The Largest X-ray Selected Sample of z>3 AGNs: C-COSMOS & ChaMP
aa r X i v : . [ a s t r o - ph . GA ] A ug Mon. Not. R. Astron. Soc. , 1–16 (2014) Printed 16 July 2018 (MN L A TEX style file v2.2)
The Largest X-ray Selected Sample of z > AGNs: C-COSMOS &ChaMP
E. Kalfountzou , ⋆ , F. Civano , , M. Elvis , M. Trichas , P. Green Harvard Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138, USA Centre for Astrophysics, Science & Technology Research Institute, University of Hertfordshire, Hatfield, Herts, AL10 9AB, UK Yale Center for Astronomy and Astrophysics, 260 Whitney ave, New Haven, CT Airbus Defence & Space, Gunnels Wood Road, Stevenage, SG1 2AS, UK
Received May 29, 2014; accepted August 25, 2014
ABSTRACT
We present results from an analysis of the largest high-redshift ( z > ) X-ray-selected activegalactic nucleus (AGN) sample to date, combining the Chandra
C-COSMOS and ChaMPsurveys and doubling the previous samples. The sample comprises 209 X-ray-detected AGN,over a wide range of rest frame 2-10 keV luminosities log L X = 43 . − . − .X-ray hardness rates show that ∼
39% of the sources are highly obscured, N H > cm − ,in agreement with the ∼
37% of type-2 AGN found in our sample based on their opticalclassification. For ∼
26% of objects have mismatched optical and X-ray classifications. Usingthe /V max method, we confirm that the comoving space density of all luminosity ranges ofAGNs decreases with redshift above z > and up to z ∼ . With a significant sample ofAGN ( N = 27 ) at z > , it is found that both source number counts in the 0.5 -2 keV bandand comoving space density are consistent with the expectation of a luminosity dependentdensity evolution (LDDE) model at all redshifts, while they exclude the luminosity anddensity evolution (LADE) model. The measured comoving space density of type-1 andtype-2 AGN shows a constant ratio between the two types at z > . Our results for both AGNtypes at these redshifts are consistent with the expectations of LDDE model. Key words:
Galaxies: active - evolution - surveys - X-rays: galaxies
Active galactic nuclei (AGN) evolution at high redshifts, beforetheir density peak, illuminates the role of AGN in the formationand co-evolution of galaxies and their central supermassive blackholes (SMBHs) during the time of rapid SMBH growth. The so-called ‘downsizing’ evolution has been revealed for both AGN(e.g. Ueda et al. 2003; Hasinger et al. 2005; Aird et al. 2010) andgalaxies (e.g. Cowie et al. 1996; Kodama et al. 2004; Damen et al.2009). Supporting this idea, X-ray surveys have shown that thenumber density of luminous AGN peaks at higher redshifts thanless luminous ones (e.g. Ueda et al. 2003; Aird et al. 2010). Thissort of cosmological ‘co-evolution’ scenario is inferred from thetight correlation exists locally between SMBH mass and galacticbulge properties (e.g. Magorrian et al. 1998; Ferrarese & Merritt2000; Gebhardt et al. 2000; McConnell & Ma 2013).To elucidate the co-evolution of SMBH and galaxies (e.g.Granato et al. 2001, 2004; Croton et al. 2006; Hopkins et al. 2006;Menci et al. 2008; Trichas et al. 2009, 2010; Kalfountzou et al.2011, 2012, 2014), the accretion activity in the Universe has to ⋆ Email: [email protected] be studied both at high redshifts and for low luminosities. Thisrequires large samples of AGN spanning wide ranges of proper-ties. While many optical surveys have investigated the space den-sity of high-redshift AGN (e.g. Richards et al. 2006; Jiang et al.2009; Willott et al. 2010; Glikman et al. 2011; Ikeda et al. 2011;Ross et al. 2013), the results are still controversial due to their in-evitable incompleteness, especially at the faint luminosity end dueto the host contamination, and the bias against obscured sources.Contrary to the optical surveys, X-ray observations are less con-taminated by the host galaxy emission and include AGN popula-tions with a wide range of neutral hydrogen column density.For the investigation of the absorption evolution (e.g.Ueda et al. 2003; Hasinger 2008; Draper & Ballantyne 2010),X-ray selected samples include all types of AGN (e.g. type-1/unobscured and type-2/obscured) and provide reduced obscura-tion bias in comparison with optically selected AGN. Although X-ray surveys have inferred the existence of an anti-correlation be-tween the obscured AGN fraction and the luminosity, several ofthese studies have suggested an increase of the fraction towardhigher redshift from z = 0 to z ∼ with limited samples at z > (e.g. La Franca et al. 2005; Ballantyne et al. 2006; Treister & Urry2006; Ballantyne 2008; Hiroi et al. 2012). c (cid:13) E. Kalfountzou et al.
However, the evolution of AGN is still rife with uncertainty.On the basis of hard X-ray surveys many studies agreed that theXLF of AGN is best described by a luminosity dependent densityevolution (LDDE) model (e.g. Ueda et al. 2003; Gilli et al. 2007;Silverman et al. 2008; Ueda et al. 2014). Aird et al. (2010) pre-ferred instead a luminosity and density evolution model (LADE).In LADE, the shift in the redshift peak of the AGN space densityversus X-ray luminosity is much weaker than in LDDE models, yetgives a similarly good fit to their data. While the z < downsiz-ing behavior is common to both models, quite different numbers ofAGN are predicted at higher redshifts ( z > ).X-ray surveys (2-10 keV) are now sensitive enough to sam-ple the bulk of the z > AGN population. Two studies have beenperformed on high-redshift AGN exploiting the deep X-ray surveysin the Cosmological Evolution Survey (COSMOS) field carried outwith XMM-Newton ( N AGN = 40 ; Brusa et al. 2009) and
Chandra ( N AGN = 81 ; Civano et al. 2011), limited to − keV luminosi-ties L − > . erg s − and . erg s − , respectively.A more recent study based on the 4 Ms Chandra
Deep Field South(CDF-S, Xue et al. 2011) was able to investigate the evolution of z > AGN down to L X ∼ erg s − ( N AGN = 34 ; Vito et al.2013). These results are consistent with a decline of the AGN spacedensity at z > , but the shape of this decline remains highly uncer-tain at z > . To overcome these limitations, in this work we com-bined the two largest samples of z > X-ray selected AGN withspectroscopic redshifts, both derived from Chandra X-ray Observa-tory (Weisskopf et al. 2002) surveys: the wide but shallow ChaMPsurvey (Kim et al. 2007; Green et al. 2009), and the deeper but nar-rower C-COSMOS survey (Elvis et al. 2009). This combination re-sults in the largest X-ray AGN sample with N AGN = 211 at z > and N AGN = 27 at z > . At the same time, by combining twosurveys with different flux limits, we are able to determine the den-sity evolution of both low luminosity ( L X < erg s − ) andhigh luminosity AGN. Our sample includes both obscured and un-obscured AGN, and their separate evolution has been determined.The paper is structured as follows. In Section 2, we discussthe the data sets used in this work and the selection of the high- z sample. In Section 3, we present the optical and X-ray propertiesof the selected high- z AGN sample and we explain the AGN typeclassification using X-ray or optical data. In Sections 4 and 5, thenumber counts and space density of the sample are compared withmodel predictions. Section 6 summarizes the conclusions. A cos-mological model with Ω o = 0 . , λ o = 0 . , and a Hubble constantof
70 km s − Mpc − is used throughout (Spergel et al. 2003). Er-rors are quoted at the σ level. The high redshift AGN sample used in this work has been selectedfrom the C-COSMOS X-ray catalog, combining the spectroscopicand photometric information available from the identification cat-alog of X-ray C-COSMOS sources (Civano et al. 2011, 2012) andthe ChaMP X-ray catalog using only the 323 ChaMP obsids over-lapping with SDSS DR5 imaging. In Figure 1, we show the skycoverage (the area of a survey that is sensitive to sources above agiven X-ray flux) using the observed soft band (0.5-2 keV) sourcedetections for the two surveys, and their sum. This corresponds to2-8 keV rest frame for z > .A schematic diagram of the sample selection with the detailednumber of sources for each step is presented in Figure 2. -16 -15 -14 S(0.5-2keV) [erg cm -2 s -1 ]0.11.010.0 A r ea [ d e g ] a pp li e d f l ux c u tl og N - l og S c u t C-COSMOSChaMP s u m Figure 1.
Sky area vs. X-ray flux sensitivity curves for the C-COSMOS(blue solid line) and ChaMP/SDSS (red solid line) samples and the totalarea (black dashed line). The vertical blue dashed line indicates the fluxcorresponding to 10% of the total C-COSMOS area (see section 4). Thevertical red dashed line indicates the ChaMP X-ray flux limit with > %completeness from SDSS/UKIDSS/WISE (see section 2.2). The total area,after the applied cuts, used for this work is represented by the shadowedgrey area. The
Chandra -COSMOS survey (C-COSMOS; Elvis et al. 2009;Civano et al. 2012) covers the central 0.9 deg of the COSMOSfield up to a depth of 200 ksec in the inner 0.5 deg , with theACIS-I CCD imager (Garmire et al. 2003) on board Chandra . TheC-COSMOS X-ray source catalog comprises 1,761 point-like X-ray sources detected down to a maximum likelihood thresholddetml=10.8 in at least one band. This likelihood threshold cor-responds to a probability of ∼ × − that a catalog sourceis instead a background fluctuation (Puccetti et al. 2009). Giventhis likelihood threshold, the flux limit reached in the surveyis . × − erg cm − s − in the Full band (0.5-10 keV), . × − erg cm − s − in the Soft band (0.5-2 keV) and . × − erg cm − s − in the Hard band (2-10 keV).The z > C-COSMOS sample, as presented by (Civano et al.2011), comprises 107 X-ray detected sources with available spec-troscopic (32) and photometric (45) redshifts plus 30 sources witha formal z phot < but with a broad photometric redshift probabil-ity distribution, such that z phot + 1 σ phot > . All of the spec-troscopic C-COSMOS sources have a quality flag 3 (2 sources)or 4 corresponding respectively to a secure redshift with two ormore emission or absorption lines and a secure redshift with twoor more emission or absorption lines with a good-quality, highS/N spectrum (see Lilly et al. 2007, 2009 for thorough explanationof quality flags). Tuned photometric redshifts for the C-COSMOSsources have been computed and presented in Salvato et al. (2011).Due to the large number of photometric bands and the sizablespectroscopic training sample spanning a large range in redshiftand luminosity the estimated photometric redshifts are expectedto be quite robust at z > . even at the fainter magnitudes( i AB > . ). The COSMOS photometric redshifts for X-ray se-lected sources have an accuracy of σ ∆ z/ (1+ z spec ) = 0 . witha small fraction of outliers ( < %), considering the sample as awhole at i < . . At fainter magnitudes, the dispersion increases c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs C-COSMOS1,761
X-ray detections
ChaMP15,202
X-ray detections(exclude target sources) z spec z phot z spec z phot SDSS (S); UKIDSS (U);WISE (W) counterparts without (S,U,W)counterparts1,164 > flux cut122 high- z AGN 32 z spec > z phot + σ > z phot > > fluxcut44 z spec > good > . > flux cut28high- z z phot > z phot + σ > > flux cut1,508point-likegood photometry177S+U+W526S+W 44S+U16U+W 367S359W 19U15high- z z phot > z phot + σ > high- z X-ray selectedAGN z spec > z phot > z phot + 1 σ > opticaldropouts Figure 2.
