Generalized Padé-type approximants to continuous functions (Q2494441)

The author extends ideas from a series of previous papers to define a generalized Padé-type approximant for functions of class C\(^{\infty}\) on compact subsets of \(\mathbb R^n\). In section 1 the compact set \(E\neq \emptyset\) and positive measure \(\mu\) on \(E\) satisfy Markov's inequality \(M_2\): for any \(\alpha\in\hbox{N}^n\) there exists a constant \(M=M(\alpha)\) and two integers \(r,m>1\) such that \[ \|D^{\alpha}Q\|_E\leq (M\,\text{deg}\,Q)^{m+r|\alpha|} \left(\int_E\,|Q(x)|^2\,d\mu(x)\right)^{1/2}, \] for any polynomial \(Q\) on \(\mathbb R^n\); \(|\alpha|=\alpha_1+\cdots \alpha_n\). Assuming that the orthonormal family of polynomials \(\{\varphi_j\}\) in \(\mathbb L^2(E,\mu)\) satisfies several conditions (self summability, span\(\{\overline{\varphi_0},\ldots, \overline{\varphi_m}\}\) is a Haar space, certain determinants different from zero), the generalized Padé-type approximant of a function is defined using the coefficients of the Fourier expansion of \(f\) with respect to the orthonormal family. This is followed in section 2 by an integral representation of this Padé-type approximant. In section 3 the set \(E\) satisfies Markov's property \(M_{\infty}\): for any \(\alpha\in\text{N}^n\) there exists a constant \(M=M(\alpha)\) and an integer \(m\geq 1\) such that \[ \|D^{\alpha}Q\|_E\leq M(\text{deg}\,Q)^{m|\alpha|} \|Q\|_E, \] for any polynomial \(Q\) on \(\mathbb R^n\). The generalized Padé-type approximant is again defined using the Fourier expansion with respect to an orthonormal system \(\{\psi_j\}\) on a Hilbert space \(H\), whose existence follows from \textit{A. Zeriahi} [Inégalités de Markov et le développment en série de polynômes orthogonaux de fonctions \(C^{\infty}\) et \(A^{\infty}\), Mittag Leffler Institut Djursholm (1991)].