Catenoidal layers for the Allen-Cahn equation in bounded domains (Q523794): Difference between revisions

The following singularly perturbed boundary value problem for the Allen-Cahn equation is considered: \[ \alpha^2 \Delta u+u(1-u^2)=0\;\text{in}\;\Omega,\quad\frac{\partial u}{\partial n}=0\;\text{on}\;\partial \Omega, \] where \(\alpha > 0\) is a small parameter, \(\Omega \subset \mathbb{R}^N\) is a smooth bounded domain, and \(n\) denotes the inward unit normal vector to \(\partial \Omega\). This problem models phase transitions and has the associated energy functional \[ J_\alpha(u)=\int_{\Omega}^{}\frac{\alpha}{2}|\nabla u|^2+\frac{1}{4 \alpha}(1-u^2)^2,\quad u\in H^1_0(\Omega). \] Since the work of Modica and Mortola, it is well known that, as \(\alpha\to 0\) (possibly along a sequence), local minimizers with uniformly bounded energies converge in \(L^1(\Omega)\) to a step function of the form \(\chi_E-\chi_{\Omega \setminus E}\), where \(E\) is a set that minimizes perimeter in \(\Omega\) and cuts \(\partial \Omega\) orthogonally. This result uncovered a deep relation between solutions of the Allen-Cahn equation and minimal surface theory. In particular, it provided the motivation for constructing solutions to the Allen-Cahn equation whose energy is concentrated around a given minimal surface (not necessarily area minimizing). This direction has been recently pursued by the authors and their collaborators to produce important results in the case of the whole space problem, most notably a counterexample to De Giorgi's conjecture in high dimensions. Related results for the bounded domain case were already available if the given minimal surface was a line in two dimensions or a plane in three dimensions, cutting the boundary orthogonally and being non-degenerate in some sense. In the latter three-dimensional situation, some partial rotational symmetry of the domain had also to be imposed. Here, the authors consider a complete embedded minimal surface \(M\) of finite total curvature in \(\mathbb{R}^3\) that cuts the boundary of \(\Omega\) orthogonally. Under suitable geometric non-degeneracy assumptions involving \(M\) and \(\Omega\), they prove the existence of solutions in three dimensions that have a steep transition layer across \(M\). More precisely, orthogonally to the interface \(M\), the solution looks like a squeezed one-dimensional hyperbolic tangent profile. Their approach is to use a gluing technique to reduce the problem to a boundary value problem on \(M\cap \Omega\) where the main role is played by the Jacobi operator of \(M\).