Aidan,
Thanks for checking for me. I don’t see a slow drift away from 0 momentum. The linear and angular momentum stay at 0 for the first ~8.3 million time steps, and then suddenly receives a relatively powerful kick, violating both its linear and angular momentum conservation and abruptly ending the simulation. Are you sure that it was a slow drift in your case? My input file has a line you can comment in for DCD output if that helps.
The technique that I was using originally is STMD, which does conserve linear and angular momenta. Here’s the paper:
URL:http://link.aps.org/doi/10.1103/PhysRevLett.97.050601
DOI:10.1103/PhysRevLett.97.050601
Gil
Yes, I'm sure. Nothing strange happened wit NVT. I suggest you
rebuild with the latest version of LAMMPS.
Attached is the input script I used, including COM output....
in.lj_cluster (1.39 KB)
linear and angular momentum stay at 0 for the first ~8.3 million time
steps, and then suddenly receives a relatively powerful kick
I haven't followed all of this discussion, but I want to call attention
to what Aidan said earlier:
Also, angular momentum is never truly conserved in a periodic
system. Each time an atom i crosses a periodic boundary, there is a
kick of order L x p_i.
If one of the 5 atoms crosses a PBC at 8.3 million steps
then you will see the kick. But only in angular momentum,
not in linear. If that is happening in linear then something
else is going wrong.
Steve