1.1. Where To Find Basic Information
Active Galactic Nuclei (AGN) is a term encompassing a variety of energetic phenomena in galactic centers which are thought to be powered directly or indirectly by accretion of matter onto central supermassive black holes. There are many reviews of this field, including several of book length. Perhaps the most technical and broad book-length review is that of J. Krolik, which was published in 1998. The classic text on spectral analysis of gaseous nebulae and AGN has been updated in 2006 (Ferland and Osterbrock); it's a wonderful book, but it is biased towards the optical and ultraviolet regions of the spectrum. A new edition (2009) of An Introduction to Radio Astronomy by Burke and Graham-Smith is also noteworthy.
Few theoretical predictions have been borne out in this field, and understanding is still semi-quantitative at best (Alloin et al. 1985; Antonucci 1988, 2002a; Courvoisier and Clavel 1991; Koratkar and Blaes 1999; Blaes 2007). This review will refer to some fairly general theoretical ideas, but it will mainly organize some observational information on radio galaxy central engines that has become clear over recent years. It will not include a general introduction to AGN (see above references), or even to orientation effects in AGN, but will address the nature of the central engines in radio galaxies via thin electromagnetic output, which is a topic tied up as a practical matter with orientation effects on observations.
The general topic of orientation effects ("Unified Models") is reviewed in detail in Antonucci 1993. The material in that review is almost entirely "still true." However, it has been updated and elucidated in several more recent (but generally narrower) reviews (Urry & Padovani 1995; Dopita 1997; Cohen et al. 1999; Wills 1999; Axon 2001; Tadhunter 2008).
In a nutshell, prior to the mid-1980s, it seemed that radio loud Quasars (with their powerful thermal optical/UV light) were quite distinct from radio galaxies; the latter are analogous to Quasars in general radio properties, but apparently lacked the strong optical/UV electromagnetic luminosity (Big Blue Bump). In a landmark review of bright extragalactic radio sources, Begelman, Blandford and Rees (1984) posited that "The ratio of mass accretion rate to the mass of the hole may determine whether a [radio loud] active galactic nucleus will be primarily a thermal emitter like an optical Quasar or a nonthermal object like a radio galaxy."
Through optical spectropolarimetry and other means, it was subsequently determined that many radio galaxies, especially the most powerful ones, with strong high-ionization narrow emission lines, actually harbor hidden Quasars surrounded by opaque dusty tori, so that the observational appearance depends on the inclination of the radio axis to the line of sight. But now we know that for many radio galaxies, any hidden Quasar must be very weak.
The radio quiet and radio loud objects with high ionization are virtually all visible (Type 1) or hidden (Type 2) Seyferts or Quasars (e.g., Antonucci 2002b). At lower radio luminosities of all radio types, we find mostly LINERs 6 (Low Ionization Nuclear Emission Regions). A recent and comprehensive review of LINERs is that of Ho (2008).
1.2. Nature of Geometrical Unified Models
In the low-redshift universe, there is a near-perfect correspondence between "radio loud" objects - with L(1.5 GHz) loosely extending from perhaps ~ 1028 - 1036 erg sec-1 Hz-1 - and Elliptical hosts. (This paper uses H0 = 70 km sec-1 Mpc-1, matter = 0.3, and = 0.7.) For a fiducial L ~ -1 spectrum, for which there is equal power per logarithmic frequency interval - also referred to as "per dex" - the parameter L gives the power integrated over an interval of 0.30 dex (powers of 10). Thus if one integrates such a spectrum over the "radio region" 30 MHz-300 GHz, the corresponding luminosity is 13.3 times as great; a look at various radio AGN in the NASA Extragalactic Database (http://nedwww.ipac.caltech.edu) shows that the luminosity per dex of the dominant optically thin synchrotron component tends to be lower outside this range, so depending on the application, one may consider this as a crude "radio bolometric correction." Resulting radio powers can exceed 1 × 1045erg/s. Estimates of energy tied up in 100-kpc scale radio lobes are as high as 1 × 1061 ergs or more, even using the particle/magnetic field minimum-energy assumption and assuming a lack of a dominant proton contribution. (The minimum-energy assumption posits that energy is apportioned between relativistic electrons and magnetic field in such a way as to minimize lobe energy content for a given synchrotron luminosity.)
