Discrete anisotropic curve shortening flow in higher codimension
Klaus Deckelnick, Robert Nürnberg
TL;DR
This work develops a mathematically rigorous and computationally practical framework for anisotropic curve shortening flow of curves in arbitrary codimension. By applying a DeTurck-type reformulation, the authors obtain a strictly parabolic, divergence-form PDE that admits a variational finite element discretization. They prove unconditional stability for the fully discrete scheme and establish optimal error bounds: a in L2-norm of order $h^2$ in the H1 estimate and order $h^4$ in L2, under a mild coupling between space and time discretization. Numerical experiments in plane and in 3D validate convergence rates and demonstrate robust behavior across isotropic and anisotropic, including crystalline-like, anisotropies, while also illustrating how tangential motion yields well-distributed mesh points and the method’s ability to handle singularities up to the limits of the theory.
Abstract
We introduce a novel formulation for the evolution of parametric curves by anisotropic curve shortening flow in ${\mathbb R}^d$, $d\geq2$. The reformulation hinges on a suitable manipulation of the parameterization's tangential velocity, leading to a strictly parabolic differential equation. Moreover, the derived equation is in divergence form, giving rise to a natural variational numerical method. For a fully discrete finite element approximation based on piecewise linear elements we prove optimal error estimates. Numerical simulations confirm the theoretical results and demonstrate the practicality of the method.
