These packages are written and maintained by other groups. They are downloaded
and compiled together with LAMMPS to add new pair styles, fixes, and other
capabilities. See the build documentation
for how to include external packages, and the
Packages list in the manual for the
packages that ship with LAMMPS itself. Want to add one? Drop a file in
content/download/packages/ (or email the developers).
External Packages
MLMOD-PYTORCH
Machine-learning methods for data-driven modeling and simulation in LAMMPS via PyTorch.
MLMOD-PYTORCH (author: Paul Atzberger, UC Santa Barbara) is a Python/C++ package
for using machine-learning methods and data-driven modeling in LAMMPS
simulations. It provides time-step integrators for dynamics and interactions
using general ML model classes — neural networks, kernel regression, and others
— with models trained and exported from PyTorch or other ML frameworks. It is
organized as a standalone library libmlmod.so with a lightweight interface to
LAMMPS via the USER-MLMOD patch, and installs via pip with pre-compiled binaries
or Docker images.
More details: arXiv:2107.14362.
MANGO-SELM
Fluctuating hydrodynamics with implicit solvent (Stochastic Eulerian Lagrangian Method).
SELM (Stochastic Eulerian Lagrangian Method; author: Paul Atzberger, UC Santa Barbara) is a set of numerical methods using a mixed Eulerian description for hydrodynamic fields coupled to a Lagrangian description of coarse-grained degrees of freedom. Fluctuating hydrodynamic equations account for both hydrodynamic flow and thermal fluctuations consistent with statistical mechanics; for implicit-solvent coarse-grained models the methods capture momentum transfer through the missing solvent. It is provided as a USER-SELM package for LAMMPS implementing several thermostats (full inertial dynamics, strong coupling, overdamped/quasi-steady-state, and Lees-Edwards shear).
USER-MLIP (MTP)
Machine-learning moment tensor potentials (MTP) interface for LAMMPS.
USER-DEEPMD
DeePMD-kit machine-learning potentials; can also be built as a LAMMPS plugin.
USER-MLIP (linearized)
Linearized machine-learning interatomic potentials for LAMMPS.
USER-AENET
Use artificial neural network (ANN) atomic potentials from aenet in LAMMPS.
USER-AENET (author: Hideki Mori, College of Industrial Technology, Japan) lets
LAMMPS users run MD and MM simulations with ANN atomic potentials generated by
aenet. The package adopts the conventional
pair_style/pair_coeff formats (like EAM) for flexible use.
USER-MESO
GPU-enabled version of the DPD-MESO package in LAMMPS.
LAMMPS Plugin Collection
External LAMMPS styles updated and built as runtime-loadable plugins.
With the plugin command it is possible to load additional LAMMPS styles into an executable at runtime, if compiled accordingly. The lammps-plugins repository contains source code for several external LAMMPS styles, updated for recent versions of LAMMPS and combined with a plugin loader and a CMake build system to compile them into plugins (see the repo README for the included packages). The USER-DEEPMD package can also be configured and compiled as a plugin. For more information, see the manual.