Elliptic versus parabolic regularization for the equation of prescribed mean curvature (Q1361818)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Elliptic versus parabolic regularization for the equation of prescribed mean curvature |
scientific article |
Statements
Elliptic versus parabolic regularization for the equation of prescribed mean curvature (English)
0 references
29 June 1998
0 references
It is well-known that the Dirichlet problem for the equation of prescribed mean curvature \[ \text{div}((1+|Du|^2)^{-1/2}Du)+ h(x)= 0\quad\text{in }\Omega\subset\mathbb{R}^n,\;u=\Phi\quad\text{on }\partial\Omega\tag{1} \] has no classical solution if the curvature function \(h\) is too big. The authors investigate for this case what happens to the parabolic and elliptic regularizations of (1): \[ u_t= \text{div}((1+|Du|^2)^{-1/2} Du)+ h(x)\text{ in }\Omega\times (0,\infty),\tag{2} \] \[ u(x,t)= \Phi \text{ on } \partial\Omega\times (0,\infty),\;u(x,0)= u_0(x)\text{ on }\Omega \] and \[ \text{div}((1+ |Du^\varepsilon|^2)Du^\varepsilon+ \varepsilon Du^\varepsilon)+ h(x)= 0\quad\text{in }\Omega,\;u^\varepsilon= \Phi\quad\text{on }\partial\Omega.\tag{3} \] More precisely, they study the asymptotic behavior of the solutions of (2) and (3) for \(t\to\infty\) and \(\varepsilon\to 0\), respectively. It is shown that \(u(x,t)/t\) converges in \(L_2\) to a function \(v\) (called `parabolic growth function') which is the unique minimizer in \(\text{BV}(\overline\Omega)\cap L^2(\Omega)\) of the functional \(F(u)= \int_\Omega|Du|+ \int_\Omega(\textstyle{{1\over 2}} u^2-hu)dx\), whereas \(\varepsilon u^\varepsilon\) converges in \(W^{1,2}_0(\Omega)\) (it is assumed that \(\Phi=0\)) to an ``elliptic growth function'' \(w\) which is the unique minimizer in \(W^{1,2}_0(\Omega)\) of \[ J(u)= \int_\Omega|Du|dx+\int_\Omega({1\over 2}|Du|^2- hu)dx. \] Furthermore, the authors analyze the subset \(\Omega^*\) of \(\Omega\) on which \(v\) takes on its maximal value \(\lambda^*= \text{ess sup }v\). It is expected that \(\Omega^*\) is the set on which the solution \(u\) of (2) grows fastest and at constant rate \(\lambda^*\). What they show under the assumption \(\lambda^*>0\) is that \(\Omega^*\) has positive Lebesgue measure and that \(\Omega^*\) maximizes \((-P(G)+ \int_G h(x)dx)/|G|\) among all measurable subsets \(G\) of \(\Omega\) with maximum value equal to \(\lambda^*\), where \(|G|\) and \(P(G)\) denote the Lebesgue measure and the perimeter of \(G\), respectively. Here one should remember that \(|\int_G h(x)dx|< P(G)\) is a necessary condition for the classical solvability of (1). Defining \(\widetilde\Omega\) and \(\widetilde\lambda\) for the elliptic growth function \(w\) in an analoguous manner as \(\Omega^*\) and \(\lambda^*\) before and assuming that \(\widetilde\lambda>0\) and \(\widetilde\Omega\) has a nonempty interior, the authors prove \(\int_{\widetilde\Omega} h(x)dx\geq P(\widetilde\Omega)\) and \(\int_G h(x)dx\leq P(G)\) for all \(G\) compactly contained in \(\text{int}(\widetilde\Omega)\). For \(n= 2\), \(\Omega\) a rectangle and \(h=\text{const.}\), the parabolic growth function \(v\) is explicitly calculated. In the last chapter, the authors describe a number of numerical experiments.
0 references
equation of prescribed mean curvature
0 references
classical solvability
0 references
numerical experiments
0 references