Orthogonal polynomials for self-dual weights (Q2689105)

Given a finite set \(X_{n}= \{x_{1}, \dots, x_{n}\}\) of points of the real line, a discrete probability measure \(d\mu(x)= \sum_{k=1}^{n} w_{k} \delta(x-x_{k})\), where \(w_{k}= \frac{\prod_{i\neq k} |x_{i} - x_{k}|^{-1}} {W}\), with \(W= \sum_{j=1}^{n} \prod_{i\neq j} |x_{i} - x_{j}|^{-1}\) is defined. This kind of discrete probabilities appear in random matrix theory providing an equivalent description of a physical ensemble in terms of particles or in terms of holes. They are characterized by remarkable self-duality under the duality definition introduced in [\textit{C. de Boor} and \textit{E. B. Saff}, Linear Algebra Appl. 75, 43--55 (1986; Zbl 0614.65035)]. The sequences of orthogonal associated with \(d\mu)\) have been studied in [\textit{V. X. Genest} et al., J. Math. Anal. Appl. 450, No. 2, 915--928 (2017; Zbl 1375.65059)], by characterizing the above weights via the mirror-symmetry of the corresponding Jacobi matrix. Therein the spectral decomposition of the Jacob matrix is used to develop a highly efficient algorithm for solving inverse spectral problem. In the contribution under review, the author deals with the case where \(X_{n}\) is a random sample from a symmetric density \(f(x)\) with the distribution function \(F(x)= \int_{-\infty}^{x} f(t) dt\) whose maximal domain of attraction is the Gumbel law. Asymptotic formulas for the expected values of the central terms of the three term recurrence relation that the orthogonal polynomials satisfy are deduced. The asymptotic behavior of self-dual orthogonal polynomials in the triangular scheme changing the notation from \(n\) to \(N=n-1\) in such a way \(N+1\) points \(\{x_{0,N}, \dots ,x_{N,N}\}\) are considered. The sequence of orthonormal polynomials satisfies the three term recurrence relation \(x P_{m, N} (x)= a_{m+1, N} P_{m+1, N} (x)+ b_{m,N} P_{m, N} (x) + a_{m-1, N} P_{m-1, N} (x)\), where \(a_{m+1,N}= a_{N+1-m, N}, b_{m,N}= b_{n-M,N}\). When \(x_{m,N}= m, m=0, \dots, N\), the orthonormal polynomials are the symmetric Krawtchouk polynomials (\(p=1/2\)). In [\textit{P. D. Dragnev} and \textit{E. B. Saff}, J. Anal. Math. 72, 223--259 (1997; Zbl 0898.31003)], these orthogonal polynomials are used as an example in deriving the asymptotics of the weighted norms and of the limiting distribution of zeros. If \(x_{m,N}= q^{-m}, m=0, \dots, N, 0<q<1\), then \(w_{m,N}= \frac{q^{m(m-1)/2}}{ (-1; q)_{N}} \frac{(q;q)_{N}}{(q;q)_{m} (q;q)_{N-m}}\), where \( (a;q)_{m}= (1-a)(1-aq)\dots (1-aq^{m-1}), m=1, \dots\), and \( (a;q)_{0}=1\). The limiting probability measure exists and it is given by the distribution of the infinite product \(\prod_{k=0}^{\infty} Y_{k}, \) where \(Y_{k}\) are independent random variables taking values \(q^{-1}\) with probability \(\frac{q^{k}}{1+ q^{k}}\) and \(1\) with probability \(\frac{1}{1+ q^{k}}\). A conjecture about the convergence of the discrete measure and the rate of its convergence when the nodes are linearly transformed to the corresponding symmetric interval is stated.