The understanding of AGN activity, growth and evolution can be achieved by combining shallow large area sky surveys to very deep pencil beam surveys [15, 25, 26]. Since most of the AGNs are absorbed (e.g. ) the best method to select a representative sample of the entire AGN population is to use an energy band which is the least affected by attenuation due to intrinsic absorption. As of now, this is the 2-10 keV energy band because above this energy no deep surveys exist. Recently,  compiled the largest set of X-ray surveys in the 2-10 keV band. According to Hasinger 2008, to best determine AGN evolution one should achieve a good sampling of the luminosity-redshift plane. From Fig. 2 (left panel) it is evident that while Chandra and XMM-Newton are able to perform a good sampling of the high-luminosity high-redshift part of the plane, the sampling remain sparse at low luminosities and low redshifts lacking a large sample able to constrain the properties of AGNs in the local Universe. Conversely, the > 10 keV range lacks deep surveys (able to sample fluxes of 10-13 - 10-14 erg cm-2 s-1), but on the other hand, the IBIS and BAT surveys (e.g. [28, 29, 30, 31, 32]) are detecting hundreds of AGNs in the local Universe. This is clearly shown in the right panel of Fig. 2. Moreover, as Fig. 3 shows, the > 15 keV band is the best one for an unbiased (with respect to absorption) search of AGNs. Indeed, for absorbing column densities of 1023 and 1024 atoms cm-2 the fraction of nuclear flux which is detected in the 2-10 keV band is 50% and 7% respectively. The 15-55 keV band, as an example, is unbiased up to ~ 1024 atoms cm-2. This is particularly important if the goal of the survey is the detection and understanding of the still missing Compton-thick source population.
Figure 2. Left Panel: Meta-sample compiled by  using the available X-ray surveys in the 2-10 keV energy band. Right Panel: BAT AGN sample (|b| > 15°) from ~ 3 years of all-sky survey in the 15-55 keV band. INTEGRAL samples luminosities and redshifts similar to BAT (e.g. [28, 29, 30]). A typical luminosity of 1044 erg s-1 in the 15-55 keV is equivalent to ~ 1.2 × 1044 erg s-1 in the 2-10 keV energy band.
Figure 3. Ratio of observed flux (scattered) to intrinsic nuclear flux as a function of column density along the line sight for the 2-10 keV (blue) and the 15-55 keV (red) bands. The solid and dashed lines are for the case of an AGN at redshift 0 and 1 respectively. Compton scattering as well as photoelectric absorption has been taken into account using the model of . The nuclear emission has been modeled as a power law with photon index of 2.0.
3.1. The role of absorption
The AGN unified models predict that ~ 3/4 of all AGNs should be absorbed by intervening matter along the line of sight. This prediction is based on the observations of ionization cones, in O [III] light, with apex at the nucleus (e.g. ). However, this is clearly not the case for the X-ray selected AGNs in the local Universe. Indeed, as Tab. 1 shows the fraction of absorbed AGNs is significantly lower than 75% and closer to ~ 50%. Moreover, it is now well established that the fraction of obscured AGNs is not constant, but that it evolves with luminosity (e.g. [35, 10, 27]). This yields that low-luminosity AGNs are surrounded by an obscuring medium which covers a large solid angle (80% of the sky as seen from the nucleus) while high-luminosity AGNs have a smaller covering factor and thus must be able to clean their environment. This effect can be interpreted in the framework of the receding torus model (e.g. ) as the increase of the dust sublimation radius as a function of AGN luminosity. This would produce a decrease of the solid angle which the dusty torus covers around the nuclear source, explaining, qualitatively, the observed trend with luminosity. However, the receding torus model fails to explain the exact trend of the absorbed fraction with luminosity . Another option is that the torus is not formed by a continuous distribution of dust particles, but that it comprises of several optically thick clouds orbiting around the central source (e.g. [38, 39]). In this framework, the torus consists of a large number of small, self-gravitating, dusty molecular clouds which form a clumpy medium (e.g. ). This model explains the observed trend of the absorbed fraction with luminosity very well and this effect might be produced either by a change in the torus width or in the number of clouds or in both (see e.g.  for details). In any case, the anti-correlation of luminosity and absorption has been interpreted as one of the main evidences for the breakdown of the AGN unified model.
