The Pearcey process (Q863118)

Determinantal point processes arise in random matrix theory and have an associated kernel \(K (x,y)\), so that certain distribution functions for the process can be expressed in terms of determinants involving this kernel. Such kernels typically have a parameter \(n\) associated with the size of the system, and their scaling limits are known to be universal, so that certain canonical operators \(K\) are the limits for a wide variety of kernels. When such kernels can be expressed as a double integral \[ \int_{C_1}\int_{C_2}\, f (s,t) e^{\Phi_{n}(s,t;x,y)}\,ds\,dt \] nontrivial limit laws arise by coalescence of saddle points. The simplest case is a coalescence of two saddles (fold singularity \(\Phi_{2} (z)=\frac{1}{3}z^3+\lambda z\)), which leads to the Airy kernel, or more generally to an extended Airy kernel which corresponds to the Airy process. The next case (cusp singularity \(\Phi_{3} (z)=\frac{1}{4}z^4+\lambda_2 z^2 +\lambda_1 z\)) corresponds to Pearcey functions and it leads to the Pearcey kernel, which first appeared in the work of \textit{E. Brézin} and \textit{S. Hikami} [Phys. Rev. E (3) 57, No. 4, 4140--4149 (1998); Phys. Rev. E (3) 58, No. 6, part A, 7176--7185 (1998)] in the context of a Gaussian unitary random matrix with external source, and later in a non-intersecting Brownian motion model by \textit{A. I. Aptekarev, P. M. Bleher} and \textit{A. B. J. Kuijlaars} [Commun. Math. Phys. 259, No. 2, 367--389 (2005; Zbl 1129.82014)], which is the starting point of the authors. The authors introduce and study an extended Pearcey kernel that in the non-intersecting Brownian motion model gives the limiting joint distribution at several times \(k\) near a critical time where two groups of Brownian paths split. They find a system of partial differential equations for the probability that no path passes through a set \(X_k\) at time \(k\), which to the author's surprise closes at the end of the computation. The authors also show that kernels involving higher-order singularities can arise in fictitious Brownian motion models, in which the end-points of the paths are complex numbers.