GBProject: A comprehensive database of fundamental grain-boundary properties in metals
Abstract
Artificial intelligence/machine learning (AI/ML)-driven materials design through grain-boundary (GB) engineering has emerged as a promising strategy for developing next-generation metals and alloys. However, its broader application has been limited by the lack of comprehensive GB databases needed to train robust AI/ML models for accurately predicting GB structures and properties across diverse crystallographic and chemical spaces. In this study, we develop a comprehensive GB database through high-throughput atomistic simulations of 28 elemental metals using both semi-empirical and machine-learning interatomic potentials. The current dataset contains 3519 face-centered cubic, 3910 body-centered cubic, and 315 hexagonal close-packed GBs, spanning four representative GB types: symmetric-tilt, twist, asymmetric-tilt, and mixed (tilt–twist) GBs. All GB structures are generated using a coincident-site lattice-based workflow implemented in the open-access Grain-Boundary Project website (GBProject.org). For each GB, two key interfacial properties, including GB excess free energy and free volume, are systematically calculated. Both the optimized GB structures and the associated property data are publicly available through the GBProject website. This GB database provides a reliable and reusable resource for validating theoretical models, benchmarking atomistic simulations, and evaluating state-of-the-art foundation models for large-scale GB simulations. It also establishes a solid foundation for developing advanced AI/ML models to accelerate the design of polycrystalline materials.