Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Jul 29·5 min read
Dual-Level Atomic and Coordination Geometry Learning for Crystal Property Prediction Using Graph Neural Networks
TL;DR
A new model called the Coordination Polyhedron Graph Network (CPGN) enhances crystal property prediction by jointly learning atomic and coordination geometry representations. CPGN outperforms existing models, achieving a formation-energy MAE of 0.060 eV/atom and a band-gap MAE of 0.292 eV on benchmark datasets.
✦ Why It Matters
Consider integrating CPGN into your materials research workflow for improved property predictions.
Key Takeaways
- Explore CPGN for enhanced crystal property predictions.
- Utilize multi-scale GNNs to capture complex material structures.
- Incorporate coordination polyhedra in your modeling approach.
- Leverage benchmark datasets to validate model performance.
How It Works
CPGN constructs three coupled graphs: an atom graph for elemental and bonding information, a line graph for angular interactions, and a coordination polyhedron graph that describes local environments. This multi-scale approach allows the model to learn both atomic-level and coordination-level features, improving the representation of crystal structures.