2.2. How much dark matter?
An important recent development is that DM can now be constrained to a value around 0.25 by several independent lines of evidence:
(i) One of the most ingenious and convincing arguments comes from noting that baryonic matter in clusters - in galaxies, and in intracluster gas - amounts to 0.15 - 0.2 of the inferred virial mass (White et al. 1993). If clusters were a fair sample of the universe, this would then be essentially the same as the cosmic ratio of baryonic to total mass. Such an argument could not be applied to an individual galaxy, because baryons segregate towards the centre. However, there is no such segregation on the much larger scale of clusters: only a small correction is necessary to allow for baryons expelled during the cluster formation process.
(ii) Very distant galaxies appear distorted, owing to gravitational lensing by intervening galaxies and clusters. Detailed modelling of the mass-distributions needed to cause the observed distortions yields a similar estimate. This is a straight measurement of DM which (unlike (i)) does not involve assumptions about b, though it does depend on having an accurate measure of the clustering amplitude.
(iii) Another argument is based on the way density contrasts grow during the cosmic expansion: in a low density universe, the expansion kinetic energy overwhelms gravity, and the growth of structure saturates at recent epochs. The existence of conspicuous clusters of galaxies with redshifts as large as z = 1 is hard to reconcile with the rapid recent growth of structure that would be expected if DM were unity. More generally, numerical simulations based on the cold dark matter (CDM) model model are a better fit to the present-day structure for this value of DM (partly because the initial fluctuation spectrum has too little long-wavelength power if DM is unity).
Other methods will soon offer independent estimates. For instance, DM can be estimated from the deviations from the Hubble flow induced by large-scale irregularities in the mass distribution on supercluster scales.