To characterize the quasars in our sample, we used the deep X-ray Chandra data available in the Chandra Deep Field North (CDF-N; 2 Ms; Alexander et al., 2003) and Chandra Deep Field South (CDF-S 4 Ms; Xue et al., 2011). For details on the data reduction we refer to Alexander et al. (2003), Luo et al. (2008), and Xue et al. (2011). Of our 33 quasars, 24 (~73%) are detected in the X-ray band, while 9 (~27%) remain undetected, despite being intrinsically very luminous in the MIR band. These sources are candidates to be the most heavily obscured, CT AGN.

For the sources that are detected in the X-rays, we extracted the spectra using ACIS Extract (AE; Broos et al., 2010, 2012) and analyzed them using a simple absorbed power-law model (including Galactic and intrinsic absorption) to constrain the amount of N_H. In Figure 1 we show the N_H distribution of the sample. We find that the majority of these quasars (16/24; ~67%) are obscured by columns of NH>1022 cm⁻², of which more than half (9/16) are heavily obscured (NH>2×1023 cm⁻²). Amongst these heavily-obscured sources, we identified six of them as CT quasars from the X-ray spectral analysis, using spectral models appropriate for heavily-obscured sources, such as, PLCABS (Yaqoob, 1997) and TORUS (Brightman and Nandra, 2011). The fraction of obscured quasars in our sample reaches ~76%, and 54% of heavily-obscured quasars, if we include the X-ray-undetected sources, assuming these are the most heavily CT ones. Indeed, the comparison between the intrinsic luminosity at 6 μm and the X-ray luminosity upper limit of these sources suggests that the X-ray emission is heavily suppressed compared to the intrinsic L_X − L_{6 μm} relation found for AGN (e.g., Lutz et al., 2004; Fiore et al., 2009; Gandhi et al., 2009), making them very good candidates to be heavily CT quasars.

We note that amongst the X-ray undetected quasars in the sample, there is one source, #28 (see Table 1 from Del Moro et al., 2016), that is now detected in the 7 Ms CDF-S catalog (XID 28; Luo et al., 2017). This source has a very flat effective photon index of Γ < 0.93 (compared to the typical Γ ≈ 1.8 for unabsorbed AGN) and an extremely low rest-frame X-ray luminosity (L0.5-7keV≈1.5×1041 erg s⁻¹, uncorrected for absorption) compared to its intrinsic luminosity measured in the MIR band (log ν L_{6 μm} = 45.97 erg s⁻¹), consistent with the upper limit reported in our analysis (Del Moro et al., 2016)¹, supporting our assumption that the source might be a heavily obscured, CT quasar.

These results suggest that there is a large population of heavily-obscured, intrinsically luminous quasars at high redshift, which are very elusive even for deep X-ray surveys. These sources might constitute a special phase of the BH-galaxy evolution, where the actively growing BH is embedded in large amounts of gas and dust, possibly as a result of a recent merger.

To study the characteristics of the host galaxies of these MIR-luminous quasars, we investigated their star-formation rates (SFRs) and their morphologies, in particular the disturbance or distortion of their morphology, as an indication of galaxy mergers and interactions. We derived the SFR of each galaxy from its far-infrared (FIR) luminosity (or upper limit) resulting from the SED decomposition in the IR band (see section 2 and Del Moro et al., 2016), assuming a Salpeter initial mass function (Salpeter, 1955) and the relation from Kennicutt (1998). We also calculated the average SFR separately for the unobscured/moderately obscured quasars (NH<2×1023 cm⁻²) and for the heavily-obscured quasars (NH>2×1023 cm⁻²) in three different redshift bins (see Figure 2). To estimate the average SFRs accounting for the upper limits, we used the Kaplan-Meier (KM) product limit estimator (Kaplan and Meier, 1958), a non-parametric maximum-likelihood estimator of the distribution function (see Stanley et al., 2015, for details on the method).

We find that the average SFR increases with redshift, in general agreement with the SFR main sequence of galaxies (Figure 2). Moreover, although the heavily-obscured sources seem to have a slightly enhanced SFR compared to the unobscured/moderately obscured ones (especially at z ≈ 2), these differences are not statistically significant, and therefore we find no clear dependence of the amount of SF in the galaxy with the X-ray obscuration of the quasar. This suggests that the obscuration in the X-ray band is likely confined in the nuclear regions and not related to the presence of gas on larger scales (e.g., Rosario et al., 2012; Rovilos et al., 2012).

Using the high-resolution optical HST images available in the GOODS-N and GOODS-S fields, as part of the GOODS and CANDELS projects (Giavalisco et al., 2004; Grogin et al., 2011), we visually inspected the morphology of these galaxies to identify signs of distortions or disturbances, which would indicate a recent galaxy merger/interaction event. We adopted a similar classification scheme to that used by Kocevski et al. (2012); see Del Moro et al. (2016, for details), separating the sources into “disturbed” and “undisturbed”. In Figure 3 we show the fraction of sources having disturbed and undisturbed morphologies over the total, dividing them again into unobscured/moderately obscured (blue squares) and heavily obscured (magenta circles), as in Figure 2. We find that a relatively high fraction of our sources shows signs of distortions/interactions (≈40%), higher than those typically found at low redshift (~15–20% at z < 1; e.g., Cisternas et al., 2011). This is in agreement with the trend of increasing major-merger fraction with redshift seen in previous works (e.g., Conselice et al., 2003; Treister and Urry, 2006; Kartaltepe et al., 2007). We find that, on average, the most-heavily obscured quasars tend to have more disturbed morphologies than the unobscured/moderately obscured ones (≈53 vs. ≈20%, respectively); although the errors on these fractions are large due to the small number of sources in our sample, the difference between the two quasar populations is significant at the 90% confidence level (Fisher exact probability test: P = 0.087). This trend is seen also in other studies (e.g., Kocevski et al., 2015; Ricci et al., 2017). We note, however, that these studies investigated samples in different redshift and/or luminosity ranges compared to ours, and therefore the actual fractions of sources with disturbed morphologies are not directly comparable. The smaller fraction of disturbed systems we found for the unobscured/moderately-obscured quasars could be interpreted within the SMBH-galaxy evolutionary models, as the unobscured quasars would represent a later stage of the evolution compared to the heavily-obscured sources and the distortion features due to mergers or interactions might have faded by the time these quasars are observed as unobscured, given the relaxation time of a galaxy is typically ~200–400 Myr (e.g., Lotz et al., 2010). On the other hand, for the most heavily-obscured quasars, which might represent a younger stage of evolution after a merger in this scenario, the signatures of the recent interactions are still evident in their hosts.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.