A second-order shape optimization algorithm for solving the exterior Bernoulli free boundary problem using a new boundary cost functional (Q782922)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A second-order shape optimization algorithm for solving the exterior Bernoulli free boundary problem using a new boundary cost functional
scientific article

    Statements

    A second-order shape optimization algorithm for solving the exterior Bernoulli free boundary problem using a new boundary cost functional (English)
    0 references
    0 references
    0 references
    29 July 2020
    0 references
    The Bernoulli Free Boundary Problem (FBP) is the prototype of overdetermined Free Boundary Problem. It can be stated in the following way (exterior FBP): let \(K\) be a compact subset in \(\mathbb{R}^N\) and \(c>0\) a constant, we look for a domain \(\Omega\) containing \(K\) such that the following ``overdetermined'' partial differential equation has a solution \[ \begin{cases} \Delta u =0 &\text{in } \Omega \\ u=1 &\text{on } \partial K\\ u=0 &\text{on } \partial \Omega\\ |\nabla u|=c &\text{on } \partial\Omega. \end{cases} \] There is an important literature on this classical problem, a very good review both on theoretical questions and on numerical approximations is proposed in the paper [J. Reine Angew. Math. 486, 165--205 (1997; Zbl 0909.35154)] by \textit{M. Flucher} and \textit{M. Rumpf}. In this paper, the authors propose a new numerical method to tackle this problem. They introduce \(u_N\) the solution of the Neumann problem \[ \begin{cases} \Delta u_N =0 &\text{in } \Omega \\ u_N=1 &\text{on } \partial K\\ \frac{\partial u_N}{\partial n}=-c &\text{on } \partial\Omega \end{cases} \] and \(u_R\) the solution of the Robin problem \[ \begin{cases} \Delta u_R =0 &\text{in } \Omega \\ u_R=1 &\text{on } \partial K\\ \frac{\partial u_R}{\partial n}+\beta u_R=-c &\text{on } \partial\Omega. \end{cases} \] Of course, if \(\Omega\) is a solution of the Bernoulli problem, we must have \(u_R=u_N\) and vice-versa. Thus the authors want to minimize the functional \(J(\Omega)=\int_{\partial\Omega} (u_N-u_R)^2\). This is a similar idea to the classical approach by Kohn-Vogelius, see [\textit{J. Ginibre} and \textit{G. Velo}, Ann. Inst. Henri Poincaré, Phys. Théor. 43, 399--442 (1985; Zbl 0595.35089)] (with the difference that the Kohn-Vogelius functional involves the solution of the Dirichlet and the Neumann problem and the integral takes place on the whole domain, not just on its boundary). The authors claim that some important advantages of their method are that their functional yields more regularity in the solution of the associated adjoint problem and provide faster and more stable convergence (of the approximated solution to the exact one). They obtain the first and second order shape derivative of the functional \(J\) useful to perform the optimization algorithm. They describe in details several algorithms and present four convincing examples. Nevertheless, a comparison with other known algorithms is missing.
    0 references
    Bernoulli problem
    0 references
    domain perturbation
    0 references
    free boundary
    0 references
    shape optimization
    0 references
    shape derivative
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references

    Identifiers