Differential equations for discrete Laguerre-Sobolev orthogonal polynomials (Q2343164)

The authors study orthogonal polynomials with respect to a discrete Laguerre-Sobolev bilinear form \[ \langle p,q\rangle=\int_0^{\infty}\,p(x)q(x)\mu_{\alpha-m}(x)dx+\left(p(0),p'(0),\dots,p^{(m-1)}(0)\right) M \left(\begin{matrix} q(0)\cr q'(0)\cr \vdots\cr q^{(m-1)}(0)\cr \end{matrix}\right), \] with \(M\) an \(m\times m\) matrix, \(\alpha\) and \(m\) positive integers with \(\alpha\geq m\) and \[ \mu_{\beta}=x^{\beta}e^{-x},\;x>0,\;\beta>-1. \] As stated in the introduction, they ``prove in a constructive way that these orthogonal polynomials are eigenfunctions of a differential operator with polynomial coefficients''. The results are: Theorem 1.1. For \(m\geq 1\), let \(M=(M_{i,j})_{i,j=0}^{m-1}\) be an \(m\times m\) matrix. For \(\alpha\neq m,m-1, m-2,\dots\), consider the Laguerre-Sobolev bilinear form \[ \langle p,q\rangle=\int_0^{\infty}\,p(x)q(x)\mu_{\alpha-m}(x)dx+\mathbf{P}(0)M\mathbf{Q}(0)^T,\tag{1} \] where \(\mathbf{P}(0)=\left(p(0),p'(0),\dots,p^{(m-1)}(0)\right)\) and \(\mu_{\alpha}\) is the orthogonalizing weight for the Laguerre polynomials \((L_n^{\alpha})_n\). Then the following conditions are equivalent. {\parindent=6mm \begin{itemize}\item[A.] The discrete Laguerre-Sobolev bilinear form (1) has a sequence \((q_n)_n\) of (left) orthogonal polynomials. \item[B.] The \(m\times m\) Casorati determinant \[ \Omega(n)=\det({\mathcal R}_i(n-j))_{i,j=1}^m, \] where the \({\mathcal R}_{\ell},\,\ell=1,\dots,m\) are defined below, does not vanish for \(n\geq 0\). \end{itemize}} Moreover, if one of these properties holds, the polynomials defined by \[ q_n(x)=\left|\begin{matrix} L_n^{\alpha}(x) & L_{n-1}^{\alpha}(x) & \cdots & L_{n-m}^{\alpha}(x) \cr {\mathcal R}_1(n) & {\mathcal R}_1(n-1) & \cdots & {\mathcal R}_1(n-m) \cr \vdots & \vdots & \ddots & \vdots \cr {\mathcal R}_m(n) & {\mathcal R}_m(n-1) & \cdots & {\mathcal R}_m(n-m) \cr \end{matrix}\right|,\;n\geq 0, \] are orthogonal with respect to (1) (as usual, we take \(L_n^{\alpha}=0\) for \(n<0\)). Here \[ {\mathcal R}_{\ell}(x)={\Gamma(\alpha-m+\ell)\over (m-\ell)!} (x+1)_{m-\ell}+(\ell -1)!{\Gamma(\alpha+1+x)\over \Gamma(1+x)}\sum_{i=0}^{m-1}{(-1)^iM_{\ell-1,i}\over \Gamma(\alpha+i+1)}(x-i+1)_i. \] Theorem 1.3. Assume that any of the two equivalent properties A and B in Theorem 1.1 holds, and assume -- in addition -- that \(\alpha\) is a positive integer with \(\alpha\leq m\). Then, for any polynomial \(S\), there exists a differential operator \(D_S\) (which will be explicitly constructed) of order \(2(\text{deg}\,S+\alpha-w r(M)+1)\) with respect to which the orthogonal polynomials \((q_n)_n\) are eigenfunctions (\(\alpha-w r(M)\) is the \(\alpha\)-weighted rank of the matrix \(M\) [explicitly defined in the paper, not reproduced in this review]). Moreover, up to an additive constant, the corresponding eigenvalues \((\lambda_n)_n\) of \(D_S\) are \(\lambda_n=P_S(n)\), where \(P_S\) is the polynomial defined by the first order difference equation \[ P_S(x)-P_S(x-1)=S(x)\Omega(x), \] and \(\Omega\) is the Casorati determinant defined before. In particular, the minimal order of the differential operator having the orthogonal polynomials \((q_n)_n\) as eigenfunctions is at most \(2(\alpha-w r(M)+1)\). Taking \(m=2\), the authors recover results previously given by \textit{J. Koekoek} and \textit{R. Koekoek} [Proc. Am. Math. Soc. 112, No. 4, 1045--1054 (1991; Zbl 0737.33003)], \textit{J. Koekoek} et al. [Trans. Am. Math. Soc. 350, No. 1, 347--393 (1998; Zbl 0886.33006)] and \textit{P. Iliev} [Ann. Mat. Pura Appl. (4) 192, No. 2, 203--224 (2013; Zbl 1273.33009)].