The XMM–Newton Bright Survey sample of absorbed quasars: X-ray and accretion properties

Abstract

INTRODUCTION

Active galactic nuclei (AGN) are believed to be powered by accretion of matter on to a supermassive black hole (SMBH, M_BH > 10⁶ M_⊙; Salpeter 1964; Lynden-Bell 1969). The discovery of quiescent ‘non-active’ SMBHs in the nuclei of nearby bulge galaxies, along with the presence of scaling relations between the central black hole (BH) mass and galaxy properties (e.g. bulge luminosity/mass and velocity dispersion; among others, Kormendy & Richstone 1995; Ferrarese & Merritt 2000; Marconi & Hunt 2003; Ferrarese et al. 2006; Cattaneo et al. 2009) clearly indicate that AGN are fundamental actors in the formation and evolution of galaxies and, in general, of cosmic structures in the Universe (see Kormendy & Ho 2013; Heckman & Best 2014, for two recent reviews on different aspects of the SMBH-host galaxy coevolution). Essential points to study the galaxy evolution are the identification of AGN in all states of accretion and obscuration and an accurate assessment of their physical properties, in order to infer their influence (‘feedback’) on the evolution of the host galaxy.

The ‘first-order’ model proposed to explain their characteristics, known as the Unified Model (Antonucci 1993), assumes that all the AGN have a similar internal structure. The basic idea is that most of the observed differences are caused by the angle between our line of sight and the plane of the system. In particular, the toroidal obscuring material supposed to partially cover the nuclear engine along particular lines of sight regulates the possibility of observing in the optical spectrum both broad and narrow (optically unobscured, or type 1, AGN) or only narrow (optically obscured, or type 2, AGN) emission lines. Confirmations of this schematic picture come from spectropolarimetric studies of optically obscured AGN, and from X-ray observations (for a recent review, see Bianchi, Maiolino & Risaliti 2012).

However, several authors suggested a possible evolution of the obscuration with cosmic time, connected with the SMBH growth (Hopkins et al. 2006; Lapi et al. 2014, recently updating the original model presented by Granato et al. 2004) in addition to the effects of different viewing angles. The main phase of efficient accretion would be characterized by AGN heavily obscured by gas and dust; this obscuring matter would be then swept away by the nuclear radiation, and the accreting SMBH would shine as an optically luminous quasar (QSO) for a brief period, until it runs out of gas. The final product of the sequence is a ‘dead QSO’ inside a passively evolving galaxy. In this scenario, unobscured AGN are a subsequent phase of obscured QSO.

In the framework of the Unified Model, we expect that absorbed and unabsorbed accreting nuclei share the same intrinsic properties (e.g. luminosity, BH mass, and accretion rate). Conversely, an evolutionary pattern for the two classes of AGN would imply a difference in their physical properties. Recently, a possible link between accretion and obscuration has been presented by Fabian, Vasudevan & Gandhi (2008) and Fabian et al. (2009). By discussing the effects of accretion-driven feedback on gas clouds in the innermost region of an AGN, the authors proposed that dusty absorbing structures around an AGN can be stable only for column densities N_H ≳ 5 × 10²³λ_Edd [cm⁻²], where λ_Edd is the AGN Eddington ratio, defined as the ratio between the bolometric luminosity and the Eddington luminosity.¹

The assessment of the nuclear properties of absorbed AGN is an issue that becomes more and more difficult as the amount of circumnuclear obscuring medium along the line of sight increases and does not permit a clear and direct view to the central energy source. In addition, at low nuclear luminosities a host contribution at optical–infrared (IR) wavelengths is expected to be increasingly noticeable, making even harder to isolate the nuclear component.

X-ray surveys are one of the best approaches to search for obscured AGN, since in this band the emission is less affected by absorption, if compared with the ultraviolet (UV)–optical domain. In this paper, we discuss the broad-band physical properties of all the X-ray absorbed QSOs (N_H > 4 × 10²¹ cm⁻², intrinsic L_2-10 keV > 10⁴⁴ ergs⁻¹; hereafter, XQSO2) in the XMM–Newton Bright Serendipitous Survey² (XBS), a complete sample of 400 bright X-ray sources. The XQSO2 sample considered here is unique at the moment: it consists of 14 X-ray and optically bright sources at intermediate redshift (0.35 < z < 0.9, plus one object at z ∼ 2), belonging to a complete and well-defined survey, for which multiwavelength data have been already collected. Proprietary or archival photometry from Spitzer, Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010), the Sloan Digital Sky Surveys (SDSS; Abazajian et al. 2009) and the Digitized Sky Survey (DSS), and good quality optical spectra, in addition to X-ray spectroscopic observations from XMM–Newton, allow us to properly characterize the sources in our sample. The major goal of this paper is to estimate the nuclear properties of the XQSO2s, in particular black hole masses and Eddington ratios, and to compare them with those of the sample of X-ray unabsorbed QSOs in the XBS.

The paper is organized as follows: in Section 2, we briefly describe the XBS and the XQSO2 sub-sample. Section 3 reports on the multiwavelength data used to construct the UV-to-IR spectral energy distribution (SED) for the XQSO2s studied here. We present the method adopted to deconvolve the host and AGN contributions in Section 4, while the details on the recovering of the nuclear properties are described in Section 5. We discuss the results obtained and their consequences, and compare the properties recovered for absorbed and unabsorbed sources, in Section 6. Finally, the main results are summarized in Section 7. The analysis of new XMM–Newton observations is presented in Appendix A, while details on the UV-to-mid-IR properties of the individual sources are reported in Appendix B. Throughout this paper, for consistency with the previous works on the XBS sample, a concordance cosmology with H₀ = 65 km s⁻¹ Mpc⁻¹, Ω_λ = 0.7, and Ω_m = 0.3 is adopted; when necessary, data and relations taken from the literature have been converted to our cosmology. The energy spectral index, α, is defined such that F_ν ∝ ν^−α. The X-ray photon index, defined as N_ϵ ∝ ϵ^−Γ, is Γ = α + 1. Finally, X-ray surveys demonstrated the existence of a small number of AGN showing a mismatch between the optical and X-ray-based methods of classification in terms of obscuration. In this paper, we define a source ‘absorbed’ or ‘unabsorbed’ on the basis of its X-ray properties (irrespective of the optical properties), adopting as threshold³ an intrinsic absorbing column density N_H = 4 × 10²¹ cm⁻².

