The previous sections illustrate the ubiquity of obscuration in AGN and the diversity of the associated observational signatures. As we discuss below, AGN obscuration is intimately connected to both the fueling of the SMBH (through inflows of gas) and “feedback” (produced by the radiative and mechanical power of the AGN). To understand this connection, we require knowledge of the nature of the obscuring material: the scales, densities, composition, kinematics of the obscuring clouds, and the physical processes that produce them. A number of studies have treated AGN obscuration as dominated by a single regime (most often on the scale of a torus; e.g., Davies et al. 2015, Mateos et al. 2016, 2017), but it is increasingly clear that obscuration can occur on a range of scales and physical conditions. Furthermore, time-varying obscuration has been invoked in models of SMBH-galaxy co-evolution to explain the SMBH-galaxy relationships and the observed connection between AGN and starburst activity (e.g. Di Matteo, Springel & Hernquist, 2005, Hopkins et al., 2008, Alexander & Hickox, 2012). In this section we will focus on three main regimes of obscuration illustrated in Figure 11: (1) the nuclear “torus” posited by AGN unification models, (2) circumuclear gas associated with central starbursts, and (3) galaxy-scale material associated with galaxy disks and mergers.
 |
Figure 11. A schematic representation of the different scales of AGN obscuration considered in Section 4, for a Milky Way-type galaxy in the local Universe. For the vastly simplified assumption of constant density, for a given gas mass MH, the typical column density NH toward the nucleus decreases with size scale as R−2; the heaviest obscuration thus tends to occur on the smaller scales, although larger-scale Compton-thick obscuration can occur in discrete events such as galaxy mergers (Section 5.1) or at high redshifts where the gas fraction in galaxies is large (Section 5.2). |
An important consideration when comparing these different regimes is the characteristic obscuring column densities that can be associated with each scale. In the simplified case of a hydrogen cloud with constant number density nH distributed over a sphere of radius R, the column density will be NH = nH R and the gas mass will be, in terms of the mass of the hydrogen atom mH, Mgas = mH nH 4/3 π R3, such that MH ∝ NH R2. Thus for a given mass of gas, NH ∝ R−2, suggesting that the highest column densities will occur on relatively small scales and therefore that only modest amounts of obscuration can be expected over larger scales for a reasonable mass of gas, as illustrated in the schematic in Figure 11 (see also Buchner & Bauer 2017). (We note however that instabilities and disturbances such as galaxy mergers can significantly increase the gas density on ∼100 pc to kpc scales, and temporarily produce larger columns along some lines of sight, as discussed in Section 5.1.) In what follows, we will discuss the different regimes of obscuration in the context of observational and theoretical constraints on the typical values of NH.
4.1. The nuclear torus and the unified AGN model
As discussed in Section 1, in the physical model for the AGN central engine comprises of a small-scale, broadly axisymmetric structure of dust and gas that surrounds the SMBH, accretion disk, and BLR clouds, and obscures them along some lines of sight (e.g., Antonucci, 1993, Urry & Padovani, 1995, Netzer, 2015). This “unified AGN model” has been remarkably successful at explaining a number of properties of individual AGN (such as the existence of hidden BLRs in some Seyfert 2s; see Section 2.1.2) and of the demographics of the AGN population as a whole (such as the correlation between the X-ray luminosity produced by the accreting SMBH and the mid-IR luminosity that is reprocessed in the torus). However, in recent years it has become clear that the simplest models of the torus that posit a smooth, symmetric “donut”-like structure are inconsistent with observations, as discussed in Section 2.3. Recent reviews by Netzer (2015) and Ramos Almeida & Ricci (2017) give a comprehensive treatment of the status of the unified AGN model and its successes and challenges; here we will give a brief overview of some of the key points.
