Article contents
Adaptive Space–Time Numerical Methods for Nonlinear Parabolic Equations with Moving Interfaces
Abstract
We study a space–time Galerkin method for nonlinear parabolic problems whose coefficients jump across an interface that moves in time. In space–time, the interface is a hypersurface, and the mesh is built to fit it. As a result, no mesh velocity is needed, and the time step does not have to follow the front. Inside a time slab, the discrete functions are continuous in space and time, and slabs are coupled by an upwind jump term. We derive a residual error estimator with four parts (space, time, interface geometry and nonlinear solver) and give proof sketches for its reliability and local efficiency. We also show that, in the exact norm, the thin strip between the true and the discrete interface causes an error of order δ¹ᐟ², where δ is the distance between the two. An adaptive algorithm with curvature-aware anisotropic refinement and a Newton stopping rule tied to the estimator is described. A prototype for one spatial dimension and one time slab was used for the tests. The estimator bounded the error in every run, with an effectivity index near 15 for equal diffusion coefficients. The geometry indicator decayed at the predicted rate, and adaptive refinement needed about a third as many unknowns as uniform refinement. The effectivity index grew sharply with the diffusion contrast, so the estimator is not robust in that respect.
Aims & scope
Call for Papers
Article Processing Charges
Publications Ethics
Google Scholar Citations
Recruitment