Scholarlink

<p>Given the diversity in the structures of the substrates and the types of active transition metal atoms, using machine learning to screen two-dimensional conductive MOFs for high-performance OER catalytic materials can significantly save time and cost. This paper collected data from 413 catalysts composed of 20 different carbon substrates and 26 metal elements as the training dataset, constructed a series of two-dimensional conductive MOFs structures, and predicted their OER catalytic performance using the trained model. Additionally, density functional theory was used to systematically study the adsorption of OER intermediates *OH, *O, and *OOH on TMO<sub>X</sub>N<sub>4-X</sub>-OHPTP (TM = Fe, Co, Ni;X = 0–4), and the reaction overpotentials were calculated. The results show that CoO<sub>4</sub>-OHPTP(ML/DFT:0.269&#xa0;eV/0.237&#xa0;eV), CoO<sub>3</sub>N<sub>1</sub>-OHPTP(ML/DFT:0.287&#xa0;eV/0.239&#xa0;eV), CoO<sub>2</sub>N<sub>2</sub>-I-OHPTP(ML/DFT:0.302&#xa0;eV/0.241&#xa0;eV), CoO<sub>2</sub>N<sub>2</sub>-II-OHPTP(ML/DFT:0.293&#xa0;eV/0.295&#xa0;eV), CoO<sub>1</sub>N<sub>3</sub>-OHPTP(ML/DFT:0.336&#xa0;eV/0.300&#xa0;eV), and CoN<sub>4</sub>-OHPTP (ML/DFT:0.311&#xa0;eV/0.33&#xa0;eV) have overpotentials lower than RuO<sub>2</sub>(<i>η</i><sub>*OER</sub> = 0.37&#xa0;eV), which are consistent with the predicted results by machine learning. Through model analysis, apart from the properties of metal atoms, it is found that the number and types of different atoms in the substrate and the bond lengths between TM metal atoms and surrounding coordinating atoms are important descriptors that significantly affect the overpotential. Finally, the trained model was used to predict more 2D c-MOFs(TMO<sub>X</sub>N<sub>4-X</sub>-OHPTP and TMO<sub>X</sub>N<sub>4-X</sub>-TBC), identifying 18 additional potential OER catalysts.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Screening of two-dimensional conductive MOFs as OER catalysts assisted by machine learning

  • 期刊论文

  • 发表日期：

    2025年1月22日

查看全文
  • Yuyang Liu,
  • Jianping Sun,
  • Yuxin Chai,
  • Yu Yang,
  • Zhi Yang

摘要

Given the diversity in the structures of the substrates and the types of active transition metal atoms, using machine learning to screen two-dimensional conductive MOFs for high-performance OER catalytic materials can significantly save time and cost. This paper collected data from 413 catalysts composed of 20 different carbon substrates and 26 metal elements as the training dataset, constructed a series of two-dimensional conductive MOFs structures, and predicted their OER catalytic performance using the trained model. Additionally, density functional theory was used to systematically study the adsorption of OER intermediates *OH, *O, and *OOH on TMOXN4-X-OHPTP (TM = Fe, Co, Ni;X = 0–4), and the reaction overpotentials were calculated. The results show that CoO4-OHPTP(ML/DFT:0.269 eV/0.237 eV), CoO3N1-OHPTP(ML/DFT:0.287 eV/0.239 eV), CoO2N2-I-OHPTP(ML/DFT:0.302 eV/0.241 eV), CoO2N2-II-OHPTP(ML/DFT:0.293 eV/0.295 eV), CoO1N3-OHPTP(ML/DFT:0.336 eV/0.300 eV), and CoN4-OHPTP (ML/DFT:0.311 eV/0.33 eV) have overpotentials lower than RuO2(η*OER = 0.37 eV), which are consistent with the predicted results by machine learning. Through model analysis, apart from the properties of metal atoms, it is found that the number and types of different atoms in the substrate and the bond lengths between TM metal atoms and surrounding coordinating atoms are important descriptors that significantly affect the overpotential. Finally, the trained model was used to predict more 2D c-MOFs(TMOXN4-X-OHPTP and TMOXN4-X-TBC), identifying 18 additional potential OER catalysts.

Graphical abstract