Formation ability descriptors for high-entropy diborides established through high-throughput experiments and machine learning

Abstract

Establishing formation ability descriptors is essential for facilitating the discovery and design of high-entropy diborides (HEBs) with tailorable properties. In this work, we establish an efficient combination of formation ability descriptors of HEBs through the high-throughput synthesis and calculations combined with the machine learning approach. Firstly, 70 HEB samples are synthesized by the self-developed high-throughput combustion synthesis apparatus, and 22 formation ability descriptors are computed simultaneously. According to these results, four formation ability descriptors, namely Pauling electronegativity difference (δχP), average volume, average bulk modulus, and density difference, are chosen and proposed as the optimal combination of HEBs with 97.1% validation accuracy through Pearson correlation coefficient analysis and machine learning method. In particular, the δχP is the most significant feature among them and the smaller δχP is, the more accessible synthesis of HEBs is. This study is expected to facilitate the discovery and design of HEBs with tailorable properties.