Bond/angle/dihedral extent > half of the periodic box length

I started using a newer version of LAMMPS last November and stumbled across this message for simulations
that had previously completed. I have added additional dumps to capture the image flags ix, iy, iz. It appears to me
my molecule is not rotating or translating through multiple MD cells and I do use
periodic boundary conditions for all three dimensions. So I believe my image flags are not overflowing and my initial
geometry is all located in MD cell 0,0,0 as defined by each atom’s image flag.

I was interested in finding what other conditions could cause the message to be displayed? I use the CFF91 force field
and I have several terms and cross-terms that do not have parameters. Without any resistance to forces caused by kinetic energy
in conjunction with zero valued potential energy constants creating a “wet spaghetti” molecule I could have an atom that moves
further within the MD cell than it should causing the message?

What other conditions should I be aware of?

Thanks in advance!

-Bruce

I ran into this problem earlier, but your reason might be different.
See this thread:
http://sourceforge.net/mailarchive/message.php?msg_id=30561948

Summary: If this is happening at the begging of the simulation, then
check the initial coordinates and bonds by visualizing them. If you
have a DATA file with the coordinates (not "set x y z" input script
commands), then you can use VMD+topotools to take a look at the bonds
in your system.

Make sure you assigned the right atoms to the right bonds. If it
hepls, the "README_visualize.txt" attachment (containing instructions)
is at the bottom of the page, if you scroll down all the way.

Hope this helps
Andrew

The test is in src/domain.cpp and is pretty simple.
It does not use image flags. It does a minimum image
check on the bond/angle/dihedral lengths. If one is
over half a periodic box length, you get an error.

I suggest you put some print statements in that method
and see what is causing the error.

Steve