Processor decomposition "cake slices"

Hi everyone,

I am currently trying to simulate a granular cylinder packing (picture: simulation example) with 200k particles of r = 100 - 200 um. The particles are inserted on top of the cylinder and settle inside. Unfortunately, I can't find a good way of mapping out my processors.
Right now if I just use 4 processors (picture: 2 2 1), the particles are equally distributed, but each processor has 50k particles. In order to use the full potential of my system (2 quad-core nodes), I used 8 processors (picture: * * 1 grid twolevel 4 * * 1) which should result in 25k per processor. The forum suggests: 25k/proc is a good amount.
Unfortunately, because of the orthogonal grid, the particles are not equally distributed and there is no improvement in overall simulation time when using 8 instead of 4 processors.

Is it possible to assign the processors to cylinder slices (picture: processors ideal), so that the particles are equally distributed?

I tried out most of the commands in the documentation, but I always get an orthogonal grid. I don't have the premium version of Liggghts, so that I can't assign my processors dynamically. Of course 8 processors reduce pair time and increase comm time, but the issue seems to be the unequal distribution. A decomposition in z-direction showed no improvement.

Thank you for your help,
Michael