HDNNP-LAMMPS

LAMMPS-extending program that consists of following 4 .h and .cpp files

• neural_network_potential.*
• pair_nnp.*
• symmetry_function.*

setup and compile

$ cd path_to_lammps
$ cd src/
$ ln -s path_to_this/neural_newtork_potential.h
$ ln -s path_to_this/neural_newtork_potential.cpp
$ ln -s path_to_this/pair_nnp.h
$ ln -s path_to_this/pair_nnp.cpp
$ ln -s path_to_this/symmetry_function.h
$ ln -s path_to_this/symmetry_function.cpp

then, this pair potential use Eigen library, and MKL library.
so you have to install them.
in order to use MKL in Eigen, edit a makefile in path_to_lammps/src/MAKE/*/Makefile.*.
in the makefile, add -mkl option as follows.

...
LIB = -mkl
...

finally, you can compile lammps with Neural Network Potential

$ make mpi  #=> lmp_mpi will be created in src/

how to use

all parameters are set in one potential file in txt format. sample parameter file coeff_sample is also included in this repository

LAMMPS command

2 changes in LAMMPS script,

pair_style nnp 
pair_coeff * * coeff_sample Ga N

you have to change only coeff_sample Ga N part. all parameters should be written in the potential file.

HDNNP program

HDNNP