ebook img

The evolution of the X-ray luminosity functions of unabsorbed and absorbed AGNs out to z∼ 5 PDF

36 Pages·2015·4.06 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The evolution of the X-ray luminosity functions of unabsorbed and absorbed AGNs out to z∼ 5

MNRAS451,1892–1927(2015) doi:10.1093/mnras/stv1062 The evolution of the X-ray luminosity functions of unabsorbed and ∼ absorbed AGNs out to z 5 J. Aird,1,2‹ A. L. Coil,3 A. Georgakakis,4,5 K. Nandra,4 G. Barro6 and P. G. Pe´rez-Gonza´lez7 1InstituteofAstronomy,UniversityofCambridge,MadingleyRoad,CambridgeCB30HA,UK 2DepartmentofPhysics,DurhamUniversity,DurhamDH13LE,UK 3CenterforAstrophysicsandSpaceSciences(CASS),DepartmentofPhysics,UniversityofCalifornia,SanDiego,CA92093,USA 4MaxPlanckInstitutefu¨rExtraterrestrischePhysik,Giessenbachstrasse,D-85748Garching,Germany 5IAASARS,NationalObservatoryofAthens,GR-15236Penteli,Greece D 6UniversityofCalifornia,SantaCruz,1156HighStreet,SantaCruz,CA95064,USA o w 7DepartamentodeAstrof´ısica,FacultaddeCC.F´ısicas,UniversidadComplutensedeMadrid,E-28040Madrid,Spain n lo a d e d Accepted2015May8.Received2015May8;inoriginalform2015February24 fro m h ttp s ABSTRACT ://a WepresentnewmeasurementsoftheevolutionoftheX-rayluminosityfunctions(XLFs)of ca d unabsorbed and absorbed active galactic nuclei (AGNs) out to z ∼ 5. We construct samples em containing2957sourcesdetectedathard(2–7keV)X-rayenergiesand4351sourcesdetected ic.o u atsoft(0.5–2keV)energiesfromacompilationofChandrasurveyssupplementedbywide- p .c areasurveysfromASCAandROSAT.WeconsiderthehardandsoftX-raysamplesseparately o m andfindthattheXLFbasedoneither(initiallyneglectingabsorptioneffects)isbestdescribed /m n byanewflexiblemodelparametrizationwherethebreakluminosity,normalization,andfaint- ra s endslopeallevolvewithredshift.Wethenincorporateabsorptioneffects,separatelymodelling /a theevolutionoftheXLFsofunabsorbed(20<logNH<22)andabsorbed(22<logNH<24) rticle AGNs,seekingamodelthatcanreconcileboththehard-andsoft-bandsamples.Wefindthat /45 1 theabsorbedAGNXLFhasalowerbreakluminosity,ahighernormalization,andasteeper /2 faint-endslopethantheunabsorbedAGNXLFouttoz∼2.Hence,absorbedAGNsdominate /18 9 2 at low luminosities, with the absorbed fraction falling rapidly as luminosity increases. Both /1 7 XLFsundergostrongluminosityevolutionwhichshiftsthetransitionintheabsorbedfraction 4 7 7 to higher luminosities at higher redshifts. The evolution in the shape of the total XLF is 8 4 primarilydrivenbythechangingmixofunabsorbedandabsorbedpopulations. b y g u Key words: galaxies: active–galaxies: evolution–galaxies: luminosity function, mass e s function–X-rays:galaxies. t o n 0 9 M a rc h 2 0 ManyAGNsaresurroundedbygasanddustthatcanobscuretheir 2 1 INTRODUCTION 3 emissionatcertainwavelengths.Thus,itisvitaltounderstandAGN Theluminosityfunctionofactivegalacticnuclei(AGNs)represents obscurationinordertoobtainaccuratemeasurementsofthelumi- oneofthecrucialobservationalconstraintsonthegrowthofsuper- nosityfunction.QuantifyingAGNobscurationalsorevealswhether massiveblackholes(SMBHs)overthehistoryoftheUniverse.The SMBHsundergosignificantperiodsofobscuredgrowth,whenthis shapeoftheluminosityfunctionreflectsacombinationoftheunder- takesplacewithinthelifetimesofAGNs,andhowitrelatestothe lyingdistributionoftheSMBHmassesandthedistributionoftheir triggeringandfuellingprocesses. accretionratesorEddingtonratios(e.g.Airdetal.2013a;Shankar, Obtainingaccuratemeasurementsoftheluminosityfunctionand Weinberg&Miralda-Escude´2013;Schulzeetal.2015).Thus,ac- revealingtheextentofobscurationrequireslarge,unbiasedsamples curatemeasurementsoftheshapeandevolutionoftheluminosity ofAGNsselectedoverthewidestpossiblerangeofredshiftsandlu- function provide crucial insights into the physical processes that minosities.Opticalsurveys,combinedwithfollow-upspectroscopy, driveSMBHgrowthovercosmictime. canefficientlycoverwideareas(e.g.SDSS:Yorketal.2000)but arebiasedtowardsthemostluminous,unobscuredsources.Alterna- tively,AGNscanbeidentifiedinthemid-infrared(mid-IR),which (cid:2)E-mail:[email protected] probesthereprocessedemissionfromthedusty,obscuringmaterial. (cid:3)C 2015TheAuthors PublishedbyOxfordUniversityPressonbehalfoftheRoyalAstronomicalSociety TheXLFsofunabsorbedandabsorbedAGNs 1893 Mid-IRselectionshouldnotbebiasedagainstobscuredsources,but bate (see Akylas et al. 2006; Dwelly & Page 2006), potentially thecontributionofthehostgalaxyisoftensignificantatthesewave- due to difficulties regarding the selection functions for absorbed lengths,whichlimitsmid-IRselectiontoluminoussourceswhere and unabsorbed sources. Recent work by Ueda et al. (2014) re- the galaxy light is overwhelmed by the AGN (e.g. Donley et al. examinedtheevolutionoftheXLFandthedistributionofN (the H 2008;Mendezetal.2013). ‘N -function’) using a large compilation of both soft and hard H X-raysurveyscanefficientlyidentifyAGNsoverawideluminos- X-ray surveys and found that both a luminosity and redshift de- ityrange,includinglow-luminositysourceswherethehostgalaxy pendenceoftheabsorbedfractionwererequired.Theyalsofound dominates at optical or infrared wavelengths (e.g. Barger et