4.1.1. The basic torus properties are well-constrained. It is well-established that the inner regions of the obscuring torus are relatively compact (< 1 pc), from near- and mid-IR measurements of reverberation time lags (e.g., Suganuma et al., 2006, Vazquez et al., 2015) and spatially resolved emission from dust using mid-IR interferometry (e.g., López-Gonzaga et al., 2016). The radii of these mid-IR detected structures (assumed to be the AGN torus) closely follow a relationship with AGN luminosity of rtorus ∝ L1/2 that is remarkably consistent with the predicted sublimation radius for graphite dust (e.g., Barvainis, 1987, Burtscher et al., 2013) that is expected to represent the inner edge of the torus. A natural scale for the outer edge of the torus is the gravitational sphere of influence of the SMBH (within which the SMBH dominates the gravitational field), which will broadly correspond to a radius of ∼10 pc for nearby systems (e.g., see Section 2.2 of Alexander & Hickox (2012)). Observations of the outer regions of the torus have come from mid-IR imaging (e.g. Asmus, Hönig & Gandhi, 2016) and studies using molecular lines (e.g., García-Burillo et al., 2016), although the precise outer edge of the torus may be difficult to distinguish from a nuclear starburst disk, as discussed in Section 4.2. The wide observed range of obscuring NH in X-ray studies of AGN suggest a range of column densities through the torus, although the ubiquity of Fe Kα reflection features in AGN X-ray spectra (Section 2.2) suggests that in general, AGN tori are Compton-thick along some lines of sight.
4.1.2. The torus is clumpy. As discussed in Section 2.3, a broad range of evidence points to the torus being highly inhomogeneous in density, temperature, and composition, so that its overall structure is clumpy rather than smooth. One important piece of observational evidence pointing toward a clumpy structure comes from high-resolution mid-IR imaging. Models for smooth tori consistently predict weaker mid-IR emission for edge-on (Type 2) systems due to the torus itself having a large optical depth in the mid-IR, so that its emission is anisotropic (e.g., Fritz, Franceschini & Hatziminaoglou, 2006). For a clumpy torus, we expect to observe only the surfaces of the optically-thick clumps, which can be illuminated deep within the torus due to the optically thin lines of sight through the gaps between the clumps. This scenario produces mid-IR emission that is much less dependent on orientation (e.g., Nenkova et al., 2008, Stalevski et al., 2012), in agreement with the remarkably tight observed relationship between X-ray and mid-IR luminosities that is consistent for both Seyfert 1 and 2 galaxies (e.g., Gandhi et al., 2009, García-Bernete et al., 2016). Further evidence for a clumpy torus comes from observations of Si absorption features in the mid-IR spectrum (Section 2.3.2). A common prediction of models for smooth tori viewed edge-on are deep absorption Si absorption lines (e.g., Fritz, Franceschini & Hatziminaoglou, 2006). However a study of local Compton-thick AGN by Goulding et al. (2012, Section 2.3) showed that deep Si absorption is most often associated with larger-scale structures (dust lanes or galaxy merger features) rather than a smooth, small-scale torus. By contrast, high-angular resolution nuclear spectra of face-on, isolated Type 2 Seyferts show shallower Si absorption features that can be naturally produced by clumpy torus models (e.g., Roche et al., 2006, Alonso-Herrero et al., 2016).
4.1.3. The torus can have a range of covering factors, with dependence on AGN properties. One key parameter of the torus is the opening angle, or equivalently, covering factor (fC). An estimate of fC can be obtained for individual sources from detailed modeling of the X-ray spectrum (e.g., Brightman & Nandra, 2011) and studies of the ratio of reprocessed (IR) to direct (optical or X-ray) AGN emission (e.g., Toba et al., 2014), while the the average fC for an AGN population can be inferred from the fraction of sources that are obscured for given AGN parameters (e.g., Lawrence, 1991). Individual AGN are observed with opening angles over the full range from 0 to 90 degrees (e.g., Mateos et al., 2016), and sources with larger fC are statistically more likely to be observed as obscured than unobscured (e.g., Elitzur, 2012). Even for AGN of similar mass and luminosity, a broad range of torus properties are observed (e.g., Ramos Almeida et al., 2011, Burtscher et al., 2013).
