On the importance of catalyst-adsorbate 3D interactions for relaxed energy predictions

TL;DR

This study explores predicting relaxed energies of catalyst-adsorbate systems without detailed geometric information, revealing that models can still perform well despite missing some structural details, thus aiding accelerated materials discovery.

Contribution

It demonstrates that machine learning models can predict relaxed energies without explicit adsorbate-catalyst geometric data, challenging the necessity of detailed binding site information.

Findings

Models retain decent accuracy without binding site info

Removing geometric edges impacts but does not eliminate prediction quality

Suggests potential for faster materials discovery workflows

Abstract

The use of machine learning for material property prediction and discovery has traditionally centered on graph neural networks that incorporate the geometric configuration of all atoms. However, in practice not all this information may be readily available, e.g.~when evaluating the potentially unknown binding of adsorbates to catalyst. In this paper, we investigate whether it is possible to predict a system's relaxed energy in the OC20 dataset while ignoring the relative position of the adsorbate with respect to the electro-catalyst. We consider SchNet, DimeNet++ and FAENet as base architectures and measure the impact of four modifications on model performance: removing edges in the input graph, pooling independent representations, not sharing the backbone weights and using an attention mechanism to propagate non-geometric relative information. We find that while removing binding site information impairs accuracy as expected, modified models are able to predict relaxed energies with remarkably decent MAE. Our work suggests future research directions in accelerated materials discovery where information on reactant configurations can be reduced or altogether omitted.