THE XBS SAMPLE OF X-RAY ABSORBED QSOS

The XBS consists of two flux-limited samples of X-ray selected serendipitous sources at high Galactic latitude (|b| > 20°; total sky area ∼28 sq. deg): the XMM Bright Serendipitous Survey sample (BSS; 389 sources selected in the 0.5-4.5 keV energy band, identification level of 98 per cent) and the XMM Hard Bright Serendipitous Survey sample (HBSS; 67 sources selected in the 4.5-7.5 keV energy band, identification level of 100 per cent; 56 sources are in common with the BSS sample). The flux limit is ∼7 × 10⁻¹⁴ ergcm⁻²s⁻¹ in both the selection energy bands.

The details on the XMM–Newton fields selection strategy and the source selection criteria of the XMM BSS and HBSS samples are discussed in Della Ceca et al. (2004). A description of the optical data and analysis, of the optical classification scheme, and of the optical properties of the extragalactic sources identified so far is presented in Caccianiga et al. (2007, 2008); a paper presenting an update of the spectroscopic identifications [mostly based on data taken at the Telescopio Nazionale Galileo (TNG), Gran Telescopio Canarias (GTC) and Very Large Telescope (VLT)] of the XBS survey is in preparation. The identification statistics below take into account these new identifications, along with the new X-ray spectral analysis in Corral et al. (2011).

Regarding the AGN population, the entire XBS sample comprises 320 AGN divided into 281 type 1, 32 type 2, 6 BL Lacs, and 1 unclassified AGN (following the criteria described in Caccianiga et al. 2008). For the sub-sample of 40 ‘optically elusive AGN’ (i.e. sources for which the X-ray data are the only way to identify the AGN nature; Caccianiga et al. 2007), the type 1/type 2 classification is assigned on the basis of the intrinsic absorption in their X-ray spectra (14 absorbed AGN, 25 unabsorbed AGN, and 1 AGN for which the upper limit on N_H is not stringent enough to allow a classification). All but one of the type 1 AGN belong to the BSS sample, while a selection at higher energies confirms its effectiveness in finding obscured sources, with ∼62 per cent of type 2 AGN detected in the HBSS, as opposed to only ∼15 per cent of type 1 AGN. The cosmological and statistical properties of the HBSS AGN sample are extensively discussed in Della Ceca et al. (2008).

For most of the type 1 sources in the XBS AGN sample, the optical/UV SED has been build, and the bolometric luminosity has been estimated by fitting a multicolour disc model (Marchese et al. 2012). Caccianiga et al. (2013) report their black hole masses, estimated using the ‘single epoch’ spectral method (based on measuring the broad linewidths and the continuum emission).

The X-ray properties of the XBS AGN population are presented in Corral et al. (2011). From their analysis, the authors identify 15 QSOs absorbed in the X-ray (XQSO2s, i.e. sources with intrinsic absorption N_H > 4 × 10²¹ cm⁻² and X-ray luminosity L_2-10 keV > 10⁴⁴ ergs⁻¹). In the work presented here, we excluded three sources, optically classified as unobscured objects, having N_H with error bars crossing the threshold (XBSJ102417.5+041656, XBSJ213719.6−433347, and XBSJ213820.2−142536). We included two more XQSO2s, identified after the publication of the X-ray analysis of the sample (XBSJ051413.5+794343 and XBSJ080411.3+650906); the data analysis has been performed following the same prescriptions as in Corral et al. (2011). Note that for XBSJ051413.5+794343, the quality of the X-ray spectra allows us to obtain only an upper limit to the intrinsic column density N_H. Nevertheless, this object has been included in the sample, since the optical classification as AGN2 suggests the presence of obscuration.

The final sample is then composed by 14 objects; two are found only in the HBSS, five belong only to the BSS, and seven are present in both samples. We report the distributions in redshift, X-ray luminosity, and column density of the XQSO2s in Fig. 1. The XQSO2 sample covers a redshift range 0.35 < z < 0.9, with only one object at higher redshift, z = 1.985 (XBSJ021642.3−043553, an X-ray-selected Extremely Red Object discussed in details by Severgnini et al. 2006). Basic information is listed in Table 1, together with a summary of the results of the analysis performed by Corral et al. (2011), except for the two new optical identifications, presented here for the first time. The X-ray and optical spectra for all the sources are shown in Fig. 2 (top and central panels): for the majority of the XQSO2s, the optical spectra are typical of optically type 2 AGN, with only three optically elusive AGN (XBSJ052128.9−253032, XBSJ080411.3+650906, and XBSJ134656.7+580315) and two objects (XBSJ000100.2−250501 and XBSJ144021.0+642144, 14 per cent of the XQSO2s) classified as type 1 AGN, in agreement with the fraction of ‘mixed’ classification of about ∼20 per cent reported in literature (e.g. Trouille et al. 2009; Merloni et al. 2014). For an extensive discussion on the X-ray versus optical absorption in the XBS, we refer the reader to section 4.1 in Corral et al. (2011); here, we note that the only two optically classified type 1 sources have the lowest N_H, near the adopted threshold, and the observed mismatch is probably due to the assumption of the Galactic dust-to-gas ratio as mean A_V/N_H.

Figure 1.

Distribution in redshift (left-hand panel), intrinsic X-ray luminosity (central panel), and column density (right-hand panel) of the XQSO2s in the XBS (red shaded histogram); the blue dotted lines show the distributions for the XQSO1s (see Section 5). In the inset (central panel), we show the distributions of L_2-10 keV obtained when only sources with z < 0.85 are considered (see Section 6.1). The right-pointing arrows in the right-hand panel represent upper limits to N_H.

Open in new tabDownload slide

Figure 2.