Despite this broad diversity in the tori of individual AGN, there are general trends in average fC with various AGN parameters. It has long been observed that the obscured fraction (and thus the average fC) decreases with AGN luminosity (Section 3), which has been interpreted in terms of receding torus models in which increasingly luminous AGN progressively blow away more of the obscuring material (e.g., Lawrence, 1991). Studies of optical and soft X-ray samples have suggested that the obscured fraction drops as low as ∼10% at the highest luminosities (e.g., Lawrence, 1991, Hasinger, 2008). However, recent studies including more sophisticated modeling of incompleteness and anisotropy in the IR emission indicate a much weaker luminosity dependence, with the obscured fraction remaining as high as 50% even for the highest luminosities (e.g., Stalevski et al., 2016, Mateos et al., 2017). This suggests that while the inner radius of the torus increases with luminosity along with the dust sublimation radius (Section 4.1.1), the covering factor of the torus remains broadly constant at the highest luminosities.
It has recently been suggested that the key parameter determining fC may not be luminosity but Eddington ratio (L / LEdd; Buchner & Bauer (2017), Ricci et al. (2017c)), with fC limited by radiation pressure from the AGN acting on dust (e.g., Fabian, Vasudevan & Gandhi, 2008). In a study of local hard X-ray selected AGN, Ricci et al. (2017c) found that fC ≈ 0.8 at L / LEdd < 0.02 and then drops dramatically at higher L / LEdd, independent of AGN luminosity. In this picture, the minimum fC of ≈ 30% at L / LEdd > 0.5 is set primarily by the covering factor of Compton-thick material along the equatorial plane of the torus. A strong dependence of fC on L / LEdd indicates that most of the obscuring material is within the gravitational sphere of influence of the SMBH, suggesting that (at least for the local AGN in their sample) that a compact torus-like structure is the dominant source of obscuration.
4.1.4. The torus is dynamic. Any structure surrounding an accreting SMBH exists in a complex environment of inflow and outflow. The dynamic nature of the torus is not captured by ad hoc models of smooth or clumpy tori (which are generally static with time) but appear naturally in hydrodynamical models of gas flows around the SMBH. These can produce a variety of broadly axisymmetric structures that may be associated with a torus, including a warped accretion disk (e.g., Jud et al., 2017) or the interaction of inflowing gas with AGN- or starburst-driven winds to produce nuclear structures with large scale heights (e.g., Wada, Schartmann & Meijerink 2016, Hopkins et al. 2016, Hönig & Kishimoto 2017; see Figure 12). Some observations of broad absorption line features in AGN spectra have been interpreted as being viewed through axisymmetric outflowing winds that could be interpreted as a torus-like structure (e.g., Gallimore et al., 2016).
 |
Figure 12. Visualization of the density distribution in a hydrodynamic simulation of flows around a SMBH (Wada, Schartmann & Meijerink, 2016), shown face-on (left) and edge-on (right). The simulations illustrate a dynamical radiatively-driven “fountain” that can have obscuration and dust emission properties similar to those observed for some AGN. Figure from Wada, Schartmann & Meijerink (2016), courtesy of K. Wada. |
4.1.5. The torus may extend in the polar direction, and to large scales. An axiomatic feature of most torus models is that the obscuring gas and dust is broadly symmetric along the rotation axis of the accretion flow. This is motivated the presence of ionization cones observed in NLR gas (e.g., Zakamska et al., 2006, Fischer et al., 2013) and hidden BLRs in some Seyfert 2 galaxies (see Section 2.1.2). However, recent high-resolution observations of a handful of nearby AGN tori using mid-IR interferometry revealed presence of substantial dust emission along the polar direction on pc scales (e.g., Hönig et al., 2012, Tristram et al., 2014, López-Gonzaga et al., 2016). These polar structures can extend to larger scales as shown by imaging observations (e.g., Asmus, Hönig & Gandhi, 2016). The physical origin of these features is still unclear, but may be associated with an AGN-driven outflow (e.g., Schartmann et al., 2014). In addition, recent observations have cast some doubt on the notion of a compact torus as being the sole origin of reflected X-ray emission as is often assumed in modeling of obscured AGN (e.g., Murphy & Yaqoob, 2009, Brightman & Nandra, 2011). Spatially resolved Chandra observations have found evidence for Fe Kα lines produced up to ∼kpc away from the nucleus (e.g., Bauer et al., 2015, Fabbiano et al., 2017). Taken together, these results suggest that emission features that have previously been attributed to a compact, axisymmetric torus may often originate from gas and dust with very different geometries. Obscuring material on larger scales may be associated with nuclear starbursts, which are discussed in the next subsection.
