1. THE LYMAN-BREAK TECHNIQUE AND THE HDF: HOW DID THEY COME TO BE?

Exploiting the Lyman limit as a clear spectral feature to identify distant star-forming galaxies is hardly a new idea (e.g., Partridge & Peebles 1967). However, its implementation as an observational technique had to wait some 25 years or so, until detectors became sensitive enough (particularly in the U band) to make it practical with the available telescopes. Early observations that made use of deep CCD multi-band imaging to detect very high-redshift galaxies by their faintness in the U-band (see Figure 1) and constrain the abundance of galaxies at z > 3 were obtained by Guhathakurta et al. (1990) and Songaila et al. (1990). However, these works only set approximate upper limits to the surface density of these systems that turned out to be of rather limited use.

The first to exploit the Lyman-break technique to actually single out very high-redshift galaxies from the deep counts with relatively high efficiency were Steidel & Hamilton (1993). Their first observations were obtained in rather poor seeing conditions, were only of modest depth (by modern standards), and covered a limited sky area to produce a large sample of robust candidates. Significantly deeper and better quality data were obtained in 1994 by Giavalisco et al. (unpublished), exploiting the image quality available at the ESO New Technology Telescope; and by Steidel et al. (1995) at the William Herschel Telescope. Later in 1995, additional deep images were obtained using the 200-inch Hale telescope and the large field-of-view COSMIC camera at its prime focus. Although these data resulted in relatively large samples of robust z ~ 3 Lyman-break galaxy candidates (or U-band "dropouts" in the current parlance), they had to remain such for one more year. Due to the faintness of these candidates (typically 24), 4-m class telescopes are impractical to secure follow-up spectroscopy, and only redshifts of unusual candidates were known by the end of 1995 (Giavalisco et al. 1994). No definitive proof could be provided for the existence of a population of galaxies at significant redshifts without the power of the 10-m class telescopes. Meanwhile, the HDF survey was being conceived and designed.

Figure 1. a) Color-selection of star-forming galaxies at z ~ 3. The transmittance of the UnGR filters are plotted. The histogram shows the UV SED of an unreddened continuously star-forming galaxy at z = 3.15, with the flux in an arbitrary scale. b) Color-color plot from deep UnGR imaging in the SSA22 field obtained at the 200-inch Palomar telescope with the COSMIC camera. Magnitudes are in the AB scale. The selection window for Lyman-break galaxies is marked by the dashed line.

As we have seen, U-band dropouts were already relatively well known in 1995, with a sample that counted ~ 100 galaxy candidates at z ~ 3, some with HST morphologies as well. The consistency of such properties as luminosity distribution, surface density (Steidel et al. 1995), HST morphology (Giavalisco et al. 1994b; 1996a) and also limited spectroscopic information (Giavalisco et al. 1994a) seemed powerful arguments that suggested the existence of a new and totally unexplored population of galaxies whose redshifts were likely to be well in excess of z ~ 2. Thus, transporting to the HDF the same observational techniques adopted for the ground-based survey appeared a rather sensible choice. This would provide the HDF survey with color information and at the same time opened the possibility to directly study unprecedentedly distant galaxies at very faint fluxes. However, for this to be achieved, a significant amount of the available observing time had to be spent with the poorly efficient U passband (i.e., the F300W filter of the WFPC-2 collection). Although it was initially received as bizarre spending as much as half of the available exposure time imaging at wavelengths where the throughput of HST+WFPC-2 is very low, in the end considerations of practical nature (see Ferguson and Dickinson, this volume) made the decision of including the F300W filter in the HDF survey a rather easy one to take. Paradoxically, thanks to the low orbit of HST, the observing time that the Director of ST ScI allocated to the project could be effectively used imaging at wavelengths where the throughput is relatively high (i.e., F450W, F606W and F814W) and still get the precious F300W data essentially for free during the bright part of the orbit of the telescope. The HDF survey was made sensitive to the U-band dropouts.

In September 1995, we were able to follow up the NTT and Palomar samples of U-band dropouts with the Keck 10-meter telescope and LRIS spectrograph (Oke et al. 1995), confirming the existence of a massive population of star-forming galaxies at z ~ 3 (Steidel et al. 1996a). In December 1995, the HDF was successfully carried out, and by January 15, 1995 the first versions of the reduced images and of the source catalog were released to the community. During the same month, we obtained the first spectra of U-band dropouts selected from the HDF and confirmed their high redshifts (Steidel et al. 1996b). Other spectroscopic follow-ups of high-redshift candidates came shortly afterwards (Lowenthal et al. 1997).