Localization and delocalization of eigenvectors for heavy-tailed random matrices (Q389281)

This paper deals with square symmetric \(n\times n\) matrices with independently identically distributed random entries above and on the diagonal. The authors consider an \(\alpha\)-stable symmetric distribution, and study spectral properties of such matrices as \(n\) goes to infinity. They give new bounds on the convergence rate of the empirical spectral distribution for \(0<\alpha<2\) (see Theorem 1.1). Further, they give vanishing bounds on the \(L^p\) norm of eigenvectors normalized in \(L^2\) norm for \(1<\alpha<2\) and \(p>2\) (see Theorem 1.2). Finally, they show that these eigenvectors are localized for \(0<\alpha<2/3\) (see Theorem 1.3). The paper is quite long and technical which can be, on the other hand, useful for further study of spectral properties of random matrices of the given type. All the most important results are summarized in its introductory part.