Machine Learning Accelerated Discovery of Entropy-Stabilized Oxide Catalysts for Catalytic Oxidation

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• Unary to ternary metal oxides catalysis is well-explored.
• High entropy metal oxides (HEOs) offer vast unexplored catalytic potential.
• The combinatorial complexity of HEOs hinders traditional discovery methods.

Purpose of the Study:

• To leverage machine learning for efficient discovery of HEO catalysts.
• To identify key correlations between crystal phase and catalytic activity in HEOs.
• To develop robust predictive models for HEO catalyst screening.

Main Methods:

• Analysis of crystal phase and catalytic performance in ACr2O4 systems.
• Development of high-accuracy machine learning models for element importance sequencing.
• Strategic use of negative data and curated training sets for regression models.
• Screening of HEOs using predictive models.

Main Results:

• A strong correlation was found between a single spinel phase and good catalytic activity for methane oxidation.
• Machine learning models achieved high accuracy (cross-validation score >0.7) in predicting catalytic performance.
• A novel HEO catalyst, Ni0.04Co0.48Zn0.36V0.12Cr2O4, was identified.
• The discovered HEO exhibited excellent sulfur and moisture resistance and long-term stability (>7000 h).

Conclusions:

• Machine learning is a viable and powerful tool for discovering HEO catalysts.
• The single spinel phase is a critical factor for effective methane oxidation catalysis in HEOs.
• The developed methodology enables rapid screening and identification of advanced HEO catalysts for specific applications.