Dear LAMMPS users
I am using LAMMPS to study water-methanol mixture.
I am trying to use NPT to equilibrate the system, and I find something not normal.
The system consists 600 Tip4p/2005 water and 85 methanol molecules.
If I use NPT with isotropic barostat (Nose-Hoover), the system is fine.
But when I use anisotropic barostat (Nose-Hoover), the volume fluctuates very widely.
I played with drag factor, relaxation time, and reduced time step, but they are not solving the widely fluctuation of volume.
I did the same run for pure methanol and water, and isotrobic and anisotropic barostat deliver similar results.
I am wondering is this because the dipole moment of methanol and water molecule is not the same?
I attach my input file for you guys to check
Cheers,
Tzu-Jen Lin
in.tip4p-300k-pre (2.46 KB)
Tip4p_MeOH_20wt.data (247 KB)
Dear LAMMPS users
I am using LAMMPS to study water-methanol mixture.
I am trying to use NPT to equilibrate the system, and I find something not
normal.
The system consists 600 Tip4p/2005 water and 85 methanol molecules.
If I use NPT with isotropic barostat (Nose-Hoover), the system is fine.
But when I use anisotropic barostat (Nose-Hoover), the volume fluctuates
very widely.
please define "very widely". ...and over what time step range?
I played with drag factor, relaxation time, and reduced time step, but they
are not solving the widely fluctuation of volume.
I did the same run for pure methanol and water, and isotrobic and
anisotropic barostat deliver similar results.
I am wondering is this because the dipole moment of methanol and water
molecule is not the same?
this makes not sense. what has the dipole to do with this?
more likely there is something else going on in the simulation that
causes issues.
i would remove anything from your input that is not essential, e.g.
analytic tail correction for pressure (since you already use kspace,
you can use pppm/disp/tip4p
and lj/long/tip4p/long for this and have this treated fully
accurately), i would not use berendsen for equilibration, but rather
csvr, langevin, or nose-hoover(nvt) (in that order of preference),
remove fix recenter, and use a tighter kspace convergence, say 1.0e-6.
1.0e-4 is already at the limit for being usable for constant volume,
since here pressure is only a diagnostic. also having tighter
convergence in shake has been shown to be better for energy
conservation and thus accuracy of simulations.
axel.
Sorry for the not accurate description of the problem.
What I observed is that the cell length of the box oscillated widely, from 20 A to 40 A.
The time step I used is 1fs. For rigid water model, I think the time step should be fine.
The bond of methanol is also fixed, and the setting is from TraPPE force field.
I did try Langevin before, but the dimension still oscillated widely.
I will try the method you suggest like removing recenter, increase accuracy of SHAKE, or kspace convergence.
I never saw lj/long/tip4p/long this command in LAMMPS manual. You mean lj/cut/tip4p/long ?
If so, I had already used that.
Cheers,
Tzu-Jen Lin
Sorry for the not accurate description of the problem.
What I observed is that the cell length of the box oscillated widely, from
20 A to 40 A.
The time step I used is 1fs. For rigid water model, I think the time step
should be fine.
i agree a 1fs time step is a conservative choice. the question was
rather to ask how many time steps does it take to see one fluctuation
after the system has equilibrated
The bond of methanol is also fixed, and the setting is from TraPPE force
field.
I did try Langevin before, but the dimension still oscillated widely.
still berendsen is a very bad choice for equilibration. it doesn't
dissipate any fluctuations cause by the non-equilibrated initial
configuration but rather enhances them. it may not show in this
specific case, but its fundamental flaw remains.
I will try the method you suggest like removing recenter, increase accuracy
of SHAKE, or kspace convergence.
and tail correction.
I never saw lj/long/tip4p/long this command in LAMMPS manual. You mean
lj/cut/tip4p/long ?
If so, I had already used that.
http://lammps.sandia.gov/doc/pair_lj_long.html