Annu. Rev. Astron. Astrophys. 2006. 44: xxx-xxx Copyright © 2006 by Annual Reviews. All rights reserved

5. CATCHING ELLIPTICALS IN FORMATION

Immediate precursors for cluster ETGs in the local universe were first identified with a population of blue, star-forming galaxies whose fraction in clusters increases very rapidly with redshift (Butcher & Oemler 1978, 1984, a trend known as the Butcher-Oemler effect), and therefore most such blue, star-forming galaxies had to subside to passive ETGs by z ~ 0. Dressler el al. (1997) and Fasano et al. (2000) showed that in clusters up to z ~ 0.5 the fraction of true ellipticals is fairly constant, while the fraction of (star-forming) spirals rapidly increases at the expense of the S0's. This was interpreted as evidence that galaxies in clusters that were star-forming spirals at z ~ 0.5, have changed their morphology to become passively evolving S0's by z = 0, a transformation that may account for much of the Butcher-Oemler effect. In clusters in the same redshift range, a sizable population of ETGs with strong Balmer absorption lines was also identified by Dressler & Gunn (1983), hence called K+A galaxies after the appearance of their spectrum. These ETGs were recognized as poststarburst galaxies, as further documented by Dressler et al. (1999, 2004) and Poggianti et al. (1999).

Thus, the metamorphosis from z ~ 0.5 to z = 0 of a fraction of cluster galaxies from star forming to passive is well documented by these studies. However, these changes seem to affect more spirals and S0's than true ellipticals, and if the bulk of star formation in ETGs took place at very high redshift, then we need to look further out to catch them in formation. From the low-redshift studies we have learned that ETGs are massive, highly clustered, and had to form the bulk of their stars at z 1.5-3 (depending on mass and environment) within a short time interval. Indeed, fast star formation is required by the -overabundance, and by the mere high formation redshift. At z ~ 3 the universe is only ~ 2 Gyr old, hence forming ~ 10¹¹ M of stars in one object between z = 5 and z = 3 (t ~ 1 Gyr) requires a star-formation rate (SFR) of 100 M yr^-1. Altogether, possible precursors of massive ETGs at low z could be searched among high-z massive, highly clustered, starburst galaxies with very high SFRs ( 100 M yr^-1). The rest of this section is dedicated to mentioning the main results in searching for such objects.

Lyman break galaxies (LBGs) were the first ubiquitous population of galaxies to be identified at z ~ 3, and their SFRs often in excess of ~ 100 M yr^-1, plus their relatively small size (R_e ~ 1-3 kpc), made them natural precursors to local bulges and ETGs (e.g., Giavalisco, Steidel, & Macchetto 1996, Steidel et al. 1996, Giavalisco 2002, but see also Giavalisco et al. 1995 for an even earlier attempt to identify a z = 3.4 precursor to ETGs). Recently, Adelberger et al. (2005) show that the 3D correlation length (r_o) of LBGs is such to match that of local lesser spheroids (M 10¹¹ M) when the secular increase of r_o is taken into account. However, Adelberger et al. note that the most massive and most rapidly star forming galaxies at high redshifts are likely to be lost by the Lyman break selection. From Figure 2 we see that local ETGs with M > 10¹¹ M include almost 50% of the mass in this kind of galaxies, and ~ 1/4 of the total stellar mass at z = 0. Therefore, looking at the starforming precursors of the most massive ETGs refers to a major component of the whole galaxy population. Hence the search was not limited to the LBG technique.

Other obvious candidates are the ultraluminous infrared galaxies (ULIRG, Sanders et al. 1988, Genzel & Cesarsky 2000), detected in mm or sub-mm surveys (e.g., Ivison et al. 1998, Lilly et al. 1999), a class defined for their infrared luminosity (8-1000 µm) exceeding ~ 10¹² L, and whose typical SFRs ( 200 M yr^-1) well qualify them for being precursors to massive spheroids, as does their very high (stellar) density ensuring that they would "land" on the fundamental plane as star formation subsides (Kormendy & Sanders 1992, Doyon et al. 1994). ULIRGs as ETG in formation are also advocated by Genzel et al. (2001), who from the resolved kinematics for 12 of them argue that typical ULIRGs are likely precursors to intermediate-mass ETGs rather than to giant ellipticals. However, the internal kinematics of one ULIRG at z = 2.8 indicates a mass 3 × 10¹¹ M (Genzel et al. 2003) making it a likely precursor to a very massive ETG. Blain et al. (2004) note that submillimeter-selected galaxies at z = 2-3 appear to be more strongly clustered than LBGs at the same redshifts, which makes them more attractive candidates than LBGs for being the progenitors of the most massive ETGs. Still, their space density falls short by a factor of ~ 10 compared to passive EROs at z ~ 1, and they could be the main precursors to EROs only if their duty cycle is very short. Nevertheless, Chapman et al. (2005) argue that the sub-mm galaxies may well be the dominant site of massive star formation at z = 2-3, once more making them excellent candidates for being ETG in formation.