The complete figure is available online. For guidance regarding its form, we show here the data for one XQSO2 (namely, XBSJ013240.1−133307). Top panel: XMM–Newton data and residuals (pn + combined MOS; red and black crosses and lines, respectively), from Corral et al. (2011), plotted as a function of the observed energy. Central panel: optical spectrum, plotted as a function of the observed wavelength; the strongest emission lines are marked. Bottom panel: rest-frame SED fits: the data [red filled circles, Spitzer; red open pentagons, WISE; green filled square and triangle, data in the R band and at 4400 Å from the literature; brown open star and triangle, OM magnitude and GALEX near-ultraviolet (NUV) flux], plotted as luminosity (λL_λ, normalized to the X-ray luminosity as in Table 1) versus the rest-frame wavelength, are superposed on the corresponding best-fitting template SEDs (blue dotted line, host galaxy; red dashed line, AGN; black solid line, total). In addition to the name (first row) and the redshift (second row) of the source, in the last row of the legend, we summarize the main parameters of the modelling: the morphological type of the host, the QSO template, the adopted extinction curve (see Fig. 3), and the dust extinction. See Section 4 for details.

Open in new tabDownload slide

As a follow-up programme of specific XBS sources, we obtained new XMM–Newton observations for 3 out of the 14 XQSO2s, namely XBSJ013240.1−133307, XBSJ113148.7+311358, and XBSJ160645.9+081525 (obsid 0550960101, 0550960301, and 0550960601, respectively; PI Della Ceca). A detailed analysis of the new data is reported in Appendix A. In the following, for these three sources we will assume the N_H and L_X reported in Table A2 for the new XMM–Newton data; for each source, the most significant parameters are quoted also in Table 1 (second row).

UV-TO-IR SED

In order to investigate the physical properties of the QSOs, we constructed their broad-band SED, from UV to IR wavelengths.

Spitzer observations and data reduction

Spitzer data for six XQSO2s were obtained in the Cycle-3, as part of a study of a statistically complete sample of optically absorbed AGN drawn from the XMM HBSS (Programme ID:30606; PI P. Severgnini). The observations were performed with the InfraRed Array Camera (IRAC) (3.6, 4.5, 5.8, and 8 μm) and Multiband Imaging Photometer for Spitzer (MIPS) (24 and 70 μm) instruments in photometry mode. We used a frame time of 12 s with a dither pattern of 16 points for the IRAC observations, and a frame time of 10 s with a 5-20 cycles and small field pattern for the MIPS ones. To estimate the flux densities in the IRAC and at the MIPS-24 μm bands, we used the final combined post-basic calibrated data (BCD) mosaics produced by the Spitzer Science Center (SSC) pipeline. At 70 μm, we looked at both the filtered and unfiltered post-BCD mosaics and if we saw negative side-lobes around our target in the filtered data, we re-filtered our BCD data by Ge Reprocessing Tools (gert). The gert is a contributed software from the MIPS Instrument Support Team made available to the scientific community to carry out custom off-line data reduction of MIPS-Ge (70 and 160 μm) data. In those cases, in which the re-filtering off-line process was applied, we have co-added and background subtracted the updated BCDs with the SSC Spitzer Mosaiker – mopex software (Makovoz & Marleau 2005). The resulting mosaics were made with ∼4 arcsec pixel size.

The flux densities of our targets have been measured using the iraf package and the sextractor software. For point-like sources, we estimated the aperture photometry at the position of the sources and we applied aperture corrections as derived by using the point response function. The chosen aperture radii are ∼6 arcsec for the IRAC bands, and ∼13 arcsec at 24 μm. Only in the presence of nearby optical and/or IR sources smaller apertures have been used. For all of those sources which appear extended in the IRAC and/or MIPS images, we used the curve of growth by masking possible sources nearby the target.

All the sources are detected up to 24 μm. We find flux densities ranging from 170 μJy to 4.4 mJy in the IRAC bands, and from 1 to 7 mJy at 24 μm. At 70 μm, we measure a significant emission (from 5 to 135 mJy) for four of our targets. For one additional source, we were able to estimate an upper limit to the flux using the weakest source detected in the field as limit.

Among the eight sources for which we do not have our own Spitzer data, XBSJ021642.3−043553 and XBSJ022707.7−050819 fall in one of the fields covered by the Spitzer Wide Area Infrared Extragalactic Survey (SWIRE; Lonsdale et al. 2003); from the SWIRE catalogue,⁴ we retrieved the flux densities obtained with aperture number 4, corresponding to a radius of 4.1 arcsec and 10.5 arcsec for the IRAC and MIPS data, respectively. Both sources are detected up to 24 μm.

Images in the IRAC and MIPS bands for XBSJ122656+013126 were retrieved from the archive and reduced as was done for the proprietary data; the source is detected up to 24 μm. We found imaging at 24 μm for one more object, XBSJ144021.0+642144; for the remaining four sources in our sample, we did not find either IRAC or MIPS data in the Spitzer archive.

Summarizing, including both proprietary and archival observations, we have 9 and 10 detections in the IRAC bands (at 3.6, 4.5, 5.8, and 8 μm) and at 24 μm, respectively, while at 70 μm we have four detections and 1 upper limit; the results of the Spitzer data analysis are reported in Table 2.

WISE

WISE is a NASA satellite that imaged the whole sky in four mid-IR photometric bands, centred at 3.4, 4.6, 12, and 22 μm. We refer to these bands as W1, W2, W3, and W4, respectively. The angular resolution is 6.1 arcsec, 6.4 arcsec, 6.5 arcsec, and 12.0 arcsec, respectively. We use the WISE All-Sky Data Release (2012 March), which includes all observations obtained during the fully cryogenic mission (2010 January 7 to 2010 August 6). The WISE All-Sky source catalogue has been searched by using the X-ray source position and assuming an impact parameter equal to 4 arcsec; we considered as detection only WISE magnitudes with signal-to-noise ratio S/N > 3 in the W1, W2, and W3 bands, and S/N > 5 in the W4 band; for lower S/N, we assumed the corresponding 2σ upper limits. A posteriori, we found that the selected WISE objects were closer than 2 arcsec from the input source position. Using a simulated sample of ∼1600 objects, we estimated that the probability to have a random WISE source into a circle of 2 arcsec radius is ∼1.1 per cent, implying that all the associations between the WISE objects and our XQSO2s are likely real.

