Machine-Learning Nonadiabatic Couplings for Reactive Photochemical Dynamics

Support Information for Machine-learning nonadiabatic couplings enable reactive photochemical reaction dynamics provides datasets, compiled models, and trajectory data for machine learning applications in non-adiabatic molecular dynamics. The repository includes training, validation, and test sets for models, alongside JSON files containing trajectory data from simulations like DBH with 500 and 5000 trajectories. This data supports the development and validation of ML models for simulating excited-state chemical reaction pathways.

Use Cases

• Training machine learning models for nonadiabatic couplings using provided training/validation/test sets
• Validating reactive dynamics simulations against ML-NAMD trajectory data in JSON format
• Benchmarking model performance on specific reactions like DBH across different trajectory counts (e.g., 500 vs 5000)
• Analyzing excited-state reaction pathways using the compiled molecular dynamics data

Strengths

• Dataset is substantial in size, approximately 18.2 GB, indicating detailed simulation data
• Clear CC-BY-4.0 license facilitates open reuse and modification
• Structured into distinct components: datasets for model development, compiled models, and trajectory data for validation

Limitations

• Column names and specific data schema are not provided, limiting immediate understanding of data structure
• Row counts for datasets are unspecified, making scale of individual components unclear
• Metadata is sparse, with no description of the specific machine learning models or simulation methodologies used

Provenance

Source

Leticia Adao Gomes

Collection Method

Likely generated from computational chemistry simulations and machine learning model training

Time Range

null

Freshness

Last updated 2026-03-24

Geography

null