Simulations of Grain Boundary
Evolution via Diffusion-Generated Motion
Algorithms
Matthew Elsey, CIMS
Abstract:
Many materials, including most metals and ceramics, are composed of
crystallites (often called grains), which are differentiated by their
crystallographic orientation. Multiphase weighted motion by mean
curvature arises as a classical model describing the annealing of these
materials. The distance function-based diffusion-generated motion
(DFDGM) algorithm is introduced and demonstrated to be an accurate and
efficient means for simulating this evolution in the case of pure
(equally weighted) multiphase motion by curvature. An extension of
DFDGM
to the more general weighted mean curvature model for grain growth is
presented. This extension makes use of a "minimizing motions" idea
originally proposed by Almgren, Taylor, and Wang. Results for simple
tests have good accuracy properties and are suggestive of numerical
convergence. Large-scale simulations are also presented and are shown
to
agree well with available predictions. Joint work with Selim Esedoglu
and Peter Smereka.