[lammps-users] Minimization_large_system

LAMMPS LAMMPS Mailing List Mirror
1

Hi all

I am trying to minimize a system of about 55 million atoms using conjugate gradient with 0.0 as energy tolerance and 1e-08 as the force tolerance. It is taking a lot many steps than expected. Can anybody tell me if there is any limit to the number of atoms for the minimization algorithm. I am checking the sum of fx^2+fy^2+fz^2 for all atoms at intervals of 20000 steps and this value is not changing at all. Any comments.

Thanks

Regards
Sreekanth

2

Theoretically CG takes 3N steps to converge when N is 55 million in your
case. The CG in LAMMPS tries to restart when it loses precision, but maybe
it is not restarting enough for such a large problem. If you run CG
for some number of iterations, then run it again, does it do better the
2nd time? That is the equivalent of a restart.

Steve