Two-scale convergence: obviousness of the choice of test functions? Not always (Q1713584)

The author proves on an example that the classical test-functions, which are used in homogenization theory, do not lead to the desired result. Let \(\Omega \) be a bounded and smooth domain in \(\mathbb{R}^{3}\), \(Y=(0,1)^{3}\), \(Y_{s}\) an open subset of \(Y\) and \(Y_{f}=Y\setminus \overline{Y}_{s}\). The author considers \(\Omega _{s}^{\varepsilon }=\Omega \cap \varepsilon \widetilde{Y} _{s}\) and \(\Omega _{f}^{\varepsilon }=\Omega \cap \varepsilon \widetilde{Y} _{f}\) and the variational formulation of a weakly damped wave equation \[ \int_{\Omega }\rho ^{\varepsilon }\frac{\partial ^{2}u_{\varepsilon }}{ \partial t^{2}}\cdot vdx+\varepsilon ^{2}\int_{\Omega _{f}^{\varepsilon }}b_{ijkh}\frac{\partial }{\partial t}\frac{\partial (u_{\varepsilon })^{k}}{ \partial x_{h}}\frac{\partial v^{i}}{\partial x_{j}}dx+\vartheta \int_{\Omega _{f}^{\varepsilon }}\mathrm{div} (u_{\varepsilon }) \mathrm{div} (v)dx=\int_{\Omega }f\cdot dx \] for every \(v\in V_{\varepsilon }=\{v\in H_{0}^{1}(\Omega ;\mathbb{R}^{3}):\frac{1}{2}(\frac{\partial v^{i}}{\partial x_{j}}+\frac{\partial v^{j}}{\partial x_{i}})=0\) in \(\Omega _{s}^{\varepsilon }\}\). The initial conditions \(u_{\varepsilon }(0)=0=\frac{ \partial u_{\varepsilon }}{\partial t}(0)\) are imposed. The author presents the functional framework adapted to this problem from which he derives uniform estimates on its solution. He then observes that the classical test-function \(\phi _{\varepsilon }(x)=\psi _{0}(x)+\varepsilon \psi _{1}(x, \frac{x}{\varepsilon })\) with \(\psi _{0}\in (\mathcal{D}(\Omega ))^{3}\) and \( \psi _{1}\in (\mathcal{D}(\Omega ))^{3}\otimes (C_{per}^{1}(Y))^{3}\) does not belong to \(V_{\varepsilon }\), even if \(\frac{1}{2}(\frac{\partial (\psi _{1})^{i}}{\partial y_{j}}+\frac{\partial (\psi _{1})^{j}}{\partial y_{i}})=0 \) in \(Y_{s}\). The main result of the paper proposes a modification of this test-function which satisfies the desired conditions and which allows to pass to the limit in the variational formulation using two-scale convergence arguments. The author here assumes that the source term belongs to \( C^{1}([0,T];L^{2}(\Omega ;\mathbb{R}^{3}))\) and satisfies \(f(0)=0\). For the proof, the author quotes results from the paper by \textit{G. Nguetseng} et al. [J. Funct. Spaces Appl. 8, No. 1, 17--54 (2010; Zbl 1194.35041)] and he uses direct computations. For the entire collection see [Zbl 1401.49003].