VASE: A High-Entropy Alloy Short-Range Order Structural Descriptor for Machine Learning.

Area of Science:

• Materials Science
• Computational Materials Science
• Alloy Design

Background:

• Short-range order (SRO) in high-entropy alloys (HEAs) significantly influences their properties.
• Density functional theory (DFT) calculations are vital for studying SRO but are computationally intensive.
• Machine learning (ML) offers a pathway for rapid estimation of DFT results in HEA research.

Purpose of the Study:

• To propose a novel descriptor for characterizing SRO in HEAs.
• To enhance the accuracy and efficiency of predicting HEA properties using ML.
• To investigate the atomic spatial arrangement beyond composition and interactions.

Main Methods:

• Development of the Voronoi Analysis and Shannon Entropy (VASE) descriptor.
• Integration of VASE with machine learning models for property prediction.
• Comparison of VASE with existing descriptors like Coulomb matrices and radial distribution functions.

Main Results:

• The VASE descriptor accurately captures atomic spatial arrangement information in HEAs.
• ML models trained with VASE descriptors show superior accuracy in predicting formation energy for the FeCoNiAlTiCu system.
• The VASE-based model achieved the best predictive performance for unrelaxed structures, with an error of 24.06 meV/atom.

Conclusions:

• The VASE descriptor provides an effective representation of atomic arrangement, crucial for understanding SRO in HEAs.
• This new descriptor significantly improves the predictive power of machine learning models for HEA properties.
• VASE is a powerful tool for advancing the investigation of SRO phenomena in complex alloy systems.