DISCOVERY OF STRONG IRON Kα EMITTING COMPTON THICK QUASARS AT z = 2.5 AND 2.9

The observed (obscured) hard X-ray luminosities are L_2–10 keV = 10^42.5 and 10^42.8 erg s⁻¹ for BzK 4892 and BzK 8608, respectively. Given the large N_H implied by the detection of high EW iron Kα lines, the X-ray emission is optically thick and we cannot directly estimate the intrinsic L_2–10 keV using the X-ray data. In order to estimate the intrinsic luminosities of the AGN we used a variety of independent methods.

Following Iwasawa et al. (2005), a lower limit to the X-ray luminosity can be obtained in the expectation that in typical conditions the luminosity of the iron Kα line is at most 3% of the 2–10 keV luminosity. For BzK 4892 L_Kα = 1.25 × 10⁴² erg s⁻¹ would imply an intrinsic 2–10 keV luminosity of 4×10⁴³ erg s⁻¹. The relation L_Kα/L_2–10 keV of Levenson et al. (2006) implies a higher luminosity, L$_{2\hbox{--}10\, \rm keV}\sim 10^{44.8}$ erg s⁻¹. An alternative estimate is obtained from the 6.0 μm luminosity. The 24 μm flux of 591 μJy implies L_6.0 μm = 10^45.5 erg s⁻¹ (the mid-IR spectral energy distribution (SED) of Mrk 231 is used for the small K-correction). While 24 μm at z = 2.578 might be contaminated by polycyclic aromatic hydrocarbon features, we find similar results if we use the 16 μm flux of 209 μJy (Teplitz et al. 2011) for which no contamination is expected (4.5 μm rest frame). This would suggest L_2–10 keV = 10^45±0.5 erg s⁻¹ using the Lutz et al. (2004) correlation and a lower L_2–10 keV = 10^44.4±0.5 erg s⁻¹ using Lanzuisi et al. (2009). A third estimate can be derived from the luminosity of the UV–optical emission lines. The VLT/FORS1-2 optical spectrum of BzK 4892 shows prominent Lyα, C iv, He ii, and C iii] narrow emission lines (Szokoly et al. 2004; Vanzella et al. 2006). Using the observed ratios between the line and the intrinsic 2–10 keV emission from Mulchaey et al. (1994), we infer L_2–10 keV = 10^44.5±0.5 erg s⁻¹. Using the Netzer et al. (2006) conversion for UV–optical line luminosities would give larger X-ray luminosities by ∼0.5 dex or L_2–10 keV = 10^45±0.5 erg s⁻¹. All in all, it appears that the 2–10 keV luminosity of BzK 4892 is in the QSO range, well in excess of 10⁴⁴ erg s⁻¹ and more likely of 10^44.5–10⁴⁵ erg s⁻¹. The implied column density is at the level of N_H ∼ 10^24.5–10²⁵ cm⁻².

For BzK 8608, we measure L(Kα) =2.5 × 10⁴² erg s⁻¹. This converts to L_2–10 keV ∼ 10^43.9 erg s⁻¹ (Iwasawa et al. 2005) or ∼10⁴⁵ erg s⁻¹ adopting Levenson et al. (2006). This galaxy is about 10 times fainter in the mid-IR than BzK 4892, which implies a lower L_6.0 μm. Using the Lutz et al. (2004) relation, this in turn suggests L_2–10 keV = 10⁴⁴ erg s⁻¹. We cannot obtain an independent estimate from the UV–optical emission lines given that no lines were detected in the VIMOS spectrum. The intrinsic 2–10 keV luminosity of BzK 8608 is likely ≈10⁴⁴ erg s⁻¹. The implied column density is likely in the range N_H ∼ (1–5) × 10²⁴ cm⁻², hence this object might be a borderline CT AGN. We note that both sources have estimated intrinsic 2–10 keV luminosities in the QSO regime.

