A ratio between the largest and the smallest particles

Hi Glaubiger,

I usually use a ratio Rmax/Rmin of either 4 or 5 to obtain a reasonable random packing of particles and moderate model sizes (in terms of number of particles). Generally models with Rmax/Rmin = 10 or more will be very large and consequently require a lot of CPUs to execute in a reasonable time.

Regarding larger ratios, the main constraints stem from the neighbour search algorithm. The grid spacing for the neighbour search needs to be larger than the diameter of the largest particle. I usually use a grid spacing of 2.5*Rmax. When the ratio Rmax/Rmin gets large (>10), the number of particles in each grid cell also becomes quite large. In such cases, neighbour search will become very slow. Also, the number of timesteps between neighbour table updates will be lower because the verlet distance should be less than Rmin. Generally speaking simulations with Rmax/Rmin > 20 are too slow to execute in a practical amount of time.

There really isn't any restrictions regarding ratios Rmax/Rmin < 3 except that as this ratio gets close to Rmax/Rmin = 1, you will often find clusters of particles that arrange themselves in a near-regular lattice. This is often undesirable unless you are trying to compare the simulations with experiments that also have a very narrow range of particle sizes.

Cheers,

Dion