Schematic flow diagram of the high- z sample selection. to σ ∆ z/ (1+ z spec ) = 0 . with ∼ % outliers, still remarkablygood for an AGN sample. For the z > C-COSMOS sample, anaccuracy of σ ∆ z/ (1+ z spec ) = 0 . is achieved with only 3 catas-trophic outliers ( < %). The spectral energy distributions (SEDs)of the sources with photometric redshift larger than 3 have been vi- sually inspected together with the photometric fitting and the prob-ability distribution of all the possible solutions.There are 91 sources selected in the 0.5-2 keV band, 14 in the2-10 keV, and 4 in the 0.5-10 keV bands. There are 15 C-COSMOSsources without a counterpart in the optical bands, but with a K-band and IRAC (7), only IRAC (6) or no infrared detection (2). c (cid:13) , 1–16 E. Kalfountzou et al.
Given the small number of bands in which these objects are de-tected, no photometric redshift is available for them. In X-ray se-lected samples, non-detection in the optical band has been often as-sumed to be a proxy for high redshift (e.g. Koekemoer et al. 2004),or for high obscuration, or a combination of both. Four of the 15sources have no detection in the soft band suggesting high obscura-tion, possibly combined with high redshift. More details about thesample selection can be found in Civano et al. (2011) and are alsopresented in Figure 2.
The
Chandra
Multi-wavelength Project (ChaMP) is a wide-areanon-continuous X-ray survey based on archival X-ray images ofthe high Galactic latitude ( | b | > deg) sky observed with ACISon Chandra . The flux levels (in erg cm − s − ) reached in thesurvey are . × − − . × − in the Full ( . − keV), . × − − . × − ( . − keV) in the Soft and . × − − . × − ( − keV) in the Hard band, respec-tively. The ChaMP survey includes a total of 392 fields, omittingpointings from dedicated serendipitous surveys like C-COSMOS,the Chandra
Deep Fields, as well as fields with extended ( > ′ )bright optical or X-ray sources. The list of Chandra pointingsavoids any overlapping observations by eliminating the observa-tion with the shorter exposure time. The survey has detected a to-tal of >
33 deg with ∼ Chandra sources are presented. Since theChaMP is a
Chandra archival survey, most ChaMP fields containtargeted sources selected by the target’s PI, and those targets arelikely to be biased toward special X-ray populations such as brightAGN. Of the targeted sources ∼ % have a secure spectroscopicredshift with 33 of them having at z > and 29 at z > (seeTrichas et al. 2012). The high rate of high-redshift detected sourcesclearly shows the strong selection biases that could affect our anal-ysis if we included the targeted sources. Therefore, we excludeall targeted sources (153) to reduce bias in sample properties andsource number counts.For SDSS point sources with i < and without availablespectroscopy, efficient photometric selection of quasars is possi-ble using a nonparametric Bayesian classification based on kerneldensity estimation as described in Richards et al. (2009). To selecthigh- z candidates without available spectroscopic or photometricredshift, SDSS detection is required in at least the i − and z − bands, to detect Lyman dropouts (e.g. Steidel et al. 1996).Searching the ChaMP catalog for X-ray sources within ′′ of the optical SDSS quasar coordinate (95% of the matched sam-ple has an X-ray/optical position difference of less than ′′ ; seeGreen et al. 2009), yields , unique matches ( ∼
63% of the to-tal ChaMP X-ray selected sample). We additionally searched forcross-matches in the Wide-field Infrared Survey Explorer (WISE;Wright et al. 2010) and UKIRT (UK Infrared Telescope) Infrared Deep Sky Survey (UKIDSS; Warren et al. 2000; Hewett et al.2006; Maddox et al. 2008). For a source to be included in the
WISE
All Sky Source Cata-log (Wright et al. 2010), a SNR > W , W , W or W , with centralwavelengths of roughly 3.4, 4.6, 12, and 22 µ m, and angular res-olutions of 6.1, 6.4, 6.5 and 12.0 arcsec. Because of the differentspatial resolutions, 6.0 arcsec (WISE; W ) and 1-2 arcsec (SDSS),we use ′′ as the matching radius for WISE counterparts (Wu et al.2012).Similarly, we searched the UKIDSS Large Area Survey(LAS; Lawrence et al. 2007) Data Release 10 for NIR counter-parts to ChaMP X-ray sources. The photometric system is de-scribed in Hewett et al. (2006), and the calibration is describedin Hodgkin et al. (2009). We used the LAS Y JHK
Source table,which contains only fields with coverage in every filter and mergesthe data from multiple detections of the same object. The X-raysource catalogs were then matched within ′′ of the X-ray positionseparately to each UKIDSS band: Y (0.97-1.07 µ m), J (1.17-1.33 µ m), H (1.49-1.78 µ m) and K (2.03-2.37 µ m) recovering also theareas with coverage in a single UKIDSS band. The individual bandlists were then combined. For objects not detected in a UKIDSSband we use the σ detection limits provided in Dye et al. (2006)of Y = 20 . , J = 19 . , H = 18 . , and K = 18 . . Match-ing the ChaMP catalog to WISE and UKIDSS, we find 1,103 ad-ditional WISE and/or UKIDSS counterparts which do not have aSDSS counterpart (the detailed numbers are reported in Figure 2).In summary, ∼
70% of the total ChaMP X-ray sample haveSDSS and UKIDSS/WISE photometry ( , SDSS and/or WISEand/or UKIDSS and , WISE and/or UKIDSS). The limitedfraction of optical matches shows how optical counterparts of faintX-ray sources are fainter than the SDSS magnitude limit ( i =21 . ). SDSS quasars were identified to i < . for spectroscopyby their UV-excess colors, with an extension for z > quasars to i = 20 . using ugri color criteria (Richards et al. 2002).Based on the X-ray limits, the identification completeness ofChaMP X-ray sources falls rapidly for objects with fainter opticalcounterparts. Figure 3 shows the optical (SDSS i − band counter-parts) and infrared (WISE and UKIDSS counterparts) complete-ness of the X-ray selected sample as a function of the soft (0.5-2keV) X-ray flux. This incompleteness can severely bias determina-tion of the number counts and space density, particularly at highredshifts (e.g. Barger & Cowie 2005).To address this issue, we set a relatively high X-ray flux limitin ChaMP, where spectroscopic completeness is higher, and photo-metric coverage allows good photometric redshifts. We use a softflux limit for ChaMP at S . − > × − erg s − cm − as at these brighter fluxes the completeness is higher than > %(see Figure 3). The completeness fraction as a function of flux hasbeen taken into account for the estimation of the number countsand comoving space density (see Sections 4 and 5). For sources notdetected in the soft band, the 0.5-2 keV flux has been computed byconverting the 2-10 keV flux using Γ = 1 . (see Section 3.2). Oneof the main advantages of our compilation is that we do not missthe faint high redshift population, since this is recovered by the C-COSMOS survey. In this way, the ChaMP sample is used for the The UKIDSS project is defined in Lawrence et al. (2007). UKIDSS usesthe UKIRT Wide Field Camera (WFCAM; Casali et al. 2007) and a pho-tometric system described in Hewett et al. (2006). The pipeline processingand science archive are described in Hambly et al. (2008).c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs -16 -15 -14 -13 S(0.5-2.0 keV) [erg s -1 cm -2 ]0.40.50.60.70.80.91.0 C o m p l e t e n e ss Figure 3.
