lammps/examples/MCa466230f4a9eefficient_neuronet
README
This directory has an input script that illustrates how to use LAMMPS as an energy-evaluation engine in a Monte Carlo (MC) relaxation loop. It is just an illustration of how to do this for a toy 2d problem, but the script is fairly sophisticated in its use of variables, looping, and an if-the-else statement which applies the Boltzmann factor to accept or reject a trial atomic-displacement move.
The script sets up a perfect 2d hex lattice, then perturbs all the atom positions to "disorder" the system. It then loops in the following manner:
pick a random atom and displace it to a random new position evaluate the change in energy of the system due to
the single-atom displacement
accept or reject the trial move if accepted, continue to the next iteration if rejected, restore the atom to its original position
before continuing to the next iteration
The 6 variables at the top of the input script can be adjusted to play with various MC parameters.
When the script is finished, statistics about the MC procedure are printed.
Dump file snapshots or images or a movie of the MC relaxation can be produced by uncommenting the appropriate dump lines in the script.
At some point, we will add a Python script to the python/examples directory that performs the same operation as this script. It will invoke LAMMPS as a library and use Python to implement the MC accept/reject logic more cleanly than the LAMMPS input script can do it.