2. EVIDENCE FOR HIGH REDSHIFT DUST

High-redshift objects can be selected by many ways at a variety of wavelengths. While there is some overlap in the properties of the galaxies selected by the different techniques they typically go by different names in the literature. Table 1 summarizes the beasts in the high-z zoo.

Table 1. The high-z zoo

wavelength
observed	rest	name
optical	UV	High-z quasars
		Lyman Break Galaxies (LBG)
		Lyman- galaxies
		Quasar absorption line systems
NIR	optical	Extremely Red Objects (ERO)
sub-mm/mm	FIR	Sub-mm and mm galaxies, SCUBA galaxies
radio	radio	Micro-Jansky Radio Sources

2.1. High-z quasars. Due to their high-luminosities, quasars are relatively easy to detect out to the highest known redshifts (e.g. Irwin & McMahon, 1991; Fan et al. 2001; 2003). The current record holder has z = 6.43 (Fan et al. 2003). One of the most early and direct measurements of high-z dust came from Omont et al. (1996) who observed the z = 4.7 radio quiet quasar BR1202-0725 with the IRAM array. They simultaneously detected both mm wavelength dust continuum emission as well as molecular CO (5-4) and (7-6) lines in emission. The mm observations show two sources: the quasar and a companion galaxy. The dust emission of the latter is certainly powered by star-formation. So in this system we see star-formation and it's key ingredients, dust and molecular gas.

2.2. Lyman Break Galaxies. Lyman Break Galaxies are selected using broad band filters in the optical and/or NIR. The strong spectral break at = 912Å, due to the ionization of hydrogen, is readily seen in broad-band SEDs. For example, galaxies with z ~ 2.8 will have no flux in the HST/WFPC2 F300W (U) band but can be detected at longer wavelengths, hence the term U-dropout (or B-dropout, V-dropout, etc.). LBGs are typically selected for having a very red color (or color limit) using filters that straddle the break, but relatively blue color using filters that are longwards of the Lyman break (e.g. Madau et al. 1996; Steidel et al. 2003). This ensures the selection of star forming galaxies, and selects against very old or very dust reddened galaxies. The ability to obtain optical photometry to the 25th magnitude and beyond with current technology means that LBGs are the most abundant species in the High-z zoo (Adelberger & Steidel 2000, hereafter AS00).

LBGs display a remarkable similarity to nearby UV bright starburst galaxies as defined by the IUE atlas of Kinney et al. (1993). Both types have rest-frame SEDs that are at the blue end of those found for normal galaxies (Papovich et al. 2001); strong emission lines in the rest-frame optical (Pettini et al. 1998); rest-frame UV spectra dominated by high ionization wind lines as well as strong narrow interstellar absorption lines (Tremonti et al. 2001; Shapley et al. 2003); and a net blue shift of the interstellar absorption lines with respect to photospheric stellar lines, indicative of strong outflows in the ISM (Pettini et al. 2000; Shapley et al. 2003). The main difference is that LBGs are much more luminous (e.g. AS00), typically by an order of magnitude or more even without any dust corrections. Since the two types have similar high effective surface brightnesses (Meurer et al. 1997), LBGs are also much larger, typically having effective radii of a few kpc (Giavalisco et al. 1996). In short, LBGs look like local UV bright starbursts but scaled up in size and hence luminosity.

While we know a lot about the dust content of the local UV bright starburst population (e.g. Calzetti et al. 1994; 2000; Calzetti 2001; Meurer et al. 1995; Gordon, Calzetti & Witt 1997; Meurer et al. 1999, hereafter MHC99), much less is known directly about the dust content of LBGs. Most individual LBGs are not detected in the sub-mm with the Submillimeter Common User Bolometer Array (SCUBA), although there are a few rare exceptions (AS00, Baker et al. 2001). Stacked SCUBA studies also have had mixed success in detecting LBGs (AS00; Chapman et al. 2000; Peacock et al. 2000). Nevertheless, we can infer the presence of dust in LBGs via reddening: we know that they have a strong ionizing population from their emission line spectrum, yet their colors are not as blue as expected from un-reddened stellar populations. Using broad-band SEDs Papovich et al. (2001) estimate the reddening distribution of LBGs which peaks at E(B - V) 0.15. A variety of studies starting with Meurer et al. (1997) have estimated typical UV attenuations in LBG samples using just rest-frame UV colors. Most recent studies estimate UV attenuation factors around 5 using this method. One of the most recent works in this vein is Vijh, Witt & Gordon (2003) who also present a nice compilation of attenuation estimates for LBGs. This subject is addressed in more detail in Section 3.

