Accelerating the discovery of high‐entropy hexaborides by data‐driven prediction: From equimolar to non‐equimolar

Abstract Estimating the synthesizability is the prerequisite to discovering novel high‐entropy ceramics with exotic properties. Herein, combined with the high‐throughput experiments and machine learning (ML) methods, the synthesizability of high‐entropy hexaborides (HEB 6 ) is investigated. To construct the database, 100 equimolar quinary HEB 6 samples synthesized using a self‐developed high‐throughput solid‐state reaction technique and 20 potential synthesizability descriptors calculated from fundamental parameters of constituent precursors are simultaneously collected. By employing the ML and the genetic algorithms, an optimal model consisting of five synthesizability descriptors (, , , , and ) is determined for predicting the synthesizability of equimolar HEB 6 with a high validation accuracy (93.0%), and 7 005 new equimolar quinary HEB 6 are then proposed. Moreover, the applicability of our established model on non‐equimolar HEB 6 is explored by the prediction of 30 synthesizability diagrams of non‐equimolar quinary HEB 6 . A high accuracy of 90.9% is further validated by the synthesis experiments of 11 non‐equimolar HEB 6 candidates. Our work establishes an effective ML model for assessing the synthesizability of both equimolar and non‐equimolar HEB 6 and paves a promising way to accelerate the discovery of new synthetic accessible high‐entropy ceramics.