Optical and/or infrared ChaMP survey completeness as a functionof 0.5-2 keV X-ray flux. The red dotted line indicates the ChaMP X-ray fluxcut with > % completeness from SDSS/UKIDSS/WISE. determination of the bright end of the luminosity function at highredshifts. We compiled secure spectroscopic redshifts for a total of 1,547sources. We have used 1,056 sources (excluding target sources)from existing ChaMP spectroscopy (Trichas et al. 2012) for the se-lected ChaMP fields. Additional spectroscopic redshifts are givenin the SDSS-III ( N = 91 ; Noterdaeme et al. 2012) and SDSS-DR10 quasar catalogs ( N = 145 ; Pˆaris et al. 2014). We alsosearched the literature by cross-correlating optical positions withthe NASA Extragalactic Database (NED), using a ′′ match radiuswhere we found 255 more sources with spectroscopic redshift.The high redshift spectroscopic sample consists of 44 sourceswith z > . All of these sources have a soft band X-ray detection,and only 3 sources lack a hard band detection. Among them, thereare 7 sources with z > and 1 source with z = 6 . ± . (Jiang et al. 2007). All but 6 of them have SDSS optical spec-tra with mean S/N > . (none of them has S/N < . ) withat least 2 broad emission lines (Ly α and CIV) significantly de-tected. For 5 of the remaining sources, redshifts have been ob-tained by Trichas et al. (2012) while for the source with the high-est spectroscopic redshift ( z = 6 . ) we have used the esti-mate from Jiang et al. (2007). For 30 sources of the ChaMP spec-troscopic sample there are available photometric redshifts derivedby Richards et al. (2009) (see Section 2.2.2) with an accuracy of σ ∆ z/ (1+ z spec ) = 0 . and only one catastrophic outlier. For the sources without spectroscopic redshifts, we derived pho-tometric redshifts. The criteria used in SDSS DR6 have now beenrefined to include objects redder than ( u − g ) = 1 . which maywell be high- z quasars. The resulting catalog of ∼ million pho-tometrically identified quasars and their photometric redshifts fromSDSS Data Release 6 (DR6) is described in Richards et al. (2009).Only point sources (type=6) with i − band magnitudes between14.5 and (de-reddened) 21.3 ( psfmag i > . & psfmag i − extinction i < . ; where psfmag are the point-spread-function magnitudes). They estimate the overall efficiency of the catalog tobe better than 72%, with subsamples (e.g., X-ray detected objects)being as efficient as 97%. At the faint limit of the catalog someadditional galaxy contamination is expected.There are 1,611 sources with SDSS high quality photomet-ric redshifts and no spectroscopic redshifts in ChaMP (i.e. thosewith good > . , where good is the quality flag; 6=most robust;-6=least robust; Richards et al. 2009). Among them there are 14sources with z phot > and one with z phot > , above the adoptedChaMP flux limit. All of these sources are detected in both soft andhard band. The SDSS photo- z code also gives a probability of anobject being in a given redshift range. In this way, we have not onlythe most likely redshift but also the probability that the redshift isbetween some minimum and maximum value, which is crucial fordealing with catastrophic failures. The redshift probability distribu-tion for each source is taken into account for the estimation of thenumber counts and comoving space density (see Sections 4 and 5).As for C-COSMOS selection of high-z sources, we also included13 sources having z photo + 1 σ zphoto > and z photo < . Thisadds another 10 objects to the main sample, all of them detected inboth soft and hard bands. z candidate selection and photometric redshiftestimation For the remaining 7,759 without a spectroscopic or photometricSDSS redshift, we selected the high-redshift AGN candidates us-ing their optical and/or their infrared colors. Most of these sources( ∼ %), despite being included in SDSS DR6 catalog, were re-jected from Richards et al. (2009) selection criteria. The remainingsources come from later SDSS data releases.Following the same morphological criteria as Richards et al.(2009), a candidate is required to be unresolved in images takenthrough the two redder filters (e.g. g and r for z ∼ selection). Thisminimizes contamination from low- z galaxies since even type-2AGN at z > appear point like. However, we avoid using anyfaint flux cut in order to insure that we do not miss faint high- z candidates since non-detection can imply high- z dropouts. Wereject sources with flags indicating that their photometry may beproblematic (e.g., blending of close pairs of objects, objects tooclose to the edge of the frame, objects affected by a cosmic-rayhit). Overall, we reject 5,079 non-point like sources or with prob-lematic photometry. This number ( ∼ %) is in good agreementwith the rejected number of sources by Richards et al. (2009) usingthe same criteria which explain the lack of a photometric redshiftfor these sources.Photometric redshift criteria must strike a quantifiable balancebetween completeness and efficiency, i.e., a probability can be as-signed both to the classification and the redshift. Using the SDSS,UKIDSS and WISE photometric data can help us to select quasarcandidates more efficiently than using each survey individually (seeTable 1). The photometric redshift reliability, defined by (Wu et al.2012) as the fraction of the sources with the difference between thephotometric and spectroscopic redshifts smaller than 0.2 is given inTable 1. The highest reliability can be reached only in the UKIDSSsurveyed area, which is much smaller (4,000 sq. deg) than the skycoverage of both SDSS and WISE surveys. We use the colors related to WISE W3 and W4 magnitudes only forsources lacking SDSS and/or UKIDSS detections because WISE uncertain-ties are substantially larger (Wu et al. 2004).c (cid:13) , 1–16
E. Kalfountzou et al.
Table 1.
Photometric redshift reliability defined by (Wu et al. 2012) andnumber of sources for ChaMP sources without spectroscopic or SDSS pho-tometric redshifts.Surveys a Reliability (%) N obj b N obj − lim c W 955 359U 27 19U+W 67.4 19 16S 70.4 637 367S+W 77.2 733 526S+U 84.8 71 44S+U+W 87.0 238 177Total 2,680 1,508 a S=SDSS; W=WISE; U=UKIDSS b Number of point-like objects in each combination of surveys. c Number of point-like objects in each combination of surveyswith S . − > × − erg s − cm − . Richards et al. (2002) used a 3D multi-color space to selecthigh redshift QSO candidates in SDSS: griz ( g - r , r - i , i - z ) forcandidates with z > . . Following the SDSS group, we searchfor high- z candidates in three redshift intervals ( z ≃ . − . , z ≃ . − . , z ≃ . ). The details of the selection criteriaare given in the Appendix. Our selection criteria require that oursources lie outside of a σ region surrounding the stellar locus. Westill expect the sample to be contaminated by stars and low- z galax-ies. For this reason, we use some additional criteria described byRichards et al. (2002) to exclude objects in color regions contain-ing predominantly white dwarfs, A stars and unresolved red-bluestar pairs. During the color selection process, no specific line isdrawn between optically selected quasars (type-1 AGN) and type-2AGN. Taking into account that both type-1 and type-2 AGN are un-resolved in optical images at z > and type-2 AGN should lie out-side the stellar locus due to their red optical colors, we expect thatthe above criteria efficiently select both high-redshift AGN popula-tions. We found 53 SDSS-detected high- z candidates.To increase the reliability of the photometric estimation, wealso combine the SDSS selection with the redder baselines fromUKIDSS and WISE, where the contamination of the stellar lo-cus and low-redshift galaxies is lower. We used the combinationof UKIDSS and SDSS colors in the Y − K versus g − z color-color diagram suggested by (Wu & Jia 2010) to efficiently separatequasars with redshift z < from stars. Similarly, Wu et al. (2012)suggested that z − W and g − z colors could be used to sepa-rate stars from quasars. Based on these criteria, we have rejected10 sources associated with stars based on both SDSS-UKIDSS andSDSS-WISE color-color diagrams. For sources detected only byUKIDSS we used the i = 21 . upper limit and a Y − K ver-sus i − Y color-color diagram to separate stars and low- z galaxiesfrom high- z candidates. We found 4 high- z candidates. In the caseof sources detected only by WISE there is no efficient way detailedin the literature to separate high- z quasars from stars.Photometric redshifts have been estimated for the high- z candidates by comparing the observed colors with theoreticalcolor-redshift relations derived from samples with known redshifts(Richards et al. 2002; Wu & Jia 2010; Wu et al. 2012). A standard χ minimization method is used to estimate the most probable pho-tometric redshifts. Here the χ is defined as (see Wu et al. 2004): χ = X ij [( m i,cz − m j,cz ) − ( m i,ob − m j,ob )] σ m i,ob + σ m j,ob , (1) where the sum is obtained for all four SDSS colors and/or WISEand/or UKIDSS colors, m i,cz − m j,cz is the color in the color-redshift relations, m i,ob − m j,ob is the observed color of a quasar,and σ m i,ob and σ m j,ob are the uncertainties of observed magnitudesin two bands. The uncertainty in the measurement was obtained bymapping the ∆ χ error. Since the above studies are dominated byoptically selected quasars, we would expect that the photometricredshifts uncertainties in type-1 AGN are smaller. However, sincethe Ly α break enters the g band at z ∼ . , the g − r colors quicklyredden with redshift for both populations. Alexandroff et al. (2013)found that g − r colors are indistinguishable at a 84% confidencelevel between type-1 and type-2 quasars at z > suggesting thateven in the case of type-2 AGN the photometric redshifts are reli-ably estimated. Overall, we found 8 sources with z > at greaterthan σ significance, 4 sources with z > but lower than σ sig-nificance, and 2 sources with z phot + 1 σ phot > . z sample The total z > ChaMP sample includes 87 sources with z > .Among them there are 44 sources with secure spectroscopic red-shift, 15 sources with SDSS z phot > and 13 sources with SDSS z phot + 1 σ phot > available from Richards et al. (2009), and 15sources with estimated photometric redshifts based on optical/in-frared color - redshift relations (13 with z phot > and 2 with z phot + 1 σ phot > ). Z > AGN SAMPLE
In summary, we have assembled a sample of X-ray selected AGNat z > in the C-COSMOS and ChaMP on the basis of both spec-troscopic and photometric redshifts. The total sample includes 209sources with z > . Of these, 45 are selected to be at z > fromtheir broad P ( z ) . There are also 15 C-COSMOS sources consid-ered to be at z > on the basis of their optical non-detection theseare included only in the derivation of the upper boundary of the log N − log S curve. The properties of the sample members aregiven in Table A (Appendix) and the detailed numbers are given inFigure 2. Figure 4 shows the optical and near-infrared ( i , K , and 3.6 µ m) observed magnitude distributions for the total high- z popula-tion and for sources with spectroscopic and photometric redshifts,separately. Sources selected as i -dropouts are also presented.The hard (2-10 keV rest frame) X-ray luminosity versus planeis shown in Figure 5 together with the flux limit of the C-COSMOSand ChaMP surveys (dashed line) and the applied flux cut forChaMP (dotted line). Luminosities were computed from in everycase assuming an intrinsic Γ = 1 . .The C-COSMOS & ChaMP high- z sample is a factor of 4-5 larger than all the previous individual X-ray selected samples at z > (e.g. Brusa et al. 2009; Hiroi et al. 2012; Vito et al. 2013).Most importantly, this is the first time that a significant sample of29 X-ray selected AGNs at z > is assembled. At these redshiftsprevious studies had a maximum of 9 sources. The z > X-rayselected AGNs sample also covers more than a factor of 2 of soft(2-10 keV rest frame) X-ray luminosity, and includes a significantnumber of both broad-line and non-broad line AGN.To discuss the obscured AGN fraction requires each object inour sample to be classified as obscured or unobscured. There aretwo commonly-adopted methods for classification; one is based onthe optical emission line widths (‘optical type’) or, if a spectrum c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs
15 20 25 30i-band magnitude0102030 N u m b e r o f O b j ec t s
15 20 25 30i-band magnitude0102030 N u m b e r o f O b j ec t s Totalspec-zphot-zC-COSMOS
16 18 20 22 24K-band magnitude051015202530 N u m b e r o f O b j ec t s
16 18 20 22 24K-band magnitude051015202530 N u m b e r o f O b j ec t s Totalspec-zphot-zi-dropoutC-COSMOS
14 16 18 20 22 243.6 micron magnitude0510152025 N u m b e r o f O b j ec t s
14 16 18 20 22 243.6 micron magnitude0510152025 N u m b e r o f O b j ec t s Totalspec-zphot-zi-dropoutC-COSMOS
Figure 4.
