Characteristic polynomials of sample covariance matrices: the non-square case (Q607429): Difference between revisions

The second-order correlation function of the characteristic polynomial of a sample covariance matrix is investigated. The complex sample covariance matrix \(Z(n,m)\) is defined as follows. Let \(Q\) be a distribution on the real line with expectation 0, variance 1/2, and finite fourth moment \(b\), and for given \(n, m \in {\mathbb N}\) with \(n \geq m\), let \(X:= X(n,m)\) denote the \(n \times m\) matrix whose entries are i.i.d. complex random variables whose real and imaginary parts are independent, each with the distribution \(Q\). Then the Hamiltonian \(m \times m\) matrix \(Z:= Z(n, m) = X(n, m) X^{*}(n, m)\) is called the (unrescaled) sample covariance matrix associated with the distribution \(Q\). The real sample covariance matrix is defined in a similar way with the difference that \(X(n, m)\) has real entries with the distribution \(Q\) whose variance is 1. This paper is a continuation of the author's work [J. Theor. Probab. (in press)], where the ``square'' case with \(n-m\) fixed was analyzed. Here, similar results are derived for the ``non-square'' case, where the difference \(n - m\) tends to infinity in such a way that the ratio \(m/n\) tends to some constant \(\gamma_{\infty} \in(0, 1)\) sufficiently quickly. It is then shown that the second-order correlation function of the characteristic polynomial of a complex sample covariance matrix is asymptotically given by the sine kernel in the bulk of the spectrum and by the Airy kernel at the edge of the spectrum. Similar results are valid for real sample covariance matrices.