Work

Machine-learning-driven identification of cooperative descriptors for polysulfide conversion on spinel oxides

Year: 2026
Type: article
Abstract: Spinel oxides have emerged as promising catalysts for polysulfide conversion, yet a comprehensive design strategy that elucidates the underlying structure–property relationship across diverse spinel oxides remains elusive. The complex multistep conversion processes make single descriptors insufficient to capture the full catalytic behavior, whereas identifying cooperative descriptors introduces additional complexity that challenges manual analysis. Here, we employ a machine-learning–assisted approach to uncover universal cooperative descriptors and accelerate the discovery of high-performance spinel catalysts. We reveal that the intrinsic polysulfide conversion efficiency of spinel oxides is cooperatively governed by two structural features. Further screening of promising catalysts is conducted using the extreme gradient boosting (XGBoost) algorithm, achieving rapid and accurate prediction of these features on spinel oxides. The model identifies Fe 2 CoO 4 and Fe 2 NiO 4 as superior catalysts, which are experimentally validated to exhibit excellent performance. This work provides deep mechanistic insights and a powerful data-driven framework for the rational design of spinel catalysts, with potential extension to other applications.
Language: English
Open Access status: gold
Estimated APC paid: $4,500
OpenAlex ID: W7236164813
FWCI:
Cites: 54
Cited by:
Related to: 10