AI agents for automating materials research: a case study of crystal plasticity simulations

This work introduces a multi-agent system, CrystalPlasticitySim, which leverages large language models (LLMs) to automate complex workflows involved in crystal plasticity simulations. The system substantially reduces the human effort required to learn, prepare, and execute crystal plasticity simulations, enabling workflows that previously required months of manual effort to be completed within hours of autonomous execution. Traditionally, using crystal plasticity simulations requires not only deep expertise in materials science but also technical knowledge of a specific simulation package and computational skills – forcing users to manually configure parameters, prepare input files, and troubleshoot simulations through laborious trial-and-error processes. These technical hurdles limit accessibility and slow scientific progress. To overcome these challenges, CrystalPlasticitySim employs three collaborating AI agents – the Supervisor Agent, Simulation Agent, and Computational Assistant Agent – that autonomously generate input files, execute simulations, extract results, and even perform parameter optimization. Using a case study on the anisotropic behavior of Ni₃Al single crystals during cold rolling, we demonstrate that the system can autonomously solve well-defined tasks, thereby lowering the barrier to crystal plasticity modeling and improving accessibility to advanced simulation tools to researchers across disciplines.