Direct and inverse results for multipoint Hermite-Padé approximants (Q2318063)

Given \(d\) functions \(\mathbf{f}=(f_1,\ldots,f_d)\), analytic in a region \(E\subset\mathbb{C}\), and a triangular array \(\alpha=\{\alpha_{n,k}\}\), \(k=1,\ldots,n\), \(n\in\mathbb{N}\) of (possibly repeated) interpolation points, the authors consider a multipoint Hermite-Padé approximant, which is a vector of simultaneous rational interpolants \(\mathbf{P}_{n,\mathbf{m}}/Q_{n,\mathbf{m}}\) of the form \((n-\mathbf{m}/|\mathbf{m}|)\), with \(\mathbf{m}=(m_1,\ldots,m_d)\) and with common denominator of degree \(|\mathbf{m}|=\sum_i m_i\). Classical Hermite-Padé approximation corresponds to all interpolation points \(\alpha_{n,k}\) at the origin and \(E\) a disk. In [\textit{J. Cacoq} et al., Constr. Approx. 38, No. 1, 133--160 (2013; Zbl 1271.30010)] convergence for \(n\to\infty\) was studied for that classical case. In this paper these results are generalized to the multipoint case assuming there is a certain distribution for the interpolation points. System poles need to be defined in terms of \(\mathbf{f}\) and \(\mathbf{m}\) which requires some subtleties. The direct results include exact rates of convergence of the interpolants and of their denominator, recovering the system poles of a given system \(\mathbf{f}\). The inverse results show what can be derived about the analyticity of \(\mathbf{f}\) from the behaviour of the approximants. Thus the classical Montessus de Ballore-Gonchar results are generalized to this multipoint situation.