PolyJarvis: LLM Agent for Autonomous 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.

Figures (14)

• Figure 1: PolyJarvis system architecture. The LLM agent receives natural language polymer specifications and orchestrates the simulation workflow through two MCP servers. The local server handles molecular construction using RadonPy and the remote server executes GPU-accelerated LAMMPS simulations. Arrows indicate bidirectional communication via MCP.
• Figure 2: Representative PolyJarvis agent conversation. The user submits a natural language task prompt specifying the polymer system, target properties, and computational resources. The agent responds with a proposed workflow and any clarifying questions before proceeding autonomously. After the simulation completes, the agent reports predicted property values alongside experimental references and percent error.
• Figure 3: Density versus temperature curves for the best replicate of each benchmark polymer, with bilinear fits (black lines), extracted $T_g$ (gold stars), and literature experimental ranges (green shaded area).
• Figure 4: Predicted versus experimental density at 300 K. Large symbols show ensemble means $\pm$ 1 s.d.; small symbols show individual runs. The shaded band marks $\pm$5%.
• Figure 5: Predicted versus reference bulk modulus at 300 K.