Pseudopotential multiphase method boundaries and thermal coupling

[mithdradates]

Hello. I’ve started working on a simple droplet evaporation problem, but I’ve run into several problems:
1) As far as I understand basic OpenLB (single component) Shan-Chen implementation only allows either periodic or bounce-back boundaries ? Because whenever I try to impose e.g. Dirichlet or Neumann BC the code fails. But I’ve seen e.g. Zou-He constant pressure boundary utilized together with SCMP Shan-Chen model in literature (although, no one mentions how they implemented it in order to work with SC model…). Does anyone know how should I modify the SCMP code in order for it to work with different boundaries ?

2) Second difficulty I ran into is the problem with thermal models. I run SCMP with Carnahan-Starling EOS (through interaction potential) and when I try to couple my Adv-Diff lattice through it via the modified operator() which allows temperature changes, the code fails. How do I couple the temperature from Adv-Diff with fluid then ? I don’t think Boussinesq approximation works in multiphase case because of major density variations in the liquid-vapor interface.

I’ve tried two models (Gong, Cheng, 2012 and Kamali et. al, 2013), but to no avail. Temperatures produced by them are adequate – my initial temperature is 0.8 * Tc and the variations due to spurious currents are in the range of 0.76 * Tc – 0.82 * Tc which is fine and shouldn’t lead to major instabilities (at least to my knowledge). I add additional terms from these models through source term in sourced Adv-Diff.

3) Could someone explain why source term is implemented as follows:
source[0] * (1 – 0.5 * omega) * t_i ? I’ve usually seen in literature that source terms is simply equal to source[0] * t_i, while the additional multiplication by (1 – 0.5 * omega) is used for various correction terms (e.g. to compensate for force term in fluid equation).

Thanks in advance!

• This topic was modified 5 years, 4 months ago by mithdradates.

[mgaedtke]

Hi mithdrates,

I assume you use the latest version of OpenLB, 1.3? If not, I recommend checking the latest version first.

to 1) The SC-Coupling accesses neighboring cells and calls calcRho on them. However, for boundary cells not every neighbor is a valid cell. Thus try to add the coupling procedure only on fluid cells, rather than the whole lattice. Do so by changing the line sLattice.addLatticeCoupling( coupling, sLattice ); to sLattice.addLatticeCoupling( superGeometry, 1, coupling, sLattice);. This will execute the coupling only on cells with material number 1.

to 2) If I understand correctly, what you could try is setting a constant gravitational force to the whole domain. The thermal SC-Method will then change the density for every cell according to the EOC, which will result in buoyancy. Stability with these coupled methods is not easy to achieve, I recommend to use proposed parameters from the literature first and then evolve from that to analyse the stable domain.

to 3) the sourced advection-diffusion dynamics are implemented after

Seta, T. (2013). Implicit temperature-correction-based immersed-boundary thermal lattice Boltzmann method for the simulation of natural convection. Physical Review E, 87(6), 063304.

Have a look in there for details of the source term.

Best regards,
Max

[mithdradates]

Hello, Max

Yes, I’m using the latest version of OpenLB.

1) Ok, I’ll try. Thanks for the suggestion!

2) Sorry, I don’t quite understand why I need gravitational force in the domain ? My initial idea was to use thermal SCMP SHan-Chen method as you suggested – to change density in every cell of the lattice according to their respective temperatures through EOS, but the simulation quickly destabilizies :\ Yes, I’ve tried with numbers from literature, the result is the same.

3) Okay, thanks!

[mithdradates]

I’m unable to edit my post, I’d like to add that I tried your suggestion 1 and the results are still the same. I can’t even implement a solid wall boundary, because the code instantly fails when periodicity is turned off. Any help would be appreciated.

• This reply was modified 5 years, 4 months ago by mithdradates.
• This reply was modified 5 years, 4 months ago by mithdradates.

[mathias]

Please, have a look of how it is realized with similar code and change it step by step. Also, we will have a spring school next year where we explain the code also step by step.

The cells in the envelope (ghost cells) must not be changed manually!

[mithdradates]

Thank you, I was able to implement a solid wall boundary via virtual densities method. But I’m still struggling with other boundary conditions. I have a question about envelope cells in SCMP Shan-Chen model – if e.g. I’m using SCMP for fluid only (material number 1) and fluid is constrained by boundaries on edges (material number 2), then “envelope” cells used in SCMP (newX0-1 etc.) are the boundary nodes or still the actual envelope nodes (outside of computational domain) ?

• This reply was modified 5 years, 4 months ago by mithdradates.
• This reply was modified 5 years, 4 months ago by mithdradates.

[mgaedtke]

Hi mithdradates,

that depends on your definition of the coupling. If you chose to run the coupling over material 1 cells, the envelop cells will be either other cells of material number 1 or cells with other material numbers in the direct neighborhood. So, make sure, that every value and or function, that is accessed in the coupling processor, is accessible for material number 1 cells as well as boundary cells.

Best,
Max

[mithdradates]

Okay, got it. Thanks!

[Author]

Posts