3. SEARCHES AT NONOPTICAL WAVELENGTHS

While optical and infrared techniques are usually needed to adequately image distant galaxies and measure their redshifts, the candidates may initially be located by observations in less classical wave bands. We discuss identifying galaxies at nonoptical wavelengths in the following subsections, beginning with radio selection which has the richest history, and gradually moving to higher energy photons.

3.1. High-Redshift Radio Galaxies

Powerful radio galaxies have proven to be good targets as luminous, presumably massive galaxies at large distance. They are spatially rare and usually luminous over many decades of the electromagnetic spectrum. As is clear from Table 1, until only recently, studies of the highest redshift galaxies were synonymous with studies of high-redshift radio galaxies (HzRGs).

After the pioneering radio-optical identifications by Baade, Minkowski, Wyndham, Sandage, Ryle, Kristian, Longair, and Gunn, it became clear that relatively nearby powerful radio galaxies (P_{408 MHz} > 10²⁸ W Hz^-1) were usually associated with giant elliptical (gE and cD) galaxies (Matthews, Morgan, & Schmidt 1964). Upon closer inspection, a more precise description is that powerful radio sources are identified with disturbed-looking elliptical galaxies of high luminosity (see Fig. 1) and that optical absolute magnitude (M_V) correlates with radio power (P_{1.4 GHz}) such that the brightest early-type (giant elliptical) galaxies are much more likely to host a radio source of significant power. Could this trend be exploited to find massive galaxies at high redshift?

Figure 1. Near-infrared image of the HzRG 3C 257 at z = 2.474, obtained with the NIRC camera on Keck I. 3C 257 is the highest redshift radio galaxy identified from the 3CR survey. Panel is 12".0 square, centered on the HzRG, and oriented such that the inner radio axis is parallel to the abscissa. The compass arrow in the upper left indicates northeast orientation, with north shown by the heavier arrow. The lowest contour corresponds to a surface brightness of 21.6 mag arcsec-2. Note that the morphology is well described by a de Vaucouleurs r1/4 profile, as expected for an early-type system. Figure courtesy van Breugel et al. (1998).

The first major effort concentrated on the hosts of radio sources from the Revised Third Cambridge Catalogue (3CR; Bennett 1962), representing the brightest radio sources in the Northern hemisphere (S_{178 MHz} > 9 Jy; > -5°). Follow-up identifications by Longair, Gunn, Kristian, Sandage, Spinrad, and collaborators (cf. Djorgovski et al. 1988) has lead to nearly complete spectroscopic redshift identification for the 328 radio galaxies in the 3CR Catalogue. Only one faint galaxy, the host of 3C 249, remains without a definite spectroscopic redshift. The highest redshift belongs to 3C 257, a faint system (K = 18.4) at z = 2.474 (van Breugel et al. 1998). Lilly & Longair (1982, 1984) found a good correlation between the 2.2 µm (K band) infrared magnitude and galaxy redshift for powerful radio galaxies from the 3CR. At the redshifts of the 3CR, the K-band samples long-wavelength light, thought to be less heavily affected by young stars and active galactic nuclei (AGNs); it was initially hoped that the K-band flux correlated tightly with mass and that the HzRGs were tracking the evolution of early-type galaxies. This discovery of a tight infrared Hubble, or K-z, diagram briefly suggested that radio galaxies might be good standard candles even if located at large redshift and thus become a viable tool for measuring basic cosmological parameters, such as q₀. Although this line of research later proved unfruitful, HzRGs still generally obey the K-z relation out to the highest accessible redshifts currently probed (see Fig. 2), despite significant morphological evolution (van Breugel et al. 1998) and the dramatic k-correction effects (K samples rest-frame U at z ~ 5).

