Re: [ESPResSo-devel] LB fluid with variable viscosity

[Ivan Cimrak]

Hi Michael,

> to implement this, you would need a field that stores all four
> relaxation times individually (shear, bulk, ghost even, ghost odd) for
> each cell. This of course increases the LB's memory footprint and
> reading from a field is slower than accessing a constant, so this code
> should only be used when variable viscosity is actually needed.

Exactly, I am aware that I need this fields and I understand that the 
performance would be slower. I plan to implemented it in a way that the code 
will be compiled in only if I use an appropriate switch in myconfig.hpp. 

I plan to have one field that stores information (one real number) giving the 
distance to the nearest boundary a blood cell. I will need to update it often 
when my blood cells move. We are working on an efficient algorithm to do dat. 
But according to this value I will be able to assign proper shear, bulk, ghost 
even, ghost odd relaxation times. So only one real number for each lbnode will 
be necessary to store.

But I am not sure how to implement such a field. Could you please point me out 
to some similar variable in ESPResSo, co I can implement it in a similar 
manner? Or is there a guide how to implement a global field so that 
parallelization will be taken care of? I would like to work with CPU version of 
LB first, the GPU would come later…


Thanks, 
Ivan






> I don't know how well LB with non-uniform viscosity actually works. All
> applications that I have seen so far only do that to make the algorithm
> more stable in certain extreme situations, but I don't know whether it
> produces any reasonable physics. So unless you already know that it
> works for object-in-fluid, I'd be careful with non-constant
> viscosities. The one thing that comes to my mind is Geier et al.'s
> claim that "[c]hanging the relaxation rates locally usually results in
> non-constant transport coefficients" (
> https://doi.org/10.1016/j.camwa.2015.05.001), though I don't know how
> bad that is. Since you don't want transport between your two phases of
> different viscosity, you might actually be fine. In Shan-Chen-like
> multiphase models, different viscosities are commonly used, but they do
> something more advanced at the interface.
> 
> Regards,
> Michael
> 
> On Tue, 2018-11-20 at 16:55 +0100, Ivan Cimrak wrote:
>> Dear developers,
>> 
>> I plan to implement the possibility of variable viscosity in LB
>> fluid. We need this feature in the object-in-fluid module, where the
>> inner part of a flowing object has different viscosity than the outer
>> part.
>> 
>> I would like to discuss this issue with someone that is familiar with
>> LB implementation in ESPResSo. Could anyone give me hint whom to ask?
>> 
>> 
>> Thanks, 
>> Ivan Cimrak
>