Optimisation of the lowest Robin eigenvalue in the exterior of a compact set. II: Non-convex domains and higher dimensions

From MaRDI portal
(Redirected from Publication:2183754)



Abstract: We consider the problem of geometric optimisation of the lowest eigenvalue of the Laplacian in the exterior of a compact set in any dimension, subject to attractive Robin boundary conditions. As an improvement upon our previous work (arXiv:1608.04896, to appear in J. Convex Anal.), we show that under either a constraint of fixed perimeter or area, the maximiser within the class of exteriors of simply connected planar sets is always the exterior of a disk, without the need of convexity assumption. Moreover, we generalise the result to disconnected compact planar sets. Namely, we prove that under a constraint of fixed average value of the perimeter over all the connected components, the maximiser within the class of disconnected compact planar sets, consisting of finitely many simply connected components, is again a disk. In higher dimensions, we prove a completely new result that the lowest point in the spectrum is maximised by the exterior of a ball among all sets exterior to bounded convex sets satisfying a constraint on the integral of a dimensional power of the mean curvature of their boundaries. Furthermore, it follows that the critical coupling at which the lowest point in the spectrum becomes a discrete eigenvalue emerging from the essential spectrum is minimised under the same constraint by the critical coupling for the exterior of a ball.


Isoperimetric problems for the lowest eigenvalue of the Robin Laplacian on the exterior \(\Omega^{\mathrm{ext}}\) of a bounded, smooth open set \(\Omega \subset \mathbb{R}^d\), \(d \geq 2\), are studied. Denote by \(\lambda_1^\alpha (\Omega^{\mathrm{ext}})\) the minimum of the spectrum of the negative Laplacian in \(L^2 (\Omega^{\mathrm{ext}})\) subject to the boundary condition \[ \frac{\partial u}{\partial n} = \alpha u \quad \text{on}~\partial \Omega, \] where the Robin parameter \(\alpha < 0\) is a constant and \(\frac{\partial}{\partial n}\) denotes the derivative with respect to the outer unit normal vector to \(\Omega\) (i.e.\ the normal pointing inside \(\Omega^{\mathrm{ext}}\)); in dimension \(d = 2\), \(\lambda_1^\alpha (\Omega^{\mathrm{ext}})\) is always a discrete, negative eigenvalue, while for \(d \geq 3\) this is true for all \(\alpha\) below a certain threshold. In the first main result of this article, it is shown for \(d = 2\), fixed \(\alpha < 0\) and fixed \(c > 0\) that \[ \max_{\frac{|\partial \Omega|}{N_\Omega} = c} \lambda_1^\alpha (\Omega^{\mathrm{ext}}) = \lambda_1^\alpha (B^{\mathrm{ext}}), \] where the maximum is taken over all smooth, bounded open sets \(\Omega\) consisting of a finite number of connected components (the latter number denoted by \(N_\Omega\)) such that \(\frac{|\partial \Omega|}{N_\Omega} = c\) and \(B\) is the disk with perimeter \(c\). This improves upon an earlier result by the same authors where only convex \(\Omega\) where allowed. The second main result concerns the higher-dimensional case \(d \geq 3\); here convexity of \(\Omega\) is required. With the notation \[ \mathcal{M} (\partial \Omega) := \frac{1}{|\partial \Omega|} \int_{\partial \Omega} \left( \frac{\kappa_1 + \dots + \kappa_{d - 1}}{d - 1} \right)^{d - 1}, \] where \(\kappa_1, \dots, \kappa_{d - 1}\) denote the principle curvatures of \(\partial \Omega\), the authors prove that, for each \(\alpha < 0\) and \(c > 0\), \[ \max_{\mathcal{M} (\partial \Omega) = c} \lambda_1^\alpha (\Omega^{\mathrm{ext}}) = \lambda_1^\alpha (B^{\mathrm{ext}}), \] where the maximum is taken over all convex, smooth, bounded open sets \(\Omega\) such that \(\mathcal{M} (\partial \Omega) = c\), and \(B\) is the ball with \(\mathcal{M} (\partial B) = c\). For Part I, see [the authors, J. Convex Anal. 25, No. 1, 319--337 (2018; Zbl 1401.35223)].



Cites work


Cited in
(28)








This page was built for publication: Optimisation of the lowest Robin eigenvalue in the exterior of a compact set. II: Non-convex domains and higher dimensions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2183754)