[Araist]

Dear all,

I'm running one hell of a case where I have to simulate mollecules of Helium passing through a skimmer (micro-hole).

I use an axi-symmetric geometry (using two symmetry planes at a small angle) which is composed of two regions: the inlet region, where the number density is bigger and the outlet region, where is much smaller.

The reason for the drop of the number density is that between the two regions there is a wall with a micrometer-range hole centered at the axis of symmetry. The particles can only pass from the inlet region to the outlet through this small hole.

The problem is that, within a reasonable computational time, i find it practically impossible to have enough dsmc particles in the outlet region (I'm aiming for 2-3 particles per cell at least)

I have thought about the following strategies:

1. Try to find a much more efficient parallelization procedure, I will try if scotch increases my efficiency enough. However I believe I would need at least a x20 increase of speed to be able to compute enough particles.

2. Time-averaging: I have not enough particles per cell, but if I let the program run enough time I do have enough particles that have been in that cell under a stable regime: is then time averaging (under convergence) matematically equivalent to averaging over the particles?

3. Divide the computation in two parts: first compute the fields up to the hole (skimmer), and use the results as boundary conditions for the part where I have a much smaller number of particles. Then change the nequivalentparticles in the second region.

4. The fields after the inlet seem reasonable and symmetric around the axis. However, due to the area-reduction of the inlet section near the line corresponding to radius=0, the number of inserted dsmc particles tends to 0 as well, reducing even further the number of dsmc particles that actually go across my hole. I'm uncertain if OpenFOAM is doing it in the right way or maybe the particle reduction close to the axis is too big.

The attached images show the inlet dsmc particles at different times of simulations, (em is the negative exponent ie 3_7_em6 = 3.7e-6)

5. Another option would be, considering that the mean free path after the hole is very small due to the minimum number of dsmc particles and the narrow velocity distribution, increase the size of the cells in the transversal direction. I have calculated that if the cell would be increased about 10 times, the number of dsmc particles should be enough. However I'm not sure if that cell would then not suffice the condition L<0.5*mean free path.

Have any of you worked in a similar problem? what do you do when you have regions with a much smaller number of dsmc particles? What strategy would you follow?

Any help will be very welcomed!

Thanks