Spectral rigidity of Liouville tori
Joscha Henheik, Vadim Kaloshin, Yunzhe Li, Amir Vig
TL;DR
The paper proves spectral rigidity for Liouville metrics on the 2-torus: any Laplace isospectral deformation within the conformal class ds^2_ε = (1 + f1(x1) + f2(x2) + εU(x1,x2))(dx1^2 + dx2^2) with linear ε-perturbations must have U ≡ 0. The authors combine a microlocal trace analysis, which shows isospectrality preserves rational invariant tori (hence the rotational length spectrum) with a second-variation argument that, given a preserved rational torus, forces the perturbation U to vanish along the corresponding geodesics, and then everywhere by foliation. The key innovations include a Hadamard-Riesz type parametrix for off-diagonal wave propagation, a Marvizi-Melrose style trace formula localized to horizontal tori, and the use of loop functions to connect the length spectrum and geodesic dynamics. Together these yield a two-step rigidity result: isospectrality implies rational integrability, and preserving a single rational torus under a linear perturbation implies trivial deformation, establishing spectral rigidity in this integrable closed-manifold setting with potential implications for broader inverse spectral problems in integrable dynamics.
Abstract
We show that Laplace isospectral deformations within a conformal class of generic Liouville metrics on the two-dimensional torus that are linear in the deformation parameter are necessarily trivial. Two of the main ingredients in our proof are a noncancellation result for the wave trace and an analysis of the second order variational formula for the energy functional associated to closed geodesics. Noncancellation allows us to detect parts of the length spectrum from the Laplace spectrum and conclude rational integrability for the deformed geodesic flow (Liouville metrics are folklorically conjectured to be the only Riemannian metrics with integrable geodesic flow on the torus). We then use the second variational formula to show how the preservation of a single rational torus is sufficient to conclude triviality of the deformation, assuming linearity. We also present some evidence that our hypothesis of linearity may indeed be necessary.
