Half-Heusler Structures with Full-Heusler Counterparts: Machine-Learning Predictions and Experimental Validation

• DOI: 10.1021/acs.cgd.0c00646
• Publication Date: 2020-09-01T19:44:17Z
• Authors: Alexander S. Gzyl; Anton O. Oliynyk; Arthur Mar
• Keywords: Heusler compound; Oversampling

Heusler compounds form a diverse group of intermetallic materials encompassing many compositions and structures derived from cubic prototypes, exhibiting complicated types disorder phenomena. In particular, preparing solid solutions between half-Heusler ABC full-Heusler AB2C offers means to control physical properties. However, as is typical in discovery, they represent only small fraction possible compounds. To address this problem unbalanced data sets, machine-learning model was developed using an ensemble approach involving the synthetic minority oversampling technique predict new likely adopt structures. The training set based on experimental crystal structures, including those nonstoichiometric achieved accuracy 98% validation gave excellent performance terms balanced statistical measures. A subset predicted having existing counterparts then targeted for preparation. Six seven these candidates were successfully synthesized confirmed be