Unified models assert that certain AGN classes differ only in orientation with respect to the line of sight. These models comprise two separate (though interacting) assertions, as illustrated in Fig. 1. The first to be recognized historically is the effect of relativistic beaming (aberration causing anisotropy in the observed frame) in the powerful synchrotron jets which feed particles and magnetic energy into the radio lobes. When seen at low inclinations, beaming amplifies and speeds up "core" (subparsec scale jet, usual synchrotron-self-absorbed) radio flux variability and (apparent faster-than-light) "superluminal motion." Thus a special fortuitous orientation of a nearly axisymmetric object leads to a different observational category (Blazars). The same objects, seen at higher inclination, are ordinary radio-loud galaxies and Quasars (Blandford et al 1984; Antonucci and Ulvestad 1985; Kollgaard et al 1992, etc.). We can call this the beaming unified model (or more loosely, just the beam model, though that term usually connotes some connection with actual physics).
Figure 1. A schematic diagram of the current paradigm for radio-loud AGN (not to scale). Surrounding the central black hole is a luminous accretion disk. Broad emission lines are produced in clouds orbiting outside the disk and perhaps by the disk itself. A thick dusty torus (or warped disk) obscures the Broad-Line Region from transverse lines of sight; some continuum and broad-line emission can be scattered into those lines of sight by warm electrons or dust that are outside the observing torus. Narrow emission lines are produced in clouds much farther from the central source. Radio jets, shown here as the diffuse 2-sided jets characteristic of low-luminosity, or FR I-type, radio sources, emanate from the region near the black hole, initially at relativistic speeds. (adapted from Urry & Padovani 1995)
At low redshift, radio-quiet (but not silent) AGN lie in Spiral hosts, and go by the name of Seyfert galaxies. They rarely show detectable motions in their weak radio jets - and when they do show motions, the apparent speed is usually much less than the speed of light. More luminous radio quiet objects are called Radio Quiet Quasars, or historically, Quasistellar Objects (QSOs).
But both radio quiet and many radio loud AGN widely exhibit another kind of orientation unification: many well-studied objects include energetically dominant continuum components in the optical-ultraviolet region, referred to as the Big Blue Bump, and widely attributed to thermal radiation from optical thick accretion flows. (Some confirmation of the latter can be found in Kishimoto et al 2004, 2005, 2008.) They are almost always accompanied by broad (5000-10,000 km/s) permitted emission lines. Both these components reside inside optically opaque dusty structures which to zeroth order have the shadowing properties of tori, and these structures are referred to loosely as the "AGN torus." In some cases we have a direct (low-inclination, polar) view of these compact components (in Quasars, Seyfert 1 galaxies, and Broad Line Radio Galaxies). In many other cases (high inclination, equatorial views), we can be sure these two components are present inside the tori using the technique of optical spectropolarimetry, which literally allows us to see the nuclei "from above," using ambient gas and dust as natural periscopic mirrors.
This wonderful trick of optical spectrapolarimetry (e.g. Antonucci 1982, 1983, 1984; Antonucci and Miller 1985) uses the polarization property of scattered light to separate the spectrum of hidden sources from any sources of direct light. The above references show that the Big Blue Bump and the Broad Line Region are present but hidden from direct view. The scattering polarization position angle indicates that the photons emerging from these components can only escape from the hidden nuclei if moving along the radio jet (and lobe) axis, so that the other directions must be blocked by an obscuring torus. This is illustrated in Fig. 2, from Tran et al 1995, illustrating the total and polarized fluxes, for 3CR234. The polarized photons (scattered light) contribute to both plots of course, but show up with great contrast when plotted alone.
Figure 2. Total and polarized flux for the Narrow Line Radio Galaxy and shows the first hidden Quasar, 3CR234. The polarization angle relative to the radio axis shows that photons can only stream out of the nucleus in the polar directions. The polarized flux (akin to scattered light) spectrum shows the hidden Quasar features at good contrast. (In this case, however, the scattered ray is itself reddened.)
If all objects fit in with the above descriptions of orientation unification, we would be left with as few as two independent types of "central engines" on relativistic scales, those for radio quiet and those for radio loud AGN. But we now know that many radio galaxies lack powerful visible or hidden Type 1 engines (Big Blue Bump, Broad Line Region, copious accretion flows). They occur frequently in some regions of parameter space (as delineated for example by redshift and radio flux or luminosity), and not in others. The purpose of this review is to gather the multiwavelength evidence for the last two statements, and emphasize the dependence of the distribution of central engine types on parameter space, which is the key to avoiding many errors and much confusion and even discord. We will see that a great deal of self-consistent information is available on the occurrence of the two types of radio galaxy/Quasar central engines.
1.3. Types of Powerful Radio Source; Scope for Unification by Orientation
Within the radio loud AGN arena, several types are also distinguished based not on central engine properties but on extended radio morphology. Unfortunately, they do not correspond very closely to the optical catagories!
