The general purpose of this paper is to investigate some consequences
of the randomness of the velocities of interacting rigid particles fal
ling under gravity through viscous fluid at small Reynolds number. Ran
dom velocities often imply diffusive transport of the particles, but p
article diffusion of the conventional kind exists only when the length
characteristic of the diffusion process is small compared with the di
stance over which the particle concentration is effectively uniform. W
hen this condition is not satisfied, some alternative analytical descr
iption of the dispersion process is needed. Here we suppose that a dil
ute dispersion of sedimenting particles is bounded externally by pure
fluid and enquire about the rate at which particles make outward rando
m crossings of the (imaginary) boundary. If the particles are initiall
y distributed with uniform concentration within a spherical boundary,
we gain the convenience of approximately steady conditions with a velo
city distribution like that in a falling spherical drop of pure liquid
. However, randomness of the particle velocities causes some particles
to make an outward crossing of the spherical boundary and to be carri
ed round the boundary and thence downstream in a vertical 'tail'. This
is the nature of break-up of a falling cloud of particles. A numerica
l simulation of the motion of a number of interacting particles (maxim
um 320) assumed to act as Stokeslets confirms the validity of the abov
e picture of the way in which particles leak away from a spherical clu
ster of particles. A dimensionally correct empirical relation for the
rate at which particles are lost from the cluster involves a constant
which is indeed found to depend only weakly on the various parameters
occurring in the numerical simulation. According to this relation the
rate at which particles are lost from the blob is proportional to the
fall speed of an isolated particle and to the area of the blob boundar
y. Some photographs of a leaking tail of particles in figure 5 also pr
ovide support for the qualitative picture.