[P1-59] Direct mapping from molecular dynamics to phase-field simulations for accurate prediction of grain growth
For numerically predicting grain growth, two different approaches have been used: atomistic simulations typified by molecular dynamics (MD); and continuum-based treatments including the Monte-Carlo, phase-field, and vertex methods. The former can model the spontaneous nucleation process that precedes grain growth. However, due to the large computational cost, it is difficult to simulate grain growth until the late stage using only atomistic methods. On the other hand, continuum-based models allow for relatively efficient computations. In particular, the multi-phase-field (MPF) model [I. Steinbach and F. Pezzolla, Physica D, 134 (1999) 385], which is an extension of the phase-field model to polycrystalline systems, is widely employed in recent years as a prominent tool for simulating grain growth with accuracy and efficiency. Nevertheless, the MPF model cannot directly reproduce the nucleation phenomenon. Considering the strong dependence of grain growth behaviors on the initial structure, there is a pressing need for a means of providing realistic initial structures for MPF simulation.
In this study, we aim to achieve more accurate and efficient prediction of grain growth by exploiting the merits of atomistic and continuum simulations. To this end, we propose a method to convert MD-generated atomic configurations into the MPF interfacial profiles; this enables us to perform MPF grain growth simulations in succession to MD nucleation simulation. Furthermore, using the proposed method, MPF and MD grain growth simulations from the same initial structure are directly compared, via which the difference between each simulation result is quantified. Through the detailed investigation of the causes of the difference, a way to improve the accuracy of the MPF model is discussed.
In this study, we aim to achieve more accurate and efficient prediction of grain growth by exploiting the merits of atomistic and continuum simulations. To this end, we propose a method to convert MD-generated atomic configurations into the MPF interfacial profiles; this enables us to perform MPF grain growth simulations in succession to MD nucleation simulation. Furthermore, using the proposed method, MPF and MD grain growth simulations from the same initial structure are directly compared, via which the difference between each simulation result is quantified. Through the detailed investigation of the causes of the difference, a way to improve the accuracy of the MPF model is discussed.