Asymptotic behaviour of cuboids optimising Laplacian eigenvalues (Q1693576): Difference between revisions

This paper is a continuation of [\textit{M. van den Berg} and \textit{K. Gittins}, Mathematika 63, No. 2, 469--482 (2017; Zbl 1378.35091)]. For an open set \(\Omega\subset \mathbb{R}^n\) with finite Lebesgue measure, denote by \(\{\lambda_k(\Omega)\}_{k=1}^{\infty}\) (\(\lambda_1(\Omega)>0\)) the non-decreasing sequence of eigenvalues of the Dirichlet Laplace operator \(-\Delta\) in \(L^2(\Omega)\), and by \(\{\mu_k(\Omega)\}_{k=1}^{\infty}\) (\(\mu_1(\Omega)=0\)) denote the non-decreasing sequence of eigenvalues of the Neumann Laplace operator \(-\Delta\) in \(L^2(\Omega)\), in the case when \(\partial\Omega\) is Lipschitz regular. For \(a_1\leq \cdots \leq a_n\) denote by \(R=R_{a_1,\dots,a_n}\) an \(n\)-dimensional cuboid of unit measure, and by \(Q\) denote the \(n\)-dimensional unit cube. For the cuboid \(R^D_{k}\) which minimises \(\lambda_k\), denote by \(a^{\star}_{1,k},\dots, a^{\star}_{n,k}\) its side-lengths, and, for the cuboid \(R^N_{k}\) which maximises \(\mu_k\), also denote by \(a^{\star}_{1,k},\dots, a^{\star}_{n,k}\) its side-lengths. For \(n\geq 2\), let \(\theta_n\) be an exponent such that, for each compact subset \(K\) of \(\mathbb{R}_{+}\), uniformly in \(a_1,\dots,a_n \in K\), \[ \sharp\{z\in Z^n;\,a_1^{-2}\,z_1^2+\cdots+a_n^{-2}\,z_n^2\leq t^2\}-\omega_n\,t^n\,\prod\limits_{i=1}^n\,a_i=O(t^{\theta_n}),\quad\text{as}\quad t\to \infty. \] The authors prove that \(a^{\star}_{n,k}=1+O\big(k^{(\theta_n-(n-1))/(2\,n)}\big)\) both for \(R^D_k\) and \(R^N_k\). This means that each sequence of minimising cuboids \(\big(R^D_k\big)\) for \(\lambda_k\) converges to the \(n\)-dimensional cube, as \(k\to \infty\), and that any sequence of maximising cuboids \(\big(R^N_k\big)\) for \(\mu_k\) also converges to the \(n\)-dimensional cube, as \(k\to \infty\). In particular, this proves that \(\displaystyle |\lambda(Q)-\lambda_k(R^D_k)|=O\big(k^{(\theta_n-(n-1))/(2\,n)}\big)\) and \(\displaystyle |\mu(Q)-\mu_k(R^D_k)|=O\big(k^{(\theta_n-(n-1))/(2\,n)}\big)\), as \(k\to \infty\).