al. that an LDDE parametrization was needed to describe the evo- 2003). Nevertheless, soft X-ray emission (at energies (cid:2) 2 keV) lution of the XLF (with some further modifications to describe will be absorbed by the same gas and dust that obscures the the evolution at z (cid:3) 3, see also Civano et al. 2011; Hiroi et al. AGN at optical and UV wavelengths. Thus, soft X-ray samples 2012). aregenerallydominatedbyunobscuredAGNs.Absorptionbiases Recently,thecombinationofextremelydeepX-raysurveydata are reduced at hard X-ray energies (∼2–10 keV), except in the and new approaches to X-ray spectral analysis have enabled im- mostheavily-obscured,Compton-thickAGNs(equivalentline-of- proved measurements of N at z ∼ 0.5–2 and have been used to H D sighthydrogencolumndensitiesNH(cid:3)1024cm−2).However,even identify sizable samples of Compton-thick AGNs (e.g. Comastri ow Compton-thick sources may still be identified at soft or hard X- etal.2011;Georgantopoulosetal.2013;Brightmanetal.2014). nlo a ray energies due to scattered emission, including the Compton- Buildingonthiswork,Buchneretal.(2015)usedaflexible,non- d e scattered emission (‘reflection’) from the obscuring material parametric method to estimate the space densities of AGNs as a d itself. functionofredshift,luminosity,andN ,effectivelymeasuringthe fro H m AlargenumberofdeepandwideX-raysurveyshavebeencar- XLFfordifferentcolumndensities.Thisworkalsorecoveredalu- h ried out, taking advantage of the efficiency and power of X-ray minosity and redshift dependence in the evolution of the fraction ttp s selection (see a recent review by Brandt & Alexander 2015). A ofabsorbedAGNs(althoughtheCompton-thickfractionwascon- ://a number of studies have measured the X-ray luminosity function sistentwithaconstantvalueof∼35percent).However,adetailed ca d (XLF) of AGNs out to high redshifts using these samples (e.g. comparisonofparametricmodelsfortheevolutionoftheXLFof e m Ueda et al. 2003; Barger et al. 2005; Miyaji et al. 2015). These AGNswasnotundertaken. ic .o studiesfindthatAGNsareastronglyevolvingpopulation,witha Inthispaperweaddresssomeremainingissuesinstudiesofthe u p sharpdecreaseintheirnumberdensitybetweenz∼1–2andtoday. evolution of the XLF of AGNs: the shape of the XLF and how .c o BrightX-ray-selectedAGNsarefoundtopeakinnumberdensity itevolves withredshift,theextent ofanyluminosityand redshift m at z ≈ 2, similar to optically selected QSOs. Fainter AGNs peak dependenceoftheabsorbedfraction,andtheconnectionbetween /m n laterinthehistoryoftheUniverse(z≈1)butwithamuchmilder theabsorptionpropertiesandtheevolutionoftheAGNpopulation. ra s declinetothepresentday(e.g.Hasinger,Miyaji&Schmidt2005). WecombinesamplesselectedatbothhardandsoftX-rayenergies /a These patterns have led several authors to propose a luminosity- anddeterminetheunderlyingXLFanddistributionofN thatad- rtic H le dependent density evolution (LDDE) parametrization to describe equatelydescribestheobservedfluxesinbothsamples(similarto /4 5 theevolutionoftheXLFofAGNs(e.g.Miyaji,Hasinger&Schmidt the approach of Ueda et al. 2014, cf. the X-ray spectral analysis 1 /2 2000;Uedaetal.2003).InthismodeltheXLFismodifiedbydif- usedinBuchneretal.2015). /1 8 fering degrees of density evolution that vary with luminosity and InSection2wedescribeourdatasetsthatweusetodefinelarge 9 2 redshift. This results in an XLF that changes shape over cosmic samplesofX-raysourcesselectedinthehard(2–7keV)andsoft /1 7 time. (0.5–2 keV) energy bands. We also compile deep optical, near- 47 7 InAirdetal.(2010,hereafterA10)wechallengedthisevolution- IR, and mid-IR imaging across our fields that we use to robustly 8 4 arymodelwithadetailedstudyofthehard-bandXLFthatcarefully identify counterparts to our X-ray sources and calculate photo- b y accountedfornumerousuncertaintiesandbiasesthatweregenerally metricredshifts.InSection3,wedescribeourBayesianstatistical g u notincludedinpriormeasurements.Theseincludedfluxmeasure- technique that allows us to incorporate a range of X-ray spectral es menterrors,Eddingtonbias,incompletenessofopticalidentifica- shapesandaccountfortheeffectsofabsorption.Wealsointroduce t o n tions,andtheuncertaintyinphotometricredshiftestimates.Athigh an approach to account for the contribution from normal, X-ray- 0 9 redshifts(z∼2–3)weadoptedarest-frameUVcolourpre-selection detectedgalaxiesonourmeasurements.Wethenpresentmeasure- M a technique(Airdetal.2008).Byperformingarobustmodelcom- mentsoftheXLFbasedonourhardandsoftsamplesindividually rc h parisonbasedonBayesianstatisticaltechniques,wefoundthatthe (Section4),introducinganewflexibleparametrizationoftheXLF. 