Switching from soft to lj/cut potentials causing unexpected behavior

Hello,

I’m not sure if the following behavior is supposed to happen. We’ve created an initial configuration and ran the simulation with the soft potential to avoid a system blow.
The simulation ran for 100,000 timesteps and a restart file was generated using the final timestep configuration.
From here on, we switched to the lj/cut potential, and this is where things get a little weird.

If we use the `read_restart’ command, the simulation runs fine with the lj/cut potential.
However, if we use the restart2data tool to convert the binary restart file to ASCII, and then use the read_data command along with declaring the following pair coeffs: pair_coeff 1 1 1.0 1.0 2.5, the simulation blows up (we only have 1 atom type).

I would gladly share the configuration file as well as the input script if necessary. Also, if this is any help, we are using LAMMPS version (9 Mar 2011).

I guess my question is two-fold:

  1. If you read a restart file which was generated using one potential, but now declare a new potential in your input script, which potential is actually being used? Is it the one that was used when the restart file was generated, or the one declared in your input script. (If it’s the former, this would explain things)
  2. If the latter of the above bullet applies, then would anyone care to explain why the different behaviors exist?

Thanks a lot for your help,

Amit

Hello,

I’m not sure if the following behavior is supposed to happen. We’ve created an initial configuration and ran the simulation with the soft potential to avoid a system blow.
The simulation ran for 100,000 timesteps and a restart file was generated using the final timestep configuration.
From here on, we switched to the lj/cut potential, and this is where things get a little weird.

If we use the `read_restart’ command, the simulation runs fine with the lj/cut potential.
However, if we use the restart2data tool to convert the binary restart file to ASCII, and then use the read_data command along with declaring the following pair coeffs: pair_coeff 1 1 1.0 1.0 2.5, the simulation blows up (we only have 1 atom type).

I would gladly share the configuration file as well as the input script if necessary. Also, if this is any help, we are using LAMMPS version (9 Mar 2011).

I guess my question is two-fold:

  1. If you read a restart file which was generated using one potential, but now declare a new potential in your input script, which potential is actually being used? Is it the one that was used when the restart file was generated, or the one declared in your input script. (If it’s the former, this would explain things)

every time you place a XXX_style, e.g. pair_style command in
the input, the current style will be deleted (and all settings related to it)
and then a new style will be created. in this case it would be
highly important to place the new pair_coeff command after
the pair_style command (LAMMPS should refuse to run, though).

my guess is that in your read_data version of the input, you are
missing some setting that was in the first input, so it was imported
from the restart but not through read data.

  1. If the latter of the above bullet applies, then would anyone care to explain why the different behaviors exist?

no idea. you have to post a (small!) input example that demonstrates
the problem, so that it would be easy to reproduce it and determine
the reason. somebody stole my crystal ball, so i have to see things
to find out what is wrong… :wink:

cheers,
axel.