We now turn to the question of what the radio waveband has to say about the properties of galaxies seen at high redshifts. One unique capability of radio astronomy for cosmology is the detection of neutral hydrogen via the 21 cm line. This tends to receive most attention at low redshifts via the Tully-Fisher relation and the studies of the distance scale and peculiar velocities. However, it also gives a unique way of detecting neutral gas at high redshift - even beyond the limit of z 5 where quasar absorption-line studies can probe. Particularly motivated by early `pancake' theories of galaxy formation in which purely baryonic models give a supereluster-scale coherence length to the mass distribution, there have been a number of attempts over the years to use low-frequency observations to detect neutral hydrogen at high redshifts (e.g. Davies, Pedlar & Mirabel 1978 [z = 3.3 & 4.9]; Bebbington 1986 [z = 8.4]; Uson, Bagri & Cornwell 1991 [z = 3.3]; Wieringa, de Bruyn & Katgert 1992 [z = 3.3]). These are sensitive only to rather large structures: for a Gaussian velocity dispersion v, the expected flux density is
where D is comoving distance divided by c / H0 - e.g. D = 2(1 - [1 + z]-1/2) in an = 1 model. Since sensitivities of a few mJy are typically attained, the experiment is sensitive to masses in the range 1014 - 1015 M.
Most such experiments have yielded only upper limits, but the VLA experiment of Uson, Bagri & Cornwell (1991) claimed the detection of a resolved object with a peak flux density of 10 mJy. The inferred parameters of their object were
This experiment caused much debate, particularly the authors' claim that this was an example of a Zeldovich pancake. The characteristics of the emission are certainly hard to understand in any other way. The gas mass and size of object, together with the effective volume of space surveyed, are about right for a rich cluster of galaxies. However, in addition to the minute velocity dispersion, one would also not expect to find intracluster gas in a neutral state. In hierarchical models, it is continually shock heated by new infalling clumps of mass as structure grows. The only neutral gas would be associated with individual galaxies, producing much less massive neutral condensations (e.g. Subrahmanian Padmanabhan 1993). The only possibility might be a group of unusually neutral-rich galaxies resembling the damped Lyman- absorption systems seen in quasar spectra; in this context, it is worth noting that Wolfe (1993) has shown these to lie in regions of high density (at least in terms of cross-correlation with weaker Lyman- emitters). In any case, it would still be necessary to appeal to the coincidence of seeing a cluster close to its turn-round time to explain why the velocity dispersion is so small (and even this does not solve things completely, since there will be a dispersion associated with substructure).
Only in models with an initial coherence length does the gas have time to cool and regain its neutrality following heating at the initial collapse of the cluster. Without attempting to turn history back to a time before dark matter, perhaps the least radical modification would involve warm dark matter with a coherence length of a few Mpc. This would in any case lead to the usual `top-down' chain of events for galaxy formation. Since we believe that objects of cluster mass in fact mainly formed relatively recently (Lacey & Cole 1993; see also the contribution to this volume by S. White), this would have important implications for the ages of galaxies. For this reason, it is vital that the Uson et al. object be either confirmed or shown not to exist. Van der Kruit (private communication) suggests that the Westerbork group have indeed failed to detect it, which may cause some relief to those distressed by the above discussion. Whatever the eventual outcome, such observations will continue with increasing sensitivity and will be capable of setting interesting constraints on conditions at high redshift.