Observed AB magnitude distribution of all the i -band, K -band,and 3.6 µ m band (from top to bottom) high- z objects. Black solid, bluedot-dashed, red dashed and green solid lines represent the total, spectro-scopic, photometric redshift and i -dropout samples, respectively. The i -band dropouts are not included in the i -band histogram. is not available, by the type of template that best fits the optical-infrared spectral energy distribution (SEDs) of the sources. Theother is based on the column densities, N H , in the X-ray spectra(‘X-ray type’) or, if an X-ray spectrum is unavailable, by the hard-ness ratio (HR) (e.g. Hasinger et al. 2001). X-ray absorption shouldtypically correlate with optical AGN type. In the unified scheme(e.g. Lawrence & Elvis 1982; Antonucci 1993; Urry & Padovani1995) as the narrow emission line AGNs are viewed throughthe dusty torus, and hence have higher absorption column densi-ties than broad emission line AGNs. In fact, evidence has beenmounting over the years that the optical- and X-ray-based clas- l og ( L - k e V / e r g s - ) Figure 5.
The hard X-ray luminosity (computed with
Γ = 1 . ) redshiftplane for the objects in our sample. Blue squares = C-COSMOS sam-ple. Red circles = ChaMP sample. Filled = spectroscopic redshift. Open= photometric redshift. The dashed lines represent the 2-10 keV luminos-ity limit of the surveys computed from the 0.5-2 keV limiting flux. Thedotted red line represents the completeness flux cut we have adopted at × − erg cm − s − . The dotted black lines correspond to the fluxlimits we imposed for the computation of the space density and their asso-ciated areas, purple ( . < log L X < . ), green ( . < log L X < . ), orange ( log L X > . ). sifications often give contrasting results (Lawrence & Elvis 2010;Lanzuisi et al. 2013; Merloni et al. 2014). The optical type of the sources has been determined of the mea-sured full width at half maximum (FWHM) of the permitted emis-sion lines. Those objects with emission lines having FWHM > ,
000 km s − (e.g. Stern & Laor 2012) are classified as ‘opti-cal broad-line’ (BLAGN), and all others as ‘optical nonbroad-line’(non-BLAGN) (i.e., they show narrow emission lines or absorptionlines only; as done by Civano et al. 2011, 2012).In the C-COSMOS spectroscopic z > sample, 21 of 32sources are classified as BLAGN. These are mainly associated withthe brighter optical sources ( i AB ∼ − ) of the spectroscopicsample (see Figure 7). At fainter optical magnitudes ( i AB > ),equal numbers of broad-line and non-broad line AGNs are found.The classification for the 75 AGN in C-COSMOS with photomet-ric redshifts is obtained by the Salvato et al. (2011) photometricfitting method fitting the SED via χ minimization with code Le-Phare . More details on the fitting can be found in Salvato et al.(2011). Briefly, two libraries of templates were used, dependingon morphology, optical variability, and X-ray flux of the source.The first library (defined in Salvato et al. 2009, Table 2) con-sists of AGN templates, hybrid (host + AGN) templates, and afew normal galaxies and was used for all the point-like opticalsources and for the extended sources with an X-ray flux brighterthan × − erg cm − s − . The second library (as defined inIlbert et al. 2009) includes only normal galaxy templates and it (cid:13) , 1–16 E. Kalfountzou et al. was used for the remaining sources (i.e., extended and with X-ray flux < × − erg cm − s − ). The flowchart in Figure6 of Salvato et al. (2011) summarizes the procedure. Civano et al.(2012), according to this fitting, divide the sources into obscuredAGN, galaxies and unobscured AGN. About 40 per cent (28sources) of the photometric sample is best fitted with an unobscuredquasar template, and 47 sources with an obscured quasar template.For 29 AGN with spectroscopic identification, the photometric andspectroscopic types match. Given the mismatch rate of ∼ %, weestimate that ∼ out of the 75 AGNs could have been assigned thewrong SED classification.In the ChaMP z > spectroscopic sample, as expected atthese fluxes (e.g. Brusa et al. 2009), and due to the predominantlySDSS spectroscopic target selection, only 2/44 sources are non-BLAGN. The characterization of these sources based on their SEDfittings has been obtained by Trichas et al. (2012). In order to be inagreement with the spectroscopic ChaMP sample, we followed thesame SED fitting method for the characterization of the 43 sourceswithout a spectroscopic classification. According to this fitting, 11of 43 sources are best fitted with an obscured quasar template (non-BLAGN). More details on the fitting can be found in Trichas et al.(2012) and Ruiz et al. (2010). Briefly, a total of 16 templates hasbeen used including QSO, Seyfert-2 galaxies, starburst galaxies,absorption line galaxies and composite templates that are known toharbor both an AGN and a starburst. The Ruiz et al. (2010) modelhas been adopted, which fits all SEDs using a χ minimizationtechnique within the fitting tool Sherpa (Freeman et al. 2001). Thefitting allows for two additive components, one associated with theAGN emission and the other associated with the starburst emission.The fit with the lowest reduced χ has been chosen as the best-fitmodel. Most sources in our sample have a low number of detected counts(median ∼ in the 0.5-8.0 keV full band). In this count regime,spectral fit results are not reliable, especially if more than one freeparameter is fit; even if the fit converges the uncertainties on theparameters are large. For these reasons, we use the Bayesian Es-timation of Hardness Ratios (BEHR) method (Park et al. 2006) toderive X-ray spectral type. Hardness count ratios (HR), defined as HR = ( C HB − C SB ) / ( C HB + C SB ) , where C SB and C HB are thecounts in the soft band and hard band, respectively.BEHR is particularly powerful in the low-count Poissonregime, because it computes a realistic uncertainty for the HR,regardless of whether the X-ray source is detected in both en-ergy bands. Sources with unconstrained upper or lower limits dueto non-detections (14 hard-only and 49 soft-only detections) havebeen computed by converting the σ flux upper limit in the unde-tected band into counts.To estimate the column density, curves of constant N H as afunction of redshift have been derived for two spectral slope val-ues, Γ = 1 . and Γ = 1 . . The flatter spectral slope has beenchosen to be consistent with the assumptions adopted in producingthe original X-ray catalogs (Kim et al. 2007; Puccetti et al. 2009).The steeper value is more representative of the intrinsic value if thespectrum is not affected by obscuration (Nandra & Pounds 1994).The relationship between HR and redshift of our C-COSMOS andChaMP AGN samples is shown in Figure 6. Curves of N H = 10 , , × and cm − are reported for Γ = 1 . (dashedlines) and Γ = 1 . (solid lines). We observe that C-COSMOS sam-ple tend to be more obscured as expected due to the fainter X-ray H a r dn e ss R a ti o C-COSMOSChaMP Γ =1.8 Γ =1.4202222.723 Figure 6.
Hardness ratio versus redshift. Blue squares = C-COSMOS sam-ple. Red circles = ChaMP sample. Filled = spectroscopic redshift. Open =photometric redshift. Sources with no hard band or soft band detection areshown with arrows. Four curves of constant N H ( , , × and cm − ) are reported for Γ = 1 . (dashed lines) and Γ = 1 . (solidlines). sensitivity limit, than the ChaMP sample (Lawrence & Elvis 1982;Ueda et al. 2003; Hasinger 2008; Brusa et al. 2010; Burlon et al.2011).Though the two samples (C-COSMOS and ChaMP) of z > AGN show different trends regarding their obscuration, the largeHR errors and the similarity in this redshift range of the curveswith widely different N H values for the same spectral slope, do notallow an accurate estimate of the column density for each source tobe made. Using the CIAO spectral analysis package, Sherpa , wehave simulated X-ray spectra for AGN populations at < z < inorder to quantify the evolution of X-ray spectral slopes due to thek-correction of the observed AGN spectra toward high- z . Based onthese simulations, we find that the HR distribution for the ChaMPsample peaks at Γ ∼ . − . while the HR distribution for C-COSMOS sample peaks at Γ ∼ . − . . Hereafter, to betterconstrain the column density and for the purpose of comparisonwith previous studies, we fixed the photon index to Γ = 1 . andconverted all source fluxes.In the present work, we adopt the N H = 10 cm − limitfor the X-ray unobscured ( N H < cm − ) and obscured( N H > cm − ) classification of the sources. Based on previ-ous studies (e.g. Ueda et al. 2003; Hiroi et al. 2012; Lanzuisi et al.2013), the adopted criterion provides a good agreement with theoptical classification of the AGN. The X-ray/Optical (X/O) flux ratio is a redshift dependent quan-tity for obscured AGN, given that the k-correction is negativein the optical band and positive for the X-rays (Comastri et al.2003; Fiore et al. 2003; Brusa et al. 2010). As a result, obscuredsources have higher X/O at high redshift. On the other hand, http://cxc.harvard.edu/ciao/ http://cxc.harvard.edu/sherpa/ c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs -16 -15 -14 -13 S(0.5-2.0 keV) [erg cm -2 s -1 ]16182022242628 i A B m a g BLAGNBLAGNnonBLAGNnonBLAGNFilled = spec-zFilled = spec-zOpen = photo-zOpen = photo-z -15 -14 -13 S(2-10 keV) [erg cm -2 s -1 ]16182022242628 i A B m a g BLAGNBLAGNnonBLAGNnonBLAGNFilled = spec-zFilled = spec-zOpen = photo-zOpen = photo-z
Figure 7.
