Searching for Compton-thick active galactic nuclei at z~0.1
Andy Goulding, David Alexander, James Mullaney, Jonathan Gelbord, Ryan Hickox, Martin Ward, Mike Watson
aa r X i v : . [ a s t r o - ph . C O ] S e p Mon. Not. R. Astron. Soc. , 1–16 (2010) Printed 4 October 2018 (MN LaTEX style file v2.2)
Searching for Compton-thick active galactic nuclei at z ∼ . A.D. Goulding ; D.M. Alexander ; J.R. Mullaney ; J.M. Gelbord , ; R.C. Hickox ;M. Ward & M.G. Watson Department of Physics, Durham University, South Road, Durham. Department of Astronomy and Astrophysics, 525 Davey Laboratory, The Pennsylvania State University, University Park, PA 16802. Department of Physics & Astronomy, University of Leicester, Leicester, LE1 7RH, UK
Released 2010 Xxxxx XX
ABSTRACT
Using a suite of X-ray, mid-infrared (IR) and optical active galactic nuclei (AGN) lumi-nosity indicators, we search for Compton-thick AGNs with intrinsic L X > erg s − at z ∼ . E >
10 keV) all-sky X-ray surveys. We haveused the widest
XMM-Newton survey (the serendipitous source catalogue) to select arepresentative sub-sample (14; ≈
10 percent) of the 147 X-ray undetected candidateCompton-thick AGNs in the Sloan Digital Sky Survey (SDSS) with f X /f [OIII] < ≈
50 percent of the overall Type-2 AGN population inthe SDSS–XMM overlap region. We use mid-IR spectral decomposition analyses andemission-line diagnostics, determined from pointed
Spitzer -IRS spectroscopic observa-tions of these candidate Compton-thick AGNs, to estimate the intrinsic AGN emission(predicted 2–10 keV X-ray luminosities, L X ≈ (0 . × erg s − ). On the basis ofthe optical [O iii ], mid-IR [O iv ] and 6 µ m AGN continuum luminosities we conserva-tively find that the X-ray emission in at least 6/14 ( > ≈
43 percent) of our sample appearto be obscured by Compton-thick material with N H > . × cm − . Under thereasonable assumption that our 14 AGNs are representative of the overall X-ray unde-tected AGN population in the SDSS–XMM parent sample, we find that > ≈
20 percentof the optical Type-2 AGN population are likely to be obscured by Compton-thickmaterial. This implies a space-density of log(Φ) > ≈ − . − for Compton-thickAGNs with L X > ≈ erg s − at z ∼ .
1, which we suggest may be consistent withthat predicted by X-ray background synthesis models. Furthermore, using the 6 µ mcontinuum luminosity to infer the intrinsic AGN luminosity and the stellar velocitydispersion to estimate M BH , we find that the most conservatively identified Compton-thick AGNs in this sample may harbour some of the most rapidly growing black holes(median M BH ≈ × M ⊙ ) in the nearby Universe, with a median Eddington ratioof η ≈ . Key words: galaxies: active – galaxies: evolution – galaxies: nuclei – infrared: galaxies
There is now strong observational evidence that all mas-sive galaxies ( M ∗ ≈ –10 M ⊙ ) in the nearby Uni-verse host a central supermassive black hole (SMBH;M BH ≈ –10 M ⊙ ; Kormendy & Richstone 1995). TheseSMBHs have grown through mass accretion events (e.g.,Soltan 1982; Rees 1984), during so-called active galacticnucleus (AGN) phases. The seminal discovery that themasses of SMBHs are proportional to those of their stel-lar spheroids implies a strong physical association betweenAGN activity and galaxy evolution (e.g., Magorrian et al. 1998; Ferrarese & Merritt 2000; Gebhardt et al. 2000;Tremaine et al. 2002). To fully interpret the role played byAGN in this symbiosis requires a complete census of ob-scured and unobscured AGNs across cosmic time.Unbiased deep and wide-field X-ray surveys have beeninstrumental in the identification of a large proportionof the AGN population to high redshifts ( z ∼
5; e.g.,Alexander et al. 2001; Barger et al. 2003; Fiore et al. 2003;Tozzi et al. 2006; Brusa et al. 2010). Using the exceptionalsensitivities of
XMM-Newton and the
Chandra X-ray Ob-servatory , >
80 per cent of the X-ray background (XRB)has been resolved into discrete sources at soft energies (0.5– c (cid:13) A.D. Goulding et al. ∼
50 per cent of the AGN population may be heav-ily obscured and remains undetected at
E > E ∼ . N H ∼ . × cm − ; i.e., Compton-thick absorption)very few photons are detected at E <
10 keV due to sig-nificant absorption and scattering. Moreover, for sourceswith N H > cm − , the entire high energy spectrumis down-scattered and, eventually, absorbed by the heav-ily Compton-thick material. Consequently, the observed X-ray flux in Compton-thick AGNs is often rendered so weakthat it becomes comparable to the X-ray emission aris-ing from the host-galaxy, making their detection extremelydifficult. The direct identification of mildly Compton-thickAGNs ( N H ∼ (1 . × cm − ) is possible through X-ray observations at E >
10 keV (e.g., using
Beppo -SAX,
Swift , Suzaku ) where the relatively unabsorbed high-energyemission can be detected. However, the sensitivities of cur-rent
E >
10 keV observatories are substantially limitedby high backgrounds, poor effective areas and inadequatespatial resolutions. Indeed, to date, only 18 Compton-thickAGNs have been unambiguously identified in the Universeat
E >
10 keV, mainly at z < ≈ .
01 (for a recent review, seeDella Ceca et al. 2008).In the absence of higher-energy
E >
10 keV data,the presence of a Compton-thick AGN may still be in-ferred using indirect methods: (1) from the detection of ahigh equivalent width ( > α fluorescence lineat E ∼ . > ≈
200 counts). For example, given the faint X-ray fluxes,even at low redshifts ( z ∼ . Chandra and
XMM-Newton ) are requiredto detect FeK α at a high significance. Indeed, only a further ≈
30 local ( z < .
01) AGNs have been robustly determinedto be Compton-thick AGNs in the absence of
E >
10 keVdata (see Comastri 2004; Della Ceca et al. 2008 and refer-ences there-in). Hence, although Compton-thick AGNs arepredicted to comprise a large proportion of the overall AGNpopulation ( > ≈
40 percent; Risaliti et al. 1999; Matt et al.2000), to date, only ≈
50 Compton-thick AGNs have beenrobustly identified in the nearby Universe at z < ≈ . ≈
30 candidate Compton-thick AGNs (i.e., those with highEW FeK or reflection-dominated spectra) have been identi-fied in high redshift X-ray surveys (e.g., Norman et al. 2002;Tozzi et al. 2006; Georgantopoulos et al. 2009). Given the required X-ray sensitivity to directly iden-tify Compton-thick AGNs using X-ray data alone, onlya small fraction of the population can be discovered us-ing current instrumentation. In recent years, new tech-niques have been developed to discover Compton-thickAGN candidates using complimentary wide-field opticalsurveys with pointed mid-infrared (IR) observations, al-lowing us to probe ≈ z – L X plane than using X-ray data alone. These ap-proaches are promising since the reprocessed mid-IR con-tinuum emission and high-excitation optical and mid-IRnarrow-line emission (i.e., [O iii ] λ v ] 14 . µ m;[O iv ] 25 . µ m) in AGN are relatively unaffected by theoptically-thick X-ray obscuring material in the central re-gion and, therefore, provide reliable measurements of theintrinsic luminosity of even the most heavily Compton-thick AGNs (e.g., Heckman et al. 2005; Panessa et al.2006; Mel´endez et al. 2008; Diamond-Stanic et al. 2009;Goulding et al. 2010). For example, through examinationof a local optically-selected AGN sample, Maiolino et al.(1998) and Bassani et al. (1999) find that those AGNswith X-ray–[O iii ] flux ratios of f X /f [OIII] < f X /f [OIII] < . Chandra -ACIS observations withoptical emission-line and mid-IR continuum luminosities toidentify six Compton-thick quasars ( L [OIII] > × L ⊙ )at z ∼ . Spitzer
IR spectroscopyand/or optical spectroscopy to identify high-redshift X-rayundetected Compton-thick AGNs in deep and wide-field sur-veys. Whilst each of these studies have successfully iden-tified Compton-thick AGNs using multi-wavelength analy-ses, they sample only the most luminous systems ( L X , intr > ≈ erg s − ) where the predicted space-density of Compton-thick AGNs, even at z ∼
2, is relatively low ( φ < ≈ − Mpc − ; Gilli et al. 2007). In order to clearly under-stand the evolution of these Compton-thick sources, it isvital to also identify the more modest luminosity popu-lation ( L X , intr ≈ [0 . × erg s − ), which comprisethe most energetically dominant AGNs in the nearby Uni-verse ( z ∼ .
1; e.g., Ueda et al. 2003; Ebrero et al. 2009;Aird et al. 2010).In this paper, we identify a sample of nearby ( z ∼ . Chandra
Deep c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . Figure 1.
