Random matrices: localization of the eigenvalues and the necessity of four moments (Q651861): Difference between revisions

This paper shows that if the entries have vanishing third moment, then for all \(1\leq i\leq n\) \[ E|\lambda_{i}(M_{n})-\sqrt{n}\gamma_{i}|^{2}=O({\min}(n^{-c} {\min}(i,n+1-i)^{-\frac{2}{3}}n^{\frac{2}{3}},n^{\frac{1}{3+\varepsilon }})), \] for some absolute constant \(c>0\) and any absolute constant \(\varepsilon>0\), where \(\lambda_{i}(M_{n})\) (in increasing order) are eigenvalues of a random Hermitian matrix \(M_{n}\), whose upper-triangular entries are independent with mean zero and variance one, and are exponentially decaying. In particular, for the eigenvalues in the bulk (\(\min (i,n-i)=\Theta(n)\)), \[ E|\lambda_{i}(M_{n})-\sqrt{n}\gamma_{i}|^{2}=O(n^{-c}). \] A similar result is achieved for the rate of convergence. The authors also show that the four moment condition in the four moment theorem is necessary, in the sense that if one allows the fourth moment to change (while keeping the first three moments fixed), then the mean of \(\lambda_{i}(M_{n})\) changes by an amount comparable to \(n^{-\frac{1}{2}}\) on the average. The authors make a precise conjecture about how the expectation of the eigenvalues varies with the fourth moment.