Dear all,
I would like to model flexible water, co2 and solid surfaces in my system with LAMMPS, which requires a very small time step (for flexible water), i.e. dt = 0.25 fs. I think r-respa may be one way to speed up the simulation. According to the LAMMPS document, it suggests using the following setup, so that the innermost time step is 0.5 fs:
timestep 4.0
run_style respa 4 2 2 2 inner 2 4.5 6.0 middle 3 8.0 10.0 outer 4
I will try this set up in my simulations with timestep = 2 fs so that the innermost timestep is 0.25 fs (for bonds and angles) . My question is: apart from checking the energy conservation, is there any “pitfalls” about using the r-respa method in molecular dynamics simulations that I should pay attention to? It will be great if anyone who has had experiences with the r-respa method can provide any comments, suggestions or hints. All comments are welcome.
Thank you for your time.
Regards,
Pengyu
Dear all,
I would like to model flexible water, co2 and solid surfaces in my system
with LAMMPS, which requires a very small time step (for flexible water),
i.e. dt = 0.25 fs. I think r-respa may be one way to speed up the
simulation. According to the LAMMPS document, it suggests using the
following setup, so that the innermost time step is 0.5 fs:
timestep 4.0
run_style respa 4 2 2 2 inner 2 4.5 6.0 middle 3 8.0 10.0 outer 4
I will try this set up in my simulations with timestep = 2 fs so that the
innermost timestep is 0.25 fs (for bonds and angles) . My question is: apart
from checking the energy conservation, is there any "pitfalls" about using
the r-respa method in molecular dynamics simulations that I should pay
attention to? It will be great if anyone who has had experiences with the
r-respa method can provide any comments, suggestions or hints. All comments
are welcome.
you are confusing multi-cutoff separation with bond/angle/pair separation.
you are not going to save much time with your setup from above, since
you are computing a lot of interactions multiple times.
you probably want something like this:
timestep 4.0
run_style respa 3 4 2 dihedral 1 pair 2 kspace 3
however, the biggest problem is not the flexible water, but the co2.
for that you would need a rigid integrator and those will require a
short time step anyway to run stable.
so using r-RESPA may not be as helpful as you expect.
axel.