Optical emission line diagnostic diagram presentingthe 2690 galaxies detected in the seventh data release of the SloanDigital Sky Survey with serendipitous X-ray coverage from
XMM-Newton at 0 . < z < . E > ii star-forming classification curve presented by Kauffmann etal. (2003) and the extreme starburst line of Kewley et al. (2001)are also shown with dash-dotted and dashed curves, respectively. Fields (CDFs; Giacconi et al. 2002; Alexander et al. 2003;Luo et al. 2008) at z ∼ . V ≈ . × Mpc ). We use pointed high signal-to-noise Spitzer
IR spectroscopy and 24 µ m photometry combined with theunprecedented wide-field coverage of the SDSS to explorethe ubiquity of typical Compton-thick AGNs at z ∼ . XMM-Newton
Serendipitous Survey ( ≈
100 deg ).These AGNs are all undetected to faint flux limits in E ∼ XMM-Newton observations, and based on their X-ray–[O iii ] flux ratio limits are likely to be heavily obscured(and possibly Compton thick). In section 3, we discuss thedata reduction techniques for our
Spitzer observations andpresent mid-IR AGN–starburst spectral decompositions tounderstand both the properties of the host galaxies and thecentral source in these obscured AGNs. In section 4, weuse mid-IR narrow-line emission and AGN-produced mid-IR continuum emission to determine the intrinsic luminos-ity of the obscured AGNs. We combine each of these AGNluminosity indicators in order to reliably identify whichsources are Compton-thick AGNs. We use these results tofurther constrain the ubiquity of Compton-thick AGNs at z ∼ .
1. Throughout, we adopt a standard ΛCDM cosmol-ogy of H = 71 km s − Mpc − , Ω M = 0 .
30, and Ω Λ = 0 . We select our candidate Compton-thick AGN sample onthe basis of their optical and X-ray properties. Sourcesthat are identified to be AGNs using traditional opticalemission line diagnostics (e.g., Baldwin et al. 1981) butare undetected to faint limits in wide-field
XMM-Newton observations (i.e., f X /f [OIII] <
1) are strong candidatesfor containing heavily obscured AGNs (e.g., Bassani et al.1999; Panessa & Bassani 2002; Akylas & Georgantopoulos2009). Here we provide the details behind the constructionof our sample of X-ray undetected optically identified AGNs(i.e., candidate Compton-thick AGNs).
We construct a parent sample of all optical spectroscopi-cally identified galaxies in the ≈
100 deg overlap region be-tween the seventh data release of the SDSS (Abazajian et al.2009; hereafter SDSS-DR7) and the second source catalogueof the XMM-Newton
Serendipitous survey (Watson et al.2009; hereafter 2XMMi). We define the redshift range for ourshallow wide-area sample based on the combined availablecosmological volume in the deep 2Ms “pencil-beam” CDF-North ( ≈
448 arcmin ; Alexander et al. 2003) and CDF-South ( ≈
436 arcmin ; Giacconi et al. 2002; Luo et al. 2008)surveys. At z ∼ . V ∼ . × Mpc , which is equivalent to thecomoving volume in the redshift range of z ∼ . The SDSS-DR7 is currently the largest publicly availableoptical spectroscopic catalogue ( ≈ ) containing929,555 spectroscopic source redshifts. Previous studieshave used past data releases of the survey to show thatthrough careful spectral analyses, general galaxy and AGNproperties can be derived from these large datasets (e.g.,Kauffmann et al. 2003; Heckman et al. 2004; Greene & Ho2007). We select all galaxies with well detected narrow[O iii ] λ α , [N ii ] λ > All galaxies with detected broad Balmer emission lines(here defined as a full-width half maximum >
700 km s − )are removed as these sources are unlikely to be intrin-sically obscured by a gas/dust-rich geometrically thicktorus. AGNs which are heavily obscured are often foundto be hosted in dust-rich galaxies, and thus are likelyto be strongly reddened (i.e., H α –H β ratios ≫ .
1; e.g., Throughout this manuscript we define
X-ray undetected asthose sources undetected in the hard band (
E > SDSS spectra are obtained through 3 arcsecond fibers; at themedian redshift of our sample ( z ∼ .
08) this projected aperture isequivalent to a physical region of ≈ (cid:13) , 1–16 A.D. Goulding et al.
Goulding & Alexander 2009). Hence, whilst useful in unam-biguously discriminating between the properties of galaxies(e.g., Kauffmann et al. 2003; Wild et al. 2010), we purposelydo not limit our selection to only galaxies with well-detectedH β emission. Sources are separated by classification basedon their optical emission-line ratios in a traditional diagnos-tic diagram (hereafter, BPT diagram; e.g., Baldwin et al.1981). We conservatively identify the narrow-line AGNs inthe SDSS-DR7 as those which lie above the theoretical star-burst limit of Kewley et al. (2001). See Fig. 1. The 2XMMi catalogue identifies all X-ray sources detectedin the 3491 observations made during the first ≈ XMM-Newton operations (Watson et al. 2009). Its un-precedented sky coverage (360 deg ) and sensitivity (me-dian XMM-Newton exposures of 20–50 ks) currently pro-vides an exceptional resource for the unbiased identificationof obscured AGN activity throughout the Universe. Usingan automated reduction and analysis pipeline, the 2XMMicatalogue provides source positions, exposure times, X-rayfluxes and band ratios of all detected sources which serendip-itously fall within the field-of-view of previous
XMM-Newton observations.All sources in our SDSS parent sample are matched to2XMMi using a 3.7 arcsecond radius, which is chosen as agood compromise between maximising source numbers andminimising the probability of spurious matches (e.g., Watsonet al. 2009). The matching algorithm is restricted to sourceswithin 14 arc-minutes of the aim point of each
XMM-Newton observation to minimise the likelihood of spurious matchesdue to the degradation of the X-ray PSF far off-axis. Basedon the X-ray/optical positional analysis of SDSS quasarsand the 2XMMi catalogue by Watson et al. (2009), if we as-sume no systematic offsets, then we expect our XMM–SDSSmatching to be ≈
92 percent complete. We identify all op-tical sources in the 2XMMi which have 3 σ detections at E ∼ XMM-Newton observation but are undetected in the hard-bandof the 2XMMi catalogue), we use flix to compute robust3 σ (likelihood threshold of 6.6) X-ray upper-limits in thisband. Using a sub-sample of the matched sources that areformally undetected in the hard band in 2XMMi, we testedthe use of flix to provide X-ray upper-limits. Broadly, wefind that the upper limits provided by flix are consistentwith the fluxes within ± σ given by 2XMMi. For the sake ofcomparison with previous studies, we convert these 2–12 keVupper-limits to 2–10 keV limits assuming a powerlaw spec-trum with spectral index of Γ = 1 . F ν ∝ ν − (Γ − ;Γ = 1 . z ∼ . flix is a purpose-built program provided by the XMM-Newton
Figure 2.
Intrinsic obscuration optical–X-ray diagnostic diagramfor AGNs. Contours indicate the distribution of all optical AGNsin the SDSS-DR7 which lie in the
XMM-Newton footprint (i.e.,those galaxies which lie above the Kewley et al. 2001 extremestarburst line presented in Fig. 1). These contours enclose 50%and 90% of the whole X-ray undetected parent sample; outliersare shown with small arrows. We select 14 candidate Compton-thick AGNs for
Spitzer observations based on their f X /f [OIII] ratio (filled circles). Four of these sources lie in the region exclu-sively occupied by Compton-thick AGNs ( f X /f [OIII] < .
1; e.g.,Bassani et al. 1999), and 10 are selected from the heavily obscured N H region ( f X /f [OIII] ∼ . f X /f [OIII] ratiofrom AG09 and for comparison, four well-studied local ‘bona-fide’Compton-thick AGNs (Circinus, Mrk 3, NGC 1068 and 6240; datais taken from Bassani et al. 1999; stars). with complimentary hard X-ray XMM-Newton coverage is2690 objects (272 are hard X-ray detected sources). The me-dian redshift of the sample is ≈ .
09. Of these galaxies, 334( ≈
12 percent) lie above the theoretical starburst limit andare classified as optical narrow-line (NL) AGNs (i.e., 101are X-ray detected and 233 are X-ray undetected AGNs; seeFig. 1).
Assuming the optical emission-lines and X-ray AGNemission are well-correlated (e.g., Mulchaey et al. 1994;Alonso-Herrero et al. 1997), sources which are optically clas-sified as AGNs but are undetected to faint limits in relativelydeep X-ray observations are likely to be those with heavilyattenuated X-ray emission, similar to the objects currentlymissed in deep X-ray surveys. From our well-defined parentsample of 334 optical NL AGNs, 233 are not detected inthe hard-band of the 2XMMi catalogue. In this section, weoutline the selection method for our sample of 14 hard-bandundetected candidate Compton-thick AGNs.The de-reddened [O iii ] luminosity is assumed to bea good tracer of AGN power (e.g., Heckman et al. 2005; c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . Table 1.
Basic source properties of the candidate Compton-thick AGNsID Source name α J δ J z D L log( M BH ) [O iii ] [NII] H α A V log( L [OIII] ) log( L HX )(deg) (deg) (Mpc) ( M ⊙ ) H β H α H β (mags) ( erg s − ) ( erg s − )(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)1 SDSS J094046+033930 145.19287 3 . .
94 20.55 1.11 16.61 4.86 42.94 < .
402 SDSS J094506+035551 146.27664 3 . .
68 11.93 0.51 4.68 1.20 42.73 < .
