I tried running a short simulation with fix recenter using the standard Verlet integrator. I then ran the same simulation again, but without fix recenter. For the case without fix recenter, I used equivalent post-processing to adjust the center of mass back to its initial position on every step. The results are initially essentially identical, but they eventually diverge significantly.
I expect that this is due to floating point arithmetic, but I would just like to check to see if you are aware of anything else that could be affected by this fix - e.g. do you happen to know if there is any LAMMPS code that uses position values not only from the previous timestep, but also the one before that?
Thanks,
Tom Coles
Can't think of a specific reason they would be different
I tried running a short simulation with fix recenter using the standard Verlet integrator. I then ran the same simulation again, but without fix recenter. For the case without fix recenter, I used equivalent post-processing to adjust the center of mass back to its initial position on every step. The results are initially essentially identical, but they eventually diverge significantly.
you should see the same kind of divergence,
simply by starting from translated, but otherwise
identical initial coordinates.
axel.
yes, this is a good point. If what you
are seeing is a slow divergence of the
2 simulations (or the post-processed quantity),
then that could simply be round-off due
to the coordinate shifting.
Steve
Thanks. I tried translated coordinates and saw the same thing.
Tom