Question on creating an uniform velocity -- problem with high temperature

Dear Lammps members,

I am trying to create a argon flow with uniform velocity in a channel.

The purpose is to , then, introduce a carbon nanotube (cnt) inside the channel and study the drag on the cnt.

In the description below, there is no cnt yet.

I have also enclosed the lammps code as attachment.

I am working with a small system of 800 argon atoms in a channel size 54 X 27 X 27 ( angstroms ).

This is compressed liquid argon at 1342 kg/m^3 density, and at temperature 125K.

Using periodic domain, and “atom” style for argon.

I do the computation in two parts:

  1. run a NVT simulation to equilibrium at 125K.

  2. then run NVE simulation for uniform flow.

For the NVE simulation:

  1. I partition the channel into 3 sections: entrance, center and exit sections.

The entrance and exit are 20% of the channel length each.

  1. At the entrance and exit, at each time step I reset the thermal velocity at 125K and add
    desired flow velocity. This maintains the temperature and velocity at those two sections.

  2. I expect that will establish a similar state at the center section also.

The average velocity at the center section reaches the desired value, but the temperature

is way too high around 3500K ( depending on the speed ).

If the velocity bias is not removed at the entrance/exit sections, then the temperature is about

1000K; with bias removed the temperature is 125K as desired.
So the high temp. of 3500K at center is way above the velocity bias.

Could anyone suggest the cause for this high temperature ?

  1. Is it because the speed is not properly removed in computing the temperature ?

I am using “temp/com” lammps command that is supposed to remove the velocity bias.

This command seem to work because, if I just use “temp” command the temperature

at the entrance/exit section is about 1000K and not 125K as desired.

  1. Perhaps the initial NVT run is not long enough, and so the initial energy is too high in
    subsequent NVE run ?

  2. Is the concept correct , to start with ?

concept == maintaining entrance/exit sections at a certain state and expecting the same state at the center section.

I would appreciate any help.

regards,
shankar

uniform_flow_argon.lmp (28.9 KB)

  1. At the entrance and exit, at each time step I reset the thermal velocity at 125K and add
    desired flow velocity. This maintains the temperature and velocity at those two sections.

How do you add flow velocity and how do you maintain it during a run, within
geometric regions.

Steve

How do you add flow velocity and how do you maintain it during a run,

within

no time to look thru all of that. This seems
like a very complicated scheme for maintaining
a velocity flow. I suggest just using fix addforce
on the solvent atoms to do a Poisuelle flow, as
in examples/flow/in.flow.poiss. Whatever
you do, you will have to be careful how you
measure the temperature of flowing atoms
to get a value that is only the thermal temp.

Steve

Dear Steve,

In presence of cnt, the flow is not uniform around the cnt anymore.

So I was not sure if “addforce” was appropriate. Hence the scheme I have now.

Fix addforce has a region option. You could only add a force kick
to flow atoms at the end of your box. I think this is a better way
to create a steady flow than what your complicated script is trying
to do.

Steve

Dear Steve,

I suggest just using fix addforce

I will investigate this option.

Whatever you do, you will have to be careful how you
measure the temperature of flowing atoms
to get a value that is only the thermal temp.

Is there any research paper or reference that you could suggest ?

regards
shankar

no paper to suggest. Rather, I am saying
that when you have a flowing system
and/or non-equilibrium system you
need to think carefully about what you
mean by “temperature” and use the
various thermostat and compute temp/* options
in LAMMPS judiciously.

Steve

Dear Steve,

Thank you – I have been trying various temp/* options.

The temp/profile option removing averaged velocity gives much reduced temperature

than temp/com for the same problem I was using so far.

regards
shankar