Accelerated discovery of high entropy alloys assisted by machine learning

Wang Kun
Alfred University, Alfred, NY 14802, United States

EXTENDED ABSTRACT: Herein, we proposed a strategy to design single-phase refractory high entropy alloys (RHEAs) with the assistance of machine learning algorithms. Based on an extensive dataset (1807 entries) built in this work, we applied multiple machine learning algorithms to train the dataset. After the blind test, we found that the Gradient boosting model can distinguish the single-phase-solid solution and non-single phasesolid solution alloys with a test accuracy of 96.41%. Given the GB model, we predicted over 100 equiatomic oxidation-resistance RHEAs from the composition space of eight metallic elements. After that, we synthesized ten of these predicted single-phase RHEAs by mechanical alloying. The XRD patterns show that all of them are single-phase BCC solid solution. The experimental results agree well with the prediction results, indicating the excellent performance of the machine learning model in single-phase RHEAs prediction. In addition, the most critical feature that is relevant the single-phase solid solution formation in RHEAs is δ (atomic size difference) through permutation importance analysis. The importance order of the features is shown in Figure 1. With the aid of the machine learning method, single-phase oxidation-resistant RHEAs were successfully designed. Our work presents a novel strategy with outstanding performance and evident effectiveness on the accelerated discovery of novel metallic materials used for extreme environments.

REFERENCES
[ 1 ] Yo n g g a n g Ya n , D a n L u , K u n Wa n g * ,
Computational Materials Science, 99, (2021) 110723