LubricateU model with simple shear gives particle overlaps

Dear LAMMPS users,

I am using an April 2013 version of LAMMPS built with the FLD package to simulate simple shear in monodisperse spherical particles at high volume fractions. For the test case I am using (script below), I am finding that there are particle overlaps occuring during the simulation. I was under the impression that a dissipative model such as this would prevent overlaps completely and was therefore wondering if someone who has experience with these simulations could say if I am setting up the simulation incorrectly or if I have misunderstood the lubricate model.

Many thanks,

Chris

atom_style sphere

boundary p p p

newton off

communicate single vel yes

region reg prism 0 6 0 6 0 6 0 0 0 units box

create_box 1 reg

lattice sc 0.1

create_atoms 1 box

pair_style lubricateU 0.1 1 1.6 2.0 1

pair_coeff * *

timestep 0.00001

velocity all create 1 4928459 dist gaussian

fix 1 all nve/noforce

compute 1 all erotate/sphere

compute 3 all temp/sphere

compute 4 all pressure 3

compute 5 all contact/atom

thermo_style custom step c_3 c_4 vol atoms

thermo 100

thermo_modify lost ignore norm no

compute_modify thermo_temp dynamic yes

dump id all custom 100 hello.dump id x y z vx vy vz fx fy fz c_5

run 100000

Chris,

In the Lubrication model, you prescribe a cut-off for the lubrication solution forcing. It does not diverge at contact, as the analytic model predicts for Stokes flow. If your cut-off is too large, particles may come into contact and no longer be able to ‘feel’ each other. This could be especially problematic if you are working with a dense system.

Jeremy may be able to advise. I don’t see any
obvious problem with your script.

Steve

Ok I understand what you mean, thanks for the reply. I have now changed the script to include a colloid-type potential and it seems to have had the desired effect:

pair_style hybrid/overlay colloid 1.2 lubricateU 1 1 1.1 1.7 1

Chris

Just something quick I also noticed. It appears that you have not declared a diameter for your spheres. You could probably get away without using double-layer forces(unless they are relevant to you) if you declare the cut-inner to be small enough (just larger than the minimum contact distance) and resolve the dynamics at a sufficiently short time-scale. Good luck.

Hi Chris,

Sorry for the slow response. Have you solved the problem? If not, can you send a sample of the output you’re getting? Do overlaps occur right away?

Jeremy

Hi Jeremy,

No problem, thanks for getting back to me.

I have solved the problem to some extent…

Initially I was using only the lubricate pair style, with cutinner 1.1 and cutouter 2. I would always see overlaps occurring soon after the start of the simulation.

Now I am using only the lubricate pair style, but with cutinner 1 and cutouter 2. Now I never see any overlaps, unless I force the system into a state of very high volume fraction (eg with fix deform), at which point the simulation crashes.

It seems that in the second case, as Eric says the force explodes at particle contact i.e. when the cutinner 1 is satisfied. However, in the first case the force does not diverge even when the particles are overlapping. Is this because the value of cutinner is greater than 1, so the force calculation never “sees” particles in contact, it only sees them at a separation of 1.1, even if they are in fact overlapping? This would seem to explain what is happening. In that case, could it be said that with a cutinner=1, the lubricate model alone is sufficient to prevent overlaps, but for cutinner>1, a colloidal interaction or similar would also be needed if overlaps are to be avoided?

Thanks for your time and help,

Chris

Chris, glad to hear you’ve got things going. Sorry I wasn’t much help, but I’m sure the head scratching will pay off in the end J. Answers to your Q’s below:

(1) Is this because the value of cutinner is greater than 1, so the force calculation never “sees” particles in contact, it only sees them at a separation of 1.1, even if they are in fact overlapping?

Yes

(2) This would seem to explain what is happening. In that case, could it be said that with a cutinner=1, the lubricate model alone is sufficient to prevent overlaps, but for cutinner>1, a colloidal interaction or similar would also be needed if overlaps are to be avoided?

Yes

Regards,

Jeremy