In Table 2, we report the flux densities, computed from the profile-fit photometry (obtained from the WISE All-Sky source catalogue) by assuming the magnitude zero-points of the Vega system corresponding to a power-law spectrum (f_ν ∝ ν^−α) with α = 1. The differences in the computed flux densities expected using flux correction factors that correspond to α = −1, 0 or 2 (lower than 0.8 per cent, 0.6 per cent, 6 per cent, and 0.7 per cent in W1, W2, W3, and W4, respectively) have been linearly added to the catalogued flux errors. Following the prescription in Wright et al. (2010), an additional 10 per cent uncertainty was linearly added to the 12 and 22 μm fluxes.

Other data

In Table 3, we report the SDSS u, g, r, i, and z magnitudes found for seven XQSO2s. Magnitudes in the R band were collected from the literature for seven XQSO2s (Fiore et al. 2003; Caccianiga et al. 2008); a DSS magnitude was available for one source, while we found a magnitude in the g^′ band for one more object (Stalin et al. 2010). From the Galaxy Evolution Explorer (GALEX) catalogue, we retrieved fluxes in the near-UV (at 2310 Å) for three XQSO2s, one of them having also an entry in the far-UV (at 1539 Å); for XBSJ013240.1−133307 a magnitude in the U band was obtained from the Optical Monitor (OM) telescope onboard XMM–Newton (see Appendix A). At longer wavelengths, in addition to Spitzer and WISE photometry, a K-band magnitude for XBSJ021642.3−043553 is reported by Severgnini et al. (2006).

SED DECONVOLUTION

To decompose the galaxy and AGN contributions, we adopted a very simple phenomenological approach. We have constraints from both photometry and optical spectra; to account for all of them, we made use of

• an AGN spectral template, composed by a combination of broad line and continuum emission, covering the optical energy range (Francis et al. 1991; Elvis et al. 1994);

• a template of AGN narrow lines, constructed adopting the prescriptions in Krolik (1998) for the relative intensities of the lines;

• a set of templates of QSOs and galaxies, extending from the UV up to the mid-IR wavelength bands, chosen from the template library of Polletta et al. (2007). The broad-band QSO templates have been derived by combining the SDSS QSO composite spectrum and rest-frame IR data of a sample of optically selected type 1 QSOs observed in the SWIRE programme. We considered three templates with the same optical spectrum but with three different IR SEDs: a mean IR spectrum, obtained from the average fluxes of all the measurements (‘QSO1’), a template with high IR/optical flux ratio, obtained from the highest 25 per cent measurements per bin (‘TQSO1’), and a low-IR emission SED obtained from the lowest 25 per cent measurements per bin (‘BQSO1’). The three templates are shown in Fig. 3, left-hand panel. Regarding the galaxy emission in the UV–IR, we considered 10 templates, each of them covering the wavelength range between 1000 Å and 1000 μm. They include three ellipticals and seven spirals (from early to late types, S0-Sdm; see Fig. 3, right-hand panel).

To extinguish the nuclear emission, both the extinction curves of our Galaxy, and of the Galactic Centre (Chiar & Tielens 2006), have been considered; a comparison between the two curves is shown in Fig. 3, central panel.

Figure 3.

Templates adopted in this work for the SED modelling (see Section 4). Left-hand panel: nuclear templates (0.1-1000 μm) of optically selected luminous AGN with different values of the IR/optical ratio, from Polletta et al. (2007). Central panel: extinction curves of our Galaxy (green continuous line) and of the Galactic Centre (red dashed line; Chiar & Tielens 2006). Right panel: templates of three elliptical galaxies with different ages, three early spirals, and four late spirals and irregulars (as labelled), generated with the grasil code (Silva et al. 1998), from Polletta et al. (2007).

Open in new tabDownload slide

First, we concentrated on the optical spectra, to obtain a guess for the dust extinction. XBSJ021642.3−043553 has been discussed in details in Severgnini et al. (2006); for the remaining 13 sources, the optical AGN template (previously labelled as a) has been absorbed and then summed up to the narrow line component (previously labelled as b); the total AGN template was redshifted to the redshift of the source and summed with a redshifted galaxy template. We varied the amount of dust extinction, A_V, and the relative AGN/galaxy normalization until a good reproduction of the available optical spectrum is reached. An estimate of A_V (six objects), or a lower limit to the same parameter (seven objects), have been thus obtained.

We then moved to the analysis of the UV-to-mid-IR SED. In this modelling, the A_V (or the lower limit to A_V) obtained from the spectral analysis has been used as a first guess for the photometric dust extinction; the latter has been left free to vary during the SED modelling. Similarly, as guess for the host type, we considered the Ca break at 4000 Å obtained from the optical spectra. After shifting the photometric data to the rest frame, each SED has been reproduced by the sum of a galaxy plus a QSO template (both previously labelled as c). In the modelling, the latter was absorbed to reproduce the intrinsic obscuration at the source redshift before summing up with the galaxy component. For each source and for each QSO template, different combinations of nuclear A_V and galaxy templates have been tested; we selected the combination that, from a visual match, was able to better reproduce most of the photometric points. In the SED analysis described here, we did not take into account information derived from the X-ray data. In particular, we did not assume any relation between the N_H and the dust extinction.

Note that, since a significant fraction of the flux in the UV and optical bands has been absorbed in our AGN template, the UV and optical photometric data are most likely dominated by the host galaxy template. In this sense, among the photometric points, the SDSS data are the most relevant ones to constrain the galaxy type, at least for the optically type 2 XQSO2s.

In discussing the results of our analysis, a couple of considerations about the adopted method of modelling must be taken into account. First, here we are assuming that the emission of type 2 QSOs can be reproduced using the mean disc+torus emission of the QSO1s, after screening both components with a nuclear extinction, reproducing the absorption and auto-absorption associated with the torus. A more physical approach would require a more sophisticated parametrization of the AGN torus, to properly describe the auto-absorption and re-emission balance and its dependence on the angle of view or the dust properties. In particular, observational evidences suggest that the dust within the torus is arranged in clumps instead of being smoothly distributed (Krolik & Begelman 1988; Tacconi et al. 1994; Tacconi 1996; Markowitz, Krumpe & Nikutta 2014, and references therein). However, our data coverage and spectral resolution are not sufficient to constrain the huge number of parameters present in the torus models incorporating clumpy gas (e.g. Nenkova, Ivezić & Elitzur 2002; Hoenig et al. 2006; Nenkova et al. 2008a,b; Schartmann et al. 2008; Heymann & Siebenmorgen 2012; see Hoenig 2013 for a review). Instead, an approach like the one adopted here is more than adequate for sparse photometric data point.