The ratio of the bolometric to the 2–10 keV luminosity, L_bol/L_2–10 keV, is typically of the order 30–50 for quasars (Marconi et al. 2004). Assuming such a bolometric correction, we estimate L_bol ∼ 10^46±0.5 erg s⁻¹ for BzK 4892 and ∼10^45.5±0.5 erg s⁻¹ for BzK 8608. These very high luminosities, if expressed in solar units, correspond to L_bol ∼ 10^12.4±0.5L_☉ and ∼10^11.9±0.5L_☉. When powered by star formation, similar luminosities require star formation rates (SFRs) in the host galaxies at the level of 100–1000 M_☉ yr⁻¹. It is interesting thus to compare these estimates to the inferred SFR for the two BzK galaxies, in order to compare the relative AGN and SFR contributions.

From the stellar SED (UV to the near-IR rest frame; Figure 3), we estimate SFR ∼ 70 M_☉ yr⁻¹ for BzK 4892, corrected for dust reddening based on the observed UV slope (see D07 for more details). It is quite possible though that also the stellar UV emission might be optically thick. If we interpret its mid-IR emission (591 μJy at 24 μm) as due to star formation this would imply a whopping SFR ∼10⁴ M_☉ yr⁻¹, demonstrating that the mid-IR is likely completely dominated by the AGN (this object is among the most extreme mid-IR excess galaxies in the sample of D07). The SED fit with a star-forming galaxy template (Figure 3) shows an excess emission already at 5.8 μm (observed frame) probably due to the AGN contribution. The SED in the IRAC bands is indeed showing a steep power law, extending all the way to 24 μm. Due to its faint optical counterpart (i ∼ 25 AB), BzK 4892 has high mid-infrared to optical flux ratio, F(24)/F(R) ∼ 2000, and therefore is also classified as a dust-obscured galaxy (Dey et al. 2008). It is an extreme object also in the Fiore et al. (2008) sample. Alonso-Herrero et al. (2006) report a 70 μm flux of 3.3 ± 1.8 mJy for this galaxy, which is also likely affected by the AGN emission. On the other hand, the galaxy is seen at 1.4 GHz with a flux of 80 μJy in the Very Large Array (VLA) data of Miller et al. (2008). If this measured flux is due to star formation, the radio–IR correlation would imply SFR ∼1100 M_☉ yr⁻¹. Inspecting the publicly available Apex+LABOCA 870 μm map of GOODS-S (Weiss et al. 2009), we find a 3σ signal at the position of BzK 4892 of 3.3 mJy, which would also convert into a similar SFR, ∼500 M_☉ yr⁻¹.

Zoom In Zoom Out Reset image size

Even for the most luminous quasars it is generally found that the far-IR emission in the submillimeter bands is due to star formation. If this is the case also for BzK 4892, our results suggest that the galaxy is witnessing also very powerful star formation activity at the level of 500–1000 M_☉ yr⁻¹. Only 5%–10% of this is seen directly in the UV, implying that also the UV emission from stars is optically thick, similarly to local ultraluminous infrared galaxies (e.g., Goldader et al. 2002; da Cunha et al. 2010) and to what is expected to be found in major mergers. However, we note that the radio and 870 μm derived SFRs would formally imply L_bol ∼ 10⁴⁶–10^46.5 erg s⁻¹, comparable to what inferred for the obscured AGN. Hence, the AGN is likely contributing an important fraction of the total L_bol.

The morphology of the galaxy from HST+ACS imaging is suggestive of a merger, showing two distinct clumps in the UV in the rest frame. When smoothing, after excluding the two bright knots, we detect significant faint low-level emission in the summed i+z band, extended over a diameter of about 075 (6 kpc). Overall, this is consistent with the size of a massive galaxy at z ∼2, and we cannot exclude that the two UV knots might be just luminous H ii regions inside a big disk galaxy.

