G-convergence of a class of evolution operators (Q1824763): Difference between revisions

Some relations between G-convergence of a family of monotone operators \(A_{\epsilon}(t): V\to V^*\) and G-convergence of evolution operators of the form \(d/dt+A_{\epsilon}(t): W\to X^*\) are investigated. The notations are as usual: \(V\subset H\subset V^*\), \(X=L^ p(0,T;V)\), \(X^*=L^ q(0,T;V^*)\), \(W=\{u\in X:du/dt\in X^*\}.\) The main theorem states that under suitable assumptions concerning the monotone operators \(A_{\epsilon}(t)\) (which assure the existence of solutions in W to the problems \(du/dt+A_{\epsilon}(t)u=f,\) \(u(0,x)=u^ 0\), where \(f\in X^*\), \(u^ 0\in H)\), the following is true: if \(A_{\epsilon}(t)\) G-converge to A(t) for every \(t\in [0,T]\), then we have also G-convergence of the evolution operators \(d/dt+A_{\epsilon}(t)\) to the operator \(d/dt+A(t)\) and moreover \(u_{\epsilon}-A_{\epsilon}^{-1}(t)A(t)u\to 0\) in X, as \(\epsilon\) \(\to 0\) \((u_{\epsilon}\) and u are solutions to the initial value problems as above). The problem of G-convergence for operators \(c_{\epsilon}d/dt+A_{\epsilon}(t)\) is also considered and some examples of G-convergent monotone operators are given.