Figure 2. Hubble K-z diagram for HzRGs. Small filled triangles are Keck/NIRC measurements of HzRGs from van Breugel et al. (1998); the large filled triangle with error bars represents TN J0924-2201 at z = 5.19 (van Breugel et al. 1999) currently the most distant HzRG known; and all other photometry is from Eales et al. (1997). Magnitudes are aperture corrected to a 64 kpc metric diameter using H0 = 65 km s-1 Mpc-1 and 0 = 0.3. The inset highlights this empirical relation at the highest accessible redshifts currently. Two stellar evolutionary models from Bruzual & Charlot (1993) with formation redshifts zf = 20 and normalized at z < 0.1 are plotted. Figure courtesy van Breugel et al. (1999).

To identify higher redshift HzRGs and remove the redshift-radio power degeneracy in studies of HzRGs, several contemporary efforts have concentrated on lower flux density limit samples. Recent mJy surveys include the Molonglo Reference Catalog (MRC; McCarthy, Baum, & Spinrad 1996), the MIT-Green Bank survey (Stern et al. 1999b), and Cambridge surveys (e.g., 6C; Hales, Baldwin, & Warner 1993). Additional efforts have concentrated on µJy samples such as the Leiden Berkeley Deep Survey (Neuschaefer & Windhorst 1995) and sensitive radio maps of the Hubble Deep Field (Richards et al. 1998). These studies have shown that radio luminosity weakly correlates with emission-line luminosity (Rawlings et al. 1989; Baum & Heckman 1989), that weak radio galaxies do not generally contain significant nonthermal light in their optical spectra (Keel & Windhorst 1991), that the alignment effect (discussed below) becomes less pronounced with lower radio flux density samples (e.g., Rawlings & Saunders 1991; Dunlop & Peacock 1993; Eales & Rawlings 1993), and that radio power apparently correlates with ionization state in HzRGs (Stern et al. 1999b).

To identify the highest redshift HzRGs, two techniques have recently been exploited with considerable success, yielding the first detection of an HzRG at z > 5: TN J0924-2201 at z = 5.19 (van Breugel et al. 1999). First, empirically, HzRGs with ultrasteep spectra (USS; for S , -1.3) at radio wavelengths tend to be at higher redshifts (e.g., Chambers et al. 1990). All HzRGs at z > 3.8 have been identified by targeting USS samples. A partial explanation for this technique derives from the observation that the most powerful radio galaxies locally have radio spectral energy distributions which steepen with frequency (e.g., Carilli & Yun 1999). The k-corrections therefore imply that sources at increasing redshifts exhibit steeper radio spectral indices at a given observed frequency. The second technique relies on selecting those USS radio sources whose hosts have the faintest observed K magnitudes and thus, according to the K-z diagram (Fig. 2), are likely associated with the most distant systems. De Breuck et al. (1999) have recently constructed a full-sky USS sample (_{365 MHz}^{1.4 GHz} < -1.3). TN J0924-2201 at z = 5.19, a 1".2 radio double with S_{4.85 GHz} = 8.6 ± 0.5 mJy, _{365 MHz}^{1.4 GHz} = -1.63 ± 0.08, and K = 21.3 ± 0.3 (2".1 diameter circular aperture), is among the steepest spectrum USS sources from the De Breuck et al. (1999) sample and has among the faintest K-band magnitudes of the subsample with infrared imaging (van Breugel et al. 1999). Table 2 lists several physical parameters for all HzRGs at z > 3.8 in the literature currently. At the time of this writing, the authors are aware of 22 HzRGs at z > 3, including five at z > 4 and one at z > 5. We note, however, that K-band imaging is not the Rosetta Stone of HzRG studies; spectroscopy of several HzRGs with faint K-band identifications (e.g., K 20) reveal lower redshift systems, thus defying a simple one-to-one translation of K-band magnitude to redshift (e.g., Eales & Rawlings 1996).

