The first evidence that the Universe may contain dark matter (undetected through electromagnetic emission or absorption) can be traced back to 1933, when Zwicky studied the velocities of galaxies in the Coma Cluster and concluded that the total mass required to hold the Cluster together is about 400 times larger than the luminous mass in stars. In 1937 he reinforced this analysis and noted that galaxies associated with such large amounts of mass should be detectable as gravitational lenses producing multiple images of background galaxies. These conclusions were substantially correct, but remarkably it took more than 40 years for the existence of dark matter to be generally accepted. The tide turned in the mid-1970s with papers by Ostriker et al. (1974) and Einasto et al. (1974) extending Zwicky’s analysis and noting that massive halos are required around our Milky Way and other nearby galaxies in order to explain the motions of their satellites. These arguments were supported by continually improving 21 cm and optical measurements of spiral galaxy rotation curves which showed no sign of the fall-off at large radius expected if the visible stars and gas were the only mass in the system (Roberts & Rots, 1973; Rubin et al., 1978, 1980). During the same period, numerous suggestions were made regarding the possible nature of this dark matter component, ranging from baryonic objects such as brown dwarfs, white dwarfs and black holes (e.g. White & Rees, 1978; Carr et al., 1984), to more exotic, elementary particles such as massive neutrinos (Gershtein & Zel’dovich, 1966; Cowsik & McClelland, 1972).
The suggestion that neutrinos might be the unseen mass was partly motivated by particle physics. In the 1960s and 1970s, it was noticed that grand unified theories (GUTs) permit the existence of massive neutrinos, and various attempts to measure neutrino masses in laboratory experiments were initiated. In the late 1970s, Lyubimov et al. (1980) and Reines et al. (1980) announced the detection of a mass for the electron neutrino at a level of cosmological interest (about 30 eV). Although the results were not conclusive, they caused a surge in studies investigating neutrinos as dark matter candidates (e.g. Bond et al., 1980; Sato & Takahara, 1980; Schramm & Steigman, 1981; Klinkhamer & Norman, 1981), and structure formation in a neutrino-dominated universe was soon worked out in detail. Since neutrinos decouple from other matter and radiation fields while still relativistic, their abundance is very similar to that of CMB photons. Thus, they must have become non-relativistic at the time the Universe became matterdominated, implying thermal motions sufficient to smooth out all structure on scales smaller than a few tens of Mpc. The first nonlinear structures are then Zel’dovich pancakes of this scale, which must fragment to make smaller structures such as galaxies. Such a picture conflicts directly with observation, however. An argument by Tremaine & Gunn (1979), based on the Pauli exclusion principle, showed that individual galaxy halos could not be made of neutrinos with masses as small as 30 eV, and simulations of structure formation in neutrino-dominated universes by White et al. (1984) demonstrated that they could not produce galaxies without at the same time producing much stronger galaxy clustering than is observed. Together with the failure to confirm the claimed neutrino mass measurements, these problems caused a precipitous decline in interest in neutrino dark matter by the end of the 1980s.
In the early 1980s, alternative models were suggested, in which dark matter is a different kind of weakly interacting massive particle. There were several motivations for this. The amount of baryonic matter allowed by cosmic nucleosynthesis calculations is far too little to provide the flat universe preferred by inflationary models, suggesting that non-baryonic dark matter may be present. In addition, strengthening upper limits on temperature anisotropies in the CMB made it increasingly difficult to construct self-consistent, purely baryonic models for structure formation; there is simply not enough time between the recombination epoch and the present day to grow the structures we see in the nearby Universe from those present in the high-redshift photon–baryon fluid. Finally, by the early 1980s, particle physics models based on the idea of supersymmetry had provided a plethora of dark matter candidates, such as neutralinos, photinos and gravitinos, that could dominate the mass density of the Universe. Because of their much larger mass, such particles would initially have much smaller velocities than a 30 eV neutrino, and so they were generically referred to as warm or cold dark matter (WDM or CDM, the former corresponding to a particle mass of order 1 keV, the latter to much more massive particles) in contrast to neutrinolike hot dark matter (HDM). The shortcomings of HDM motivated consideration of a variety of such scenarios (e.g. Peebles, 1982; Blumenthal et al., 1982; Bond et al., 1982; Bond & Szalay, 1983).
