EPFL researchers reproduce early precipitation behavior and phase-transition competition in Mg–Nd alloys with MLIP

1/ AI 활용 방법: 머신러닝 원자간 포텐셜(MLIP)
2/ 적용 과제: Mg–Nd 합금의 석출 초기 단계와 상전이 경쟁을 원자 수준에서 시뮬레이션
3/ 보고된 결과: 실험에서 관찰된 복잡한 초기 석출 거동을 재현하고, 질서-무질서 및 구조 변환 경쟁과 hcp↔bcc 연속 전이 가능성을 시사
4/ The EPFL researchers parameterized a machine learning interatomic potential (MLIP) for multicomponent-alloy solid-state precipitation modeling and reproduced the early precipitation behavior of dilute Mg–Nd alloys through large-scale atomistic simulation.
5/ The study used a phase-transition-path enumeration algorithm together with weighted Kendall-τ metrics to evaluate the accuracy of low-temperature thermodynamics.
6/ From a materials perspective, this result shows that order-disorder transitions and structural transformations compete in complex early precipitation processes, and that continuous transitions between hcp and bcc are possible.
7/ By reproducing experimentally observed behavior, it supports MLIP as an effective computational tool for alloy precipitation and phase-transformation path analysis.
8/ However, the summarized result is limited to a specific alloy system and simulation conditions.
9/ In addition, the low-temperature thermodynamic accuracy evaluation depends on the path enumeration and metrics used here, so further verification is needed for generalization to other compositions or temperature ranges. [1]