3.2. Compton-thick AGNs in the local Universe
Compton-thick AGNs may come in two flavors: transmission-dominated and reflection-dominated sources (see e.g. ). For trasmission-dominated objects, where a fraction of the nuclear emission (piercing through the Compton-thick torus) is detected, the spectral shape varies accordingly to the absorbing column density and the photoelectric cut-off moves progressively to higher energy. For increasing column densities (24 < LogNH < 25), the absorbing material becomes more and more efficient in suppressing the radiation below 10 keV and Compton recoil makes steeper (down-scattering) the > 10 keV part of the spectrum. Transmission-dominated sources (also called heavily Compton-thick objects) usually show a broad iron K line over a flat continuum and the ratio of the observed to the nuclear flux can be as low as a few % (e.g. [41, 42]). Studies of the local Universe in the Optical showed that Compton-thick AGNs should be as numerous as normally obscured ones (e.g. [43, 44]), however up to now only a handful bona-fide Compton-thick objects were detected [14, 45].
Results from the BAT and INTEGRAL surveys allow to shed some light on the overall picture. Tab. 1 shows an up to date status of hard X-ray surveys. It is apparent that at the fluxes currently sampled, the fraction of Compton-thick objects is bound to be < 20% and likely closer to 10%. Strictly speaking this represents a lower limit on the real fraction since, as I showed in Fig. 3, even above > 15 keV not all the source flux can be detected. This low number of Compton-thick AGNs seems, however, to be in agreement with the prediction of population synthesis models which require a substantial contribution from Compton-thick objects to explain the peak of the CXB emission (e.g. ). In other words, Compton-thick AGNs seem to exist, but even INTEGRAL and BAT are not sensitive enough to detect many of them in the local Universe. If these sources evolve similarly to the other classes of AGNs (e.g. [15, 25]), then their number density is expected to raise quickly with redshift peaking around z 1. In this respect, instruments able to sample the 10-13 - 10-14 erg cm-2 s-1 fluxes above 15 keV have more chances to detect a large number of heavily absorbed objects. In this case the k-correction plays also in their favour allowing to sample the source spectrum at an higher energy as Fig. 3 shows.
|Reference||Completeness||% Absorbed||% C-thick||Band||Instrument|
|Markwardt et al., 2005||95%||~ 64%||~ 10%||15-200||Swift/BAT|
|Beckmann et al., 2006||100%||~ 64%||~ 10%||20-40||INTEGRAL|
|Bassani et al., 2006||77%||~ 65%||~ 14%||20-100||INTEGRAL|
|Sazonov et al., 2007||90%||~ 50%||~ 10-15%||17-60||INTEGRAL|
|Ajello et al., 2008a||100%||~ 55%||< 20%||14-170||Swift/BAT|
|Tueller et al., 2008||100%||~ 50%||~ 5%||14-195||Swift/BAT|
|Paltani et al., 20082||100%||~ 60%||< 24%||20-60||INTEGRAL|
|Della Ceca et al., 2008||97%||~ 57%||0||4.5-7.5||XMM-Newton|
|1 The fraction of Compton-thick AGNs comes from .|
|2 Since the Paltani et al. sample may contain a fraction of spurious sources, we restricted their sample to a limiting significance of 6. Above this threshold all sources are identified (see Tab. 2 in ).|
Another interesting point is that the current hard X-ray surveys have not detected any new Compton-thick AGNs of the trasmission-dominated class. They have found many objects with very high column densities which are almost Compton-thick (e.g. NH 1023.8-24.0 atoms cm-2), but not completely so. Quoting , who analyzed the Tueller et al. (2008) sample,: If we take the Compton-thick definition to apply to sources whose column densities are > 1.4 × 1024 cm-2, none of the BAT-detected sources are Compton-thick. This is also due to the fact that when fitting broad-band X-ray data (e.g. 0.1-200 keV), it is difficult (with the low signal-to-noise ratio at high energy) to discriminate between alternative interpretations. Indeed, most of the time reflection, partial absorbers and cut-off components remain degenerate.
However, a major discovery was achieved by INTEGRAL and Swift and this is the detection of "buried" super-massive black holes [42, 41]. These are AGNs for which the reflection component dominates over the transmitted one. If the ratio of the normalizations of these two components is interpreted as the solid angle covered by the reflecting medium (i.e. R = / 2) then this value exceeds 1. This would imply that part of the nuclear emission is blocked by nonuniform material along the line of sight even above 10 keV. The extremely low scattering efficiency which has been found for these objects (e.g. 0.5-2% [42, 41]) implies a torus half-opening angle of < 20° in contrast to the classical 30° - 40° expected in the framework of AGN unified model (e.g. ). The large equivalent widths (~ 1 keV, see Fig. 4) of the iron K line and the absence or weakness of [O III] lines confirm this interpretation.
