PDEs satisfied by extreme eigenvalues distributions of GUE and LUE

Abstract

In this paper we study, $\textsf{Prob}(n,a,b),$ the probability that all the eigenvalues of finite $n$ unitary ensembles lie in the interval $(a,b)$. This is identical to the probability that the largest eigenvalue is less than $b$ and the smallest eigenvalue is greater than $a$. It is shown that a quantity allied to $\textsf{Prob}(n,a,b)$, namely, $$ H_n(a,b):=\left[\frac{\partial}{\partial a}+\frac{\partial}{\partial b}\right]\ln\textsf{Prob}(n,a,b),$$ in the Gaussian Unitary Ensemble (GUE) and $$ H_n(a,b):=\left[a\frac{\partial}{\partial a}+b\frac{\partial}{\partial b}\right]\ln \textsf{Prob}(n,a,b),$$ in the Laguerre Unitary Ensemble (LUE) satisfy certain nonlinear partial differential equations for fixed $n$, interpreting $H_n(a,b)$ as a function of $a$ and $b$. These partial differential equations maybe considered as two variable generalizations of a Painlev\'{e} IV and a Painlev\'{e} V system, respectively. As an application of our result, we give an analytic proof that the extreme eigenvalues of the GUE and the LUE, when suitably centered and scaled, are asymptotically independent.