Homogenization of the wave equation in composites with imperfect interface: a memory effect (Q871047)

The authors describe the asymptotic behaviour of the solution \( u_{\varepsilon }\) of the linear wave equation \(u_{\varepsilon }^{\prime \prime }- \text{div}(A^{\varepsilon }\nabla u_{\varepsilon })=f_{\varepsilon } \), posed in \(\Omega \times (0,T)\), where \(\Omega =\Omega _{1\varepsilon }\cup \Omega _{2\varepsilon }\) is a domain of \(\mathbb R^{n}\), \(\Omega _{1\varepsilon }\) being connected and \(\Omega _{2\varepsilon }\) being the union of disjoint inclusions of size \(\varepsilon \). Here \(A^{\varepsilon }(x)\) is equal to \(A(x/\varepsilon )\), where \(A\) is a periodic, bounded, symmetric and positive definite matrix. Homogeneous Dirichlet boundary conditions are imposed on the boundary \(\partial \Omega \times (0,T)\). The following boundary conditions are imposed on \(\Gamma ^{\varepsilon }\times (0,T)\), with \(\Gamma ^{\varepsilon }=\partial \Omega _{2\varepsilon }\): \( [A^{\varepsilon }\nabla u_{\varepsilon }]\cdot n_{1\varepsilon }=0\), \( A^{\varepsilon }\nabla u_{1\varepsilon }\cdot n_{1\varepsilon }=-\varepsilon ^{\gamma }h^{\varepsilon }[u_{\varepsilon }]\). Here \([\;]\) means the jump of the function across \(\Gamma ^{\varepsilon }\), \(n_{1\varepsilon }\) is the unit outward normal to \(\Omega _{1\varepsilon }\) and \(h^{\varepsilon }(x)\) is equal to \(h(x/\varepsilon )\) for some positive, bounded and periodic function \(h\). The initial data \(U_{\varepsilon }^{0}=u_{\varepsilon }(0)\) and \(U_{\varepsilon }^{1}=u_{\varepsilon }^{\prime }(0)\) are supposed to converge in some weak-\(L^{2}\) sense, together with the sources \( f_{\varepsilon }\). The authors prove the convergence of the solution \( u_{\varepsilon }\) to the solution \(u_{1}\) of the wave equation \( u_{1}^{\prime \prime }- \text{div}(A_{\gamma }^{0}\nabla u_{1})=f_{1}+f_{2}\), plus the homogeneous Dirichlet boundary conditions and some associated initial data. The matrix \(A_{\gamma }^{0}\) is proved to take different values according to the position of \(\gamma >-1\) with respect to 1. When \( \gamma =1\) a term associated to some memory effect appears in this limit matrix. The authors first prove estimates on the solution using some Galerkin method. They then prove the convergence results building appropriate test-functions.