Weak convergence of orthogonal polynomials (Q1901417): Difference between revisions

The asymptotic behaviour of integrals \(\int f(x) p_{n + k} (x; \mu_n) p_{n + j} (x; \mu_n)d \mu_n(x)\) as \(n \to \infty\) is investigated, where \(f\) is a continuous function and \(p_k (x; \mu_n)\) \((k = 0,1,2, \ldots)\) are orthonormal polynomials for a positive measure \(\mu_n\) on the real line. Such integrals are very natural in the spectral representation of birth-and-death processes and random walks, and have some other useful applications. The analysis is done via the three-term recurrence relation \[ xp_k (x; \mu_n) = a_{k + 1,n} p_{k + 1} (x; \mu_n) + b_{k,n} p_k (x; \mu_n) + a_{k,n} p_{k - 1} (x; \mu_n), \] which gives rise to a semi- infinite tridiagonal matrix, known as the Jacobi matrix. The weak convergence can then be given in terms of the limits \(\lim_{n \to \infty} a_{n + k,n} = a^0_k\) and \(\lim_{n \to \infty} b_{n + k,n} = b^0_k\) which give rise to a doubly infinite tridiagonal matrix. The spectral analysis of such a doubly infinite Jacobi matrix is reviewed (based on work by Berzanskij, Nikishin and others) and a perturbation argument is then used to obtain the required result. Several examples are worked out in detail, including Jacobi polynomials with parameters depending on \(n\), Laguerre polynomials with a parameter depending on \(n\), Wall polynomials and dual Hahn polynomials.