2 0 evolution of the XLF could be described by a simpler model in Weshowthatsignificantdiscrepanciesbetweenthemeasurements 2 3 whichtheXLFretainsthesameshapeatallredshiftsbutevolvesin atallredshiftswarrantthefurtherconsiderationofabsorptionef- bothluminosityanddensity(seealsoAssefetal.2011;Rossetal. fects. In Section 5 we separately model the XLF of unabsorbed 2013). andabsorbedAGNs(includingacontributionfromCompton-thick While A10 presented a number of important advances, ab- sources)andshowhowthecombinationofthesepopulationscan sorption effects were not explored. Other studies have attempted simultaneouslyaccountforbothourhard-andsoft-bandsamples. to measure the distribution of absorption column densities and OurresultsplaceconstraintsonthetotalXLFofAGNsandtheab- presentabsorption-correctedmeasurementsoftheXLF.Uedaetal. sorbedfractionasafunctionofluminosityandredshift.InSection6 (2003) found that the fraction of absorbed AGNs (those with wecompareourresultstopriorworkanddiscussthewiderimpli- N >1022 cm−2)wasstronglydependentonluminosity,decreas- cationsofourfindings.Section7summarizesourpaperandoverall H ingathigherluminosities.Laterstudiesfoundthatthefractionof conclusions. absorbed AGNs depends on both luminosity and redshift, drop- Giventhelengthofthispaper,acasualreadermaywishtoskip ping at high luminosities but increasing (at a given luminosity) toSection4.4,Section5,andthediscussioninSection6(andfocus to higher redshifts (e.g. La Franca et al. 2005; Hasinger 2008). onFigs7–9).Weadoptaflatcosmologywith(cid:3)(cid:4)=0.7andh= The extent of any redshift evolution has been a matter of de- 0.7throughoutthispaper. MNRAS451,1892–1927(2015) 1894 J.Airdetal. Table1. DetailsoftheX-raysurveysusedinthispaper. Field Survey RA Dec. X-ray Survey Softband Hardband exposure area NX Nctrprt Nspec−z NX Nctrprt Nspec−z (J2000) (J2000) (deg2) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) CDFS CDFS-4Ms 03:32:27.2 −27:47:55 4Ms 0.075 413 397(96.1%) 240(58%) 283 273(97%) 155(55%) CDFS E-CDFS 03:32:27.2 −27:47:55 250ks 0.181 334 328(98.2%) 115(34%) 273 268(98%) 101(37%) CDFN CDFN 12:36:49.3 +62:13:19 2Ms 0.112 384 363(94.5%) 242(63%) 286 273(96%) 176(62%) EGS AEGIS-XD 14:19:20.8 +52:50:03 800ks 0.260 698 673(96.4%) 295(42%) 552 539(98%) 233(42%) EGS AEGIS-XW 14:17:15.0 +52:25:31 200ks 0.204 334 332(99.4%) 130(39%) 274 274(100%) 110(40%) COSMOS C-COSMOS 10:00:20.3 +02:11:20 160ks 0.984 1213 1195(98.5%) 694(57%) 889 877(99%) 530(60%) Bootes XBootes 14:31:28.3 +34:28:07 5ks 7.124 754 744(98.7%) 566(75%) 257 255(99%) 196(76%) – ALSS 13:14:00 +31:30:00 – 5.800 – – – 34 34(100%) 33(97%) D – AMSS – – – 81.77 – – – 109 109(100%) 107(98%) o – ROSAT – – – 20391 221 221(100%) 221(100%) – – – wn lo Total 4351 4253 2503 2957 2902 1641 ad e d Nexoptoess.urCeotliummen;s(:6)(1t)otnaalmareeaofcothveerfieedldb;y(b2o)tnhatmheeXof-rXay-rdayatasuarnvdeythperoogprtaicmalmoerwiniftrhairnedthiimsfiageilndg;,(3ex,4c)luadpipnrgoxairmeaasteclcoesnetrteoobfrigthhetsstuarrvse;y(;7)(5n)unmobmeirnoaflXX--rraayy from sourcesdetectedinthesoft(0.5–2keV)energyband;(8)number(andfraction)ofthesoftX-raysourcesthatareassociatedwitharobustmultiwavelength h counterpart;(9)number(andfraction)ofthesoftX-raysourcesthathaveaspectroscopicredshift;(10)numberofX-raysourcesdetectedinthehard(2–7keV) ttp s energyband;(11)number(andfraction)ofthehardX-raysourcesthatareassociatedwitharobustmultiwavelengthcounterpart;(12)number(andfraction) ://a ofthehardX-raysourcesthathaveaspectroscopicredshift. c a d e m 2 DATA In each of the remaining fields we adopt data from a single ic Chandrasurvey:the2MssurveyinCDFN(Alexanderetal.2003); .ou To constrain the evolution of the XLF we require large samples the∼160-ksC-COSMOSobservations(Elvisetal.2009;Puccetti p.c of X-ray-selected AGNs. By selecting samples in both the hard o etal.2009);andthe5-ksXBootessurvey(Murrayetal.2005). m (>2keV)andsoft(0.5–2keV)observedenergybands,wecanalso TheX-raydatafromallofoursurveyswerereducedwithourown /m constrainthedistributionofNHandcorrectfortheseeffectsonthe pipelineprocedure,whichisdescribedindetailbyLairdetal.(2009) nra XLF.InthispaperwecombinealargenumberofChandraX-ray andNandraetal.(2015).Weperformedpointsourcedetectionusing s/a surveysalongwithlargerareasurveysfromASCAandROSAT.We theproceduredescribedbyLairdetal.(2009)andappliedafalse rtic givefurtherdetailsofourdatasetsbelow.Table1summarizesthe le Poissonprobabilitythresholdof<4×10−6togeneratecatalogues /4 differentsurveysandprovidesthenumberofhardandsoftX-ray- 5 of detected sources in the soft (0.5–2 keV), hard (2–7 keV), full 1 selectedsourcesfromeach. /2 (0.5–7 keV) or ultrahard (4–7 keV) observed energy bands. We /1 8 combinedthesourcelistsineachbandtocreateamergedcatalogue, 9 2 whichisusedinthecounterpartidentificationproceduredescribed /1 2.1 ChandraX-raydata inSection2.2below. 747 7 InthispaperweuseChandraX-rayobservationsfromfivedistinct Forthe5-ksXBootessurveyweappliedastricterfalseprobabil- 8 4 partsofthesky:theChandraDeepField-South(CDFS),Chandra ity cut (<10−8) in addition to a requirement of ≥5 total detected b y Deep Field-North (CDFN), Extended Groth Strip (EGS), COS- countsineachband.Thiscutreducesthesamplesizebutapplies g u MOS,andBootesfields. aneffectiveX-rayfluxlimitthathelpsraisethecompletenessofthe es In the CDFS field we identify two different ‘surveys’: (1) the spectroscopicfollow-upinthisfield.Inthisfield,wealsorestrict t o n seriesofobservationsthathavetargetedthecentral∼0.07deg2 of ouranalysistothe∼7.1deg2 oftheBootesfieldthatcorresponds 0 9 thefieldandreachatotalcombinedexposuretimeof∼4Ms(the to the 15 standard sub-fields of the AGES spectroscopic survey M a CDFS-4Mssurvey;Xueetal.2011);and(2)theseriesoffour250ks (Kochaneketal.2012).Thiscutalsoensureswehaveahighspec- rc h observationsthatsurroundthecentralarea(theExtended-CDFS,or troscopiccompleteness(seeSection2.5below). 