文献类型：期刊文献

英文题名：Interpretable Machine Learning for Optimizing Electrocatalytic Hydrogen Evolution via Rational Design of Electrolyte Environments

作者：Fan, Sihan[1]; Zhao, Hongyang[2]; Gao, Yang[1]; Pan, Likun[2]; Xuan, Fu-Zhen[1,3]

机构：[1] School of Mechanical and Power Engineering, Shanghai Key Laboratory of Intelligent Sensing and Detection Technology, East China University of Science and Technology, Shanghai, 200237, China; [2] Shanghai Key Laboratory of Magnetic Resonance, School of Physics, Institute of Magnetic Resonance and Molecular Imaging in Medicine, East China Normal University, Shanghai, 200241, China; [3] Key Laboratory of Pressure Systems and Safety, China

年份：2026

外文期刊名：SSRN

收录：EI(收录号：20260175662)

语种：英文

外文关键词：Additives - Density functional theory - Electrocatalysis - Electrodes - Electrolytes - Hydrogen - Hydrogen bonds - Hydrogen evolution reaction - Machine learning - Molecular dynamics - Optimization - Positive ions - Reaction kinetics - Shrinkage

摘要：The rational design of electrolyte configurations is pivotal for optimizing the hydrogen evolution reaction (HER), however, the quantitative deconstruction of the intricate "cation effect" remains a formidable challenge. In this work, we developed a robust machine learning (ML) model framework to predict HER overpotentials across diverse electrolyte configurations. A high-fidelity database comprising 2367 samples was established, encompassing a broad spectrum of ionic species and electrode materials. By employing the least absolute shrinkage and selection operator for feature selection and extreme gradient boosting for predictive modeling, exceptional prediction accuracy was achieved, demonstrated by a coefficient of determination of 0.974 and a mean absolute percentage error of 0.094. The integration of Shapley additive explanations analysis, molecular dynamics simulations, and density functional theory calculations identified cationic hydration radius and atomic weight as the predominant descriptors governing HER performance. It was found that "structure-breaking" cations (e.g., K+) hinder reaction kinetics by disrupting the interfacial hydrogen-bonding network and blocking active sites. In contrast, "structure-making" cations (e.g., Li+) maintain an ordered interfacial water environment, thereby facilitating proton transfer. Furthermore, the generalizability of our ML model was experimentally validated through the identification of Sr2+ as a high-performance additive due to the fact that its experimental overpotential on a Pt electrode at pH=0 was only 35.4 mV at 10 mA cm-2, showing excellent agreement with the predicted value. This work delivers a high-fidelity tool for electrolyte screening and profound insights into interfacial ion-water-electrode interactions, facilitating the highly-efficient design of advanced electrocatalytic systems. ? 2026, The Authors. All rights reserved.