393 SDSS J100328+554154 150.86636 55 . .
53 13.89 0.82 4.72 1.22 42.39 < .
364 SDSS J101757+390528 154.48708 39 . .
29 4.59 0.55 4.94 1.35 41.33 < .
195 SDSS J102142+130550 155.42455 13 . .
93 4.32 0.68 9.69 3.30 41.83 < .
236 SDSS J111521+424217 168.83784 42 . > .
63 0.57 > . > .
88 44.68 < .
157 SDSS J115658+550822 179.24117 55 . .
72 17.04 1.10 7.40 2.52 42.96 < .
628 SDSS J121355+024753 183.47861 2 . .
77 6.44 0.62 6.30 2.05 42.49 < .
239 SDSS J123026+414258 187.60876 41 . .
50 8.82 0.70 8.53 2.93 42.58 < . − − . < .
30 11.66 0.62 6.34 2.07 42.07 < . . .
85 3.48 0.77 9.78 3.32 41.48 < . . .
60 9.14 0.79 12.09 3.94 42.32 < . − − . .
14 9.43 0.43 4.41 1.02 41.61 < . . .
36 14.53 1.85 11.15 3.71 41.80 < . NOTES: (1) Source-identification; (2) SDSS source name; (3–4) J2000 positional co-ordinates from the SDSS-DR7; (5) SDSS spectroscopicredshift; (6) luminosity distance in megaparsecs calculated using our adopted cosmology; (7) logarithm of black hole mass in units ofsolar masses derived from the stellar velocity dispersion (MPA-JHU DR7 release) using the M – σ relation (Gebhardt et al. 2000); (8–10)Emission-line ratios from SDSS-DR7; (11) Implied A V in magnitudes derived from the Balmer decrement ( Hα/Hβ ; see section 2.1.3);(12) Logarithm of dust extinction corrected [O iii ] luminosity; (13) Logarithm of the 3- σ upper limit of the 2-10 keV X-ray luminosityderived from fluxes produced using flix (see footnote 4). Netzer et al. 2006; Panessa et al. 2006). In order to iden-tify obscured AGN candidates, we follow Maiolino et al.(1998) and Bassani et al. (1999) by using the flux ratio ofde-reddened (intrinsic) [O iii ] and observed (absorbed) 2–10keV X-ray emission, and compare it to the [O iii ] /H β ratio ina new diagnostic diagram analogous to a BPT diagram; seeFig. 2. Optical luminosities are corrected for dust-reddeningtowards the AGN NL-region using the Balmer decrement(i.e., the observed H α –H β ratio; Ward et al. 1987), an in-trinsic ratio of 3.1 (Osterbrock & Ferland 2006) and a stan-dard R = 3 . iii ] lu-minosities in the range L [OIII] ≈ (0 . × erg s − (median L [OIII] ≈ erg s − ). AG09 find that the averageX-ray–[O iii ] flux ratio for unobscured AGNs (i.e., Type 1s; N H < cm − ) is ≈
30, whilst heavily obscured AGNs(Type 2s) typically exhibit lower values of f X /f [OIII] . AGNswith f X /f [OIII] < . f X /f [OIII] ∼ . . ≈
63 percent) of the 233 X-ray undetected AGNs have f X /f [OIII] < . f X /f [OIII] < .
1) from whichwe select a representative sub-sample of 14 ( ≈
10 per-cent) to be further investigated using pointed mid-IR spec-troscopic and photometric observations. Our sample of 14AGNs are well-matched to the parent sample of X-ray un-detected AGNs with a redshift distribution of 0.03–0.2 (me-dian ∼ .
08) and L [OIII] ≈ (0 . × erg s − (me-dian L [OIII] ≈ × erg s − ). For completeness, we alsonote that 11/14 of our sources are detected in at least oneof the softer X-ray bands ( E <
E > σ upper-limits would only serve to reducethe current hard-band limits. The basic source properties forour sample of 14 candidate Compton-thick AGNs are shownin Table 1. We have used the
Spitzer
Infra-Red Spectrograph (IRS) andMulti-band Imaging Photometry for
Spitzer (MIPS) to ob-serve the 14 candidate Compton-thick AGNs selected inSection 2 (PID:50818; PI: D.Alexander). In this section,we present the reduction methodology and resulting spec-troscopy and photometry for these 14 targets. In order to ro-bustly assess the intrinsic luminosity of the central sources inthese AGNs, we also present mid-IR spectral decompositionanalyses to isolate the AGN continuum and star-formationemission.
Spitzer -IRS Spectral Reduction and Analysis
Each of the 14 candidate Compton-thick AGNs were ob-served in spectral staring mode with the low-resolutionmodules (short-low [SL; 5.2–14 . µ m] and long-low [LL;14.0–38 . µ m]; R ≈ Spitzer -IRS instrument(Houck et al. 2004). The sources were observed between30th November 2008 and 24th February 2009 using rampdurations of 60 seconds ×
10 (4) cycles and 120 seconds × c (cid:13) , 1–16 A.D. Goulding et al.
Figure 3.
Rest-frame low-resolution
Spitzer -IRS spectroscopy of the 14 X-ray undetected SDSS AGNs in our sample. The prominentemission-line and PAH features that may be detected are highlighted. images produced by the S18.7.0
Spitzer
Science Center(SSC) pipeline were retrieved and further analyzed usingour custom idl reduction routine (see Goulding 2010). In-dividual BCDs were rigorously cleaned of rogue ‘hot’ pixelsusing a customised version of irsclean . Next, individualrows were fit as a function of time to remove latent chargewhich exists on the detector after bright and/or long obser-vations. The cleaned BCDs were averaged in the differingnod positions, which were then used to perform alternatebackground subtractions of the source in each nod position.Spectral extraction was performed using the
Spitzer -IRS Custom Extraction ( spice ) software provided by theSSC. The spectroscopy was extracted using an optimallycalibrated 2-pixel wide spectral window to maximize thesignal-to-noise ratio of each spectrum. Flux uncertaintieswere estimated for each of the spectra using a second spec-tral window offset from the source in the spatial direction.The spectra for each of the modules for an individual objectwere corrected for their differing apertures and normalizedto the flux level of the 1st LL module. Orders were clippedof spectral noise (see the
Spitzer -IRS handbook for furtherinformation) and stitched together by fitting low-order poly-nomials to produce the final spectra.
Spitzer -MIPS Reduction and Analysis
In order to measure accurate emission-line and continuumfluxes we use
Spitzer -MIPS 24 µ m photometry to flux cal-ibrate the IRS spectroscopy. Eleven of our 14 Spitzer -IRStargets were observed with the MIPS photometer. The re-maining three sources were scheduled but were not observedbefore the depletion of the instrument’s cryogenic liquidcoolant. BCDs were retrieved and the data reduced using theSSC analysis program mopex . We post-process individualBCD frames to remove common MIPS artifacts (i.e., “jail-bars”, latents etc.) and suppress large and small-scale gradi-ents using master-flat images generated from the initial data.Processed frames were then background matched, stacked,mosaicked and median filtered using mopex to create thefinal background-subtracted reduced image.Point source extraction was performed using the SSCprovided
Spitzer
Astronomical Point Source EXtraction( apex ) software to produce 24 µ m aperture photometricfluxes of the sources in the reduced MIPS frames (see col-umn 5 of Table 2). The IRS spectra were convolved with the24 µ m MIPS response curves and compared to the photo-metric fluxes to produce absolute flux calibrated IRS spectraof each source. The average upwards photometric correctionrequired to the spectroscopy was a factor of ≈ .
3. For thethree objects lacking MIPS observations (see Table 2) we didnot attempt to correct these spectra. Hence, we consider theemission-line and continuum fluxes (derived in sections 3.3and 3.4) for these three AGNs to be less accurate and mostlikely conservative lower-limits.
