Multi-phase-field modeling and high-performance computation for predicting material microstructure evolution during sintering

As the properties of sintered products are considerably affected by the microstructures formed during sintering, numerical simulation is essential for predicting and controlling microstructures with high accuracy. Although the phase-field method can reproduce sintered microstructures with the highest accuracy, its high computational cost has limited the scale of computation. In this study, we develop a multi-phase-field (MPF) sintering model with a double-obstacle potential that is effective for large-scale simulations. We also establish an efficient algorithm on graphics processing unit (GPU) to accelerate the computations of rigid-body motions of particles, which cause densification, and implement it on multiple GPUs in parallel. The simulation method enables three-dimensional large-scale MPF sintering simulations of approximately 160,000 Al 2 O 3 particles on 1,280 3 grid points, which is sufficiently large to reproduce a bulk sintering behavior. For the large-scale simulation results, the difference between near-surface and bulk sintering behaviors is discussed. The large-scale MPF sintering simulation method developed in this study is expected to contribute to the accurate prediction and control of sintered microstructures significantly.