenvalue $z$ and the associated right eigenvector ${\bf v}$ (normalized with ${\bf v}^*{\bf v}=1$) for non-Hermitian random matrices of a given size $N\times N$. The approach is essentially based on the Kac-Rice counting formula applied to the associated characteristic polynomial combined with a certain integral identity for the Dirac delta function of such a polynomial. To illustrate utility of the general method we derive ${\cal P}_N(z,{\bf v})$ in the two particular cases: (i) one-parameter family of matrices interpolating between complex Ginibre and real Ginibre ensembles and (ii) a complex Ginibre matrix additively perturbed by a general fixed matrix. In particular, in the former case we analyze the formation of an excess of eigenvalues in the vicinity of the real axis on approaching the real Ginibre limit, which eventually gives rise to the existence of a new scaling regime of "weak non-reality" as $N\to \infty$. In the second case we further analyze non-Hermitian Rosenzweig-Porter model which recently attracted considerable interest in physics literature. In addition, we provide new insights into eigenvalue and eigenvector distribution for a general rank one perturbation of complex Ginibre matrices of finite size $N$, and in the structure of an outlier emerging as $N\gg 1$. Finally we discuss a generalization of the proposed method which is expected to be suitable for analysis of JPD involving both left- and right eigenvectors.