*(English)*Zbl 0944.42013

In this important and interesting paper, the authors again illustrate the awesome power of the Deift-Zhou steepest descent method associated with Riemann-Hilbert problems. Let $V:\mathbb{R}\to \mathbb{R}$ be real-valued and analytic on $\mathbb{R}$, with

For $n\ge 1$, let ${\left\{{p}_{k}(x;n)\right\}}_{k=0}^{\infty}$ denote the sequence of orthonormal polynomials for the weight $exp(-nV)$, so that

The authors derive Plancherel-Rotach asymptotics for ${p}_{n}(z;n)$ as $n\to \infty $, valid in every region of the plane. The precision of the asymptotics on and off the real line is remarkable. The proofs involve the Fokas-Its-Kitaev identify for the orthonormal polynomials as solutions of a Riemann-Hilbert problem, followed by the Deift-Zhou steepest descent technique. The details are intricate, but are clearly presented, and the main ideas are summarized to guide the reader through the proofs.

As an application, the authors prove universality limits that arise in random matrix theory. These involve the weighted reproducing kernel

and have the form

uniformly for $s$, $t$ in compact subsets of $\mathbb{R}$. Here $a$ lies in a subinterval of the support of the equilibrium measure ${\mu}_{V}$ associated with the field $V$, and $\psi $ is the density of that equilibrium measure.

##### MSC:

42C05 | General theory of orthogonal functions and polynomials |

15A52 | Random matrices (MSC2000) |

41A60 | Asymptotic approximations, asymptotic expansions (steepest descent, etc.) |

82B41 | Random walks, random surfaces, lattice animals, etc. (statistical mechanics) |