Eigenvectors of Sample Covariance Matrices: Universality of global fluctuations

Abstract: In this paper, we prove a universality result of convergence for a bivariate random process defined by the eigenvectors of a sample covariance matrix. Let $V_n=(v_{ij}){i \leq n,\, j\leq m}$ be a $n\times m$ random matrix, where $(n/m)\to y > 0$ as $ n \to \infty$, and let $X_n=(1/m) V_n V^{*}_n $ be the sample covariance matrix associated to $V_n :$. Consider the spectral decomposition of $X_n$ given by $ U_n D_n U_n^{*}$, where $U_n=(u{ij}){n\times n}$ is an eigenmatrix of $X_n$. We prove, under some moments conditions, that the bivariate random process $<B{s,t}^{n} = \underset{1\leq j \leq \lfloor nt \rfloor}{\sum_{1\leq i \leq \lfloor ns \rfloor}} <|u_{i,j}|^2 - \frac{1}{n}> >_{(s,t)\in[0,1]^2} $ converges in distribution to a bivariate Brownian bridge. This type of result has been already proved for Wishart matrices (LOE/LUE) and Wigner matrices. This supports the intuition that the eigenmatrix of a sample covariance matrix is in a way "asymptotically Haar distributed". Our analysis follows closely the one of Benaych-Georges for Wigner matrices, itself inspired by Silverstein works on the eigenvectors of sample covariance matrices.