Harmonic bond_style not using periodic boundaries

I tried initialising the velocities to zero, but the result was unaffected. Luckily zero-velocity is indeed default, but it was good to check it just in case.

Derek,
You have a system with no pairwise interactions and LAMMPS uses those
to deal with periodic images. Look into the communicate command for a
solution.
A line like this for example: "communicate single cutoff 2"
will do the trick.
Carlos

Thank you! Communicate helped alleviate the problem. I didn’t realize that bonds required this additional command for periodicity.

Thank you! Communicate helped alleviate the problem. I didn't realize that
bonds required this additional command for periodicity.

it usually is a non-issue, since inputs have non-bonded pairwise
interactions with a cutoff that is 5-10x the length of a bond, and this
cutoff is used as a default for selecting the communication cutoff.

axel.

If you set neigh_modify cluster yes it will do an error check for this.

The reason this happens is that LAMMPS finds the bond partner
by looking for the closest atom (real or ghost) that has the right
atom ID. If there are no ghosts (due to no pair cutoff) then the
"closest" is the one far away, when running on 1 proc. If you ran
on more procs you would get a missing bond atom error. The fix
is, as indicated, to force some ghost atoms to be communicated.

Steve