X-ray flux (soft-left, hard-right) vs. the i -band magnitude for all the X-ray sources with an i -band counterpart. The grey shaded region representsthe locus of AGNs along the correlation X / O = 0 ± . Sources with secure spectroscopic redshifts are represented by filled symbols and sources witha photometric redshift by open symbols. Orange circles and black squares represent non-BLAGN and BLAGN, respectively. Green upper limits represent i -band dropouts and black left pointing arrows represent soft and hard X-ray flux upper limits for undetected sources in each band. The C-COSMOS sampleis represented by the open big blue circles. unobscured sources have similar k-corrections in the two bands,and the distribution in X/O is not correlated with the redshift(Civano et al. 2012). Usually, the r - or i -band flux is used (e.g.Brandt & Hasinger 2005) while a soft X-ray flux was used origi-nally used for this relation with the majority of luminous spectro-scopically identified AGNs in the Einstein and ASCA surveys char-acterized by X / O = 0 ± (e.g. Schmidt et al. 1998; Stocke et al.1991; Lehmann et al. 2001). The same relation has been used alsoin the hard band, without really accounting for the X-ray bandused or the change in spectral slope (e.g. Alexander et al. 2001;Fiore et al. 2003; Brusa et al. 2003; Civano et al. 2005; Laird et al.2009; Xue et al. 2011).Figure 7 shows the distribution of X-ray soft (left) and hard(right) flux versus optical magnitude to illustrate the parameterspace spanned by the broad-line and nonbroad-line populations.The X/O ratio (Maccacaro et al. 1988) is defined as: X / O = log ( f X /f opt ) = log ( f X ) + C + m opt / . (2)where f X is the X-ray flux in a given energy range, m opt , is themagnitude at the chosen optical wavelength, and C is a constantwhich depends on the specific filter used in the optical observations.For both X-ray bands, the X / O = ± locus (grey area) has beendefined using as C ( i ) = 5 . (Civano et al. 2012), which was com-puted taking into account the width of the i -band filters in Subaru,CFHT (Canada-France-Hawaii Telescope), or for bright sourcesSDSS. In the hard band, the locus is plotted taking into accountthe band width and the spectral slope used to compute the X-rayfluxes ( Γ = 1 . ). The majority of BLAGNs with a secure spectro-scopic redshift, follow the trend of − < log ( f X /f i ) < . How-ever, given the variation in α OX with luminosity (e.g. Vignali et al.2003; Young et al. 2010; Trichas et al. 2013), there can be someshift in the locations of QSOs with luminosity within the so-calledBLAGN region. This shift is consistent with the X/O relation be-ing originally calibrated on soft-X-ray-selected sources, bright inthe optical and also in the X-rays. This might explain the mild shiftbetween the ChaMP and C-COSMOS BLAGN.Apart from the AGN population found in the BLAGN region, there is also a significant population that lie at log ( f X /f i ) > suggesting obscured nuclei. The main characteristics of this sam-ple are: 1) lack of spectroscopic redshifts (open symbols), 2) non-BLAGN optical classification (green symbols) and 3) low X-rayluminosities ( erg s − < L − < erg s − ) with N H > cm − for ∼ of them, which is consistent withprevious studies finding that mild obscuration is common at theseluminosities (e.g. Silverman et al. 2010). Furthermore, nearly 75%of all the sources with X/O > are obscured, thus confirming thatselections based on high X/O ratio are efficient in finding samplesof obscured AGN. In order to compare our optical classification to the expected ob-scuration of BLAGNs and nonbroad-line AGNs based on the uni-fied scheme we have separated our total sample into broad-line andnonbroad-line AGNs (as described in Section 3.1).In the case of the BLAGN, the X-ray classification criterion( N H = 10 cm − ) gives 28/124 X-ray obscured sources for Γ = 1 . . Half of these sources have a spectroscopic redshift andall but 3 come from C-COSMOS sample. If we also take into ac-count the HR errors, then for the lower HR limits, 11 BLAGN areclassified as X-ray obscured sources and 41 are classified X-rayobscured sources for the upper HR limits. In the non-BLAGN sub-set, the above criterion gives 49/71 X-ray obscured sources (de-tected in both soft and hard bands) for Γ = 1 . . The 27 soft bandsources in non-BLAGN sample with no detection in the hard band(reported as downward arrows in Figure 6) have very high upperlimits on the HR, due to the conservative flux upper limit computedby Puccetti et al. (2009), but most of them do not thereby satisfythe N H > cm − criterion.For the total sample, we find agreement between the opticaland X-ray classification for ∼ %: ∼ %for the BLAGN and ∼ % for the non-BLAGN. These rates are consistent with recentstudies (e.g. Lanzuisi et al. 2013; Merloni et al. 2014). Possible ex-planation can be a misclassification of faint type-1 with strong op-tical/IR contamination from host galaxy light.To improve the statistics and gain information on the aver- c (cid:13) , 1–16 E. Kalfountzou et al.
Table 2.
Comparison of optical and X-ray types. The upper and lower limitshave been estimated taking into account only the error ranges in HR.Number Unobscured Obscuredof sources N H < cm − N H > cm − BLAGN +16 − +14 − non-BLAGN +13 − +9 − H a r dn e ss R a ti o N H = 10 cm -2 N H = 5x10 cm -2 N H = 10 cm -2 N H = 10 cm -2 Figure 8.
The mean hardness ratio as a function of redshift for BLAGN(black squares) and non-BLAGN (orange circles) z > AGN subsam-ples. The error bars represent the 68% dispersion. Only sources with bothsoft and hard band detections are taken into account for the estimation ofthe mean HR in each bin. Undetected sources in one of these bands are in-cluded only for estimation of the upper and lower limits (dashed areas). The z phot = 6 . source with an upper limt HR up = 0 . has been shifteddown to HR=0.2 in order to be included in the figure. age properties of the two subclasses, we compared their mean HRvalues as a function of redshift (Figure 8). Despite the ∼ %misclassification for the individual sources, the mean propertiesof the BLAGN and non-BLAGN seem to agree with the N H ∼ cm − division. These results does not change even if we useonly sources with spectroscopic redshifts. The upper and lower lim-its detected only in the soft or the hard band were used to computethe upper and lower boundary of the shaded area. We discuss theresults in Section 5. log N- log S OF THE
Z > AGN
We derived the soft band number counts of the z > and z > samples by folding the observed flux distribution throughthe sky coverage area versus flux curve of the C-COSMOS sur-vey (Puccetti et al. 2009) and the ChaMP’s 323 fields (Green et al.2009). Additionally, we have corrected the number counts forChaMP incompleteness in the spectroscopic/photometric coverageas a function of X-ray flux (i.e. Figure 3).To minimize the error associated with the most uncertain part of the sensitivity curve, we truncate the C-COSMOS sample at theflux corresponding to 10% of the total area (blue dashed line inFigure 1). Following Civano et al. (2011), all the sources with a . − keV flux above × − erg cm − s − have been con-sidered (73 objects out of the 81 soft band detected). The flux limitapplied to the sample is consistent with the signal-to-noise ratiothresholds chosen by Puccetti et al. (2009), on the basis of exten-sive simulations, to avoid the Eddington bias in the computationof the number counts of the entire C-COSMOS sample. Thus, byapplying a flux limit cut, we also reduce the Eddington bias af-fecting our sample. For ChaMP this would be at S . − > × − erg cm − s − , below the flux limit already applied.The binned logN-logS relations for two redshift ranges ( z > ; orange points and z > ; blue points, with associated errors)are plotted in Figure 9 (top panel). In integral form, the cumulativesource distribution is represented by: N ( > S ) = N S X i =1 i (3)where N ( > S ) is the number of sources with a flux greater than Sand Ω i , is the limiting sky coverage associated with the i th source.The associated error is the variance: σ = N S X i =1 ( 1Ω i ) (4)The grey shaded area represents an estimate of the maximum andminimum number counts relation at z > obtained by consideringthree different effects:1) the σ uncertainty in the sky coverage area for each sourceusing the sky coverage as a function of flux (see Figure 1) and the σ uncertainty in the flux;2) the 14 sources from C-COSMOS with no-optical detection(seen in the soft band);3) the sources with photometric redshift z photo < but z photo + σ zphoto > .To compute the upper boundary of the shaded area, we in-cluded all the sources in the main sample plus the sources with nooptical detection at their flux +1 σ error. For the lower boundary,we used the flux − σ error only for the sources with z spec > .Under these assumptions, the number counts estimation would fol-low the lower boundary curve only in the very unlikely case that allthe photometric redshifts are overestimated, while the observationswould be described by the upper boundary if the selection of highredshift X-ray sources based on the lack of optical detection was a100% reliable proxy.We have compared our number counts with previous X-ray surveys that span the range from deep, small area (CDFS,at 464.5 arcmin with a soft band flux limit of ∼ . × − erg cm − s − , Xue et al. 2011), to moderate area and mod-erate depth (Chandra-COSMOS at 0.9 deg with a 0.5-2 keV fluxlimit of ∼ . − erg cm − s − , Elvis et al. 2009) and finally tomoderate area and shallower depth (XMM-COSMOS, at 2 deg and a soft band flux limit of ∼ . × − erg cm − s − ,Cappelluti et al. 2009). The binned logN-logS relations are plot-ted in Figure 9 (top panel), together with the XMM-COSMOS(Brusa et al. 2009, open circles), C-COSMOS (Civano et al. 2011,open squares) and 4 Ms CDF-S number counts (Vito et al. 2013,filled triangles).Good agreement exists among the comparison surveys pre- c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs N ( > S ) d e g - z > 3 Civano+11Vito+12Brusa+09 z > 4 ( N / N e xp ) [ > S ] LDDE+expLADECT AGNs LDDE -16 -15 -14 S(0.5-2.0 keV) [erg cm -2 s -1 ]1 ( N / N e xp ) [ > S ] LDDE+exp LADE CT AGNs
Figure 9.