The survey and characterization of sub-mm galaxies are currently limited by the modest sensitivity and resolution of existing sub-mm facilities, hence optical/near-IR selections are still the most efficient way of identifying large samples of massive star-forming galaxies at high redshift. For the star-forming BzK-selected objects, Daddi et al. (2004) estimate <SFR> 200 M yr^-1, which is typical of ULIRGs, and most of them are clearly mergers on ACS images. Indeed, out of 131 non-AGN (i.e., non X-ray emitter) star-forming BzKs with K < 20 in the GOODS-North field (Dickinson et al. in preparation), 82% were individually detected with Spitzer/MIPS at 24 µm (Daddi et al. 2005a). Moreover, by stacking the fluxes of the 131 objects ( <z> = 1.9) from radio to X-rays (i.e., VLA 1.4 GHz, SCUBA 850 and 450 µm, MIPS 24 µm, IRAC 8-3.5 µm, near-IR and optical bands, and Chandra's 0.5-8 keV) Daddi and colleagues showed that the resulting composite SED is an excellent match to that of a template ULIRG with L_IR = 1.7 × 10¹² L and <SFR> 250 M yr^-1, in agreement with the typical SFRs derived from the extinction-corrected UV flux. Two of these BzKs have also been detected at 1.2 mm with MAMBO, implying a SFR ~ 1000 M yr^-1 (Dannerbauer et al. 2006). So, the BzK selection proves to be an excellent way of finding large numbers of ULIRGs at high redshift, whose space density at z ~ 2 (~ 1-2 × 10^-4 Mpc^-3) is about three orders of magnitudes higher than the local density of ULIRGs, and a factor of 2-3 higher than that at z = 1. Moreover, the number of star-forming BzKs with M > 10¹¹ M is close to that of passive BzKs of similar mass, and added together nearly match the space density of massive ETGs at z = 0 (Kong et al. 2006). Hence, it is tantalizing to conclude that as star formation subsides in star-forming BzKs the number of passive ETGs will approach their local density. Finally, worth mentioning is that the majority among samples of J - K > 2.3 DRGs in the Extended HDF-South field (Webb et al. 2006) and GOODS-South field (Papovich et al. 2006) have been recently detected at 24 µm with Spitzer/MIPS, indicating that the majority of DRGs are likely to be dusty starburst precursors to ETGs, rather than having already turned into passive ETGs themselves. Moreover, DRGs appear to be distributed over a very broad and nearly flat redshift distribution, from less than 1 to over ~ 3.5 (Reddy et al. 2006).

Moderate redshift precursors to local ETGs are not necessarily star-forming. They may also be less massive ETGs that will merge by z = 0, an event now called "dry merging". The merger rate since z ~ 1.2 has been estimated by Lin et al. (2004) as part of the DEEP2 survey, concluding that only ~ 9% of present-day M^* galaxies have undergone a major merger during the corresponding time interval. However, Bell et al. (2006) searched for dry merger candidates over the GEMS field, and based on 7 ETG-ETG pairs estimated that each present-day ETG with M_v < -20.5 has undergone 0.5-1 major dry merger since z ~ 0.7. This may be at variance with the estimate based on the 3D two-point correlation function of local ETGs in the overwhelming SDSS database. Indeed, each local ETG is found to have less than 1% probability per Gyr of merging with another ETG, hence the dry merging rate appears to be "lower, much lower than the rate at which ETG-hosting DM halos merge with one another" (Hogg 2006), at least for z 0.36 (Masjedi et al. 2006).

The history of star formation in ETGs has been deduced from the properties of their passively evolving stellar populations in low and high redshifts galaxies, and we know that ETGs and bulges hold at least ~ 50% of the stellar mass at z = 0. Threfore, it is worth addressing here one last issue, even if only in a cursory way: is such an inferred history of star formation consistent with the direct measurements of the star-formation density and stellar mass density at high redshifts? Based on the estimate that the bulk of stars in spheroids formed at z 3, it has been suggested that at least ~ 30% of all stars (and metals) have formed by z = 3 (Renzini 1999). This appears to be at least a factor of ~ 3 higher than the direct estimate based on the HDF-North, according to which only 3%-14% of today's stars were in place by z = 3 (Dickinson et al. 2003). Data from HDF-South give the higher value 10%-40% (Fontana et al. 2003), most likely as a result of cosmic variance affecting both HDF fields. Based on the ~ 10 times wider field of the K20 survey, ~ 30% of the stellar mass appears to be in place by z ~ 2 (Fontana et al. 2004), but the corrections for incompleteness are large. Drory et al. (2005) find that over the ~ 200 arcmin² area of the combined GOODS-South and FORS Deep Field, ~ 50%, ~ 25%, and at least ~ 15% of the mass in stars is in place, respectively at z = 1, 2, and 3. So, no gross discrepancy has emerged so far between the mass density at z ~ 3 as directly measured and as estimated from the fossil evidence at lower redshift. The same holds for the comparison between the star-formation densities as a function of redshift, as inferred from the distribution of stellar ages of ETGs on the one hand, and as directly measured by observations on the other hand (e.g., Madau et al. 1996, Steidel et al. 1999, Giavalisco et al. 2004b). Errors on both sides are still as large as a factor of 2 or 3, but this should rapidly improve thanks to the deep and wide surveys currently under way.