Moreover, in the mid/far-IR band the empirical QSO1 templates from Polletta et al. (2007) in principle could suffer of contamination due to the host galaxy emission. However, a comparison with the results presented by Mateos et al. (2012, 2013) on a sample of X-ray selected type 1 AGN SED strongly suggests that at luminosities greater than 10⁴⁴ ergs⁻¹ such a contribution, if present, would not affect our results. The selection technique applied by the authors, based on the mid-IR colours, suggests a total mid-IR emission dominated by the AGN component. The median SED of the Mateos et al. type 1 AGN is well represented by the same templates adopted here, therefore we expect that a similar result can be applied to our XQSO2s.

In Fig. 2, bottom panel, we report the resulting SEDs normalized to the de-absorbed rest-frame 2-10 keV luminosity derived from the X-ray spectral analysis (as in Table 1). For most of the sources, each of the tested templates of QSO (i.e. ‘TQSO1’, ‘QSO1’, and ‘BQSO1’) provides an acceptable description of the observed SED, with slightly different relative normalizations of the templates and amount of dust extinction: of course, this implies different values for the physical parameters estimated from the modelling (see Section 5). Details on the SED modelling of each object are reported in Appendix B; in the following, for each parameter we will quote the value estimated assuming the mean template, ‘QSO1’, while the associated range corresponds to the minimum and maximum values obtained when all the possible modellings (with all the QSO templates and extinction curves) are considered.

For XBSJ052128.9−253032 and XBSJ080411.3+650906, the available data do not allow us to infer both the intensity of the nuclear emission and the level of optical absorption. In these cases, we carried out the modelling several times, with A_V fixed to different values ranging between the lower limit recovered by the analysis of the optical spectrum (A_{V, opt}) and the highest value obtained from the column density provided by the X-ray analysis (A_{V, X}), assuming a Galactic dust-to-gas ratio, A_V/N_H = 5.27 × 10⁻²² mag/cm⁻² (Bohlin et al. 1978) to investigate the effects of different levels of obscuration. For each physical parameter, we assume the value provided by the modelling with A_{V, opt} as reference.

When the spectral analysis provides estimates of A_V, these differ from the values derived from the SED by a factor lower than 1.5. Regarding the seven sources for which we obtained only a lower limit to the dust extinction, in four cases the SED-derived values agree with these limits. For three sources, namely XBSJ022707.7−050819, XBSJ134656.7+580315, and XBSJ160645.9+081525 (which optical spectra suggest A_V > 2-3 mag), the SEDs require an A_V in good agreement with the N_H derived from the X-ray analysis (see Table 4). For what concerns the two optically type 1 AGN, XBSJ000100.2−250501 and XBSJ144021.0+642144, both the spectral analysis and the SED modelling give us results in agreement with the optical classification. At the same time, the quite different A_V recovered reflects the difference observed in the optical spectra (see Fig. 2), with XBSJ000100.2−250501 (A_V = 1.5 mag, as derived from the SED) being clearly redder than XBSJ144021.0+642144 (for which we find A_V = 0.2 mag).

From our analysis, 9 out of the 14 XQSO2s reside in ellipticals or early spirals (S0 or Sa). Our findings are quite in agreement with previous results on luminous type 2 AGN, such as the SDSS sample by Zakamska et al. (2006), who found mainly elliptical hosts. Recently, Mignoli et al. (2013) linked the lower fraction of ellipticals found in their sample of |${[{\rm Ne\,\small {\rm V}}]}$|-selected z ∼ 1 obscured AGN to the range in nuclear luminosity covered, extended to an order of magnitude lower than SDSS type 2 QSOs. For five XQSO2, instead, the SED modelling suggests a late-type host galaxy (Sd and Sdm). In three cases, this result is quite robust thanks to the availability of SDSS and/or GALEX data. Although any comparison should be done with caution, due to the different methods of classification (X-ray based or optically based), this is an interesting result since a paucity of AGN hosted in late-type spiral and irregular morphologies has been claimed, at least in optically selected samples (e.g. Mignoli et al. 2013).

RECOVERING THE NUCLEAR PROPERTIES

Disentangling the AGN and host galaxy emission as described in Section 4 provides detailed information on the host-galaxy bulge and nuclear components. In this section, we detail how we use this information to derive the BH masses, the bolometric luminosities, and the Eddington ratios for the XQSO2s.

Black hole masses

Note that adopting these relations implies assuming that the absolute magnitude of the bulge strictly mirrors the mass in old stars, M_bulge, which is the quantity primarily related to the BH mass. Moreover, we are implicitly assuming that the M_bulge–M_BH relation is already imprinted at z ∼ 1, driving the main episode of accretion.

We found rather high masses, narrowly distributed between ∼1.7 × 10⁸ and ∼1.6 × 10⁹ M_⊙ (see Table 4 and Fig. 4, left-hand panel).

Figure 4.

Nuclear properties of the XQSO2s studied here, derived as described in Section 5 (red shaded histograms): distributions in BH mass (left-hand panel), X-ray bolometric correction (central panel), and Eddington ratio (right-hand panel). The arrows mark the two objects for which we are able to put only limits to the nuclear emission (namely, XBSJ052128.9−253032 and XBSJ080411.3+650906). The blue dotted lines show the corresponding distributions for the XQSO1s. In the insets, we show the corresponding distributions obtained when only sources with z < 0.85 are considered (see Section 6.1).

Open in new tabDownload slide

Bolometric luminosities and Eddington ratios

The next step is to investigate the nuclear bolometric luminosity of the XQSO2s. In unabsorbed sources, this quantity can be computed as the sum⁵ of the accretion disc optical/UV luminosity integrated at wavelengths shorter than 1 μm and the intrinsic X-ray luminosity between 0.1 and 500 keV rest frame, L_bol = L_opt-UV+L_X.

