Using Machine Learning to Predict Oxygen Evolution Activity for Transition Metal Hydroxide Electrocatalysts

Area of Science:

• Materials Science
• Electrochemistry
• Machine Learning

Background:

• Electrocatalytic water splitting generates hydrogen and oxygen for clean energy, but the oxygen evolution reaction (OER) is kinetically challenging.
• Hydroxides show promise as OER electrocatalysts due to high activity and surface area, yet precise composition design is hindered by prediction difficulties.
• Quantitative prediction of hydroxide electrocatalytic performance across vast compositional spaces remains a significant challenge.

Purpose of the Study:

• To develop a machine learning model for predicting OER activity in hydroxide electrocatalysts.
• To enable performance-oriented design of hydroxide catalysts by overcoming prediction challenges.
• To explore extensive doping spaces for novel hydroxide electrocatalyst discovery.

Main Methods:

• Employed a random forest algorithm to model the relationship between catalyst composition, morphology, phase, electrolyte pH, electrode type, and OER overpotential.
• Trained and validated the machine learning model on a diverse dataset of hydroxide catalysts.
• Designed and synthesized a novel hydroxide catalyst based on model predictions.

Main Results:

• The machine learning model accurately predicted OER activity with a mean relative error of 4.74%.
• A new hydroxide catalyst, Ni_0.77Fe_0.13La_0.1, was rationally designed and synthesized.
• The designed catalyst exhibited an ultra-low overpotential of 226 mV at 10 mA cm^-2 for OER.

Conclusions:

• This work presents a novel machine learning approach for predicting hydroxide electrocatalyst performance in OER.
• The predictive model facilitates efficient catalyst design and composition optimization.
• The developed method accelerates the discovery of high-performance electrocatalysts for water splitting and clean energy applications.