LAMMPS - A flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales

This paper serves as the definitive reference for LAMMPS, the classical molecular dynamics simulator released as open source in 2004 and now a widely used tool for particle-based modeling of materials at length scales ranging from atomic to mesoscale to continuum. It describes several of the code's fundamental algorithms, including neighbor lists, parallel spatial decomposition, parallel FFTs for long-range Coulombic interactions, and Stormer-Verlet symplectic time integration, along with the design strategies that have made LAMMPS flexible for both users and developers.

That flexibility explains the code's popularity and growth from fifty thousand lines in 2004 to a million lines today, with hundreds of contributors, a wide variety of particle interaction models, and portability from a single CPU core to the largest accelerator-equipped supercomputers. The paper highlights recently added capabilities enabled by this design, including dynamic load balancing, on-the-fly visualization, magnetic spin dynamics models, and quantum-accuracy machine learning interatomic potentials.