Skip to main content
QUICK REVIEW

[Paper Review] Localized Eigenfunctions: Here You See Them, There You Don't

Steven Heilman, Robert S. Strichartz|ArXiv.org|Sep 4, 2009
Quantum chaos and dynamical systems24 references20 citations
TL;DR

This paper presents numerically computed, highly localized Neumann eigenfunctions on planar domains with symmetry-induced localization at low eigenvalues. Using finite element methods on symmetric 'rooms and passages' domains like the 'smiley' and 'cow' shapes, the authors demonstrate that eigenfunctions can be strongly localized in one region despite being global solutions, with $L^2$ and $L^∞$ norms off the main region decaying as power laws in the connection width $h$, revealing unexpected localization without high-frequency asymptotics.

ABSTRACT

This expository note explores Laplacian eigenfunction localization for compact domains. We work in the context of a particular numerically determined, localized, low frequency eigenfunction.

Motivation & Objective

  • To investigate the existence and nature of localized eigenfunctions for the Laplacian on planar domains with low eigenvalues, challenging the assumption that localization only occurs at high frequencies.
  • To demonstrate that symmetry—specifically skew-symmetric eigenfunctions on a symmetric subdomain—can lead to strong spatial localization in the eigenfunction's support.
  • To provide numerical evidence of localization using finite element methods on domains like the 'smiley' and 'cow' shapes with narrow connecting passages.
  • To quantify the degree of localization via $L^2$ and $L^\infty$ norms on the complement of the primary domain, showing power-law scaling with passage height $h$.
  • To challenge the prevailing view that eigenfunction localization is exclusive to high-frequency or chaotic systems, showing it can occur robustly at low frequencies via geometric and symmetry constraints.

Proposed method

  • The study uses the finite element method in MATLAB to numerically compute Neumann eigenfunctions on planar domains with two symmetric rooms connected by a narrow passage.
  • Domains are constructed such that one room $\Omega_1$ has a line of symmetry $L$, and eigenfunctions on $\Omega_1$ are chosen to be skew-symmetric with respect to reflection across $L$.
  • Such skew-symmetric eigenfunctions vanish along $L$, and if $L$ intersects the boundary at a corner, the function and its gradient vanish at that point, reducing amplitude near the junction.
  • The localization is quantified by measuring the $L^2$ and $L^\infty$ norms of the eigenfunction on $\Omega \setminus \Omega_1$, the complement of the primary domain.
  • Log-log plots of these norms versus the passage height $h$ are used to infer power-law scaling relationships, with best-fit power laws extracted from numerical data.
  • The analysis focuses on eigenfunctions with low to moderate eigenvalues, contrasting with typical high-frequency localization phenomena in quantum chaos.

Experimental results

Research questions

  • RQ1Can eigenfunctions of the Laplacian be significantly localized in space at low eigenvalues due to geometric and symmetry constraints?
  • RQ2To what extent does the presence of a symmetric subdomain with skew-symmetric eigenfunctions lead to localization in the full domain?
  • RQ3How does the degree of localization scale with the geometric parameter $h$, the height of the connecting passage between two rooms?
  • RQ4Are there measurable power-law relationships between localization norms ($L^2$ and $L^\infty$) and the passage width $h$ in symmetric domains?
  • RQ5Why do such localized eigenfunctions not appear in generic domains, and what structural features (e.g., symmetry, corner singularities) are essential?

Key findings

  • For the 'smiley' domain, the $L^2$ norm of the fifth eigenfunction on $\Omega \setminus \Omega_1$ scales as $y = 11.254x^{3.9087}$ with respect to passage height $h$, indicating strong localization.
  • In the same domain, the $L^\infty$ norm of the fifth eigenfunction decays as $y = 4.1735x^{3.2959}$, further confirming uniform localization.
  • For the twelfth eigenfunction in the 'smiley' domain, the $L^2$ norm on the complement scales as $y = 249.06x^{2.4636}$, showing a weaker but still significant power-law decay.
  • In the 'cow' domain, the $L^2$ norm of the fourth eigenfunction on the complement decays as $y = 119.65x^{3.0889}$, indicating strong localization.
  • The $L^\infty$ norm of the eleventh eigenfunction in the 'cow' domain decays as $y = 676.08x^{2.5700}$, demonstrating uniform localization that is highly sensitive to passage width.
  • The power-law exponents for localization vary significantly across different eigenfunctions and domains, indicating no universal scaling law, but consistent decay trends across all tested cases.

Better researchstarts right now

From paper design to paper writing, dramatically reduce your research time.

No credit card · Free plan available

This review was created by AI and reviewed by human editors.