Top: The binned logN-logS relation (with associated errors) of the z > (orange circles) and z > (blue squares) QSOs population. The greyshaded area represents the maximum and minimum number counts underthe assumptions described in Section 4. The blue and orange curves corre-spond to the prediction based on the LDDE+exp (thick solid), LDDE (dot-ted), LADE (dashed) and CT AGNs (dash cross)models for each redshiftrange, respectively. The small open circles represents the number counts es-timated by Brusa et al. (2009), the small open squares are from Civano et al.(2011) work and small filled triangles from Vito et al. (2013). Middle: Theratio of the observed number counts for z > relative to the LDDE+expmodel (thick solid line at the top panel). The thick solid line represents theLDDE+exp model N/N exp = 1 , the dotted line represents the LDDE rel-ative to the LDDE+exp model, the dashed lines the LADE relative to theLDDE+exp model and the dash cross lines the CT AGNs model relative tothe LDDE+exp model. Bottom: The ratio of the observed number countsfor z > relative to the LDDE+exp model. Symbols are similar to themedium panel. The ratio of the LDDE relative to the LDDE+exp model is N/N exp > . and is not presented. sented here. At z > and fainter X-ray fluxes ( S . − < × − erg cm − s − ) our points confirm the agreement withthe model predictions, previously found by Brusa et al. (2009);Civano et al. (2011); Vito et al. (2013). At the same redshift rangebut brighter X-ray fluxes, where only XMM-COSMOS samplehave available points (Brusa et al. 2009) based on 4 sources, wereduce the uncertainties by factor of 4 using a sample of 66 sourceswith ( S . − > × − erg cm − s − ). Notably it is the firsttime that data points at the bright end at z > are included. At red-shift z > , where the XMM-COSMOS sample had only 4 sources,the 4 Ms CDF-S 9 sources and the C-COSMOS 14 sources, theC-COSMOS & ChaMP sample has 27 sources (2-5 times larger),making it possible to compare the slope of the counts with models.A comparison with the AGN number counts from three differ-ent phenomenological model predictions is also presented in Fig-ure 9 with the different types of curves (orange color for z > andblue color for z > ):1) The thick solid lines correspond to the predictions of theXRB synthesis model of Gilli et al. (2007), based on the X-rayluminosity function observed at low redshift (e.g. Hasinger et al.2005), parametrized with a luminosity dependent density evolu-tion (LDDE) and a high redshift exponential decline with thesame functional form adopted by (Schmidt et al. 1995) ( Φ( z ) =Φ( z ) × − . z − z ) and z = 2 . ) to fit the optical luminosityfunction between z ∼ . − (Fan et al. 2001), corresponding toone e-folding per unit redshift (hereafter referred to as LDDE+exp).2) The dotted curves correspond to the predictions of theLDDE model without the high- z decline (Gilli et al. 2007),obtained extrapolating to high-z the best-fit parameters of(Hasinger et al. 2005).3) The dashed line is the luminosity and density evolutionmodel (LADE; Aird et al. 2010) which fits the hard X-ray lumi-nosity function derived by (Aird et al. 2010) using the 2Ms Chan-dra Deep Fields and the AEGIS-X (200 ksec) survey to probe thefaint end ( log L X <
43 erg s − ) and the high-z ( z ∼ ) range.4) The dash-crossed lines correspond to the Treister et al.(2009) X-ray background population synthesis predictions (CTAGNs). While at z > the two model predictions are very close, at z > , where the models have different slopes, the errors on thedata of the previous studies do not allow a firm preference of oneof the two models, highlighting the advantage of our sample withrespect to previous surveys. In this work we find that the LDDEmodel (Gilli et al. 2007) without decline clearly overestimates theobserved counts even in the most optimistic scenario (upper bound-ary) in both the z > and z > redshift ranges. Our results forthe z > sample (orange color) are in good agreement with bothLDDE+exp (thick solid line) and LADE (dashed line) model pre-dictions but only up to flux ∼ × − erg cm − s − , wherethe difference of the two models is < %. However, the main ad-vantage of our sample becomes clear at brighter fluxes; our resultsstrongly exclude the LADE model. This is in contrast to previousstudies which could not distinguish between the two models due totheir large uncertainties.At z > (blue color) our results are in good agreement withLDDE+exp predictions. We can not clearly exclude the LADEmodel if we take into account the upper and lower boundaries. Model predictions from the work of Treister et al. (2009)for a range of input values are publicly available athttp://agn.astroudec.cl/j agn/main.htmlc (cid:13) , 1–16 E. Kalfountzou et al.
However, we can point out for first time that there is no sign of theexpected decline to higher fluxes. LDDE is fully ruled out at z > .While fainter samples would be also useful for a better descriptionof the model, considering only the z > subsample by Vito et al.(2013) (4 Ms CDF-S, filled triangles), the data lie between theLDDE+exp and LADE models prediction. Vito et al. (2013) haveincluded the presence of 3 sources at < z < . , whose redshiftsare determined on the basis of relatively uncertain photometric in-formation. If these sources were placed at < z < , a good agree-ment would be obtained with the LDDE+exp model (see Vito et al.2013, Figure 9). In this case, the Vito et al. 2013 z > samplewould be consisted only by 5 sources. To investigate the cosmological evolution of AGN at z > thecomoving space densities were calculated from our sample utiliz-ing the /V max method (Schmidt 1968). This method takes intoaccount the fact that more luminous objects are detectable over alarger volume and is readily adapted to the case in which the sur-vey area depends on flux.The maximum available volume, over which each source canbe detected, was computed by using the formula: V max = Z z max z min Ω( f ( L X , z, N H )) dVdz dz (5)where Ω( f ( L X , z, N H )) is the sky coverage at the flux f ( L X , z ) corresponding to a source with absorption column density N H andobserved luminosity L X , and z max is the maximum redshift atwhich the source can be observed at the flux limit of the survey.If z max > z up,bin , where z up,bin is the maximum redshift of theredshift bin, then the z max is the upper boundary of the redshift binused for computing V max . In the case of the ChaMP sample, z max is estimated using both the X-ray and optical survey limits and is se-lected to be the minimum of the two estimates so the source can beobserved at the flux limit of both surveys. We computed the spacedensity using the luminosities derived with Γ = 1 . . The contri-bution of sources with photometric redshift to the space density isweighted for the fraction of their P ( z ) at z > .After calculating the V max for each source, we sum the valuesin each redshift bin: φ = z min Figure 10. The comoving space density in 3 different 2-10 keV X-ray lumi-nosity ranges. The solid lines corresponds to the X-ray selected AGN spacedensity computed for the same luminosity limit from the Gilli et al. 2007LDDE+exp model. The dashed curve corresponds to the space density de-rived from the LADE model of Aird et al. 2010. The colors respond to theshaded areas in Figure 5 and the shaded area represents the maximum andminimum space density under the assumptions described in the text. Whenonly one source is included in the bin it has been plotted as an upper limit(at σ ). The small black symbols and dotted lines correspond to the comov-ing space density data points and model derived from Ueda et al. 2014 forsimilar X-ray luminosity ranges. which emit more at harder energies, without having to introduceany further correction or assumption.The resulting comoving space densities are shown in Fig-ure 10. To reduce the effects of incompleteness and to have a com-plete sample over a given redshift range, we divided the sample inthree luminosity intervals (see Figure 5, shaded areas):1) At low luminosities (purple shading) we computed thespace density in 3 redshift bins ( z = 3 − . ) at . < log( L X / erg s − ) . .2) At intermediate luminosities (green shading) we computedthe space density in 3 redshift bins ( z = 3 − . ) at . < log( L X / erg s − ) . 3) At high luminosities (orange shading) we computedthe space density in 5 redshift bins ( z = 3 − . ) at log( L X / erg s − ) > . .The X-ray flux errors have been taken into account for the es-timation of the space density upper and lower limits. For example,if the flux of a source is lower than the applied flux limit but itsflux +1 σ is higher than this limit, then the source will be includedin the upper boundary sample. Similarly, if the flux of a source ex-ceed the applied flux limit but its flux − σ is lower than this limit,then taking into account the lower limit, this source will be by thelower boundary sample. The shaded areas include the above uncer-tainties affecting the computation of the space density, i.e., the fluxerrors and thus errors on the maximum volume associated to eachsource.As explained in Section 2.1, the 15 sources with no opti-cal band detection from C-COSMOS survey have not been in-cluded in the space density boundaries. However, we computedthe space density assuming that all the 15 sources were at the red-shift corresponding to the first bin, then to the second bin and so c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs on (Civano et al. 2011). The space density values computed in thiscase, in the first three bins, are within the shaded areas.The space density in the three luminosity ranges is comparedwith the predictions, at the same luminosity threshold, from thesame three models discussed in Section 4. The LDDE+exp model(Gilli et al. 2007) used for the logN-logS (solid lines), including inthe model all the sources up to a column