A random cover of a compact hyperbolic surface has relative spectral gap \(\frac{3}{16}-\varepsilon\) (Q2671212)

Call a 2-dimensional Riemannian manifold (without boundary) with constant curvature \(-1\) a hyperbolic surface. If \(X\) is a compact, connected, hyperbolic surface, then for each \(n\in \{1,2,3,\ldots\}\), \(X\) has finitely many degree-\(n\) covers, which we label by their monodromy maps \(\phi\) and denote by \(X_{\phi}\). For each \(n\), we put the uniform probability measure on the set of degree \(n\) covers of \(X\). We let \(\Delta_X\) (respectively \(\Delta_{X_{\phi}}\)) be the Laplacian on \(X\) (respectively \(X_{\phi}\)), with the sign convention that it is a non-negative operator on \(L^2\). Because \(X\) is connected and compact, \(\Delta_X\) has a pure point spectrum, \(\mathrm{spec}\left( \Delta_{X} \right)\), consisting of eigenvalues, written with multiplicity, satisfying \[ 0=\lambda_0(X) < \lambda_1(X)\leq \lambda_2(X) \leq \cdots \] and similarly for any degree-\(n\) cover \(X_{\phi}\). Moreover, any eigenfunction of \(\Delta_X\) lifts to an eigenfunction of \(\Delta_{X_{\phi}}\) with the same eigenvalue. Given this, we call an eigenvalue of \(\Delta_{X_{\phi}}\) new if it is not an eigenvalue of \(\Delta_X\), or if it appears with higher multiplicity (which is consistent with the fact that we listed our eigenvalues above with multiplicity). Inspired by analogous conjectures for random (covers of) graphs by Alon and Friedman (which have since been proven), the main theorem of the present paper is as follows. Let \(X\) be a compact, connected, hyperbolic surface, and pick \(\varepsilon>0\). Then the probability of the event \[ \mathrm{spec}\left( \Delta_{X_{\phi}}\right) \cap \left[ 0,\frac{3}{16}-\varepsilon \right] = \mathrm{spec}\left( \Delta_{X}\right) \cap \left[ 0,\frac{3}{16}-\varepsilon \right] \] (with the same multiplicities on both sides) goes to 1 as \(n\) (the degree of the random cover \(X_{\phi}\)) goes to infinity. That is, the probability of new eigenvalues below \(\frac{3}{16}-\varepsilon\) goes to 0. Further (based in part on the spectral gap of the hyperbolic plane and on Selberg's eigenvalue conjecture), the authors conjecture that the \(3/16\) can be improved to \(1/4\). The introduction gives a thorough overview of the development and and current state of the area, including the analogous results for random graphs and degree \(n\)-covers of graphs (as mentioned above), for other classes of surfaces (such as infinite area, convex, co-compact, hyperbolic surfaces and finite area, non-compact, hyperbolic surfaces), and for other probability measures on hyperbolic surfaces (namely the one induced by the Weil-Petersson volume form on the moduli space and a random model introduced by Brooks-Makover). Section 2 shows how the main theorem follows from the Selberg trace formula and an effective asymptotic formula for the expected number of fixed points of certain monodromy elements. The remaining three sections are devoted to proving this asymptotic formula, which requires significant effort.