Convergence rate of eigenvector empirical spectral distribution of large Wigner matrices (Q2423201)

Let $\mathbf{W}_n=\frac{1}{\sqrt{n}}\left(X_{i,j}\right)_{i,j=1}^n$ be an $n\times n$ Hermitian random matrix whose diagonal entries are (suitably normalized) IID real-valued random variables, and whose entries above the diagonal are (suitably normalized) IID complex-valued random variables satisfying some boundedness condition on their moments; the main results of the paper assume either bounded eighth or tenth moments. Write the spectral decomposition of $\mathbf{W}_n$ as $\mathbf{W}_n=\mathbf{U}_n\mathbf{\Lambda}_n\mathbf{U}_n^*$, where $\mathbf{\Lambda}_n=\text{diag}(\lambda_1,\ldots,\lambda_n)$ is a diagonal matrix containing the eigenvalues of $\mathbf{W}_n$ arranged in decreasing order. For an arbitrary unit vector $\mathbf{x}_n$, let $\mathbf{y}_n=(y_1,\ldots,y_n)^\prime=\mathbf{U}_n^*\mathbf{x}_n$. Then, the eigenvector empirical spectral distribution of $\mathbf{W}_n$ is defined by \[ H^{\mathbf{W}_n}(x)=\sum_{i=1}^n|y_i|^2I(\lambda_i\leq x). \] The main results of the paper give explicit rates of convergence of $H^{\mathbf{W}_n}$ to the semicircular law. The authors show that the expected eigenvector empirical spectral distribution converges to the semicircular law with rate of convergence $O(n^{-1/2})$. They also show that convergence rates of the eigenvector empirical spectral distribution to the semicircular law in probability and almost surely are $O(n^{-1/4})$ and $O(n^{-1/6})$, respectively. A simulation study is also used to investigate these convergence results.