Large \(n\) limit of Gaussian random matrices with external source. III: Double scaling limit (Q883046): Difference between revisions

The authors continue their study of the Gaussian random matrix model with external source, \[ \frac1{Z_ n}e^{-n\operatorname{Tr}(\frac 12 M^ 2-A M)}\,dM, \] where \(d M\) is the Lebesgue measure on the Hermitean \((n\times n)\)-matrices, and \(A\) is a diagonal matrix having \(n/2\) entries equal to a parameter \(a>0\) and \(n/2\) entries equal to \(-a\), and \(n\) is even. The matrix \(A\) is interpreted as an external source. This matrix ensemble was introduced by Brézin and Hikami as a simple model for a phase transition that is expected to exhibit universal properties in the large-\(n\) limit. In fact, for \(a>1\), the eigenvalues accumulate in two intervals, while for \(a<1\), they accumulate in one interval only. The limiting mean density of the eigenvalues can be described by the Pastur equation in terms of a certain three-sheeted Riemann surface, a feature that cannot be realized by a usual unitary random matrix ensemble without external source. This phase transition can also be reflected in terms of a related model of non-intersecting Brownian paths. Brézin and Hikami also showed that the eigenvalues of the above matrix model are distributed according to some determinantal point process. The authors of the present paper showed in their previous work that, after some appropriate scaling, for \(a\not=1\), the corresponding kernel exhibits the well-known behavior of the Gaussian unitary example (GUE), that is, it scales to the sine kernel in the bulk and to the Airy kernel at the edge of the spectrum. In the present paper, the authors study the case of \(a=a_ n\) tending to the critical value 1 at a critical velocity, which is \(O(1/\sqrt n)\). The main result is that, in an appropriate scaling, the Pearcey kernel pops up in the limit as \(n\to\infty\). (This kernel was first described by Brézin and Hikami and was recently investigated by Tracy and Widom.) The main tools in the proof are the Deift/Zhou steepest decent method and the solution to an associated Riemann-Hilbert problem. For Part I, cf. the authors, Commun. Math. Phys. 252, No. 1--3, 43--76 (2004; Zbl 1124.82309) and Part II, \textit{A. I. Aptekarev} and the authors, Commun. Math. Phys. 259, No. 2, 367--389 (2005; Zbl 1129.82014).