Introduction to the theory of two-scale convergence (Q460518): Difference between revisions

Let \(\Omega\) be a bounded domain in \(\mathbb R^n\) and let \(\square=[0,1]^n\) be the cell of periodicity. For a 1-periodic function \(\psi(y)\), denote by \(\langle\psi\rangle=\int_\square\psi(y)\,dy\) the mean value of \(\psi\). The symbol \(L^2(\square)\) is used not only for functions defined on \(\square\) but also for their 1-periodic extensions to \(\mathbb R^n\). One denotes by \(C^\infty_{per}(\square)\) the space of infinitely differentiable 1-periodic functions on \(\mathbb R^n\). If \(b\in L^2(\square)\) is 1-periodic on \(\mathbb R^n\), then \(b(\varepsilon^{-1}x)\) converges weakly in \(L^2(\Omega)\) to \(\langle b\rangle\) as \(\varepsilon \to 0\), for every bounded domain \(\Omega\subset \mathbb R^n\). A sequence \((u_\varepsilon)\) in \(L^2(\Omega)\) is called weakly two-scale convergent if it is bounded and there exists \(u\in L^2(\Omega\times\square)\) such that \[ \lim_{\varepsilon\to 0}\int_\Omega u_\varepsilon(x)\varphi(x)b(\varepsilon^{-1}x)\,dx=\int_{\Omega\times\square} u(x,y)\varphi(x)b(y)\,dx\,dy \] for all \(\varphi\in C_0^\infty(\Omega)\) and \(b\in C^\infty_{per}(\square)\). This convergence is denoted by \(u_\varepsilon(x) \rightharpoonup_2 u(x,y)\). The authors prove a compactness principle for this type of convergence: any bounded sequence \((u_\varepsilon)\) in \(L^2(\Omega)\) contains a weakly two-scale convergent subsequence (Theorem 2). They consider also the strong two-scale convergence of a sequence \(u_\varepsilon\) in \(L^2(\Omega)\), meaning that it is bounded and there exists \(u(x,y)\in L^2(\Omega\times\square)\) such that \[ \lim_{\varepsilon\to 0}\int_\Omega u_\varepsilon(x) v_\varepsilon(x)\,dx=\int_{\Omega\times\square} u(x,y)\varphi(x)b(y)\,dx\,dy, \] for any sequence \(v_\varepsilon\) in \(L^2(\Omega)\) satisfying \(v_\varepsilon(x) \rightharpoonup_2v(x,y)\). Obviously, strong two-scale convergence implies weak two-scale convergence and one shows that weak two-scale convergence plus the condition \[ \lim_{\varepsilon\to 0}\int_\Omega |u_\varepsilon(x)|^2\, dx=\int_{\Omega\times\square} |u(x,y)|^2\,dx\,dy \] imply strong two-scale convergence. The authors present in this paper ``the basic properties of these notions in Lebesgue and Sobolev spaces and the results are applied to the homogenization of elliptic operators and to spectral problems''.