Table 2. Physical Parameters of High-Redshift Radio Galaxies at z > 3.8

Galaxy	z	365 MHz1.4 GHz	K (mag)	S1.4GHz (mJy)	LLy	WLyrest (Å)	Reference
TN J0924-2201...	5.19	-1.63	21.3	73	0.8	115	1
VLA J123642+621331...	4.42	-0.94	21.4	0.432	0.1	50	2
6C 0140+326...	4.41	-1.15	20.7	92	7.7	700	3
8C 1435+63...	4.25	-1.31	20.1	498	2.3	670	4
TN J1338-1942...	4.11	-1.31	20.3	123	14.4	200	5
4C 41.17...	3.80	-1.25	20.7	266	9.0	100	6
NOTES. - Ly luminosity, LLy, is in units of 1043 h50-2 ergs s-1. WLyrest is the rest-frame Ly equivalent width. The discovery papers for 8C 1435+63 and 4C 41.17 were Lacy et al. 1994 and Chambers et al. 1990, respectively. References. - (1) van Breugel et al. 1999; (2) Waddington et al. 1999; (3) Rawlings et al. 1996; (4) Spinrad, Dey, & Graham 1995; (5) De Breuck et al. 1999; (6) Dey et al. 1997.

The radio morphologies of HzRGs often have a double-lobed structure. At low redshift, optical images of radio galaxies often show signs of recent mergers, as evidenced by multiple nuclei and subtle shells around some symmetric early-type galaxies to dramatic tidal tails around others (e.g., Rigler et al. 1992). At z 0.7, the optical axis of the host galaxy is typically aligned with the radio axis (Chambers, Miley, & van Breugel 1987; McCarthy et al. 1987). Optical and radio major axes usually differ by less than 20°. This alignment is most noticeable in emission-line images: nearly all of the emission-line regions of HzRGs in the 3CR sample are well aligned with their radio axis ( McCarthy, Spinrad, & van Breugel 1995). At z > 0.7 the rest-frame UV continua are also strongly aligned. At observed near-infrared wavelengths, the alignment is less pronounced. Lower redshift radio galaxies often have regular r^1/4 profiles in the K band (e.g., Best, Longair, & Röttgering 1996), while at the high-redshift end, van Breugel et al. (1998) find peculiar K-band morphologies - faint, large-scale (~ 50 kpc) emission often surrounding multiple compact components aligned with the radio axis.

One interpretation of the alignment effect is that it is dominated by a blue component, which has variously been attributed to emission from young, hot stars, scattered light from an AGN (a buried quasar or a buried miniquasar), or nebular continuum emission from clouds excited by an obscured nucleus (Dickson et al. 1995). Spectropolarimetric studies have convincingly demonstrated that extended, aligned UV emission in HzRGs at z ~ 1-2 is often strongly polarized, with the electric vector perpendicular to the major axis of the UV emission (e.g., Jannuzi et al. 1995; Dey et al. 1996; Cimatti et al. 1996, 1997). These observations strongly indicate that much of the observed UV light is AGN light which has been scattered into our line of sight by dust and/or electrons in the ambient medium. Other galaxies, notably, 4C 41.17 at z = 3.80 (Dey et al. 1997), exhibit no polarized continuum in the aligned emission, but do show stellar absorption features (e.g., S V 1502), indicating that (jet-)induced star formation is important in some HzRGs. The true scenario is likely a combination of these processes.

The observed infrared, sampling the rest-frame visible or UV for large redshifts, is thought to be dominated by starlight from the host galaxy; deductions made from the H or K bands might then be compared to other large galaxies of stars. The robustness of this conclusion is still unproven (e.g., see McCarthy 1999). To settle the point we need to observe stellar spectral features in the near-infrared, but spectroscopy of faint (K 19) galaxies at 2 µm is technically difficult, even with the largest telescopes. However, the uniformity of the K-z diagram (Fig. 2) suspiciously traces the gradual evolution of a massive galaxy of stars at high redshift, similar to expectations for the formation and evolution of early-type galaxies.

