The boundary Carathéodory-Fejér interpolation problem (Q549802)

From MaRDI portal
scientific article
Language Label Description Also known as
English
The boundary Carathéodory-Fejér interpolation problem
scientific article

    Statements

    The boundary Carathéodory-Fejér interpolation problem (English)
    0 references
    0 references
    0 references
    0 references
    18 July 2011
    0 references
    Given a point \(x\in\mathbb{R}\), a non-negative integer \(n\) and a system \(a=(a^{-1},a^0,\dots,a^n)\) of complex numbers, the boundary Carathéodory-Fejér interpolation problem (\(\partial CF\mathcal{P}\)) entails finding a function \(f\) in the Pick class \(\mathcal{P}\) (defined as the set of maps which are analytic and have non-negative imaginary part on the upper half-plane) such that \(f\) is analytic in a deleted neighborhood of \(x\) and \[ L_k(f,x)=a^k,\quad k=-1,0,\dots,n, \] where \(L_k(f,x)\) is the \(k\)th Laurent coefficient of \(f\) at \(x\). It is the aim of the paper under review to provide new solvability criteria for this problem. The approach uses only elementary methods making it easily accessible to engineers working in applied areas (such as signal processing) where these problems arise. The proposed criteria involve certain positivity conditions for the associated Hankel matrix \(H_m(a):=[a^{i+j-1}]_{i,j=1}^m\), where \(m\leq\frac{n+1}{2}\) is a positive integer, \(a^{-1}\leq 0\) and \(\text{Im}a^0\geq 0\). More precisely, if \(S(a):=\{k=-1,0,\dots,n:\text{Im}a^k\neq 0\}\neq\emptyset\), then the Problem \(\partial CF\mathcal{P}\) is solvable if and only if \(\rho(a):=\min S(a)\) is even, \(H_m(a)>0\) and \(\text{Im}a^{2m}>0\), where \(m=\frac{1}{2}\rho(a)\). When \(a\in\mathbb{R}^{n+2}\), one can distinguish two different situations. In the first case, when \(n=2m-1\) is odd, the problem \(\partial CF\mathcal{P}\) is solvable if and only if \(H_m(a)\) is either positive definite or southeast-minimally positive (i.e., \(H_m(a)\geq 0\) and, for every \(\epsilon>0\), \(H_m(a)-\text{diag}(0,\dots,0,\epsilon)\) is not positive). In the second case, when \(n=2m\) is even, the Problem \(\partial CF\mathcal{P}\) is solvable if and only if either \(H_m(a)\) is positive definite or both \(H_m(a)\) is southeast-minimally positive and \(a^n=[a^m,a^{m+1},\dots,a^{m+r-1}]H_r(a)^{-1}[a^{m+1},a^{m+2},\dots,a^{m+r}]^{\text{t}}\), where \(r=\text{rank}H_m(a)\). The proofs are based on a reduction procedure due to \textit{G. Julia} [Acta Math. 42(1919-20), 349--355 (1919; JFM 47.0272.01)] and \textit{R. Nevanlinna} [Math. Zeitschr. 13, 1--9 (1922; JFM 48.0322.02)]. A linear fractional parametrization of the set of solutions of the interpolation problem is also given for the situation when the target values \(a^{-1},a^0,\dots,a^n\) are all real.
    0 references
    Pick class
    0 references
    boundary interpolation
    0 references
    Carathéodory-Fejér interpolation
    0 references
    Hankel matrix
    0 references
    Schur complement
    0 references
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references