Dear Lammps users,
I have a system which includes a huge surface containing ~18000 atoms. To increase the simulation speed, I want to fix the surface. I used “fix freeze surface setforce 0.0 0.0 0.0” (surface is defined in group command).
But when I run the system the energies increases tremendously from -1357243.4 to 1.1113726e+11. Energy breakdown shows that vdw energy and pressure are both very high.
Am I missing anything in fixing the surface?
Thank you
Fatima
Dear Lammps users,
I have a system which includes a huge surface containing ~18000 atoms. To
18,000 huge? not for lammps!
steve is getting the code ready to handle more than 2 billion atoms. 
increase the simulation speed, I want to fix the surface. I used "fix freeze
surface setforce 0.0 0.0 0.0" (surface is defined in group command).
the main cost of computation is not the integration of the equations of motion,
but (usually) the calculation of the forces. that would still happen, so there
is no significant gain.
But when I run the system the energies increases tremendously
from -1357243.4 to 1.1113726e+11. Energy breakdown shows that vdw energy and
pressure are both very high.
Am I missing anything in fixing the surface?
to have atoms not move, you can just ignore them in the
time integration. i.e. apply fix nve or nvt only to the group
of mobile atoms.
but see my point from above, that this will not speed up
the calculation significantly.
axel.
Thanks Axel for your reply,
I learnt how to fix the surface.
About the speed, this 18000 atoms cost me 13 days to run 1ns on 8 processors. Of course I can run it on 32 processors and reduce 13 days to 3 but I have many of these simulations … More importantly, what is studied here is the interaction of some molecules with the surface. The surface itself is not that important for me which is currently responsible for the expensive simulations. Is there anyway that I can exclude the force calculation of surface atoms from the simulation?
Regards,
Fatima
You can use the neigh modify exclude command to exclude
certain interactions from being calculated. However, as Axel
says, unless that is a large fraction of your system, it is unlikely
to provide a dramatic speed-up.
Steve