.png){kind=icon}
.png){kind=icon}
Return
A relaxation-free training framework for task-oriented modular potentials in multi-principal element alloys
.png){kind=icon}
PRE
AI
F
F
J
W
P
DOI：10.1007/s11431-025-3298-1.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Medium entropy alloys (MEAs) exhibit exceptional mechanical properties that are closely governed by stacking-fault-related energetics, which are commonly described using the generalized stacking fault energy (GSFE). However, efficient and accurate evaluation of GSFE in chemically complex MEAs remains challenging due to chemical disorder and strong local relaxation effects. Here, we propose an efficient relaxation-free training framework for developing modular deep potential (DP) models that enables task-oriented potential construction tailored to specific physical domains by decomposing complex material behaviors into physically relevant subspaces and optimizing the potential for each targeted task. As a representative case, a DP model is trained on density functional theory datasets of face-centered-cubic (FCC) and stacking-fault configurations in CoCrNi, using a training strategy that enables direct prediction of relaxed energies from unrelaxed structures. In this work, particular attention is paid to the intrinsic stacking fault (ISF), which corresponds to a critical local minimum on the GSFE landscape and governs partial dislocation behavior and therefore represents a key descriptor for defect-mediated plasticity. The trained potential achieves high accuracy for both energies and atomic forces, as validated against first-principles reference data. Molecular dynamics (MD) simulations demonstrate that the DP model reproduces the target FCC and ISF energies and generalizes effectively to related defect configurations governed by stacking fault mechanisms. These results establish the feasibility of modular potential construction within a relaxation-free training framework for chemically complex alloy systems and bridge first-principles accuracy with large-scale simulations. The framework provides an efficient route to analyze defect energetics and their coupling with chemical short-range order, while also establishing a basis for quantitative assessments of twinning propensity and other defect-mediated deformation processes in chemically complex alloys.
Keywords:
medium entropy alloys
generalized stacking fault energy
deep potential
density functional theory
chemical short-range order
Journal
IF:
4.9
Papers:
4.9K
Citations:
9.9K
Organization
Cited Papers