Scholarlink

<p>The metal–support interaction (MSI) critically influences the performance of supported nanocatalysts and their long-term stability, yet the factors governing MSIs are multifaceted and challenging to sort out. Here we combine first-principles neural network molecular dynamics (NN-MD) simulations with interpretable machine learning (iML) to shed light on the factors determining MSIs for Pt nanoparticles on diverse metal–oxide supports. Our approach reveals the atomic-scale dynamics of sintering mechanisms and identifies key features of oxide supports governing MSI. We find that the surface energy, surface oxygen bond order, surface dipole and work function of the support are dominant in Pt–oxide interactions. Leveraging these insights, we screened promising sinter-resistant supports for Pt nanoparticles from over 10,000 metal–oxide surfaces and validated some cases by Monte Carlo simulations and experiments. This work integrates iML with NN-MD to accelerate the understanding and discovery of stable supported nanocatalysts, and should be broadly applicable to numerous catalytic applications.</p><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

Predictive model for the discovery of sinter-resistant supports for metallic nanoparticle catalysts by interpretable machine learning

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • Chenggong Jiang,
  • Bill Yan,
  • Bryan R. Goldsmith,
  • Suljo Linic

摘要

The metal–support interaction (MSI) critically influences the performance of supported nanocatalysts and their long-term stability, yet the factors governing MSIs are multifaceted and challenging to sort out. Here we combine first-principles neural network molecular dynamics (NN-MD) simulations with interpretable machine learning (iML) to shed light on the factors determining MSIs for Pt nanoparticles on diverse metal–oxide supports. Our approach reveals the atomic-scale dynamics of sintering mechanisms and identifies key features of oxide supports governing MSI. We find that the surface energy, surface oxygen bond order, surface dipole and work function of the support are dominant in Pt–oxide interactions. Leveraging these insights, we screened promising sinter-resistant supports for Pt nanoparticles from over 10,000 metal–oxide surfaces and validated some cases by Monte Carlo simulations and experiments. This work integrates iML with NN-MD to accelerate the understanding and discovery of stable supported nanocatalysts, and should be broadly applicable to numerous catalytic applications.