The most luminous giant ( 100 kpc) double radio sources are designated FR II ("Classical Double"), for Fanaroff and Riley (1974); those authors noted that a whole suite of properties change together fairly suddenly over a critical radio luminosity (referring to the roughly isotropic diffuse emission). Sources above ~ 2 × 1032 erg/sec/Hz at 1.5 GHz show edge-brightening and hot spots, where the radio jets impinge on an external medium, and shocks partially convert bulk kinetic energy to particle and field energy. They also tend to have strong side-to-side asymmetry (generally attributed to relativistic beaming) of the jets over scales from the relativistic region up to tens of kpc. The lower luminosity giant (also 100 kpc) objects (FR I galaxies) also have strong side-to-side jet asymmetry, but only on 1-1000 pc scales in most cases. An important refinement to the FR classification scheme is that the dependence of the exact radio luminosity cutoff depends on the optical luminosity of the host galaxy (Owen and Ledlow 1994, but see Best 2009, Fig. 4a).
The "FR" types of radio galaxy generally have sizes of 25-1000 kpc, but large populations of smaller sources exist, and they can still be very powerful. (The small sizes mean that their lobe energy content is very much lower, however.) They are denoted in an inconsistent way, according to their means of discovery. Optically thin, steep-spectrum radio sources which were historically unresolved on arcminute scales, were (and are) called Compact Steep Spectrum (CSS) sources to distinguish them from opaque beamed synchrotron cores; their spectra peak at ~ 100 MHz, and they are generally defined to be in the size range 1 kpc - 15 or 25 kpc. Sometimes they are crudely defined to be sources smaller than typical host galaxies. Sources whose extent is less than ~ 1kpc, with even more compact substructure, are often dominated by synchrotron components which are self-absorbed up to ~ GHz frequencies 7, and are called Gigahertz-Peaked-Spectrum sources (GPS). (Even tinier sources are being sought by selecting for self-absorption peaks at even higher frequencies.) These classes are reviewed by O'Dea (1998). Since that paper was written, much evidence has accumulated from VLBI proper motions that the sources are small because they are very young (~ 1000-100,000 years! 8). In at least some cases it is known from faint extended emission that these very young ages refer only to a recent phase of activity however. Statistically, only a very small fraction of small and very short-lived sources can grow to be huge bright long-lived sources.
The relevant properties of classes will be discussed in turn, generally starting with the radio data and proceeding upwards in frequency. We will discover that some FR II radio galaxies at z 0.5-1.0 lack powerful hidden Quasars. These objects may have hidden Type 1 nuclei, but they are constrained to be much weaker than those of the "matched" 9 visible Quasars, and thus they are not "unified" (identified) with them via orientation with respect to the line of sight. At low redshifts (z <~ 0.5) it is probable that only a minority of FR II radio galaxies in the 3CR catalog host hidden Quasars.
Next we will tackle the less powerful (FR I) giant radio galaxies, which are by selection nearby in almost all cases. The 3CR catalog flux cutoff of 10 Jy at 178 MHz (Laing, Riley and Longair 1983) corresponds to the FR I vs. II radio luminosity separation at z ~ 0.2. Most (but by no means all) of these objects have nuclear spectral energy distributions dominated by synchrotron radiation, with no evidence for visible or hidden "Type 1" central engines.
Finally the small, young GPS and CSS sources will be discussed. Very recent information from the ISO and Spitzer infrared satellites has greatly increased our knowledge of "shadowing unification" 10 at various radio luminosities.
A major caveat of this paper, and of this field, is that most of the information derives from the brightest radio sources, especially those in the 3CR catalog, so no implication is made for unexplored regions of parameter space! Another major caveat is that while we discuss thermal vs. nonthermal galaxies 11, relatively little evidence is presented that any parameter is bimodal, so that there could be a continuum of properties.
1.4. The Infrared Calorimeter
Radiation absorbed by the dusty torus is largely reradiated as infrared, and many studies have concluded that in reasonably luminous AGN (so that the IR is not dominated by a normal host galaxy), at least the near- and mid-infrared reradiation ( 1-40 microns) is dominated by reprocessed nuclear optical/UV/X-ray light. In specific populations, the entire IR seems to be radiated by the torus, because the colors are warm throughout, and there is evidence for only weak star formation (e.g. PAHs) or synchrotron radiation (e.g., strong radio-mm emission). Thus the infrared reradiation of nuclear light can potentially be used as a calorimetric indicator for the luminosity of any hidden AGN.
