A parabolic free boundary problem modeling electrostatic MEMS (Q2436331)

The authors study a model which accounts for the description of the behaviour of microelectromechanical systems (MEMS). The equation which describes the evolution of the unidirectional displacement \(u\) of the membrane is written as \(\partial _{t}u-\partial _{x}^{2}u=-\lambda (\varepsilon ^{2}\left| \partial _{x}\psi (t,x,u)\right| ^{2}+\left| \partial _{z}\psi (t,x,u)\right| ^{2})\), in \( (-1,1)\times (0,\infty )\). The homogeneous Dirichlet boundary conditions \( u(t,\pm 1)=0\) and the initial condition \(u(0,x)=u^{0}(x)\) on \(I=(-1,1)\) are added. Here \(\lambda \) represents the voltage and \(\varepsilon \) is the aspect ratio of the device before scaling. The dimensionless electrostatic potential \(\psi \) is the solution of Laplace's equation \(\varepsilon ^{2}\partial _{x}^{2}\psi +\partial _{z}^{2}\psi =0\) in \(\Omega (u(t)):=\{(x,z)\in I\times (-1,\infty ):-1<z<u(t,x)\}\). The boundary conditions \(\psi (t,x,z)=\frac{1+z}{1+u(t,x)}\) are imposed at \(\Omega (u(t))\), \(t>0\). Here \(z=0\) (resp. \(z=-1\)) corresponds to the undeflected membrane (resp. a rigid ground plate). The authors first prove a local well-posedness result for the system, for every \(\lambda >0\), assuming that \(u^{0}\in W_{q,D}^{2}(I)\) for \(q\in (2,\infty )\) and \(u^{0}>-1\) on \(I\). They then prove a global in time existence result assuming that \(u^{0}\in \{u\in W_{q,D}^{2}(I);\left\| u\right\| _{W_{q,D}^{2}(I)}<1/\kappa \), \( -1+\kappa <u\) on \(I\}=:S_{q}(\kappa )\) for some \(\kappa \in (0,1)\) and \( \lambda \in (0,\lambda _{\ast })\) for some \(\lambda _{\ast }(\kappa ,\varepsilon )>0\). They also prove that if \(\lambda >\lambda ^{\ast }(\varepsilon )>0\) then the solution exists only for \(t\leq T_{m}^{\varepsilon }<\infty \), a maximal time of existence. For the proof of these existence results, the authors first introduce a diffeomorphism \(T: \overline{\Omega (v)}\rightarrow \overline{I\times (0,1)}\) which brings the moving domain to a fixed rectangle but which changes the Laplace operator to a \(v\)-differential operator, whose properties are studied. The authors quote results from \textit{P. Laurençot} and \textit{C. Walker} [Arch. Ration. Mech. Anal. 207, No. 1, 139--158 (2013; Zbl 1268.78004)] and they use compactness results and an analytic semigroup framework. For the non-existence of a global in time solution, the authors essentially use energy estimates. In the last parts of their paper, the authors analyze the asymptotic steady states of the problem in the case of a global in time solution and they describe the behaviour of the solution \((u_{\varepsilon },\psi _{\varepsilon })\) when the aspect ratio \( \varepsilon \) goes to 0, first proving that the maximal time of existence \( T_{m}^{\varepsilon }\) is always greater than a fixed \(\tau >0\). The proofs are based on a principle of linearized stability and on the analytic semigroup framework, through the derivation of uniform estimates and the use of compactness results.