On rational approximation of Markov functions on finite sets (Q2255302)

The theory of rational approximation of Markov functions is developed from the viewpoint of discrete Hankel operators. The classes \(\mathcal{P}_n\) of all polynomials with real coefficients of degree at most \(n\) and the corresponding class of rational functions \(\mathcal{R}_{m,n}\) are considered. For a real-valued function \(f\) on a finite set \(e_n\) of \(2n+1\) points, the error \(\rho_{n-1,n}(f;e_n)\) in the best approximation of \(f\) by rational functions from \(\mathcal{R}_{n-1,n}\) is determined. Analogously to the approach of the theory of Adamyan, Arov and Krein, an appropriate discrete Hankel operator \(B_f\) acting on polynomials is taken into account. The developed relation between the smallest singular value of \(B_f\) and the error \(\rho_{n-1,n}(f;e_n)\) forms a version of the AAK theorem. The Markov function \(\hat \mu\), the Cauchy transform of a positive Borel measure \(\mu\) with the compact support, is considered. The properties and the value of the smallest singular value of \(B_{\hat \mu}\) are investigated. This results in explicit integral formulas for the error in the best rational approximation \(\rho_{n-1,n} (\hat\mu; E_N)\), with \(E_N\) being a set of \(N\geq 2n+1\) points.