Asymptotic behaviour of cuboids optimising Laplacian eigenvalues

TL;DR

This paper proves that as the eigenvalue index increases, the optimal cuboids for Laplacian eigenvalues in any dimension tend to become perfect cubes, with additional stability and shape optimization results for eigenvalue means.

Contribution

It extends known results to higher dimensions, showing asymptotic convergence of optimal cuboids to the cube and providing stability and shape optimization insights.

Findings

Optimal cuboids converge to the cube as eigenvalue index increases

Stability results for eigenvalues as index tends to infinity

Shape optimization for Riesz means of eigenvalues

Abstract

We prove that in dimension $n \geq 2$, within the collection of unit measure cuboids in $\mathbb{R}^n$ (i.e. domains of the form $\prod_{i=1}^{n}(0, a_n)$), any sequence of minimising domains $R_k^\mathcal{D}$ for the Dirichlet eigenvalues $\lambda_k$ converges to the unit cube as $k \to \infty$. Correspondingly we also prove that any sequence of maximising domains $R_k^\mathcal{N}$ for the Neumann eigenvalues $\mu_k$ within the same collection of domains converges to the unit cube as $k\to \infty$. For $n=2$ this result was obtained by Antunes and Freitas in the case of Dirichlet eigenvalues and van den Berg, Bucur and Gittins for the Neumann eigenvalues. The Dirichlet case for $n=3$ was recently treated by van den Berg and Gittins. In addition we obtain stability results for the optimal eigenvalues as $k \to \infty$. We also obtain corresponding shape optimisation results for the Riesz means of eigenvalues in the same collection of cuboids. For the Dirichlet case this allows us to address the shape optimisation of the average of the first $k$ eigenvalues.