Dropping the independence: singular values for products of two coupled random matrices (Q2630731): Difference between revisions

It is very well known [the first author and \textit{Z. Burda}, J. Phys. A, Math. Theor. 45, No. 46, Article ID 465201, 18 p. (2012; Zbl 1261.15041)] that the eigenvalues of products of two independent coupled rectangular complex Gaussian matrices with i.i.d. standard complex entries constitute a determinantal point process in \(\mathbb C\). The behavior of singular values for such products was analyzed in [the first author et al., ``Products of rectangular random matrices: singular values and progressive scattering'', Phys. Rev. E 88, No. 5, Article ID 052118, 13 p. (2013; \url{doi:10.1103/PhysRevE.88.052118})]. Indeed, the squared singular values also constitute a determinantal process in the positive real semiaxis. The correlation kernels of the above determinantal processes can be written in terms of Meijer \(G\)-functions, with suitable choices of the parameters. In the paper under review, the authors deal with similar problems for parameter-dependent Gaussian complex matrices. This concept is defined from parameter-dependent coupled Gaussian random variables. If \(X_{1}\) and \(X_{2}\) are \(\mu\)-dependent Gaussian complex matrices of size \(N\times M\) and \(M\times N,\) respectively, and assuming \(M\geq N\) the joint probability density function for the squared singular values of \(Y=X_{1}X_{2}\) is given in terms of the product of two determinants involving the modified Bessel functions of the first and second kind, respectively. The matrix \(Y\) represents an interpolating biorthogonal ensemble in the sense that it interpolates between the ensemble of two independent complex Gaussian matrices and the Laguerre ensemble of a single complex Gaussian matrix. The exact formula for the correlation kernel \(K_{N}(x,y)\) is given as well as a Christoffel-Darboux (CD) formula. This last result is a direct consequence of the fact that the above biorthogonal polynomials satisfy a five term recurrence relation. A contour integral representation for those polynomials in terms of hypergeometric \(_{0}F_{1}\) and \(_{1}F_{2}\) functions is obtained. Using the CD formula and the above contour integral representations, the scaling limit, when \(N\) tends to infinity, of \(K_{N}(x,y)\) near the origin (hard edge) is given in terms of a Meijer \(G\)-kernel involving two Meijer \(G\)-functions with a suitable choice of parameters. On the other hand, from the five term recurrence relation, a central limit theorem for the linear statistics of singular values of \(Y\) is derived and the expression of the limiting variance is given. For the proof, the authors use the results by \textit{J. Breuer} and \textit{M. Duits} [J. Am. Math. Soc. 30, No. 1, 27--66 (2017; Zbl 1351.15022)] based on the asymptotic behavior of the recurrence coefficients of those biorthogonal polynomials.