Boosting Heterogeneous Catalyst Discovery by Structurally Constrained Deep Learning Models

TL;DR

This paper enhances graph neural network models for catalyst discovery by integrating Voronoi tessellation-based features and atomic properties, significantly improving energy prediction accuracy on catalyst datasets.

Contribution

It introduces a novel GNN embedding method using Voronoi tessellation and atomic features, advancing catalyst energy prediction accuracy.

Findings

Improved mean absolute error to 651 meV per atom on Open Catalyst dataset.

Achieved 6 meV per atom error on intermetallics dataset.

Data selection influences prediction accuracy, surpassing physical thresholds.

Abstract

The discovery of new catalysts is one of the significant topics of computational chemistry as it has the potential to accelerate the adoption of renewable energy sources. Recently developed deep learning approaches such as graph neural networks (GNNs) open new opportunity to significantly extend scope for modelling novel high-performance catalysts. Nevertheless, the graph representation of particular crystal structure is not a straightforward task due to the ambiguous connectivity schemes and numerous embeddings of nodes and edges. Here we present embedding improvement for GNN that has been modified by Voronoi tesselation and is able to predict the energy of catalytic systems within Open Catalyst Project dataset. Enrichment of the graph was calculated via Voronoi tessellation and the corresponding contact solid angles and types (direct or indirect) were considered as features of edges and Voronoi volumes were used as node characteristics. The auxiliary approach was enriching node representation by intrinsic atomic properties (electronegativity, period and group position). Proposed modifications allowed us to improve the mean absolute error of the original model and the final error equals to 651 meV per atom on the Open Catalyst Project dataset and 6 meV per atom on the intermetallics dataset. Also, by consideration of additional dataset, we show that a sensible choice of data can decrease the error to values above physically-based 20 meV per atom threshold.