Machine Learning Chemical Guidelines for Engineering Electronic Structures in Half-Heusler Thermoelectric Materials.

Area of Science:

• Materials Science
• Solid State Physics
• Computational Chemistry

Background:

• Half-Heusler materials are promising for thermoelectric devices due to high power factors, linked to electronic band structure valley degeneracy.
• Over 50 semiconducting half-Heusler phases exist, but the impact of chemical composition on electronic structure remains unclear.
• While n-type structures have minima at Γ or X points, p-type structures exhibit diverse valence band maxima at Γ, L, or W points.

Purpose of the Study:

• To compare valence bands across known half-Heusler compounds using high-throughput computation and machine learning.
• To establish chemical guidelines for promoting the W-point to the valence band maximum, enhancing valley degeneracy.
• To develop predictive rules for the valence band maximum location in half-Heusler phases.

Main Methods:

• High-throughput computation was employed to analyze the electronic band structures of numerous half-Heusler compounds.
• Machine learning algorithms were utilized to identify patterns and develop predictive models.
• An "orbital phase diagram" was constructed to categorize electronic structures based on contributing atomic orbitals.

Main Results:

• The study clustered diverse electronic structures into groups using an orbital phase diagram.
• New chemical rules were developed using machine learning to predict valence band maximum locations.
• These rules facilitate the engineering of band structures with improved band convergence and valley degeneracy.

Conclusions:

• The developed chemical rules provide a pathway to design half-Heusler materials with optimized electronic structures for thermoelectric applications.
• This work advances the understanding of structure-property relationships in half-Heusler compounds.
• The findings enable targeted material discovery for high-performance thermoelectric energy conversion.