The photometric information is of lower quality for BzK 8608 due to its overall faintness. The SFR inferred from the UV is also 70 $M_{\odot } \,\rm yr^{-1}$. This object is also a mid-IR excess source in D07 sample (by a less extreme factor of six), with S₂₄ = 45 μJy (∼1/10 of BzK 4892), although it is not a dust obscured galaxy. The SED (Figure 3) might also be consistent with pure stellar emission, but a deviation from the star formation template is observed at the 8μm band. This galaxy is not detected at 70 μm and 870 μm, suggesting a lower AGN luminosity and/or SFR. Inspecting the VLA 1.4 GHz data (Miller et al. 2008), we detect a faint 3σ source at its position with a flux of 17.5 ± 6.5 μJy. This corresponds to a luminosity 3 × 10¹² L_☉ and SFR ∼300 M_☉ yr⁻¹. If the radio signal is real and not due to AGN, also in this case the stellar emission appears to be heavily obscured, i.e., optically thick in the UV, as expected for very active starbursts and mergers. The HST+ACS imaging of this galaxy (Figure 3) is not particularly telling.

For both galaxies we derive similar estimates of the stellar masses, at the level of 5 × 10¹⁰ M_☉. Our results thus support the picture that the luminous, CT AGNs that we discovered are hosted by fairly massive galaxies at z ∼ 2.5–3, which at the same time also host vigorous star formation activity heavily obscured by dust. This is relevant in the AGN–host galaxy co-evolution scenario, in which a phase of rapid, heavily obscured black hole growth accompanied by intense obscured star formation is predicted (Silk & Rees 1998; Fabian 1999; Granato et al. 2004).

We derive a crude value of the volume density of CT AGN with high EW iron Kα emission, based on our two detections. We conservatively use the full redshift range explored by the BzK selection (z =1.4–3), finding a space density of 3 × 10⁻⁶ Mpc⁻³. This is consistent with the predictions of the Gilli et al. (2007) XRB synthesis model, for CT AGN with L$_{2\hbox{--}10 \, \rm keV}>10^{44}$ erg s⁻¹. However, we note that our sampling of L$_{2\hbox{--}10\, \rm keV}>10^{44}$ erg s⁻¹ AGNs with CT absorption at 1.4 < z < 3 might be substantially incomplete due to selection effects, and the luminosity of at least one of the BzK galaxies in this Letter might be substantially higher than the adopted 10⁴⁴ erg s⁻¹ limit.

We emphasize that the sources presented here have X-ray spectra that are the high-luminosity analogous of those of local prototype CT AGNs. This discovery represents one of the first clear-cut evidence for this class of objects at high redshift. Our results confirm that indeed there is a population of CT QSO among massive, star-forming, dust obscured galaxies, as suggested by several studies (D07; Fiore et al. 2008; Treister et al. 2010; Alexander et al. 2008; Lanzuisi et al. 2009) and as predicted by models of merger-driven AGN/galaxy co-evolution (Silk & Rees 1998; Fabian 1999; Di Matteo et al. 2005). Indeed, these luminous CT quasars appear to be hosted by galaxies with optically thick UV emission, which are a relative small minority among BzK s (Daddi et al. 2007a) and imply the presence of a dense starburst usually connected with mergers. This seems to agree with evolutionary scenarios, where mergers induce both rapid accretion onto supermassive black hole and intense star formation activity.

The two sources presented here have high intrinsic X-ray and bolometric luminosity, and bright iron Kα line, allowing their identification. They might represent just the tip of the iceberg of a much larger population of highly obscured AGNs with somewhat lower intrinsic luminosity, whose iron lines are still remaining hardly detectable even in the deepest hard X-ray surveys. For example, the Gilli et al. (2007) model predicts a ∼10× higher space density of CT sources with L_2–10 keV>10⁴³ erg s⁻¹. Detecting iron lines in galaxies with such luminosities might be challenging with current facilities, even if they have similar EW as in BzK 4892 and BzK 8608.