Determining system poles using row sequences of orthogonal Hermite-Padé approximants

Approximation in the complex plane (30E10) Padé approximation (41A21) Rate of convergence, degree of approximation (41A25) Inverse theorems in approximation theory (41A27) Simultaneous approximation (41A28)

Abstract: Given a system of functions

e x t u p e x t b f F = (F_{1}, l d o t s, F_{d}),

analytic on a neighborhood of some compact subset

E

of the complex plane with simply connected complement, we define a sequence of vector rational functions with common denominator in terms of the orthogonal expansions of the components

F_{k}, k = 1, l d o t s, d,

with respect to a sequence of orthonormal polynomials associated with a measure

m u

whose support is contained in

E

. Such sequences of vector rational functions resemble row sequences of type II Hermite-Pad'e approximants. Under appropriate assumptions on

m u,

we give necessary and sufficient conditions for the convergence with geometric rate of the common denominators of the sequence of vector rational functions so constructed. The exact rate of convergence of these denominators is provided and the rate of convergence of the simultaneous approximants is estimated. It is shown that the common denominator of the approximants detect the location of the poles of the system of functions.

Review text

The aim of the paper is to give a generalization of the Montessus de Ballore Gonchar theorem (Theorems 1.4 and 1.7 from Constr. Approx. 38 (2013),133--160; Zbl. 1271.30010) in the direction that Taylor expansions of the underlying \(d\)-tuples (\(d\) an integer, \(d\geq 1\)) of functions are replaced by expansions in orthogonal polynomials. The main result is Theorem 1.2. Let \({\mathbf F}=(F_1,F_2,\ldots,F_d)\in{\mathcal H}(E)^d,\;{\mathbf m}\in {\mathbb N}^d\) be a multi-index, and \(\mu\in {\mathbf Reg}_{1,2}^{|{\mathbf m}|}(E)\). Then the following two assertions are equivalent: (a) \({\mathbf F}\) has exactly \(|{\mathbf m}|\) system poles with respect to \({\mathbf m}\), counting multiplicities. (b) The polynomials \(Q_{n,|{\mathbf m}|}^{\mu}\) of \({\mathbf F}\) are uniquely determined for all sufficiently large \(n\), and there exists a polynomial \(Q_{|{\mathbf m}|}\) of degree \(|{\mathbf m}|\) such that \[ \limsup_{n\rightarrow\infty}\,||Q_{n,|{\mathbf m}|}^{\mu}-Q_{|{\mathbf m}|}||^{1/n}=\theta<1. \] Moreover, if either (a) or (b) takes place, then \(Q_{|{\mathbf m}|}=Q_{\mathbf m}^F\), \[ \theta=\max\left\{{|\Phi(\xi)|\over \rho_i({\mathbf F},{\mathbf m})}:\xi\in\mathcal{P}({\mathbf F},{\mathbf m}) \right\}, \] and for any compact subset \(K\) of \(D_{i,0}^{\ast}({\mathbf F},{\mathbf m})\setminus{\mathcal P}({\mathbf F},{\mathbf m})\), \[ \limsup_{n\rightarrow\infty}\,||R_{n,{\mathbf m},i}^{\mu}-F_i||_{K}^{1/n}\leq {||\Phi||_K \over \rho_{i,0}^{\ast}({\mathbf F},{\mathbf m})}, \] where \(||\cdot||_K\) denotes the sup-norm on \(K\) and if \(K\subset E\), then \(||\Phi||_K\) is relaced by \(1\). It would take too much space to give \textit{all} the necessary notations and definitions in detail here. But an important role is played by the \((n,|{\mathbf m}|)\) simultaneous Fourier-Padé approximants of \({\mathbf F}=(F_1,F_2,\ldots,F_d)\) with respect to \(\mu\) (the defining equations use power series with \(z^n\) replaced by the \(n\)-th degree orthogonal polynomial with respect to \(\mu\)): \[ \tilde{{\mathbf R}}_{n,{\mathbf m}}^{\mu}=({\tilde R}_{n,{\mathbf m},1}^{\mu}, \ldots, {\tilde R}_{n,{\mathbf m},d}^{\mu})=({\tilde P}_{n,{\mathbf m},1}^{\mu}/{\tilde Q}_{n,{\mathbf m},1}^{\mu}, \ldots, {\tilde P}_{n,{\mathbf m},d}^{\mu}/{\tilde Q}_{n,{\mathbf m},d}^{\mu}). \] The interest lies specifically in the fact that the limit \(n\rightarrow\infty\) denotes convergence along \textbf{row} sequences; most results up to now deal with diagonal or near diagonal sequences. Also, \(Q_{\mathbf m}^{F}\) have as zeros just the so-called \textit{system poles} of \({\mathbf F}\) with respect to \({\mathbf m}\), taking into account their order. These system poles are defined as follows: \[ {\mathbf F}=(F_1,F_2,\ldots,F_d)\in{\mathcal H}(E)^d,\;{\mathbf m}=(m_1,m_2,\ldots,m_d)\in{\mathbb N}^d, \] then \(\xi\in{\mathbf C}\) is a system pole of order \(\tau\) of \({\mathbf F}\) with respect to \({\mathbf m}\) if \(\tau\) is the largest positive integer such that for each \(t=1,2,\ldots,\tau \) there exists at least one polynomial combination of the form \[ \sum_{i=1}^d\,v_i F_i,\;deg\,v_i<m_i\;(1\leq i\leq d), \] which is holomorphic in a neighborhood of \(\overline{D}_{|\Phi(\xi)|}\), except for a pole at \(z=\xi\) of exact order \(t\). (\(D_{\rho}= E\cup \{z\in{\mathbf C}:|\Phi(z)|<\rho\}\) with \(\Phi\) the unique conformal mapping from \(\overline{{\mathbf C}}\setminus E\) onto \(\{\overline{{\mathbf C}}\setminus\{w\in{\mathbf C}:|w|\leq 1\}\) satisfying \(\Phi(\infty)=\infty\), \(\Phi'(w)>0\)).

Recommendations

Cites work

Cited in

(9)

This page was built for publication: Determining system poles using row sequences of orthogonal Hermite-Padé approximants

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1749038)