基于DFT和机器学习的Pt单原子合金电催化剂的理性设计:应用于NO至NH_(3)的转化  

Rational design of Pt-anchored single-atom alloy electrocatalysts for NO-to-NH3 conversion by density functional theory and machine learning

作　　者：刘洁宇 郭海强 熊钰林 陈星 于一夫[2] 王长洪 Jieyu Liu;Haiqiang Guo;Yulin Xiong;Xing Chen;Yifu Yu;Changhong Wang(College of Engineering,Hebei Provincial Key Laboratory of Information Fusion and Intelligent Control,Hebei Normal University,Shijiazhuang 050024,Hebei,China;Institute of Molecular Plus,School of Science,Tianjin University,Tianjin 300072,China)

机构地区：[1]河北师范大学中燃工学院,河北省信息融合与智能控制重点实验室,河北石家庄050024 [2]天津大学理学院,分子+研究院,天津300072

出　　处：《Chinese Journal of Catalysis》2024年第7期243-253,共11页催化学报（英文）

基　　金：河北自然科学基金(B2022205013,B2022205029).

摘　　要：电催化NO还原反应(NORR)合成氨,作为一项新兴技术,旨在减少NO污染并生产高附加值产品.因此,寻找合适的NORR电催化剂至关重要.单原子合金催化剂(SAACs)因其独特的催化性能和明确的成键环境,是研究催化剂结构-活性关系的理想对象.本文依托密度泛函理论(DFT)计算和机器学习(ML)算法,深入研究了Pt掺杂的SAACs在NORR合成氨反应中的催化机制,并提出了SAACs的理性设计原则.通过ML算法,成功识别了与催化活性密切相关的原子和电子特征,并定量分析了这些特征之间的协同作用.DFT计算和ML算法的联合运用不仅加深了对复杂NORR机理的理解,还为理性设计具有特定活性中心的高效SAACs提供了一种切实可行的策略.本文研究了13种Pt掺杂的SAACs(Pt/TM,TM=Sc,Ti,V,Cr,Mn,Fe,Co,Ni,Cu,Zn,Ru,Rh和Pd)和14种纯过渡金属(包括Sc,Ti,V,Cr,Mn,Fe,Co,Ni,Cu,Zn,Ru,Rh,Pd和Pt)在NORR合成氨反应中的催化机制.首先,考察了Pt/TM的热力学稳定性和在电化学条件下的稳定性.结果表明,Pt/TMSAACs展现出了良好的热力学稳定性.在电化学条件下,当外加电压等于极限电势时,单质Zn在pH>6时稳定存在,而其他纯金属和Pt/TM SAACs则在整个pH范围内均能保持稳定.随后,探讨了这些催化剂对NO的吸附及活化能力.DFT计算发现,NO能够在所有催化剂表面稳定吸附,并且吸附后N-O键长与NO得电子数之间存在显著的线性相关性,这两者可共同作为NO活化的重要指标.接着,考察了NORR的反应机理,并筛选出7种具有较高过电势(UL>-0.25V)的催化剂,包括Cu,Zn,Rh,Pt以及Pt/Cu,Pt/Zn,Pt/RhSAACs.考虑到H2和N2副产物对氨选择性的影响,排除了Zn和Pt/Zn作为潜在的优秀催化剂.在此基础上,在排除贵金属(Rh,Pt和Pt/Rh)基底的SAACs后,筛选出Cu和Pt/Cu作为NORR合成氨反应的高效催化剂.最后,采用随机森林(RF)和基于压缩感知原理的数据驱动(SISSO)方法分析了SAACs的内在特性与NORR性能之间的关Electrochemical NO reduction reaction(NORR)toward NH3 synthesis emerges as a promising approach to eliminate NO pollution and generate high-value-added products simultaneously.Therefore,exploring suitable NORR electrocatalysts is of great importance.Here,we present a design principle to evaluate the activity of single-atom alloy catalysts(SAACs),whose excellent catalytic performance and well-defined bonding environments make them suitable candidates for studying structure-activity relationships.The machine learning(ML)algorithm is chosen to unveil the underlying physics and chemistry.The results indicate that the catalytic activity of SAACs is highly correlated with the local environment of the active center,that is,the atomic and electronic features.The coeffect of these features is quantitatively verified by adopting a data-driven method.The combination of density functional theory(DFT)and ML investigations not only provides an understanding of the complex NORR mechanisms but also offers a strategy to design highly efficient SAACs with specific active centers rationally.

关 键 词：NO还原反应 合成氨 单原子合金催化剂 机器学习 密度泛函理论 

分 类 号：TQ113.2[化学工程—无机化工] TQ426