There are at least two ambiguities in using this infrared emission as an AGN calorimeter. The first is that some of the nuclear radiation may not reach the torus, either because of the intrinsic latitude-dependence expected for many models of the Big Blue Bump (e.g. Netzer 1985), or preferentially planar Broad Line Region absorption (Maiolino et al 2001c; Gaskell et al 2007). Both of these effects add noise to the dust reradiation calorimeter, and tend to make hidden Quasars look dimmer than visible Quasars in the infrared for a given opening angle. Nevertheless most studies show infrared luminosity about as expected, from detailed studies of individual objects (Carleton et al 1984; Storchi-Bergmann et al 1992) and for populations of Type 1 and Type 2 objects within isotropically selected samples (Keel et al 1994). Carleton et al (1984, see Fig. 3) shows how the infrared calorimeter works, based on the first spectropolarimetric hidden AGN, 3CR234.
Figure 3. 3CR234 continuum observations plus model fits. The ordinate is linear in L, and the area under any portion of a curve is proportional to the luminosity in that portion of the spectrum. The hidden optical/UV component (Big Blue Bump) has been reprocessed into the infrared in this hidden Quasar.
Also, although the covering factor deduced from dividing infrared re-emission by Big Blue Bump luminosities are therefore lower limits, they tend to be high (~ 0.1-1) so that they can't be too far off! 12
Another concern with the infrared calorimeter is the expected anisotropy of the thermal dust emission due to the large dust column densities (Pier and Krolik 1992, 1993). Many Type 2 AGN have X-ray columns of 1 × 1024 cm-2; absorption of mid-IR lines, molecular maps, and the great difference in the average X-ray columns between Type 1 and Type 2 AGN suggest that a commensurate dust extinction is present. (See Maiolino et al 2001a, b, c for arguments which affect this line of reasoning quantitatively but not qualitatively.)
There is a limit on the anisotropy of the ratio [O III] 5007 / F(60µ) in Seyferts from Fig. 3 of Keel et al 1994. The figure shows that in their well-selected sample (60µ flux with a mild 25µ - 60µ warmth criterion), Type 1 and Type 2 objects have indistinguishable distributions of L(60µ), L, and of course their ratio. The 5007 line is produced outside the torus in most objects like these: it doesn't appear in polarized flux along with the broad emission lines and Big Blue Bump. Thus it's not significantly hidden inside the torus, is likely quite isotropic in this parameter space, and is nearly isotropically selected 13, fairly powerful Seyferts.
Torus models do predict that the optical depths will be small in the far-infrared in general. For AGN-dominated infrared SEDs, that means there is an elegant and detailed method of deriving the degree and wavelength-dependence of the dust emission anisotropy. For isotropically selected samples, we can divide composite or representative Type 1 SEDs by those for Type 2, tying them together at 60µ.
In order to make the infrared calorimeter accurate, we need to account for the common anisotropy of the near- and mid-thermal dust emission. There is general agreement on near isotropy past ~ 30µ, as was first predicted by Pier and Krolik (1992, 1993). The main basis for their prediction of anisotropy at shorter wavelengths was that many Seyfert 2's are Compton-thick, including NGC 1068 - which is completely opaque and for which N(H) 1025 according to the X-ray spectrum (e.g., Pounds and Vaughan 2006). In Galactic dusty gas this corresponds to A(V) ~ 1000. (See e.g., Maiolino et al 2001a, b for a quantitative correction, which however doesn't greatly affect the discussion of gas vs. dust columns below.) As an aside, I do think that dust-free atomic gas can contribute to the X-ray absorption in some cases (e.g., Risaliti et al 2011; Antonucci et al 2004), but the X-ray columns of Type 2 AGN are on average 100 × those of Type 1, so most of the column is connected with the Type 2 classification. Also in some cases, the dust column can be constrained to be similar to the very high X-ray columns. For example, Lutz et al (2000) used the lack of Pf- at 7.46µ to show that A(V) > 50 in NGC 1068.
Recall that one could pick spectral Type 1 (broad-line) radio loud Quasars and radio galaxies by some fairly isotropic AGN-related luminosity (e.g., lobe-power), and divide the two spectra to produce a spectrum of diminuation from anisotropy. We were able to do this for the z > 1 3CR (Fig. 4) because all of the radio galaxies seem to host hidden Quasars, but only out to ~ 15µ in the rest frame. The correction for anisotropy is around a factor of 10 in the near-IR, 1.5-2 on both sides of the silicate feature, and about three in the silicate feature, but varies significantly from object to object. As expected, it flattens out at ~ 1 at long wavelengths.
Figure 4. The infrared SEDs of z > 1 3CR Quasars and the matched radio galaxy composite, from Hönig et al (2011). Since these radio galaxies are all edge-on Quasars, this quotient spectrum measures the indication dependence at each wavelength.
