The maximization of the first eigenvalue for a two-phase material (Q2152589)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The maximization of the first eigenvalue for a two-phase material |
scientific article |
Statements
The maximization of the first eigenvalue for a two-phase material (English)
0 references
8 July 2022
0 references
In the present article the author addresses the maximisation problem for the first Dirichlet eigenvalue of a composite of a two-phase material. More precisely, given \(\Omega\subseteq \mathbb{R}^N\) open, bounded, \(N\geq 2\). Let \(0<\alpha<\beta\) and \(0<\kappa<|\Omega|\), \(|\Omega|\) the Lebesgue measure of \(\Omega\). For \(\omega\subseteq \Omega\), \(|\omega|\geq \kappa\), consider the Rayleigh-coefficient \[ \lambda_1(\omega) := \min_{u\in H_0^1(\Omega), u\neq 0} \frac{\int_{\Omega}(\alpha 1_{\omega}+\beta 1_{\Omega\setminus\omega})|\nabla u|^2}{\|u\|_{L_2(\Omega)}^2}. \] Then the main question is to find \(\omega\) satisfying the above conditions such that \(\lambda_1(\omega)\) is maximised. As similar problems are known to be ill-posed in general, a relaxation is considered, where the maximisation is not run over characteristic functions but rather over \(L_\infty\). That is to say, the author rather considers the relaxed problem \[ \max_{\theta} \min_{u} \int_{\Omega}(\alpha \theta+\beta (1-\theta))|\nabla u|^2 / (\|u\|_{L_2(\Omega)}^2) \] where the maximum runs over \(\theta \in L_\infty(\Omega;[0,1])\) subject to \(\int_\Omega \theta\geq \kappa\) and the minimum runs over all non-zero \(u\in H_0^1(\Omega)\). Restricting to \(u\) non-negative functions from the unit sphere of \(L_2(\Omega)\), the author shows that it is possible to interchange max and min in the relaxed formulation. Moreover, the author establishes a uniqueness result for \(u\), where max and min above are interchanged. The eigenvalue function, that is, \(\theta\mapsto \min_{u} \int_{\Omega}(\alpha \theta+\beta (1-\theta))|\nabla u|^2 / (\|u\|_{L_2(\Omega)}^2)\) is shown to be concave. Finally, for \(\Omega\) smooth, it is shown that the unrelaxed problem has no solution \(\omega\subseteq \Omega\); a posteriori justifying the relaxation considered in the first place. The theoretical results just mentioned are complemented by some numerical results and experiments on the ball and the square.
0 references
two-phase material
0 references
eigenvalue
0 references
optimal design
0 references
relaxed formulation
0 references
gradient method
0 references
0 references
0 references
0 references