With 418 galaxies, the spectroscopic sample of Lyman-break galaxies is a rich source of information on the physical properties of star-forming galaxies at high redshifts. In addition to the optical data, the current sample also includes 5 near-IR (K-band) spectra obtained with the UKIRT telescope and the CGS4 spectrograph by Pettini et al. (1998, in preparation). In most cases, the S/N of the "discovery spectra," i.e., of the optical spectra obtained to secure the redshift of the galaxies, is relatively modest. Furthermore, the near-IR spectra, being obtained with a 4-m class telescope, have limited depth. Nevertheless, the amount of information contained in the data is sufficient to discuss important properties of the galaxies, as we shall now review.
3.1. The rest-frame UV spectra
At z ~ 3 the optical spectra sample the rest-frame far-UV radiation, approximately between 1000 and 2000 Å, carrying information on massive stars and on the interstellar gas. Overall, the spectra are all qualitatively similar and also bear a rather close resemblance with those of local star-forming galaxies, consistent with a relatively small dispersion of properties of the powering source and of the physical state of the interstellar absorbing medium. However, a closer inspection reveals differences, both between the Lyman-break galaxies themselves and with the local systems, particularly in the kinematics of the interstellar absorption lines and in the Lyman-alpha properties.
Figure 3 shows two examples of spectra that illustrate the variety observed among the Lyman-break galaxies along with the UV spectrum of NGC 4214, a Wolfe-Rayet galaxy observed with HST and the GHRS by Leitherer et al. (1996). As mentioned above, in both cases the similarity between the spectra of the Lyman-break galaxies and the local galaxies is striking. In each case, the dominant characteristics of the far-UV spectra are (1) the flat continuum, whose spectral index is approximately f ~ 0 or slightly redder; (2) the weak or absent Ly emission, whose equivalent width is, in general, substantially smaller than the predictions of the radiation-bound case-B recombination theory; (3) the strong interstellar absorption lines due to low-ionization species of C, O, Si and Al; (4) the prominent high-ionization stellar lines of HeII, CIV, SiIV, NV.
Figure 3. Two examples of Keck spectra of Lyman-break galaxies plotted together with the spectrum of a star-forming knot in the Wolfe-Rayet galaxy NGC 4214 (Leitherer et al. 1996) placed at the same redshift. The flux zero point of the high-z galaxies is offset relative to the zero point for the scaled spectrum of NGC 4214 (indicated with the lower dotted line). Some of the identified spectral features (both stellar and interstellar) have been labeled for comparison. These two cases give an idea of the variety of spectra observed among the LLG population. In the top panel Ly is in absorption together with strong interstellar absorption lines. The CIV P-Cygni feature is also visible. The galaxy in bottom panel has Ly in emission, albeit with rest-frame equivalent width Ew = 8 Å and less strong but broader interstellar features. There is a strong absorption feature at the wavelength of the CIV, but no emission.
The two spectra, do show differences, however. The first has no detected Ly emission and has a broad absorption feature at its wavelength together with deep and relatively narrow interstellar absorption lines. The second, on the contrary, is characterized by broader and shallower interstellar lines and by the Ly emission line, whose equivalent width, however, is modest, 10 Å in the rest frame.
The S/N of these "discovery spectra" is insufficient to detect other stellar features. However, it is very interesting that whenever we could follow-up at higher S/N and dispersion some bright Lyman-break galaxy, we have always detected weak stellar features such as the photospheric lines of SV1502 and OIV1343, and SiIII1417, characteristic of O stars.
The observed UV continua result from the integrated light of massive O and early-B stars, A galaxy at z = 3 with ~ 24.5 has a far-UV luminosity density ~ 1 x 1041 (2.5 x 1041) h-250 erg s-1 Å-1 at ~ 1500 Å, with q0 = 0.5 (0.1), respectively (we will follow this convention throughout this paper, unless explicitly indicated). This is about three order of magnitudes the luminosity density of the brightest local star clusters studied with HST (see, e.g., Leitherer et al. 1996). Given the short life-time of O and B stars, it is relatively straightforward to convert the observed absolute UV luminosities into instantaneous rates of star formation once an IMF has been assumed. Adopting a continuous star-formation activity with a Salpeter IMP from 0.1 to 100 M, a galaxy at z = 3 with ~ 24.5 forms stars at the rate of 12 (29) h-250 M yr-1 if q0 = 0.5 (0.1).
These are almost certainly lower limits, since dust extinction and a shorter burst age result in larger rates (a burst with an age of 107 yr implies a SFR 1.7x larger than the values above).
Calzetti et al. (1994) have shown that the far-UV spectral index is a robust indicator of dust reddening because it depends weekly on the exact shape of the IMF and the age of the burst. Both models and empirical templates of starburst galaxies shows that the spectral index expected in absence of dust obscuration is significantly bluer than what is typically observed in the Lyman-break galaxies, as is schematically illustrated in Figure 4. Although such a relatively red spectrum can be produced by stellar populations with a significant deficiency of massive stars (due either to a shallower IMF or an evolved burst), this possibility seems unlikely since even in the relatively red spectrum of Figure 4 itself we were able to directly see the photospheric lines of O stars. Thus, the presence of dust obscuration (as opposed to a deficiency of very massive stars) seems the explanation for the observed UV continua.
Figure 4. Spectral indexes of models of burst and continuous star-formation rates and of one observed Lyman-break galaxy. This last is characterized by the presence of photospheric lines of O stars, suggesting that dust obscuration, as opposed to lack of massive stars, is responsible for its reddening. The flux is in an arbitrary scale. The models are from Leitherer et al. (1995).
A more detailed description of the amount of dust obscuration and the correcting factors to recover the intrinsic star-formation rates are discussed in Dickinson's paper. Such corrections are in the range 3x - 7x, depending on the exact shape of the extinction curve, and although of modest amount, they are nevertheless of sufficient amplitude to affect our conclusions about the nature and possible evolutionary history of both the single Lyman-break galaxies and of the overall population in relation to the general problem of galaxy formation, as we will discuss later.
The Ly emission of the LBGs is very often weak or absent despite the rather high star-formation rates implied by the UV continua, in agreement with the generally null results of search for "primeval galaxies" based on the presence of this emission line (e.g., Thompson et al. 1995). Such a relatively weak Ly is very likely the result of resonant scattering in presence of dust in an outflowing interstellar medium (Charlot & Fall 1993). The Ly, when detected, is generally redshifted respect to the interstellar absorption lines by several hundred km s-1 and its profile is often asymmetric. The redshifts of both the Ly and of the interstellar lines are not likely to reflect the systemic redshift of the galaxies. In the few cases of high S/N spectra where photospheric lines from O stars have been observed (SV1502 and OIV1343) and also when the optical nebular lines ([OII], H and [OIII]) have been observed with near-IR spectra, the redshifts of these features have been found to exceed that of the interstellar lines by up to ~ 500 km s-1. Or, in other words, if the bulk of the stars define the systemic redshift of the galaxies the interstellar lines are blueshifted respect to them. Taken together, the phenomenology of the Ly the interstellar, stellar and nebular lines is consistent with the presence of large-scale outflows in the interstellar medium of the Lyman-break galaxies, very likely the consequence of the injection of kinetic energy by stellar winds and supernova events (Pettini et al. 1997). Large-scale motions of expanding shells are very likely the main reason for the large strengths of the interstellar lines, which typically have equivalent widths of a few Angstroms. Since these lines are saturated, these equivalent widths are much more sensitive to the velocity fields in the gas than to its metallicity.