Attractors for a nonlinear parabolic problem with terms concentrating on the boundary (Q487942)

Let \(\Omega \subset \mathbb{R}^{2}\) be a bounded domain with \(C^{2}\) boundary \(\partial \Omega \) and let \(g_{\epsilon }^{\eta }\) be a function satisfying \(0<g_{0}\leq g_{\epsilon }^{\eta }\leq g_{1}\) for fixed and positive constants \(g_{0},g_{1}\), \(\epsilon \) and with \(\eta \) being positive. The function \(g_{\epsilon }^{\eta }\) will be taken as \(g_{\epsilon }^{\eta }(s)=g(s,s/\epsilon ^{\eta })\), where \(g\) is a positive and smooth function which is \(l(x)\)-periodic with respect to its second variable for every \(x\). The authors consider the thin layer \(\omega _{\epsilon }=\{\xi \in \mathbb{R} ^{2}:\xi =\zeta (s)-tN(\zeta (s))\), \(s\in [ 0,T]\), \(0\leq t<\epsilon g(s,s/\epsilon ^{\eta })\}\), where the curve \(\zeta (s)=(x(s),y(s))\) is a \( C^{2}\) parametrization of the boundary \(\partial \Omega \) with \(\left\| \zeta ^{\prime }(s)\right\| =1\) and \(N(\zeta (s))=(y^{\prime }(s),-x^{\prime }(s))\) is the unit outward normal to \(\partial \Omega \). The authors consider the evolution problem \[ \frac{\partial u_{\epsilon }}{ \partial t}-\Delta u_{\epsilon }+\lambda u_{\epsilon }+\frac{1}{\epsilon } \chi _{\omega _{\epsilon }}V_{\epsilon }u_{\varepsilon }=\frac{1}{\epsilon } \chi _{\omega _{\epsilon }}f(u_{\varepsilon }) \text{ in } \Omega \times (0,+\infty), \] where \(\chi _{\omega _{\epsilon }}\) is the characteristic function of the thin layer \(\omega _{\epsilon }\), \(\lambda \in \mathbb{R}\), \(V_{\epsilon }\in L^{\infty }(\Omega )\) satisfying \(\frac{1}{\epsilon }\int_{\omega _{\epsilon }}(V_{\epsilon })^{2}dxdy\leq C\) and \(f\) is a \(C^{2}\) function on \(\mathbb{R}\) which satisfies \(\lim \sup_{|s|\rightarrow \infty }f(s)/s<0\). The homogeneous Neumann boundary condition \(\partial u_{\epsilon}/\partial N=0\) is imposed on \(\partial \Omega \times (0,+\infty )\) and the solution starts from an initial data \(\phi ^{\epsilon }\in H^{1}(\Omega ) \). Using an abstract semigroup framework, the authors describe the asymptotic behavior of the solution \(u_{\varepsilon }\) to this problem. The main result of the paper proves that the limit problem is \[ \frac{\partial u_{0}}{\partial t}-\Delta u_{0}+\lambda u_{0}=0 \text{ in } \Omega \times (0,+\infty ) \] with the nonlinear boundary condition \[ \frac{\partial u_{0}}{ \partial N}+V_{0}u_{0}=\mu f(u_{0}) \text{ on } \partial \Omega \times (0,+\infty ), \] where \[ \mu (s)=\frac{1}{l(s)}\int_{0}^{l(s)}g(s,\tau )d\tau \text{ for every } s\in (0,T). \] For the proof, the authors first study the behavior of the integral \(\frac{1}{\epsilon }\int_{\omega _{\epsilon }}V_{\epsilon }u\varphi d\xi \) for every \(u\in H^{s}(\Omega )\) and \(\varphi \in H^{\sigma }(\Omega )\) with \(s>1/2\), \(\sigma >1/2\) and \(s+\sigma >3/2\). Then, they study the operators \(A_{\epsilon }\) from \(H^{1}\subset H^{1}(\Omega )\) into \(H^{-1}(\Omega )\) and defined through \[ \left\langle A_{\epsilon }u,v\right\rangle =\int_{\Omega }\nabla u\nabla vdxdy+\lambda \int_{\Omega }uvdxdy+\frac{1}{\epsilon }\int_{\omega _{\epsilon }}V_{\epsilon }uvdxdy. \] They prove that these operators are continuously invertible from \(H^{2-\alpha }(\Omega )\) into \(H^{-\alpha }(\Omega )\) for \(1/2<\alpha \leq 1\) and for \(\epsilon \) small enough. The last part of the paper describes some properties of the attractors \(\mathcal{A} _{\epsilon }\) and \(\mathcal{A}_{0}\) associated to these problems. The authors first prove that the limit attractor \(\mathcal{A}_{0}\) attracts \( \cup _{\epsilon \in (0,\epsilon _{0})}\mathcal{A}_{\epsilon }\). The authors also prove an error estimate concerning the semigroups and an upper semicontinuity property for the attractors \(\{\mathcal{A}_{\epsilon }\}_{\epsilon \in [ 0,\epsilon _{0}]}\) at \(\epsilon =0\).