2 0 E-CDFS survey; Lehmer et al. 2005). The X-ray data reduction WedeterminedX-raysensitivitymapsandareacurvesforeach 2 3 and source detection procedures were carried out independently bandasdescribedinGeorgakakisetal.(2008),accountingforthe foreachofthesesurveys.Tocombinethetwosurveys,wedefine stricterfalseprobabilitycutandminimumcountsrequirementfor a central region where the CDFS-4Ms survey takes precedence, theXBootessurvey.Weconvertthesensitivitymapstoareacurves roughlycorrespondingtotheareawithin∼9arcminofthecentreof as a function of flux by assuming a fixed X-ray spectral slope of thefield.Outsidethiscentralregionweadoptthesourcesdetected (cid:5)=1.4.Wenotethatthefluxcalculatedwiththisfixedconversion intheE-CDFSsurveyonly.Thisprocedureensureswehaveawell- factorscalesdirectlywiththecountrate;inSection3belowwede- definedsamplewherewecanaccuratelydeterminethesensitivity. scribeourproceduretoconvertbetweenthecountrateandintrinsic In the EGS field we also identify two distinct surveys: (1) the quantities(suchasluminosity),whichallowsforamorecomplex series of eight pointings that reach a nominal depth of ∼200 ks X-ray spectrum and accounts for uncertainties in this conversion and were presented in Laird et al. (2009), which we refer to as factor.Thesensitivitymapcalculationislimitedtothefootprintof the AEGIS-XW(ide) survey; and (2) the AEGIS-XD(eep) survey the multiwavelength photometry for each field. Fig. 1 shows the (Nandra et al. 2015) which took three of the original 200 ks to a correspondingareacurvesforeachofourX-rayfields. depth of ∼800 ks. We adopt the deeper AEGIS-XD observations WenotethatpreviouslypublishedX-raysourcecatalogues,often whenavailable. includingmutiwavelengthcounterpartinformation,areavailablefor MNRAS451,1892–1927(2015) TheXLFsofunabsorbedandabsorbedAGNs 1895 stricter(i.e.lower)falseprobabilitythreshold.Thus,ourcatalogues aremoreconservative. 2.2 Multiwavelengthcounterpartsandphotometry IneachofourChandrafields,weidentifymultiwavelengthcoun- terpartstoourX-raysourcesusingthelikelihoodratio(LR)method (e.g. Ciliegi et al. 2003; Brusa et al. 2007; Civano et al. 2012), matchingtomultipleoptical,near-IR,andmid-IRbandstoensure a high completeness and reliability. We also compile multiwave- lengthphotometryfromalargernumberofbands,whichweuseto calculatephotometricredshifts(seeSection2.6below). D o w 2.2.1 CDFS,CDFN,andEGS n lo a d In three of our fields – the CDFS, CDFN, and EGS – we iden- e d tify counterparts and extract the multiwavelength photometry us- fro ing a custom version of the Rainbow Cosmological Surveys m . Database1(Pe´rez-Gonza´lezetal.2005,2008;Barroetal.2011a,b), http whichprovidesacompilationofthevariousphotometricdatasets. s AppendixAlistsallthedifferentphotometricimagingdatasetsfor ://a c eachfieldthatareusedinthispaper.Wenotethatfullcoverageof ad e the entire field is not always available in each photometric band. m ic All the images are registered to a common astrometric reference .o u frameandphotometryisperformedinconsistentaperturestopro- p .c duce spectral energy distributions (SEDs) that span from the UV o m to mid-IR. The X-ray source matching procedure is described in /m detailinNandraetal.(2015),butwebrieflysummarizethemethod n ra here. s /a First,wererantheRainbowphotometriccode,extractingallpo- rtic tentialcounterpartswithin3.5arcsecoftheX-raypositionsinanyof le /4 thebandscoveredbyRainbow(usinginitialSEXTRACTORcatalogues 5 1 ineachofthebands).Thecounterpartswerethencross-matchedus- /2 /1 inga2-arcsecsearchradiustocreateasinglemultibandcatalogue. 8 9 We obtained consistent photometry by applying a single aperture 2 /1 acrossallopticalandnear-IRbands.WealsoextractedIRACpho- 7 4 tometry,applyingtheproceduredescribedinPe´rez-Gonza´lezetal. 77 8 (2008) and Barroet al. (2011a) to deblend the IRAC photometry 4 b whenasingleIRACsourceisassociatedwithmultipleoptical/near- y g IRcounterparts. u e Figure1. X-rayareacurves(sensitiveareaversusX-rayflux)foreachof Next, we calculated the LR for all candidate counterparts de- st o oursurveysinthesoftandhardbands.Theblacklineindicatesthetotalarea tectedintheIRAC3.6μmbandanddeterminedanLRthreshold n 0 curveforourstudy.Welimitoursensitivityanalysistotheareafallingwithin that maximizes the sum of the completeness and reliability (see 9 thefootprintofourmultiwavelengthphotometry,excludingareascloseto Luo et al. 2010). A candidate counterpart that exceeded this LR Ma brightstars.TheROSATareacurveincludestheROSATbrightsurveyand thresholdwasdeemedasecurecounterpart(takingthecounterpart rch SA–N survey (see Section 2.3). The EGS area curve includes both the withthehighestLRvalueincasesof>1securecandidate).Wethen 20 AEGIS-XD800ksdataandtheadditionalareaat200ksfromAEGIS-XW. 2 repeatedtheentireLRmatchingprocessforthebandsindicatedin 3 TheCDFSareacurvecombinestheCDFS4Msdatawiththeflanking250ks