Towards large-scale quantum-informed molecular dynamics simulations: implementing a machine learning surrogate for many-body dispersion in polymer melts [Dataset]

A supplement dataset for Chapter 4 of Z. SHEN's doctoral thesis. Several JAX-MD NVT simulation results of a PE melt have been provided. The force field model incorporates TraPPE and a trimmed-SchNet machine learning surrogate model for many-body dispersion (MBD). The datasets primarily consist of raw MD trajectory snapshots stored in npz format together with the corresponding final configurations (.gen) that can be used for simulation restart. Atomic coordinates are reported in Angstrom, while atomic velocities are not included. Depending on the intended post-processing workflow, these datasets can support prospective studies on how MBD affects polymer dynamics, including analyses such as radial distribution functions (RDF), radius of gyration, and velocity power spectra. No dedicated post-processing scripts are provided in this repository. Detailed descriptions of the file naming convention, simulation conditions, and test objectives can be found in README.txt. The dataset for fine-tuning the machine learning model is also provided, which is in the same format as described in the pretrain dataset.