Spectral convergence for a general class of random matrices (Q633054): Difference between revisions

Motivated by applications to array signal processing, the authors consider a random matrix model of the form \[ B=A+R^{1/2}XTX^HR^{1/2}, \] where: {\parindent6.5mm \begin{itemize}\item[(1)] \(X\) is an \(m\times n\) random matrix satisfying a finite-moment condition and such that its entries are independent identically distributed complex random variables with mean 0 and variance \(1/n\); \item[(2)] \(A\) is \(m\times m\) Hermitian non-negative definite; \item[(3)] \(R\) is \(m\times m\) Hermitian non-negative definite; \item[(4)] \(T\) is \(n\times n\) diagonal with real entries; \item[(5)] \(\Theta\) is \(m\times m\). \end{itemize}} The four matrices \(A\), \(R\), \(T\), \(\Theta\) are required to be uniformly (in \(n,m\)) bounded in the spectral norm. The dimensions \(m,n\) are required to vary in such a way that there exist constants \(0<c_1<c_2<1\) with \(c_1 n \leq m < c_2 n\). Then, for any \(z\in\mathbb{C}\setminus\mathbb{R}\), the authors prove that the asymptotic behaviour of \[ \text{Tr}[\Theta(B-z I_m)^{-1}] \] is almost surely independent of \(X\). The deterministic limit is established explicitly.