A path integral approach to sparse non-Hermitian random matrices

TL;DR

This paper develops a path integral method to analyze sparse non-Hermitian random matrices, deriving modified spectral laws and extending classical results to sparse and non-Gaussian cases.

Contribution

It introduces a novel path integral framework for sparse non-Hermitian matrices, deriving generalized spectral laws and handling non-Gaussian statistics.

Findings

Derived modified elliptic and semi-circle laws for sparse matrices

Established a non-Hermitian Marchenko-Pastur law

Extended analysis to non-Gaussian ensembles

Abstract

The theory of large random matrices has proved an invaluable tool for the study of systems with disordered interactions in many quite disparate research areas. Widely applicable results, such as the celebrated elliptic law for dense random matrices, allow one to deduce the statistical properties of the interactions in a complex dynamical system that permit stability. However, such simple and universal results have so far proved difficult to come by in the case of sparse random matrices. Here, we perform an expansion in the inverse connectivity, and thus derive general modified versions of the classic elliptic and semi-circle laws, taking into account the sparse correction. This is accomplished using a dynamical approach, which maps the hermitized resolvent of a random matrix onto the response functions of a linear dynamical system. The response functions are then evaluated using a path integral formalism, enabling one to construct Feynman diagrams, which facilitate the perturbative analysis. Additionally, in order to demonstrate the broad utility of the path integral framework, we derive a generic non-Hermitian generalization of the Marchenko-Pastur law, and we also show how one can handle non-negligible higher-order statistics (i.e. non-Gaussian statistics) in dense ensembles.