Qualitative behaviour of solutions in two models of thin liquid films (Q341174)

Starting from the Stokes flow \(\nabla p=\mu\nabla^2u+f\), \(\nabla\cdot u=0\) occurring between two plates \(z=\pm H\) and considering the Hele-Show regime, the authors obtain the system \(u=\nabla p\frac{z^2-H^2}{2\mu}\), \(\nabla^2p=0\), \(\nabla p\cdot \overrightarrow{n}=0\). They finally end with the thin-film equation \(h_{t}+(|h|^{n}(f(h_{x})h_{x})_{xx})_{x}=0\) in \( Q_{T}=\Omega \times (0,T)\) with \(\Omega =(-a,a)\). Here \(f(z)=(1+z^{2})^{-1/2} \). The boundary conditions \(h_{x}(\pm a,t)=(f(h_{x})h_{x})_{xx}(\pm a,t)=0\) are added together with the initial condition \(h(x,0)=h_{0}(x)\). The authors define the notion of nonnegative generalized weak solution as a function \( h\in C(\overline{Q_{T}})\cap L^{\infty }(0,T;W_{1}^{1}(\Omega ))\), \(h_{t}\in L^{2}(0,T;(H^{1}(\Omega ))^{\prime })\), \(|h|^{n}(f(h_{x})h_{x})_{xx}\in L^{2}(\overline{Q_{T}}\setminus (\{h=0\}\cup \{t=0\})\), satisfying \( \int_{0}^{T}\left\langle h_{t},\phi \right\rangle _{(H^{1})^{\prime },H^{1}}dt-\int \int_{\overline{Q_{T}}\setminus (\{h=0\}\cup \{t=0\})}|h|^{n}(f(h_{x})h_{x})_{xx}\phi _{x}dxdt=0\) for every \(\phi \in L^{2}(0,T;H^{1}(\Omega ))\), and \(h(.,t)\rightarrow _{t\rightarrow 0}h_{0}(.)\) strongly in \(L^{2}(\Omega )\). The first main result of the paper proves the existence of a nonnegative generalized weak solution to the thin-film problem, assuming appropriate hypotheses on the initial data \(h_{0}\). Further this weak solution satisfies some regularity properties. For the proof, the authors introduce an approximate parabolic problem, for which they prove an existence result. They conclude deriving uniform estimates with respect to the two small parameters involved in this auxiliary problem and passing to the limit. The second main result proves an asymptotic decay of this weak solution that is \(\int_{\Omega }(\sqrt{1+h_{x}^{2}}-1)dx\leq C_{0}(1+t)^{-1}\) and \(h(x,t)\rightarrow _{t\rightarrow \infty }\frac{M}{ \left| \Omega \right| }\), where \(C_{0}>0\) only depends on \(h_{0}\) and \(\left| \Omega \right| \). Moreover if \(M>\left| \Omega \right| ^{2}\) \(\int_{\Omega }(\sqrt{1+h_{x}^{2}}-1)dx\leq \exp (-C_{1}t)\int_{\Omega }(\sqrt{1+h_{0x}^{2}}-1)dx\)., for a constant \(C_{1}>0\) which only depends on \(h_{0}\) and \(\left| \Omega \right| \). The proof is obtained through direct computations after the introduction of the function \( L_{\sigma }(z)=\sqrt{1+z^{2}}+\frac{\sigma }{z^{2}}\) with \(\sigma >0\) and assuming the existence of a unique positive and classical solution to the thin-film problem. The last result of the paper proves the existence of a classical solution \(h\in C_{x,t}^{4,1}(\overline{Q_{T}})\) to the thin-film problem, assuming that the positive initial data \(h_{0}\) belongs to \( C^{4+\alpha }(\overline{\Omega })\). Moreover the corresponding arclength dissipates \(\int_{\Omega }\sqrt{1+h_{x}^{2}}dx+C(\delta )\int \int_{Q_{T}}f^{3}(h_{x})h^{n}h_{xxx}^{2}dxdt\leq \int_{\Omega }\sqrt{ 1+h_{0x}^{2}}dx\), under hypotheses on \(n\) and \(h_{0}\). The authors finally illustrate this arclength dissipation property with results of numerical computations.