Abstract:We provide an answer to a question raised by Levine and Weinberger in their $1986$ paper concerning the difference between Dirichlet and Neumann eigenvalues of the Laplacian on bounded domains in $\R^{n}$. More precisely, we show that for a certain class of domains there exists a sequence $p(k)$ such that $\lambda_{k}\geq \mu_{k+ p(k)}$ for sufficiently large $k$. This sequence, which is given explicitly, grows with $k^{1-1/n}$ as $k$ goes to infinity, which we conjecture to be optimal, and may be chosen independently of the domain. We also prove the existence of a sequence, now not given explicitly and only of order $k^{1-3/n}$ but valid for bounded Lipschitz domains in $\R^{n} (n\geq4)$, for which a similar inequality holds for all $k$. From these results and the analysis of some particular examples we formulate a conjecture for general Euclidean domains.
What problem does this paper attempt to address?
This paper mainly discusses the gap between the eigenvalues of the Dirichlet and Neumann Laplace operators on finite domains. The author attempts to answer the question raised by Levine and Weinberger in 1986, whether it is possible to improve the inequality λk > µϕ(n, k), where λk and µk are the Dirichlet and Neumann eigenvalues, and ϕ(n, k) is an increasing sequence related to k.
The author first shows that for a specific category of regions that satisfy non-periodic conditions, there exists an exponential function p(k) = k^1-1/n, such that λk ≥ µk + p(k) holds for sufficiently large k. Here, p(k) is independent of the region and grows at a rate of k^1-1/n as k increases. In addition, it is also proved that for n-dimensional Lipschitz domains (n>3), there exists a sequence p(k) such that λk ≥ µk + p(k) holds for all k, but the asymptotic behavior of p(k) is slightly worse and depends on an unspecified constant.
Two conjectures are also proposed in the paper: firstly, for general Euclidean domains, there exists an optimal exponent of 1-1/n such that λk ≥ µk + p(k) holds for all k; secondly, for the two-dimensional case, an inequality related to the perimeter and area of rectangles is proposed.
The author further supports these results by studying the particular case of two-dimensional rectangles and points out that this type of inequality may not hold or even reverse for non-Euclidean backgrounds such as spherical surfaces.
In summary, the core issue of the paper is the gap between the Dirichlet and Neumann eigenvalues, and an improved form of inequality is provided, along with conjectures about general Euclidean domains.