Active Learning Guided Discovery of High Entropy Oxides Featuring High H₂-production

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• High entropy oxides (HEOs) are complex solid solutions with vast potential but challenging design due to high-dimensional composition spaces.
• Traditional HEO design relies on expert knowledge and intuition, limiting efficient exploration of compositional possibilities.

Purpose of the Study:

• To introduce an active learning (AL) strategy for efficient exploration of HEO compositional space.
• To accelerate the discovery of novel HEOs with desirable catalytic properties.

Main Methods:

• Implemented a closed-loop active learning system involving iterative "Training, Prediction, and Experiment" stages.
• Utilized X-ray photoelectron spectroscopy and density functional theory for material characterization and analysis.

Main Results:

• Identified four novel, stable high entropy oxides through the AL strategy.
• Achieved a hydrogen evolution rate of 251 μmol g_cat^-1 min^-1, outperforming established catalysts like Pt/γ-Al₂O₃ and Cu/ZnO/Al₂O₃.
• Observed a loss of elemental identity in the discovered HEOs, confirmed by spectroscopy and DFT.

Conclusions:

• The active learning strategy effectively accelerates the discovery of high entropy oxides.
• Machine learning provides a powerful tool for navigating complex compositional spaces and identifying advanced catalytic materials.
• The discovered HEOs demonstrate significant potential for applications in hydrogen evolution reactions.