[lammps-users] Monte Carlo Minimization

Hi Steve,

Is there a min_style for "minimize" that equilibrates via Monte Carlo rules using a Boltzmann acceptance criterion? I'd like something similar to the "bond/swap" fix except I want to randomly nudge atoms between trial energy calculations instead of swapping bonds.

Nicholas J. Lee

More info:

My system gets stuck at local minima when using conjugate gradient. I've discovered this by comparing CG runs with "fix viscous". But damping is only useful when a structure is very close to equilibrium and there are no significant potential energy barriers between the initial and lowest energy configurations; ...but there are large pot barriers in this situation. So I want to use a Monte Carlo method since it has a chance of stochastically slipping through these barriers to descend into lower minima as long as it runs.

nope - that is non-trivial to do in parallel - how would
you make simultaneous moves on multiple procs without
conflicting? The next release of our MC code SPPARKS
will have this capability for a limited set of problems (atoms,
not molecules), but there's no plan to do it in LAMMPS