Free boundary regularity in the multiple membrane problem in the plane (Q6161164)

Suppose one has a domain $\Omega \subset \mathbb{R}^2$, $\{\omega_k\}_{k=1}^N \subset \mathbb{R}^+$, and bounded functions $f_k$, $k =1,2,\dots,N$. In the article, the authors investigate minimizers of \[ \mathbf{u}:=(u_1,u_2,\dots,u_N) \mapsto \int_{\Omega} \sum \omega_k\Big(\frac{1}{2}|\nabla u_k|^2+f_ku_k \Big) dx \] with boundary data and \[ u_1 \ge u_2 \ge \cdots \ge u_N\text{ in }\Omega. \] In the multiple membrane problem, $f_k$ is the force acting on the $k^{th}$ membrane, $\omega_k$ is its weight, and its height is given by $u_k$. One also initially assumes \[ f_1>f_2>\cdots>f_N. \] The information given by the minimization is \[ (v_i-u_i)\Delta u_i \le (v_i-u_i)f_i, \] supposing $\mathbf{v} \in H^1(\Omega)$ satisfies \[ v_1 \ge v_2 \ge \cdots \ge v_N\text{ in }\Omega; \] \[ \sum \omega_i \Delta u_i = \sum \omega_i f_i; \] \[ \sum \omega_i u_i\Delta u_i = \sum \omega_i u_i f_i. \] In addition, the contact region is defined to be $\{u_k=u_{k+1}\}$. The authors study the smoothness of the free boundary $\Gamma_k= \partial \{u_k>u_{k+1}\}$. Assuming normalizations, and \[ 0 \in \bigcap \Gamma_k, \] the first theorem is uniqueness of the blow-ups: \[ \mathbf{p}(x)=\lim_{r \rightarrow 0^+}\mathbf{u}(r x)/r^2, \] \[ \mathbf{u}(x)=\mathbf{p}(x)+O(|x|^2(-\log |x|)^{-1}). \] The second theorem is about the free boundary regularity at points where the blow-up has the least energy: Denote $\mathbf{p}_0(x_2):=\frac{1}{2}(x_2^+)^2\mathbf{f}$, and suppose that there is some constant $\epsilon>0$ such that with \[ |\mathbf{u}-\mathbf{p}_0| \le \epsilon\text{ in }B_1, \] than the free boundary $\Gamma_k$ is $C^{1,\log}$ in $B_{1/2}$. The tangential geometry of the intersection of the free boundaries around points with positive density is one interesting element in the paper. Furthermore, the discrete multiple membrane problem contains an interesting game theoretic interpretation: there are initially $N$ players and $N$ tickets, each has one. There is a token which randomly moves to neighbors on a lattice in $\Omega$. Each round, players interchange tickets. The player with the first ticket can choose any ticket; the second can select one from the $N-1$ remaining tickets, etc. The pay to keep the first ticket is $f_1$, to keep the second is $f_2$, etc. Supposing the token leaves $\Omega$, the boundary data $\phi_k$ is the payoff for the $k^{th}$ player. The solution $\mathbf{u}$ at the location of the token is the expected payoff in the domain when all players optimize the strategies with unit weights, and the contact regions have the optimal strategies for the interchange part.