Matrix Christoffel functions (Q1885372)

This paper extends to a matrix setting many classical results from the scalar theory of orthogonal polynomials and Christoffel functions. Namely, the authors consider an \(N\times N\) matrix weight \(W\) (that is, a positive definite matrix of measures supported on \(\mathbb R\)), for which they define the sequence of matrix-valued orthonormal polynomials \(P_n\) and the Christoffel function \(\lambda_{n,X}(z)\), solution of an \(L^2\) norm minimization problem in the set of polynomials of degree \(n\) normalized to the identity matrix at \(z\), and whose leading coefficient belongs to the left ideal \(X\). In the scalar case the Christoffel function can be expressed in terms of the orthogonal polynomials, as well as has a connection with another extremal problem (mass maximization at a point among all the measures with identical first moments). The authors prove the analogues of these results for the matrix weight \(W\). This is not a mere transcription of the scalar situation, multi-dimensionality does play a relevant role, introducing extra parameters (for instance, the leading coefficient of a matrix polynomial can be non-zero but singular). Last section of the paper is devoted to relative asymptotics of the Christoffel function. The matrix analogue of the well-known result of Nevai is the existence of limit of the expression of the form \[ (P_n^*)^{-1} (\lambda _{n, X_n})^{-1} P_n^{-1}, \] where \(X_n\) is a certain left ideal associated with \(P_n\). The authors establish convergence in the matrix Nevai class and describe the limit as a solution of a certain matrix non-linear equation. Under additional assumptions on the class they show that this limit is in fact the Cauchy transform of a Chebyshev weight of the second kind.