Matrix-valued Allen-Cahn equation and the Keller-Rubinstein-Sternberg problem (Q6101147)

This paper is devoted to establishing rigorously the sharp interface limit of a matrix-valued Allen-Cahn equation \[ \partial_tA_\varepsilon=\Delta A_\varepsilon-\varepsilon^{-2}\left(A_\varepsilon A_\varepsilon ^TA_\varepsilon -A_\varepsilon \right), \quad A_\varepsilon: \Omega\subset\mathbb{R}^m\mapsto \mathbb{R}^{n\times n}. \tag{1} \] This parabolic equation is the gradient flow of the energy functional \[ E_\varepsilon(A)=\int \frac{1}{2}|\nabla A|^2+\frac{1}{\varepsilon^2} F(A), \quad F(A)=\frac{1}{4}\|AA^T-I\|^2. \tag{2} \] Notice that the potential energy \(F\) has two wells (global minima sets), \(O^\pm(n)\) (the set of orthogonal matrices with determinant \(\pm 1\)). The limit of (1) is a two-phase flow: there exist two disjoint domains \(\Omega_\pm\), where \(A_\pm\), the limit of \(A_\varepsilon\), lie in \(O^\pm(n)\) respectively and evolve according to the harmonic map heat flow (into \(O^\pm(n)\)), while the interface between \(\Omega^+\) and \(\Omega^-\), \(\Gamma_t\), evolves according to the motion by mean curvature. There is also a coupling between \(A_+\) and \(A_-\) via some intriguing boundary conditions of them on \(\Gamma_t\). The main result of this paper states that, if the initial data is well-prepared and the limiting problem has a local smooth solution, then the solution of (1) converges to the limit in a smooth way. This type of problems was introduced by \textit{J. Rubinstein} et al. [SIAM J. Appl. Math. 49, No. 6, 1722--1733 (1989; Zbl 0702.35128)]. The rigorous derivation of the sharp interface limit was also known in several simpler cases, e.g. the scalar Allen-Cahn equation or if \textit{the fully minimally paired condition} is satisfied, which requires for every pair \(p_\pm\) in the two wells of \(F\), there exists a minimal connecting orbit. However, (1) is only partially minimally paired, i.e. there may not exist minimal connecting orbit between some \(p_\pm\in O^\pm(n)\). A notion of \textit{quasi-minimal connecting orbit} is then introduced in this paper to replace the role of minimal connecting orbit. The proof of the main result consists of two steps: first an approximate solution of (1) with error controlled up to sufficiently high order is constructed, where a modified matched asymptotic expansion method (gluing of outer and inner expansions) in the framework of Hilbert expansion is used; and then a spectral lower bound for the linearized operator around this approximate solution is established, which is used to control the error between the approximate solution and the real solution.