The Open Catalyst 2022 (OC22) Dataset and Challenges for Oxide Electrocatalysts

Abstract

The development of machine learning models for electrocatalysts requires a broad set of training data to enable their use across a wide variety of materials. One class of materials that currently lacks sufficient training data is oxides, which are critical for the development of Oxygen Evolution Reaction (OER) catalysts. To address this, we developed the Open Catalyst 2022 (OC22) dataset, consisting of 62,331 Density Functional Theory (DFT) relaxations (∼9,854,504 single point calculations) across a range of oxide materials, coverages, and adsorbates. We define generalized total energy tasks that enable property prediction beyond adsorption energies; we test baseline performance of several graph neural networks; and we provide predefined dataset splits to establish clear benchmarks for future efforts. In the most general task, GemNet-OC sees a ∼36% improvement in energy predictions when combining the chemically dissimilar Open Catalyst 2020 Data set (OC20) and OC22 datasets via fine-tuning. Similarly, we achieved a ∼19% improvement in total energy predictions on OC20 and a ∼9% improvement in force predictions in OC22 when using joint training. We demonstrate the practical utility of a top performing model by capturing literature adsorption energies and important OER scaling relationships. We expect OC22 to provide an important benchmark for models seeking to incorporate intricate long-range electrostatic and magnetic interactions in oxide surfaces. Data set and baseline models are open sourced, and a public leaderboard is available to encourage continued community developments on the total energy tasks and data.

Molecular Understanding and Practical In Silico Catalyst Design in Computational Organocatalysis and Phase Transfer Catalysis—Challenges and Opportunities

“…Several general-purpose MLPs suitable for organocatalysis have been reported in the literature, for instance, ANI-2x [ 225 ], DimeNet++ [ 226 ], GemNet [ 227 ], and SchNet [ 228 ]. The Open Catalyst project also provides state-of-the-art MLPs that can be used out-of-the-box, although these MLPs are targeted toward heterogeneous and electro-catalysis [ 18 , 19 ].…”

“…Practical is a relative term that depends on the computational resources available to the research group. However, it is reasonable to assume that most research groups would not have exclusive access to computing resources equivalent to the one used to generate the open-catalyst project data [ 18 , 19 ]. As such, continued synergetic development in hardware and software is still required to reach a practical in silico catalyst design accessible to the average research group.…”

WhereWulff: A Semiautonomous Workflow for Systematic Catalyst Surface Reactivity under Reaction Conditions

“…At this stage ( Figure 1 -I), WhereWulff exhibits another case of decision-making ability: it prioritizes the slab models that show the highest contribution to the nanoparticle shape. For each of the prioritized slab models, WhereWulff leverages the adsorption site finder 44 ( Figure 1 -J) to identify potential adsorption sites on the clean surface. This is accomplished by exploiting the Wyckoff and equivalent positions in the bulk and overlaying them onto the surface to locate the exposed TMs at the interface based on a decrease in the coordination number.…”

Quantifying the Limits of Methane Activation in Cu-exchanged Zeolites using Reactive and Interpretable Machine Learning based Potentials

“…Although computational screening studies are now routine for metals and alloys [70][71][72][73] , similar studies for zeolites are not as straightforward due to the diversity of local chemical environments. For instance, while sampling high symmetry sites (e.g., ontop, bridge, fcc, hcp) is often sufficient to study the catalytic properties of metals and alloys, even a simple CHA unit cell consists of 31 unique [CuOCu] 2+ configurations.…”

“…[79][80][81] Similarly, the computational catalysis community has reported several ML models to predict adsorption energy-based descriptors to approximate the kinetics of surface-mediated reactions. 72,73 However, to the best of our knowledge, this work represents the first example of using reactive MLPs to identify the transition state geometries and screen the catalytic performance of thousands of zeolite-based active sites at DFT accuracies.…”