Orthogonal polynomials and analytic functions associated to positive definite matrices (Q819025): Difference between revisions

A semi-infinite positive definite matrix \(A=(a_ij)_{i,j\geq 0}\) defines naturally an inner product \(\langle \cdot, \cdot \rangle_A\) in the space of finite sequences, \[ \langle \alpha , \beta \rangle_A= (\alpha _0, \alpha _1, \dots ) \begin{pmatrix} a_{0,0} & a_{0,1} & \dots \\ a_{1,0} & a_{1,1} & \dots \\ \vdots & \vdots & \ddots \end{pmatrix} \begin{pmatrix} \overline{\beta_{0}} \\ \overline{\beta _{1}} \\ \vdots \end{pmatrix}\,, \] as well as the \(A\)-norm, \(\| \alpha \| _A^2 = \langle \alpha , \alpha \rangle_A\). If we apply the Gram-Schmidt process to the standard basis \(e_i\), \(i\geq 0\), then we obtain the orthonormal basis (with respect to the \(A\)-inner product) \(p_i\), \(i\geq 0\). The operator \(T(e_i)=p_i\) can be extended linearly to the whole space of finite sequences. If we denote by \(A_n\) the \(n \times n\) principal minor of \(A\), a similar construction can be carried out with \(A_n\). A natural question is the connection between the finite and infinite cases. For instance, if \(\lambda _{1, n}\) denotes the smallest eigenvalue of \(A_n\), then as \(n\to \infty\) it tends to \(\| T\| ^{-2}\) (the norm induced by the \(A\)-inner product), see Theorem 1.1. These problems have an alternative natural reading from the point of view of orthogonal polynomials if we associate each element of the standard basis \(e_i\) with the monomial \(t^i\). Then vectors \(p_i\) are just coefficients of the orthogonal polynomials (with the \(A\)-inner product) with respect to the basis \(1, t, t^2, \dots\). In particular, a consequence of Theorem 1.1 is that if the sequence \(\lambda _{1, n}\) is uniformly bounded away from zero, the sequence of polynomials \(\{ p_n(z)\} \in \ell^2\) for suitable values of the argument \(z\). This has obvious connections with the determinacy of the moment problem associated with these polynomials. In particular, if \(A\) is a Hankel matrix, the authors characterize the finite index of indeterminacy in terms of the boundedness of the operator \(T\) in certain subspaces. Another consequence for the block Hankel matrices \(A\) is a characterization of complete indeterminacy for matrix weights in terms of the behavior of \(\{ \lambda _{1, n}\} \) as \(n\to \infty\).