Gaussian fluctuations for non-Hermitian random matrix ensembles (Q874731): Difference between revisions

An ensemble of \(N \times N\) non-Hermitian matrices in which all entries are independent identically distributed complex random variables of mean zero and mean-square one is considered. \textit{Z. D. Bai} [Ann. Probab.~25, 494--529 (1997; Zbl 0871.62018)] has proved so far the most general result that the circular law holds for ensembles of independent entries such that the common entry distribution possesses a bounded density and finite \(({\text 4}+ \epsilon)\) moment. In this paper the fluctuations from the circular law are studied in the class of non-Hermitian matrices for which higher moments of the entries obey a growth condition. Attention is paid to linear spectral statistics of the form \(X_N(f) = \sum_{k=1}^N f(\lambda_k)\), where \(\lambda_1, \lambda_2, \dots, \lambda_N\) are ensemble eigenvalues. \(X_N(f)\) is given by \(X_N(f) = N \int f(\lambda) \mu_N (d \lambda)\), where \(\mu_N (d \lambda)\) is the empirical spectral distribution and the test test function \(f\) is analytic in a neighborhood of the disk \(|z| \leq {\text 4}\). The central limit theorem for such an ensemble is established. The proof is based on the work of \textit{Z. D. Bai} and \textit{J. W. Silverstein} [Ann. Probab.~32, 553--605 (2004; Zbl 1063.60022)] and the used method is fairly robust in terms of the underlying distribution of the matrix entries but does not require the analyticity of the test function \(f\) in \(X_N(f)\).