Many N-body simulations have shown how galaxies disperse and cluster as the universe expands. We shall just examine some representative examples. The 4000-body simulations of Aarseth and his collaborators (see bibliography for references) were designed to see how different initial conditions and cosmological models, within reasonable ranges, alter the clustering of galaxies. Experiments whose initial velocities had a pure Hubble flow behaved much like those which also had peculiar Maxwellian velocities smaller than the expansion velocity. Gravitational graininess built up the self-consistent velocity field after about one initial Hubble expansion in either case.

Somewhat more important is the galaxies' mass distribution. In the computation this does not change with time, although small groups which form can mimic the distant effects of larger mass galaxies at later stages of clustering. Various simulations have shown that the more massive galaxies have a greater tendency to cluster. Their two-point correlation amplitude, for example, increases faster than for less massive galaxies. Linear analysis (Equation (24.1)) leads us to expect this, at least during the initial clustering when the velocity distribution does not depend strongly on mass. Experiments also show that many results, especially regarding spatial distribution, do not depend strongly on detailed properties of the mass spectrum, so long as it continuously spans a range of at least an order of magnitude.