|Annu. Rev. Astron. Astrophys. 2005. 43:
Copyright © 2005 by Annual Reviews. All rights reserved
At the time this review was written, most of the detailed informations on dusty galaxies in the 0.5 z 1.5 redshift range comes from galaxies selected with the ISOCAM cosmological surveys and the multi-wavelength analysis of detected sources. The ISOCAM extragalactic surveys were performed with two filters, LW2 (5-8.5 µm) and LW3 (12-18 µm) centered at 6.75 and 15 µm, respectively. However, because of star contamination and because the stellar light dominates the flux from galaxies in the 6.75 µm band above redshift 0.4, only the 15 µm surveys are relevant here. Shallow, deep and ultra-deep surveys were performed in various fields including the Lockman hole, Marano, northern and southern Hubble Deep Field (HDF), and Canada-France Redshift Survey (CFRS) (e.g., Aussel et al. 1999; Flores et al. 1999; Lari et al. 2001; Gruppioni et al. 2002; Mann et al. 2002; Elbaz & Cesarsky 2003; Sato et al. 2003). Deeper images have been made in the direction of distant clusters (e.g., Metcalfe et al. 2003). Finally, the bright end of the luminosity function was explored by the ELAIS survey (e.g., Oliver et al. 2000). The deepest surveys reach a completeness limit of about 100 µJy at 15 µm (without lensing). The most relevant data to this section are the deep and ultra-deep surveys.
4.1. Detailed Properties
To find out the nature and redshift distribution of the 15 µm deep survey sources, many followup observations have been conducted including HST imaging and VLT spectroscopy. With a point-spread function full width at half of maximum of 4.6 arcsec at 15 µm, optical counterparts are easily identified. Redshifts are found using emission and/or absorption lines. From field to field, the median redshift varies from 0.52 to 0.8, a quite large variation due to sample variance. Each field clearly exhibits one or two redshift peaks, with velocity dispersion characteristic of clusters or galaxy groups. Most of ISOCAM galaxies have redshifts between ~ 0.3 and 1.2, consistent with Figure 3. About 85% of the ISO galaxies show obvious strong emission lines (e.g., [OII] 3723, H, H, [OIII] 4959, 5007). These lines can be used as a diagnostic of the source of ionization and to distinguish the HII-region like objects from the Seyferts and LINERs. Most of the objects are found to be consistent with HII regions, e.g., from Liang et al. (2004) and exhibit low ionization level ([OIII] / H < 3). From emission lines studies, the AGN fraction is quite low, ~ 20%. This is consistent with X-ray observations of ISOCAM sources showing that AGNs contribute at most 17 ± 6% of the total mid-infrared flux (Fadda et al. 2002). Assuming template SEDs typical of star-forming and starburst galaxies, 15 µm fluxes can be converted into total infrared luminosities, LIR (between 8 and 1000 µm). About 75% of the galaxies dominated by the star formation are either LIRGs or ULIRGs. The remaining 25% are nearly equally distributed among either "starbursts" (1010 < LIR < 1011 L) or "normal" (LIR < 1010 L) galaxies. The median luminosity is about 3 × 1011 L. ULIRGs and LIRGs contribute to about 17% and 44% to the CIB at 15 µm, respectively (Elbaz et al. 2002). This suggests that the star formation density at z < 1 is dominated by the abundant population of LIRGs. As will be shown later, this has important consequences for the evolution of galaxies. Because of large extinction in LIRGs and ULIRGs, the infrared data provide more robust SFR estimate than UV tracers. The extinction factor in LIRGs averages to Av ~ 2.8 at z ~ 0.7 (Flores et al. 2004). It is much higher than that of the local star-forming galaxies for which the median is 0.86 (Kennicutt 1992). Assuming continuous burst of age 10-100 Myr, solar abundance, and a Salpeter initial mass function, the SFR can be derived from the infrared luminosities (Kennicutt 1998):
Thus typical LIRGs form stars at 20 M year-1. The median SFR for the 15 µm galaxies is about 50 M year-1, a substantial factor larger than that found for faint-optically selected galaxies in the same redshift range.
