
The industry-standard engine for massively parallel molecular dynamics and AI-driven materials discovery.

LAMMPS is a versatile, classical molecular dynamics (MD) code designed for high-performance computing environments. Originally developed by Sandia National Laboratories, it has evolved into the dominant engine for simulating materials at the atomic, meso, and continuum scales. Its technical architecture relies on spatial decomposition, allowing it to scale linearly across millions of CPU/GPU cores. In 2026, LAMMPS is at the forefront of the 'AI for Science' movement, providing the primary inference engine for Machine Learning Interatomic Potentials (MLIPs) like SNAP, NequIP, and DeepMD. It supports a vast array of force fields, from simple Lennard-Jones to complex reactive models (ReaxFF). Its modular C++ design allows for extensive customization, while its Python interface enables seamless integration into AI workflows and automated high-throughput screening pipelines. Whether modeling crack propagation in alloys or the self-assembly of lipid bilayers, LAMMPS remains the critical infrastructure for researchers bridging the gap between quantum accuracy and macroscopic behavior.
LAMMPS is a versatile, classical molecular dynamics (MD) code designed for high-performance computing environments.
Explore all tools that specialize in energy minimization. This domain focus ensures LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) delivers optimized results for this specific requirement.
A performance-portable library that allows LAMMPS to run efficiently on CPUs, GPUs (NVIDIA/AMD/Intel), and ARM processors with a single codebase.
Native support for neural network and Gaussian process potentials through the ML-SNAP, ML-PACE, and DeepMD-kit packages.
Highly optimized reactive force field integration for simulating chemical reactions within large-scale MD.
Partitions the simulation box into 3D sub-domains assigned to individual processors with ghost atom communication.
A Python wrapper that allows LAMMPS to be treated as a library, enabling dynamic parameter updates and machine learning loops.
A modular system where 'fixes' modify system state (e.g., thermostats) and 'computes' perform on-the-fly analysis.
Dynamic adjustment of sub-domain boundaries to account for density fluctuations in the simulation box.
Download the stable source code or clone the GitHub repository.
Install prerequisites including C++ compiler, MPI library, and CMake.
Configure build using CMake, selecting necessary packages (e.g., KOKKOS for GPU, ML-SNAP for AI).
Compile the binary using 'make' or 'ninja' to generate the lmp executable.
Define the simulation system in a data file specifying atomic coordinates and topologies.
Write an input script (.in) defining force field parameters, ensembles (NPT/NVT), and temperature controls.
Run a test simulation on a single core to verify force field convergence.
Execute parallel production runs using mpirun or srun on HPC clusters.
Monitor thermo output in real-time to ensure system stability.
Export dump files for post-processing and visualization in Ovito or VMD.
All Set
Ready to go
Verified feedback from other users.
"Universally praised for scalability and modularity. Some users find the input script syntax steep, but the community support is unparalleled."
Post questions, share tips, and help other users.
No direct alternatives found in this category.