Shape optimization on the lateral boundary for multi-phase stefan problems (Q4030863)

This paper is concerned with the shape optimization on the lateral boundary for multiphase Stefan problems with Dirichlet-Neumann boundary conditions. The enthalpy formulation of the Stefan problem is described as follows: \[ SP(\Gamma)\begin{cases} u_ t-\Delta\beta(u)=f &\qquad\text{in} \qquad Q,\\ u(0,\cdot)=u_ 0 &\qquad\text{in}\qquad \Omega,\\ \beta(u)=g &\qquad\text{on} \qquad (0,T)\times\Gamma,\\ \partial\beta(u)/\partial n=h&\qquad\text{on} \qquad (0,T)\times(\partial\Omega - \Gamma),\end{cases} \] where \(\Omega\) is a bounded domain in \(\mathbb{R}^ N\) (\(N\geq 2\)) with smooth boundary \(\partial\Omega\), \(0<T<\infty\), \(Q:=(0,T)\times\Omega\) and \(\Sigma:=(0,T)\times\partial\Omega\), \(\Gamma\) is a closed subset of \(\partial\Omega\) with positive surface measure, \(\beta: \mathbb{R}\to\mathbb{R}\) is a given function, with \(\beta(0)=0\), which is nondecreasing and Lipschitz continuous on \(\mathbb{R}\), \(f\) is a given function on \(Q\), \(g\), \(h\) on \(\Sigma\) and \(u_ 0\) on \(\Omega\), \(u:(0,T]\times\Omega\to\mathbb{R}\) is the unknown state function (\(u\) represents the enthalpy and \(\beta(u)\) the temperature). The cost functional \(J(\Gamma)\) is given by the formula (for the shape optimization of \(\Gamma\)) \[ J(\Gamma)={1\over 2} \int_ Q |\beta(u_ \Gamma)-\beta_ d|^ 2 dx dt+{1\over 2} \int_{(0,t)\times (\partial\Omega-\Gamma)} | h|^ 2 d\sigma dt \] for each \(\Gamma\in\Pi\), where \(\Pi=\{\Gamma\subset\partial\Omega\), \(\Gamma\) is compact in \(\partial\Omega\) and the surface measure \(\sigma(\Gamma)\) is positive\}, \(u_ \Gamma\) is the solution of \(SP(\Gamma)\) and \(\beta_ d\) is a given function in \(L_ 2(Q)\). For a given nonempty subset \(\Pi_ C\) of \(\Pi\) the optimization problem \[ P(\Pi_ c):\qquad\qquad\qquad\qquad\Gamma_ *\in\Pi_ C; \qquad J(\Gamma_ *)=\inf_{\Gamma\in\Pi_ c} J(\Gamma) \] is considered. The aim of this important paper is to establish an existence result for \(P(\Pi_ c)\). The second aim of this paper is to discuss approximations of \(SP(\Gamma)\) and \(P(\Pi_ c)\) by smooth problems from some numerical points of view. Main result: 1) The specification (the most important step) of a topology on \(\Pi\) which gives continuous dependence of the solution \(u_ \Gamma\) of \(SP(\Gamma)\) upon \(\Gamma\in\Pi_ c\). 2) The authors' main idea is to approximate \(\beta\) by smooth function \(\beta^ \varepsilon\), \(0<\varepsilon\leq 1\), with \((d/dr)\beta^ \varepsilon(r)\geq\varepsilon\) and at the same time approximate the mixed boundary condition by a smooth Neumann boundary condition of the following type: \[ \partial\beta^ \varepsilon (u)\partial n=(-\chi_ \Gamma^ \nu/\mu)(\beta^ \varepsilon (u)-g)+(1-\chi_ \Gamma^ \nu)h \qquad \text{on}\qquad \Sigma, \] where \(\nu,\mu\in(0,1]\) are parameters and \(\chi_ \Gamma^ \nu\) is a smooth function on \(\partial\Omega\) which approaches \(\chi_ \Gamma\) as \(\nu\to 0\).