Machine learning guided discovery of super-hard high entropy ceramics

In the present work, we used machine learning (ML) to explore new compositions of high entropy ceramic (HEC) materials that show improved hardness. Starting from a dataset containing hardness, loads and compositions of 557 ceramic materials including HECs, a ML model was built using random forest (RF) algorithm. The RF-based model successfully reproduced experimental load-hardness behavior of Al2O3, (Hf0.2Zr0.2Ti0.2Ta0.2Nb0.2)B2 and (Hf0.2Zr0.2Ti0.2Ta0.2Mo0.2)B2. Accordingly, the built model was employed to find super-hard HECs.