# [lammps-users] Couette's flow with polymer walls

As far as creating a starting configuration, you are mostly on your own. LAMMPS does have some functions for generating lattices of atoms (maybe some other stuff too, I don't know), but I generally make my systems "by hand". That is, writing some code to do it or finding bits of code on the internet that do what I need. When you do this you will need to pay careful attention to the way bond and angle information is included in the configuration input.

I'm not very familiar with Poiseuille flow. The few papers I have read simply add a constant force in the desired flow direction. LAMMPS has an addforce fix that can do this. I'm not sure how realistic this is, though.

If your system is only periodic in two directions - like a liquid confined between two infinite, parallel planes - you will need to use a 2d summation for the electrostatics. Look at kspace_modify slab in the documentation.

You need to be particularly careful with the thermostat. If you thermostat in the flow direction, it will kill the flow. LAMMPS considers the flow "temperature". There is some functionality in LAMMPS for subtracting the translation due to a flow profile, but you need to know the velocity profile to begin with.

But before you build a system I suggest you start with the work Denis Evans has done with NEMD. http://rsc.anu.edu.au/~evans/index.php It should save you a lot of false starts and useless simulations. Good luck.

Matt

Quoting "Romashevsky S." <[email protected]...>:

All of Matt's advice is good. I just add one point:

You need to be particularly careful with the thermostat. If you
thermostat in the flow direction, it will kill the flow. LAMMPS
considers the flow "temperature". There is some functionality in
LAMMPS for subtracting the translation due to a flow profile, but you
need to know the velocity profile to begin with.

There is a compute temp/profile, which when used as a thermostat,
will subtract out the current actual T profile and treat the remaining
velocity as "thermal". This is different than compute temp/deform
which assumes a profile that matches the box deformation (for NEMD).

Steve