Sharp-interface limit of a multi-phase spectral shape optimization problem for elastic structures (Q6139958)

The authors consider a bounded and Lipschitz domain \(\Omega \subset \mathbb{R }^{d}\), whose boundary is split into two disjoint parts: \(\Gamma _{D}\) with strictly positive \((d-1)\)-dimensional Hausdorff measure and \(\Gamma _{0}\). They consider the eigenvalue problem: \(-\nabla \cdot \lbrack \mathbb{C} (\varphi )\mathcal{E}(w^{\varepsilon ,\varphi })]=\lambda ^{\varepsilon ,\varphi }\rho (\varphi )w^{\varepsilon ,\varphi }\), posed in \(\Omega \), with the boundary conditions \(w^{\varepsilon ,\varphi }=0\) on \(\Gamma _{D}\), \([\mathbb{C}(\varphi )\mathcal{E}(w^{\varepsilon ,\varphi })]n=0\), on \( \Gamma _{0}\). The elasticity tensor \(\mathbb{C}:\mathbb{R}^{N}\rightarrow \mathbb{R}^{d\times d\times d\times d}\) is a fourth-order tensor such that \( \mathbb{C}_{ijkl}\in C_{\mathrm{loc}}^{1,1}(\mathbb{R}^{N};\mathbb{R})\), \(\mathbb{C} _{ijkl}=\mathbb{C}_{jikl}=\mathbb{C}_{ijlk}=\mathbb{C}_{klij}\), for \( i,j,k,l=1,\ldots ,d\), \(\mathbb{C}\) is coercive: for any fixed \(\varepsilon >0 \), there exists \(\theta _{\varepsilon }>0\) such that \(\mathbb{C}(\varphi ) \mathcal{B}:\mathcal{B}\geq \theta _{\varepsilon }\left\vert \mathcal{B} \right\vert ^{2}\), for all \(\varphi \in \mathbb{R}^{N}\) and all symmetric matrices \(\mathcal{B}\in \mathbb{R}^{d\times d}\). The strain tensor \( \mathcal{E}\) of a vector-valued function \(u\in H_{D}^{1}(\Omega ;\mathbb{R} ^{d})\) is given as \(\mathcal{E}(u)=\frac{1}{2}(\nabla u+\nabla u^{T})\). The density function \(\rho :\mathbb{R}^{N}\rightarrow \mathbb{R}\) belongs to \( C_{\mathrm{loc}}^{1,1}(\mathbb{R}^{N};\mathbb{R})\) and is uniformly positive: for any fixed \(\varepsilon >0\), there is a constant \(\rho _{0,\varepsilon }>0\) such that \(\rho (\varphi )\geq \rho _{0,\varepsilon }\) for all \(\varphi \in \mathbb{R}^{N}\). The authors finally introduce the minimization problem: \( min_{\varphi \in \mathcal{G}^{m}}J_{l}^{\varepsilon }(\varphi )\), where \( \mathcal{G}^{m}=\{\varphi \in H^{1}(\Omega ;\mathbb{R}^{N})\mid \varphi (x)\in G=\mathbb{R}_{+}^{N}\cap \{\xi \in \mathbb{R}^{N}\mid \sum_{i=1}^{N}\xi _{i}=1\}\) for almost all \(x\in \Omega \), \(\frac{1}{ \int_{\Omega }dx}\int_{\Omega }\varphi (x)dx=1\}\), \(J_{l}^{\varepsilon }(\varphi )=\Psi (\lambda _{n_{1}}^{\varepsilon ,\varphi },\ldots ,\lambda _{n_{l}}^{\varepsilon ,\varphi })+\gamma E^{\varepsilon }(\varphi )\), \( E^{\varepsilon }\) being the Ginzburg-Landau energy: \(E^{\varepsilon }(\varphi )=\int_{\Omega }(\frac{\varepsilon }{2}\left\vert \nabla \varphi \right\vert ^{2}+\frac{1}{\varepsilon }\psi (\varphi ))dx\), the potential \( \psi \) enforcing the phase-field \(\varphi \) to attain its values only in the Gibbs simplex \(\{\xi \in \mathbb{R}^{N}\mid \sum_{i=1}^{N}\xi _{i}=1\}\), and \(\gamma >0\) is a fixed constant related to surface tension. The authors first recall the optimality conditions they proved in [\textit{H. Garcke} et al., ESAIM, Control Optim. Calc. Var. 29, Paper No. 10, 57 p. (2023; Zbl 1512.35411)], as a gradient inequality. The first main result proves the existence of a solution \( \varphi \in H^{2}(\Omega ;\mathbb{R}^{N})\) to this gradient inequality, under a suitable regularity assumption on the eigenfunctions involved in the minimization problem. For the proof, the authors introduce a regularization problem, regularizing the potential \(\psi \) and rewriting the constraints. They prove the existence of a solution to this regularized problem in \( H^{2}(\Omega ;\mathbb{R}^{N})\). They prove uniform estimates on this solution with respect to the regularizing parameter, which allow passing to the limit. They then build inner and outer asymptotic expansions to produce a state equation and a gradient equality in the sharp-interface limit. They compare the leading order terms of the inner and outer asymptotic expansions in the state equation and in the gradient equality. They state the complete sharp-interface problem, passing to the limit in the state equation and in the first-order optimality condition. They specialize the sharp-interface optimality system in the case of only one material. They compare their results to that obtained by \textit{G. Allaire} and \textit{F. Jouve} in [Comput. Methods Appl. Mech. Eng. 194, No. 30-33, 3269--3290 (2005; Zbl 1091.74038)] using shape calculus. The paper ends with the presentation of numerical results.