4.3. Emission-Line Searches for Distant Galaxies
The stellar population synthesis models of Charlot & Fall (1991) predict rest-frame Ly equivalent widths of 50-200 Å for young, dust-free galaxies. An unattenuated Ly equivalent width of 200 Å corresponds to 5% of the bolometric luminosity of the protogalaxy being emitted as Ly photons. For a constant star formation history, the Ly luminosity and equivalent width are only somewhat dependent on the star formation rate and are greatest at times less than 10 Myr after the onset of the burst. Although this line emission is quite vulnerable to extinction from dust, little dust might be expected in the first phase of star formation. Indeed, Thommes et al. (1998) suggest requiring strong Ly emission as a necessary criterion for primeval galaxies. Alternatively, dust production can proceed remarkably quickly in supernovae remnants, implying that most "primordial" galaxies may contain substantial dust by the time we observe them. We note that recent UV studies of low-redshift Ly suggest that the kinematics of neutral gas may be the more important attenuator of line emission (Kunth 1998a).
Several programs are currently underway to search for high-redshift primeval galaxies through deep narrowband imaging. The previous generation of surveys failed to detect any field Ly-emitting protogalaxy candidates (Thompson & Djorgovski 1995; Pritchet 1994). One of the new programs is the Calar Alto Deep Image Survey (CADIS; Thommes et al. 1998) which uses a Fabry-Pérot interferometer to search for emission lines in three windows of low night sky emission corresponding to Ly at z = 4.75, 5.75, and 6.55. The survey will eventually sample 0.3 deg2 of sky to a flux density limit of Slim(5 ) 5 × 10-17 ergs cm-2 s-1. The technique employs an arsenal of narrowband "veto" filters to distinguish foreground emission-line galaxies from distant Ly emitters. Medium-band filters are also used to discriminate foreground objects from Ly-emitting protogalaxy candidates at high redshift on the basis of spectral energy distributions. Six early candidates were reported in Thommes et al. (1998). However, follow-up spectroscopy with the Keck II telescope has not confirmed a Ly interpretation for any of these sources; Thommes (1999) present a revised candidate surface density of 0.2 Ly emitters arcmin-2 (unit-z)-1 at z = 5.75 to the relatively bright CADIS survey flux-density limit.
Another program to identify strong Ly-emitting star-forming galaxies recently has been started at the Keck II telescope (Cowie & Hu 1998; Hu, Cowie, & McMahon 1998). Using a combination of narrowband interference filter imaging and broadband imaging, they discriminate high-redshift Ly emitters on the basis of high equivalent width (Wobs > 77 Å) and broadband color. The survey probes to a flux density limit of Slim(5 ) 1.5 × 10-17 ergs cm-2 s-1. Preliminary results, covering 46 arcmin2 with a 5390/77 Å filter, suggest a surface density of 3 Ly emitters arcmin-2 (unit-z)-1 at z ~ 3.4 (Cowie & Hu 1998). The Hawaii group has narrowband filters tuned to gaps in the telluric OH emission, corresponding to redshifts z = 3.4, 4.5, 5.8, and 6.5.
Serendipitous searches on deep slit spectra, as discussed in the following subsection (Section 4.4), are also sensitive to line emission. A technique combining narrowband filters and spectroscopy is discussed in Section 4.5.
Of course, identification of a strong emission line alone does not necessarily imply the detection of high-redshift Ly. Arguments based upon the equivalent width of the line, lack of other emission lines, the line profile, and associated continuum decrements are typically used to determine the line identity (e.g., Stern et al. 1999a). However, selecting objects on the basis of strong emission samples a different galaxy population from the traditional magnitude-limited surveys: emission-line surveys are much more sensitive to active galaxies and objects undergoing massive bursts of star formation. Distinguishing the redshift and source of line emission is challenging; comparison to field surveys selected on the basis of continuum magnitude is perhaps inappropriate.
For example, Stern et al. (1999a) recently reported the serendipitous detection of an emission line at 9185 Å with an observed frame equivalent width greater than 1225 Å (95% confidence limit). The spectral atlas of nearby galaxies by Kennicutt (1992) shows that the rest-frame equivalent width of the H + [N II] complex rarely exceeds 200 Å, that of [O III] 5007 rarely exceeds 100 Å, H rarely exceeds 30 Å, and [O II] 3727 (hereafter [O II]) rarely exceeds 100 Å. Field surveys of moderate-redshift, star-forming galaxies substantiate that [O II] rarely has a rest-frame equivalent width exceeding 100 Å (e.g., Songaila et al. 1994; Guzmàn et al. 1997; Hammer et al. 1997; Hogg et al. 1998). Preliminary analysis would strongly suggest that this source was a Ly emitter at z = 6.56, for which the implied rest-frame Ly equivalent width would be consistent with confirmed sources at z > 5 (Dey et al. 1998; Weymann et al. 1998). However, a long-exposure Keck/LRIS spectrogram revealed a source 2".7 away with strong [O II] emission offset by only 7 Å spectrally, persuasively arguing that this is an unusual, likely active, [O II]-emitting system at z = 1.46.
Infrared programs, similar to these optical programs, have begun to yield some candidates, most likely at intermediate redshift thus far. Malkan, Teplitz, & McLean (1995) and Teplitz, Malkan, & McLean (1998), using the NIRC camera on the Keck I telescope, detect H (but possibly [O II] or [O III]) emission using the 2.16 µm narrowband CO filter. These initial searches were targeted; i.e., fields were selected to search for spatially correlated emission-line galaxies around known quasars or damped quasar absorption systems.
McCarthy et al. (1999) report on blank-sky grism searches obtained during parallel time with the NICMOS camera on HST. The observations sample 1.1-1.9 µm and are unaffected by atmospheric water absorption bands. They find a surface density of single emission-line galaxies (most likely H) of 0.5 galaxies arcmin-2 to a limiting flux density of 2 × 10-17 ergs cm-2 s-1. No variation with wavelength (redshift) is statistically significant in their data set, which corresponds to H emission over the redshift range 0.7 z 1.9. Recombination theory coupled with an assumed stellar luminosity function can be used to relate H luminosity to the star formation rate (e.g., Kennicutt 1983; Madau, Pozzetti, & Dickinson 1998). Yan et al. (1999) find an average star formation rate of = 21 h50-2 M yr-1 for this redshift range.