Using Machine-Learning Interatomic Potentials with LAMMPS

Use Claude Code to run LAMMPS simulations, work with machine-learning potentials, and develop custom features for materials research.

Using Machine-Learning Interatomic Potentials with LAMMPS in support of Claude Code Abstract :

LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) is a classical molecular dynamics (MD) code with a focus on materials modeling. Given the current trend of using machine-learning interatomic potentials (MLIAPs) for MD research, LAMMPS has become one of the most popular software packages as it provides an easy-to-use interface for using these models in MD simulations with native scalable parallelization and GPU support. At the same time, the rapid adoption of AI agents such as Anthropic Claude Code in research activities opens new opportunities to quickly learn how to use LAMMPS and to implement custom features specific to your projects.

This workshop explains how to use Claude Code to dig deep into the LAMMPS architecture to understand how to set up and run MD simulations with foundation models. We also discuss how to supervise Claude Code to develop new features specific to your research projects. The approach presented in this workshop can be applied to other software packages using AI coding agents.

Objectives:

Participants will understand: How LAMMPS works with MLIAPs under the hood. How LAMMPS is coupled with other codes via Python and C++ interfaces. How to modify and extend LAMMPS features for their research projects. How to contribute their codes to the LAMMPS community Presenter: Trung Nguyen

Level: Advanced

Duration: 2 hours

Prerequisite: Experience with LAMMPS in past or ongoing projects is required. Having access to Claude Code Pro via personal or university subscription is required.