We now ask a more difficult question: what can the study of distant radio galaxies tell us about the youthful galaxy population in general? Until recently, HzRGs were the only stellar systems known at z > 3. However, modern photometric studies of deep imaging fields (discussed in detail in Section 4) have led to the discovery of a population of normal, star-forming galaxies at comparably large distances. At z ~ 3 the HzRGs are rather larger and more luminous than these field galaxies: a typical powerful radio galaxy at z ~ 3 has K 19, while the normal, star-forming, field galaxies at z ~ 3 typically have K 22, some 3 mag fainter. At z ~ 3 the HzRGs are typically spatially extended by 1"-2" at K, while the young, field galaxies are generally barely resolved in space-based (optical) images (half-light radii of 0".2-0".3; Giavalisco, Steidel, & Macchetto 1996). Assuming the observed K-band light is dominated by stars in both cases, it would take ~ 15 young field galaxies to match the luminosity of a powerful radio galaxy at z ~ 3. Could it be that the HzRGs were the first objects to form in the early universe? If so, what is the significance of the massive black hole that is thought to reside at the HzRG nucleus and power the synchrotron emission at radio wavelengths? Is it primordial, or the result of galaxy evolution in the early universe? Extending the frontiers of distant galaxy studies of both the young, star-forming systems and the HzRGs to earlier cosmic epochs will be a valuable enterprise. Eventually, as we probe to the epoch in which the (proto-)elliptical hosts of HzRGs are being constructed by galaxy mergers, we might expect to see diminishing systematic differences in the long-wavelength luminosities of HzRGs and star-forming galaxies.

Optical spectroscopy of a few (lower power) HzRGs at z 1.5 have shown spectra devoid of the narrow, high equivalent width emission lines which typically dominate HzRG spectra. Detailed studies of two galaxies (LBDS 53W091 and LBDS 53W069: Dunlop et al. 1996; Spinrad et al. 1997; Dey et al. 1999b) show that their observed optical spectra are well represented by an evolved stellar population of age 3.5 Gyr (see Fig. 3), with distinctive continuum decrements evident at 2640 and 2900 Å whose amplitudes are similar to those seen in late F-type stars. Although the modeling of the rest-frame UV spectra are somewhat uncertain currently - different spectroevolutionary models of a solar metallicity system with a delta function star formation history yield significantly different inferred ages (see Spinrad et al. 1997) - the hypothesis that these systems have ages a substantial fraction of the Hubble time at their observed redshifts seems unassailable. These systems are among our best examples of giant elliptical galaxies at z > 1.

Figure 3. Keck spectra of the two oldest known high-redshift galaxies. The upper panel shows the spectrum of 53W091 (histogram) overlaid with the mean spectrum of an F6 V star from the Wu et al. (1991) IUE Spectral Atlas. The bottom panel shows the spectrum of 53W069 (histogram) compared with the mean spectrum of an F9 V star. Since the UV spectrum of coeval populations is dominated by light from main-sequence turnoff stars, these fits imply ages of 3.5 Gyr for 53W091 and 4.5 Gyr for 53W069. These ages agree with those derived from more detailed population synthesis. Figure courtesy Dey (1999).

The modest surface density of faint-/moderate-strength radio sources reminds us of their rarity at any redshift: the surface density of radio sources with flux densities S_{1.4 GHz} > 1 mJy (i.e., a moderately weak radio flux) is 0.03 sources arcmin^-2 (Becker, White, & Helfand 1995). The comoving density of HzRGs is estimated to be 10^-8 h₅₀³ galaxies Mpc^-3 at z ~ 3 (e.g., Dunlop & Peacock 1990). For reference, the comoving density of early-type galaxies locally is ~ 2 × 10^-4 h₅₀³ Mpc^-3 (Cheng & Krauss 1999).

Even if HzRGs prove not to be identified with an early stage in the formation of all early-type galaxies, they remain convenient beacons for large, overdense regions, populated by smaller normal galaxies and are likely to be important for studies of the formation and evolution of large-scale structure. Several galaxy clusters at z 1 have been identified around distant radio galaxies (e.g., Dickinson 1995; Deltorn et al. 1997), with associated high-density regions located even out to z = 2.4 (Pascarelle et al. 1996; Francis et al. 1996; Pentericci et al. 1997; Carilli et al. 1998).

For a review of HzRGs, see McCarthy (1993). We now consider high-redshift galaxies selected at other nonoptical/infrared wavelengths. Selecting targets at other wavelengths has the potential to pick out the rare beast in the cosmos, which, from a Galilean world view, is apt to be far away.