Recent observational advances in understanding the co-evolution of galaxies and black holes point to the existence of multiple modes. Thus, taking a `macro' view of both the galaxy and the active galactic nucleus (AGN) host galaxy population is critical in any attempt to disentangle the large number of physical parameters such as mass, environment, morphology, and star formation rate. Large samples and high-quality multi-wavelength data are thus essential if we are to map out the evolutionary pathways that lead to co-evolution and feedback.
In order to understand co-evolution, we need to take such samples and assess two fundamental questions.
It is important to address both questions as it is possible to imagine particular phases of galaxy evolution leading to accretion, but not feedback. In this case, the assumption that a high AGN fraction in this population is indicative of a high impact of AGN on the host galaxies is misleading.
3.1. The Local Universe
The advent of large samples from the Sloan Digital Sky Survey (York et al. 2000) has enabled large studies of AGN host galaxies. These studies showed that AGN host galaxies appear to be an `intermediate' population: they are neither blue and actively star-forming, nor are they red and passively evolving; rather, they reside in the `green valley' in between the two general galaxy populations. Their morphology also appeared to be in-between: large bulges with some disk component. These observations have led to the interpretation of AGN host galaxies as galaxies undergoing transformation from blue star-forming galaxies to red and dead ellipticals (Kauffmann et al. 2003).
Recently, a more complex picture has emerged. The local AGN host galaxy population is in fact a composite of two distinct groups. The fraction of the local AGN host galaxy in major mergers is negligible (Darg et al. 2010), though the incidence of mergers in the Swift BAT sample is significantly higher (Koss et al. 2011); this is most likely due to the very shallow flux limit of the BAT sample which biases it to high luminosity objects. The remainder is divided between early-type galaxies (~ 10%) and late-type galaxies (~ 90%). Both of these two populations are very specific subsets of their parent populations (early-type and late-type galaxies, respectively) indicating that there are two fundamentally different modes of black hole fueling and co-evolution in the local universe (Schawinski et al. 2010a and Figure 1):
Figure 1. The distribution of the fraction of galaxies that host AGN on the color-mass diagram together with example images of an early-type (or elliptical) galaxy (top right) and a late-type (or spiral) galaxy (bottom right). The contours represent the normal galaxy population while the green shaded contours trace the fraction of galaxies with growing central black holes. The active fraction in a specific sub-population is a proxy for the duty cycle of AGN in that population; it reveals which galaxy populations have a high black hole growth duty cycle and illustrates the importance of morphology for understanding the role of black hole growth in galaxy evolution (Schawinski et al. 2010a). The morphological classifications were made by the citizen scientists taking part in the Galaxy Zoo project (galaxyzoo.org, Lintott et al. 2008, Lintott et al. 2011)
3.2. Early-type AGN Host Galaxies - Merger-driven Migration from Blue to Red
In the early-type galaxy population, black hole growth occurs in the least massive (109.5-10.5 M) early-type galaxies with the bluest host galaxy colors amongst the early-type population. They reside in the `green valley,' and their stellar populations show that they feature post-starburst objects migrating from the blue cloud to the red sequence at roughly fixed stellar mass, thus populating the low-mass end of the red sequence (Schawinski et al. 2007). They may represend a `downsized' version of the formation process undergone at high redshift by their more massive counterparts (Thomas et al. 2005, Thomas et al. 2010). Deep imaging of early-type galaxies along the evolutionary sequence migrating from blue to red indicate that this migration is initiated in at least half, and perhaps all cases, by a merger or interaction (Schawinski et al. 2010b).
However, the concentration of AGN host galaxies in the green valley also challenges the traditional picture of the quenching of star formation by AGN feedback. Even assuming instantaneous quenching, stellar evolution dictates that the migration from the blue cloud to the green valley takes at least ~ 100 Myr (roughly the lifetime of OB stars). Detailed stellar population age-dating by Schawinski et al. (2007) shows that the typical post-starburst timing of high-Eddington Seyfert activity ranges from ~ 270 Myr to ~ 1 Gyr, i.e. long after the quenching event. Thus, the black hole growth in the green valley and the associated energy that is liberated cannot be responsible for the shutdown of star formation. An unbiased search for AGN host galaxies using hard X-rays yields no `missing' powerful AGN in blue host galaxies (Schawinski et al. 2009a).
This observation does not entirely negate AGN as the agent responsible for the shutdown of star formation, it merely shows that the radiatively efficient Seyfert phase in the green valley is not the accretion phase responsible for the quenching of star formation. It may simply be the `mopping up' phase. At earlier times along the transition from the blue cloud to the red sequence, early-type galaxies do lose their molecular gas - the fuel for star formation - very rapidly (Schawinski et al. 2009b). This destruction of the molecular gas reservoir occurs before the high-Eddington Seyfert phase but coincides with weak AGN photoionization being present in the optical spectrum combined with still on-going star formation. Could this be the phase where a radiatively inefficient AGN is destroying the molecular gas reservoir?
Simple modeling of the depletion of molecular gas reservoirs following the Schmidt law for star formation (Schmidt 1959) by Kaviraj et al. (2011) shows that in the absence of an extra forcing mechanism, galaxies cannot rapidly quench their star formation since star formation is a self-regulated process. Every dynamical time, tdyn, they will convert a fraction of their available gas reservoir Mgas into stars with some efficiency (canonically ~ 0.02) resulting in a depletion timescale for gas-rich galaxies of many Gyrs - precisely what is observed for star-forming spirals. In order to rapidly quench star formation and enable the migration to the red sequence within ~ 1 Gyr, some process beyond star formation alone must destroy or make unavailable the present molecular gas reservoir; the best candidate for this process is (kinetic) AGN feedback.
