Analysis of a second order discontinuous Galerkin finite element method for the Allen-Cahn equation and the curvature-driven geometric flow

TL;DR

This paper develops and analyzes a second-order, energy-stable discontinuous Galerkin finite element method for the Allen-Cahn equation, providing sharper error bounds and demonstrating convergence to mean curvature flow.

Contribution

It introduces a novel second-order IPDG method with unconditional energy stability and improved error estimates for the Allen-Cahn equation.

Findings

The method is unconditionally energy-stable.

Error bounds depend polynomially on 1/ε, not exponentially.

Numerical experiments confirm the method's effectiveness.

Abstract

The paper proposes and analyzes an efficient second-order in time numerical approximation for the Allen-Cahn equation, which is a second order nonlinear equation arising from the phase separation model. We firstly present a fully discrete interior penalty discontinuous Galerkin (IPDG) finite element method, which is based on the modified Crank-Nicolson scheme and a mid-point approximation of the nonliner term $f(u)$. We then derive the stability analysis and error estimates for the proposed IPDG method under some regularity assumptions on the initial function $u_0$. There are two key works in our analysis, one is to establish unconditionally energy-stable scheme for the discrete solutions. The other is to use a discrete spectrum estimate to handle the midpoint of the discrete solutions $u^m$ and $u^{m+1}$ in the nonlinear term, instead of using the standard Gronwall inequality technique. This discrete spectrum estimate is not trivial to obtain since the IPDG space and the conforming $H^1$ space are not contained in each other. We obtain that all our error bounds depend on reciprocal of the perturbation parameter $\epsilon$ only in some lower polynomial order, instead of exponential order. These sharper error bounds are the key elements in proving the convergence of our numerical solution to the mean curvature flow. Finally, numerical experiments are also provided to show the performance of the presented approach and method.