The matrix Bochner problem
W. Riley Casper, Milen Yakimov
Abstract
A long standing question in the theory of orthogonal matrix polynomials is the matrix Bochner problem, the classification of $N\times N$ weight matrices $W(x)$ whose associated orthogonal polynomials are eigenfunctions of a second order differential operator. Based on techniques from noncommutative algebra (semiprime PI algebras of Gelfand-Kirillov dimension one), we construct a framework for the systematic study of the structure of the algebra ${\cal D}(W)$ of matrix differential operators for which the orthogonal polynomials of the weight matrix $W(x)$ are eigenfunctions. The ingredients for this algebraic setting are derived from the analytic properties of the orthogonal matrix polynomials. We use the representation theory of the algebras ${\cal D}(W)$ to resolve the matrix Bochner problem under the two natural assumptions that the sum of the sizes of the matrix algebras in the central localization of ${\cal D}(W)$ equals $N$ (fullness of ${\cal D}(W)$) and the leading coefficient of the second order differential operator multiplied by the weight $W(x)$ is positive definite. In the case of $2\times 2$ weights, it is proved that fullness is satisfied as long as ${\cal D}(W)$ is noncommutative. The two conditions are natural in that without them the problem is equivalent to much more general ones by artificially increasing the size of the matrix $W(x)$.