Applications of multiple orthogonal polynomials with hypergeometric moment generating functions

DOI10.1016/J.AAM.2024.102709MaRDI QIDQ6564065zbMATH OpenFDO

Authors

Publication date 28 June 2024

Published in Advances in Applied Mathematics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/J.AAM.2024.102709

Diophantine approximation hypergeometric series Mellin transform multiple orthogonal polynomials random matrices

Diophantine approximation in probabilistic number theory (11K60) Random matrices (algebraic aspects) (15B52) Orthogonal polynomials and functions of hypergeometric type (Jacobi, Laguerre, Hermite, Askey scheme, etc.) (33C45) Other hypergeometric functions and integrals in several variables (33C70) Padé approximation (41A21) Orthogonal functions and polynomials, general theory of nontrigonometric harmonic analysis (42C05) Integral transforms of special functions (44A20) Convolution as an integral transform (44A35) Random matrices (probabilistic aspects) (60B20)

Review text

Multiple orthogonal polynomials of type I and type II with respect to a vector of measures \((\omega_{1}, \dots, \omega_{r})\) have been studied in the seminal monograph [\textit{E. M. Nikishin} and \textit{V. N. Sorokin}, Rational approximations and orthogonality. Transl. from the Russ. by Ralph P. Boas. Transl. ed. by S. Gelfand. Providence, RI: American Mathematical Society (1991; Zbl 0733.41001)], in the framework of Hermite-Padé rational approximations to a vector of functions. The systems of weights in which all multi-indices are normal, i.e., type I and II orthogonal polynomials attain their maximal degree, are said to be perfect. Among them, one has Angelesco-, AT- and Nikishin-systems. Examples of multiple orthogonal polynomials associated with classical weights have been exhaustively studied in the literature, see Chapter 23 in [\textit{M. E. H. Ismail}, Classical and quantum orthogonal polynomials in one variable. With two chapters by Walter Van Assche. Cambridge: Cambridge University Press (2005; Zbl 1082.42016)], among others.\N\NIn this paper, the author studies several families of multiple orthogonal polynomials associated with weights for which the moment generating functions are hypergeometric series with slightly varying parameters. Indeed, the moment generating functions of the weights are (scaled) hypergeometric functions \[f_{0} (z; \vec{a}, \vec{b})= \sum_{k\geq0} \frac{\Gamma ( k+ \vec{a}+1)}{\Gamma ( k+ \vec{b}+1)} z^{k},\] where \(\vec{a}\in ( -1, \infty)^{p}\) and \(\vec{b}\in ( -1, \infty)^{q}.\)\N\NThe weights are supported either on the unit interval with \(p=q=r, a_{j} <b_{j},\) \(\omega_{j} (x; \vec{a}, \vec{b})= \omega_{0} (x; \vec{a},\vec{ b}+ \vec{e}_{j}), j=1, \dots ,r,\) where \(\omega_{0}\) can be constructed by taking the Mellin convolution of the \(r\) beta densities \(\mathcal{B}^{a_{j}, b_{j}}, 1 \leq j\leq r, \) (Jacobi-like setting), or in the positive real line with \(p >q, a_{j} <b_{j}\) and \(p\) weights \(\{\omega_{j}(x; \vec{a}, \vec{b})\}_{j=1}^{q}\bigcup \{v_{j} (x; \vec{a}, \vec{b})\}_{j=1} ^{p-q},\) where \(\omega_{0}\) can be constructed by considering the Mellin convolution of \(q\) beta densities \(\mathcal{B}^{a_{j}, b_{j}}, 1 \leq j\leq q \) and \(p-q\) gamma densities \(\mathcal{G}^{a_{j}}, q+1 \leq j \leq p.\) The remaining weights are defined as \(v_{j}( x; \vec{a}, \vec{b})= (xD)^{j-1}[ w_{0}(x; \vec{a}, \vec{b})]\) (Laguerre-like setting). Finally, if \(p<q, a_{j} <b_{j}, 1\leq j\leq p,\) and the \(q\) weights are given as \(\omega_{j} (z; \vec{a}, \vec{b})= f_{0} (z, \vec{a},\vec{ b}+ \vec{e}_{j}), j=1, \dots ,q,\) supported on the unit circle, onme has the Bessel-like setting. Explicit formulas for the type I and type II multiple orthogonal polynomials are deduced for multi-index \(\vec{n}\) near the diagonal, i.e., \(|n_{j}-n_{i}| \leq 1\), as well as some of their properties are carefully analyzed. In particular, the asymptotic distribution of the (scaled) zeros of the type II multiple orthogonal polynomials via the free convolution is deduced in the first two cases.\N\NAs pointed above, essential to the overall approach is the use of the Mellin transform. In particular, the Mellin convolution plays a key role. This methodological approach is the most remarkable novelty of this contribution.\N\NThe connection between the multiple orthogonal polynomials in the Jacobi-like setting and Laguerre-like setting and products of truncated Haar distributed unitary random matrices and Ginibre matrices, i.e., matrices whose entries are independent standard complex Gaussian random variables is pointed out. On the other hand, the authors use the relation between multiple orthogonal polynomials and Hermite-Padé approximants to study the Diophantine properties of values of certain hypergeometric series as well as to simultaneously approximate certain hypergeometric series and to provide an explicit proof of their \(\mathbb{Q}\)-linear independence. The Diophantine quality and the approximation quality are emphasized.

Recommendations

Cites work

Cited in

(4)

This page was built for publication: Applications of multiple orthogonal polynomials with hypergeometric moment generating functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6564065)