(84j) Predicting Surface Coverage Effects in Heterogeneous Catalysis Via an Interaction-Counting Approach

Accurately calculating adsorbate-adsorbate interactions is crucial for predicting in-situ properties of catalysts and reaction mechanisms at the catalyst interface. Numerous efforts have been made to accurately predict adsorbate-adsorbate interactions at the surface of transition metals, such as through the use of cluster expansion (CE) methods,^1,2 thereby improving the ability to describe reaction energetics under varying chemical potentials. Although CE has relatively high prediction accuracy (0.01 – 0.001 eV), its computational cost (often thousands of DFT calculations) remains a key barrier to its broadest implementation in high-throughput screening studies.

In this work, we provide a novel, computationally efficient (10-100 DFT calculations), and generalizable interaction-counting approach to more effectively include adsorbate interactions in catalyst screening methods. Our model enables prediction of interaction energies of simple adsorbates, including CO, O, CH₃, and CCH₃, on various transition metal surfaces with sufficient accuracies (~0.01-0.10 eV) for implementation in applications relying on general interaction correlations (e.g., microkinetic modeling). The predicted interaction parameters can be correlated to the nature of the host transition metal. Finally, we leverage the use of leave-one-out cross validation to enable our methodology to predict generalizable and reduced training sets (~15-50 configurations) for extensions to new/untrained materials. This methodology is further extended for predicting interaction energies of adsorbates with different identities, including O-CO interactions on Au(100) and Au(111) surfaces.

References:

1. 1. Sanchez, J. M. et al. Phys. A: Stat. Mech. Appl. 128, 334-350 (1984).
2. 2. Schmidt, D. J. et al. J. Chem. Theory Comput. 8, 264–273 (2011).
3. 3. Roling, L. T. et al. J. Phys. Chem. C 121, 23002-23010 (2017).