TWO ROADS TO 1/4: ARITHMETIC RIGIDITY VERSUS TYPICAL GENERICITY IN THE SPECTRAL GAP PROBLEM FOR HYPERBOLIC SURFACES

Abstract. The spectral gap λ1 of a hyperbolic surface – the first non-zero eigenvalue of its Laplacian – is bounded above by 1/4 in the large-genus limit, by a classical theorem of Huber. Two largely independent research programs have tried to determine when this bound is actually approached. The first, going back to Selberg’s 1965 theorem λ1 ≥ 3/16 for congruence subgroups of SL2(Z), is tied to deep instances of the Langlands functoriality conjecture, and has not moved past Kim and Sarnak’s bound λ1 ≥ 975/4096 ≈ 0.238 since 2003. The second, concerning typical hyperbolic surfaces – sampled from Mirzakhani’s Weil–Petersson measure, or built as random covers of a fixed surface – has, remarkably, been fully resolved only very recently: after Magee–Naud–Puder, Wu– Xue and Lipnowski–Wright matched Selberg’s 3/16 bound for generic surfaces, Hide–Magee and then a multi-year program of Anantharaman and Monk (with an independent, quantitative route by Hide, Macera and Thomas) pushed this all the way to the optimal 1/4−ϵ, in direct imitation of Friedman’s 2003 proof of Alon’s second eigenvalue conjecture for random regular graphs. We survey both roads, explain why the second program’s machinery does not transfer to the first, and propose – as an explicitly unproven working hypothesis – a hybrid strategy for the arithmetic problem based on measuring the “distance to genericity” of a congruence cover inside the much larger space of covers whose typical member is now known to be near-optimal.