The reduced and flux-calibrated mid-IR spectra producedin the previous sections were further analyzed (i.e., fittingof emission-lines and polycyclic aromatic hydrocarbon fea-tures) using the
Spitzer spectral analysis program, smart (Higdon et al. 2004). Typical AGN dominated emission linespresent in the spectra of these sources included [Ne v ] c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . ( λλ . . µ m) and [O iv ] +[Fe ii ] ( λλ . . µ m). Additionally, AGN and star-formation produced lines suchas [Ne ii ] ( λ . µ m) and [Ne iii ] ( λ . µ m) were alsopresent. See Table 2 for the Spitzer -IRS derived propertiesand see Fig 3 for the final
Spitzer -IRS spectra.Mid-IR high-excitation narrow-line emission such as[Ne v ] and [O iv ] (ionisation potentials of 97.1 eV and 54.9 eV,respectively) have been shown to be excellent extinction-free unambiguous indicators of the bolometric luminosity ofan AGN (e.g., Mel´endez et al. 2008; Diamond-Stanic et al.2009; Goulding et al. 2010). From analysis of our Spitzer -IRS spectroscopy, we find that six of the AGNs in our sam-ple have detected [Ne v ] emission and all 14 have detected[O iv ] +[Fe ii ]. The [Ne v ] and [O iv ] luminosities for the sixAGNs with detected [Ne v ] are well correlated and lie withinthe intrinsic scatter of Equation 2 of Goulding & Alexander(2009) and the more recent calibration of Weaver et al.(2010); for the sources with [Ne v ] 3 σ upper-limits, we findthat the majority of the fluxes are also consistent with theserelationships. Hence, we confirm that the [O iv ] emission inthese particular sources is likely to be a good indicator ofthe intrinsic luminosity of the AGN. For AGNs with strongcontributions from star-formation, the [Fe ii ] emission maycontaminate the measured [O iv ] flux measured from low-resolution Spitzer -IRS spectroscopy (Mel´endez et al. 2008).For those galaxies which we find to be dominated by starformation at mid-IR wavelengths (i.e., AGN contributionsof <
50 percent; see Section 3.4), we conservatively apply asmall downwards correction factor of ≈ . iv ] flux to account for the [Fe ii ]contamination. Our final adopted [O iv ] luminosities coverthe range, L [OIV] ≈ (0 . × erg s − . The low-resolution mid-IR spectra of typical Type-2 AGNsat rest-frame λ ≈ µ m are composed of three primarycomponents: 1) a power-law like thermal AGN dust con-tinuum; 2) a star-formation component which arises fromthe super-position of PAH features; and 3) a silicate ab-sorption feature at λ ≈ . µ m produced by the hot dustcontinuum being absorbed by cooler dust on parsec scales(e.g., Goulding & Alexander 2009; Gallimore et al. 2010;Mullaney et al. 2010; Tommasin et al. 2010). Therefore, asexpected, we find that the mid-IR spectra for the majority ofour candidate Compton-thick AGNs contain both an AGNproduced continuum and strong polycyclic aromatic hydro-carbon (PAH) features, which are associated with starburstactivity in the circumnuclear photodissociation regions ofthe host galaxy. Here we outline our spectral decompositionroutine to determine the relative contributions of starburst(SB) activity and the AGN continuum in our 14 Compton-thick AGN candidates (i.e., the SB:AGN ratio), and measurethe intrinsic luminosity of the central source from the AGNproduced mid-IR continuum at 6 µ m (e.g., Lutz et al. 2004). We note that due to the spectral resolution of the LL modules,the [O iv ] and [Fe ii ] emission lines cannot be individually resolved. We note that the identification of some AGNs even in the verynearby Universe can often require extremely high signal-to-noise,high-resolution mid-IR spectroscopy (e.g., Satyapal et al. 2008;Goulding & Alexander 2009).
Using a purpose-built idl -based routine, we fit the IRSspectroscopy for each of the 14 candidate Compton-thickAGNs with a combined standard starburst template and anAGN power-law component (with spectral index, k ) con-volved with a Draine & Li (2007) extinction curve ( ρ ( λ )) ofthe form, f AGN ( λ ) = aλ k exp[ − bτ ρ ( λ )] (1)where a , b and k are constants, and τ is the optical depth. Within the fitting we use four possible starburst templateswhich cover a realistic range of physical and theoreticalscenarios: 1) low-resolution
Spitzer -IRS spectroscopy of thearchetypal starburst galaxy, M82; 2) a combined
Spitzer -IRS starburst template of local pure star-forming galaxiespresented in Brandl et al. (2006);
3) a theoretical radia-tive transfer model of a pure circumnuclear starburst regionat r ≈ L IR ≈ L ⊙ (Siebenmorgen & Kr¨ugel2007; hereafter, SK07); and 4) a theoretical radiative trans-fer model of a nuclear star cluster at r < .
35 kpc with L IR ≈ L ⊙ (SK07).The best resulting model parameters derived from theminimum Chi-squared fit to the IRS data are given in Ta-ble 2 and shown in Fig. 4. We note that none of the AGNsin our sample have mid-IR spectral features which are con-sistent with the theoretical nuclear star cluster model, andhence, the best-fit spectral model for each are that of anAGN combined with one of the three circumnuclear star-burst templates. Based on the mid-IR spectral-fits, we alsoderive the approximate contribution of the AGN to the mid-IR emission for each of the sources; see column (7) of Table 2.We find that although these AGNs were selected to be strong[O iii ] and [N ii ] emitters (i.e., optically-dominated Seyfertgalaxies), the mid-IR spectra of ≈
50 percent of the sourcesare consistent with being dominated by star-formation activ-ity. Indeed, on the basis of these spectral decompositions, themid-IR spectroscopy for one source (SDSS J102142+130550)is consistent with there being no AGN component, despitethis source clearly being identified as an AGN at opticalwavelengths.In order to estimate the intrinsic AGN luminosity andhence place limits on the X-ray absorption in these sources,we use the measured AGN power-law parameters to derive6 µ m luminosities ( L µm ). The uncertainties of these 6 µ mfluxes are established by considering a weighted spread inthe measured 6 µ m fluxes from all statistically valid star-burst template fits (i.e., we reject all statistically poor fits atthe 95 per cent level). The mean 1 σ uncertainty is ≈ . µ m continuum luminosities for 13of our 14 candidate Compton-thick AGNs and conserva-tively estimate an upper limit for the 6 µ m continuumflux of < − mJy for SDSS J102142+130550. We findusing our adopted cosmology, that the AGNs cover more Our idl routine is based on that used by Mullaney et al. (2010b)and makes use of the Markwardt 1-dimensional Chi-squared anal-ysis library, see http://cow.physics.wisc.edu/ ∼ craigm/idl/ forfurther details. We note that as we require only starburst emission in thesetemplates, we do not include galaxies in the combined Brandlstarburst template which have any previous evidence for AGNactivity (i.e., Mrk 266, NGC 660, 1097, 1365, 3628 and 4945).c (cid:13) , 1–16
A.D. Goulding et al.
Figure 4.
Spectral decompositions of the
Spitzer -IRS spectra (blue solid curve) for our candidate Compton-thick AGNs produced byour spectral analysis program as described in section 3.4. The grey shaded region indicates the 1 σ flux uncertainties to the observedspectrum. The best-fit absorbed power-law and starburst template are shown with dotted and dashed curves, respectively. See Table 2for best-fitting parameters. The total best-fit spectrum (i.e., power-law+starburst+emission-lines) is shown with a solid red curves.c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . Figure 4 – continued than 2 decades in 6 µ m luminosity, with νL µm ≈ (0 . × erg s − . We have selected a sample of 14 [O iii ] bright, X-ray unde-tected AGNs from the ≈
100 deg overlap region betweenthe SDSS-DR7 and 2XMMi surveys. These sources lie at z ∼ . L [OIII] ≈ (0 . × erg s − (i.e.,similar to those of typical nearby Seyfert galaxies). Our14 targets all have f X /f [OIII] <
1, implying strong intrin-sic absorption of their X-ray flux and many (possibly all)are likely Compton-thick AGNs. In the absence of X-rayspectroscopic data, in section 3 we derived AGN-produced emission line and continuum luminosity measurements inorder to independently constrain the intrinsic luminosity ofthese candidate Compton-thick AGNs. In this section, weuse these intrinsic luminosities in conjunction with X-rayconstraints from
XMM-Newton data to test whether theseobjects are indeed Compton-thick AGNs. We then use theseresults to place new constraints on the space density andrelative mass-accretion rates of Compton-thick AGNs in thenearby Universe ( z ∼ . z ∼ . iii ] λ c (cid:13) , 1–16 A.D. Goulding et al.
Table 2.
Measured AGN propertiesID [Ne v ] λ .
32 [Ne v ] λ .
32 [O iv ] λ . S µ m SB AGN S µ m log( L X , [OIV] ) log( L X , µ m ) log( η ) C-thick?(erg s − cm − ) (erg s − cm − ) (erg s − cm − ) (mJy) Model cont. (mJy) ( erg s − ) ( erg s − ) [O iv ] 6 µ m(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)1 < . < .
77 3 . ± .
81 3 . ± .
04 2 40% 1 . ± .
14 42.67 42.74 -2.24 y y2 10 . ± .
43 6 . ± .
09 29 . ± .
05 17 . ± .
05 2 55% 3 . ± .
19 44.42 43.58 -0.94 Y y3 4 . ± .
78 4 . ± .
78 6 . ± .
03 8 . ± .
03 3 92% 2 . ± .
02 43.63 43.30 -1.02 y ?4 3 . ± .
55 3 . ± .
54 12 . ± .
64 - 3 38% 1 . ± .
76 42.76 42.08 -2.28 Y ?5 < . < .
00 6 . ± .
62 8 . ± .
05 3 < . < .
01 42.80 - - Y -6 < . < .
78 5 . ± .
91 6 . ± .
03 2 72% 1 . ± .
17 43.88 43.34 - Y y7 3 . ± .
98 2 . ± .
95 10 . ± .
38 10 . ± .
10 2 53% 2 . ± .
31 43.26 42.74 -1.92 Y y8 7 . ± . < .
19 19 . ± .
03 65 . ± .
06 3 38% 5 . ± .
99 43.26 43.01 -0.63 Y Y9 < . < .
47 5 . ± .
90 - 3 66% 8 . ± .
33 43.45 43.69 -0.50 Y Y10 < . < .
32 6 . ± .
92 6 . ± .
04 1 85% 1 . ± .
11 42.91 42.52 -2.75 y ?11 < . < .
39 12 . ± .
92 39 . ± .
03 3 12% 3 . ± .
71 42.26 42.06 -1.87 y y12 < . < .
63 4 . ± .
76 6 . ± .
03 3 9% 0 . ± .
40 43.35 41.91 -2.79 ? ?13 57 . ± .
