.png){kind=icon}
.png){kind=icon}
Return
Accelerating Discovery in Computational Electrocatalysis: From Machine Learning Potentials to AI-Driven Workflows
.png){kind=icon}
PRE
AI
DOI：10.1016/j.coelec.2026.101848.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Machine learning (ML) is rapidly reshaping computational electrocatalysis. Conventional Density Functional Theory (DFT) cannot simultaneously provide quantum-level accuracy and the large spatiotemporal scales required to model electrified interfaces under operando conditions. This review surveys recent advances in overcoming these limitations, with a primary focus on machine learning potentials (MLPs). We highlight constant-potential MLPs that enable grand-canonical simulations of electrified interfaces, capturing dynamic phenomena such as catalyst restructuring, cation effects, and nuclear quantum effects. We further discuss how the emerging convergence of large language models, multi-agent systems, and automated experimentation is opening new horizons for accelerating electrocatalysis research. By critically evaluating current achievements and remaining challenges, we outline how integrating MLP-driven simulations with AI-driven automation may enable the rational design of next-generation electrocatalysts.
Keywords:
machine learning potentials
computational electrocatalysis
grand-canonical simulations
AI-driven workflows
electrified interfaces
Journal
IF:
6.9
Papers:
1.7K
Citations:
8.6K