4.2. Obscuration by nuclear starbursts
A starburst disk on < 100 pc scales is a natural consequence of the significant inflow of gas into the central regions of the galaxy that is required to produce rapid accretion onto the SMBH (e.g., Thompson, Quataert & Murray, 2005, Davies et al., 2009). On the scales of the entire galaxy, far-IR observations of AGN have shown that there is a relatively weak correlation between the AGN luminosity and current (or recent) star formation (e.g., Rosario et al., 2012, Stanley et al., 2015). However, these relationships are found to become tighter when measured over smaller spatial scales (e.g. Diamond-Stanic & Rieke, 2012, Esquej et al., 2014), confirming that accreting SMBHs often have a substantial reservoir of gas within the central 100 pc that can fuel a starburst disk.
Such gas is generally kinematically decoupled from the larger galaxy disk, and radiation pressure can cause the starburst disk to expand to a large scale height (e.g., Thompson, Quataert & Murray, 2005, Hopkins et al., 2016). Sampling all lines of sight, starburst disks can produce NH distributions that are broadly consistent with observations of the AGN population (Ballantyne, 2008, Hopkins et al., 2016, Gohil & Ballantyne, 2017). As per the discussion in Section 4.1.1, Compton-thick obscuration in these models is generally limited to small-scale structures (< 1 pc for a 3 × 107 M⊙ SMBH) that are difficult to distinguish from a torus. However, Compton-thin obscuration by starburst disks on larger (> 10 pc) scales may contribute significantly to the total population of obscured AGN.
4.3. Obscuration by galaxy-scale material
In addition to structures directly related to accretion flows onto the SMBH, obscuration can be produced by gas on the scales of the entire galaxy (> kpc). In a cosmological context, large-scale obscuration is common to models in which SMBH-galaxy co-evolution is driven by galaxy mergers, whereby the gas flows onto the SMBH are connected to galaxy-scale disturbances associated with merger-driven torques (e.g., Di Matteo, Springel & Hernquist, 2005, Hopkins et al., 2008, Alexander & Hickox, 2012). In this “evolutionary” picture, the earliest phases of rapid SMBH growth are surrounded by powerful starbursts and shrouded in dust clouds produced by the merger, followed by a “blowout” due to radiative feedback of the AGN that produces an unobscured quasar. Motivated by these theoretical expectations, a number of observational studies have explored the question of whether AGN obscuration can be associated with galaxy-scale structures rather than a nuclear torus or starburst disk.
One approach to identifying galaxy-scale obscuration in AGN is to search for links between obscuration and disturbed or merger morphologies of the host galaxies. From an observational perspective, the merger-AGN connection has been controversial, with some studies suggesting a strong connection, others showing no relationship, and some suggesting a dependence on AGN luminosity (e.g., Koss et al., 2010, Treister et al., 2012, Villforth et al., 2017). Recent results indicate clearly that merging galaxies are more likely to host AGN than isolated galaxies with otherwise similar properties (e.g., Ellison et al., 2013, Weston et al., 2017, Goulding et al., 2017), but whether these mergers are associated with obscuration remains unsettled. For low-luminosity nearby systems, the excess of AGN in mergers is significantly stronger for AGN selected in the IR with WISE than for (presumably less-obscured) optically-selected AGN (Satyapal et al., 2014), although the WISE color-selected AGN may suffer contamination from low-metallicity starbursts (Hainline et al., 2016). Studies of IR-selected quasar hosts at z ∼ 1–2 show no clear connection between merger morphology and obscuration (Farrah et al. 2017), and similar results were found for a X-ray selected AGN at z ∼ 2 (Schawinski et al., 2011, Kocevski et al., 2012). However, recent studies report a possible connection between galaxy mergers and Compton-thick AGN obscuration (Kocevski et al., 2015, Ricci et al., 2017a). While the obscuring material in these studies is usually modeled to have a small-scale torus geometry, in principle the characteristic X-ray features might be produced by reflection off clouds on larger scales associated with the merger (e.g., Levenson et al. 2002, see Section 4.1.5). For the population of reddened quasars (which exhibit a visible but highly reddened AGN continuum and represent some of the most luminous known AGN; see Section 2.1) a very high fraction (∼ 80%) are associated with galaxy mergers and disturbances (e.g., Glikman et al., 2015). Powerful, heavily-obscured WISE-selected quasars at z ∼ 2 (e.g., Assef et al., 2015) also exhibit a large fraction of mergers (Fan et al., 2016). These results are suggestive of a link between mergers and powerful obscured AGN, but further work is needed to confirm this phenomenon.
