Inverse results on row sequences of Hermite-Padé approximation (Q1701889)

In this paper some analytic properties of a vector of formal power expansion about the origin are studied in terms of the asymptotic behavior of the sequence of the common denominators of the type II Hermite-Padé approximants. Two conjectures to questions posed by A. A. Gonchar for row sequences of Padé approximants are restated. In order to have a comprehensive approach to the contents of this contribution, let us introduce some basic background. Let \(\mathbf f= (f_1,\ldots, f_d)\) be a system of \(d\) formal Taylor expansions. For a fixed multi-index \(\mathbf m= (m_1, \ldots, m_d)\in \mathbb Z^d_+\setminus\{\mathbf 0\}\), let denote \(|\mathbf m|= m_1+\cdots + m_d\). Then for each \(n\geq \max\{n_1,\ldots, n_d\}\) there exist polynomials \(q(z)\) and \(p_k(z), k= 1,\ldots, d\), such that {\parindent=0.7cm\begin{itemize}\item[--] deg \(p_k\leq n-m_k, k=1,\ldots ,d\), with deg \(q\leq|\mathbf m|\), and \(q\) not identically zero. \item[--] \(q(z) f_k(z)- p_k(z)= A_k z^{n+1} + \cdots\) \end{itemize}} The vector rational function \(R_{n,\mathbf m}(z)= (\frac{p_1(z)}{q(z)},\ldots,\frac{p_d(z)}{q(z)})\) is said to be an \((n, m)\) type II Hermite-Padé approximant of \(\mathbf f\). In contrast with Padé approximation, Hermite-Padé approximants are, in general, not uniquely determined. Given \((n,\mathbf m)\), let consider one particular solution and let write \(R_{n,\mathbf m}(z)= \frac{(p_{n,\mathbf m,1}(z), \ldots, p_{n,\mathbf m,d}(z))}{q_{n,\mathbf m}(z)}\), where the denominator is coprime with all the entries of the vector of polynomials in the numerator. For fixed \(\mathbf m\), the sequences \((R_{n,\mathbf m})_{n\geq|\mathbf m|}\) are called \textit{row sequences}. A point \(\zeta\in \mathbb{C}\setminus\{\mathbf 0\}\) 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 \(s= 1, \ldots, \tau\) there exists at least one polynomial combination \(F=\sum_ {k=1}^d s_kf_k\), with \(s_k\) polynomial of degree less than \(m_k, k=1,\ldots, d\), which is analytic on a neighborhood of the closed disk \(\bar{D}(0,|\zeta|)\) except for a pole at \(z=\zeta\) with exact order \(s\). \(\mathbf f\) cannot have more than \(|\mathbf m|\) system poles with respect to \(\mathbf m\). In a similar way, a system singularity point of \(\mathbf f\) is defined. For a fixed \((\mathbf f,\mathbf m)\) and \(\zeta\in \mathbb{C}\setminus\{\mathbf 0\},\) let denote by \(\zeta_{n,1}, \ldots, \zeta_{n,r_{n}},\) the zeros of \(q_{n,\mathbf{m}}\) indexed according their increasing distance from \(\zeta\). Following \textit{A. A. Gonchar} [Mat. Sb., Nov. Ser. 115(157), 590--613 (1981; Zbl 0477.30021)]], for such a \(\zeta\) two characteristic values \(\nu= \lambda(\zeta)\) and \(\mu =\mu(\zeta),\) defined in terms of the lim sup of the distance of the above zeros to \(\zeta\) and the \(n\)-th root of such a distance, respectively, are introduced. If \(\zeta\) is a system pole of \((\mathbf{f},\mathbf{m})\) of order \(\tau\), then \(\mu(\zeta) \geq \tau\). This is a direct type result. The authors claim two conjectures concerning sufficient conditions to have either a system pole point or a system singularity point in terms of the above characteristic values. An inverse result (Theorem 5.1). Let assume that \(\lim_{n\rightarrow\infty} q_{n,\mathbf{m}}= q_{\mathbf{m}},\) \(\operatorname{deg}q_{\mathbf{m}}=|\mathbf{m}|\), \(q_{\mathbf{m}}(0)\neq0,\) and their zeros are simple. For a fixed \(m^\ast\), with \(1\leq m^\ast\leq \max\{m_{k}, k=1, \ldots, d\}\), let consider \(F= \sum_ {k=1}^{d} s_{k}f_{k},\) \(s_{k}\) arbitrary polynomials of degree \(\leq m_{k}- m^\ast\). Then the closure of the disk \(D_{m^\ast-1}\) centered at the origin and radius defined in terms of the difference between two consecutive incomplete Padé approximants of \(F\), contains at least \(m^\ast\) singular points of \(F\) which are zeros of \(q_{\mathbf{m}}\) and those lying in the interior of the disk are simple poles of \(F\). In particular, all such zeros are system singularities of \(\mathbf{f}\). Auxiliary results based on incomplete Padé approximants and the regularization of the sequence of coefficients of a Taylor series are carefully analyzed in order to prove the last theorem.