16 42 . ± .
52 180 . ± .
01 106 . ± .
55 1 81% 16 . ± .
20 44.24 43.23 -0.73 Y Y14 < . < .
96 3 . ± .
51 - 2 62% 0 . ± .
09 42.09 41.44 -2.10 Y ?
NOTES: (1) SDSS source-identification. (2-4) Fluxes and their 1 σ uncertainty for the measured AGN-produced mid-IR emission linesin units of 10 − erg s − cm − . (5) MIPS 24 µ m flux density in units of mJy. (6) Best-fit starburst model in the mid-IR spectralmodeling analysis (see section 3.4) - 1: M82; 2: Combined SB (Brandl et al. 2006); 3: Circumnuclear starburst at 3 kpc (SK07). (7)Inferred AGN contribution to the mid-IR emission at λ ∼ µ m. (8) Unabsorbed AGN 6 µ m flux and 1 σ uncertainty in units ofmJy. (9-10) Logarithm of the estimated 2–10 keV luminosity from the AGN-produced [O iv ] and 6 µ m emission (see sections 3.3, 3.4and 4.1 for further details). (11) Logarithm of implied Eddington ratio derived from the estimated 2–10 keV luminosity using 6 µ memission, and the SMBH masses presented in Table 1. (12-13) Indication for whether the source is consistent with being Compton-thick ( N H > . × cm − ) on the basis of the mid-IR indicators: ‘Y’ - conservatively identified to be Compton-thick, ‘y’ -less-conservatively identified to be Compton-thick, ‘?’ - inconclusive due to depth of current X-ray data (i.e., they are upper-limits; seesection 4.1). their intrinsic luminosity ( L AGN ; e.g., Mulchaey et al.1994; Alonso-Herrero et al. 1997; Heckman et al. 2005;Panessa et al. 2006). However, such emission may also bereadily excited by strong star formation as well as beingsubject to significant dust extinction within the host galaxy.By contrast, mid-IR high-excitation narrow-line emission(e.g., [Ne v ]; [O iv ]) is an excellent extinction-free indicatorof L AGN (see Section 3.3) and, when combined with sensi-tive X-ray data, can provide good first order constraints onwhether an AGN is Compton thick.In Fig. 5a we present the observed 2–10 keV X-rayupper-limit luminosities from the
XMM-Newton data versusthe mid-IR [O iv ] luminosity for our candidate Compton-thick AGNs and compare them to the intrinsic proper-ties found for local ‘bona-fide’ Compton-thick AGNs. Wefind that the candidate Compton-thick AGNs are spreadover a wide range of [O iv ] luminosities, L [OIV] ≈ (0 . × erg s − ; for the sources in our sample which wefind to be dominated by SF at mid-IR wavelengths (column7 of Table 2), [O iv ] fluxes have conservatively been adjustedfor contamination from [Fe ii ] emission (see section 3.3). Wefind that based on the observed X-ray upper-limits, noneof the objects in our sample are consistent with the localintrinsic relation of AGNs from Goulding et al. (2010), sug-gesting that the X-ray emission is heavily obscured. How-ever, as we illustrate in Fig. 5a, the observed L X /L [OIV] for our sample (mean ratio ≈ .
6) is consistent with theobserved L X /L [OIV] ratio for a sample of well-studied lo-cal ‘bona-fide’ Compton-thick AGNs (i.e., Circinus, Mrk 3,NGC 1068 and NGC 6240). Furthermore, the L X /L [OIV] luminosity ratio of these four Compton-thick AGNs is con-sistent with the L X – L [OIV] relationship of Goulding et al.(2010) when the X-ray data is corrected for the absorptionimplied from high-quality X-ray spectroscopy. Assuming the[O iv ] emission is indeed an isotropic AGN indicator (e.g.,Melendez et al. 2008; Diamond-Stanic et al. 2009; Gouldinget al. 2010), this suggests that by comparing the observedX-ray upper-limit to the intrinsic X-ray luminosity as pre-dicted by our [O iv ] measurements ( L x , [OIV] ), we may inferwhether the sources in our sample are indeed Compton-thickAGNs.Based on Compton reflection models, Alexander et al.(2008) predict that the observed–intrinsic X-ray flux ratioin the 2–10 keV band for a Compton-thick AGN with N H ∼ . × cm − is f X , intr /f X , obs ≈
15. Hence, we predict thatsources with f X , intr /f X , obs > ≈
15 are likely to be obscuredby Compton-thick material. For the 14 candidate Compton-thick AGNs in our sample, we calculate f X , [OIV] using thelocal X-ray–[O iv ] relation of Goulding et al. (2010). Wepredict intrinsic X-ray luminosities of L X, predict ≈ (0 . × erg s − (see Column 9 of Table 2). We find that13 ( ≈
90 percent) of the sources exhibit f X , [OIV] /f X , obs > ≈ ≈ . ≈
65 percent) of our candidate Compton-thick AGNs could be genuine Compton-thick AGNs. It isalso prudent to note that as the observed X-ray fluxes forall of these sources are upper-limits, we cannot exclude the c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . Figure 5. Left; (a):
Rest-frame 2–10 keV X-ray luminosity versus mid-IR [O iv ] ( λ . µ m) luminosity ( L [OIV] ) for our sample ofCompton-thick AGN candidates (filled circles). We show the local X-ray–[O iv ] relation of Goulding et al. (2010) and find on the basisof L [OIV] that the majority of our sample are consistent with the observed X-ray luminosity being absorbed by a factor >
15 (i.e., N H > . × cm − ). Right (b):
Rest-frame 2–10 keV X-ray luminosity versus the mid-IR AGN continuum luminosity at 6 µ mfor the Compton-thick AGN candidates with measured 6 µ m luminosities (filled circles). A comparison sample of z ∼ µ m relation of Lutz et al. (2004) and theluminosity-dependent relation of Fiore et al. (2009) to predict the region of parameter space where Compton-thick AGN lie. We find thaton the basis of 6 µ m luminosities, many of the sources in our sample are likely to be Compton-thick AGNs. For comparison, in both (a)and (b) we additionally highlight the observed (filled stars) and absorption-corrected (open stars) X-ray luminosities for 4 well-studied‘bona-fide’ Compton-thick AGNs (Circinus, Mrk 3, NGC 1068 and 6240). In both figures, the observed X-ray luminosities for these‘bona-fide’ Compton-thick AGNs occupy roughly the same region of parameter space as our sample but our objects are ≈ possibility that all of the sources in our sample are Compton-thick AGNs as the implied f X , [OIV] /f X , obs ratio is a lower-limit.By combining multiple indirect AGN luminosity indi-cators, particularly those which probe different regions ofthe central engine, we can place even stronger constraintson whether the AGNs in our sample are Compton thickthan using narrow-line emission alone. The 6 µ m contin-uum luminosity has been shown to provide a good proxyfor the intrinsic AGN luminosity (e.g., Lutz et al. 2004;Maiolino et al. 2007; Treister et al. 2008; Fiore et al. 2009).In Fig. 5b we again present the observed 2–10 keV X-rayupper-limit luminosities from XMM-Newton data but nowcompare these luminosities to the AGN continuum lumi-nosity at 6 µ m derived in Section 3.4 and the luminosity-dependent (Fiore et al. 2009) and luminosity-independent(Lutz et al. 2004) relations derived using high-quality X-raydata and mid-IR Spitzer
IRAC photometry and ISO spec-troscopy, respectively. As noted in Section 3.4, one of our14 candidate Compton-thick AGNs is consistent with therebeing little or no mid-IR emission from an AGN continuumat λ ∼ µ m, and we remove this AGN from furtheranalyses in here.We conservatively adopt the slightly lower-luminosityX-ray–6 µ m relationship of Lutz et al (2004) to infer the intrinsic X-ray luminosities of the candidate Compton-thick AGNs. We estimate intrinsic X-ray luminosities of L X, predict ≈ (0 . × erg s − (see Column 10 of Ta-ble 2). Eight out of the 13 ( ≈
60 percent) 6 µ m detectedsample members lie in the region expected for Compton-thick AGNs (i.e., N H > ≈ . × cm − ; see Column 13of Table 2). However, if we were to adopt the relation-ship of Fiore et al. (2009) this Compton-thick AGN fractionwould increase to 9/13 sources ( ≈
80 percent; i.e., consis-tent with that found when using [O iv ] as a N H diagnos-tic). Furthermore, if we account for the intrinsic scatter ofwithin the Lutz et al (2004) relationship ( ≈ . ≈
20 percent)sources must be Compton thick.In Columns 12 and 13 of Table 2 we summarise whetherwe identify the sources to be Compton-thick AGNs on thebasis of their combined X-ray and mid-IR properties. Weconsider those AGNs which are conservatively identified asCompton-thick AGNs (i.e., those which lie below the regionof intrinsic scatter derived from the L X – L [OIV] and L X – L µm relationships) in at least one of the mid-IR diagnosticsand are also below the standard Compton thick thresholdin the other mid-IR diagnostic to be genuine Compton-thick c (cid:13) , 1–16 A.D. Goulding et al.
