[Andre van der Walt]

HI all, does anyone know the subcell division of dsmcFoam works? Is the subcells divided automatically so that the correct number of molecules per subcell is obtained. IS the grid that is initially specified only for sampling purposes? IS the time step varied over the domain, so that the time step is smaller at regions of high number densities? It is not clear how all this works, the selection of the number of equivalent dsmc molecules, the specification of the grid, and the specification of the timesteps.
I would appreciate it if anyone could point mein the right direction.
Andre

[TomS]

Hi Andre,

I believe the sub-cell division is carried out automatically in every sampling cell in order to promote nearest-neighbour collisions. The cell centroid acts as a locator to split any cell into 8 quadrants (sub-cells). The initial, un-divided grid is used for sampling.

The "correct" number of dsmc simulators per sub-cell is difficult to achieve and will change as the solution progresses. Good dsmc practice suggests an initial number of 10-20 particles per sampling cell. Particle number sensitivity studies should always be conducted in any case. The parameter nEquivalentParticles should be set to produce the initial 10-20 particles at the beginning of a solution using the utility dsmcInitialise.

The time-step is fixed (normally based on the mean free time or likely particle residence time). The grid is also non-adaptive and fixed (cell size based on mean free path) but this is not normally a restrictive issue.

Regards,
Tom

[TomS]

A review of the validation of dsmcFoam can be found at:

Scanlon, Roohi, White, Darbandi and Reese, " An open-source, parallel, DSMC code for rarefied gas flows in arbitrary geometries", Computers and Fluids, 39, 2078-2089, 2010.

Regards,
Tom