ARTFEED — Contemporary Art Intelligence

AI Agent Automates Atomistic Simulations for Materials Design

ai-technology · 2026-07-29

Within the URSA (Universal Research and Scientific Agent) framework, researchers have created an agent-based system that automates atomistic simulations utilizing the LAMMPS tool. This system independently chooses interatomic potentials, designs and executes simulations, and incorporates iterative error recovery in a closed-loop process. Its scientific validity was evaluated against LAVA, a high-throughput toolkit for LAMMPS and VASP computations. By minimizing the need for manual input and reducing trial-and-error, the framework enhances the rigor, reproducibility, and scalability of atomistic modeling.

Key facts

  • The agent-based system is embedded within the URSA framework.
  • It automates design, execution, and validation of atomistic simulations.
  • The system uses the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS).
  • It autonomously selects interatomic potentials and constructs simulations.
  • Iterative error recovery is performed within a closed-loop workflow.
  • Benchmarking was done against LAVA, a toolkit for LAMMPS and VASP.
  • The framework reduces manual intervention and trial-and-error.
  • It improves rigor, reproducibility, and scalability of atomistic modeling.

Entities

Institutions

  • URSA (Universal Research and Scientific Agent) framework
  • Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS)
  • LAVA
  • Vienna Ab initio Simulation Package (VASP)
  • arXiv

Sources