1. INTRODUCTION

QSO absorption systems with 21 cm lines have attracted attention because they exhibit physical conditions found in the HI disks of spiral galaxies (see Briggs 1988). The resemblance to HI disks is not surprising. The large HI column densities, N(HI) 10²¹ cm^-2, low velocity dispersions, 10 km s^-1, and low spin temperatures, T_S 300 °K, typical of HI disks lead to 21 cm optical depths, ₂₁ 1. Moreover, the fact that the signal-to-noise ratio of a resolved 21 cm feature with velocity width, v, is proportional to N(HI) / (T_S x sqrt[v]) indicates that HI layers which are cold and quiescent are the easiest to detect. Consequently, 21 cm observations provide a limited field of view that focuses on absorbers with the essential properties of HI disks. In order to decide whether or not these really are the disks of high-redshift galaxies, we need a broader perspective, one that reveals how the 21 cm systems fit into the wider context of the universe at large redshifts.

For these reasons we began a survey for a large, unbiased sample of absorbers with 21 cm properties. The survey was carried out at optical rather than radio wavelengths mainly because of the limited frequency response of the most sensitive line feeds at Arecibo. At present Arecibo is the only radio antenna capable of detecting a statistically meaningful sample of HI layers with ₂₁ 0.01, the threshold set by the detection of some very weak 21 cm lines at large redshifts (see Briggs 1988). The problem is that the phase-correcting line feeds available at Arecibo provide a 21 cm redshift path given by z₂₁ = 0.1 - 0.2. On the other hand, z_opt 1 for prominent resonance lines detected at optical wavelengths. Thus the optical surveys should be more efficient, provided a suitable signature of the 21 cm systems is found.

The ideal signature is an absorption feature whose strength is sensitive to N(HI), but insensitive to v; i.e., a feature that stands out despite the low velocity dispersion expected in HI disks. The advantage of metal line doublets such as MgII is that they are easy to recognize redward of Ly- emission. The disadvantage with this, and other doublets arising in abundant metals, is that the lines have low equivalent widths, W(MgII) 0.3 Å, because they are saturated transitions that form in gas with low velocity dispersion. Moreover, it is probable that most MgII lines do not form in HI gas, but rather arise in ionized gas which need not be related to material in HI disks (Wolfe 1986a). On the other hand, Ly- lines that form in HI disks will be very strong. The combination of large N(HI) and low velocity dispersion assures that Ly- will be broadened by radiation damping. In that case W, the rest-frame equivalent width, is uniquely determined by N(HI) according to W(Ly-) = 7.3 x [N(HI) / 10²⁰ cm^-2] Å. For the N(HI) that typify the 21 cm systems, W(Ly-) 10 Å. Thus while the absorber is hardly noticeable in the metal lines, it stands out amongst the `confusion noise' of the Ly- forest. The appearance of damped Ly-, 30 times stronger than associated low-ion transitions, is a unique spectroscopic imprint not found in any other type of absorption-line cloud.