High‐throughput methods using composition and structure spread libraries

M any properties of multicomponent materials of interest in chemical engineering, e.g., viscosity, catalytic activity, stability, etc. . ., depend on the material composition and the environment in which the material is used. The ability to predict quantitatively how these properties depend on the composition of the material and on environmental variables is essential to enabling the optimal design of materials and operating conditions, as well as prediction of material’s performance under suboptimal conditions. Unfortunately, the necessary measurements of material properties vs. composition are not commonly available and the ability to calculate their properties at arbitrary compositions is extremely challenging. Traditionally, it has been time-consuming and costly to map materials properties across composition and environmental conditions. This is a combinatorial problem, which quickly becomes intractable because there are many relevant degrees of freedom (material components, environmental components, temperature, pressure, etc. . .) which are enumerated by continuous variables. The conventional approach to mitigate this problem is to limit the number of components, compositions or conditions, and to perform discrete experiments to sample the relevant degrees of freedom. Nearly all progress in chemical engineering, materials science and other materials related disciplines has resulted from this approach, but there is a growing recognition that it is slow and expensive. As technologies advance, the expectations of materials properties become increasingly stringent and the types of materials that meet these expectations become increasingly complex and difficult to design. It takes 10–20 years to translate a new material from discovery into commercial applications. The recently announced Materials Genome Initiative (MGI) is a large effort funded by the US government to stimulate the acceleration of new materials from discovery to commercialization at twice the traditional speed and at a fraction of the cost. There are several approaches to the MGI, including computational, experimental, and hybrid approaches that use both experiments and computations. These are having a significant impact on corporate R&D processes and they will impact research across the Chemical Engineering discipline. This perspective describes the application of combinatorial libraries that are continuous composition spreads of metallic alloys, or spreads of another property such as surface structure or nanoparticle size, and is not intended to be a comprehensive review of high-throughput methods. We refer interested readers to the following published reviews in materials, heterogeneous catalysis, homogeneous catalysis, and automotive applications. For high-throughput computational approaches to these areas we refer to these published reviews. The work described in this perspective differs from most of these reviews in the utilization and spatial characterization of samples with continuous or near-continuous gradients in some property, e.g., composition, structure, or particle size. In this perspective, we describe innovations in experimentation and computations by our research groups and others that have the potential to contribute significantly to the development of multicomponent materials for engineering applications. We describe new experimental methods and instrumentation that enable the synthesis and characterization of samples with gradients in alloy composition, surface structure, nanoparticle size, etc. . . and their characterization across composition, structure and size space using spatially resolved analytical tools. This reduces the need to make numerous discrete samples, perform discrete experiments, and simultaneously results in richer and higher quality datasets defining structure/composition/property relationships. We will illustrate the utility of these methods with examples from studies of alloy corrosion/oxidation resistance, alloy catalysis and electrocatalysis. These examples show how properties vary across composition, surface structure, and nanoparticle size spaces, and in some cases under different environmental conditions. We will also show how computational simulation has been integrated into these studies, providing guidance on what properties to study, and confirmation of how to interpret datasets that span entire alloy composition spaces. We end with our thoughts on the opportunities and the needs of high throughput methodologies as applied in chemical engineering. Correspondence concerning this article should be addressed to J. R. Kitchin at jkitchin@andrew.cmu.edu.