Painlevé kernels in Hermitian matrix models (Q485297)

This article reviews the different universal determinantal processes and their kernels arising in matrix models with the rigorous approach of Riemann-Hilbert problems and orthogonal polynomials. The first half of the paper summarizes known result about scaling limits of hermitian one-matrix models and the corresponding kernels both in the regular cases (sine and Airy kernel) and the singular cases: exterior singular point, interior singular point (Painlevé II kernel), singular edge point (Painlevé I kernel) and Pearcey kernel (external source). The results are summarized in a very pedagogical way with the illustration of their corresponding quartic potentials. \newline Advanced readers will find the second half of the paper much more challenging since hermitian two-matrix models are introduced. In this context, the classification of all types of singular points is still missing. The starting-point of the analysis is the connection with biorthogonal polynomials and higher-dimensional Riemann-Hilbert problems. The main contribution of the paper is to produce the \(4\times 4\) Riemann-Hilbert problem associated with the highly singular case for the quartic/quadratic potentials: \[ V(x)=\frac{1}{2}x^2 \text{ and } W(y)=\frac{1}{4}y^4+\frac{\alpha}{2}y^2 \] around \(\tau^2=2+\alpha\) and \(\alpha \tau=1\) where \(\tau\) is the strength of the interaction between the two matrices (\(\tau M_1 M_2\)). In particular, the specific kernel \(K_{cr}(x_1,x_2)\) is given in terms of the Riemann-Hilbert components. The reductions for specific values of its parameters to the other kernels (Pearcey, Painlevé II) is also provided. Eventually, the author proposes a possible generalization towards more singular cases as open problems. As a conclusion, this article is a good survey of results regarding determinantal processes and their kernels in matrix models using Riemann-Hilbert problems approach. People interested in the establishment of the rigorous classification of the various universal local behaviors arising in matrix models in connection with integrable systems will find it particularly enlightening.