On the Christoffel-Darboux kernel for random Hermitian matrices with external source (Q1045724)

Let \(A\) be a fixed Hermitian matrix and \(V(x)\) a function on \({\mathbb R}\) which increases sufficiently fast. Then, the ensemble of \(n \times n\) Hermitian matrices with the density function defined by \(p(M) = c_n \cdot \exp(- \text{Tr}(V(M) - AM))\), where \(c_n\) is a normalization constant, is considered. Due to the unitary-invariance of the density function it can be assumed without loss of generality that \(A\) is a diagonal matrix \(A = \text{diag}(a_1, a_2, \dots, a_n)\), where \(a_1 \geq a_2 \geq \dots \geq a_n\). When \(A = 0\), this ensemble is the standard unitary-invariant ensemble in the theory of random matrices. The case when \(A \not= 0\) is called a random matrix model with external source. When \(A = 0\), the correlation functions of eigenvalues are expressible in terms of the Christoffel-Darboux kernel of the orthogonal polynomials with respect to the weight \(e^{-V(x)}\). When \(A \not= 0\), it was shown by \textit{P. Bleher} and \textit{A. Kuijlaars} [Int. Math. Res. Not. 2004, No.~3, 109--129 (2004; Zbl 1082.15035)] that the correlation functions of the eigenvalues can be expressed in terms of a kernel involving a sum of multiple orthogonal polynomials. This summation was further simplified by \textit{E. Daems} and \textit{A. Kuijlaars} [J. Approximation Theory 130, No.~2, 190--202 (2004; Zbl 1063.42014)] yielding a Christoffel-Darboux formula for multiple orthogonal polynomials. Their expression of the Christoffel-Darboux formula is in this paper expressed newly in terms of the usual orthogonal polynomials. Some special cases of models of random Hermitian matrices with an external source when \(n\) tends to infinity are briefly discussed.