On maximal weight solutions in a truncated trigonometric matrix moment problem (Q617791): Difference between revisions

Let \({\mathbf C}=\{C_0,C_1,\dots,C_n\}\in \mathbb C^{q\times q}\), and let \(\mathcal M({\mathbf C})\) stand for the set of all non-negative Hermitian \(q\times q\) Borel measures \(F\) on the circumference \(\mathbb T\subset \mathbb C\) such that \(C_l=\int_{\mathbb T}z^{-l}F(dz)\) for any \(l=0,\dots,n\). For fixed \(u\in\mathbb T\), consider the family of non-negative matrices \(F(\{u\})\) as \(F\) varies over the measures in \(\mathcal M({\mathbf C})\), and let \(F_u\) stand for the measure such that \(F_u(\{u\})\) is maximal in that family, with respect to the partial ordering of Löwner in the cone of non-negative Hermitian matrices (see Theorem 2.2.2 in [\textit{V. K. Dubovoj, B. Fritzsche} and \textit{B. Kirstein}, Matricial version of the classical Schur problem. Teubner-Texte zur Mathematik. 129. Stuttgart etc.: B.G. Teubner (1992; Zbl 0754.47013)]). For \(l=0,\dots, n\), put \(C_{-l}:=C_l^*\) and assume that the \((n+1)\times (n+1)\) Toeplitz block-matrix \(T=(C_{j-k})_{j,k=0,\dots, n}\) is strictly positive. For \(v\in\mathbb C\), \(u\in\mathbb C\), \(I = q\times q\text{ identity matrix}\), let \[ A_n(u)=(I\; v I\; v^2 I\; \cdots v^n I)T^{-1}\left(\overline{v}I \; \overline{v}^2I \; \cdots \; \overline{v}^n I\right)^t, \] where \(x^t\) denotes the transposed of the vector \(x\). The main theorem of this paper is the following: For each \(F\in {\mathcal M}({\mathbf C})\), we have \(A_n(u)^{-1}\geq F(\{u\})\), where equality holds if and only if \(F=F_u\). The paper finishes with results concerning strict inequality in the previous estimate, and some special cases concerning equality.