PolyJarvis: LLM Agent for Autonomous Polymer MD Simulations

2026-04-02

AI Summary

This work proposes PolyJarvis, the first large language model–driven, end-to-end autonomous agent for all-atom molecular dynamics (MD) simulations of polymers, which significantly lowers the expertise barrier and procedural complexity traditionally associated with such workflows. By seamlessly integrating with the RadonPy platform via the Model Context Protocol, PolyJarvis autonomously executes the entire pipeline—from natural language instructions to monomer construction, force field parametrization, GPU-accelerated MD simulations, and prediction of key properties including density, bulk modulus, and glass transition temperature (Tg). Validation on four polymers demonstrates excellent accuracy: density predictions exhibit errors of 0.1–4.8%, Tg for PMMA deviates by only +10–18 K, and five out of eight predicted properties meet stringent acceptance criteria, thereby enhancing both accessibility and reproducibility in polymer MD simulations.

Abstract

All-atom molecular dynamics (MD) simulations can predict polymer properties from molecular structure, yet their execution requires specialized expertise in force field selection, system construction, equilibration, and property extraction. We present PolyJarvis, an agent that couples a large language model (LLM) with the RadonPy simulation platform through Model Context Protocol (MCP) servers, enabling end-to-end polymer property prediction from natural language input. Given a polymer name or SMILES string, PolyJarvis autonomously executes monomer construction, charge assignment, polymerization, force field parameterization, GPU-accelerated equilibration, and property calculation. Validation is conducted on polyethylene (PE), atactic polystyrene (aPS), poly(methyl methacrylate) (PMMA), and poly(ethylene glycol) (PEG). Results show density predictions within 0.1--4.8% and bulk moduli within 17--24% of reference values for aPS and PMMA. PMMA glass transition temperature (Tg) (395~K) matches experiment within +10--18~K, while the remaining three polymers overestimate Tg by +38 to +47K (vs upper experimental bounds). Of the 8 property--polymer combinations with directly comparable experimental references, 5 meet strict acceptance criteria. For cases lacking suitable amorphous-phase experimental, agreement with prior MD literature is reported separately. The remaining Tg failures are attributable primarily to the intrinsic MD cooling-rate bias rather than agent error. This work demonstrates that LLM-driven agents can autonomously execute polymer MD workflows producing results consistent with expert-run simulations.

Problem

• molecular dynamics
• polymer simulation
• force field
• property prediction
• expertise barrier

Innovation

• LLM agent
• autonomous molecular dynamics
• polymer simulation
• Model Context Protocol
• property prediction