Singular values distribution of squares of elliptic random matrices and type B Narayana polynomials (Q2412520)

The authors consider Gaussian elliptic random \(N\times N\)-matrices \((X_{ij})\), where the bivariate vectors \((X_{ij}, X_{ji})\), \(1\leq i\leq j\), are mutually independent zero-mean Gaussian with \(\mathbb E X_{ij}^2= \mathbb E X_{ji}^2 = 1\) and \(\mathbb E X_{ij} X_{ji} = \rho\), \(i < j\). Here, \(\rho \in (-1,1)\) is some parameter. The main result states that the expected empirical distribution function of the singular values of the matrix \(\frac 1N X^2\) converges to a limit \(F\) as \(N\to\infty\). The limit distribution is uniquely determined by its moments for which the authors provide recurrence relations. Furthermore, the authors prove that if \(\xi\) is a random variable with distribution function \(F\), then the distribution of \(\varepsilon \sqrt{\xi}\) (where \(\varepsilon\) is independent of \(\xi\) and takes values \(\pm 1\) with probability \(1/2\) each) is uniquely determined by its free cumulants that are given by Narayana polynomials of type B: \(c_{2n} = \sum_{k=0}^n \binom nk^2 \rho^{2k}\). The main tool is the method of moments.