Homogenization of a Dirichlet semilinear elliptic problem with a strong singularity at $u=0$ in a domain with many small holes

TL;DR

This paper studies the homogenization of a semilinear elliptic problem with a strong singularity at zero in domains with many small holes, introducing a new solution concept to handle the singularity and analyzing the limit behavior.

Contribution

It introduces a new notion of solution for strongly singular problems and extends homogenization results to cases where solutions do not belong to the usual Sobolev space.

Findings

A new solution concept ensures existence and uniqueness.

Homogenization leads to a 'strange term' in the limit equation.

Extension of previous results to strongly singular cases.

Abstract

We perform the homogenization of the semilinear elliptic problem \begin{equation*} \begin{cases} u^\varepsilon \geq 0 & \mbox{in} \; \Omega^\varepsilon,\\ \displaystyle - div \,A(x) D u^\varepsilon = F(x,u^\varepsilon) & \mbox{in} \; \Omega^\varepsilon,\\ u^\varepsilon = 0 & \mbox{on} \; \partial \Omega^\varepsilon.\\ \end{cases} \end{equation*} In this problem $F(x,s)$ is a Carath\'eodory function such that $0 \leq F(x,s) \leq h(x)/\Gamma(s)$ a.e. $x\in\Omega$ for every $s > 0$, with $h$ in some $L^r(\Omega)$ and $\Gamma$ a $C^1([0, +\infty[)$ function such that $\Gamma(0) = 0$ and $\Gamma'(s) > 0$ for every $s > 0$. On the other hand the open sets $\Omega^\varepsilon$ are obtained by removing many small holes from a fixed open set $\Omega$ in such a way that a "strange term" $\mu u^0$ appears in the limit equation in the case where the function $F(x,s)$ depends only on $x$. We already treated this problem in the case of a "mild singularity", namely in the case where the function $F(x,s)$ satisfies $0 \leq F(x,s) \leq h(x) (\frac 1s + 1)$. In this case the solution $u^\varepsilon$ to the problem belongs to $H^1_0 (\Omega^\varepsilon)$ and its definition is a "natural" and rather usual one. In the general case where $F(x,s)$ exhibits a "strong singularity" at $u = 0$, which is the purpose of the present paper, the solution $u^\varepsilon$ to the problem only belongs to $H_{\tiny loc}^1(\Omega^\varepsilon)$ but in general does not belongs to $H^1_0 (\Omega^\varepsilon)$ any more, even if $u^\varepsilon$ vanishes on $\partial\Omega^\varepsilon$ in some sense. Therefore we introduced a new notion of solution (in the spirit of the solutions defined by transposition) for problems with a strong singularity. This definition allowed us to obtain existence, stability and uniqueness results.