Until a few years ago the existence of obscured QSOs (QSO2s) was under question. QSO2s are difficult to find because both absorbed and rare. However, recently a large number of QSO2s have been found in by systemic surveys over large sky areas, in the optical [Zakamska et al. (2003), Ptak et al. (2006)], in the X-rays [Norman et al. (2002), Fiore et al. (2003), Barger et al. (2005), Silverman et al. (2005), Maccacaro et al. (2004), Maiolino et al. (2006)], in the radio [Donley et al. (2005), Belsole et al. (2006)] and in the infrared [Martíxnez-Sansigre et al. (2005), Polletta et al. (2006), Alonso-Herrero et al. (2006), Franceschini et al. (2005)].
The large number of newly discovered QSO2s has allowed the investigation of the fraction of obscured AGNs as a function of luminosity. By using the results from hard X-ray surveys, various authors have found evidence for a decreasing fraction of obscured AGNs with increasing luminosity [Ueda et al. (2003), La Franca et al. (2005), Barger et al. (2005), Akylas et al. (2006)], although this result has been questioned [Dwelly & Page(2006)]. The same trend was found by [Simpson(2005)] among optically selected AGNs. If confirmed, these results can be interpreted within the scenario of the so-called "receding torus" [Lawrence(1991)], where the dependence of the dust sublimation radius with luminosity causes the covering factor of the absorbing medium to decrease with luminosity. Alternatively, [Lamastra et al. (2006)] suggested that the dependence of the covering factor with luminosity is an indirect consequence of the gravitational effects of the black holes, which is on average more massive in more luminous AGNs, because of selection effects.
One should keep in mind that, at least in hard X-ray surveys, the census is limited to the Compton thin sources. Indeed, the faintness of Compton thick AGNs even in the hard X-rays prevents their detection at cosmological distances (except for a tail of the population, [Tozzi et al. (2006)]). To infer the fraction of Compton thick sources at high redshift we have to rely on other, indirect indicators. One possibility is to exploit the shape of the hard X-ray background. [Gilli et al. (2006)] show that the population of (Compton thin) AGNs that resolve the X-ray background at 2 - 10 keV fall short to account for the peak of the latter at 30 keV. The additional contribution by a population of Compton thick AGNs with 1024 < NH < 1025 cm-2 is required to match the shape and intensity of the X-ray background at energies higher than 10 keV. The required proportion of Compton thick AGNs with 1024 < NH < 1025 cm-2 relative to Compton thin AGNs must be 1:2, i.e. as observed in local AGNs. Note however, that the X-ray background is insensitive to the population of totally Compton thick AGNs with NH > 1025 cm-2 (since their emission is totally suppressed at any energies), which therefore remains unconstrained.
An alternative method to identify Compton thick AGNs at high redshift is by means of mid-IR data. Indeed, recent Spitzer observations have discovered a large population of high-z AGNs (identified through a mid-IR, AGN-like excess) that do not have hard-X counterpart even in deep X-ray observations, and therefore are likely Compton thick AGNs [Alonso-Herrero et al. (2006), Polletta et al. (2006), Donley et al. (2005)]. Many of the Spitzer studies on high-z AGNs are still ongoing, therefore this field is currently in continuous evolution. However, results published so far suggest that the Compton thick AGNs at high-z (including those with NH > 1025 cm-2) are as numerous as Compton thin AGNs, i.e. matching the same relative fractions observed in the local universe.