(616b) General Prediction of Reaction Pathway Energetics on Alloy Surfaces Using a Latent-Variable Machine Learning Architecture

Most approaches to high-throughput screening of catalysts rely on simple descriptors to estimate catalytic performance. While a descriptor-based approach can be quite useful, descriptors can oversimplify the design problem, and take effort to develop. The goal of this work is to predict the energetics of entire reaction pathways on many alloy surfaces, which allows elucidation of pathways and direct calculation of rates and selectivities using kinetic models. However, most machine learning models are focused on predicting the adsorption energies of just a few species. Here, we develop and apply a machine learning architecture that allows the prediction of many reaction intermediates and transition states on many alloy surfaces, allowing prediction of the energetics of entire reaction pathways.

To develop a general model, we used a new machine learning architecture. This architecture partially decouples the adsorbed species from the surface by introducing latent variables, which are a set of implicit surface properties that control how the surface interacts with all adsorbates. Each adsorbate's stability has a different dependence on the latent variables. These latent variables are learned during the fitting process. We create separate sub-models for predicting these latent variables for each surface element, and separate sub-models for how different adsorbates respond to these latent variables. The sub-models are all fit simultaneously. Our framework takes advantage of the fact that elements are discrete entities, and greatly simplifies the huge combinatorial challenge of considering many possible adsorbed species and many possible alloy surfaces. In the end, our method allows efficient prediction of the energetics of entire reaction pathways on alloy surfaces, and can be reused in new contexts (i.e., for a new reaction). We apply this framework to multiple reactions and show how it can quickly elucidate pathways and increase the efficiency of computational catalysis studies.