A detailed understanding of the interaction of oxygen with Pt and other transition metal surfaces is essential for understanding the catalytic activity and deactivation in oxidizing environments. Using first-principle density functional theory (DFT) calculations, we can calculate the stable phases of oxygen and other adsorbates on late transition metal catalysts. Cluster expansion (CE) models have become a valuable complement to traditional first-principle studies by allowing the study of a wider range of system compositions and more efficient sampling of configurational space than DFT alone. However, the process of parameterizing these CE models from first-principle data remains a challenge, both in determining which clusters to include in the CE and in ensuring the quality of predictions for configurations outside of the fitting set. We show that CE fitting methods can be studied independently of first-principle calculations in order to determine the best CE fitting method for a model system, representing atomic adsorbates on the closed-packed (111) surface of a face-center-cubic transition metal. Comparing three different approaches the structure inversion method fitting algorithm in the ATAT software, a genetic algorithm, and a steepest descent (SD) fitting algorithm, we find that the SD algorithm was able to achieve the lowest overall fitting error independent of the complexity of the model system. We then apply the SD fitting method with first-principle data to study oxygen adsorbates on a close-packed (111) Pt surface.

Comparison of cluster expansion fitting methods for the study of oxygen on Pt(111)

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.