Re: [ESPResSo-users] Simulation of fluid using LB-method

[Stefan Kesselheim]

Hi Markus,

On Jan 30, 2014, at 3:03 PM, "Wink, Markus" <address@hidden> wrote:

> Hello everybody,
> 
> I tried to think about of the real position of the wall: 
> "The (0, 0, 0) node of the LB lattice in Espresso is placed at (agrid/2, 
> agrid/2, agrid/2)." 
> In my case, I chose a_grid =1, so I put the walls at 1.5 and 51.5 in the 
> simulation box. Thus the length of the simulation box is 53 in my case, 
> gibing a channel width of 50. 
> Am I right, or do you detect any mistake in my thoughts?

No, this is not quite it. With our bounce-back boundary conditions the effect 
hydrodynamic walls are located between two lattice nodes. Therefore the right 
way to go is putting walls to integer multiples of the lattice constant, and 
then you get it where it is supposed to be. If you put it at 1.5 it depends if 
some programmer put a ">" or a ">=" into that piece of code which determines if 
a certain node is "inside" or "outside" the wall. So, please put your walls at 
some integer position and check if it becomes right. For the a=0.5 case it 
should be OK, but I'm not sure why there are still deviations. Did change the 
parameters to the lbboundaries?

> 
> I am also puzzled with the negative slip velocity. Attached, you will find 
> the simulation for a_grid = 0.5. I am not sure, but probably that helps.

By negative slip, Ulf is referring to velocities which are smaller than the 
theory curve. Try cranking up the viscosity to 10 with a time-step of 0.1, then 
you should see a positive slip :-). Interestingly this can even lead to the 
following: For a=1 everything looks perfect, but for a=0.5 the viscosity in LB 
units goes up, and the boundaries start producing slip and everything is f**ed 
up.       

Cheers
Stefan