AI reveals how magnesium thermoelectrics improve through natural self-optimization

Magnesium-based thermoelectric compounds are valued for their eco-friendly properties and wide availability, but finding high-performance candidates has been slow due to reliance on trial-and-error experimentation. Researchers at Beihang University have now combined high-throughput computational screening with machine learning to accelerate discovery.

Their breakthrough shows that thermal expansion itself significantly enhances thermoelectric efficiency. As magnesium crystals heat and expand, atomic spacing increases, boosting lattice anharmonicity and lowering thermal conductivity. At the same time, electron band structures become more concentrated, raising carrier effective mass and improving the Seebeck coefficient. Together, these effects drive up the figure of merit ZT.

To uncover this mechanism, the team mined magnesium-based crystal structures from the open quantum materials database, then conducted density functional theory simulations to create a large dataset. Five machine learning models were evaluated, with XGBoost chosen for its predictive accuracy and fast screening capacity. This workflow provides a reliable tool for optimizing magnesium-based thermoelectrics.

The study not only identifies how thermal expansion governs performance but also sets out a quantitative pathway for future design strategies. Beyond magnesium systems, the findings could guide development of a wide range of advanced thermoelectric materials.

Research Report:Machine Learning-Guided Design of High-Performance Mg-based Thermoelectrics: Insights into Thermal Expansion Effects