materialsmodeling/lammps-ase repository overview

LAMMPS is a classical molecular dynamics code that supports a very large number of interaction models. The full documentation can be found here⁠. This repository provides images with precompiled lammps versions that are set up to allow LAMMPS to be run both from the command line and as a calculator via ase.

When inside the container, the LAMMPS installation files can be found in /lammps/. This includes potential input files, which reside in /lammps/potentials/ and examples in /lammps/examples, which you can use to test the image.

⁠Running LAMMPS interactively

Running the following command will launch an interactive shell

docker run -it -v $PWD:/data materialsmodeling/lammps

Here, the current directory has been mounted in order to enable reading from and writing data to disk during the session. One can now simply run LAMMPS as follows

lmp -in <input_file>

Here <input_file> is the main LAMMPS input file, which must reside in the directory from which the container is launched along with all other pertinent input files.

As a practical test you can try the following

cp -r /lammps/examples/indent/ .
cd indent
lmp -in in.indent

To utilize MPI parallelization run

mpirun -np <n_cores> lmp -in <input_file>

To utilize OpenMP parallelization set the number of threads before starting LAMMPS

export OMP_NUM_THREADS=4

⁠Running LAMMPS directly

LAMMPS can also be run without launching a shell via

docker run -v $PWD:/data materialsmodeling/lammps <command>

Here, <command> is the usual command used to launch, as for example in

docker run -v $PWD:/data materialsmodeling/lammps lmp -in my_input_file.in

or

docker run -v $PWD:/data materialsmodeling/lammps mpirun -np4 lmp -in my_input_file.in

⁠Running LAMMPS via ase

LAMMPS can also be run as a calculator via ase⁠. The following script serves as a simple example

from ase.build import bulk
from ase.calculators.lammpslib import LAMMPSlib

calc = LAMMPSlib(lmpcmds=['pair_style sw',
                          'pair_coeff * * /lammps/potentials/Si.sw Si'],
                 log_file='lammps.log',
                 keep_alive=True,
                 atom_types={'Si': 1})
atoms = bulk('Si')
atoms.set_calculator(calc)
energy = atoms.get_potential_energy()
print(f'energy: {energy}')

This script (ase-lammpslib-sw.py) can be launched as follows

docker run -v $PWD:/data materialsmodeling/lammps-ase:patch_8Feb2019 python3 ase-lammpslib-sw.py