shear simulation issue

Submitted by jagan1mohan on Sat, 06/06/2020 - 22:00

Hello Team,

Greetings. I hope, you are all doing good. I'm working on homogeneous simple shear simulation using the following two methods.

In first method, I'm taking a cube of side 7.5 and filling it with 483 particles each of diameter 1 and density 1000 so that packing / volume fraction is 0.6. There is only 1 atom_type and all six faces are periodic. Particle fiction is 0.1 and there is no gravity for time being. I apply fix deform xy with a known value of dimensioned shear rate, say 1/sec, so that the top plane (ymax plane) moves along positive x-direction with a velocity of 7.5. Script writes out VTK files and I observe that the simulation box / region tilts correctly and at steady state, using spatial averaging I compute x - velocity gradient which closely matches with applied shear rate. This I have tried with lower as well as higher shear rates and also for different particle friction values and it works well.

In second method, I want to mimic same phenomenon by moving top and bottom walls in opposite directions (and hence not tilting box and not using fix deform). Rest all settings are same. I first used fix wall/gran on top (ymax) and bottom (ymin) planes to create walls of same atom_type and used shear keyword along with +3.75 and -3.75 units as velocity, respectively. The particles, close to the wall move very very slowly with velocity several orders of magnitude less. This could be due to the less friction between the wall and the particles. In next attempt, I created walls of particles of same diameter, density in a staggered arrangement such that the particles just touch each other. I'm using fix move on these two walls and I observe that a bit more momentum is transferred as compared to fix wall/gran but falling short by 40%.

1. If I use only fix wall/gran and declare walls as type 2, then what should be the wall-particle friction value? Perhaps, LIGGGHTS can take any value and can this value be as high as possible?

2. In particle wall method, I'm not integrating the wall particles using nve/sphere as I'm already moving them with fixed velocity. How can I improve this setup to achieve the correct shear rate?

Please let me know if you have any additional questions. I can share the scripts if you want more details.

Thank you,
Jagan Mohan.

Forums:

mschramm | Tue, 06/09/2020 - 22:15

DISCLAIMER...
I mostly do direct shear simulations... (like https://www.youtube.com/watch?v=Ti0_LxGVN94 (not me))

Hello,
1) I do not believe there is a limit on how high your friction values are. The exact value depends on the simulation and may take some trial and error to determine the correct coefficient.

2) again, trial and error on the part of the friction force between your "wall" atoms and your remaining atoms.

Just because your wall moves at a specific speed does not guarantee that the atoms next to your wall have to move at the same speed. This is where your friction value becomes important.

jagan1mohan | Thu, 06/11/2020 - 16:02

I'd try the "primitive wall" method and update you both.

tjleps | Wed, 06/10/2020 - 02:02

You may want to try creating the particle wall out of your random packing a couple particles deep instead of using a regular crystal lattice. that way there will be holes to mechanically capture non wall particles and you won't be shearing across a crystal plane.

RonMexico | Mon, 12/28/2020 - 21:39

1) Use moving walls with friction on the top and bottom.
2) Use a layer of moving particles on the top and bottom bounded by an elastic solid wall.

The second seems to work better since steady-state can be reached. In both cases, you need to determine the shear rate in the simulation by taking du/dy since the applied rate via moving particles or wall will not match.

tjleps | Mon, 12/28/2020 - 22:30

I strongly recommend, when using particles as a moving boundary for shear simulations, to fill your simulation box fully with a random packing, then choose a region from the boundary of your random packing as your moving boundary. This prevents having a low friction, highly organized crystal structure sliding across the bulk of your particles which are relatively randomly packed. The exception would be if you expect large scale crystal structure in your particle packing for whatever reason.

RonMexico | Fri, 01/08/2021 - 17:53

I had not explicitly stated that, but I create my moving boundaries with as high of a randomly packed layer of particles as LIGGGHTS' insert/pack allows. The randomness allows the actual particles of interest more of a chance to interact with the moving layer. However, I did not notice a significant difference when seeding the moving layer to the same volume fraction as the domain of interest. As long as particle counts were high enough, it did not matter. I think the solid walls on the top and bottom are especially critical for this reason.