Development of a deep potential model for FePt alloys: DFT-level accuracy in high-temperature mechanical simulations

KY Chen and CH Lin and HQ Li and CY Hou and YZ Wu and L Ma and JY Ye and J Rong and XH Yu and ZH Liu and J Feng, MATERIALS TODAY COMMUNICATIONS, 48, 113401 (2025).

DOI: 10.1016/j.mtcomm.2025.113401

FePt alloys with L10-ordered structures are promising candidates for next-generation hypersonic thermal-end components due to their excellent high-temperature stability and oxidation resistance. However, the lack of atomic scale understanding of their deformation and failure mechanisms under extreme thermal environments hinders their practical application. Conventional interatomic potentials such as Embedded Atom Method (EAM) face limitations in accurately describing high-temperature phase transitions, dislocation evolution, and interfacial behavior, making them insufficient for predictive modeling under complex conditions. The Deep Potential (DP) method, combining first-principles accuracy with molecular dynamics efficiency, offers a powerful alternative. In this work, a DP model for FePt alloys is constructed over a wide temperature range (300-1200 K) using first-principles calculations and the DP-GEN active learning framework. The model is rigorously validated and used to investigate high-temperature mechanical properties. Results show that FePt alloys achieve a tensile strength of 18 GPa at 1200 K, with a 35-percent-point lower strength degradation rate compared to MEAM predictions. The DP model also predicts a fracture strain of 32 %, significantly higher than traditional potentials and consistent with experimental ductility. This work provides new atomic- level insight into FePt alloy mechanics and supports their optimization for extreme thermal environments.

Return to Publications page