image flags for a single bonded periodic structure

hi
I’m sure I’ll slap my forehead when I read the answer, but how do I set the image flags for a bonded structure that extends from one end of the system box to the other? E.g. a 1 D chain that is really a ring with atoms 1… N where N is bonded to N-1 and 1. I can’t think of a set of image flags that don’t put a cut in the unwrapped coordinates such that at least one bond is essentially the length of the box. I have read that LAMMPS uses a minimum image convention but does this supersede the image flags? The warning seems to imply “no”. Or perhaps the image flags are only used for atoms within a cutoff distance of the box boundary??
thanks
Reese
ps the reason I ask is that I am reading in a data file (without image flags) for a 2D sheet like structure with charges that eventually loses atoms and goes unstable. The inconsistent image flag warning is the only thing I see for the cause. I am using neigh_modify 1 and comm_modify cutoff 16.1 (i.e. a large value )

Reese (Jones) 925 294-4744 [email protected]…33…3…

hi
  I’m sure I’ll slap my forehead when I read the answer, but how do I set
the image flags for a bonded structure that extends from one end of the
system box to the other? E.g. a 1 D chain that is really a ring with atoms
1.. N where N is bonded to N-1 and 1. I can’t think of a set of image flags
that don’t put a cut in the unwrapped coordinates such that at least one
bond is essentially the length of the box. I have read that LAMMPS uses a
minimum image convention but does this supersede the image flags? The
warning seems to imply “no”. Or perhaps the image flags are only used for
atoms within a cutoff distance of the box boundary??

​hi reese,​

for the initial structure, you should have all image flags set to zero. ​in
the case of an infinite periodic chain (or an infinitely large ring, if you
want to call it that),​ you *have* to get the "Inconsistent image flags"
warning. but you need not worry. when LAMMPS assembles the bond list, it
looks for the closest image, regardless of the value of the image flags of
each of the constituent atoms of a bond. this approach will only break
down, if you have an extremely small system (so that the "wrong" image
would be closer) or an extremely long bond (which also causes the "wrong"
image to be closer). there is no way for LAMMPS to detect whether such a
bond is intentional or a problem, so the consensus was to rather print a
warning and tell the few people with bonds across PBC to not worry too
much, if they have properly checked their geometry.

thanks

Reese
ps the reason I ask is that I am reading in a data file (without image
flags) for a 2D sheet like structure with charges that eventually loses
atoms and goes unstable. The inconsistent image flag warning is the only
thing I see for the cause. I am using neigh_modify 1 and comm_modify cutoff
16.1 (i.e. a large value )

​i would carefully check the bonds for atoms in the corners that cross
periodic boundaries. perhaps do a short run of that system with compute
bond/local and verify that all bonds are of the expected length.​

axel.