Local limits of high energy eigenfunctions on integrable billiards

TL;DR

This paper investigates whether high energy eigenfunctions in integrable billiards resemble random waves locally, finding that most integrable polygons and ellipses do not exhibit this inverse localization property, contrasting with chaotic systems.

Contribution

It provides the first rigorous analysis of local eigenfunction limits in integrable billiards, showing that inverse localization generally fails beyond flat tori.

Findings

Many integrable polygons show good inverse localization

Most integrable polygons and ellipses do not exhibit inverse localization

Local eigenfunction limits differ from random waves in generic integrable billiards

Abstract

Berry's random wave conjecture posits that high energy eigenfunctions of chaotic systems resemble random monochromatic waves at the Planck scale. One important consequence is that, at the Planck scale around "many" points in the manifold, any solution to the Helmholtz equation $\Delta\varphi+\varphi =0$ can be approximated by high energy eigenfunctions. This property, sometimes called inverse localization, has useful applications to the study of the nodal sets of eigenfunctions. Alas, the only manifold for which the local limits of a sequence of high energy eigenfunctions are rigorously known to be given by random waves is the flat torus $(\mathbf{R}/\mathbf{Z})^2$, which is certainly not chaotic. Our objective in this paper is to study the validity of this "inverse localization" property in the class of integrable billiards, exploiting the fact that integrable polygonal billiards are classified and that Birkhoff conjectured that ellipses are the only smooth integrable billiards. Our main results show that, while there are infinitely many integrable polygons exhibiting good inverse localization properties, for "most" integrable polygons and ellipses, this property fails dramatically. We thus conclude that, in a generic integrable billiard, the local limits of Dirichlet and Neumann eigenfunctions do not match random waves, as one might expect in view of Berry's conjecture. Extensions to higher dimensions and nearly integrable polygons are discussed too.