Distribution of the first particle in discrete orthogonal polynomial ensembles (Q1400910): Difference between revisions

The authors study discrete transformed probability measures on \(\mathbb R_+^k\) of the type \[ (x_1,\dots,x_k)\mapsto\frac 1 Z \prod_{1\leq i<j\leq k} (x_i-x_j)^2\prod_{i=1}^k w(x_i), \] where \(w\colon \mathcal X\to(0,\infty)\) is a function with finite moments, i.e., \(\sum_{s\in \mathcal X}|s|^n w(s)<\infty\) for every \(n\in\mathbb N\). Here \(Z\) is the normalizing constant, and \(\mathcal X\subset(0,\infty)\) is a finite or at most countable set without accumulation points. Measures of this type recently appeared in the description of random growth models and non-colliding random walks, first-passage percolation, and in some models in combinatorics. Continuous variants of this measure appear in the theory of random matrices, most prominently in the description of the joint distribution of the eigenvalues of the Gaussian Unitary ensemble, GUE. The distribution of the largest particle in the limit \(k\to\infty\) for some of these ensembles has been identified in terms of the so-called \textit{Tracy-Widom distribution}, which is based on the solution of one of the well-known six Painlevé equations. The paper under review treats the discrete case introduced above and derives, for fixed \(k\in\mathbb N\), recursion formulas for the distribution of the largest particle in the ensemble. The main assumptions on the function \(w\) are the following. Write \(\mathcal X=\{\pi(s)\colon s\in\mathbb N_0, s< N\}\) (where \(N\leq \infty\)) with either \(\pi(s)<\pi(s+1)\) for all \(s\) or \(\pi(s)>\pi(s+1)\) for all \(s\). Then it is assumed that (1) there is an affine transformation \(\sigma\colon\mathbb R\to\mathbb R\) such that \(\sigma(\pi(s+1))=\pi(s)\), and (2) there are polynomials \(P\) and \(Q\) such that \(w(\pi(s-1))/w(\pi(s))= P(\pi(s))/Q(\pi(s))\) for all \(s\). Under these assumptions, a recurrence relation is established for certain matrix polynomials, such that the quantity \(D_s={\text{Prob}}(\max_{i=1}^k x_i<\pi(s))\) may be expressed in terms of an explicit rational function of the matrices. After a change of variables, the resulting relation is identified with some of the so-called \textit{difference Painlevé equations}. A number of explicit examples (Charlier, Meixner, Krawtchouk, \(q\)-Charlier, Little \(q\)-Jacobi and others) and the associated Painlevé equations are demonstrated, and numerical evidence is given. The proofs of the main results are based on the use of orthogonal polynomials and discrete Riemann-Hilbert problems.