DMREF: Collaborative Research: Design of surface functionality through surface composition and structure

多组分材料，如合金，服务于国家基础设施、制成品、硬件技术和国防设备的关键方面。常见的例子是黄铜（铜和锌的混合物）和不锈钢（铁、碳和其他元素如铬和镍的混合物）。为特定应用设计合金的主要挑战是了解其有用特性（成本、硬度、密度、耐腐蚀性、化学性质等）如何取决于其特定的元素混合。随着元素数量的增加，测试可能组合的范围需要制备和研究呈指数增长的合金样品。该项目将开发和应用现代研究工具，以大大提高合金性能可以通过实验测量和/或通过计算模拟预测的速度。感兴趣的应用是催化生产环氧丙烷，这是一种价值数十亿美元的商品化学品。研究人员将使用高通量实验方法同时收集100种不同的二元或三元合金催化剂组成的催化反应数据，而不是一次收集一种催化剂组成。这些实验方法将与通过快速计算模拟预测催化活性的机器学习方法相结合。一旦这些计算方法的性能达到实验基准，它们就可以用于设计具有更复杂成分、结构和形态的合金催化剂。更重要的是，一旦它们在催化材料设计中的实用性得到确立，这些工具就可以应用于各种各样的应用，加速各种工程、产品设计和硬件技术的多组分合金的设计。本DMREF项目将在CuxAgyAu（1-x-y）催化剂表面上使用丙烯环氧化制环氧丙烷，以改进催化剂表面设计中合金的选择和优化。这里研究的系统将能够有效地剪裁变量，如表面取向和结构；由于偏析，块状和表面元素组成的关系；以及作战环境对地表结构和组成的影响。合金催化研究的高通量方法将利用成分扩散合金膜（csaf）作为材料库，包含平行于其表面的Cu， Au和Ag的梯度，从而在单个膜上发现许多成分。再加上高通量多通道微反应器阵列，催化活性将在100种合金成分下同时测量，在一系列温度和反应物组成下。此外，将测量合金的表面成分，以确定偏析对反应性的影响。这些数据将用于验证催化活性的高通量计算模拟。将开发机器学习方法来加速基于密度泛函理论衍生的原子间势的原子模拟，这有望加速过滤合金表面催化活性位点的许多可能的原子构型的过程。这些建模方法为催化活性的进一步实验测量和跨越合金成分空间的表面的原位表征提供了信息；CuxAgyAu（1-x-y）与0x1和0y1-x。研究人员希望这些研究能够产生新的计算工具来预测合金的催化活性，从而实现严格和有效的催化剂优化。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。

Multicomponent materials, such as alloys, serve in critical aspects of the nation's infrastructure, manufactured goods, hardware technologies, and defense apparatus. Common examples are brass (mixture of copper and zinc) and stainless steels (mixtures of iron, carbon and other elements such as chromium and nickel). The main challenge of designing an alloy for a particular application is understanding how its useful characteristics (cost, hardness, density, corrosion resistance, chemical properties, etc.) depend on its specific mix of elements. As the number of elements increases, testing the range of possible combinations requires the preparation and study of an exponentially increasing number of alloy samples. The project will develop and apply modern research tools to vastly increase the rate at which the properties of alloys can be measured experimentally and/or predicted by computational simulation. The application of interest is the catalytic production of propylene oxide, a multibillion-dollar commodity chemical. The investigators will use high-throughput experimental methods to collect catalytic reaction data from 100 different binary or ternary alloy catalyst compositions concurrently, rather than one catalyst composition at a time. These experimental methods will be integrated with machine-learning methods that predict catalytic activity via rapid computational simulations. Once the performance of these computational methods has been benchmarked against experiment, they can be deployed for design of alloy catalysts with more complex compositions, structures and morphologies. More importantly, once their utility in the design of catalytic materials is established, such tools can be applied to a wide variety of applications, accelerating the design of multicomponent alloys for a wide variety of engineering, product design, and hardware technologies.This DMREF project will use the propylene epoxidation to propylene oxide on CuxAgyAu(1-x-y) catalyst surfaces in order to improve alloy selection and optimization in the context of catalytic surface design. The system investigated here will enable efficient tailoring of variables such as surface orientation and structure; the relationship between elemental compositions in the bulk and on the surface due to segregation; and the influence of operational environments on surface structure and composition. High-throughput methods for alloy catalysis study will make use of composition spread alloy films (CSAFs) as material libraries, containing gradients of Cu, Au, and Ag parallel to their surfaces, such that many compositions are found over a single film. Coupled with a high throughput multichannel microreactor array, the catalytic activity will be measured at 100 alloy compositions concurrently, across a range of temperatures and reactant feed compositions. In addition, the surface composition of the alloy will be measured to determine the effects of segregation on reactivity. These data will serve to verify high-throughput computational simulations of catalytic activity. Machine-learning methods will be developed to expedite atomistic simulations based on interatomic potentials derived from Density Functional Theory, which is expected to accelerate the process of filtering the many possible atomistic configurations of catalytically active sites at an alloy surface. These modeling methods inform further experimental measurements of catalytic activity and in-situ characterization of surfaces spanning alloy composition space; CuxAgyAu(1-x-y) with 0x1 and 0y1-x. The investigators expect these studies to yield new computational tools for predicting catalytic activity of alloys, enabling rigorous and efficient catalyst optimization.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.