Large language models (LLMs) are becoming increasingly applied beyond natural language processing, demonstrating strong capabilities in complex scientific tasks that traditionally require human expertise. This progress has extended into materials discovery, where LLMs introduce a new paradigm by leveraging reasoning and in-context learning, capabilities absent from conventional machine learning approaches. Here, we present a Multi-Agent-based Electrocatalyst Search Through Reasoning and Optimization (MAESTRO) framework in which multiple LLMs with specialized roles collaboratively discover high-performance single atom catalysts for the oxygen reduction reaction. Within an autonomous design loop, agents iteratively reason, propose modifications, reflect on results and accumulate design history. Through in-context learning enabled by this iterative process, MAESTRO identified design principles not explicitly encoded in the LLMs' background knowledge and successfully discovered catalysts that break conventional scaling relations between reaction intermediates. These results highlight the potential of multi-agent LLM frameworks as a powerful strategy to generate chemical insight and discover promising catalysts.
Multi-Task Mixture Density Graph Neural Networks for Predicting Cu-based Single-Atom Alloy Catalysts for CO2 Reduction Reaction
Tsinghua University(清华大学);Tsinghua-Berkeley Shenzhen Institute(清华-伯克利深圳学院);Tsinghua Shenzhen International Graduate School(清华大学深圳国际研究生院);Shenzhen Geim Graphene Center(深圳盖姆石墨烯中心);The Chinese University of Hong Kong(香港中文大学);Tencent(腾讯);Zhejiang University(浙江大学);Zhejiang Lab(之江实验室)
Coordination Engineering of Dual-Atom Catalysts for Overall Water Splitting: Mechanistic Insights from Constant-Potential First-Principles and Machine Learning
Anhui Province Key Laboratory of Intelligent Computing and Applications, College of Physics and Electronic Engineering, Huaibei Normal University(淮北师范大学物理与电子工程学院安徽省智能计算与应用重点实验室);Huaibei Key Laboratory for Advanced Thin Film Materials and Technology, Huaibei Normal University(淮北师范大学淮北先进薄膜材料与技术重点实验室)
Structural flexibility dictates reactivity of single-atom catalysts
CEITEC – Central European Institute of Technology, Brno University of Technology(中欧技术学院,布尔诺理工大学);Faculty of Mechanical Engineering, Brno University of Technology(布尔诺理工大学机械工程学院)
A non-equilibrium strategy for the general synthesis of single-atom catalysts
State Key Laboratory of Nuclear Physics and Technology, School of Physics, Peking University(北京大学物理学院核物理与技术国家重点实验室);School of Materials Science and Engineering, Peking University(北京大学材料科学与工程学院);Department of Chemistry, Capital Normal University(首都师范大学化学系)
Integrating Atomic Scale Catalyst Design with Transport Engineering for Stable and Efficient CO2 Electrolysis to CO in a Membrane Electrode Assembly
