Homogenization of some types of Neumann and Dirichlet problems (Q2569874)

The authors study the asymptotic behaviour as \(\varepsilon\rightarrow 0\) of solutions \(u_\varepsilon^{\tau,\sigma}\) of the equation \(-\Delta u=f\) in a perforated domain \(\Omega_\varepsilon^{\tau,\sigma}\) obtained from a bounded Lipschitz domain \(\Omega\) in \({\mathbb R}^n\) by removing closed, smoothed cubes well contained in \(\Omega\), and of diameter \(r(\varepsilon)\leq \varepsilon\), with suitable boundary conditions imposed on the boundary of each hole. More precisely, the situation is studied in which two kinds of holes are removed from \(\Omega\). The first (Neumann) kind is periodically distributed with period \(\varepsilon\) in \({\mathbb R}^n\) and scaled as \(r(\varepsilon)=\varepsilon\), and the second (Dirichlet) kind is scaled as \(\varepsilon^{n/(n-2)}\) if \(n\geq 3\) and as \(\exp(-\varepsilon^{-2})\) if \(n=2\). Each element of the second family is assumed to move towards an element of the first family with speed \(\tau\varepsilon^\sigma\) if \(n\geq 3\) and with speed \(\tau\exp(-\varepsilon^{-\sigma})\) if \(n=2\) for suitable values of parameters \(\tau, \sigma\). Homogeneous Dirichlet boundary conditions are imposed on the exterior boundary \(\partial\Omega\) and on the boundary of each Dirichlet hole, while a nonhomogeneous Neumann condition \({{\partial u_\varepsilon}\over{\partial{\mathbf n}}}=g_\varepsilon\) is imposed on the boundary of each Neumann hole, where \(g_\varepsilon\) is obtained by rescaling and periodicizing an \(L^2\) function \(g\) on \(\partial Q\) such that \(\int_{\partial Q}g\,dx=0\), where \(Q\) is a standard reference hole. The main result is the strong convergence in \(L^2(\Omega_\varepsilon^{\tau,\sigma})\) as \(\varepsilon\rightarrow 0\) of \(u_\varepsilon^{\tau,\sigma}\) to a solution \(u^{\tau,\sigma}\) of the limit problem \[ -\text{div}\, ({\mathcal A}\,\nabla u^{\tau,\sigma}) + {1\over 2} \mu^{\tau,\sigma}u^{\tau,\sigma} = \theta f \quad\hbox{in}\quad\Omega, \qquad u^{\tau,\sigma}=0 \quad\hbox{on}\quad\partial\Omega, \] for some constants \(\mu^{\tau,\sigma}\) and \(\theta\), and a constant matrix \({\mathcal A}\) determined by the parameters \(\tau,\sigma\). Similar problems have been studied by various authors, for example, \textit{A. Corbo Esposito, C. D'Apice} and \textit{A. Gaudiello} [Asymptotic Anal. 31, No. 3--4, 297--316 (2002; Zbl 1043.35027)] and \textit{G. Cardone, C. D'Apice} and \textit{U. De Maio} [NoDEA, Nonlinear Differ. Equ. Appl. 9, No. 3, 325--346 (2002; Zbl 1046.35007)].