Unified Graph Neural Network Force-field for Materials Across the Periodic Table

A graph neural network force-field for predicting properties of solid-state materials across the periodic table. The dataset is associated with the ALIGNN model and was created by Kamal Choudhary. It is available under an Open Access license.

Use Cases

• Predicting material stability and formation energy based on atomic structure graphs.
• Simulating molecular dynamics for solids using learned interatomic potentials.
• Accelerating materials discovery by screening candidate compounds for target properties.
• Benchmarking graph neural network architectures on materials science tasks.

Strengths

• The model is designed to cover the entire periodic table for solids, suggesting broad applicability.
• The associated ALIGNN model architecture is specifically tailored for atomistic line graph representations.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset scale are unknown, which may limit suitability assessment.

Provenance

Source

Kamal Choudhary, associated with the ALIGNN GitHub repository.

Collection Method

Likely generated from computational simulations or aggregated from existing materials databases.