Using "fix bond/break" made ERROR Angle atoms ___ ___ ___ missing on proc ___ at step ___.

Dear Lammps Users,

I am using lammps/10Dec12 version on Lonestar linux cluster in Texas Advanced Computing Center (TACC). This is the latest version on this machine.

I created crosslinked structure using “fix bond/create” command. Then, I saved the system using “write_restart” command.

After using restart2data tool to create LAMMPS data file, I assigned a new bond type for newly created bonds.

Then, I tried to delete and adjust hydrogens using “fix bond/break” and “delete_atoms” commands.

Several error messages came out when I ran simulations.

I fixed most problems. For example,

When I had an error like “out of range atoms - cannot compute PPPM”, I added “delay 0 every 1 check yes” in neighbor_modify.

One problem is now encountered.

ERROR Angle atoms ___ ___ ___ missing on proc ___ at step ___.

I have tried several solutions 1) decreasing time-step, 2) decreasing Nevery in “fix bond/break” command, 3) increasing “cutoff distance” and “neighbor skin” with inserting “page” and “one” numbers in neighbor_modify, 5) decreasing temperature in “fix nve and npt” 6) adding “fix temp/berendsen” and 7) finding a possible solution in mail list archive and in Lammps manual.

None of these solutions were helpful.

I tested creating new bonds with the same restart data file using “fix bond/create” command. There is no problem.

I would like to attach the input file and data file. However, the size of those files is above the limitation of 500K on this mailing list.

I appreciate any comments.

Thanks.

Best regards,
Changwoon Jang

Dear Lammps Users,

    I am using lammps/10Dec12 version on Lonestar linux cluster in Texas
Advanced Computing Center (TACC). This is the latest version on this
machine.

there is nothing keeping you from compiling your own.
center staff provide pre-compiled versions of applications
like LAMMPS as a convenience to their users, but i
often found that those are not done very well, since the
people compiling LAMMPS didn't know anything about it.

    I created crosslinked structure using "fix bond/create" command. Then, I
saved the system using "write_restart" command.

After using restart2data tool to create LAMMPS data file, I assigned a new
bond type for newly created bonds.

Then, I tried to delete and adjust hydrogens using "fix bond/break" and
"delete_atoms" commands.

Several error messages came out when I ran simulations.

I fixed most problems. For example,

When I had an error like "out of range atoms - cannot compute PPPM", I added
"delay 0 every 1 check yes" in neighbor_modify.

One problem is now encountered.

ERROR Angle atoms ___ ___ ___ missing on proc ___ at step ___.

I have tried several solutions 1) decreasing time-step, 2) decreasing Nevery
in "fix bond/break" command, 3) increasing "cutoff distance" and "neighbor
skin" with inserting "page" and "one" numbers in neighbor_modify, 5)
decreasing temperature in "fix nve and npt" 6) adding "fix temp/berendsen"
and 7) finding a possible solution in mail list archive and in Lammps
manual.

when you make changes to your system that are as massive
as what you describe, you are likely to introduce "hot spots",
local areas of very high potential energy. you have to properly
equilibrate your system again and particularly remove the
unphysical hot spots. neither of the steps you mention help,
and fix temp/berendsen doing a global rescale actually makes
it worse(!).

the simplest option is to switch to using fix nve in combination
with fix langevin with a rather short time constant. that will put
a damping on everything and also dissipate energy by assigning
random velocity changes. you may need to reduce your timestep
for a bit and/or using fix nve/limit instead of just fix nve. once your
system has equilibrated enough, you can switch back to a
more realistic thermalization using fix nvt or npt and a larger
time step.

you can also do a sequence of quenches, i.e. do a quick
minimization for a small number of steps, re-assign velocities,
run for a little bit, and repeat until the system behaves well again.

None of these solutions were helpful.

I tested creating new bonds with the same restart data file using "fix
bond/create" command. There is no problem.

I would like to attach the input file and data file. However, the size of
those files is above the limitation of 500K on this mailing list.

you can always try the same thing with a (much smaller)
test system. in fact, that is often a much better way to try
something like this out and develop a suitable protocol, as
it requires much less waiting and computational resources.
remember, that such test calculations do not need to result
in publishable resultsl they simply are a way of developing
a workflow with less effort.

axel.

Changwoon,

Have you removed angles and dihedrals for deleted H atoms? You could check the delete_bonds command to remove topology by atom_type (or other types). Also the weighting for nonbond 1-2, 1-3, 1-4 interactions may need to be updated after changing topology. delete_bonds has a “special” flag to update these weights.

cheers,

Tim

Dear Axel and Tim,

Thank you for Axel’s details about possible reasons and Tim’s suggestion for lammps command. I am testing all your comments now and I think that these commands will work.

Thanks.

Best regards,
Changwoon Jang