# barostat rescaling problem

Dear all,
I am simulating a lubricant confined between two rigid plates (evolved as rigid bodies), the top one is mobile and driven along y through a spring, the idea is to study the tribological properties of the system.
Along x the lubricant can flow in/out of the plates into “lateral reservoirs”. To keep the reservoir pressure constant I applied along x a Berendsen barostat.
At the same time, to keep the plates aligned I constrained their x component of the force to be zero.
At this point I was expecting no motion of the top plate along x, however it moves! The problem disappears as I switch off the barostat. I think this is a problem due to the barostat rescaling, however with the option “dilate partial” only the lubricant particles should be rescaled, am I right? Would it be possible to completely suppress the rescaling also in the lubricant and just move the cell walls to change the volume letting the system to find the correct density?

Attached you will find input and output of the calculations with and without the barostat.
Column 5 of the output is the x position of the top plate, constant with no barostat and variable with barostat.
Column 5 of the output is the x component of the force acting on the top plate, this is correctly set to 0 in both cases.

The initial velocity of the top should be automatically set to zero.
initpos.dat contains the initial coordinates and cell size.

Thanks for your help, best regards.
Andrea

input_barost (3.39 KB)

input_nobarost (3.39 KB)

output_barost (1.42 KB)

output_nobarost (1.42 KB)

initpos.dat (421 KB)

Dear all,
I am simulating a lubricant confined between two rigid plates (evolved as
rigid bodies), the top one is mobile and driven along y through a spring,
the idea is to study the tribological properties of the system.
Along x the lubricant can flow in/out of the plates into "lateral
reservoirs". To keep the reservoir pressure constant I applied along x a
Berendsen barostat.

why do you use a barostat on such a system? there is no need.
use a fixed size box. do not time integrate the particles that you
want to keep immobile and use fix aveforce to on the second, mobile
plate. you can only allow movement in that direction through setting
the velocities and forces for the rest to zero. and then just time
integrate.

At the same time, to keep the plates aligned I constrained their x component
of the force to be zero.
At this point I was expecting no motion of the top plate along x, however it
moves! The problem disappears as I switch off the barostat. I think this is
a problem due to the barostat rescaling, however with the option "dilate
partial" only the lubricant particles should be rescaled, am I right? Would
it be possible to completely suppress the rescaling also in the lubricant
and just move the cell walls to change the volume letting the system to find
the correct density?

yes. see above. using a barostat is not a good approach for your system.

axel.

Indeed Axel is correct. This is a typical procedure for such systems.
(Personal blurb) See:
http://www.apmaths.uwo.ca/~ccampana/Dec_2_seminar/shear_flow.mpg

for a movie that does exactly what Axel suggested. Yet, unless you are

looking at some generic FF to simulate the lubricant and walls, I would

suggest to put some extra effort into finding good parameters for the
lubricant-wall interactions if you suspect relevant action to occur there.
They can make quite a different in some cases such as the one portrait
in the movie where you have hexene molecules confined between Al
surfaces. If interested, the movie aimed to reproduce results from the paper:
http://dx.doi.org/10.1088/0965-0393/18/3/034004

Carlos

Carlos,

After reading your response and Axel's I still don't know exactly how
you allow the wall-spacing to fluctuate. Your paper just says "In all
simulations, we keep the normal pressure,P⊥, constant, i.e. the upper
wall is allowed to ﬂuctuate normal to the interface." Are you applying
fix aveforce and fix addforce in the normal direction to the upper
wall?

Aidan

Hi Aidan,

One correction: The paper I listed wasn’t mine. In my movie, I use fix aveforce to set the force on the upper-wall atoms according to the desired pressure given by Total_force/Total_area.

Carlos

Maybe I should add that the reason why you see the wall fluctuating “a bit too much” in my movie is because that movie is intentionally started when the upper wall start sliding thus the extra energy injected into the system is compensated by a reduction in the lubricant density (walls separate).

Carlos

Hello,

You may also have a look at the "DPD thermostated liquid of LJ chains
sheared between (111) fcc walls" script on lammps website ("user
scripts" page). It does something very similar to what you have in
mind.

Best regards,
Laurent