[Paper Review] Concentration of eigenfunctions of the Laplacian on a closed Riemannian manifold

Kei Funano, Yohei Sakurai|arXiv (Cornell University)|Dec 25, 2017

Advanced Mathematical Modeling in Engineering19 references3 citations

TL;DR

This paper establishes exponential concentration of eigenfunctions of the Laplacian on closed Riemannian manifolds around their nodal sets, proving that the volume measure concentrates exponentially near the zero set as eigenvalues grow. Under a lower Ricci curvature bound, it derives restricted exponential concentration inequalities and refined $L_p$ moment estimates, showing that eigenfunctions decay rapidly outside large subsets, improving upon classical Chebyshev-type bounds and Sogge's global estimates.

ABSTRACT

We study concentration phenomena of eigenfunctions of the Laplacian on closed Riemannian manifolds. We prove that the volume measure of a closed manifold concentrates around nodal sets of eigenfunctions exponentially. Applying the method of Colding and Minicozzi we also prove restricted exponential concentration inequalities and restricted Sogge-type $L_p$ moment estimates of eigenfunctions.

Motivation & Objective

• To investigate the global concentration behavior of eigenfunctions of the Laplacian on closed Riemannian manifolds.
• To quantify how much the volume measure concentrates near the nodal set of eigenfunctions as eigenvalues increase.
• To establish refined $L_p$ moment estimates for eigenfunctions restricted to large subsets, improving upon classical bounds.
• To extend results from the smooth setting to include curvature-dependent concentration inequalities under Ricci curvature lower bounds.
• To provide exponential decay estimates for the measure of superlevel sets of eigenfunctions on large subsets, refining Chebyshev-type bounds.

Proposed method

• Utilizes the method of Colding and Minicozzi to derive exponential concentration inequalities for eigenfunctions on large subsets of the manifold.
• Applies a key concentration inequality around the boundary of domains with Dirichlet eigenvalues, adapted to nodal sets via geometric and analytic estimates.
• Employs the Cauchy-Schwarz inequality and $L^2$-norm control to bound the gradient and pointwise values of eigenfunctions near nodal sets.
• Derives a restricted exponential concentration inequality by combining volume concentration near nodal sets with curvature-dependent spectral estimates.
• Applies the Cavalieri principle and change of variables to transform measure bounds into $L_p$ moment estimates.
• Uses the gamma function and Stirling's approximation to bound the $L_p$-norm of eigenfunctions restricted to large subsets.

Experimental results

Research questions

• RQ1 How rapidly does the volume measure of a closed Riemannian manifold concentrate around the nodal set of an eigenfunction as the eigenvalue increases?
• RQ2 Can exponential concentration inequalities be established for eigenfunctions restricted to large subsets under Ricci curvature lower bounds?
• RQ3 To what extent do $L_p$ moment estimates of eigenfunctions improve when restricted to large subsets, compared to global Sogge-type bounds?
• RQ4 How does the decay rate of superlevel sets of eigenfunctions depend on the $L^2$-norm and subset measure under curvature constraints?
• RQ5 Can the concentration near nodal sets be quantified using geometric and spectral techniques, particularly in relation to the first Dirichlet eigenvalue?

Key findings

• The volume measure of a closed Riemannian manifold concentrates exponentially around the nodal set: $ m_g(M \setminus B_r(\varphi_\lambda^{-1}(0))) \leq \exp(1 - \sqrt{\lambda}\,r) $ for all $ r > 0 $.
• Under a lower Ricci curvature bound $ \operatorname{Ric}_g \geq -(n-1) $, for $ \lambda \geq \mathcal{C}_n $, there exists a subset $ \Omega \subset M $ with $ m_g(\Omega) \geq 1 - \xi $ such that $ m_g(\Omega \cap \{ |\varphi_\lambda| > r \}) \leq \exp\left(1 - \frac{C_n \sqrt{\xi}}{\|\varphi_\lambda\|_2} r\right) $.
• The restricted $ L_p $ moment estimate satisfies $ \left( \int_\Omega |\varphi_\lambda|^p \, dm_g \right)^{1/p} \leq e \, \Gamma(p+1)^{1/p} \, \frac{\|\varphi_\lambda\|_2}{C_n \sqrt{\xi}} $, which improves upon the Chebyshev bound and removes $ \lambda $-dependence found in Sogge's global estimate.
• The decay rate of the superlevel set measure is exponential in $ r $, with a rate proportional to $ \sqrt{\xi}/\|\varphi_\lambda\|_2 $, indicating strong concentration on large subsets.
• The constant $ C_n $ in the bounds depends only on the dimension $ n $, and the results are sharp in the sense that the $ \Gamma(p+1)^{1/p} $ factor asymptotically matches $ p $ via Stirling's formula.
• The results refine both classical $ L^2 $-based inequalities and Sogge's global $ L_p $ estimates by localizing to large subsets and achieving exponential decay in the tail bounds.