物理化学学报 >> 2026, Vol. 42 >> Issue (8): 100227.doi: 10.1016/j.actphy.2025.100227

论文 上一篇    下一篇

机器学习指导筛选多主元合金作为析氢反应电催化剂

伍昊1,2, 李凤麒1, 石欣伟1, 卞海峰1, 周庆1, 贾顺顺1, 马玉洁1, 顾坚1, 张靖梓3,*(), 何水剑2,*(), 孟祥康1,*()   

  1. 1 南京大学现代工程与应用科学学院, 固体微结构物理全国重点实验室, 人工微结构科学与技术协同创新中心, 江苏 南京 210093
    2 南京林业大学材料科学与工程学院, 林业资源高效加工利用协同创新中心, 林产化学与材料国际创新高地, 江苏 南京 210037
    3 哈尔滨工业大学(深圳), 计算机科学与技术学院, 广东 深圳 518055
  • 收稿日期:2025-10-14 修回日期:2025-12-03 录用日期:2025-12-04 发布日期:2026-06-11
  • 通讯作者: Email: zjzhang@hit.edu.cn (张靖梓)shuijianhe@njfu.edu.cn (何水剑)mengxk@nju.edu.cn (孟祥康)

Machine-learning guides discovery of multi-principal element alloys as electrocatalyst for hydrogen evolution reaction

Hao Wu1,2, Fengqi Li1, Xinwei Shi1, Haifeng Bian1, Qing Zhou1, Shunshun Jia1, Yujie Ma1, Jian Gu1, Jingzi Zhang3,*(), Shuijian He2,*(), Xiangkang Meng1,*()   

  1. 1 National Laboratory of Solid State Microstructures, Collaborative Innovation Center of Advanced Microstructures, College of Engineering and Applied Sciences, Nanjing University, Nanjing 210093, Jiangsu Province, China
    2 Co-Innovation Center of Efficient Processing and Utilization of Forest Resources, International Innovation Center for Forest Chemicals and Materials, College of Materials Science and Engineering, Nanjing Forestry University, Nanjing 210037, Jiangsu Province, China
    3 School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen 518055, Guangdong Province, China
  • Received:2025-10-14 Revised:2025-12-03 Accepted:2025-12-04 Published:2026-06-11
  • Contact: Email: zjzhang@hit.edu.cn (Jingzi Zhang)shuijianhe@njfu.edu.cn (Shuijian He)mengxk@nju.edu.cn (Xiangkang Meng)

RichHTML

17

PDF

403

赞

补充材料

1

摘要：

多主元合金凭借组分间的协同作用，展现出卓越的物理化学性质，成为极具潜力的析氢反应(HER)电催化剂候选材料。然而，多主元组分结构的复杂性及系统性机器学习(ML)筛选方法的缺乏，使得电催化剂组分的最佳配比无法确定，这制约了多主元合金电催化剂的合理设计与开发。本研究通过Light Gradient Boosting模型从601种候选合金中筛选出了NbZnCo2多主元合金作为最优候选材料，与Pt/C相比，其成本缩减了约34倍，同时HER活性更优。结合密度泛函理论(DFT)计算与实验验证，证实了ML模型的可靠性。微米级NbZnCo2催化剂在10 mA cm−2电流密度下仅需20 mV超低过电位，并保持60 h的稳定运行。此外，纳米级NbZnCo2颗粒仍保持了优异HER性能，验证了NbZnCo2合金作为HER电催化剂的普适性。本研究构建了"机器学习-密度泛函理论-实验"框架，筛选出高性能HER电催化剂，该方法可扩展至其他电催化反应，为可持续能源转换技术提供了更广阔的应用前景。

关键词: 多主元合金, 析氢反应, 机器学习, 密度泛函理论

Abstract:

Owing to synergistic interactions among their components, multi-principal element alloys manifest remarkable physicochemical properties that render them highly promising candidates for hydrogen evolution reaction (HER) electrocatalysts. Despite extensive experimental investigations, the intricate composition of multi-principal components and the absence of systematic machine learning (ML) screening poses significant challenges in identifying optimal elemental configurations for electrocatalysts, thereby constraining the rational design and development of multi-principal alloy electrocatalysts. In this work, the NbZnCo2 multi-principal component alloy emerges as the optimal candidate from a pool of 601 candidate alloys using the Light Gradient Boosting model, demonstrating approximately 34-fold cost efficiency enhancement over Pt/C while surpassing HER activity. Combined density functional theory (DFT) calculations and experimental validation confirmed the ML model's reliability, with the micrometer NbZnCo2 catalyst achieving an ultralow overpotential of 20 mV at 10 mA cm−2 and remarkable stability over a period of 60 h. Furthermore, the NbZnCo2 nanoparticle retained exceptional HER properties, validating the universality of NbZnCo2 element composition. Our work establishes a synergistic "ML-DFT-Experiment" framework for the precise design of high-performance HER electrocatalysis. This methodology exhibits extensibility to diverse other electrocatalytic processes, thereby broadening the applicability in sustainable energy conversion technologies.

Key words: Multi-principal element alloys, Hydrogen evolution reaction, Machine learning, Density functional theory