Running machine-learned force fields in LAMMPS

LAMMPS is a very popular molecular dynamics (MD) package which implements lots of advanced simulation methods. Although VASP itself also offers MD simulations LAMMPS provides more flexibility and additional methods (e.g. thermostats/barostats, grouping atoms, etc.). This page describes how to make pre-trained VASP machine-learned force fields available within LAMMPS. On a technical level this is achieved by patching the LAMMPS source code in such way that a new pair_style named vasp is available in the LAMMPS script language. The patched LAMMPS source is compiled and then linked to the VASPml library. The first section describes in detail how to patch and compile LAMMPS. The second section explains how the VASP ML_FF file can be loaded in a LAMMPS script.

Warning: The VASPml library available as of VASP 6.5.0 and therefore also the LAMMPS interface are experimental features. Please carefully check results and, if feasible, compare against similar VASP MD simulations.

Building LAMMPS with VASPml patch

Before starting the LAMMPS build please either compile VASP with the VASPml library, or alternatively, keep your VASP installation untouched and perform a standalone build of the VASPml library. To confirm that everything is ready we can check for the presence of the following files:

• For a VASP build with VASPml:
  • /path/to/vasp/build/VERSION/vaspml/lib/libvaspml.a
  • /path/to/vasp/build/VERSION/vaspml/include/InterfaceLAMMPS.hpp

where VERSION is any of std, gam or ncl. It does not matter which VASP build version is used because the VASPml library will be identical in all three cases.

• Or, alternatively, for a VASPml standalone build please check:
  • /path/to/vaspml/lib/libvaspml.a
  • /path/to/vaspml/include/InterfaceLAMMPS.hpp

Next, we need to obtain the LAMMPS source code with the patch for VASPml. Please clone the following repository to your hard disk and change into the LAMMPS root directory:

git clone https://github.com/vasp-dev/lammps
cd lammps

At this point looking around in the git repository will show the files from the develop branch (the main development branch of LAMMPS). However, the patch files are located in another branch, hence we need to switch with this command:

git checkout vasp-mlff

Then, while still in the LAMMPS root directory we need to create a symbolic link from lammps/lib/vasp/vaspml to the actual location of the VASPml folder.

• In case of a VASP build with VASPml:

ln -s /path/to/vasp/build/std/vaspml lib/vasp/

Again, it is not important which of the three VASP build directories (std, gam or ncl) is used because the VASPml library is the same.

• Alternatively, for a standalone VASPml build:

ln -s /path/to/vaspml lib/vasp/

For the remaining steps we need to change into the src directory:

cd src

The source files for the pair_style vasp are part of the optional package ML-VASP which can be activated with this command:

make yes-ml-vasp

Before we can finally start the build process it is necessary to adapt the LAMMPS makefile in MAKE/Makefile.mpi. Modify the CC and LINK variables to match the compiler used during the VASPml build process (CXX_ML variable). Also match the compiler flags in CCFLAGS and LINKFLAGS to CXXFLAGS_ML. Flags do not need to be identical but at least compatible compiler features should be used. Since VASPml requires BLAS and LAPACK we may also need to complete the LIB variable with the required libraries and their paths. For the GNU compiler with openBLAS it may look like this:

...
CC =    mpic++
CCFLAGS = -g -O3 -std=c++17
...
LINK =    mpic++
LINKFLAGS = -g -O3 -std=c++17
LIB = -L${OPENBLAS_ROOT}/lib -lopenblas

Finally, compile the LAMMPS source code with this command:

make mpi

Optionally, add the -j flag to perform a parallel build (faster). If the build process succeeds the LAMMPS executable lmp_mpi will be located in the lammps/src directory.

Setting up a LAMMPS MD run

Warning: Please always review the current limitations and bug reports.

LAMMPS comes with its own powerful script language which allows the user to specify all relevant MD simulation parameters in a single file. Please consult the LAMMPS documentation for details. Within the LAMMPS script language the commands pair_style and pair_coeff are responsible for selecting a force field. The ML-VASP package introduces a new pair_style called vasp. The pair_style vasp command does not have any additional arguments, all configurable settings are given as arguments to the pair_coeff command in this format:

pair_style vasp
pair_coeff * * file types

The pair_coeff command must be followed by * *, then followed by the name of the VASP force field file, typically ML_FF. Finally, there comes a mapping from LAMMPS atom types to VASP force-field types, e.g., H O Na Cl means that LAMMPS types 1, 2, 3 and 4 are mapped to VASP types H, O, Na and Cl, respectively. A valid example may look like this:

pair_style vasp
pair_coeff * * ML_FF Pb Br Cs

This will map the LAMMPS atom types 1, 2 and 3 in the input data file to the types Pb, Br and Cs for which a pre-trained machine-learned force field should be present in the ML_FF file in the execution directory. A summary of the type mapping is provided in the screen output and the log.lammps file, e.g. for the example above it looks like this:

   LAMMPS       pair_coeff      VASP      |             VASP force field
    types       names           subtypes  |     types       names        subtypes
----------------------------------------- | -------------------------------------
        1 <---> Pb        <---> 0         |         0 <---> Pb     <---> 0        
        2 <---> Br        <---> 1         |         1 <---> Br     <---> 1        
        3 <---> Cs        <---> 2         |         2 <---> Cs     <---> 2  

On the left side we find the mapping, the right side gives an overview of types present in the force field file. In this example, there is a one-to-one mapping, hence, the table looks pretty obvious and contains somewhat redundant information. However, it is also possible to leave out a mapping from specified LAMMPS types by supplying NULL instead of a valid VASP type name. This can be helpful when multiple force fields should be combined, see pair_style hybrid. Furthermore, multiple LAMMPS types may be mapped to the same VASP types. Finally, the force field file may contain types which are not used in the current MD simulation. Therefore, a more complicated example may look like this:

pair_coeff * * vasp ML_FF NULL Cs NULL Br Pb Br

and the corresponding table could contain this information:

   LAMMPS       pair_coeff      VASP      |             VASP force field
    types       names           subtypes  |     types       names        subtypes
----------------------------------------- | -------------------------------------
        1 <---> unmapped! <---> unmapped! |         0 <---> Ca     <---> unused!
        2 <---> Cs        <---> 2         |         1 <---> Pb     <---> 0        
        3 <---> unmapped! <---> unmapped! |         2 <---> O      <---> unused!
        4 <---> Br        <---> 1         |         3 <---> Br     <---> 1        
        5 <---> Pb        <---> 0         |         4 <---> Cs     <---> 2        
        6 <---> Br        <---> 1         |

Mind: Always ensure that the type mapping is correctly set up because mixed-up types may not immediately result in errors. An MD simulation may still run and only post-processing may ultimately reveal inconsistencies which can be tedious to trace back to type-mapping mistakes.

The pair_style vasp expects input coordinates to be in the units of Ångström and returns energies and forces with the energy unit of eV. Hence, it is only compatible with the LAMMPS setting units metal in the input script, otherwise an error will occur.