Title: Phase transitions and edge scaling of number variance in Gaussian random matrices

Abstract: Using a Coulomb gas technique, we study the large $N$ behavior of the probability $\mathcal{P}_{\scriptscriptstyle N,L}(N_L)$ that a $N\times N$ Gaussian random matrix has a number $N_L$ of eigenvalues within the box $[-L,L]$, while the Wigner semi-circle has support on $[-\sqrt{2},\sqrt{2}]$. This probability scales as $\mathcal{P}_{\scriptscriptstyle N,L}(N_L=\kappa_L N)\approx\exp\left(-{\beta} N^2 \psi_L(\kappa_L)\right)$, where $\beta$ is the Dyson index of the ensemble and $\psi_L(\kappa_L)$ is a $\beta$-independent rate function that we compute exactly. We identify three regimes as $L$ is varied: (i) $\, N^{-1}\ll L<\sqrt{2}$ (bulk), (ii) $\ L\sim\sqrt{2}$ on a scale of $\mathcal{O}(N^{-{2}/{3}})$ (edge) and (iii) $\ L > \sqrt{2}$ (tail). We find a dramatic non-monotonic behavior of the number variance $V_N(L)$: after a logarithmic growth $\propto \ln (N L)$ in the bulk, $V_N(L)$ decreases abruptly as $L$ approaches the edge of the semi-circle before it decays, with $L$, as a stretched exponential for $L > \sqrt{2}$. This falloff of $V_N(L)$ at the edge is described by a scaling function $\tilde V_{\beta}$ which smoothly interpolates between the bulk (i) and the tail (iii). For $\beta = 2$ we compute $\tilde V_2$ explicitly in terms of the Airy kernel. The analytical results are verified by numerics.