Representations of central matrix-valued Carathéodory functions in both nondegenerate and degenerate cases (Q1764263): Difference between revisions

A sequence \(\{\Gamma_j\}_{j=0}^n\) of complex \(q\times q\)-matrices is said to be a Carathéodory sequence if the Toeplitz matrix \[ S_n=\begin{pmatrix}\Gamma_0 &\dots &0\\ \vdots & \ddots & \vdots\\ \Gamma_n & \dots &\Gamma_0\end{pmatrix} \] has a nonnegative real part \(T_n=\text{Re}\, S_n\geq 0\). The matrix Carathéodory interpolation problem \({\mathcal C}_q[{\mathbb D},\,\{\Gamma_j\}_{j=0}^n]\) consists in the following: given a sequence \(\{\Gamma_j\}_{j=0}^n\) of complex \(q\times q\)-matrices, find a Carathéodory \(q\times q\)-matrix function \(\Omega(z)\) with the Taylor coefficients \(\{\Gamma_j\}_{j=0}^n\). It is known that the problem \({\mathcal C}_q[{\mathbb D}\), \(\{\Gamma_j\}_{j=0}^n]\) is solvable if and only if \(\{\Gamma_j\}_{j=0}^n\) is a Carathéodory sequence. In this case the set of all \(\Gamma_{n+1}/2\) such that \(\{\Gamma_j\}_{j=0}^{n+1}\) is also a Carathéodory sequence forms a matrix ball \(\{M_{n+1}+L_{n+1}KR_{n+1}:\| K\| \leq 1\}\) with the center \(M_{n+1}\) and the left and the right radii \(L_{n+1}\) and \(R_{n+1}\) being some \(q\times q\)-matrices. The function \(\Omega_c(z)=\sum_{k=0}^\infty\Gamma_kz^k\) is called the central Carathéodory function if \(\Gamma_{n+k}=2M_{n+k}\), \(k\in{\mathbb N}\) [\textit{B. Fritzsche} and \textit{B. Kirstein}, ``An extension problem for non-negative Hermitian block Toeplitz matrices'', Math. Nachr. 130, 121--135 (1987; Zbl 0624.15011)]. In the present paper quotient representations \(\Omega_c(z)=a_n(z)b_n(z)^{-1}=d_n(z)^{-1}c_n(z)\) of the central Carathéodory function are investigated both in the nondegenerate (\(\det T_n\neq 0\)) and the degenerate (\(\det T_n= 0\)) cases. In the nondegenerate case matrix polynomials \(a_n(z)\) and \(b_n(z)\) occur in the Arov-Krein resolvent matrix which describes the set of all solutions of the problem \({\mathcal C}_q[{\mathbb D}\), \(\{\Gamma_j\}_{j=0}^n]\), see \textit{D. Z.~Arov} and \textit{M. G.~Krein} [Funct. Anal. Appl. 15, 123--126 (1981); translation from Funkts. Anal. Prilozh., 15, No. 2, 61--64 (1981; Zbl 0484.46060)]. The authors found recurrent formulas for the Arov-Krein resolvent matrix, which lead to its factorization corresponding to the step-by-step algorithm of the solution of the problem \({\mathcal C}_q[{\mathbb D},\,\{\Gamma_j\}_{j=0}^n]\).