However, in obscured sources L_opt-UV is almost completely absorbed due to the high dust extinction. Thus, any estimate of the accretion disc luminosity would strongly depend on the template adopted to reproduce the intrinsic optical–UV emission of the QSO. Moreover, from the analysis of the optical/UV SED of the type 1 AGN in the XBS presented in Marchese et al. (2012), we found that the effective disc temperature ranges from kT ∼ 1 to kT ∼ 8 eV, and on average the disc emission has a 〈kT〉 ∼ 3.7 eV. Therefore, the limit at 0.1 μm of the template of AGN would provide a change in an estimated disc luminosity lower than the intrinsic one by a factor up to ∼57 per cent.

We therefore adopted a different approach: we self-calibrated the intrinsic nuclear emission using the SED of the X-ray unabsorbed QSOs drawn from the XBS.

We considered the 40 AGN in the XBS (hereafter, XQSO1s) with (i) intrinsic absorption (detected, or upper limit) low enough to minimize the possible effects of absorption on the estimate of the L_bol (N_H or upper limit to N_H lower than 5 × 10²⁰ cm⁻²; Fanali et al. 2013); and (ii) X-ray luminosity above the threshold assumed for the XQSO2s (L_2-10 keV > 10⁴⁴ ergs⁻¹). The redshift, X-ray luminosity, and column density distributions of the XQSO1s are reported as dashed blue lines in Fig. 1. A Kolmogorov–Smirvov (KS) test indicates that this sub-sample is not statistically different from the sample of 95 X-ray unabsorbed (N_H < 4 × 10²¹ cm⁻²) AGN with L_2-10 keV > 10⁴⁴ ergs⁻¹ belonging to the XBS for what concerns the distributions in redshift, X-ray and bolometric luminosities, BH mass, and Eddington ratio.

We want to use this sample of 40 XQSO1s, for which a bolometric luminosity has been already computed, to obtain an IR bolometric correction to be applied to the sample of XQSO2. In particular, we use the WISEW3 magnitude since in this band (a) observations with good S/N are available, and (b) we are confident that in luminous unobscured QSOs the data are dominated by the AGN (note that all the XQSO1s are also optically classified as type 1). From our SED modelling, when the QSO emission is corrected for the absorption (thus reproducing the SED of an unabsorbed QSO), we expect that the galaxy contributes less than 10 per cent to the total luminosity observed in the W3 band.

As done for the XQSO2s, we looked for a WISE counterpart of the 40 XQSO1s, finding an IR detection in WISE for all but one; here, we considered the 36 WISE counterparts with S/N > 3 in W3. For these XQSO1s, we computed the luminosity in the W3 band, L_{WISE, 3} = ν_W3 × L_W3, where L_W3 and ν_W3 are the monochromatic luminosity (derived from the catalogued WISE magnitude as detailed in Section 3) and the central frequency correspondent to the W3 band, observed frame (corresponding to ∼ 4.5-9.1 μm rest frame, depending on the redshift). Since early studies based on IRAS data (e.g. Spinoglio & Malkan 1989), it is known that the AGN flux emitted at these wavelengths is a fraction almost constant of the bolometric flux.

Figure 5.

Comparison between the bolometric luminosity (estimated from the optical/UV SED by Marchese et al. 2012) and the luminosity derived from the WISE W3 data for the XQSO1s with a detection in WISE. The green thick solid line is the result of the least-squares fitting when L_bol is assumed as dependent variable (corresponding to the relation quoted in the plot; the thinnest solid green lines mark the 1σ dispersion). The grey dotted line corresponds to the least-squares fitting when L_{WISE, 3} is assumed as dependent variable, while the long-dashed line is the bisector of the two least-squares fitting lines.

Open in new tabDownload slide

Figure 6.

Bolometric luminosities versus black hole masses for the sample of XQSO2s (red filled circles); the bars identify the ranges of values obtained when all the possible modelling (with all the QSO templates and extinction curves) are considered. The arrows mark the two objects for which we are able to put only limits to the nuclear emission (namely, XBSJ052128.9−253032 and XBSJ080411.3+650906). The blue open squares mark the positions of the XQSO1s. Grey long-dashed, dashed and dotted lines define the locus for sources emitting at the Eddington limit, at 1/10 of it, and at 1/100 of it.

Open in new tabDownload slide

From the black hole mass and the bolometric luminosity, we can estimate the accretion rate normalized to the Eddington luminosity, λ_Edd (as defined in Section 1). The values are reported in Table 4; as shown in Fig. 4 (right-hand panel), λ_Edd ranges from 0.01 and 0.73; for 6 out of 14 QSOs (among them, the two sources for which we are only able to put limits to the nuclear emission) we found λ_Edd < 0.1.

DISCUSSION

XQSO1s versus XQSO2s in the XBS

In the previous sections, we described how we recovered L_bol, M_BH, and λ_Edd for the complete sample of XQSO2s in the XBS. Thanks to the analysis previously performed for the XQSO1s (Marchese et al. 2012; Caccianiga et al. 2013; Fanali et al. 2013), we are now able to compare the physical properties of X-ray absorbed and X-ray unabsorbed QSOs on a statistical basis.

Using the project for statistical computing⁷R (R Core Team 2013), we performed a set of two-samples KS tests; we compared for various parameters of interest the cumulative distributions obtained for the 40 XQSO1s and for the 12 XQSO2s for which we were able to infer the intensity of the nuclear emission (i.e. excluding XBSJ052128.9−253032 and XBSJ080411.3+650906). There is no evidence that the two samples are drawn from different populations in terms of any of the tested quantities (see the first row in Table 5, where the results are summarized), although the KS probability for the Eddington ratio is quite low, p ∼ 4 per cent.

While the lower end of the two redshift distributions are similar (z = 0.32 and z = 0.35 for XQSO1s and XQSO2s, respectively), all but one the XQSO2s have z < 0.85, while 40 per cent of the XQSO1s have a redshift between 0.85 and 1.64 (see Fig. 1, left-hand panel). In Table 5, second row, we quote the results of the KS tests performed on the two sub-samples with z < 0.85 (their distributions in terms of intrinsic X-ray luminosity, black hole mass, X-ray bolometric correction and Eddington ratio are reported in the insets of Fig. 1, central panel, and Fig. 4): also with this cut applied, the hypothesis of the same original population is confirmed, although with a lower significance in L_2-10 keV. The KS probability for M_BH increases significantly when we restrict to z < 0.85; 10 out of the 13 XQSO1s with M_BH ≥ 1.6 × 10⁹ M_⊙, that is, the maximum value found for the XQSO2s (see Fig. 4, left-hand panel, and Fig. 6) have z > 1.

