Open Catalyst 2020 (OC20) Dataset and Community Challenges

Catalyst discovery and optimization is key to solving many societal and\nenergy challenges including solar fuels synthesis, long-term energy storage,\nand renewable fertilizer production. Despite considerable effort by the\ncatalysis community to apply machine learning models to the computational\ncatalyst discovery process, it remains an open challenge to build models that\ncan generalize across both elemental compositions of surfaces and adsorbate\nidentity/configurations, perhaps because datasets have been smaller in\ncatalysis than related fields. To address this we developed the OC20 dataset,\nconsisting of 1,281,040 Density Functional Theory (DFT) relaxations\n(~264,890,000 single point evaluations) across a wide swath of materials,\nsurfaces, and adsorbates (nitrogen, carbon, and oxygen chemistries). We\nsupplemented this dataset with randomly perturbed structures, short timescale\nmolecular dynamics, and electronic structure analyses. The dataset comprises\nthree central tasks indicative of day-to-day catalyst modeling and comes with\npre-defined train/validation/test splits to facilitate direct comparisons with\nfuture model development efforts. We applied three state-of-the-art graph\nneural network models (CGCNN, SchNet, Dimenet++) to each of these tasks as\nbaseline demonstrations for the community to build on. In almost every task, no\nupper limit on model size was identified, suggesting that even larger models\nare likely to improve on initial results. The dataset and baseline models are\nboth provided as open resources, as well as a public leader board to encourage\ncommunity contributions to solve these important tasks.\n