Method of moving frames to solve (an)isotropic diffusion equations on curved surfaces (Q461258): Difference between revisions

The present paper is devoted to the formulation of the method of moving frames (MMF) to solve numerically partial differential equations of diffusion type on surfaces with curvature and possibly some degree of anisotropy. It can be considered as the second part of the previous work [J. Sci. Comput. 53, No. 2, 268--294 (2012; Zbl 1273.65138); erratum ibid. 53, No. 2, 481 (2012; Zbl 1273.65139)], in the sense that it uses the same ideas for the implementation of the method. Roughly speaking, the method of moving frames is used to solve partial differential equations by constructing a pair of orthogonal Euclidean axis at every point of the curved surface and expand in this system vectors, and the gradient and divergence operators. To avoid singularities and to ensure the convergence of the schemes, it is assumed that every curved element of the surface is a two-dimensional manifold and that it is locally Euclidean. It is shown that if the divergence is computed exactly on the surface, then it is possible to formulate the problem in such a way that it is equivalent to the Laplace-Beltrami operator, although the direct computation of the divergence is not straightforward (due in particular to the possible existence of singularities on the surface). If this direct computation is not performed, then the error in the divergence dominates, although this effect can be ameliorated either by using higher-order differentiation schemes or by appropriately aligning the moving frames. In this setting, anisotropy is represented just by a rescaling of the moving frames, so that no additional complications are introduced with respect to the isotropic case. The direction and magnitude of the moving frames and their contribution of the MMF error in this setting are also analyzed in detail, as well as the effect of the aligning of the moving frame on a given line curvature on the overall error of the scheme. Finally, to illustrate the applications of the MMF in practical situations, the so-called atrial reentry phenomenon in electrophysiology is simulated with the new schemes.