4.4. On the nature of the dark matter in clusters
The early papers by Zwicky [512, 513] and Smith  did not convince the astronomical community that dark matter existed in clusters. Most astronomers favoured the alternative hypothesis, cluster instability, until the 21cm measurements proved the galaxy rotation curves to be non-Keplerian (see Section 4.2 and the excellent reviews of Sidney van den Bergh [481, 482]). However, many astronomers took the dark matter hypothesis very seriously and tried to elucidate its nature.
When Zwicky  discovered the missing matter problem, his first reaction was to question the validity of Newton's gravitational law. Later he turned his attention to possible forms of dark matter, that could also provide IC obscuration and thus explain the non-uniform sky distribution of clusters [520, 522, 524]. During the 50's he could not find much observational evidence for a significant quantity of IC matter , so he  again suggested abandoning the general theory of relativity.
In 1956 Heeschen , motivated by Stone 's theoretical argument, searched for and detected HI emission from Coma. The detected emission implied a mass of 5 × 1013 solar masses. Heeschen's detection was however shown to be spurious by Muller , three years later. From a theoretical point of view, Limber  noted that clusters are likely to contain IC gas, because the galaxy formation process is unlikely to be 100 % efficient, and because galaxy-galaxy collisions sweep gas out of galaxies. He pointed out that if this IC gas remained undetected at 21 cm, it could be hot and ionized. Extensive searches for intergalactic material by Zwicky & Humason  did not prove very successful. In 1961 the total amount of IC cold gas was constrained by Penzias  to be less than a tenth of the virial mass in the Pegasus I cluster. Penzias remarked that the integrated 21 cm emission from individual cluster galaxies could well account for the total cluster emission.
Dark matter was also searched for in the form of diffuse optical luminosity [521, 490, 296, 457] and dwarf galaxies [227, 366, 220, 221, 385], but these components were found to account for less than half the total cluster luminosity (see Section 4.3).
In 1960 de Vaucouleurs  summarized the observational situation by noting that the missing mass could neither consist of cold HI, nor of dust, nor of diffuse (optical) luminosity. The total mass of all these components only makes a small fraction of the total cluster mass, and therefore ``a large number of essentially dark bodies must also be assumed.'' If the existing observations could not establish the nature of dark matter, they were anyway not in conflict with the hypothesis that only a small fraction of the matter in the Universe is in bright galaxies (Layzer ).
The idea that galaxies have dark halos gained a hold upon the astronomical community in a very short time, from 1969 to 1974, through the work of Arp & Bertola , de Vaucouleurs , Vorontsov-Velyaminov , Freeman , Lewis , Ostriker & Peebles , Einasto et al.  and Ostriker et al.  (see Section 4.2). Rood  had already demonstrated in 1965 that not all cluster dark matter can be attached to galaxies, or relaxation processes could produce much more energy equipartition (and hence, luminosity segregation) than observed (Rood's early finding was later confirmed by White ). Moreover, Lecar  pointed out that galaxies in clusters should loose their halos via tidal stripping.
Figure 28. The logarithm of the mass-to-light ratios for rich clusters vs. the logarithm of their velocity dispersions. From Rood (1974b).
The idea of a scale-dependent mass-to-light ratio took a step forward through the works of Zwicky & Humason , Karachentsev , Rood et al. , Rood  - see Fig. 28 - Ostriker et al. , Einasto et al. , Bahcall  and Davis et al. . The mass-to-light ratio seemed to increase from galaxies to groups and to rich clusters. This evidence seemed to indicate that the dark matter does not follow the distribution of bright galaxies. Dressler  however noted that including the IC gas mass would reduce the mass discrepancy in galaxy clusters, and destroy the evidence for a scale dependence of the mass-to-light ratio. BAHCALL (these proceedings) has recently shown that the mass-to-blue light ratio increases with scale up to the size of galaxy clusters, but not beyond. The trend can be explained as an age effect (galaxies in high density environments form earlier, so that their blue luminosity fades earlier).
The apparently different distribution of dark matter and bright galaxies strengthened the idea that the dark matter consists of diffuse gas. A significant amount of IC HI had been ruled out by observations. Astronomers then started looking for ionized gas. In 1967 Woolf  put the first constraints on diffuse ionized gas, by looking at H and H emission from clusters. He concluded that if the cluster dark matter was in the form of ionized gas, the temperature of this gas had to be below 106 K. Three years later, Turnrose & Rood  confirmed Woolf's estimate, using X-ray data. When diffuse X-ray cluster emission was detected (Meekins et al. , Gursky et al. ) it was immediately clear that the IC gas could not account for all the cluster missing mass.
It was at this point that astronomers really started to grope in the dark. In 1969, van den Bergh  considered massive collapsed objects (of 108-1012 M each) as dark matter candidates, but ruled them out on the basis of the limited tidal distortion of galaxies in Virgo. Peebles  suggested frozen HI snowballs as dark matter candidates, a possibility never really ruled out (see, e.g., Wright et al. ). Another form of baryonic dark matter was proposed by Tarter & Silk  (M8 dwarf stars), but they also frankly remarked that ``nothing better'' could be said on this topic than had already said thirty years earlier by Zwicky. A scaled-down version of Tarter & Silk's dark-matter candidates were Napier & Guthrie 's 10-2 M ``black dwarfs''. Tarter & Silk's dark matter candidates were later suggested by Sarazin & O'Connell  to be the end-products of the cooling flows onto cD galaxies (see Section 5.4). In 1981 Gott  proposed a gravitational lensing experiment to detect an hypothetical huge population of low-mass stars in galaxy halos, thus anticipating the recent AGAPE , EROS  and MACHO  projects.
Baryons as candidate for dark matter are however ruled out by the theory of primordial nucleosynthesis (see, e.g., Cavaliere et al. ), and therefore more exotic dark-matter candidates were proposed. Here is a short list of them: a variable G (Lewis ); MOdified Newtonian Dynamics (Milgrom ); vacuum strings (Vilenkin ); magnetic monopoles (Hoyle ); heavy neutrinos (Cowsik & McClelland , Szalay & Marx , Doroshkevich et al. ) - eventually unstable (Sciama ); gravitinos (Pagels & Primack ), axions (Stecker & Shafi ), and cold dark matter in general (Bond et al. ).
Recent observations of the cosmic microwave background (de Bernardis et al. ) have added considerable constraints on the nature of the missing mass, which is now thought to consist of a mixture of cold dark matter and dark energy (in the form of a cosmological constant or quintessence, see, e.g., Bahcall et al. ). It is nevertheless wise to close this section with a statement of Alan Dressler :
``The answer to the mass discrepancy problem awaits more data and more inspiration, not necessarily in that order.''