CAREER: Understanding metal/support interactions in catalysis with statistical learning

由支撑在氧化物表面的金属纳米颗粒组成的非均相催化剂对于促进燃料和化学品的快速高效制造以及控制环境污染物的排放至关重要。催化剂的研究和开发历来主要依赖于实验方法——通常涉及时间和资源密集的材料筛选。然而，近年来，随着理论、计算方法和人工智能的进步，人们越来越多地致力于先进催化剂技术的预测设计。该项目开发并应用了基于机器学习的先进模拟工具，该工具将用于制定控制金属/支持物相互作用的策略，以提高催化剂的稳定性、活性和选择性。用这些工具建立的模型将加速研究工作，以设计由孤立的金属原子或分散在高表面积金属氧化物载体上的小簇组成的新型催化剂。除了提高催化剂的反应活性、产品选择性和稳定性外，该项目还将有助于发现和设计催化剂，以减少广泛用于化学制造和污染控制的昂贵和战略性金属（如铂、钯和铑）的数量。除了改进催化技术外，该项目还将通过莱斯大学的塔皮亚卓越与教育公平中心为服务欠缺的社区建立一个外展项目，从而扩大教育工作。已知金属纳米颗粒和氧化物载体之间的协同相互作用会影响催化剂的性能。该项目的目标是开发一个理论框架，该框架使用基于统计的机器学习和密度泛函理论来理解多相催化中的金属/支撑相互作用（msi）。第一个目标是应用计算效率高的统计学习方法来识别金属在氧化物载体上结合的物理描述符，这反过来可以用来预测金属烧结速率和簇形态。第二个目标是通过构建msi与吸附质结合能和相关的动力学障碍之间的模型来预测催化活性和选择性，这些动力学障碍控制了典型模型反应（如一氧化碳氧化）的机制。第三个目标是应用热力学分析来了解战略性地引入以增强msi的支持修改的稳定性。因此，整个方法不仅将通过调整msi来确定控制催化行为所需的表面修饰，而且还将预测这种修饰最容易实现的环境。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Heterogeneous catalysts consisting of metal nanoparticles supported on oxide surfaces are essential for promoting rapid and energy efficient manufacturing of fuels and chemicals, as well as controlling emissions of environmental pollutants. Catalyst research and development has historically relied primarily on experimental methods – often involving time- and resource-intensive materials screening. In recent years, however, increasing effort has been directed toward predictive design of advanced catalyst technologies, enabled by advances in theory, computational methods, and artificial intelligence. The project develops and applies advanced simulation tools, based on machine learning, which will be used to establish strategies for controlling metal/support interactions to enhance catalyst stability, activity, and selectivity. Models built with these tools will accelerate research efforts to design new types of catalysts consisting of either isolated metal atoms or small clusters dispersed on high-surface area metal oxide supports. In addition to improving catalyst reactivity, product selectivity, and stability, the project will enable discovery and design of catalysts that reduce the amount of expensive and strategic metals (such as platinum, palladium, and rhodium) used widely in chemical manufacturing and pollution control. Beyond improving catalytic technologies, the project will extend educational efforts by establishing an outreach program for underserved communities through the Tapia Center for Excellence and Equity in Education at Rice University. Synergistic interactions between metal nanoparticles and oxide supports are known to influence catalyst performance. The goal of the project is to develop a theoretical framework that uses statistical-based machine learning, together with density functional theory, to understand metal/support interactions (MSIs) in heterogeneous catalysis. The first objective is to apply computationally efficient statistical learning methodologies to identify the physical descriptors of metal binding on oxide supports, which in turn can be used to predict metal sintering rates and cluster morphology. The second objective is to predict catalytic activity and selectivity by constructing models that relate MSIs to adsorbate binding energies and the associated kinetic barriers that control mechanisms for prototypical model reactions, such as carbon monoxide oxidation. The third objective is to apply thermodynamic analyses to understand the stability of support modifications that are strategically introduced to enhance MSIs. Thus, the overall methodology not only will identify desirable surface modifications for controlling catalytic behavior by tuning MSIs, but also will predict the environments in which such modifications can be achieved most readily.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.