IR emission and L_bol

In Section 5.2, we derived the bolometric luminosity of the XQSO2s using the properties of the XQSO1s. The SED decomposition presented in Section 4 allows us to check these estimates. The IR luminosity is considered an indirect probe of the accretion disc optical/UV luminosity; using an approach similar to Pozzi et al. (2007, 2010) and Vasudevan et al. (2010), the bolometric luminosity can be computed as the sum of the total IR luminosity and the X-ray luminosity, L_bol = L_IR+L_X, where the former can be obtained from the nuclear luminosity between 1 and 1000 μm.

We calculate the L_{1 − 1000} μm by integrating the QSO template between 1 and 1000 μm after correcting for absorption. Following the works of Pier & Krolik (1992) and Pozzi et al. (2007, see also Lusso et al. 2011), L_{1 − 1000} μm can be converted into the nuclear accretion disc luminosity when corrected for geometrical effects related to the covering factor f of the torus. Following Pozzi et al. (2007, 2010), we assumed f ∼ 0.67, corresponding to an angle θ ∼ 48° between the axis of the disc and the edge of the torus. This is of course a mean value, and the derived luminosities must be assumed just as a first-order estimate.

The bolometric luminosities as derived from equation (3) are compared to the 1-1000 μm-based bolometric luminosities in Fig. 7. The 12 sources with well-constrained nuclear component in our SED modelling fall around the equal-luminosity relation, with a mean and a maximum scatter of 0.07 dex and 0.16 dex, respectively. The agreement of the two estimates of L_bol implies that absorbed and unabsorbed QSOs in the XBS have the same direct-to-reprocessed luminosity ratio. Moreover, this confirms that equation (3) can be confidently used to derive a first-order estimate of the bolometric luminosity, at least for unabsorbed high-luminosity QSOs, where the host galaxy is not expected to contaminate the emission in the WISE W3 band.

Figure 7.

IR-based L_bol derived for the XQSO2s following the prescriptions by Pier & Krolik (1992) and Pozzi et al. (2007) versus the WISE-based L_bol derived by applying the relation between the WISE W3 data and the bolometric luminosity recovered from the analysis of the XQSO1s in the XBS; the bars identify the ranges of values obtained when all the possible modelling (with all the QSO templates and extinction curves) are considered. The arrows mark the two objects for which we are able to put only limits to the nuclear emission (namely, XBSJ052128.9−253032 and XBSJ080411.3+650906). The dotted line marks the one-to-one relation between the luminosities.

Open in new tabDownload slide

The effective Eddington limit

The studies of Fabian et al. (2008) and Fabian et al. (2009) identify an ‘effective Eddington limit’ for dusty gas in the N_H–λ_Edd plane (see Fig. 8). Starting from the model presented in Fabian, Celotti & Erlund (2006), the authors suggest a connection between the radiation pressure exerted by the AGN on its proximity (i.e. where the influence of the BH is dominant) and the structure of the surrounding material. Briefly, for a fixed nuclear luminosity, the incident radiation would result on an easier sweeping away of the absorber when the medium is more dusty. Assuming a dust-to-gas ratio constant during the process of ‘cleaning’, this translates in a ‘forbidden region’ at high N_H and high ‘classical’ λ_Edd (i.e. derived by assuming a distribution of gas composed only of ionized hydrogen). In this region, absorbing dusty gas clouds are unstable to radiation and absorption may be transient or variable. For lower ‘classical’ Eddington ratios, also quite dense absorbing clouds can survive to the radiation pressure (‘long-lived clouds region’). Finally, sources with low N_H (≲3 × 10²¹ cm⁻²) can accrete at high λ_Edd, being the absorption possibly associated with dust lanes, far away from the nucleus, so as to be retained by a gravitational mass large enough to balance also an intense nuclear emission.

Figure 8.

Hydrogen column density versus Eddington ratio for the XQSO2s analysed in this work (red filed circles); the bars identify the ranges of values obtained when all the possible modelling (with all the QSO templates and extinction curves) are considered. The horizontal arrows mark the two objects for which we are able to put only limits to the nuclear emission (namely, XBSJ052128.9−253032 and XBSJ080411.3+650906). With the blue open symbols, we show the whole sample (i.e. N_H < 4 × 10²¹ cm⁻²) of X-ray unabsorbed high-luminosity (L_2-10 keV > 10⁴⁴ ergs⁻¹) AGN in the XBS (from Marchese et al. 2012). The lines and the different regions in the plot summarize the ‘effective Eddington limit paradigm’ (adapted from Fabian et al. 2009, see the details in the text): the grey continuous line marks the effective Eddington limit for standard ISM and enclosed mass of M_BH; the grey long-dashed line show how the effective Eddington limit moves when 1/10 of standard ISM grain abundance is assumed; if in addition the enclosed mass is twice the M_BH and the incident radiation is reduced by a factor of 2.5 (accounting for an anisotropic emission from the disc; see the text for details), the limit can further increase (grey dashed line).

Open in new tabDownload slide

In Fig. 8, we plot the XQSO2s studied here (red filled circles) in the N_H–λ_Edd diagram; the blue open squares mark the positions of the X-ray unabsorbed AGN in the XBS with L_2-10 keV > 10⁴⁴ ergs⁻¹. Unlike in the previous sections, to properly fill the plane we do not restrict the comparison to AGN with N_H < 5 × 10²⁰ cm⁻², but we consider the whole sample studied by Marchese et al. (2012, N_H < 4 × 10²¹ cm⁻²). As evident from the figure, our XQSO2s populate the ‘forbidden region’ proposed by Fabian and collaborators (the white part of the plot). The agreement found in Section 6.2 between the different estimates of L_bol guarantees that this result does not depend on the method adopted to derive the bolometric luminosity. In particular, the small scatter found between the bolometric luminosities derived from equation (3) and from L_{1 − 1000} μm is not enough to move any source outside the ‘forbidden region’.

