Tutorials on Machine Learning Tools for Simulations in Electrocatalysis
TL;DR
This tutorial-style article demonstrates seven ML approaches that accelerate or augment simulations relevant to the oxygen reduction reaction on platinum-based catalysts, including machine-learned exchange–correlation functionals, Gaussian-process optimizers, Bayesian configuration search, and a range of ML interatomic potentials capable of geometry optimization and molecular dynamics at near-DFT accuracy.
Abstract
Machine learning (ML) is becoming an integral part of modern electrocatalysis, complementing and extending traditional density functional theory (DFT) simulations. Yet practical guidance on how these methods operate within realistic catalytic workflows remains limited. In this tutorial-style article we demonstrate seven ML approaches that accelerate or augment simulations relevant to the oxygen reduction reaction on platinum-based catalysts. These include machine-learned exchange–correlation functionals, Gaussian-process optimizers, Bayesian configuration search, and a range of ML interatomic potentials capable of geometry optimization and molecular dynamics at near-DFT accuracy. Each approach is illustrated on a chemically meaningful structural model and implemented in a reproducible Python notebook format. The tutorial highlights where ML currently provides clear benefit, where its accuracy is bounded by the underlying training data, and how these tools integrate with standard DFT workflows. Thus, the tutorial offers a compact and accessible introduction to usable ML tools for electrocatalysis and outlines the opportunities and practical limitations of their application in simulations of catalysts.