We consider the product of n complex non-Hermitian, independent random matrices, each of size N × N with independent identically distributed Gaussian entries (Ginibre matrices). The joint probability distribution of the complex eigenvalues of the product matrix is found to be given by a determinantal point process as in the case of a single Ginibre matrix, but with a more complicated weight given by a Meijer G-function depending on n. Using the method of orthogonal polynomials we compute all eigenvalue density correlation functions exactly for finite N and fixed n. They are given by the determinant of the corresponding kernel which we construct explicitly. In the large-N limit at fixed n we first determine the microscopic correlation functions in the bulk and at the edge of the spectrum. After unfolding they are identical to that of the Ginibre ensemble with n = 1 and thus universal. In contrast the microscopic correlations we find at the origin differ for each n > 1 and generalize the known Bessel law in the complex plane for n = 2 to a new hypergeometric kernel 0Fn - 1.
Journal of Physics A Mathematical General
- Pub Date:
- November 2012
- Mathematical Physics;
- Condensed Matter - Statistical Mechanics
- 20 pages, v2 published version: typos corrected and references added