On the Matricial Truncated Moment Problem. II

Abstract

We continue the study of truncated matrix-valued moment problems begun in arXiv:2310.00957. Let $q\in\mathbb{N}$. Suppose that $(\mathcal{X},\mathfrak{X})$ is a measurable space and $\mathcal{E}$ is a finite-dimensional vector space of measurable mappings of $\mathscr{X}$ into $\mathcal{H}_q$, the Hermitian $q\times q$ matrices. A linear functional $\Lambda$ on $\mathcal{E}$ is called a moment functional if there exists a positive $\mathcal{H}_q$-valued measure $\mu$ on $(\mathcal{X},\mathfrak{X})$ such that $\Lambda(F)=\int_\mathcal{X} \langle F, \mathrm{d}\mu\rangle$ for $F\in \mathcal{E}$. In this paper a number of special topics on the truncated matricial moment problem are treated. We restate a result from (Mourrain and Schm\"udgen, 2016) to obtain a matricial version of the flat extension theorem. Assuming that $\mathcal{X}$ is a compact space and all elements of $ \mathcal{E}$ are continuous on $\mathcal{X}$ we characterize moment functionals in terms of positivity and obtain an ordered maximal mass representing measure for each moment functional. The set of masses of representing measures at a fixed point and some related sets are studied. The class of commutative matrix moment functionals is investigated. We generalize the apolar scalar product for homogeneous polynomials to the matrix case and apply this to the matricial truncated moment problem.