3.3. Late-type AGN Host Galaxies - Secular Evolution and Stochastic Feeding
Most (up to 90%) of local AGN host galaxies are massive spirals. They show no evidence for post-starburst stellar populations (Wong et al. 2011) or morphological disturbances indicating a recent catastrophic interaction or change in star formation rate. In fact, the typical late-type AGN host galaxy has the physical parameters of the Milky Way (Schawinski et al. 2010a, Mutch et al. 2011): a massive spiral with a low specific star formation rate - hence the green valley host galaxy colors, in contrast to the early-types whose color is due to post-starburst stellar populations. Since there is no evidence for significant external forcing of the system, the most likely explanation for the accretion seen in these late-type host galaxies is stochastic feeding of the black hole via secular processes (Kormendy & Kennicutt 2004).
Since the typical local AGN host galaxy is similar to the Milky Way, the Galactic Center makes an excellent case study for what precisely leads to black hole feeding and feedback. While quiescent at the moment, observations show that the black hole in the Galactic Center was a low luminosity AGN as recently as 300 years ago as seen in hard X-ray light echos traveling across the molecular clouds surrounding the black hole (Revnivtsev et al. 2004, Nobukawa et al. 2008, Nobukawa et al. 2011). The recently-discovered Fermi Bubble may also be a remnant of recent accretion (Su et al. 2010).
Figure 2. Hubble Space Telescope WFC3/IR images of typical z ~ 2 moderate luminosity AGN host galaxies. These images are taken with the infrared channel of the Wide Field Camera 3 and show for the first time the rest-frame optical morphologies of z ~ 2 AGN. Analysis of the surface-brightness profiles of these galaxies show disk-like profiles in 80% of cases with the remainder composed of bulges and mergers. This means that a significant fraction of cosmic black hole growth in typical galaxies occurred in disk galaxies and are therefore likely driven by secular processes rather than major mergers as suggested by simulations by theorists (Schawinski et al. 2011a).
3.4. The High Redshift Universe
The bulk of both black hole growth and star formation occurs at high redshift with the peak epoch for both occurring around z ~ 2. This peak epoch has been difficult to study due to the lack of deep, high quality observations. Deep X-ray surveys over the last decade have captured black hole growth out to very high redshift and down to relatively low luminosities and have revealed a large population of moderate-luminosity AGN at 1 < z < 3 where normal mass black holes grow (see reviews by Brandt & Hasinger 2005, Treister & Urry 2011). This growth is slow as the Eddington ratios are moderate (Simmons et al. 2011).
Rest-frame optical spectroscopy from the ground is extremely challenging even with 8-10m telescopes (e.g. Treister et al. 2009) while space-based infrared spectroscopy is only now becoming possible with the new Hubble Space Telescope Wide Field Camera 3 (WFC3) which has a slitless (grism) mode (e.g. Straughn et al. 2011, Schawinski et al. 2011b).
Recent imaging observations with WFC3/IR have revealed that the majority of the moderate luminosity AGN at 1 < z < 3 feature disk light profiles rather than being massive bulges or major mergers (Schawinski et al. 2011a confirmed by Kocevski et al. 2011). This implies that the high redshift AGN host galaxy population is in fact very similar to that in the nearby universe: mostly disk galaxies, a few spheroids and virtually no major mergers. Thus, black hole growth at z ~ 2 is likely driven by stochastic fueling and secular processes. Observations from Herschel by Mullaney et al. (2011) support this picture as the far-infrared derived specific star formation rates of X-ray selected AGN host galaxies up to z ~ 3 are indistinguishable from the underlying galaxy population.
Combining this observation with what we know from the local universe, we arrive at a picture where a significant fraction of cosmic black hole growth can be attributed to secular processes (Schawinski et al. 2011a estimate ~ 40%) and that major mergers as a driver for black hole growth may be restricted to only a small fraction of the black hole growth in normal mass galaxies. Only at the highest luminosities do highly disturbed morphologies indicative of major mergers begin to appear (e.g. Urrutia et al. 2008, Zakamska et al. 2006).
A large caveat on any conclusions drawn from X-ray selected sample is that even the deepest Chandra deep field surveys miss the most obscured black hole growth phases even when the underlying bolometric luminosity of the AGN is high. The X-ray emission from these AGN is so heavily absorbed that they are only betrayed by `excess' mid-infrared emission from hot dust in the nucleus, which even large amounts of extinction cannot remove (e.g. Treister et al. 2004, Daddi et al. 2007, Fiore et al. 2008, Treister et al. 2009, Fiore et al. 2009, Treister et al. 2010). Expectations from simulations (e.g. Di Matteo et al. 2005, Hopkins et al. 2005a, Hopkins et al. 2005b) indicate that heavily obscured, high luminosity AGN (quasars) at high redshift should reside in galaxies undergoing major mergers, and that it is during this obscured phase preceding the classical unobscured quasar phase that the quasar begins to blow the gas, thus quenching the star formation in the host galaxy. Near future observation with Hubble may shed light on these poorly studied obscured AGN and whether they conform to the picture expected from simulations.