2.3. Lyman- galaxies. Initially Ly emission was thought to be one of the best ways to detect the first epoch of galaxy formation (Partridge & Peebles 1967). However, after many disappointing surveys it was realized that there was something wrong with the original predictions (e.g. Pritchet 1994). Spectroscopic follow-up of LBGs showed that they have low rest-frame Ly equivalent widths ( 20Å) and fluxes lower than expected for their UV continuum strength (e.g. Steidel et al. 1996a, b; Lowenthal et al. 1997). This is due to the effects of resonant scattering of the Ly photons through the ISM of the galaxies, which greatly increases the total path required for the photons to escape the system. Hence even a small amount of dust is enough to greatly attenuate Ly emission compared to the neighboring continuum. In an expanding dusty ISM, such as a galactic wind, Ly photons can escape by back scattering out the far-side of the outflow resulting in a distinctly asymmetric line profile characterized by a sharp blue side cutoff. Such Ly profiles are indeed observed in both nearby starbursts (Kunth et al. 1998) as well as LBGs (Shapley et al. 2003).

Recent high-z Ly surveyors have learned their lessons and are going deeper and wider, and hence are becoming more successful (e.g. Rhoads et al. 2000). In fact the current most distant "normal" galaxies (z ~ 6.5) have been found using narrow band imaging targeting Ly (Kodaira et al. 2003). Spectroscopic confirmation of these and other blank-field Ly emitters inevitably shows the asymmetric profiles indicating the presence of a dusty expanding ISM (Kodaira et al. 2003; Rhoads et al. 2000; 2003).

2.4. Quasar Absorption Line Systems. The absorption lines in the spectra of quasars probe the gas phase of intervening systems, which need not necessarily be galaxies (self-gravitating conglomerations of stars, gas and dark matter). The most commonly observed feature seen is Ly absorption. In the literature, Ly absorption lines are referred to (in order of decreasing log(N_HI)) as "damped Ly absorption systems" (DLAS) with log(N_HI[atoms cm^-2]) 20, "Lyman limit systems" with 17 log(N_HI) 20, and "Lyman forest" clouds 14 log(N_HI) 17.

Metal absorption lines are also seen, albeit much less frequently. These include C, Si, Mg, S, Zn. Metals are also seen in the IGM out to z ~ 5, with little evolution in the cosmic density in Civ absorption for 1.5 z 5.5 (Songaila 2001; Pettini et al. 2003). This lack of evolution is somewhat puzzling. It may indicate a massive pollution event to the IGM at z > 5.5 (Songaila 2001), or alternatively may be due to IGM features being correlated with star formation which also evolves only weakly with redshift (Adelberger et al. 2003).

It is less clear that the IGM contains significant quantities of dust. We expect a bias against detecting dusty IGM clouds in optically selected quasar samples - the dust would diminish the flux of the background quasar (Fall & Pei, 1993). However, a study of a radio selected sample of quasars shows that the dust bias is at most a factor of two in absorption line systems having 2 z 3 (Ellingson et al. 2001). Prochaska et al. (2003) provide some evidence suggestive of dust in a DLAS at z = 2.6: the elemental abundance pattern of this system scales well to the solar abundance after correction for depletion onto dust grains. This on its own is not convincing proof of dust in DLASs, nor does it follow that lower column density sight lines of the IGM contain dust.

2.5. Extremely Red Objects. Elston et al. (1988) pointed out the existence of an interesting new population of galaxies having very red optical - NIR colors, R - K 5. These "Extremely Red Objects" have cropped up in numerous other deep NIR surveys although the exact selection limits vary. Detailed studies of the multi-wavelength SEDs of EROs show that they are a mixed bag, with roughly half being dusty starbursts and the other half being passively evolving (presumably dust-free) ellipticals at z ~ 1 (e.g. Smail et al. 2002a). Likewise, Ivison et al. (2002) find that about half of the radio-confirmed bright SCUBA sources have ERO or very red counterparts showing the strong overlap between the SCUBA and ERO populations.

