On the density of polynomials in some $L^2(M)$ spaces

Abstract

In this paper we study the density of polynomials in some $L^2(M)$ spaces. Two choices of the measure $M$ and polynomials are considered: 1) a $(N\times N)$ matrix non-negative Borel measure on $\mathbb{R}$ and vector-valued polynomials $p(x) = (p_0(x),p_1(x),...,p_{N-1}(x))$, $p_j(x)$ are complex polynomials, $N\in \mathbb{N}$; 2) a scalar non-negative Borel measure in a strip $\Pi = \{(x,\phi):\ x\in \mathbb{R}, \phi\in [-\pi,\pi) \} $, and power-trigonometric polynomials: $p(x,\phi) = \sum_{m=0}^\infty \sum_{n=-\infty}^\infty \alpha_{m,n} x^m e^{in\phi}$, $\alpha_{m,n}\in \mathbb{C}$, where all but finite number of $\alpha_{m,n}$ are zeros. We prove that polynomials are dense in $L^2(M)$ if and only if $M$ is a canonical solution of the corresponding moment problem. Using descriptions of canonical solutions, we get conditions for the density of polynomials in $L^2(M)$. For this purpose, we derive a model for commuting self-adjoint and unitary operators with a spectrum of a finite multiplicity.