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.

Computational materials discovery has been successful in predicting novel, technologically relevant materials. However, it has remained focused almost exclusively on finding ground-state structures. Now that the lower-hanging fruit has been found in many fields of application, materials exploration is moving toward metastable materials: higher energy phases that are stable at practical time scales. Because of the challenges associated with predicting which phases are realistic, this class of materials has remained relatively unexplored, despite numerous examples of metastable structures with unmatched properties (e.g., diamond). This article highlights recent advances in developing computational and theoretical methods for predicting useful and realizable metastable materials. Topics discussed cover (1) the latest strategies for identifying potential metastable phases, (2) methodologies for assessing which phases can be realized experimentally, and (3) current approaches to estimate the lifetime of metastable materials.

To extend rational materials design and discovery into the space of metastable polymorphs, rapid and reliable assessment of their lifetimes is essential. Motivated by the early work of Buerger (1951), here we investigate the routes to predict kinetics of polymorphic transformations using solely crystallographic arguments. As part of this investigation we developed a general algorithm to map crystal structures onto each other and construct transformation pathways between them. The developed algorithm reproduces reliably known transformation pathways and reveals the critical role that the dissociation of chemical bonds along the pathway plays in choosing the best (low-energy barrier) mapping. By utilizing our structure-mapping algorithm we were able to quantitatively demonstrate the intuitive expectation that the minimal dissociation of chemical bonds along the pathway, or in ionic systems, the condition of coordination of atoms along the pathway not decreasing below the coordination in the end compounds, represents the requirement for fast transformation kinetics. We also demonstrate the effectiveness of the structure-mapping algorithm in combination with the coordination analysis in studying transformations between polymorphs for which a detailed atomic-level picture is presently elusive.