Ratio asymptotic of Hermite-Padé orthogonal polynomials for Nikishin systems. II (Q927678)

In the theory of orthogonal polynomials the theorem of \textit{E. A. Rakhmanov} [Math. USSR Sb. 32, 199--213 (1977); translation from Mat. Sb., N. Ser. 103(145), 237-252 (1977; Zbl 0373.30034) and Math. USSR Sb. 46, 105--117 (1983); translation from Mat. Sb., N. Ser. 118(160), 104--117 (1982; Zbl 0509.30028)] on ratio asymptotics for the monic orthogonal polynomials is well known: \[ \lim_{n\in\mathbb{Z}_{+}}\,{Q_{n+1}(z)\over Q_n(z)}={\varphi(z)\over \varphi'(\infty)} \] uniformly in \(z\) on each compact subset of \(\mathbb{C}\setminus\Delta\). Here the \(Q_n\) are orthogonal with respect to a finite positive Borel measure \(s\) supported on a bounded real interval \(\Delta\) with \(s'>0\) a.e. on its support; furthermore, \(\varphi(z)\) denotes the conformal mapping of \(\overline{\mathbb{C}}\setminus\Delta\) onto \(\{w\,:\,| w| >1\}\), such that \(\varphi(\infty)=\infty,\;\varphi'(\infty)>0\). This result was extended in a paper by \textit{A. I. Aptekarev, G. López Lagomasino} and \textit{I. A. Rocha} [Mat. Sb. 196, 1089--1107 (2005; Zbl 1077.42015)] to orthogonal polynomials for Nikishin systems (introduced by \textit{E. M. Nikishin} in [Math. USSR Sb. 41, 409--426 (1982); translation from Mat. Sb., N. Ser. 113(155), 499--519 (1980; Zbl 0456.30009)]) with certain conditions on the underlying measures. Starting with a system \((\sigma_1,\ldots,\sigma_m)\) of finite Borel measures on the real line with constant sign and compact support containing infinitely many points, let Co(supp\((\sigma_k))=\Delta_k\) denote the smallest interval which contains supp\((\sigma_k)\) and assume \[ \Delta_k\cap\Delta_{k-1}=\emptyset,\;k=1,\ldots,m-1. \] The \textit{Nikishin system} of measures \(S=(s_1,\ldots,s_m)\) is then given by \[ \begin{aligned} s_1=\sigma_1,\,s_2=\langle\sigma_1,\sigma_2\rangle,\,\ldots\,,\,s_m=\langle\sigma_1,\langle\sigma_2,\ldots, \sigma_m\rangle\rangle,\;\text{where}\\ \langle\sigma_1,\sigma_2\rangle=\int\,{d\sigma_2(t)\over x-t}\,d\sigma_1(x)=\hat{\sigma}_2(x)d\sigma_1(x).\end{aligned} \] For a fixed multi-index \(\mathbf{n}=(n_1,\ldots,n_m)\in\mathbb{Z}^m_{+}\) the monic polynomial \(Q_n\) is called the \textbf{n}th \textit{multiple orthogonal polynomial} with respect to \(S\) if it is not identically zero, deg\(Q_n\leq| \mathbf{n}| =n_1+\cdots +n_m\), and \[ \int\,Q_{\mathbf{n}}(x)x^{\nu}\,ds_k(x)=0,\;\nu=0,\ldots,n_k-1,\;k=1,\ldots,m. \] The main result now reads as Let the Nikishin system satisfy supp\((\sigma_k)={\tilde{\Delta}}_k\cup e_k\;(1\leq k\leq m)\) where \({\tilde{\Delta}}_k\) is a bounded interval on the real line, \(| \sigma'_k| >0\) a.e. on \({\tilde{\Delta}}_k\) and \(e_k\) is a set without accumulation points in \(\mathbb{R}\setminus{\tilde{\Delta}}_k\). Let \(\Lambda\subset\mathbb{Z}^m_{+}(\ast)\) be an infinite sequence of distinct multi-indices with the property that \[ \max_{\mathbf{n}\in\Lambda}\,(\max_{1\leq k\leq m}\,mn -| \mathbf{n}| )<\infty. \] Assume that there exists an \(\ell\in\{1,\ldots,m\}\) and a fixed permutation \(\tau\) of \(\{1,\ldots,m\}\) such that for all \(\mathbf{n}\in\Lambda\) we have \(\mathbf{n},\text\textbf{n}_{\ell}\in\mathbb{Z}^m_{+}(\ast,\tau)\). Then \[ \lim_{\mathbf{n}\in\Lambda}\,{Q_{\mathbf{n}_{\ell}}(z)\over Q_{\mathbf{n}}}=G_0^{(\tau^{-1}(\ell))}(z), \] uniformly in \(z\) on compact subsets of \(\mathbb{C}\setminus\text{supp}(\sigma_1)\). Here \[ \begin{aligned}\mathbb{Z}^m_{+}(\ast)=\{\mathbf{n}\in\mathbb{Z}^m_{+}\,:\,{\not \exists}\;1\leq i<j<k\leq m,\;\text{with} \;n_i<n_j<n_k\},\\ \mathbb{Z}^m_{+}(\ast,\tau)=\{\text\textbf{n}\in\mathbb{Z}^m_{+}(\ast)\,:\, \tau_{\text\textbf{n}}=\tau\},\end{aligned} \] where \(\tau_{\mathbf{n}}\) is the permutation of \((1,\ldots,m)\) derived from the multi index \(\mathbf{n}\) in the following way \[ \tau_{\mathbf{n}}\,(i)=j\;\text{if}\;\begin{cases} n_j>n_k &for\;k<j,\;k\not\in\{\tau_{\mathbf{n}}(1), \ldots,\tau_{\mathbf{n}}(i-1)\}, \cr n_j\geq n_k &for\;k>j,\;k\not\in\{\tau_{\mathbf{n}}(1), \ldots,\tau_{\mathbf{n}}(i-1)\}\end{cases}. \] For \(\mathbf{n}\in\mathbb{Z}^m_{+}\) and \(\ell\in\{1,\ldots,m\}\) \[ \mathbf{n}_{\ell}=(n_1,\ldots,n_{\ell-1},n_{\ell}+1,n_{\ell +1},\ldots,n_m). \] \(G^{(\ell)}\) is the conformal mapping of the \((m+1)\)-sheeted Riemann surface \(\mathcal R\) \[ {\mathcal R}={\overline{\cup_{k=0}^m\;{\mathcal R}_k}} \] onto \(\overline{\mathbb{C}}\) such that \[ G^{(\ell)}(z)=z+{\mathcal O}(1),\;z\rightarrow\infty^{0}, \;G^{(\ell)}(z)=C/z+{\mathcal O}(1/z^2),\;z\rightarrow\infty^{\ell}, \] and \(G^{(\ell)}_k\) is the branch of \(G^{(\ell)}\) onto \({\mathcal R}_k\). The construction of the \(\mathcal R\) uses \[ {\mathcal R}_0=\overline{\mathbb{C}}\setminus{\tilde{\Delta}}_1, \;{\mathcal R}_k=\overline{\mathbb{C}}\setminus({\tilde{\Delta}}_k\cup{\tilde{\Delta}}_{k+1}),\;k=1,\ldots,m-1, \;{\mathcal R}_m=\overline{\mathbb{C}}\setminus{\tilde{\Delta}}_m. \] There is also an interesting result on ratio asymptotics for the ``polynomials of the second kind'' that is proved using intermediate results on the location of zeros; the interested reader is referred to the paper for the (notational) details, stating all formulae is outside the scope of a review.