A relatively simple python tool to create any pair potential to use with the LAMMPS pair_style table option https://docs.lammps.org/pair_table.html, and provides an automatic check against the pair_write output.
Benefits:
* One only needs to define a pair function. The force is computed automatically through automatic differentiation.
* It (the code) automatically makes a plot that compares the original table, with what LAMMPS sees through pair_write.
* It generates a table containing the relative differences between the original table of energies and forces against pair_write.
* Users can then simply copy commands from the -- automatically generated -- pair_write input script into their simulation script.
The automatic check involves automatically running LAMMPS with pair_write and plotting a comparison of your original table (of potential energy and force) against what LAMMPS sees (see below for results of the examples). A text file containing the relative differences is generated (again automatically), helping you to gain a numerical estimate on reliability.
This tool has been tested with style = linear and spline. This tool supports 'R' (linear) and 'RSQ' (squared) distances.
- If you have used this tool and want to add your potential, then send the python code for such a potential along with the make_your-potential.py code and potential and force test plots to luke.davis@ucl.ac.uk. Then I will add it to the potentials_for_lammps.py. Please keep to the style of naming your pair potential function pair_name.
Linux, python3
-
autograd info: https://autograd.readthedocs.io/en/latest/introduction.html pip: https://pypi.org/project/autograd/
-
numpy https://numpy.org/
-
scipy https://scipy.org/
-
matplotlib (with pdf backend - which it should come with automatically) https://matplotlib.org/
-
For the example_pair_rep_coh_smooth_linear_MPI.ipynb example you need mpi4py https://mpi4py.readthedocs.io/en/stable/
You need a lammps executable (you can specify the exact path to this in your make_pairstyle.py code (see the other make<...>.py examples)).
- Install the package with
pip install . - Ensure the path to the lammps executable is specified, the first time you run this code you'll receive instructions, or see the jupyter notebook in the examples directory.
- Specify functional form with a pair_keyword (if the potential is supported by eptl)
or specify a custom function and pass to
eptl.create_lammps_pairstyle_table.make_table_for_lammps(see the jupyter notebook in the examples directory for guidance). Just requires the pair-wise distance r to be the first argument. You can add an arbitrary number of other arguments and keywords to be passed. - Check plot (if plot=True is on) and/or the relative differences data to see if the potential will work - as expected - in lammps.
- Copy the pair_style table and pair_coeff commands from the generated pair_write lammps input script into your actual simulation input scripts.
- Run your simulations!
This example code is basic and if you understand how it works you can then do what you want.
Note you can specify rather involved potentials, and as long as you use numpy functions etc. the force can be gotten automatically. You can also look at example_pair_rep_coh_smooth_linear_MPI.ipynb for a more involved example which computes many pair tables using the Message Passing Interface (MPI).