AGNs (i.e., in the nomenclature of Table 2, only those AGNswith Y-Y, Y-y or y-Y). This is a reasonable and conservativeassumption to make if we consider the ‘bona-fide’ Compton-thick AGNs shown in Fig. 5; all of these AGNs would beidentified to be Compton-thick AGNs (y/Y) on the basis of[O iv ] emission and 3/4 on the basis of 6 µ m emission. Hence,using our adopted definition, we find that 6/14 ( ≈
43 per-cent) of our candidate Compton-thick AGNs are very likelygenuine Compton-thick AGNs on the basis of their com-bined mid-IR properties. Under the reasonable assumptionthat our sample of candidate Compton-thick AGNs is a rep-resentative sub-sample of the parent X-ray undetected pop-ulation of AGNs in the SDSS-DR7 (i.e., in both redshiftand luminosity parameter space; see section 2), these resultsimply that at least ≈ ±
21 percent of the sources with f X /f [OIII] < We note that on the ba-sis of our most conservative and most optimistic Compton-thick diagnostic thresholds, ≈ f X /f [OIII] < f X /f [OIII] < . > ≈
15. Hence, these are very likely to be Compton-thickAGNs. Whilst the [O iv ] emission from the fourth AGN(SDSS J221742+000908) is consistent with a Compton-thickAGN (see Fig. 5a) we find little evidence for this on thebasis of its 6 µ m continuum luminosity. Indeed, the ob-served L X, − keV appears to be comparatively unabsorbedon the basis of AGN continuum luminosity. However, wenote that we would also find a similar result for the ‘bona-fide’ Compton-thick AGN, Mrk 3. We find that five of the10 AGNs with f X /f [OIII] ∼ . iv ] and 6 µ m luminosities suggestingstrong absorption of the X-ray emission, as well as evi-dence for silicate absorption at 9 . µ m. By contrast, forthe AGN which does not appear to be clearly Compton-thick on the basis of either of our mid-IR AGN indicators(SDSS J142931+425149), we find that the underlying AGNcontinuum in this sources is consistent with an unabsorbedpower-law (i.e., no evidence for silicate absorption). z ∼ . > ≈ ±
21 percent)of our sample of 14 X-ray undetected optical narrow-lineAGNs with L X /L [OIII] < N H > ≈ . × cm − (i.e., they areCompton-thick AGNs). Assuming that our sample of can-didate Compton-thick AGNs is representative of the parentpopulation, we may use our derived Compton-thick AGNfraction to infer at least a lower limit for the number, andhence space density, of Compton-thick AGNs at z ∼ . Uncertainties are calculated using standard Poisson countingstatistics.
Figure 6.
Space density of Compton-thick AGNs compared withthe XRB synthesis models of Gilli et al. (2007; solid curves) andTreister et al. (2009; dashed curves) for intrinsic X-ray luminosi-ties of L X > ,10 ,10 and 10 erg s − . Data points referto this work (solid circle; L X > erg s − ) and those fromcomparable studies of z ∼ L X > erg s − ) and luminous Compton-thick quasars at z ∼ . L X > . × erg s − ). The total number of Type-2 AGNs in the SDSS-2XMMioverlap region is 334 at z ∼ . ≈
45 per-cent) of these AGNs are X-ray undetected in 2XMMi with L X /L [OIII] < ≈ ±
31 of the sources in our parent sam-ple with L X /L [OIII] < ≈
20 percent of all Type-2 AGNs in the SDSS-2XMMi over-lap region. The comoving volume encompassed by our sur-vey is V ≈ . × Mpc . Hence, we estimate a space-density of Compton-thick AGNs at z ∼ . ≈− . +0 . − . Mpc − for luminosities of L X > ≈ erg s − . How-ever, given that both our sample selection and methodfor identifying Compton-thick AGNs are based on X-rayflux upper-limits (i.e., those sources with L X /L [OIII] > L X /L [OIII] < Spitzer -IRS sample may still be Comptonthick), the space-density found here should be considereda strict lower-limit to the number of Compton-thick AGNsin our parent sample.In Fig. 6 we compare our estimated space-density tothe similar study of SDSS-selected Compton-thick quasarsof Vignali et al. (2010) and to those derived from theXRB synthesis models of Gilli et al. (2007) and Treisteret al. (2009). Both models invoke a large population ofCompton-thick AGNs in order to fit the peak of the XRB The XRB synthesis models of Gilli et al. (2007) and Treisteret al. (2009) are defined using a slightly different cosmology ( H =70 km s − ) to that considered here ( H = 71 km s − ). However,c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . at E ∼
30 keV. The differences in the required space densi-ties of Compton-thick AGNs are derived from the normal-isations to the XRB, as well as the inclusion of the localCompton-thick AGNs identified by
INTEGRAL and
Swift (green triangle in Fig. 6) in the model of Treister et al.(2009). We find that at the median redshift of our parentsample, our estimated space density is systematically lowerthan those predicted by the Gilli et al. (2007) and Treis-ter et al. (2009) models by a factor of ≈ L X > ≈ erg s − . It is important to note thatour estimated space density is a lower limit, and is there-fore consistent with both model predictions. However, if wewere to assume the highly unlikely situation that all of ourparent sample of 334 AGNs are Compton thick (i.e., a pre-dicted space density of log(Φ) ≈ − . . < z < .
2. In the SDSS-DR7, only the most lu-minous galaxies are targeted for spectroscopic follow-up,resulting in very few ( < ≈
10 percent) galaxies being in-cluded in the SDSS with M ∗ < ≈ M ⊙ (i.e., those likelyto host relatively low mass SMBHs with M BH < ≈ M ⊙ )to z ∼ . L X / M BH distribution of our sample (SMBH masses for our sampleare estimated from stellar velocity dispersions; see Section4.3 for further details; see Column 7 of Table 1), ≈ (0 . × erg s − M ⊙− , at the luminosity completenesslimit ( L X > ≈ erg s − ) and assuming a similar distribu-tion in Eddington ratios, the smallest SMBH which couldconceivably be in our sample is M BH ≈ × M ⊙ ; this isa factor ≈
10 below the lowest mass Compton-thick AGNidentified here (M BH ≈ × M ⊙ ). We may now estimatehow many Compton-thick AGNs may contain SMBHs in themass region M BH ≈ (0 . × M ⊙ , and hence constrainthe number of Compton-thick AGNs not included in ourspace-density estimate due to the lower mass limit of theSDSS.If the five most conservatively identified Compton-thickAGNs with M BH estimates contained SMBHs which were afactor ≈
10 smaller in mass but had the same L X / M BH ratio,four ( ≈
80 percent) would still have L X > ≈ erg s − andwould therefore be included in our estimate of the spacedensity of Compton-thick AGNs as shown in Fig. 6. Basedon a simple extrapolation of the SMBH mass function ofMarconi et al. (2004), AGNs hosting SMBHs with M BH ≈ (0 . × M ⊙ are a factor ≈ . BH ≈ (0 . × M ⊙ . Hence, based on thissimplistic formalism, we estimate that approximately halfof all Compton-thick AGNs with L X > ≈ erg s − maycontain SMBHs with M BH ≈ (0 . × M ⊙ which arenot included in our parent sample. this has little or no effect on the conclusions drawn from thecomparisons. Obscuration and host-galaxy contamination may fur-ther prevent us from identifying all Compton-thick AGNsin our considered volume. Obscured AGNs can be mis-classified when the host-galaxy over-shines the nuclear emis-sion. In the absence of extinction, a NL-AGN with L X ≈ × erg s − can be almost totally diluted by a star-formation rate of 10 M ⊙ yr − (i.e., > ≈
95 percent of theobserved Hβ emission is produced in H ii regions; e.g.,Yan et al. 2010). The contribution of these sources to ourobserved space density is difficult to quantify. However, itis predicted that as many ≈
50 percent of AGNs may showno evidence for AGN activity in their optical spectroscopy(e.g., Maiolino et al. 2003; Goulding & Alexander 2009), andwould therefore not be included in our optically selectedAGN sample. Allowing for the incompleteness within ouroptical parent sample and the possibility that many more ofthe AGNs studied here may be Compton thick, we suggestthat our derived space density can be broadly consistentwith the XRB models. z ∼ . η ∼ L AGN /L Edd ; where L Edd ≈ . × (M BH / M ⊙ ) erg s − ) for the Compton-thick AGNsidentified in our sample with publicly available black-holemass (M BH ) estimates. Stellar velocity dispersion measure-ments have been computed for 13 of the 14 AGNs in oursample, at least five of which we conservatively identifyas Compton-thick AGNs. These measurements are publiclyavailable in the MPA-JHU release of SDSS-DR7 and arederived from the fitting of stellar population synthesis mod-els to the SDSS 1-D spectra. Using the M – σ relation ofGebhardt et al. (2000) we convert the stellar velocity dis-persions to M BH (see Column 7 of Table 1) in order to cal-culate L Edd for these sources. The median SMBH mass forour sample is M BH ≈ × M ⊙ (i.e., these AGNs hostSMBHs which are similar to those identified in the opticalstudy of Heckman et al. 2004).In order to estimate η for our Compton-thick AGNs, weuse L µm as a proxy for L AGN and we assume the bolometriccorrections of Marconi et al. (2004). The use of the 6 µ mcontinuum emission to infer L AGN has the advantage thatit is an independent measure of the intrinsic luminosity ofthe AGN, whilst the NL [O iv ] emission arises from a similarregion to that of [O iii ] which was used for the selection ofthe sources considered here.Twelve of our 14 sources have both M BH estimates and6 µ m measurements. We find that our sample of AGNs are (cid:13) , 1–16 A.D. Goulding et al. spread over a wide-range of Eddington ratio, η ≈ . ≈ . η ≈ . . ≈ . Similarly, we findthat the well-studied local Compton-thick AGNs (Circinus,Mrk 3, NGC 1068 and NGC 6240) have a similar range inEddington ratio, η ≈ . ≈ . µ m relation combinedwith a possible Eddington ratio dependent bolometric cor-rection (e.g., Vasudevan & Fabian 2007) could yield an un-certainty factor of the order > ≈
10 for the highest Eddingtonratio sources. None of the AGNs in our sample appear to beEddington-limited on the basis of the 6 µ m luminosity, andany uncertainties would apply equally to all of the AGNsconsidered here, hence our finding of systematically higherEddington ratios for Compton-thick AGNs, to first-order,appears to be relatively robust. However, we suggest thatthis result may be driven by our selection of [O iii ]-brightAGNs as well as our sensitivity towards the identification ofCompton-thick AGNs. For example, with deeper X-ray datawe may identify further Compton-thick AGNs in our samplewhich have lower values of η , and hence reducing the medianEddington ratio for the Compton-thick AGN subsample.By comparison, for the total population of Type-2AGNs identified from optical emission-line diagnostics inthe SDSS, Heckman et al. (2004) find that based on the useof [O iii ] emission to infer L AGN , < . BH ≈ × M ⊙ are accret-ing above η ≈ .
