Minimization of the \(k\)-th eigenvalue of the Robin-Laplacian (Q2420442)

The Robin eigenvalue problem for the Laplace operator reads \[ \begin{cases} -\Delta u=\lambda u & \text{in }\Omega,\\ \partial_\nu u=\beta u & \text{on }\partial\Omega, \end{cases} \] and by classical spectral theory it is known that for any \(\beta\in\mathbb{R}\) the spectrum is bounded from below and consists of eigenvalues of finite multiplicity diverging to plus infinity. An interesting question that has gained a lot of attention in the last decades is that of optimizing the \(k\)-th eigenvalue under some type of constraint, and in particular if the volume of the domain \(|\Omega|\) is fixed: \[ \max_{|\Omega|=c}\lambda_k(\Omega)\quad\text{ or }\quad \min_{|\Omega|=c}\lambda_k(\Omega). \] These types of question have been extensively studied for eigenvalues of the Laplacian with a lot of interesting results (see [\textit{A. Henrot}, Extremum problems for eigenvalues of elliptic operators. Basel: Birkhäuser (2006; Zbl 1109.35081)]). The present paper deals with the issue of existence of minimizers for the Robin problem when \(\beta>0\). Inspired by the celebrated paper [\textit{G. Buttazzo} and \textit{G. Dal Maso}, Arch. Ration. Mech. Anal. 122, No. 2, 183--195 (1993; Zbl 0811.49028)] (which dealt with the Dirichlet Laplacian) they devise a suitable relaxation of the minimization problem which is then shown to admit (at least) a solution. Then the question of existence for the original problem is translated into a regularity issue for the relaxed one. As the standard way of considering quasi-open set is not effective in the Robin case, the authors provide a new free discontinuity formulation which they proved to be a generalization in [Arch. Ration. Mech. Anal. 198, No. 3, 927--961 (2010; Zbl 1228.49049)]. Then, the proof of existence of minimizers for the relaxed problem is heavily technical, involving many properties of functions of bounded variation as well as ideas coming form the theory of free discontinuity functionals. In particular, it is interesting to see that, even though the Dirichlet and the Robin problem for the Laplacian are strictly linked, the relaxed problems and the corresponding techniques used to prove existence of minimizers are radically different.