thetableinAppendixA,retaininganyadditionalsecurecounter- datafromtheE-CDFS.Theseareacurvesassumeasingle,fixedconversion factorbetweentheX-raycountrateandflux,correspondingtoanX-ray partsidentifiedinthesebands.Thisprocedureenablesustoidentify spectrumwithaphotonindex(cid:5) =1.4andGalacticabsorptiononly;the securecounterpartsforahighfraction(>90percent)oftheX-ray effectofdifferentX-rayspectralshapes–andtheresultinguncertainties sources in the CDFS, CDFN, and EGS fields (see Table 1). The intheconversionfactors–areaccountedforinourBayesianmethodology vastmajority(∼92percent)ofthesecurecounterpartswereiden- describedinSection3. tifiedinthe(deblended)IRAC3.6μmcatalogue.Matchingtothe additionalbandsallowsustoidentifycounterpartswhentheIRAC theallofourChandrafields(e.g.Alexanderetal.2003;Brandetal. candidate is faint, blended, or non-existent. No additional cross- 2006; Goulding et al. 2012). Adopting our own X-ray reductions matchingisrequiredasthefullmultibandphotometryisprovided andsourcedetectionproceduresensureswecanaccuratelydeter- throughmatchedaperturesforallsourcesintheRainbowdatabase. minethesensitivityinaconsistentmanner,whichisessentialfor our Bayesian analysis of the XLF. Our catalogues contain ∼10– 25percentfewersourcesthanXueetal.(2011)andPuccettietal. (2009)intheCDFS-4MsandC-COSMOSareas,mainlyduetoour 1https://rainbowx.fis.ucm.es MNRAS451,1892–1927(2015) 1896 J.Airdetal. 2.2.2 COSMOS 2.3 Large-areasurveys IntheCOSMOSfield,whichisnotcurrentlyincludedintheRain- CombiningourfiveChandrafieldsprovidesasampleofover4000 bow surveys data base, we matched directly between our X-ray soft-band detections and over 2800 hard-band detections from a source lists and two multiwavelength catalogues: (1) the COS- totalareaof∼9deg2.However,toaccuratelyconstrainthebright MOSIntermediateandBroadBandPhotometryCatalogue2008,2 endoftheXLFrequiressamplesofhigherluminosityX-raysources which is based on detection in the deep Subaru i+ imaging of identifiedfromlarger-areasurveys.Wethussupplementoursample the entire COSMOS field (Capak et al. 2007); and (2) the S- withsourcesfromlarge-areasurveyscarriedoutwithASCA(inthe COSMOS IRAC 3.6 μm based catalogue (Sanders et al. 2007). hardband)andROSAT(inthesoftband). Unlike the Rainbow catalogues, the S-COSMOS IRAC 3.6 μm Forourlarge-areahard-bandsample,weincludethe34sources cataloguehasnotbeendeblended.Thus,wefirstidentifiedsecure fromtheASCALargeSkySurvey(ALSS:Uedaetal.1999),which matches (using the LR method) from the higher-resolution Sub- coversacontiguousareaof5.8deg2 nearthenorthGalacticpole. aru i+ catalogues. We found secure counterparts for 1348 of the We adopt the optical identifications from Akiyama et al. (2000): 1621X-raysources(83percent)inouroverallC-COSMOScata- 2 sources are optically identified as galaxy clusters, 1 is a star, 1 logue.Next,weappliedtheLRmethodtomatchbetweentheX- source remains unidentified, and the remaining 30 are associated Do w rayandS-COSMOScatalogues.Weidentifiedsecurecounterparts withAGNs,allofwhichhavespectroscopicredshifts.Wealsose- n for an additional 178 sources, taking our overall completeness to lect sources from the ASCA Medium Sensitivity Survey (AMSS; loa d 94percent. Ueda et al. 2001), which combines data from a large number of e d Finally,wecross-matchedbetweenoursecurepositionsandthe ASCAobservationsathighGalacticlatitudesoveranarea∼82deg2. fro originalcatalogues,againusingtheLRmethodtoensureonlysecure We includesourcesfromtheAMSSnsub-sample,selectedinthe m h associationsareconsidered.For1499ofthe1526securecounter- hard (2–10 keV) band, with optical identifications presented by ttp pFaorrtsth,weseeesnodurucpewswitheboobtthaianSpuhboatroumi+etaryndinSu-CpOtoS1M8ObSrocaodu-nbtaenrdpafirlt-. Ahaksiy1a0m0paeertcaeln.t(s2p0e0c3t)r.osTchoepiscamcopmleplientcelnuedsess.W87eXin-craluydseoaudrdceitsioannadl s://ac a tersspanningfromtheUVtonear-IR,aswellas15intermediate-or sources from the AMSSs sub-sample (Ueda & Akiyama, private d e narrow-bandopticalfilters,fromtheSubarui+catalogue(seeIlbert communication),whichincludes20AGN;twosourcesinthissam- m ic etal.2009;McCrackenetal.2010).Thephotometrywasextracted pleremainunidentified.WeadoptareacurvesfromtheALSSand .o u in3-arcsecdiameteraperturesfromPSF-matchedimages.Wealso AMSSfromAkiyamaetal.(2003)andUedaetal.(2003),respec- p adoptIRACphotometryinthe3.6,4.5,5.8,and8.0μmbandsfrom tively. .co m theS-COSMOScatalogue,applyingtheaperturecorrectionsgiven Foroursoft-bandsamplewecombinesamplesfromtheROSAT /m inIlbertetal.