Lower thermal velocities result in the survival of fluctuations of galactic scale (for WDM and CDM) or below (for CDM). The particles decouple from the radiation field long before recombination, so perturbations in their density can grow at early times to be substantially larger than the fluctuations visible in the CMB. After the baryons decouple from the radiation, they quickly fall in these dark matter potential wells, causing structure formation to occur sufficiently fast to be consistent with observed structure in today’s Universe. M. Davis et al. (1985) used simulations of the CDM model to show that it could provide a good match to the observed clustering of galaxies provided either the mass density of dark matter is well below the critical value, or (their preferred model) that galaxies are biased tracers of the CDM density field, as expected if they form at the centers of the deepest dark matter potential wells (e.g. Kaiser, 1984). By the mid-1980s, the ‘standard’ CDM model, in which dark matter provides the critical density, Hubble’s constant has a value ∼ 50 km s−1Mpc−1, and the initial density field was Gaussian with a Harrison–Zel’dovich spectrum, had established itself as the ‘best bet’ model for structure formation .
In the early 1990s, measurements of galaxy clustering, notably from the APM galaxy survey (Maddox et al., 1990a; Efstathiou et al., 1990), showed that the standard CDM model predicts less clustering on large scales than is observed. Several alternatives were proposed to remedy this. One was a mixed dark matter (MDM) model, in which the universe is flat, with ∼ 30% of the cosmic mass density in HDM and ∼ 70% in CDM and baryons. Another flat model assumed all dark matter to be CDM, but adopted an enhanced radiation background in relativistic neutrinos (
). A third possibility was an open model, in which today’s Universe is dominated by CDM and baryons, but has only about 30% of the critical density (OCDM). A final model assumed the same amounts of CDM and baryons as OCDM but added a cosmological constant in order to make the universe flat (
) .
Although all these models match observed galaxy clustering on large scales, it was soon realized that galaxy formation occurs too late in the MDM and τCDM models, and that the open model has problems in matching the perturbation amplitudes measured by COBE. ΛCDM then became the default ‘concordance’ model, although it was not generally accepted until Garnavich et al. (1998) and Perlmutter et al. (1999) used the distance–redshift relation of Type Ia supernovae to show that the cosmic expansion is accelerating, and measurements of small-scale CMB fluctuations showed that our Universe is flat (de Bernardis et al., 2000). It seems that the presentday Universe is dominated by a dark energy component with properties very similar to those of Einstein’s cosmological constant.
At the beginning of this century, a number of ground-based and balloon-borne experiments measured CMB anisotropies, notably Boomerang (de Bernardis et al., 2000), MAXIMA (Hanany et al., 2000), DASI (Halverson et al., 2002) and CBI (Sievers et al., 2003). They successfully detected features, known as acoustic peaks, in the CMB power spectrum, and showed their wavelengths and amplitudes to be in perfect agreement with expectations for a ΛCDM cosmology. In 2003, the first year data from WMAP not only confirmed these results, but also allowed much more precise determinations of cosmological parameters. The values obtained were in remarkably good agreement with independent measurements; the baryon density matched that estimated from cosmic nucleosynthesis, the Hubble constant matched that found by direct measurement, the dark-energy density matched that inferred from Type Ia supernovae, and the implied largescale clustering in today’s Universe matched that measured using large galaxy surveys and weak gravitational lensing (see Spergel et al., 2003, and references therein). Consequently, the model has now established itself firmly as the standard paradigm for structure formation. With further data from WMAP and from other sources, the parameters of this new paradigm are now well constrained (Spergel et al., 2007; Komatsu et al., 2009).
