IDENTIFICATION OF STABLE NEAR SURFACE ALLOY SYSTEMS USING DENSITY FUNCTIONAL THEORY AND DATA SCIENCE

Recently, a new class of catalysts, called single-atom alloy (SAA), has been discovered, synthesized, and studied. These heterogeneous catalysts contain a single, highly active promoter metal that sits within the surface of a less reactive host metal. The promising potential of these SAAs was demonstrated by Sykes and co-workers for selective hydrogenation reactions using a single palladium atom in copper. Our research group has also found multiple varieties of SAAs that are promising candidates for many industrial processes. However, there are only a few studies on the stabilities of these SAAs. We used Density Functional Theory (DFT) and the Atomic Simulation Environment (ASE) to refine and expand the work by Ruban et al. and determined if a given promoter metal will cluster on the surface, diffuse to the subsurface, stay as an ad-atom, or form an SAA within any other host metal. Then, we used data science and statistical analyses to extract heuristic rules from our database of the stability of SAAs. By using the open source machine learning package TensorFlow, an artificial neural network has been trained to classify the alloys as stable/unstable systems, showing the relation between the alloy’s stability and its component’s properties. These guidelines provide fundamental insight into the formation of stable surface structures of binary alloys where one component is a minority species, helping scientists and researchers to focus their efforts on stable SAAs in their quest to develop new catalyst for industrial applications.