I can run my simulation of a 569 atom oligomer using my 10 Feb, 2012 build of LAMMPS
and I do not received the “Error: Bond/angle/dihedral extent > half of periodic box length”,
however, I get this message when I run the same simulation with the 21 Nov, 2012 build. I’m continuing to
look for “pilot error” on my part. I was wondering if this message was added to LAMMPS since
I built it last in February or if It’s been here a while.
Thanks in advance for any information,
-Bruce
hi bruce,
I can run my simulation of a 569 atom oligomer using my 10 Feb, 2012 build
of LAMMPS
and I do not received the "Error: Bond/angle/dihedral extent > half of
periodic box length",
however, I get this message when I run the same simulation with the 21 Nov,
2012 build. I'm continuing to
look for "pilot error" on my part. I was wondering if this message was added
to LAMMPS since
I built it last in February or if It's been here a while.
according to "git blame" this message (and the associated checks)
were added in mid june 2012 in an effort to tighten checks on incorrect
assignment of bonded coordinates to periodic images.
in many cases, the trigger for this message is the incorrect
specification of image flags in a data file (or lack of it). in
older versions of lammps, the simulation would continue as
expected in most cases, due to enforcing minimum image
conditions inside bonded interaction calculations. this limitation
has been lifted since.
cheers,
axel.
Dr. Kohlmeyer,
I did some further reading in the LAMMPS manual and I have the following questions:
I have only used periodic boundary conditions in my simulations.
LAMMPS is setting my initial geometry to image zero by default?
I am seeing the “Error: Bond/angle/dihedral extent > half of periodic box length” error below
due to the image flag for at least one atom wrapping back to zero? When the image flag wraps
back to zero for an atom, LAMMPS does not know where the atom is bonded since the image
information is lost? Or does LAMMPS lose the atom’s bonding information due to it being beyond the midpoint
in the next periodic box as per the error message?
In a well behaved periodic boundary condition-based simulation atoms in image zero should only
traverse the box boundary to the adjoining box/image? What I mean is a simulation where you expect the
material to not be a projectile in motion. In my case I calculate moduli and watch water migrate
through an epoxy matrix.
The final box configuration at the end of a fix is composed of atoms from your original box plus atoms that entered
the box from neighboring boxes, which are the atoms that left the original box that now have different image flag values?
So LAMMPS is doing a better job of atom accounting in the final analysis?
Thanks,
-Bruce
Dr. Kohlmeyer,
I did some further reading in the LAMMPS manual and I have the following
questions:
I have only used periodic boundary conditions in my simulations.1. LAMMPS is setting my initial geometry to image zero by default?
yes, unless you provide a different setting from reading a data file.
2. I am seeing the "Error: Bond/angle/dihedral extent > half of periodic box
length" error below
due to the image flag for at least one atom wrapping back to zero? When
the image flag wraps
back to zero for an atom, LAMMPS does not know where the atom is bonded
since the image
information is lost? Or does LAMMPS lose the atom's bonding information
due to it being beyond the midpoint
in the next periodic box as per the error message?
no. the error message stems from having inconsistent settings.
when checking bonds, angles, and dihedrals, LAMMPS checks,
whether the image flags are consistent with the bonding pattern.
it flags an error, when on of the coordinates is the "wrong" image.
that usually is an error in the input. in rare cases, it can also be
an indication of bad dynamics.
3. In a well behaved periodic boundary condition-based simulation atoms in
image zero should only
traverse the box boundary to the adjoining box/image? What I mean is a
simulation where you expect the
material to not be a projectile in motion. In my case I calculate moduli
and watch water migrate
through an epoxy matrix.
if atoms migrate, their image counters are incremented
or decremented as needed. that atoms in the center of
the system need not be in image zero.
4. The final box configuration at the end of a fix is composed of atoms from
your original box plus atoms that entered
the box from neighboring boxes, which are the atoms that left the
original box that now have different image flag values?
i don't understand this question.
So LAMMPS is doing a better job of atom accounting in the final analysis?
i don't understand what you are asking here, either.
axel.