A structured dataset of polymers in molecular dynamics simulations extracted by LLM–assisted text mining

This literature-mined dataset compiles polymer materials studied using molecular dynamics (MD) simulations, spanning diverse polymers, force fields, and simulation-derived properties reported in peer-reviewed publications from 1995-2025. The dataset was constructed using a large language model (LLM)-assisted extraction and curation workflow, with each record traceable to its source publication via DOI. The repository contains two Excel files: polydatabase_dataset.xlsx: Primary extracted dataset of polymer molecular dynamics simulation records. Each row links a source publication (DOI) to a standardized polymer name, the reported force field, one of six target properties (density, viscosity, diffusion coefficient, glass transition temperature, radius of gyration, or Young's modulus), the corresponding value in standardized units, and an "Extra Information" field capturing reported simulation conditions (e.g., temperature, pressure, chain length or molecular weight, number of chains, ensemble/equilibration protocol, and system type). polydatabase_dataset_augmented.xlsx: Extended version of polydatabase_dataset.xlsx that includes all extracted fields plus a force field type classification (All-Atom, United-Atom, or Coarse-Grained) and additional polymer-level descriptors (e.g., abbreviation, source monomer, trade name, chain structure, architecture, thermal type, and origin type) added to support filtering and search in the online database. These augmented descriptors do not originate from the source publications; they are assigned through LLM-based classification and provided as convenience metadata.