Improved material descriptors for bulk modulus in intermetallic compounds via machine learning

Abstract Bulk modulus is an important mechanical property in the optimal design and selection of intermetallic compounds. In this study, bulk modulus datasets of intermetallic compounds were collected, and the features affecting the bulk modulus of intermetallics were screened via feature engineering. Three features B cal , d B avg , and TIE (corresponding to calculated bulk modulus, mean bulk modulus, and third ionization energy, respectively) were found to be the dominant factors influencing bulk modulus and can be extended to other multi‐component alloys. Particularly, we predicted the bulk modulus with an accuracy of 95% using surrogate machine learning models with the selected features, and these features were also demonstrated to be effective for high‐entropy alloys. Moreover, symbolic regression provided an expression for the relationship between bulk modulus and the screened features. The machine learning models provide a new approach for optimizing and predicting the bulk moduli of intermetallic compounds.