density of cm − . Wealso compare with the LADE model (Aird et al. 2010; dashed line).The LDDE model is fully ruled so it is not included in the follow-ing comparisons. In agreement with the results obtained from thenumber counts, the LDDE+exp model provides an excellent repre-sentation of the observed data, although the LADE model cannotbe rejected taking into account the upper and lower boundaries. Weconfirm that the shape of the space-density evolution of X-ray se-lected luminous AGN is consistent with that derived from opticalquasar surveys within current uncertainties.The results from the Ueda et al. (2014) for AGNs with thesame X-ray luminosity ranges are also plotted for comparison(small black symbols). As can be seen, our results are consistentwith those of Ueda et al. (2014) within the statistical errors, indi-cating a significant decline in the AGN space density from z = 3 tohigher redshifts. To take into account the observed decline in theirLDDE model (Figure 10; black dotted lines), Ueda et al. (2014)introduced another (luminosity-dependent) cutoff redshift abovewhich the model declines. Their model indicates an ‘up-sizing’evolution instead of the global ‘downsizing’ evolution, where moreluminous AGN have their number density peak at higher redshiftscompared with less luminous ones. In the cases of lower and in-termediate X-ray luminosities the Ueda et al. (2014) LDDE modelslightly overestimates our results but it is within the upper bound-aries. In the case of the higher X-ray luminosities, where our datasignificantly reduce the uncertainties of the Ueda et al. (2014) sam-ple, their model underestimates our result while a our data may in-dicate a flatting up to redshift z ∼ . Comparing the high redshift evolution of optical- and X-ray- se-lected AGN samples, the same decline profile has been revealed.Considering that X-ray selected samples provide reduced obscura-tion bias in comparison with optically selected AGN, this similaritysuggests no significant cosmological evolution of obscured AGNfraction at least at higher redshifts. Supporting these result, previ-ous studies have been concluded that the obscured AGN fractionincreases with redshift z = 1 to z = 2 (Ballantyne et al. 2006)but decreases at higher redshifts (Hasinger 2008). To investigatefurther the cosmological evolution of type-1 and type-2 AGN andtheir fraction with in the redshift range of z = 3 − , we calculatethe co-moving space density for the two sub-classes of AGN in oursample following the same method as described in Section 5.The co-moving space density is shown in Figure 11. The up-per and lower boundaries are estimated similarly to Section 5 takinginto account the X-ray flux errors and the 15 C-COSMOS sourcewith no optical band detection. We have used both optical (largesymbols) and X-ray (small symbols) AGN classification and theyseem to be in a good agreement. To estimate the upper and lowerboundaries, we also take into account the ∼ % of mismatchesinto the optical and the X-ray classification of AGN types, in or-der to estimate the upper and lower boundaries. Specifically, in thecase of optically classified type-1 AGN, for the upper boundarywe also include in each bin the sources classified as unobscured( N H < cm − ) even if they are defined as type-2 AGN based on their optical classification. For the lower boundary, we excludefrom each bin the type-1 AGN which have N H > cm − . Thesame method is also applied for the type-2 AGN. So, the upperand lower boundaries include also the uncertainties due to the mis-classification of sources.The space density in the three luminosity ranges is comparedwith the predictions, at the same luminosity threshold, from thesame LDDE+exp Gilli et al. (2007) model, including in the modelall the sources with N H cm − for the case of type-1AGN and all the sources with N H = 10 − cm − for thecase of type-2 AGN. In agreement with our previous results, theLDDE+exp model provides an excellent representation of bothtype-1 and type-2 AGN. Since both type-1 and type-2 AGN fol-low the same decline profile it suggests that there is no significantcosmological evolution for their fraction above z > . The sameresults are obtained even if we follow the X-ray classification ofthe AGN sample into obscured and unobscured sources based (Fig-ure 11; small black symbols).The LDDE+exp model for these three luminosity ranges andthe N H = 10 cm − division predicts a fraction of objects clas-sified as type-2 over the type-1 AGN sources of 0.63, 0.5 and 0.48at low luminosities . < log( L X / erg s − ) . , intermedi-ate luminosities . < log( L X / erg s − ) . and high lumi-nosities log( L X / erg s − ) > . respectively. For the same lu-minosity bins, we have calculated the mean co-moving space den-sity ratio for the two types Φ type / Φ type = 0 . ± . , . ± . 04 and 0 . ± . from low to high luminosities. These are ingood agreement with the LDDE+exp model predictions.Recently, Hiroi et al. (2012) estimated the X-ray type-2 AGNfraction, classified based on the N H > cm − criterion, to be . +0 . − . at z = 3 . − . and in the luminosity range of log L X =44 . − . while for their optical selection of type-2 AGN theyfound a fraction of . ± . . Their estimates are somewhat largerthan our result of . ± . , although they agree within the errors.This difference could be easily explained from the fact that theirsample is significantly smaller (30 sources) than ours and containsonly 4 sources with z > from which the 2 classified as type-2AGN have only photometric redshifts. We have presented the results of the largest X-ray selected sampleof z > AGN to date, compiled from the C-COSMOS and ChaMPsurveys. The large body of C-COSMOS and ChaMP data and theircombination allowed us to devise a robust method to build a sizablesample of X-ray selected AGN and control the selection effects in-cluding both type-1 (unobscured) and type-2 (obscured) AGN, at z > . Our sample consists of 209 detections in the soft and/orhard and/or full band. We find:1) The average HR of the type-1 and type-2 AGN samplesis consistent with the N H < cm − criterion for the classifi-cation of the X-ray unobscured AGN and the N H > cm − criterion for the classification of the X-ray obscured AGN, respec-tively.2) For the individual sources there is a mis-match of ∼ %between optical and X-ray classification. The contribution fromstarburst emission in the soft band, or misclassification of fainttype-1 with strong optical/IR contamination from host galaxy lightcould possibly explain the differences between the two classifica-tions.3) The number counts derived in this work (Figure 9) are con- c (cid:13) , 1–16 E. Kalfountzou et al. -8 -7 -6 -5 Φ ( M p c - ) LDDE+expN H < 10 cm -2 Optical ClassificationX-ray Classification -8 -7 -6 -5 Φ ( M p c - ) LDDE+expN H > 10 cm -2 Optical ClassificationX-ray Classification Figure 11. The comoving space density in 3 different 2-10 keV X-ray luminosity ranges for type-1 (left) and type-2 (right) AGN. The solid lines correspondsto the X-ray selected AGN space density computed for the same luminosity limit from the Gilli et al. 2007 model and for N H cm in the case oftype-1 AGN and N H > cm for type-2 AGN. The colors respond to the shaded areas in Figure 5 and the shaded area represents the maximum andminimum space density under the assumptions described in Section 5. The small black symbols represent the comoving space density in the 3 different 2-10keV X-ray luminosity ranges for sources classified as unobscured (left) and obscured (right) AGN based on the X-ray classification and the N H = 10 cm limit. sistent with previous determinations from the literature, yet signif-icantly reduce the uncertainties especially at bright fluxes and athigh redshifts ( z > ). The number counts of our combined C-COSMOS/ChaMP sources are consistent with the trend that thespace density significantly declines at higher redshifts, similarlyto XMM (Brusa et al. 2009) and C-COSMOS (Civano et al. 2011)results at similar fluxes at least within their errors and CDF-S(Vito et al. 2013) at fainter fluxes, and are better described by theLDDE+exp model (Gilli et al. 2007).In contrast to the previous studies, and due mainly to the largesample and wide flux coverage, our results exclude the Aird et al.(2010) LADE model, at the brighter fluxes. At fluxes × − F . − × − erg cm − s − the predictions of thismodel are very similar to the Gilli et al. (2007) LDDE+exp model,but at fainter and brighter fluxes the two models deviate signif-icantly. The Vito et al. (2013) results trace the Gilli et al. (2007)LDDE+exp model well at fainter fluxes, but only for z > , whileour sample give the same results at the brighter fluxes where (dueto the low expected counts) a large sample is required.4) In agreement with the number counts, the space density iswell-described with the LDDE+exp model at all X-ray luminos-ity bins and redshifts, while the LADE model fails to fit the data.These results confirm the declining space density as observed in theoptical wavelengths.5) Taking into account both optical and X-ray classifica-tions, we derived the space density for type-1 and type-2 (ob-scured and unobscured) AGN separately. In both cases the resultsare in agreement with the LDDE+exp model suggesting that thehigh-redshift evolution of obscured AGNs is similar to that of un-obscured AGN. For each luminosity bin, we derived the type-2AGN fraction among the total AGN sample to be . ± . , . ± . and . ± . at z > in the luminosity rangesof L X = 10 . − . , . − . , . − . erg s − .The last result should have a significant impact on our under-standing of the galaxy and black hole co-evolution. Either or bothline-of-sight orientation or evolutionary phase (e.g., covering fac- tor of high column obscuration) can affect the apparent obscuration(and therefore classification) of AGN. Given that orientation doesnot evolve preferentially