By selecting purely in the NIR it is possible to select the most extreme galaxies - the Hyper Extremely Red Objects or HEROs (Totani et al. 2001) with colors J - K 3 so red that they can not be produced by pure passive evolution - some dust is required. Totani et al. find that the these are best modeled as very dusty starbursts at z ~ 3.

2.6. Sub-mm/mm Galaxies. The advent of SCUBA on the 15m James Clerk Maxwell Telescope made it possible to survey for dust emission from high-luminosity, high-z galaxies. SCUBA has been particularly effective at 850 µm, where "negative K-corrections" result in sources with fixed star formation rate having nearly constant flux as a function of z in the range of ~ 0.5 to 5 (Guiderdoni et al. 1997). The 850 µm confusion limit for SCUBA is ~ 2 mJy (Hughes et al. 1998) corresponding to Bolometric luminosities of ~ 2 × 10¹² L, about that of Arp 220. Hence only ultra-luminous (L_bol > 10¹² L) and hyper-luminous galaxies (L_bol > 10¹³ L) are detectable with SCUBA in blank fields. Similar star formation rate detection levels are also possible at 1.2mm, with using the MAMBO detector on the 30m IRAM telescope (Dannerbauer et al. 2002). Staring at strong lensing clusters allows the detection limit to be pushed down by a typical factor of ~ 3 (Smail et al. 2002b). The resulting lens amplification corrected number counts indicates that the 850 µm background is nearly completely resolved at sub-mJy levels.

The large beam sizes of SCUBA (15") and MAMBO (11") make identification of optical counterparts difficult. The optical counterparts are usually faint, and often not the most obvious galaxy in the sub-mm beam; Frayer et al. (2003) show an example of such an identification. Radio synthesis follow-up studies allows the counterparts of sub-mm and mm galaxies to be pinpointed to sub-arcsec accuracy. Ivison et al. (2002) find that 60% of their 850 µm "8 mJy sample" have robust radio identifications, and that 90% of those identified in the radio have near-infrared (NIR) and optical counterparts. Hence, over half of the brightest SCUBA sources have rest frame UV and optical counterparts. The success rate for finding optical and NIR counterparts for fainter highly magnified lensed SCUBA sources appears to be lower (Smail et al. 2002b), although the radio detection limits tend not to be that deep in those cases. The high dust luminosity and faint rest-frame UV and optical fluxes indicate that sub-mm/mm galaxies are similar to local Ultra Luminous Infrared Galaxies (ULIRGs; e.g. Goldader et al. 2002), but scaled up in luminosity.

Details and further information on sub-mm galaxies can be found in the excellent and comprehensive review of Blain et al. (2002).

2.7. Micro-Jansky Radio Sources. The radio emission of local star forming galaxies correlates very well with the FIR emission, although the physics behind the correlation is less clear (Helou et al. 1985; de Jong et al. 1985; Lisenfeld et al. 1996). Hence, radio observations are an excellent means to peer through the dust in a galaxy and observe star formation. For sources with very low fluxes, f(3.5 cm) 35 µJy, the frequency domain spectral slope (f `) becomes more steep ( ~ - 0.7) indicating that star forming galaxies are dominating over AGN (Fomalont et al. 2002). In contrast to the sub-mm, star forming galaxies dim considerably with redshift. Hence, spectroscopic, follow-up studies show that the majority of µJy sources are at redshifts z < 1.5. Simulations show that star forming galaxies should not be detectable for z > 3 in the deepest radio images currently available without evolution to include sources much more luminous than Arp 220 (Chapman et al. 2002).

The µJy sources with optically faint counterparts (I 24) are most likely to have z > 2 (Richards et al. 1999; Chapman et al. 2003a). The main evidence for dust in these is their high detectability rate with SCUBA at 850 µm (e.g. Barger et al. 2000; Chapman et al. 2003a). In addition, the optical counterparts tend to be redder than other field galaxies with the same I magnitudes and become progressively redder towards fainter magnitudes which also suggests the presence of dust in the host (Chapman et al. 2003a). Unlike the SCUBA galaxies, µJy radio galaxies usually have optical counterparts when one looks hard enough (Richards et al. 1999). Since these are rather faint, redshifts are difficult to obtain in the optical. Redshifts can be crudely estimated using the ratio of SCUBA and radio fluxes (e.g. Carilli & Yun 1999), however there are nasty degeneracies with dust temperature to contend with (Blain 1999).