(2009). BrightSurvey(RBS;Fischeretal.1998;Schwopeetal.2000)and nra In 27 cases we have an i+ counterpart but do not find an S- theSelected-Area–Northsurvey(SA–N:Appenzelleretal.1998), s/a COSMOScounterpart,eitherduetothelimiteddepthoftheIRAC removing duplicate sources. We include sources with significant rtic imaging or because the i+ source is blended at the IRAC resolu- detections in the 0.5–2 keV band and adopt the unabsorbed flux le /4 tion. For these sources we simply ignore the IRAC bands in our estimates(correctedforGalacticabsorption).Wecutoursampleat 5 1 photometricredshiftestimates(seeSection2.6below). fluxlimitsoff >3.6×10−12ergs−1cm−2fortheRBSsample /2 In23cases,weidentifyanS-COSMOScounterpart,butnoSub- andf >0.51-2×keV10−12ergs−1cm−2fortheSA–Nsample.These /18 0.5-2keV 9 arui+source.Forthesesourcesweextractedphotometryin3-arcsec highfluxlimitsensureoursourcesallliewellabovethesensitivity 2/1 diameteraperturesattheS-COSMOSpositioninthePSF-matched limitsofthesurveys,allowingustoadoptsimplesensitivitycurves 74 7 imagesforalltheUVtonear-IRbands,generallyobtainingonly thatcorrespondtotheentireareaofeachsurveyandcutoffsharply 7 8 upperlimitsfortheseopticallyfaintsources. ateachofthefluxlimits(seeFig.1).Allthesourcesabovethese 4 b fluxlimitshavespectroscopicclassificationsandweidentifyatotal y g of221AGNswithspectroscopicredshifts(excludingBLLactype u e s 2.2.3 Bootes objects). t o FortheBootesfield,wecompiledmultiwavelengthcataloguesfrom Wenotethat,incontrasttoourChandrafields,ourareacurves n 0 forourlarge-areasurveysdonotaccountforthePoissonnatureof 9 the NOAO Deep Wide-Field Survey (NDWFS; Jannuzi & Dey M the detection. As we restrict the samples from the large-area sur- a 1999) DR3, Sloan Digital Sky Survey (SDSS) DR9 (Ahn et al. veys to highly significant detections, this simplification will have rch 2012),GALEXGR7,3FLAMINGOSExtragalacticSurvey(Elston 2 aminimaleffectonourXLFmeasurements.However,differences 0 etal.2006),andtheSpitzerDeep-WideFieldSurvey(Ashbyetal. 2 in the assumed spectral shape can have a significant impact on 3 2009). We use the LR method to match our X-ray catalogues to theestimateofafluxandthustheassumedsensitivity.Uncertain- theappropriateselectionbandforeachofthesesurveys,assigning ties in the spectral shape and the resulting differences in sensi- securematchesfromthesurveysintheorderofpriorityindicated tivity are accounted for by our Bayesian methodology described inthetableinAppendixA.Thevastmajorityofoursecurematches below. areidentifiedintheNDWFSIband.Weidentifysecurecounterparts for 95.8percent of our X-ray sources. Finally, we cross-matched between the original catalogues and our secure counterpart posi- 2.4 Identificationandmaskingofstars tions,againapplyingtheLRmethod,toidentifycommonsources. The combined surveys provide photometry in up to 17 different Bright stars can contaminate our photometry, leading to issues bands(seeAppendixA). withcounterpart identificationand photometricredshiftestimates intheseregions.Wehavethereforemaskedoutareasclosetobright stars from all of our Chandra fields in a consistent manner. We 2http://irsa.ipac.caltech.edu/data/COSMOS/datasets.html searchedforstarsbrighterthanV=15intheHSTGuideStarCat- 3http://galex.stsci.edu/GR6/ alog2.3(Laskeretal.2008).Wemaskedallareaswithinaradius, MNRAS451,1892–1927(2015) TheXLFsofunabsorbedandabsorbedAGNs 1897 r,givenby alsosearchedforspectroscopicredshiftsfromtheSDSS(Yorketal. 2000). r =(16−V)×6arcsec, (1) Wematchedthespectroscopiccataloguestothesecurecounter- whereVistheV-bandmagnitudefromtheGuideStarCatalog.We partsofourX-raysourcesusinga2-arcsecsearchradius,corrected set a maximum masking radius of 40 arcsec. We have removed foranyoverallastrometricoffset,andrepeatedthematchingwith anyX-raysourceswithinthisradiusfromoursamples(whichcan a1arcsecradius.Weonlyadoptthosespectroscopicredshiftsthat includethestaritselforanearbysource).Wealsoexcludedthese are flagged as high-quality, reliable redshifts in the original cata- regionswhencalculatingtheX-raysensitivityandareacurves. logues.Table1givesthenumberofX-raysourcesinourhard-and We have also identified stars with fainter magnitudes that are soft-band-selectedsampleswithreliablespectroscopicredshifts. detectedatX-raywavelengthsandremovedthemfromoursamples. In the fields with Rainbow coverage (CDFS, CDFN, EGS) stars 2.6 Photometricredshifts were identified by a range of colour and morphology criteria, as described in Barro et al. (2011a). In the COSMOS and Bootes The levels of spectroscopic completeness vary over our Chandra fields we applied a single colour criterion based on the region of surveysfrom∼35percent(inourE-CDFSarea)to∼75percent(in D o colour–colourspaceoccupiedbystarsinIlbertetal.(2009), theBootesfield).FortheremainingX-raysourceswemustresort w n tophotometricredshiftestimates,whicharedeterminedbyfittinga lo R−[3.6]<3.0×(R−I)−1.2, (2) a setoftemplatespectratotheobservedSEDsofoursources.Such d e wthheemreag[3n.i6tu]dies itnhethmeaSgunbitaurduerin+thoerNIRDAWCF3S.6Rμfimlteirm,aangdinIg,isRthies rfaeidnsthXift-sracyasnoubrecehsi.gAhlykeuynacdevratanitna,gepaorftiocuurlaBrlayyewsihaennacnoanlyssidiser(isnege d from malsaognreitquudiereidntthhaetSthuebaXru-raiy+soourrNceDsWexFhSibIitfialtleorw.FXor-raalyl-fitoe-lodpst,iwcael Sphecottioomne3tr)icisrethdashtiwftsebayreaadbolpetitnogapcrcoobuanbtilfitoyrdthisetruibnucteiortnasinftoiersthine https fluxratio,logf /f <−1,wheretheratioiscalculatedas redshift,p(z),ratherthanasingleredshiftestimate.Wethusrequire ://a X opt thatourphoto-zapproachrecoversap(z)distributionthataccurately ca logffoXpt =logf0.5−2keV+5.4+ 2I.5. (3) reflWecetscathlceuulantceeprthaointotimesetirnicourerdrsehdisfhtsifutseisntgimthateesE.AZYphoto-zcode demic.o ThiscutensuresthatwedonotexcludebrightQSOsfromoursam- (Brammer, van Dokkum & Coppi 2008). We