Figure 4. Example of Suzaku X-ray spectra of two buried AGNs. Note the dominating reflection component (which is curved at high energy) and the strong iron line. For both objects (SWIFT J0601.9-8636 and NGC 5728 left and right respectively) the scattering efficiency is very low (0.2-2%). Adapted from  (left) and  (right).
In order to recognize these objects as such, one needs sensitive and broad-band X-ray coverage. Thus it might well be that a few (or many) of these objects are hiding among the INTEGRAL and Swift/BAT survey sources. According to  these objects might be a relevant fraction, ~ 25%, of the total AGN population of the local Universe. The question whether this is the missing source population (i.e. the one needed to explain the totality of the CXB emission at its peak) is a difficult one to ask. The answer might come only from a sample of buried AGN large enough to derive its luminosity function. Until that time  derived the first luminosity function of Compton-thick AGNs as a difference between the Optical luminosity function of AGNs derived by  and their X-ray luminosity function of AGNs. They find that the space density of Compton-thick objects is twice the density of Compton-thin AGNs with an hard upper limit to four times this density given by the limit imposed by the local black hole mass density .
3.3. Compton-thick AGNs in the high-z Universe
As already seen, selecting Compton-thick objects is extremely difficult even in hard X-rays. There is however a possibility to recover obscured AGNs thanks to the reprocessing of the AGN UV emission in the infrared (IR) band. Thus, selecting bright mid-IR sources which are faint in the Optical, and thus likely to be obscured, might be rewarding (e.g. [52, 53]). Such window of opportunity has been recently opened by the Spitzer telescope. Indeed, using Spitzer  has found that bright mid-IR sources (F(24 µm) > 0.3 mJy) with faint optical and near-IR counterparts are likely to be highly obscured type 2 QSOs. This approach has been extensively applied to Spitzer observations of the Chandra Deep Fields [54, 55]. In these fields, the X-ray coverage is deep and therefore it allows to probe the nature of these mid-IR "excesses". A few of these sources have a direct X-ray detection which might indicate that they are obscured sources 2. The average X-ray properties of a class of sources can be studied using the stacking technique (i.e. summing the signal of different sources). The stacked X-ray spectrum (shown in the left panel of Fig. 5) of mid-IR excesses is compatible with the one of a Compton-thick AGNs [54, 55]. Since most of these sources have either a photometric or a measured redshift, it becomes possible to estimate their volume density (see right panel of Fig. 5). This, at the average redshift of these samples (z ~ 2), turns out to be of the same order of that of X-ray detected AGNs. This would imply that the mid-IR selection is a powerful technique to recover the population of Compton-thick objects which is not detected even in the deepest X-ray surveys.
Figure 5. Left: Stacked X-ray spectrum of mid-IR excesses detected in the Chandra deep fields. The spectrum is compatible with the spectrum of a Compton-thick AGN. Adapted from . Right:Volume density of mid-IR selected sources in the COSMOS field . The solid lines are predictions from the model of .
However, the complex selection criteria leave some uncertainty about the true nature of these sources. Indeed, mid-IR excesses might be produced by powerful sturburst galaxies (e.g. ). In order to remove all uncertainties one would need to obtain an IR and an X-ray spectra for all these sources. Given the large redshift and the number of sources this is not always feasible. At least in the case of HDF-OMD49, a Spitzer object at z = 2.21, the IR and X-ray spectra show that this source is very likely a Compton-thick AGN . However, recently , using a compbination of IR bands and mid-IR spectroscopy, determined that the majority of the dust obscured galaxies 3 are dominated by star formation rather than AGN activity. Thus, while mid-IR selection might represent a powerful tool to recover Compton-thick AGNs at large redshifts, it might be that the contamination due to powerful starburst galaxies is at the moment affecting the estimates of the space density of highly-obscured AGNs.
2 Given the signal-to-noise ratios it is very difficult to say whether these sources are simply obscured (NH > 1022 cm-2) or are Compton-thick (NH > 1024 cm-2). Back.
3 Dust obscured galaxies are selected in such a way that 90% of them would meet the selection criteria of . Back.