Simulation of water in nanochannels using Moltemplate files and couette flow example

LAMMPS LAMMPS Mailing List Mirror
1

Dear All,

I changed the components of moltemplate-files in “Carbon-Nanotube Capillary” to generate the molecules of water between two plates. Based on couette flow example, I also added the top and bottom walls to the input file and defined the velocity for the top wall (Velocity= 10 Ang/femto). The velocity distribution shows that the velocity of water in the channel is almost zero and the velocity of the last bin (wall) is only 10. I was wondering why the velocity of liquid inside of channel is almost zero? I have used similar input file without data file for Argon before and I got accurate velocity distribution. The input, dump file and velocity profile were attached to this email.

Best regards,
Hamed.

Image.tif (846 KB)

velocity.tif (12 KB)

traj_nvt.lammpstrj (1.64 MB)

vel.profile (7.64 KB)

2

Dear All,

I changed the components of moltemplate-files in "Carbon-Nanotube Capillary"
to generate the molecules of water between two plates. Based on couette flow
example, I also added the top and bottom walls to the input file and defined
the velocity for the top wall (Velocity= 10 Ang/femto). The velocity
distribution shows that the velocity of water in the channel is almost zero
and the velocity of the last bin (wall) is only 10. I was wondering why the
velocity of liquid inside of channel is almost zero? I have used similar
input file without data file for Argon before and I got accurate velocity
distribution. The input, dump file and velocity profile were attached to
this email.

this really isn't a question about LAMMPS but about the science. what
you are asking about is to understand what makes these processes work
and drives them at a fundamental level. there is a difference between
argon and water and how they interact with each other and your wall. i
suggest you discuss with with your adviser and or more experienced
colleagues and perhaps also have a look into suitable (comparative)
simulation studies and text books and what drives the formation of a
velocity profile and its shape.

if you want to be successful in doing research, you will have to
figure out such things and/or come up with a hypothesis that attempts
and explanation which can be confirmed or disproved by further
simulation studies.

axel.

3

My apologies for the long delay. (Also, please disregard the other email I sent you today privately. This email is more complete.) Here are some suggestions:

  1. It sounds like you want to move one plate at a constant velocity in one direction, while keeping the other plate stationary in order to exert a shear force on the liquid. There are several other ways to do this. My favorite is “fix move”:

http://lammps.sandia.gov/doc/fix_move.html

several other fixes come to mind, although I think “fix move” is probably the best choice (because you can leave the force-field between carbon atoms unspecified):

http://lammps.sandia.gov/doc/fix_deform.html
http://lammps.sandia.gov/doc/fix_addforce.html
http://lammps.sandia.gov/doc/fix_momentum.html
http://lammps.sandia.gov/doc/Section_commands.html#fix-styles
Comment: I don’t think you should use “addforce” or “momentum” in this case, unless you have
defined a suitable force-field to keep the carbon atoms in place. (Moltemplate’s graphene example does not bother to define a force-field between the carbon atoms, but it does not allow them to move, so it does not matter in that case. I’m not sure if this is helpful, but the AIREBO force-field has been used in LAMMPS to simulate bond-breakage in carbon nanotubes, although it might require using a smaller timestep.)

Additional comment: If you decide to use “fix move”, then you should get rid of the
line which applies “fix nvt” to these boundary atoms:

fix 1 boundary nvt temp 300.0 300.0 100.0

(From the documentation: “The atoms affected by this fix should not normally be time integrated by other fixes (e.g. fix nve, fix nvt), since that will change their positions and velocities twice.”)

  1. Also keep in mind that you should probably pick the spacing between the carbon layers, and the number of water molecules between them carefully so that the density of water is realistic at this pressure. (Probably the best way to do this is by using an NPT simulation, although we warned that this can tricky to do correctly in LAMMPS.)

  2. You need to think more carefully about the units you are using and the timescales of your system. (Axel has a point.)

Why aren’t the water molecules moving? How long will it take for the water molecules to reach uniform shear-flow? It definitely won’t happen immediately. You only ran your simulation for 9000 time steps. (On top of that you were using a timestep of 0.01fs (which is 200 times smaller than the recommended timestep). I would not expect anything to happen on a 90fs timescale. If I remember correctly, aqueous hydrogen bonds last longer than that. Perhaps you can use a more reasonable timestep if you slow down the movement of the walls. 10 Angstroms per femto-second is ridiculously fast, compared to thermal movement at 300K.

Use “fix move” instead of “velocity”, use a more sensible velocity (I don’t know, how about 0.002 Angstroms per fs), a larger timestep (1fs or 2fs), and then run a long simulation (say 200000 timesteps) and visualize the movement of the atoms using OVITO or VMD.

I hope that helps a little bit.

Andrew
P.S.
It’s probably a bad idea to post a trajectory/dump file to the LAMMPS mailing list. Thankfully yours was small.