Asymptotic distribution of smoothed eigenvalue density. I. Gaussian random matrices (Q4940238)

Random matrices of large dimension are now of considerable interest because of their applications in various fields of theoretical physics such as statistical nuclear physics, statistical mechanics of disordered systems, quantum chaos theory, quantum field theory, etc. In the spectral theory of random matrices the asymptotic behaviour of local spectral statistics can be regarded as the most interesting and challenging problem. In this paper the authors study the smoothed density of the eigenvalue counting function of \(N\times N\) real symmetric matrices \(H_{N}\) with independent entries, namely, the so-called Gaussian orthogonal ensemble of random matrices. The smoothing is determined by the resolvent \(G_{N}(z)=(H_{N}-z)^{-1}.\) It is stated that if the entries of \(H_{N}\) have joint Gaussian distribution, then the random variable \[ \xi_N^{\alpha}(\lambda)=N^{-1}ImTrG_{N}(\lambda+iN^{-\alpha}), \] where \(\alpha\in (0,1)\), converges as \(N\to +\infty\) to the Wigner semicircle density. The authors also show that \[ \gamma_{N}^{\alpha}(\lambda)=N^{1-\alpha}[\xi_{N}^{\alpha}(\lambda)-E\xi_{N}^{\alpha}(\lambda)] \] converges to a Gaussian random variables as \(N\to+\infty\), find explicitly the limit of the correlation function of \(\gamma_{N}^{\alpha}(\lambda)\), and discuss the relations between their results and the universality conjecture for large random matrices.