use EAZY in two- up template mode, allowing for combinations of a galaxy and AGN .c plethatmaysatisfytheotherstellarcriteria.Whenaspectroscopic o template. For the galaxy templates we adopt the ‘pegase13’ tem- m classificationisavailable(seeSection2.5)thisoverridesourpho- /m tioonumroeuptrhroifictvocemlaCesthsriaificncdparrtaoiocfinee.dluOdrsfe,.t4h0e(5800sppeercctreonstc)owpeicrealalylscolaidsseinfiteifidesdtabrys psddyleuansstcttehyrie’sbtteeiectdmpgbpraylolavatGxiedsyreaadztreieamwniniptcelhaltutEeadAsle.ZddY(r2.ba0yTw0ahn6pi)sp.frltAyoeimmdndgpitlthtaihoeteenraCaslane‘ltgzsetecatortoni-ffseoitpsraamtsrl.aino(m2gfe02ta0e5n0r9ds) nras/article reddeninglawforarangeofdifferentextinctionstoasub-setofthe /4 5 2.5 Spectroscopicredshifts galaxytemplates.WeadoptsevenAGNtemplatesfromtheSalvato 1/2 et al. (2009) template set,4 namely the Sey 1.8, Sey 2, Mrk231, /1 8 AllofourChandrasurveyshavebeenthesubjectofintensespectro- pl_TQSO1, pl_QSOH, pl_QSO, and the S0-10_QSO2-90 hybrid 92 scopiccampaigns.Thesecampaignsincludelarge-scalefollow-up template. We also include the Type-2 ‘Torus’ template from the /17 ofthegeneralgalaxypopulation,inadditiontothosedirectlytar- Polletta et al. (2007) library,5 which is not included in the final 47 7 getingX-raysources. Salvatoetal.(2009)set.Wenotethatsomeofthesetemplateswill 84 IntheCDF-Swefirstsearchedforspectroscopicredshiftsinthe includehostgalaxycontributions(particularlytheSey1.8,Sey2, by catalogueofXueetal.(2011).Themajorityofthespectroscopic Mrk231, and S0-10_QSO2-90 templates). This is not a major is- gu redshiftsforourCDFS-4Mssurveyaretakenfromthiscatalogue, sueaswewanttodeterminethedistributionofpossibleredshifts, es as well as some of our E-CDFS survey (in the area that overlaps ratherthanperformanaccuratehost-AGNdecomposition.Thetwo- t on withthe4Msdata).Wealsosearchedforspectroscopicredshifts templatemodeinEAZYallowsanypossiblecombinationofoneof 09 fromtheArizonaCDFSEnvironmentSurvey(Cooperetal.2012), ourAGNtemplatesandoneofourgalaxytemplatesandthusallows M a thespectroscopicsub-sampleofsourcesfromtheMUSYCsample foralargeamountofflexibilityinthefittedtemplateSEDs. rc h (Cardamoneetal.2010a),andPRIMUS(Coiletal.2011). AnotheradvantageoftheEAZYphoto-zcodeisthatitallowsfor 20 In the CDFN we used spectroscopic redshifts from DEEP3 theinclusionofa‘templateerrorfunction’.Thisfeatureaccounts 23 (Cooperetal.2011)aswellasthesurveysofTrouilleetal.(2008), foradditionaluncertaintyinthetemplateSEDasafunctionof(rest- Barger, Cowie & Wang (2008), Reddy et al. (2006), Wirth et al. frame)wavelengthandthusallowsforthefactthatourtemplateset (2004),Cowieetal.(2004),andSteideletal.(2003). maynotaccuratelyrepresentthetruediversityofSEDshapes.This IntheEGSwecompiledspectroscopicredshiftsfromanumber uncertaintyinthetruerangeoftemplateSEDsisparticularlyuse- ofsurveysincludingDEEP2,DEEP3,theCanada–FranceRedshift fulasourobservedSEDsextendintotheUVandmid-IR,where Survey, and MMT follow-up of X-ray sources. See Nandra et al. thetemplatesarepoorlycalibrated,especiallyforAGNs.Inaddi- (2015)andreferencesthereinforfulldetails. tion,astheopticalemissionfroman(unobscured)AGNcanvary InCOSMOSweinitiallysearchedforspectroscopicredshiftsof ontime-scalesofmonths-to-years,ourobservedSEDsmaynotbe X-raysourcesintheC-COSMOScatalogueofCivanoetal.(2012). well-matchedbyasingleunderlyingtemplate.Thetemplateerror Wealsosearchforadditionalspectroscopicredshiftsfromthebright functioncanalsoallowforanyoverallcalibrationuncertaintiesin zCOSMOSsurveycatalogue(Lillyetal.2009)andPRIMUS(Coil etal.2011). In Bootes we adopt spectroscopic redshifts from the AGN and 4http://www.mpe.mpg.de/∼mara/PHOTOZ_XCOSMOS/ GalaxyEvolutionSurvey(Kochaneketal.2012).Forallfieldswe 5http://www.iasf-milano.inaf.it/∼polletta/templates/swire_templates.html MNRAS451,1892–1927(2015) 1898 J.Airdetal. D o w n lo a d e d fro m h ttp s ://a c a d e m ic .o u p .c o m /m n ra s aFsigthuereb2es.tCesotmimpaatreisoofnthoefpphhoottoommeettrriiccrreeddsshhiiffttss((zbplhaoct)kacnidrcslepse)c;terorrsocropbiacrsreidnsdhicifattse(tzhsepec9)5fpoerrocuernfitvceenCtrhaalncdornafifideelndcse.Wineteprvloatl.thTehemdeaasnhoefdthlienep(izn)ddiciasttreisbuati1o:n1 /artic relation,whereasthedottedlinescorrespondto(cid:6)z/(1+zzspec)=±0.15(sourceswherethebestestimateofthephoto-zliesoutsidethisrangeareflaggedas le/4 outliers).InthelegendforeachpanelwegivethenumberofX-raysourceswithreliablespectroscopicredshifts(nzspec),theaccuracybasedonthenormalized 51 mcaetadsitarnopahbiscofluaitleurdeesvwiaittihona(rσedNMtrAiaDn)g,lteh.eOfvrearctfioounrooffoouutrlifieersld(sfo(uCtlDierF),Sa,nCdDtFhNe,frEaGctSio,nanodfCcaOtaSsMtroOpSh)icwfeaiolubrteasin(facactaosntrsoipshtiecn);tsaececuteraxctyfoorfdσeNtaMilAsD.W≈e0h.0ig5h,lwigithht /2/189 ∼15percentoutliersand∼5percentcatastrophicfailures.IntheBootesfield–ourlargestarea,shallowestfield–wehavepooreraccuracyandahigheroutlier 2 fraction,reflectingthemorelimitedphotometricimaginginthisfield.Whileourphoto-zhaveapooreraccuracy(σNMAD)andahigheroutlierratethaninsome /174 priorworks,wechoosetoadoptourestimatesastheyhavebeencalculatedinaconsistentmanneracrossallfiveofourChandrafieldsandhaverepresentative 7 7 errorsthatwecanfullytrackviathep(z)inourBayesianmethodology(seetextforfurtherdiscussion). 