Spin-Aware DFT Trajectories for Heterogeneous Catalysis

13.5 million DFT calculation trajectories model approximately 5,000 materials and 47,000 intermediate-catalyst systems. The dataset provides high-fidelity data for developing machine learning potentials in heterogeneous catalysis, created by SandboxAQ and released in October 2025.

Use Cases

• Train spin-aware machine learning potentials using the 13.5 million DFT trajectories for catalyst property prediction.
• Model reaction pathways in heterogeneous catalysis by analyzing intermediate-catalyst systems across the 47,000 identified configurations.
• Benchmark computational chemistry models against high-fidelity DFT calculations for the 5,000 included materials.

Strengths

• 13.5 million DFT calculation trajectories provide a substantial data volume for model training.
• Covers approximately 5,000 distinct materials and 47,000 intermediate-catalyst systems, offering broad chemical diversity.
• Designed for high-fidelity calculations, complementing existing large-scale datasets with potentially greater accuracy.

Limitations

• Specific column definitions and data schema are not provided, complicating direct analysis.
• The dataset's focus on DFT trajectories may require specialized domain knowledge in computational chemistry for effective use.

Provenance

Source

SandboxAQ

Collection Method

Contains Density Functional Theory (DFT) calculation trajectories.

Freshness

Last updated on 2025-10-30.