A real space convolution‐based approximate algorithm for phase field model involving elastic strain energy

Abstract

Phase field models have been employed extensively in the study of microstructure evolution in materials. Elasticity plays an important role in solid‐state phase transformation processes, and it is usually introduced into phase field models in terms of the elastic strain energy by applying Khachaturyan–Shatalov microelasticity theory. Conventionally, this energy is derived in the reciprocal space and results in full‐space Fourier transformation in practice, which becomes bottle‐neck in large‐scale and massively‐parallel applications. In this article, we propose an error‐controlled approximation algorithm for scalable and efficient calculation of the elastic strain energy in phase field models. We first derive a real‐space convolutional representation of the elastic strain energy by representing the equilibrium displacements in the Khachaturyan–Shatalov microelasticity theory using Green's function. Then we propose an error‐controlled truncation criterion to approximate the corresponding terms in the phase field model. Finally, a carefully designed parallel algorithm is presented to carry out large‐scale simulations. The accuracy and efficiency of the proposed algorithm are demonstrated by real‐world large‐scale phase field simulations.