High-Throughput Calculations and Machine Learning for Discovering Halide-Type Solid-State Electrolyte Materials
Zhenming Xu, Yinghui Xia, Yinghui Xia, Huiyu Duan, Jianhao Li, Yixi Lin, Yuqiao Jiang, Xiangmin Feng, Zhenhui Liu, Mingbo Zheng, Yongyao Xia, Yongyao Xia
Abstract
Oxide- and sulfide-based solid-state electrolytes face various intrinsic defect challenges. Developing solid-state electrolyte materials with superior overall performance is critical for the practical implementation of solid-state alkali metal ion batteries. Herein, we utilized high-throughput calculation technique to screen halide-type solid-state electrolytes and resolve the structure–function relationships of halides. A 4 BC 7 -type halides with F 4̅3 m space group possess the Tet-Oct-Tet ion migration channels, always showing the high ionic conductivities. Additionally, halides are much more chemically compatible with some oxide-type cathodes than the common sulfide-type solid-state electrolytes. Machine learning models resolve that halogen anion ionic radius is the most important factor negatively affecting the formation energy of halides, while electronegativity difference between cation and halogen anion is the key positively determining band gap of halides. This work finally screened out six halide-type solid-state electrolytes, including Li 4 YBr 7, Li 4 LaBr 7, Li 4 PrBr 7, Na 5 LuF 8, K 5 LaCl 8, and K 5 NdCl 8, which would well match with oxide cathode and alkali metal anode to construct the high-performance all-solid-state batteries. This work not only further expands the chemistry, structure, and component space of halides but also provides a top-down paradigm and shines a light for the development of halide-type solid-state electrolytes.