Recall that torus shape was inferred from the high 14 typical polarization of the reflected light in hidden-Broad Line Region objects, which is generally perpendicular to the radio axis, meaning that photons can only escape the nucleus to scatter into the line of sight if they leave the nucleus in the polar directions.
The obscuring dusty tori invoked in the shadowing aspect of the unified model are active, not just passive, components. Modulo factors of order unity for geometry and dust cloud albedo, the tori will reprocess almost all of the incident Big Blue Bump/Broad Line Region luminosity into the infrared. Thus the ratio of re-emitted infrared emission to Big Blue Bump (and often considerable absorbed X-ray) emission tells us the approximate covering factor of the dusty gas which is idealized as a torus shape. 15
The covering fraction should agree with the fraction of Type 2 (hidden) nuclei in a sample which is selected by any isotropic AGN property. What are the results of this comparison? I'm crudely leaving out intermediate types, and the substantial minority of Seyfert 2s whose Type 1 nuclei are hidden by dust in the host galaxy plane (Keel 1980; Lawrence and Elvis 1982) rather than by a nuclear torus. Keel et al (1994) found 80 far-IR selected Seyfert 1s and 141 Seyfert 2s (plus some H II galaxies that they weeded out). So one expects a dust covering factor of ~ two-thirds in this parameter space. While the paper describes dust and gas covering factors as "broadly consistent" with the model, it would still be very valuable at this point to take advantage of this sample and actually measure covering factor for the Type 1s based on existing spectral energy distributions.
Figure 4 (kindly supplied by C. Leipski and S. Hoenig) shows the quotient SED for z > 1 individual Quasars and the radio galaxy composite, which however only covers up to 15 microns in the rest frame. 16 Again this plot purports to give us directly the anisotropy as a function of wavelength.
6 A few LINERs have strong broad lines, e.g., Filippenko and Halpern 1984. Many LINERs have very inconspicuous broad H components, but it isn't clear to me that they are strictly analogous to those in low-luminosity Seyferts. For example, we do not know whether they vary rapidly (L. Ho, 2011, private communication). Back.
7 It is proposed by Begelman (1999) that the weak fluxes at low frequencies result from free-free rather than inverse synchrotron absorption. Encouraging follow-up work can be found in Stawarz et al 2008 and Ostorero et al 2010. Back.
8 This material was reviewed recently by Giroletti 2008. Back.
9 In radio flux and redshift. Back.
10 As a reminder, unification of broad line and narrow line AGN by orientation of a toroidal nuclear obscurer is here called "shadowing unification." Ascription of superluminal motion and other relativisitic effects in subpopulations to orientation is called "beaming unification." Back.
11 Recall that thermal vs. non-thermal refers not to the radio emission itself, but to the presence of an energetically dominant optical/UV source thought to arise from an accretion disk. Back.
12 Although Maiolino et al 2007 didn't actually integrate over the SED and should therefore be viewed with much caution, it is interesting that they actually find some covering factors nominally greater than one for low-moderate luminosity objects. This was predicted qualitatively by Gaskell et al (2004), who argued that a preponderance of large grains leads to less apparent absorption (spectral curvature) in the optical/UV region than otherwise, and thus one may fall into the trap of inferring very low extinction. The inner torus is expected to lack small grains, most robustly perhaps on the grounds that the torus sublimation radius must be larger for small grains according to radiative equilibrium. The lack of small grains at the sublimation radius is supported by the data of Suganuma et al 2006: see discussion in Kishimoto et al 2007. Back.
13 Isotropic selection means selection on an isotropic AGN property, which avoids powerful biases in comparisons between classes. It is essential in order to produce intelligible results. Back.
14 A common error is to use the percent polarization of the continuum (after correcting for bulge or Elliptical starlight) as the scattering polarization: that is usually contaminated by hot stars, and only the broad lines themselves can be used to find, or more often place a lower limit on, the percent polarization of the scattered nuclear light (Antonucci 2002b). Back.
15 There is at least one obscured Quasar with a thin obscuring disk - the radio-loud, mini-MgII BAL OI287 (Goodrich and Miller 1988; Rudy and Schmidt 1988; Ulvestad and Antonucci 1988; Antonucci, Kinney, and Hurt 1993); however, even that one seems to have an upturn longward of 1µ at least according to the NED figures. Back.
16 Buchanan et al (2006) does this type of division for a 12µ (better than optical, but not ideal) selected sample of Seyfert galaxies. As expected, their anisotropy curve flattens at long wavelengths, but oddly not at a value of unity. Back.