The ‘effective Eddington limit’ drawn in Fig. 8 (grey continuous line) has been calculated for a grain abundance typical of the interstellar medium (ISM), and under the hypothesis that the only inward gravitational force of importance is that from the central BH (see Fabian et al. 2009). Different assumptions can modify the shape of the ‘forbidden region’. The lower is N_H, the stronger are the effects of a decreasing of the grain abundance (e.g. long-dashed grey line in Fig. 8). Ratios of optical extinction A_V to X-ray column density N_H lower than the Galactic dust-to-gas ratio are not unusual in AGN (Maiolino et al. 2001); in 9 out of the 12 XQSO2s, for which we were able to properly model the nuclear properties, we find a dust-to-gas ratio lower than the Galactic one (we stress again that in Section 4 the dust extinction from the SED modelling was derived independently of the value of N_H from the X-ray spectra).

As suggested by Fabian et al. (2009), the presence of stars inwards from the dusty gas clouds close to the nucleus would increase the effective Eddington limit by an amount proportional to the total enclosed mass, regardless of the amount of obscuration. Near-IR high-resolution data for nine Seyfert galaxies (Davies et al. 2007) have revealed evidence for recent, intense, short-lived starbursts on few pc scale distance from the BH, leaving behind massive and vertically thick nuclear star clusters.

Qualitatively, a combination of low grain abundance (e.g. of the order of one tenth of the ISM abundance), enclosed mass higher than M_BH (e.g. twice the mass of the central BH), and anisotropy of the disc emission, would reduce the proposed ‘forbidden region’ in the N_H–λ_Edd plane as marked by the dashed grey line in Fig. 8. More quantitative analyses support the role of the anisotropic disc radiation in populating the ‘forbidden region’, as discussed by Liu & Zhang (2011).

SUMMARY

In this paper, we presented a detailed analysis of the accretion properties of a complete sample of 14 XQSO2s (column density N_H > 4 × 10²¹ cm⁻² and X-ray luminosity L_2-10 keV > 10⁴⁴ ergs⁻¹; 0.35 < z < 0.9, plus one object at z ∼ 2) drawn from the XBS.

The bright flux of these sources guarantees a high detection rate in a number of multiwavelength catalogues (Spitzer, WISE, SDSS, GALEX), thus allowing us (in combination with proprietary observations) to construct photometric broad-band SED from the optical range up to ∼ 25 μm (observed frame), or even ∼ 70 μm (observed frame). By analysing the SED with a combination of empirical templates, we were able to separate the nuclear and the host-galaxy emission. For all the sources, we obtained the masses of the central black hole from the luminosity of the host galaxy, using the well-known scaling relations linking the BH mass and the host properties.

To derive the nuclear bolometric luminosities, we used the properties of the X-ray unabsorbed high-luminosity AGN (N_H < 5 × 10²⁰ cm⁻²and L_2-10 keV > 10⁴⁴ ergs⁻¹; XQSO1s) in the XBS. We compared their WISE luminosity at 12 μm (observed frame, a wavelength range where a powerful QSO is expected to dominate over the host galaxy emission) with the L_bol, derived from an optical/UV SED modelling (an approach that does not include the IR emission; Marchese et al. 2012). We found that in these high-luminosity unabsorbed QSOs the bolometric luminosity and the W3 luminosity are well-correlated over ∼2 dex. The best-fitting relation has been then applied to the XQSO2s, adopting as W3 luminosity the value derived from the nuclear component after correcting for the estimated absorption.

The main results of our work can be summarized as follows.

• Being the unabsorbed population in the XBS previously analysed (Marchese et al. 2012; Caccianiga et al. 2013; Fanali et al. 2013), we could compare the two samples of XQSO1s and XQSO2s, looking for possible statistically significant differences. We found that X-ray-selected absorbed and unabsorbed QSOs share the same intrinsic properties (L_X, L_bol, k_X, M_BH, and λ_Edd).

• The WISE-based L_bol have been compared with the IR-based L_bol, derived from the intrinsic (i.e. absorption-corrected) L_{1 − 1000} μm as computed from the SED modelling. We found a good agreement between the two estimates. We conclude that the absorbed and unabsorbed QSOs in the XBS sample have similar observed torus-reprocessed luminosity-to-bolometric luminosity ratio.

• The XQSO2s analysed here populate the proposed ‘forbidden region’ in the N_H versus λ_Edd diagram. This result could imply a grain abundance much lower than in the local ISM or the presence of stars inwards of the obscuring material with mass comparable to the mass of the BH. In addition, we propose that the anisotropy in the nuclear emission can also play a significant role.

In conclusion, our work suggests that the hypothesis of different evolutionary states for absorbed and unabsorbed QSOs in addition to the basic Unified Model is not required, at least up to z ∼ 1.

We thank the referee for her/his constructive comments that improved the paper. The research leading to these results has received funding from the European Commission Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 267251 ‘Astronomy Fellowships in Italy’ (AstroFIt). The authors acknowledge financial support from the Italian Ministry of Education, Universities and Research (PRIN2010-2011, grant no. 2010NHBSBE) and from ASI (grant no. I/088/06/0). SM and FJC acknowledge financial support by the Spanish Ministry of Economy and Competitiveness through grant AYA2012-31447. SM acknowledges support from the ARCHES project (Seventh Framework of the European Union, no. 313146).

We thank F. Pozzi, A. Feltre and E. Lusso for the useful discussions; T. Maccacaro and L. Maraschi for their helpful suggestions; and R. Paladini for her assistance with the WISE data and catalogue.

Based on observations obtained with XMM–Newton (an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA, NASA). Based on observations made under the proposals GTC18-10B and GTC44-11A with the Gran Telescopio Canarias (GTC), installed at the Spanish ‘Observatorio del Roque de los Muchachos’ of the Instituto de Astrofísica de Canarias, in the island of La Palma. This work is based in part on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA.

Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the US Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. This publication makes use of data products from the WISE, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This research has made use of NASA's Astrophysics Data System.

REFERENCES