1; by contrast, we find that > ≈
30 percentof our sample have η > ≈ .
1. A particular advantage to a di-rect comparison with the study of Heckman et al. (2004) isthat selection processes and biases are likely to be identicalbetween both studies. Hence, these results suggest that, onaverage, the Compton-thick AGNs identified here may har-bour some of the most rapidly growing black holes in thenearby Universe. This would further suggest that not tak-ing account of Compton-thick AGNs in deep-field X-ray sur-veys may exclude the most rapid growth phases of SMBHs,as predicted by many theoretical models (e.g., Fabian 1999;Granato et al. 2006; Hopkins et al. 2008).
We have presented a sample of 14 local ( z ∼ . iii ] emission line ratios (i.e., f X /f [OIII] < .
0; e.g., Bassani et al. 1999; see Section 2). Wehave employed a suite of optical (e.g., [O iii ] emission-line)and mid-IR (e.g., [O iv ] emission-line; 6 µ m AGN contin-uum) diagnostics to infer the intrinsic AGN luminosity inthese sources. Assuming any deficit in X-ray flux comparedto these estimates is due to Compton-thick absorption, we We note that when using [O iv ] emission to infer L AGN , tofirst-order, we find a similar result (i.e., systematically higher Ed-dington ratios for the Compton-thick AGNs). assess the ubiquity of Compton-thick AGN activity in thenearby Universe. Our main findings are the following:(1) Using
Spitzer -IRS low resolution spectroscopy, we findthat six of our 14 candidate Compton-thick AGNs have > ≈ σ detections of [Ne v ] and all 14 have > ≈ σ detections of [O iv ].We performed mid-IR spectral decompositions of our sam-ple to establish 6 µ m AGN continuum luminosities. Usingestablished X-ray to mid-IR continuum and emission-line re-lationships, we infer the intrinsic X-ray luminosity of theseAGNs and conservatively find that 6/14 ( ≈ ±
21 percent)of the sources in our sample appear to be heavily obscuredwith N H > ≈ . × cm − (i.e., are Compton-thick AGNs).See sections 3.3, 3.4 and 4.1.(2) We used our results to infer the ubiquity of Compton-thick AGNs in our SDSS–2XMMi parent sample. We predictthat on the basis of the analyses presented here that at least > ≈
20 percent of the 334 optical Type-2 AGNs in the SDSS-DR7 at z ∼ . > ≈ − . − for Compton-thick AGNs with L X > ≈ erg s − at z ∼ . µ mcontinuum luminosity to infer the L AGN and the stellar ve-locity dispersion to estimate M BH , we find systematicallyhigher Eddington ratios for the most conservatively iden-tified Compton-thick AGNs ( η ≈ . BH ≈ × M ⊙ e.g., Heckman et al. 2004), we findthat Compton-thick AGNs selected in the SDSS may har-bour some of the most rapidly growing black holes in thenearby Universe ( z ∼ . L X – z plane than can be achieved using X-ray spectroscopy alone. Using the next generation of X-raysatellites (e.g., NuStar ; IXO ; WFXT ), high-quality X-rayspectroscopy and
E >
10 keV detections will allow us todirectly and unambiguously identify which of these sourcesare Compton-thick AGNs.
ACKNOWLEDGMENTS
We thank the anonymous referee for their considered andthorough report which has improved the quality of this pa-per. We would like to acknowledge useful conversations withJ Geach and T Roberts. We thank R Gilli and E Treisterfor kindly providing the XRB space density tracks. We alsothank the Science & Technologies Facilities Council (ADG;RCH), the Royal Society (DMA) and the Leverhulme Trust c (cid:13) , 1–16 earching for Compton-thick AGNs at z ∼ . (DMA; JRM) for funding. This research has made use ofthe Sloan Digital Sky Survey data archive and the NASA Spitzer
Space Telescope which is operated by the Jet Propul-sion Laboratory, California Institute of Technology undercontract with the National Aeronautics and Space Admin-istration. This research has also made use of data obtainedfrom the Leicester Database and Archive Service at the De-partment of Physics and Astronomy, Leicester University,UK.
REFERENCES
Abazajian K. N., et al. 2009, ApJS, 182, 543Aird J., Nandra K., Laird E. S., Georgakakis A., AshbyM. L. N., Barmby P., Coil A. L., Huang J., KoekemoerA. M., Steidel C. C., Willmer C. N. A., 2010, MNRAS,401, 2531Akylas A., Georgantopoulos I., 2009, A&A, 500, 999Alexander D. M., Bauer F. E., Brandt W. N., SchneiderD. P., Hornschemeier A. E., Vignali C., Barger A. J.,Broos P. S., Cowie L. L., Garmire G. P., Townsley L. K.,Bautz M. W., Chartas G., Sargent W. L. W., 2003, AJ,126, 539Alexander D. M., Brandt W. N., Hornschemeier A. E.,Garmire G. P., Schneider D. P., Bauer F. E., GriffithsR. E., 2001, AJ, 122, 2156Alexander D. M., Chary R., Pope A., Bauer F. E., BrandtW. N., Daddi E., Dickinson M., Elbaz D., Reddy N. A.,2008, ApJ, 687, 835Alonso-Herrero A., Ward M. J., Kotilainen J. K., 1997,MNRAS, 288, 977Awaki H., Koyama K., Inoue H., Halpern J. P., 1991, PASJ,43, 195Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP,93, 5Barger A. J., Cowie L. L., Capak P., Alexander D. M.,Bauer F. E., Fernandez E., Brandt W. N., Garmire G. P.,Hornschemeier A. E., 2003, AJ, 126, 632Bassani L., Dadina M., Maiolino R., Salvati M., Risaliti G.,della Ceca R., Matt G., Zamorani G., 1999, ApJS, 121,473Bauer F. E., Yan L., Sajina A., Alexander D. M., 2010,ApJ, 710, 212Brandl B. R., Bernard-Salas J., Spoon H. W. W., DevostD., Sloan G. C., et al. 2006, ApJ, 653, 1129Brusa M., Civano F., Comastri A., Miyaji T., Salvato M.,et al. 2010, ApJ, 716, 348Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345,245Carrera F. J., Ebrero J., Mateos S., Ceballos M. T., CorralA., Barcons X., Page M. J., Rosen S. R., Watson M. G.,Tedds J. A., Della Ceca R., Maccacaro T., Brunner H.,Freyberg M., Lamer G., Bauer F. E., Ueda Y., 2007, A&A,469, 27Comastri A., 2004, in A. J. Barger ed., Supermassive BlackHoles in the Distant Universe Vol. 308 of Astrophysics andSpace Science Library, Compton-Thick AGN: The DarkSide of the X-Ray Background. pp 245–+Della Ceca R., Severgnini P., Caccianiga A., Comastri A.,Gilli R., Fiore F., Piconcelli E., Malaguti P., Vignali C.,2008, Memorie della Societa Astronomica Italiana, 79, 65 Diamond-Stanic A. M., Rieke G. H., Rigby J. R., 2009,ApJ, 698, 623Donley J. L., Rieke G. H., Alexander D. M., Egami E.,P´erez-Gonz´alez P. G., 2010, ApJ, 719, 1393Draine B. T., Li A., 2007, ApJ, 657, 810Ebrero J., Carrera F. J., Page M. J., Silverman J. D., Bar-cons X., Ceballos M. T., Corral A., Della Ceca R., WatsonM. G., 2009, A&A, 493, 55Fabian A. C., 1999, MNRAS, 308, L39Ferrarese L., Merritt D., 2000, ApJ, 539, L9Fiore F., Brusa M., Cocchia F., Baldi A., Carangelo N.,Ciliegi P., Comastri A., La Franca F., Maiolino