In an evolutionary scenario, the same galaxy-scale dust and gas that obscures the AGN may also be expected to produce enhanced star formation. Any distinction between the star-forming properties of obscured and unobscured AGN immediately rules out the simplest unified AGN models, in which obscuration is purely due to orientation of the dusty torus. Far-IR and submm studies of luminous quasars show that obscured sources exhibit stronger emission from cold dust; this conclusion holds for obscuration measured in X-rays (e.g., Page et al., 2004, Page et al., 2011) and also from IR-optical SEDs (e.g., Chen et al., 2015). Chen et al. (2015) furthermore showed that the fraction of quasars that are obscured increased strongly with far-IR luminosity (Figure 13), consistent with a picture in which obscuration in luminous AGN is frequently associated with galaxy-scale dust. We emphasize, however, that these studies focused primarily on powerful quasars; for less luminous AGN classified as obscured or unboscured in the X-rays or optical, no comparable difference in average far-IR emission is observed (Merloni et al., 2014). These results suggest that a connection between obscuration and galaxy-scale star-forming material may be most prevalent in powerful AGN.
 |
Figure 13. The connection between obscuration and star formation in mid-IR luminous quasars, adapted from Chen et al. (2015). The left panel shows fits to the optical–FIR SEDs (including Herschel data) of an unobscured and an obscured quasar identified using mid-IR and optical photometry (Hickox et al., 2007, Chen et al., 2015). Obscured quasars in this sample exhibited stronger far-IR (cold dust) emission compared to their unobscured counterparts; the right panel shows that the fraction of AGN that are obscured rises significantly with far-IR luminosity, suggesting a connection between AGN obscuration and larger-scale star-forming dust, as discussed in Section 4.3. Figures from Chen et al. (2015), courtesy of C.-T. Chen. |
A final piece of the evolutionary puzzle comes from spatial correlation studies, which provide a robust statistical measure of the large-scale structures (i.e., dark matter halos) in which galaxy and AGN reside, independent of systematics in measurements of galaxy or AGN properties (e.g., Berlind & Weinberg, 2002). Differences in the host halo masses between AGN types would rule out the simplest unified AGN models. Comparisons of obscured and unobscured AGN clustering have engendered significant debate, with some studies showing stronger clustering for obscured AGN, others for unobscured AGN, and still others showing no difference (e.g., Hickox et al., 2011, Allevato et al., 2011, Mendez et al., 2016). The large samples of > 105 quasars identified with WISE have enabled high-precision measurements, consistently showing stronger clustering for the obscured population (e.g., Donoso et al., 2012, DiPompeo et al., 2014). This difference has been confirmed through independent cross-correlations of the quasar positions with lensing maps derived from the cosmic microwave background (e.g. DiPompeo et al., 2015, DiPompeo et al., 2017a). These results can be explained qualitatively with a model in which the obscured quasars have SMBHs that are undermassive relative to their halos and are “catching up” to their final mass, consistent with an evolutionary scenario (DiPompeo et al., 2017b).