Orthogonal matrix polynomials: zeros and Blumenthal's theorem (Q1907515): Difference between revisions

In this paper the authors investigate orthogonal matrix polynomials \(P_n\) from their matrix three-term recurrence relation \[ tP_n(t)= D_{n+1} P_{n+1} (t)+E_n P_n(t)+ D^*_n P_{n-1}(t), \] where \(D_n\) and \(E_n\) are \(N\times N\) matrices. First, some results regarding the zeros of such matrix polynomials are given, then a quadrature formula is constructed. The relationship between the zeros and weights in this quadrature formula and eigenvalues and eigenvectors of block tridiagonal matrices is used extensively. The second part of the paper deals with Blumenthal's theorem for matrix polynomials. Here one assumes that the recurrence coefficients \(E_n\) and \(D_n\) converge to matrix limits \(E\) and \(D\). Assuming that these matrix limits are finite Toeplitz matrices, the authors show that the support of the matrix measure for the orthogonal matrix polynomials (i.e., the support of the trace measure) is a compact interval determined by the entries in the matrices \(D\) and \(E\), possibly together with real numbers outside this interval, with accumulation points at the endpoints of this interval. An example is given, showing that the Toeplitz character of the limits \(D\) and \(E\) is needed for the support to be an interval. An interesting open problem is to study Blumenthal's theorem when \(D\) and \(E\) are not Toeplitz.