Strong convergence of the empirical distribution of eigenvalues of sample covariance matrices with a perturbation matrix (Q968484)

Let \(\{X_{ij}^{(n)}\), \(i=1,\dots,n\), \(j=1,\dots,p\}\) be independent real random variables with common mean \(\mu\) and \({\text{Var}(X_{11}^{(n)})} = 1\). Let \({\mathbf X}_n = \frac {1}{\sqrt{n}} (X_{ij}^{(n)})_{n \times p}\) and \({\mathbf B}_n = {\mathbf A}_n + {\mathbf X}_n {\mathbf T}_n {\mathbf X}_n^T\), where \({\mathbf A}_n\) and \({\mathbf T}_n\) are both random symmetric matrices and \({\mathbf X}_n\), \({\mathbf A}_n\) and \({\mathbf T}_n\) are independent. Then, it is proved, using the Stieltjes transform of the limiting spectral distribution of the empirical spectral distribution (ESD) of the matrix \({\mathbf X}_n {\mathbf T}_n {\mathbf X}_n^T\), that the ESD of the eigenvalues of random symmetric matrices \({\mathbf B}_n\) converges almost surely to a non-random distribution.