From Synthesis to Machine Learning Validation for the Thermoelectric Zintl Phase Ba1-xEuxZn2-yAlySb2 System

Abstract

Five new Zintl phase compounds in the Ba1-xEuxZn2-yAlySb2 system were synthesized using the molten Pb-flux method for thermoelectric application. All compounds crystallized in the orthorhombic BaCu₂S₂-type phase with a phase selectivity determined by the r⁺/r⁻ radius ratio criterion. Detailed structural analysis revealed a complex three-dimensional anionic framework of [(Zn/Al)Sb4] tetrahedra combined with the Ba/Eu mixed-cations occupying the central voids within the frameworks. Density functional theory calculations proved that the particular amounts of Eu and Al co-substitutions stabilized the given structure type and altered the electronic structural environment, resulting in reduced carrier mobility and ultra-low thermal conductivity. Thermoelectric property measurements showed that the enhanced Seebeck coefficients and the minimized thermal conductivities yielded a maximum ZT of 0.50 at 653 K. Three machine learning models were developed by XGBRegressor algorithms to predict electrical conductivity, Seebeck coefficient, and thermal conductivity, respectively. These models were initially trained using a customized dataset of 5,798 TE properties and demonstrated high predictive accuracy (R² up to 0.87), which closely matched the trends of experimental results and highlighted the impact of structural features. Herein, we established a powerful workflow for accelerating thermoelectric materials discovery by integrating experimental synthesis with density functional theory analysis and machine learning validation. Therefore, the currently investigated approach successfully demonstrated a vital synergy where machine learning was used to validate and interpret complex experimental results, thus successfully closing the loop from laboratory findings to data-driven understanding.