Early investigations with hard X-ray satellites clearly revealed an excess of absorption in type 2 Seyferts, in agreement with the expectations from the unified model. However, such early studies could identify only very few Compton thick sources, suggesting that the latter is a very rare class of objects. Yet, later, deeper surveys, adopting careful selection criteria and exploiting a wider energy range, discovered a much larger fraction of Compton thick AGNs [Maiolino et al. (1998), Bassani et al. (1999)]. By extracting a subsample selected in [OIII]5007 flux, assumed to trace the intrinsic AGN flux, [Risaliti et al. (1999)] could determine a first, unbiased distribution of NH among Seyfert nuclei (Fig. 1). Among optically obscured Seyfert nuclei the distribution of NH is nearly flat, and Compton thick Seyfert 2s (NH > 1024 cm-2) are found to be as numerous as Compton thin ones. Generally there is a good correspondence between optical classification and X-ray absorption: Sy1s tend to have little or no absorption, "strict" type 2 Seyferts tend to be heavily absorbed (NH > 1023 cm-2), while intermediate type 1.8 - 1.9 Seyferts are absorbed by intermediate NH (~ 1022 - 1023 cm-2).
The column density distribution has been extracted also for other samples of AGNs selected in different ways. [Markwardt et al. (2005)] and [Bassani et al. (2006)] give the NH distribution of AGNs selected in the 10 - 100 keV energy range. These samples are less biased against obscured AGNs, with respect to samples selected at lower energies, but they are still biased against Compton thick AGNs. Indeed, as discussed above, AGNs with NH > 1025 cm-2 are absorbed at all energies, while partially Compton thick AGNs with 1024 < NH < 1025 cm-2 do show a transmitted component at E > 10 keV, but still significantly absorbed with respect to Compton thin AGNs. Indeed, the samples presented by [Markwardt et al. (2005)] and [Bassani et al. (2006)] show the absence of AGNs with NH > 1025 cm-2 and a paucity of AGNs with 1024 < NH < 1025 cm-2.
Radio emission is another selection which should be free of absorption biases. Indeed, early X-ray studies of radio-loud AGNs revealed a large fraction of obscured AGNs [Sambruna et al. (1999)]. The absorbing column density is also found to anti-correlate with the radio core dominance parameter R (a measure of the jet orientation), in agreement with the expectations from the unified model [Grandi et al. (2006)]. However, there is a puzzling shortage of Compton thick AGNs, which is confirmed also in more recent studies [Evans et al. (2006), Hardcastle et al. (2006)]. A possible explanation is that the radio jet contributes (or dominates) the X-ray luminosity, which is therefore inefficiently obscured by a compact, circumnuclear medium.
Recently, [Cappi et al. (2006)] measured the column density with XMM in a distance limited sample of Seyferts (pre-selected through optical spectroscopy, [Ho et al. (1997)]). Within the statistical uncertainties, the resulting NH distribution is similar to that obtained in previous optically selected samples.
One of the major limitations of the previous surveys is that in most of them AGNs were pre-selected to have a Seyfert-like optical spectrum. However, hard X-ray observations [Vignati et al. (1999), Guainazzi et al. (2000), Della Ceca et al. (2002)] have revealed the presence of heavily obscured, relatively luminous Seyfert nuclei (L2-10 keV > 1042 erg s-1) in galaxies optically classified as starburst (HII) or LINER (here with the latter term we refer to emission due to shocks in starburst superwinds, not nuclear LINERs). Similar results have been obtained through near- and mid-IR spectroscopy [Imanishi et al. (2006), Risaliti et al. (2006)]. [Maiolino et al. (2003)] investigated the statistical properties of such optically elusive Seyfert nuclei, and found that most of them are Compton thick. Therefore, the actual fraction of Compton thick AGNs is higher than inferred in samples where objects are pre-selected through the optical spectrum. Note that evidence for relatively powerful Seyfert nuclei hosted in galaxies which are apparently normal in the optical has also been found in high redshift surveys (e.g. [Comastri et al. (2002), Szokoly et al. (2004), Barger et al. (2005), Cocchia et al. (2007)]). However, in several of these cases the optical mis-classification may simply be due to dilution of the nuclear light by the host galaxy [Moran et al. (2002)] or to inappropriate (rest-frame) spectral coverage [Severgnini et al. (2003)].