AIREBO potential problem in parallel

Hi all,

I am simulating the penetration of He atoms through graphene-based surfaces. I use the AIREBO potential for the interaction between carbons and a Lennard-Jones potential for de C-He interaction.

My problem is that when I run the simulation in parallel using two processors (with mpirun -np 2), the dynamic is a little bit different to the case of using a single processor. But when I turn off the torsion term of the AIREBO potential, the dynamic is the same using one or two processors. I don't know the reason of that. I think I should obtain the same dynamic using one or two processors even when the torsion term of AIREBO is turned on.

Thanks.

How did you assign your pair_style and pair_coeff commands?

Ray

Hi all,

I am simulating the penetration of He atoms through graphene-based
surfaces. I use the AIREBO potential for the interaction between carbons
and a Lennard-Jones potential for de C-He interaction.

My problem is that when I run the simulation in parallel using two
processors (with mpirun -np 2), the dynamic is a little bit different to
the case of using a single processor. But when I turn off the torsion
term of the AIREBO potential, the dynamic is the same using one or two
processors. I don't know the reason of that. I think I should obtain the
same dynamic using one or two processors even when the torsion term of
AIREBO is turned on.

​some questions about this:​

- ​can you please let us know what specific version ​of LAMMPS you are
using?
- can you also look at the output before starting a run and tell us what is
the ghost atom cutoff?
- how often you rebuild (and check) the neighbor lists?
- do you get any any "dangerous builds"?
- would you be able to post a small but complete test input?

thanks,
     axel.

  • I am using the 7 Dec 2015 version.
  • The ghost atom cutoff is 12.2 angstrom.
  • For neighbor lists, I use default values (update every 1 steps, delay 10 steps, check yes).
  • No, I don’t get any dangerous build.

Here I post a test input to simulate the transmission of He atoms through a graphdiyne sheet. The initial positions and velocities of the He and C atoms are read from the data file.
The differences in dynamics and energy when I use 1 or 2 processors appears when the graphdiyne sheet starts to oscillate in the perpendicular direction, due to the interaction with He with AIREBO potential.

####################################
# 3d simulation of graphdiyne-He
####################################

#------------------------------------------------------------------------------
# setup of the simulation
#------------------------------------------------------------------------------
units metal
atom_style atomic
boundary p p f

#------------------------------------------------------------------------------
# read positions and velocities from the data file. Define pair interactions
#------------------------------------------------------------------------------
read_data data.graphdiyne-He

pair_style hybrid/overlay airebo 3.0 lj/cut 8.0
pair_coeff * * airebo …/potentials/CH.airebo C NULL
pair_coeff 1 2 lj/cut 0.001209 3.20291 8.0
pair_coeff 2 2 lj/cut 0.0 3.2 8.0

#------------------------------------------------------------------------------
# computations and outputs
#------------------------------------------------------------------------------
fix 1 all nve

thermo 100
thermo_style custom step etotal
thermo_modify norm no

timestep 0.00025
run 20000

Thanks,

Alfonso.