The other fundamental parameter characterizing the sources of the peak of the infrared background is their stellar mass content that traces the integral of the past star formation activity in the galaxies and is a natural complement to the instantaneous rate of star formation. The stellar masses can be obtained using spectral synthesis code modeling of the UV-optical-near infrared data or, more simply using the mass-to-luminosity ratio in the K-band. The derived stellar masses for the bulk of ISOCAM galaxies range from about 1010 to 3 × 1011 M, compared to 1.8 × 1011 M for the Milky Way. As expected from the selection based on the LW3 flux limit - and thus on the SFR - masses do not show significant correlation with redshift (Franceschini et al. 2003). An estimate of the time spent in the starburst state can be obtained by comparing the rate of ongoing star formation (SFR) with the total mass of already formed stars: t[years] = M / SFR. Assuming a constant SFR, t is the timescale to double the stellar mass. For LIRGs at z > 0.4, t ranges from 0.1 to 1.1 Gy with a median of about 0.8 Gyr (Franceschini et al. 2003; Hammer et al. 2005). From z = 1 to z = 0.4 (i.e., 3.3 Gyr), this newly formed stellar mass in LIRGs corresponds to about 60% of the z = 1 total mass of intermediate mass galaxies. The LIRGs are shown to actively build up their metal content. In a detailed study, Liang et al. (2004) show that, on average, the metal abundance of LIRGs is less than half of the z ~ 0 disks with comparable brightness. Expressed differently, at a given metal abundance, all distant LIRGs show much larger B luminosities than local disks. Assuming that LIRGs eventually evolve into the local massive disk galaxies, Liang et al. (2004) suggest that LIRGs form nearly half of their metals and stars since z ~ 1.
Finally, morphological classification of distant LIRGs is essential to understand their formation and evolution. Zheng et al. (2004) performed a detailed analysis of morphology, photometry, and color distribution of 36 (0.4 < z < 1.2) ISOCAM galaxies using HST images. Thirty-six percents of LIRGs are classified as disk galaxies with Hubble type from Sab to Sd; 25% show concentrated light distributions and are classified as Luminous Compact Galaxies (LCGs); 22% display complex morphology and clumpy light distributions and are classified as irregular galaxies; only 17% are major ongoing mergers showing multiple components and apparent tidal tails. This is clearly different from the local optical sample of Nakamura et al. (2004) in the same mass range in which 27%, 70%, < 2%, 3% and < 2% are E/S0, spirals, LCGs, irregulars and major mergers respectively. Consequences for galaxy evolution will be given in Section 4.3. For most compact LIRGs, the color maps reveal a central region strikingly bluer than the outer regions. These blue central regions have a similar size to that of bulges and a color comparable to that of star-forming regions. Because the bulge/central region in local spiral is relatively red, such a blue core structure could imply that the galaxy was forming the bulge (consistent with Hammer et al. 2001). It should be noticed that they find all LIRGs distributed along a sequence that relates their central color to their compactness. This is expected if star formation occurs first in the center (bulge) and gradually migrate to the outskirts (disk), leading to redder colors of the central regions as the disk stars were forming.
