One of the most important developments of the last decade has been the systematic study of z ~ 3 galaxies carried out be Steidel and his co-workers (see e.g., Steidel et al. 1996). Using largely ground-based UGR photometry to find z ~ 3 candidates from U-band ``drop-outs'' Steidel and his collaborators have obtained redshifts with Keck and confirmed some 700 2.5 < z < 3.5 galaxies. This is a remarkable accomplishment, when one considers that only five 5 years ago the number of known galaxies at z > 2 was but a handful.
By extending the photometry to the red, with GRI images, Steidel et al. (1999) have obtained a sample of ``blue drop-outs'' (strictly G-band drop-outs, but usually referred to as ``blue'' as opposed to ``uv'' dropouts) that have already yielded a large number of z ~ 4 galaxies. The redshift-confirmed sample of 3.5 < 4 < 4.5 galaxies is already nearly 50 objects (Steidel et al. 1999).
Figure 3. The composite Keck LRIS spectrum of the Lowenthal z ~ 3 galaxies in the HDF. Note the weak Ly emission, the damped absorption at Ly, the ISM and stellar wind absorption lines, and the UV continuum slope.
When HST data is available, the z = 3-4 galaxies can be analyzed for structural characteristics and length scales. A study that showed the value of combining redshifts with HST imaging was that of Lowenthal et al. (1997) for a small sample of z ~ 3 galaxies in the HDF. The composite spectrum of the galaxies that were measured with Keck is shown in Figure 3. The spectral characteristics are very similar to low redshift starbursting systems (typical SFR in this HDF sample are around 10 M yr-1). The characteristic weak Ly, as well as the many absorption lines from the ISM and stellar winds can be seen in this composite spectrum. The galaxies that have been detected in this sample (and in the Steidel et al. survey) typically have luminosities L* or greater, and have quite small half-light radii (~ 0.3''). Such measurements allow us to compare these objects with low redshift galaxies. As can be seen in Figure 4 (from Lowenthal et al. 1997), they are typically more luminous than low redshift galaxies of comparable size, but it is instructive to note that the stellar population from such a young burst will dim by some five magnitudes, and so these objects could well be the precursors of local sub-L* galaxies (of the bulges?). However, much can, and likely will happen to these objects in the 10 Gyr between z ~ 3 and now and so it is not obvious what such objects will become in individual cases - though since most stars will end up residing in bulges, as was noted above from the results of Fukugita, Hogan and Peebles (1998), it is almost a no-brainer to note that these are likely, on average, to be the precursors of present day bulges.
Figure 4. The absolute magnitude vs half-light radius for a large sample of nearby galaxies of many morphological types, overplotted with the HDF z ~ 3 sample (from Lowenthal et al. 1997). The z ~ 3 galaxies are luminous, but quite compact.
One other conclusion that can be drawn from the composite spectrum of Figure 3 is that the typical z ~ 3 starburst is extincted. The slope of the UV continuum around 1500 Å can be used to estimate the level of extinction. It turns out, as might be expected, that the extinction correction is an important one, with E(B - V) typically being found to be ~ 0.3 (Steidel et al. 1999). The resulting correction for the star formation rate does depend sensitively, however, on the form of the dust extinction law that is used. For example, Dickinson (1998) shows that the ratio of the mean corrected SFR to the uncorrected SFR for z ~ 3 galaxies was a factor 7 (!) for a Calzetti reddening law (Calzetti, Kinney and Storchi-Bergmann 1994), but only a factor 2 for an SMC reddening law, with the Calzetti law giving much greater dispersion due to its rather grey behavior. Since the Calzetti reddening law was derived for (low redshift) starburst galaxies it might seem more appropriate to use it for the high redshift starbursts, but metallicities, and the (possibly) different environment may lead one to a different reddening law - but it will be difficult to verify what the different law should be.
The net effect of both the larger sample of galaxies at high redshift and the corrections for extinction has been to raise significantly the star formation rate at z > 2 where the ``drop-out'' sample have been the dominant source of constraints on the SFR vs redshift. With no extinction corrections the rates have increased with the larger samples so that there is only a modest factor of 2 increase from the SFR at z ~ 4 to the peak SFR at z ~ 1.5, in contrast to the original factor of nearly an order-of-magnitude from the HDF data alone (small number statistics?). Correcting for extinction, based on a Calzetti reddening law, leads to a further increase of the SFR at redshifts z > 2 to where the SFR(z) relation is essentially unchanged at higher redshifts, from redshift z ~ 4-5 to z ~ 1.5. The ``Madau plots'' incorporating these new data are shown in Figure 5 (from Steidel et al. 1999).
Figure 5. The recent derivation by Steidel et al. (1999) of the ``Madau plot'' showing the effect of using their much larger sample, and of the importance of the corrections for extinction. The HDF (lower symbols at z = 3 and z = 4) contained fewer high-redshift galaxies than the typical fields of Steidel et al., leading to an underestimate of the SFR at z ~ 3-4 (as in Figure 1).