Empirical spectral distribution of a matrix under perturbation (Q2312772)

This paper is devoted to the spectral study of a large \(n \times n\) Hermitian matrix with a small additive random perturbation. The empirical spectral distribution of the original matrix is assumed to converge to a limiting measure with density and the entries of the small perturbation Hermitian matrix are independent in the eigenvector basis of the first matrix with asymptotic variance depending on the location on or out of the diagonal. If the order of magnitude of the perturbation is less than \(n^{-1}\), a perturbation regime will be in force with the appearance of a centered Gaussian field in the leading term of the expansion of the empirical spectral distribution. If the order of magnitude lies between \(n^{-1}\) and 1, a semi-perturbative regime will appear with the introduction of a non-random function determined by the asymptotic parameters. Some precise details on the expansion of the empirical spectral distribution are given when the magnitude is of order \(n^{-1}\) or between \(n^{-1}\) and \(n^{-1/3}\).