4.2 Cosmic Evolution
Another remarkable property of the 15 µm sources is their extremely high rates of evolution with redshift exceeding those measured for galaxies at other wavelengths and comparable to or larger than the evolution rates observed for quasars. Number counts at 15 µm show a prominent bump peaking at about 0.4 mJy. At the peak of the bump, the counts are one order of magnitude above the non-evolution models. In fact, data require a combination of a (1 + z)3 luminosity evolution and (1 + z)3 density evolution for the starburst component at redshift lower than 0.9 to fit the strong evolution. Although it has not been possible with ISOCAM to probe in detail the evolution of the infrared luminosity function, Spitzer data at 24 µm give for the first time tight constraints up to redshift 1.2 (Le Floc'h et al. 2005; Pérez-González et al. 2005). A strong evolution is noticeable and requires a shift of the characteristic luminosity L⋆ by a factor (1 + z)4.0 ± 0.5. Le Floc'h et al. (2005) find that the LIRGs and ULIRGs become the dominant population contributing to the comoving infrared energy density beyond z ~ 0.5-0.6 and represent 70% of the star-forming activity at z ~ 1. The comoving luminosity density produced by luminous infrared galaxies was more than 10 times larger at z ~ 1 than in the local Universe. For comparison, the B-band luminosity density was only three times larger at z = 1 than today. Such a large number density of LIRGs in the distant Universe could be caused by episodic and violent star-formation events, superimposed on relatively small levels of star formation activity. This idea has emerged in 1977 (Toomre 1977) and is fully developed in Hammer et al. (2005). These events can be associated to major changes in the galaxy morphologies. The rapid decline of the luminosity density from z = 1 is only partially due to the decreasing frequency of major merger events. Bell et al. (2005) showed that the SFR density at z ~ 0.7 is dominated by morphologically normal spiral, E/S0 and irregular galaxies ( 70%), while clearly interacting galaxies account for < 30%. The dominent driver of the decline is a strong decrease in SFR in morphologically undisturbed galaxies. This could be due, for example, to gas consumption or to the decrease of weak interactions with small satellites that could trigger the star formation through bars and spiral arms.
Locally 0.5% of galaxies with Lv > 1010 L have SEDs typical of LIRGs. This changes dramatically at higher redshift: in deep surveys, ISO detect about 15% of the MB -20 galaxies (LIRGs, Hammer et al. 2005) and Spitzer detect about 30% of field galaxies (Starbursts and LIRGs, Bell et al. 2005). Thus the two populations (optical and LIR galaxies) overlap more at high z.
4.3. Towards a Scenario of Recent Bulge and Disk Formation in Intermediate-Mass Spirals
Because a significant part of recent star formation takes place in LIRGs, any overall picture of galaxy evolution requires a detailed panchromatic study. Optical/spectral properties of LIRGs are similar to those of other galaxies and only infrared measurements are able to describe how the star formation is distributed between the different galaxy types. Thus a complete study has to link the star formation revealed in the infrared to the morphological changes seen in the optical. This has been done by Hammer et al. (2005) using HST, ISO, VLA and VLT observations of the CFRS. A detailed comparison of the morphologies of distant (0.4 < z < 1.2) galaxies with the local galaxies shows the complete vanishing of the LCGs in the local Universe (by a factor ~ 10) and the decrease of mergers and irregulars (by a factor ~ 4). Almost all distant galaxies have much bluer central colors than local bulges, probably as a result of active star formation in the 1kpc central region of most distant spirals. This supports a relatively recent formation of bulges in many present-day spirals. This simultaneous changes in galaxy morphologies and central colors of distant galaxies together with the observed lower metallicities (Liang et al. 2004) and overall higher star-formation rates at high z are the ingredients for an updated scenario of bulge and disk formation in spirals. Hammer et al. (2005) propose three different phases of galaxy evolution: the mergers/interacting, the compact galaxy and finally the growth of disk phase. During the last 8 Gyrs, most luminous galaxies are expected to experience a major merger that suppresses the disk as matter is falling to the mass barycenter. This phase is associated with short (1 Gyr) and strong peaks of star formations. Most of galaxies in this phase are LIRGs. Then, the compact phase corresponds to a decrease over 0.6-2 Gyr of the enhanced star formation due to merging. A significant fraction of stars form in bulges and additional occurrence of gas infall may subsequently wrap around the bulge to form a new disk-like component. Finally, the star formation spreads over all the forming disk as fed by large amounts of gas infall. In this scenario, about half of the bulge stellar content was made earlier in their progenitors, before the last major phase of accretion. More than a third of the present-day stellar mass is formed at z < 1. This scenario is in very good agreement with the hydrodynamical numerical simulations of Scannapieco & Tissera (2003) in which mergers, through secular evolution and fusions, transform galaxies along the Hubble sequence by driving a morphological loop that might also depend on the properties of the central potential wells, which are also affected by mergers. This very attractive scenario links in a simple way the distant and local galaxies; it will be confronted to the new panchromatic studies of Spitzer galaxies. Note that another possibility of buildup of dense central component in disk galaxies is internal secular evolution, as reviewed by Kormendy & Kennicutt (2004).