Predicting the stability of ternary intermetallics with density functional theory and machine learning

Area of Science:

• Materials Science
• Computational Chemistry
• Solid State Physics

Background:

• Ternary compounds with AB2C2 stoichiometry are crucial in materials science.
• Exploring novel stable compounds within known structural prototypes is an ongoing challenge.
• High-throughput calculations and machine learning offer powerful tools for materials discovery.

Purpose of the Study:

• To explore the potential for novel stable ternary compounds with the AB2C2 composition.
• To investigate the feasibility of using machine learning combined with high-throughput density-functional theory (DFT) for materials discovery.
• To identify promising candidates for new metallic and non-magnetic materials.

Main Methods:

• Utilized high-throughput density-functional theory (DFT) calculations.
• Employed machine learning techniques to accelerate the screening process.
• Focused on two common intermetallic prototypes: tI10-CeAl2Ga2 and tP10-FeMo2B2.

Main Results:

• Identified approximately 10 times more stable ternary compounds than previously known within the studied prototypes.
• The majority of the newly discovered stable compounds are predicted to be metallic and non-magnetic.
• Machine learning reduced computational costs by approximately 75%.

Conclusions:

• The combination of machine learning and DFT is effective for discovering new stable ternary compounds.
• Significant potential exists for novel materials within the AB2C2 composition, particularly metallic and non-magnetic ones.
• Limitations in machine learning predictions were noted for compounds involving second-row elements or magnetic properties.