Deep Learning-Driven Molecular Dynamics for the Prediction of 3d Transition Metal Multi-principal Alloys Structure
摘要
Multi-principal element alloys (MPEAs) have attracted much attention in recent years due to the optimization of performance by adjusting the contents and ratios of different elements. Since the phase structure of MPEAs that determine the macroscopic properties are directly affected by the solidification process, the construction of an accurate solidification model is the basis and key to the prediction and optimization of alloy properties. Therefore, this work establishes a microscopic model that can enable to predict the solidification structure of 3d transition metal MPEAs exactly. We have calculated the interatomic interactions between 3d transition MPEAs using first principle calculation as a dataset for deep potential molecular dynamics to construct a new force field which provide an accurate potential function for the molecular dynamics to achieve an accurate prediction of the atomic configurations of 3d transition metal MPEAs after solidification. The results show that laminar dislocations were formed near the Co-rich and Fe-rich body-centered cubic phases along the densely arranged surfaces. The FeCoNi laminar dislocation energy is lower than that of pure Ni and NiFe, which is more likely to form laminar dislocations. It suggests that the compromise between accuracy and cost can be effectively solved by our strategy established by solidification structure microscopic modeling, which provides an effective method for phase formation prediction of MPEAs, and contributes to the design and further development of MPEAs.