Universal framework for efficient estimation of stability in multi-principal element alloys.

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Design

Background:

• Predicting the synthetic accessibility of multi-principal element alloys (MPEAs) is a significant challenge in materials science.
• Exploring the vast chemical space of MPEAs requires efficient predictive models.

Purpose of the Study:

• To develop and validate a physical model for predicting the synthesizability and stability of MPEAs.
• To enable accurate predictions across broad compositional and structural spaces.

Main Methods:

• Utilizing a physical model that expresses the total energy of MPEAs as a linear combination of energies from lower-dimensional subsystems.
• Validating the model with a large computational dataset (135,791 MPEAs) derived from density functional theory calculations.
• Comparing model accuracy with state-of-the-art deep learning models and cluster-expansion theory.

Main Results:

• Achieved mean absolute errors near or below 7 meV/atom in predicting MPEA energies.
• Demonstrated high accuracy comparable to deep learning models while retaining interpretability.
• Successfully predicted the stability of high-entropy alloys, indicating a flatter energy landscape.

Conclusions:

• The developed physical model offers a reliable and interpretable method for predicting MPEA synthesizability and stability.
• This approach facilitates accelerated discovery and design of novel MPEAs.
• The model's accuracy and interpretability provide valuable insights into the factors governing alloy formation.