Continuum limits of random matrices and the Brownian carousel (Q834796)

The eigenvalue distribution of Gaussian orthogonal and unitary ensembles plays an important role not only in mathematics [see e.g. \textit{P. A. Deift}, Orthogonal Polynomials and Random Matrices: A Riemann-Hilbert Approach. Courant Lecture Notes in Mathematics. 3. New York, NY: Courant Institute of Mathematical Sciences. Providence, RI: American Mathematical Society (AMS). (2000; Zbl 0997.47033)]. For dimension \(n\), the ordered eigenvalues \(\lambda_1\leq \dots\leq\lambda_n\in \mathbb{R}\) have joint density \[ {1\over{Z_{n,\beta}}}e^{-\beta\sum_{k=1}^n\lambda_k^2/4}\prod_{j<k}|\lambda_j-\lambda_k|^\beta, \tag{1} \] where \(\beta=1, 2\) for the Gaussian orthogonal and unitary ensembles, respectively. The family of distributions defined by the density (1) is called the \(\beta\)-ensemble. The density (1) makes sense for any \(\beta\geq 0\), and the point process is often called Coulomb gas in Gaussian potential at inverse temperature \(\beta\). The purpose of the paper is to study its \(n\to\infty\) point process limit away from the spectral edge. It is shown that at any location away from the spectral edge, the eigenvalues of the Gaussian unitary ensemble and its general \(\beta\) siblings converge to \(\text{Sine}_{\beta}\), a translation invariant point process. This process has a geometric description in terms of the Brownian carousel, a deterministic function of Brownian motion in the hyperbolic plane. The proof of this result is based on the tridiagonal matrix model introduced by \textit{H. F. Trotter} [Adv. Math. 54, 67--82 (1984; Zbl 0562.15005)] and \textit{I. Dumitriu} and \textit{A. Edelman} [J. Math. Phys. 43, No. 11, 5830--5847 (2002; Zbl 1060.82020)]. The Brownian carousel, a description of the a continuum limit of random matrices, provides a convenient way to analyse the limiting point processes. It is shown that the gap probability of \(\text{Sine}_{\beta}\) is continuous in the gap size and \(\beta\), and its asymptotics for large gaps is computed. Moreover, the stochastic differential equation version of the Brownian carousel exhibits a phase transition at \(\beta=2\).