towards us, this work allows us to to saythat the obscured and unobscured evolutionary phases do evolvesimilarly. ACKNOWLEDGMENTS The authors thank J. Aird and E. Glikman for sharing their lumi-nosity functions and Y. Ueda for providing his space density esti-mations. The authors would like to thank the referee F. Bauer forthe helpful and constructive report. This work was supported byNASA Chandra archival grant, designated code 617817 and con-tract number AR2-13010X. REFERENCES Aird J. et al., 2010, MNRAS, 401, 2531Alexander D. M., Brandt W. N., Hornschemeier A. E., GarmireG. P., Schneider D. P., Bauer F. E., Griffiths R. E., 2001, AJ,122, 2156Alexandroff R. et al., 2013, MNRAS, 435, 3306Antonucci R., 1993, ARA&A, 31, 473Ballantyne D. R., 2008, ApJ, 685, 787Ballantyne D. R., Everett J. E., Murray N., 2006, ApJ, 639, 740Barger A. J., Cowie L. L., 2005, ApJ, 635, 115Brandt W. N., Hasinger G., 2005, ARA&A, 43, 827Brusa M. et al., 2010, ApJ, 716, 348Brusa M. et al., 2009, ApJ, 693, 8Brusa M. et al., 2003, A&A, 409, 65Burlon D., Ajello M., Greiner J., Comastri A., Merloni A.,Gehrels N., 2011, ApJ, 728, 58Cappelluti N. et al., 2009, A&A, 497, 635Casali M. et al., 2007, A&A, 467, 777 c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs Civano F. et al., 2011, ApJ, 741, 91Civano F., Comastri A., Brusa M., 2005, MNRAS, 358, 693Civano F. et al., 2012, ApJS, 201, 30Comastri A., Brunetti G., Dallacasa D., Bondi M., Pedani M.,Setti G., 2003, MNRAS, 340, L52Cowie L. L., Songaila A., Hu E. M., Cohen J. G., 1996, AJ, 112,839Croton D. J. et al., 2006, MNRAS, 365, 11Damen M., Labb´e I., Franx M., van Dokkum P. G., Taylor E. N.,Gawiser E. J., 2009, ApJ, 690, 937Draper A. R., Ballantyne D. R., 2010, ApJ, 715, L99Dye S. et al., 2006, MNRAS, 372, 1227Elvis M. et al., 2009, ApJS, 184, 158Fan X. et al., 2001, AJ, 121, 54Ferrarese L., Merritt D., 2000, ApJ, 539, L9Fiore F. et al., 2003, A&A, 409, 79Freeman P., Doe S., Siemiginowska A., 2001, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series,Vol. 4477, Astronomical Data Analysis, Starck J.-L., MurtaghF. D., eds., pp. 76–87Garmire G. P., Bautz M. W., Ford P. G., Nousek J. A., Ricker, Jr.G. R., 2003, in Society of Photo-Optical Instrumentation Engi-neers (SPIE) Conference Series, Vol. 4851, X-Ray and Gamma-Ray Telescopes and Instruments for Astronomy., Truemper J. E.,Tananbaum H. D., eds., pp. 28–44Gebhardt K. et al., 2000, ApJ, 539, L13Gilli R., Comastri A., Hasinger G., 2007, A&A, 463, 79Glikman E., Djorgovski S. G., Stern D., Dey A., Jannuzi B. T.,Lee K.-S., 2011, ApJ, 728, L26Granato G. L., De Zotti G., Silva L., Bressan A., Danese L., 2004,ApJ, 600, 580Granato G. L., Silva L., Monaco P., Panuzzo P., Salucci P., DeZotti G., Danese L., 2001, MNRAS, 324, 757Green P. J. et al., 2009, ApJ, 690, 644Hambly N. C. et al., 2008, MNRAS, 384, 637Hasinger G., 2008, A&A, 490, 905Hasinger G. et al., 2001, A&A, 365, L45Hasinger G., Miyaji T., Schmidt M., 2005, A&A, 441, 417Hewett P. C., Warren S. J., Leggett S. K., Hodgkin S. T., 2006,MNRAS, 367, 454Hiroi K., Ueda Y., Akiyama M., Watson M. G., 2012, ApJ, 758,49Hodgkin S. T., Irwin M. J., Hewett P. C., Warren S. J., 2009, MN-RAS, 394, 675Hopkins P. F., Somerville R. S., Hernquist L., Cox T. J., RobertsonB., Li Y., 2006, ApJ, 652, 864Ikeda H. et al., 2011, ApJ, 728, L25Ilbert O. et al., 2009, ApJ, 690, 1236Jiang L. et al., 2009, AJ, 138, 305Jiang L., Fan X., Vestergaard M., Kurk J. D., Walter F., KellyB. C., Strauss M. A., 2007, AJ, 134, 1150Kalfountzou E., Jarvis M. J., Bonfield D. G., Hardcastle M. J.,2012, MNRAS, 427, 2401Kalfountzou E. et al., 2014, MNRAS, 442, 1181Kalfountzou E., Trichas M., Rowan-Robinson M., Clements D.,Babbedge T., Seiradakis J. H., 2011, MNRAS, 413, 249Kim M. et al., 2007, ApJS, 169, 401Kodama T. et al., 2004, MNRAS, 350, 1005Koekemoer A. M. et al., 2004, ApJ, 600, L123La Franca F. et al., 2005, ApJ, 635, 864Laird E. S. et al., 2009, ApJS, 180, 102Lanzuisi G. et al., 2013, MNRAS, 431, 978 Lawrence A., Elvis M., 1982, ApJ, 256, 410Lawrence A., Elvis M., 2010, ApJ, 714, 561Lawrence A. et al., 2007, MNRAS, 379, 1599Lehmann I. et al., 2001, A&A, 371, 833Lilly S. J. et al., 2009, ApJS, 184, 218Lilly S. J. et al., 2007, ApJS, 172, 70Maccacaro T., Gioia I. M., Wolter A., Zamorani G., Stocke J. T.,1988, ApJ, 326, 680Maddox N., Hewett P. C., Warren S. J., Croom S. M., 2008, MN-RAS, 386, 1605Magorrian J. et al., 1998, AJ, 115, 2285McConnell N. J., Ma C.-P., 2013, ApJ, 764, 184Menci N., Fiore F., Puccetti S., Cavaliere A., 2008, ApJ, 686, 219Merloni A. et al., 2014, MNRAS, 437, 3550Nandra K., Pounds K. A., 1994, MNRAS, 268, 405Noterdaeme P. et al., 2012, A&A, 547, L1Pˆaris I. et al., 2014, A&A, 563, A54Park T., Kashyap V. L., Siemiginowska A., van Dyk D. A., ZezasA., Heinke C., Wargelin B. J., 2006, ApJ, 652, 610Puccetti S. et al., 2009, ApJS, 185, 586Richards G. T. et al., 2002, AJ, 123, 2945Richards G. T. et al., 2009, ApJS, 180, 67Richards G. T. et al., 2006, AJ, 131, 2766Ross N. P. et al., 2013, ApJ, 773, 14Ruiz A., Miniutti G., Panessa F., Carrera F. J., 2010, A&A, 515,A99Salvato M. et al., 2009, ApJ, 690, 1250Salvato M. et al., 2011, ApJ, 742, 61Schmidt M., 1968, ApJ, 151, 393Schmidt M. et al., 1998, A&A, 329, 495Schmidt M., Schneider D. P., Gunn J. E., 1995, AJ, 110, 68Silverman J. D. et al., 2008, ApJ, 679, 118Silverman J. D. et al., 2010, ApJS, 191, 124Spergel D. N. et al., 2003, ApJS, 148, 175Steidel C. C., Giavalisco M., Pettini M., Dickinson M., Adel-berger K. L., 1996, ApJ, 462, L17Stern J., Laor A., 2012, MNRAS, 426, 2703Stocke J. T., Morris S. L., Gioia I. M., Maccacaro T., Schild R.,Wolter A., Fleming T. A., Henry J. P., 1991, ApJS, 76, 813Treister E., Urry C. M., 2006, ApJ, 652, L79Treister E., Urry C. M., Virani S., 2009, ApJ, 696, 110Trichas M., Georgakakis A., Rowan-Robinson M., Nandra K.,Clements D., Vaccari M., 2009, MNRAS, 399, 663Trichas M. et al., 2013, ApJ, 778, 188Trichas M. et al., 2012, ApJS, 200, 17Trichas M. et al., 2010, MNRAS, 405, 2243Ueda Y., Akiyama M., Hasinger G., Miyaji T., Watson M. G.,2014, ApJ, 786, 104Ueda Y., Akiyama M., Ohta K., Miyaji T., 2003, ApJ, 598, 886Urry C. M., Padovani P., 1995, PASP, 107, 803Vignali C., Brandt W. N., Schneider D. P., 2003, AJ, 125, 433Vito F. et al., 2013, MNRAS, 428, 354Warren S. J., Hewett P. C., Foltz C. B., 2000, MNRAS, 312, 827Willott C. J. et al., 2010, AJ, 139, 906Wright E. L. et al., 2010, AJ, 140, 1868Wu X.-B., Hao G., Jia Z., Zhang Y., Peng N., 2012, AJ, 144, 49Wu X.-B., Jia Z., 2010, MNRAS, 406, 1583Wu X.-B., Zhang W., Zhou X., 2004, ChjAA, 4, 17Xue Y. Q. et al., 2011, ApJS, 195, 10Young M., Elvis M., Risaliti G., 2010, ApJ, 708, 1388 c (cid:13) , 1–16 E. Kalfountzou et al. APPENDIX A: QSO candidates at redshift ≃ . − . satisfied the following cuts: σ r < . 13 AND u > . u − g > . g − r < . r − i < . i − z > − . g − r < . u − g ) − . . (A1)For the redshift range ≃ . − . this selection becomes A) σ r < . u − g > . OR u > . g − r > . g − r > . OR r − i < . g − r ) − . i − z < . 25 AND i − z > − . , (A2)in the combination A AND B AND C AND D AND E. For theredshifts above ≃ . we use u > . g > . r − i > . i − z > − . i − z < . r − i ) − . . (A3)This paper has been typeset from a TEX/ L A TEX file prepared by theauthor. c (cid:13) , 1–16 he Largest X-ray Selected Sample of z > AGNs Table A1. Properties of the high - redshift AGN sample.Survey a R.A. Dec. z spec z phot z phot b z phot c Optical X-ray S soft d log Lxe HR f HR HR N H g (deg) (deg) lower upper Type h Type i low j up k < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . a b Photometric redshift lower limit c Photometric redshift upper limit d The 0.5-2 keV flux in units of − erg cm − s − converted to Γ = 1 . . For sources undetected in soft band a negative symbol is given and the flux estimatedby the hard or full band detection. e The 2-10 keV luminosity in units of erg s − . f Hardness ratio defined as HR = (H + S)/(H - S), S: 0.5-2.0 keV count rate and H: 2.0-10 keV count rate. g The absorption column density in units of cm − h 1: optical type-1 AGNs, and 2: optical type-2 AGNs i 1: unobscured AGNs, and 2: obscured AGNs; based on N H = 10 cm − limit. j Hardness ratio lower limit k Hardness ratio upper limitc (cid:13) , 1–16 E. Kalfountzou et al. Survey R.A. Dec. z spec z phot z phot z phot Optical X-ray S soft log Lx HR HR HR N H (deg) (deg) lower upper Type Type lower upper1 195.445 -0.112 3.7 ... ... ... 1 2 0.160 45.17 -0.16 -0.41 0.09 20.01 196.534 3.940 6.016 ... ... ... 1 1 0.031 44.93 -0.69 -0.75 -0.61 < . < . < . < . < . < . < . < . 11 208.2111 33.4822 ... 2.755 2.32 3.07 1 1 0.056 44.54 -0.462 -0.57 -0.30 0.11 212.7671 52.2988 ... 2.815 2.66 3.13 1 1 0.097 44.99 -0.42 -0.48 -0.36 0.11 214.2243 44.6294 ... 2.795 2.52 3.12 1 1 0.134 44.94 -0.54 -0.66 -0.43 < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . > > < . > < . > < . < . > < > > < . < . > . c (cid:13)000 11 208.2111 33.4822 ... 2.755 2.32 3.07 1 1 0.056 44.54 -0.462 -0.57 -0.30 0.11 212.7671 52.2988 ... 2.815 2.66 3.13 1 1 0.097 44.99 -0.42 -0.48 -0.36 0.11 214.2243 44.6294 ... 2.795 2.52 3.12 1 1 0.134 44.94 -0.54 -0.66 -0.43 < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . < . > > < . > < . > < . < . > < > > < . < . > . c (cid:13)000 , 1–16 he Largest X-ray Selected Sample of z > AGNs Survey R.A. Dec. z spec z phot z phot z phot Optical X-ray S soft log Lx HR HR HR N H (deg) (deg) lower upper Type Type lower upper2 149.8894 1.9662 ... 3.053 2.2 3.12 2 1 0.004 43.47 ... ... -0.32 < . < . > < . < . < . > < . < . > < . > . > . < . < . > < . < . > > > . > . < . > . < . < . < . < . < . < . < . < . > > . < . > > < . < . c (cid:13) , 1–16 E. Kalfountzou et al. Survey R.A. Dec. z spec z phot z phot z phot Optical X-ray S soft log Lx HR HR HR N H (deg) (deg) lower upper Type Type lower upper2 150.3007 2.3007 3.498 3.434 3.41 3.46 2 2 0.003 43.55 ... ... 0.70 < . < . < . > < . < . > < . < . > . < . > . > < . < . < . < . > . > > (cid:13)000