error with PBC with triclinic cell

Hi All,

I am running lammps-1Feb2014, and I’ve seemingly run into a bug when simulating a triclinic crystal configuration of TATB, where the atomic coordinates get garbled somewhat upon applying periodic boundary conditions. I’ve attached three files with this email: a lammps input script (in.reaxc.tatb) with the ‘run’ variable set to zero, the data file with the initial coordinates (data.head), and the final coordinates transformed to xyz format for visualization (lmp_traj.xyz). This input script was run in the ‘lammps-1Feb14/examples/reax’ directory, where all of the needed reax/c input files can be found. As you can see, after running lammps the configuration gets noticeably changed and bonds appear to have been broken/formed. I notice this issue when running with different pair styles as well, and trying to optimize the before and after configurations in a quantum code indicates that they are indeed different chemically. Also, I can make the issue go away by unwrapping the atomic coordinates within lammps (i.e., by making calls such as domain->unmap(x[i],image[i],unwrap); in a pair potential file).

Thanks in advance for the help,

Nir

in.reaxc.tatb (1.43 KB)

lmp_traj.xyz (5.08 KB)

data.head (5.13 KB)

Hi All,

I am running lammps-1Feb2014, and I've seemingly run into a bug when
simulating a triclinic crystal configuration of TATB, where the atomic
coordinates get garbled somewhat upon applying periodic boundary conditions.
I've attached three files with this email: a lammps input script
(in.reaxc.tatb) with the 'run' variable set to zero, the data file with the
initial coordinates (data.head), and the final coordinates transformed to
xyz format for visualization (lmp_traj.xyz). This input script was run in
the 'lammps-1Feb14/examples/reax' directory, where all of the needed reax/c
input files can be found. As you can see, after running lammps the
configuration gets noticeably changed and bonds appear to have been
broken/formed. I notice this issue when running with different pair styles
as well, and trying to optimize the before and after configurations in a
quantum code indicates that they are indeed different chemically. Also, I
can make the issue go away by unwrapping the atomic coordinates within
lammps (i.e., by making calls such as domain->unmap(x[i],image[i],unwrap);
in a pair potential file).

NOOOOOO! this is *such* a bad, bad, idea. LAMMPS expects local
coordinates to be properly wrapped. otherwise your simulation will be
messed up badly.

what you see is simply the effect of outputting wrapped coordinates,
as you have requested with your dump command.
if you want unwrapped coordinates, you need to output them accordingly.

axel.

Hi Axel,

Thank you for your note. I don't think I was quite clear in my email. The
issue seems to be that LAMMPS garbles the atomic coordinates of my input
triclinic cell AFTER wrapping the atoms. Offhand, it seems that there is
some error in how my input deck is being processed. I'm hopeful to get
some help in looking into the matter. I experimented with unwrapping the
coordinates as I wrote below simply to see if there could be some error
with how LAMMPS was wrapping the atoms for this configuration -- I wasn't
using the unwrap call to actually run a simulation. Hopefully that
clarifies things.

Best,

Nir

Hi Axel,

Thank you for your note. I don't think I was quite clear in my email. The
issue seems to be that LAMMPS garbles the atomic coordinates of my input
triclinic cell AFTER wrapping the atoms. Offhand, it seems that there is
some error in how my input deck is being processed. I'm hopeful to get
some help in looking into the matter. I experimented with unwrapping the
coordinates as I wrote below simply to see if there could be some error
with how LAMMPS was wrapping the atoms for this configuration -- I wasn't
using the unwrap call to actually run a simulation. Hopefully that
clarifies things.

i don't quite see any irregularities with my 20March2014 version here.
if i use write_dump after running just one step and compare it to the
initial data file, the coordinates are changed, but the visualization
is practically identical. the .xyz you provide is useless, since it
doesn't provide any usable cell geometry information.

so at this point, i have to assume, that your problem is a
visualization problem,not a LAMMPS problem. i would need something
more tangible and convincing to dig deeper.

axel.