R., MattG., Molendi S., Mignoli M., Perola G. C., Severgnini P.,Vignali C., 2003, A&A, 409, 79Fiore F., Puccetti S., Brusa M., Salvato M., Zamorani G.,et al. 2009, ApJ, 693, 447Gallimore J. F., Yzaguirre A., Jakoboski J., StevenoskyM. J., Axon D. J., Baum S. A., Buchanan C. L., ElitzurM., Elvis M., O’Dea C. P., Robinson A., 2010, ApJS, 187,172Gebhardt K., Bender R., Bower G., Dressler A., FaberS. M., Filippenko A. V., Green R., Grillmair C., Ho L. C.,Kormendy J., Lauer T. R., Magorrian J., Pinkney J.,Richstone D., Tremaine S., 2000, ApJ, 539, L13Georgantopoulos I., Akylas A., Georgakakis A., Rowan-Robinson M., 2009, A&A, 507, 747Ghisellini G., Haardt F., Matt G., 1994, MNRAS, 267, 743Giacconi R., Zirm A., Wang J., Rosati P., Nonino M., TozziP., Gilli R., Mainieri V., Hasinger G., Kewley L., BergeronJ., Borgani S., Gilmozzi R., Grogin N., Koekemoer A.,Schreier E., Zheng W., Norman C., 2002, ApJS, 139, 369Gilli R., Comastri A., Hasinger G., 2007, A&A, 463, 79Goulding A. D., 2010, Ph D. thesis, University of Durham,U.K.Goulding A. D., Alexander D. M., 2009, MNRAS, 398, 1165Goulding A. D., Alexander D. M., Lehmer B. D., MullaneyJ. R., 2010, MNRAS, 406, 597Granato G. L., Silva L., Lapi A., Shankar F., De Zotti G.,Danese L., 2006, MNRAS, 368, L72Greene J. E., Ho L. C., 2007, ApJ, 667, 131Heckman T. M., Kauffmann G., Brinchmann J., CharlotS., Tremonti C., White S. D. M., 2004, ApJ, 613, 109Heckman T. M., Ptak A., Hornschemeier A., KauffmannG., 2005, ApJ, 634, 161Hickox R. C., Markevitch M., 2006, ApJ, 645, 95Hickox R. C., Markevitch M., 2007, ApJ, 661, L117Higdon S. J. U., Devost D., Higdon J. L., Brandl B. R.,Houck J. R., Hall P., Barry D., Charmandaris V., SmithJ. D. T., Sloan G. C., Green J., 2004, PASP, 116, 975Hopkins P. F., Hernquist L., Cox T. J., Kereˇs D., 2008,ApJS, 175, 356Houck J. R., Roellig T. L., et al. 2004, ApJS, 154, 18Kauffmann G., Heckman T. M., Tremonti C., BrinchmannJ., Charlot S., White S. D. M., Ridgway S. E., BrinkmannJ., Fukugita M., Hall P. B., Ivezi´c ˇZ., Richards G. T.,Schneider D. P., 2003, MNRAS, 346, 1055Kauffmann G., Heckman T. M., White S. D. M., CharlotS., Tremonti C., Brinchmann J., Bruzual G., Peng E. W.,Seibert M., Bernardi M., et al. 2003, MNRAS, 341, 33Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A.,Trevena J., 2001, ApJ, 556, 121Kormendy J., Richstone D., 1995, ARA&A, 33, 581 c (cid:13) , 1–16 A.D. Goulding et al.
LaMassa S. M., Heckman T. M., Ptak A., HornschemeierA., Martins L., Sonnentrucker P., Tremonti C., 2009, ApJ,705, 568Levenson N. A., Weaver K. A., Heckman T. M., Awaki H.,Terashima Y., 2004, ApJ, 602, 135Luo B., Bauer F. E., Brandt W. N., Alexander D. M.,Lehmer B. D., Schneider D. P., Brusa M., Comastri A.,Fabian A. C., et al. 2008, ApJS, 179, 19Lutz D., Maiolino R., Spoon H. W. W., MoorwoodA. F. M., 2004, A&A, 418, 465Magorrian J., Tremaine S., Richstone D., Bender R., BowerG., Dressler A., Faber S. M., Gebhardt K., Green R., Grill-mair C., Kormendy J., Lauer T., 1998, AJ, 115, 2285Maiolino R., Comastri A., Gilli R., Nagar N. M., BianchiS., B¨oker T., Colbert E., Krabbe A., Marconi A., MattG., Salvati M., 2003, MNRAS, 344, L59Maiolino R., Salvati M., Bassani L., Dadina M., della CecaR., Matt G., Risaliti G., Zamorani G., 1998, A&A, 338,781Maiolino R., Shemmer O., Imanishi M., Netzer H., OlivaE., Lutz D., Sturm E., 2007, A&A, 468, 979Marconi A., Risaliti G., Gilli R., Hunt L. K., Maiolino R.,Salvati M., 2004, MNRAS, 351, 169Matt G., Brandt W. N., Fabian A. C., 1996, MNRAS, 280,823Matt G., Fabian A. C., Guainazzi M., Iwasawa K., BassaniL., Malaguti G., 2000, MNRAS, 318, 173Mel´endez M., Kraemer S. B., Armentrout B. K., Deo R. P.,Crenshaw D. M., Schmitt H. R., Mushotzky R. F., TuellerJ., Markwardt C. B., Winter L., 2008, ApJ, 682, 94Mel´endez M., Kraemer S. B., Schmitt H. R., CrenshawD. M., Deo R. P., Mushotzky R. F., Bruhweiler F. C.,2008, ApJ, 689, 95Mulchaey J. S., Koratkar A., Ward M. J., Wilson A. S.,Whittle M., Antonucci R. R. J., Kinney A. L., Hurt T.,1994, ApJ, 436, 586Mullaney J. R., Alexander D. M., Huynh M., GouldingA. D., Frayer D., 2010, MNRAS, 401, 995Netzer H., Mainieri V., Rosati P., Trakhtenbrot B., 2006,A&A, 453, 525Norman C., Hasinger G., Giacconi R., Gilli R., KewleyL., Nonino M., Rosati P., Szokoly G., Tozzi P., Wang J.,Zheng W., Zirm A., Bergeron J., Gilmozzi R., Grogin N.,Koekemoer A., Schreier E., 2002, ApJ, 571, 218Osterbrock D. E., Ferland G. J., 2006, Astrophysics ofgaseous nebulae and active galactic nuclei. Astrophysicsof gaseous nebulae and active galactic nuclei, 2nd. ed. byD.E. Osterbrock and G.J. Ferland. Sausalito, CA: Univer-sity Science Books, 2006Panessa F., Bassani L., 2002, A&A, 394, 435Panessa F., Bassani L., Cappi M., Dadina M., Barcons X.,Carrera F. J., Ho L. C., Iwasawa K., 2006, A&A, 455, 173Rees M. J., 1984, ARA&A, 22, 471Risaliti G., Maiolino R., Salvati M., 1999, ApJ, 522, 157Satyapal S., Vega D., Dudik R. P., Abel N. P., HeckmanT., 2008, ApJ, 677, 926Siebenmorgen R., Kr¨ugel E., 2007, A&A, 461, 445Soltan A., 1982, MNRAS, 200, 115Tommasin S., Spinoglio L., Malkan M. A., Fazio G., 2010,ApJ, 709, 1257Tozzi P., Gilli R., Mainieri V., Norman C., Risaliti G.,Rosati P., Bergeron J., Borgani S., Giacconi R., Hasinger G., Nonino M., Streblyanska A., Szokoly G., Wang J. X.,Zheng W., 2006, A&A, 451, 457Treister E., Krolik J. H., Dullemond C., 2008, ApJ, 679,140Treister E., Urry C. M., Virani S., 2009, ApJ, 696, 110Tremaine S., Gebhardt K., Bender R., Bower G., DresslerA., Faber S. M., Filippenko A. V., Green R., GrillmairC., Ho L. C., Kormendy J., Lauer T. R., Magorrian J.,Pinkney J., Richstone D., 2002, ApJ, 574, 740Ueda Y., Akiyama M., Ohta K., Miyaji T., 2003, ApJ, 598,886Vasudevan R. V., Fabian A. C., 2007, MNRAS, 381, 1235Vignali C., Alexander D. M., Gilli R., Pozzi F., 2010, MN-RAS, 404, 48Ward M. J., Geballe T., Smith M., Wade R., Williams P.,1987, ApJ, 316, 138Watson M. G., et al. 2009, A&A, 493, 339Weaver K. A., Mel´endez M., Mushotzky R. F., KraemerS., Engle K., Malumuth E., Tueller J., Markwardt C.,Berghea C. T., Dudik R. P., Winter L. M., Armus L.,2010, ArXiv e-printsWild V., Heckman T., Sonnentrucker P., Groves B., ArmusL., Schiminovich D., Johnson B., Martins L., LaMassa S.,2010, ArXiv e-printsWorsley M. A., Fabian A. C., Bauer F. E., AlexanderD. M., Hasinger G., Mateos S., Brunner H., Brandt W. N.,Schneider D. P., 2005, MNRAS, 357, 1281Yan R., Ho L. C., Newman J. A., Coil A. L., WillmerC. N. A., Laird E. S., Georgakakis A., Aird J., BarmbyP., et al. 2010, ArXiv e-printsThis paper has been typeset from a TEX/ LaTEX file preparedby the author. c (cid:13)000