Machine-learning interatomic potential for W-Mo alloys.

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Development

Background:

• Tungsten-molybdenum (W-Mo) alloys are crucial for high-temperature applications.
• Accurate simulation of their properties requires reliable interatomic potentials.
• Existing potentials may not capture the full range of compositional and defect behaviors.

Purpose of the Study:

• To develop a novel machine-learning interatomic potential for W-Mo random alloys.
• To accurately predict key material properties across all compositions.
• To investigate the influence of alloying on fundamental defect properties.

Main Methods:

• Utilized the Gaussian Approximation Potential (GAP) framework.
• Trained the potential using density functional theory (DFT) data from the Vienna Ab initio Simulation Package (VASP).
• Incorporated an adjusted Ziegler-Biersack-Littmarck potential for short-range interactions, fitted with all-electron DFT data.

Main Results:

• The developed machine-learning potential accurately predicts elastic properties, melting points, and point defect behavior for W-Mo alloys.
• Investigated the effect of alloying on threshold displacement energies.
• Observed a significant dependence of threshold displacement energies on the local chemical environment and the primary recoiling atom's element.

Conclusions:

• The new machine-learning potential provides a robust tool for simulating W-Mo alloys.
• Alloying in W-Mo systems significantly influences radiation damage resistance via threshold displacement energies.
• Understanding these dependencies is critical for designing advanced W-Mo based materials.