Pourbaix Machine Learning Framework Identifies Acidic Water Oxidation Catalysts Exhibiting Suppressed Ruthenium...

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Green hydrogen production relies on efficient oxygen evolution reaction (OER) catalysts.
• Iridium, a critical OER catalyst component, faces supply concerns.
• Developing stable and active iridium-free OER catalysts is crucial.

Purpose of the Study:

• To identify novel, stable, and active OER catalysts using a machine learning-aided computational approach.
• To screen a large database of mixed metal oxides for potential OER applications.
• To experimentally validate promising catalyst candidates.

Main Methods:

• Machine learning pipeline trained on >36,000 mixed metal oxides to predict Pourbaix decomposition energy.
• Screening of 2070 new metallic oxides for stability under acidic conditions.
• Experimental characterization of Ru$_{0.6}$Cr$_{0.2}$Ti$_{0.2}$O$_{2}$ including overpotential, durability tests, and in situ/ex situ spectroscopy.

Main Results:

• Identified Ru$_{0.6}$Cr$_{0.2}$Ti$_{0.2}$O$_{2}$ as a promising OER catalyst candidate.
• Achieved an overpotential of 267 mV at 100 mA cm$^{-2}$ with sustained operation for >200 h.
• Demonstrated reduced Ru dissolution (20×) and suppressed lattice oxygen participation (>60%) compared to RuO$_{2}$.

Conclusions:

• Machine learning effectively screens for stable and active OER catalysts.
• Ru$_{0.6}$Cr$_{0.2}$Ti$_{0.2}$O$_{2}$ offers enhanced activity and durability for green hydrogen production.
• Ti incorporation improves stability, while Cr enhances catalytic activity.