A QCAR-approach to materials modeling

Abstract

Little is known about the relationship between the function and structure of materials. Materials (solids with a function) are complex entities and a better knowledge of the parameters that contribute to function is desirable. Here, we present modeling approaches that correlate chemical composition with function of heterogeneous catalysts. The complete composition space of the mixed oxides of Ni–Cr–Mn and of Ni–Co–Mo–Mn (10% spacing) have been measured for the oxidation of propene to acroleine. The data have been collected, visualized and modeled. Different mathematical approaches such as Support Vector Machines, multilevel B-splines approximation and Kriging have been applied to model this relationship. High-throughput screening data of ternary and quaternary composition spreads are approximated to locate catalysts of high activity within the search space. For quaternary systems, slice plots offer a good tool for visualization of the results. Using these approximation techniques, the composition of the most active catalysts can be predicted. The study documents that distinct relationships between chemical composition and catalytic function exist and can be described by mathematical models.