The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices (Q531808): Difference between revisions

Let \(X_n\) be an \(n\times n\) Hermitian or symmetric random matrix. Let \(P_n\) be an \(n\times n\) Hermitian or symmetric matrix of rank \(r\). The authors study the behaviour of the eigenvalues and eigenvectors of perturbations of \(X_n\) by \(P_n\), namely \(X_n+P_n\), \(X_n(I_n+P_n)\), \((I_n+P_n)^{1/2}X_n(I_n+P_n)^{1/2}\). Almost sure convergence of the extreme eigenvalues and of the projections of the corresponding eigenvectors on the eigenspaces of \(P_n\) are proven. The limiting eigenvalue is shown to depend explicitly on the limiting eigenvalue distribution of \(X_n\). A threshold is found where the limit as \(n\to\infty\) of the extreme eigenvalues of the perturbed matrix differ from those of \(X_n\) if and only if the eigenvalues of \(P_n\) are above that threshold. An analogous phase transition is found for the eigenvectors.