arXiv · 2602.04730
Semiclassical Structure of the Advection--Diffusion Spectrum in Mixed Phase Spaces
Abstract
We examine the spectral structure of the two-dimensional advection-diffusion operator in flows with mixed phase space at very large Peclet number. Using Fourier discretization combined with symmetry reduction and Krylov-Arnoldi methods, we compute on the order of one hundred leading eigenpairs reliably in the asymptotic, weak-diffusion regime. While the principal eigenvalue is asymptotically diffusive and localized on the largest regular region, the broader spectrum exhibits a rich organization controlled by local Lagrangian phase-space geometry. In particular, exponential mixing in chaotic regions rapidly suppresses correlations, whereas algebraic mixing in integrable regions generates long-lived coherent structures that dominate the slow and intermediate parts of the spectrum. We identify three distinct classes of eigenmodes: advective modes associated with transport on invariant tori, diffusive modes and, within the duffusive branch, tunneling modes arising from weak coupling between dynamically separated regular regions. Drawing on a semiclassical analogy, we assign quantum-number-like labels to these families and predict the appearance, scaling, and ordering of their sub-spectra directly from the Hamiltonian phase-space structure. The coexistence of these families implies that no uniform control of the spectral gap exists across the full spectrum: although the slowest mode is diffusive, arbitrarily small gaps arise between competing families at higher mode numbers. As a result, finite-time advection-diffusion dynamics is generically governed by persistent modal competition rather than single-mode dominance, even at asymptotically large Peclet number.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Christopher Amey, Bala Sundaram, Andrew C. Poje. 2026-02-04. Semiclassical Structure of the Advection--Diffusion Spectrum in Mixed Phase Spaces. https://arxiv.org/abs/2602.04730
Cite the original work for its findings. Save a collection to share your selection of sources.