The Lyman-break technique consists of a set of color criteria to identify star-forming galaxies at high-redshift through multi-band imaging across the 912 Å Lyman-continuum discontinuity. At high redshifts (z 2.5 for ground-based telescopes and z 2 for the HDF) the Lyman limit is shifted far enough into the optical window - into the U passband in particular - to be detectable with broad-band photometry. By placing filters on either side of the redshifted break (see Figure 1) one can identify high-redshift objects by their faintness in the U band and by an otherwise blue spectral energy distribution. Using redder passbands, the technique can be used to look for galaxies at even higher redshifts. For example, one can define similar criteria for B-band dropouts which would be galaxies candidates at redshifts 3.5 z 4.5 and so on.
More details can be found in Mark Dickinson's contribution in this volume, who describes the Lyman-break technique applied to the HDF. Here we will simply review the salient features of the ground-based survey. For this, we have adopted a custom photometric system, named UnGR, optimized for selecting LBGs with redshifts around z ~ 3 (Steidel & Hamilton 1993). An initial selection region in the [G - R, Un - G] plane where to expect high-redshift candidates was identified based on the expected colors of moderately unreddened star-forming galaxies computed using stellar population synthesis codes (Bruzual & Charlot 1998, in prep.), and including the effects of the opacity of interstellar gas and intervening absorption by HI (Steidel et al. 1995; see also Madau 1995). Our selection criteria were subsequently verified and refined after extensive spectroscopy with the Keck telescopes.
An object is considered a Lyman-break galaxy if its colors satisfy
with an additional requirement < 25.5 imposed to produce a reasonably complete sample that is suitable for spectroscopic follow-up (magnitudes are in the AB scale of Oke & Gunn, 1983). Figure 1b shows an example of the color diagram and selection window that we have used to identify the high-redshift candidates. At the time of this writing, the ground-based survey consists of more than 1,200 U-dropouts brighter than ~ 25.5, of which 418 have been spectroscopically confirmed in the range 2 < z < 4. Figure 2a shows the redshift distribution N(z) of the current spectroscopic sample. In comparison, the HDF sample is significantly smaller, including 30 candidates with V606 < 25.5, 22 of which have secured redshifts (see Dickinson's paper). However, it is considerably deeper, with 187 candidates at V606 < 27, a flux level where its fractional completeness is approximately equivalent to the ground-based sample. Unfortunately, the redshift distribution of this faint sample is empirically poorly constrained.
Figure 2. a) The redshift distribution function N(z) of 418 LBGs obtained with the Keck telescope and the LRIS spectrograph. The bin size is z = 0.2, and the median redshift is z = 3.01. The interval 2.4 z 3.6 contains 90% of the galaxies. b) The cosmic star-formation activity as a function of time in linear scale. Filled symbols are uncorrected data. Open symbols are data corrected for dust obscuration (see text). The triangle is the H point from Gallego et al. (1995). Squares are the CFRS points (Lilly et al. 1995). Pentagons are the photometric redshift from Connolly et al. (1997). Pentagons and circles are the HDF and UnGR LBGs points, respectively. H0 = 50 km-1 Mpc-1 and q0 = 0.5.
As it turns out, the Lyman-break selection is particularly efficient to single out high-redshift galaxies from the deep counts, and we have found that at least 75% of the objects meeting the criteria of Eqn. (1) are indeed high-redshift galaxies. The median value of the redshift distribution in Fig. 2a is = 3.01 and the standard deviation is z = 0.29, with 90% of the objects having redshifts in the range 2.4 z 3.6. About 5% of the objects meeting these criteria are stars, essentially all brighter than ~ 24. The remaining 20% of objects remain unidentified because of the low S/N of their spectra. However, it is important to note that these are still consistent with being at high-redshifts, either the same as the confirmed galaxies or slightly smaller (e.g., 1.8 z 2). Also, note that our success in obtaining a redshift has no obvious dependence on luminosity or color.