[lammps-users] restart with hybrid dihedral_style class2


In a restart script I need to set the hybrid dihedral_coeff section
since it is not saved in the restart file. However it appears that with
class2 the additional sections

MiddleBondTorsion Coeffs
EndBondTorsion Coeffs
AngleTorsion Coeffs
AngleAngleTorsion Coeffs
BondBond13 Coeffs

can only go into a data file which am not able to read (I can't issue a
read_data command before or after the read_restart (below). My restart
script is as follows:

That's a deficiency of the code currently. I suppose no
one has ever wanted to use class2 styles underneath
a dihedral_style hybrid command.

For now, I suggest you convert your restart file to
a data file (tools/restart2data) and put the class2 coeffs
back into that data file.


Thanks. One related question. I am using reset_timestep N to continue
the simulation from N in the previous run. However I am also using the fix

fix 1 all npt temp 298.15 298.15 100.0 iso 1.0 1.0 1000.0

and I am not sure how to adjust it during a restart. I guess it depends
whether the previous run achieved the desired temperature and pressure.
Say it did, then should I just use a small value for the Tdamp and Pdamp
parameters say: 10 timesteps? and if the previous run did not achieve
the thermostat/barostat final values, something bigger?

Thanks for comments,

Actually the restart with the npt fix can cause the job to quit. For
instance, if I use

    fix 1 all npt temp 298.15 298.15 10.0 iso 1.0 1.0 10.0

with a configuration that it is pretty much at the desired temperature
and pressure, I get:

    Out of range atoms - cannot compute PPPM

If I change the fix to

    fix 1 all npt temp 298.15 298.15 10.0 iso 1.0 1.0 100.0

the job runs.

Insight appreciated.


The fix npt command has nothing to do with restarting.
You can use whatever fix npt command you want
after restarting.

If it doesn't perform well, it's is because you picked
bad parameters - again that has nothing to
do with restarting.

You should never use 10 timesteps for Tdamp or Pdamp.
More like 100-1000 timesteps for Tdamp and 1000 or
more for Pdamp.


Just posted a patch (17 Nov) that allows
you to now specifiy these cross-term coeffs
in an input script for class2 potentials,
as well as in the data file.

So please try it out, and see if it works
as you expect.