Re: [ESPResSo-users] Setting up Lattice Boltzmann

[Stefan Kesselheim]

Dear Vincent,

I' like to ask a few questions back:

On May 6, 2013, at 3:12 AM, Vincent Ustach <address@hidden> wrote:

> Hello everyone,
> 
> I have a working script of spheres diffusing from a source to a sink through 
> a pore with pore walls simulated by particles that are fixed in space. The 
> box size is (30*sigma)^3. Rather than partial periodic conditions (x and y 
> periodic, z not periodic), the system has another wall of fixed particles at 
> z = 29*sigma so there is no diffusion from source to sink except through the 
> pore. The thermostat and solvent are simulated through dissipative particle 
> dynamics. I would like to add the Lattice Boltzmann system to the simulation 
> to model the hydrodynamic interactions and I have a few questions. 

* Why would u use DPD+LB? Are you referring to the (conservative) soft particle 
interaction of DPD type or the thermostat? Is it clear to you that inter_dpd is 
not the conservative DPD interaction?
* Are you familiar with the pore constraint? It might be a replacement for your 
particles. The wall constraint can probably replace the other walls.

> 1) Are there reasonable parameters to use as a starting point for lbfluid? 
> The fluid should be quiescent water with no external force. The wall particle 
> size is (2.55*sigma) and the diffusing sphere size is (0.1*sigma) as dictated 
> by the LJ interaction parameters which might influence the choice for the 
> agrid parameter. The time step for MD is 0.001. I am confused by the term "LB 
> units" in the user guide which is why I don't know where to start. 

LB intrinsically uses a different unit system, in which the (LB time-step is 
1), however the interface is always made so that you interact with it in 
(usual) MD units.
The right choice of LB parameters however is a tricky thing. It depends on the 
effects you expect in your system. Are you expecting collective transport (e.g. 
diffusophoretic effects), such as fluid transport due to a concentration 
gradient? Or are pairwise interactions important?

From my experience a good starting point are the LB parameters:
density 1 viscosity 0.8 friction 10 agrid 1 tau 0.01

Then the hydrodynamic radius of the particles will be 0.45 sigma. I'd recommend 
to scale up the rest of the system.

> 2) Will the walls (fixed particles) serve as hydrodynamic boundaries for the 
> fluid or do I need to add the LB boundaries as dictated by the pore 
> dimensions?

This is not clear yet. lbboundaries use the link bounce back method, which is 
considered a "good" method for boundaries. If a set of particles is a "good" 
boundary depends e.g. on the friction, on the density etc. Usually the 
combination of a "constraint" and an identical "lbboundary" is the way to go, 
but implementing own shapes is tedious.

Cheers and good luck
Stefan