Department of Chemical Engineering, McMaster University(麦克马斯特大学化学工程系);Canadian Center for Electron Microscopy(加拿大电子显微镜中心);Department of Applied Chemistry and Center for Emergent Functional Matter Science, National Yang Ming Chiao Tung University(国立阳明交通大学应用化学系与新兴功能物质科学中心)
Potential-Induced Dynamic Coordination of Nonmetal Atoms Directly Bound to Metal Centers in Graphene-Embedded Single-Atom Catalysts and Its Implications
Anhui Provincial Collaborative Innovation Center for Advanced Functional Composite Materials, College of Physics and Electronic Engineering, Huaibei Normal University(淮北师范大学物理与电子工程学院安徽省先进功能复合材料协同创新中心)
Defect-Engineered h-BN as a Platform for Single-Atom HER Catalysts: Descriptor Screening Refined by Electrochemical Stability Analysis
University of Belgrade − Faculty of Physical Chemistry(贝尔格莱德大学物理化学学院);Luleå University of Technology(吕勒奥理工学院);Serbian Academy of Sciences and Arts(塞尔维亚科学与艺术院)
Accelerating the Discovery of g-C$_3$N$_4$-Supported Single Atom Catalysts for Hydrogen Evolution Reaction: A Combined DFT and Machine Learning Strategy
由单原子催化(SAC)支撑的二维材料有望取代铂,用于可持续制氢的大规模工业可扩展性。在这里,一系列的金属(Al、Sc、Ti、V、Cr、Mn、Fe、Ni、Cu、Zn)和非金属通过DFT计算和六种机器学习(ML)算法筛选出嵌入g-C $_3$ N $_4$不同活性位点的(B,C,N,O,F,Si,P,S,Cl)单原子(支持向量回归,梯度提升回归,随机森林回归,AdaBoost回归,多层感知器回归,岭回归)。基于形成能、吉布斯自由能和能带分析的结果表明,B、Mn和Co的单原子锚定在g-C $_3$ N $_4$上可以作为高效的产氢活性中心。基于支持向量回归(SVR)的ML模型表现出最好的性能,准确和快速地预测吉布斯氢吸附自由能($Δ$ GH)的测试集与较低的平均绝对误差(MAE)和高决定系数(R $^2$)的0.45和0.81,分别。基于SVR模型的特征选择突出了前五个主要特征:形成能、键长、沸点、熔点和价电子作为关键描述符。总之,通过DFT计算结合ML模型用于有效筛选g-C $_3$ N $_4$基单原子催化的潜在析氢反应(HER)的多步工作流程可以显著地有助于催化剂设计和制造。
Transition States Energies from Machine Learning: An Application to Reverse Water-Gas Shift on Single-Atom Alloys
Center for Interstellar Catalysis, Department of Physics and Astronomy, Aarhus University(星际催化中心,物理天文系,奥胡斯大学);Aarhus Institute of Advanced Studies, Aarhus University(奥胡斯先进研究所,奥胡斯大学)
In Situ Detection of Active Edge Sites in Single-Layer MoS$_2$ Catalysts
MoS 2纳米颗粒是加氢脱硫和析氢等工艺的成熟催化剂,但在催化工作条件下解开其原子级结构仍然具有重大挑战性。环境压力X-射线光电子能谱学(AP-XPS)使我们能够跟踪原位形成的催化相关的二硫化钼边缘网站在其活性状态。XPS指纹被描述为独立的贡献的Mo 3d核心水平的光谱,其相对强度是敏感的热力学条件。采用密度泛函理论(DFT)对Au(111)表面的三角形MoS 2颗粒进行了模拟,并确定了其边缘位置的硫化状态。一个一致的图片出现在边缘钼原子的核心能级位移的发展反直觉向更高的结合能时,减少活跃的边缘。这种转变的解释是一个令人惊讶的改变,在金属字符的边缘网站,这是一个独特的光谱特征的二硫化钼边缘工作条件下。
Combining first-principles modeling and symbolic regression for designing efficient single-atom catalysts in Oxygen Evolution Reaction on Mo$_2$CO$_2$ MXenes
该研究结合第一性原理建模、ESSI 分析与符号回归,筛选 Mo$_2$CO$_2$ MXenes 上的 OER 单原子催化剂,发现 Cu 过电位最低,并提取关键描述符与过电位公式。
The 3d Transition Metal Single Atom Catalysis for Prominent Adsorption Performance Towards Cr(VI) : A First Principle Study
通过合适的氧化剂和纳米催化剂在污染土壤和水中将Cr(VI)还原为Cr(III)的还原反应因其环境效益而受到广泛关注。单原子催化剂作为一种新兴的催化剂,在同时吸附去除Cr(VI)方面显示出巨大的潜力。在这项工作中,我们研究了3d过渡金属(从Sc到Zn)SAC的吸附效率,通过理论方法去除Cr(VI)。采用密度泛函理论(DFT)研究了Cr(VI)(H2 CrO 4和Cr2 O 72-)在活性炭上的吸附行为. Ti-N-C对H2 CrO 4和Cr2 O 72-具有良好的去除性能,对水的净化具有重要的环保意义。此外,我们的研究结果表明,H2 CrO 4吸附的SAC系统的结合能被发现是远远小于为Cr2 O 72-吸附,表明还原Cr(VI)在Cr2 O 72-是积极有利的比,Cr(VI)在H2 CrO 4。