84 b y thediversesetsofphotometricobservationsusedtoconstructour panelsfortheEGSandCDFSfieldscombinethedeepsurveyareas gu e observedSEDs.Wederiveatemplateerrorfunctiononafield-by- (AEGIS-XD,CDFS-4Ms),wherethebestphotometryisavailable, s fieldbasis–toensurethatitrepresentsthecalibrationuncertainties with the larger-area shallow surveys (AEGIS-XW, E-CDFS). We t on in a given data set – using the basic procedure laid out in Bram- includeallX-raysourceswithahigh-qualityspectroscopicredshift 09 meretal.(2008).First,weattempttofittheobservedSEDswith in these plots and do not apply any cuts based on the estimated M a ourtemplates,fixingtheredshiftatthespectroscopicvalue,where quality of the photo-z. Thus, we include sources with extremely rc h available.Wethencalculate broadp(z)distributions,whichareoftenflaggedasunreliableand 2 0 (cid:6)fj = FjF−Tj, (4) excInluedaecdhwpahneenlaosfsFeisgs.in2gwtheeprseuscecnetsasnoufmphboertoo-fzstuemchmnaiqryuessta.tistics: 23 j whereFjindicatestheobservedfluxinafilterforsourcej,andTjis (i) σNMAD:theaccuracybasedonthenormalizedmedianabsolute thefluxfromthebest-fittingtemplate.Wecalculate(cid:6)fjasfunction deviation between the best photo-z and the spectroscopic value, of rest-frame wavelength for all sources and filters and take the definedasσ =1.48×median(|z −z |/(1+z )). NMAD phot spec spec medianofevery400individualmeasurementsacrosstherest-frame (ii) f : the fraction of outliers, defined as the fraction of outlier wavelength range. We subtract the median photometric error, in sourceswhere|z −z |/(1+z )>0.15; phot spec spec quadrature,toestimatethecontributionfrom‘templateerror’tothe (iii) f :thecatastrophicoutlierrate,calculatedasthefrac- catastrophic uncertainty as a function of rest-frame wavelength. The value of tionofsourceswherelessthan5percentoftheintegratedp(z)lies thetemplateerroristypicallyaround10percent(influx)butvaries within−0.15<(z−z )/z <+0.15. spec spec between∼4and∼20percentdependingonthewavelengthandthe datainagivenfield. Inoverfourofourfields(CDFS,CDFN,EGS,andCOSMOS)we InFig.2wecompareourphoto-zestimatesforX-raysourcesto obtainaconsistentaccuracyofσ ≈0.06,with∼15percent NMAD securespectroscopicredshiftsacrossourfiveChandrafields.The outliers.Ourapproachensuresweassignanappropriateuncertainty, MNRAS451,1892–1927(2015) TheXLFsofunabsorbedandabsorbedAGNs 1899 etal.2010b;Hsuetal.2014).Manyofthesestudiestakeadditional stepstoimprovethequalityofthephoto-zestimates.Thesestepscan includeoptimizingthetemplateset(e.g.Luoetal.2010),attempting tocorrecttheobservedphotometryforvariability(e.g.Salvatoetal. 2009), or applying priors based on the source morphology and X-ray flux (e.g. Salvato et al. 2011). These studies often achieve a higher accuracy and lower outlier rate than our own photo-z analysis.However,theseadditionalstepscanleadtounderestimates ofthetrueuncertaintiesinthephoto-z.Conversely,weretainalarge set of possible templates to ensure we produce p(z) distributions thataccountforthelargeuncertaintiesintheredshiftandtemplate degeneracies. While we have higher outlier rates, our fraction of catastrophicfailuresremains(cid:2)5percent,indicatingthatourp(z) distributionsarerepresentingtheuncertainties.Thenominalerrors D (i.e. the 68percent confidence intervals) on our photo-z are also ow Figure3. RedshiftdistributionofX-raysourcesinourfiveChandrafields comparabletotheresidualsbetweenourbestphoto-zestimateand nlo a withhigh-qualityspectroscopicredshifts(blacksolidline)andthosewhere theavailablespectroscopicredshifts,incontrasttomostprevious d e wtioenaodfopthtep‘hboetsotm’pethroictor-ezdeshstiifmtsa.tTehse(mbleuaendoafshtheedpli(nze)dinisdtircibatuetsiotnh)e,dwishterribeaus- wLuoorketwahle.r2e0e1r0r;oHrssmuaeytabl.e2u0n1d4e)r.eFstuirmthaetermdobrye,awfeacrteoqru∼ire2–th6e(efu.gll. d fro m the orange dotted line shows the distribution obtained by combining the p(z)distribution,whichmostpriorstudiesdonotprovide.Wethus h individualp(z)distributions. choosetouseourownphotometricredshifts:ourpooreraccuracy ttp s andhigheroutlierratesareaccountedforandcompensatedbyour ://a tracedbythep(z),tothebulkofoursourcesandonly∼5percentof Bayesiananalysisthatincorporatesthefullp(z)information. ca d sourcesarethusflaggedascatastrophicfailures.IntheBootesfield e m –ourlargestarea,shallowestfield–wehavepooreraccuracyanda ic muchhigheroutlierfraction,reflectingthemorelimitedphotomet- 3 BAYESIAN METHODOLOGY .ou p ricimaginginthisfield.However,thecatastrophicoutlierfraction .c InthispaperweexpandontheBayesianmethodologydevelopedby o isonly∼1percent,indicatingthatthisadditionaluncertaintyisrep- m resentedbyourp(z)distributions.Wealsonotethatinthisfieldwe A10,incorporatingthedistributionofX-rayabsorptionproperties /m andaccountingfortheeffectsontheinferredshapeandevolutionof n havethehighestspectroscopiccompleteness(∼75percent)andso ra theXLF.Ourmethodalsoaccountsfortheuncertaintyinthemea- s onlyusephotometricredshiftsforarelativelysmallfractionofthe /a sourcesinoureventualanalysisoftheXLF;whenwedoresortto suredX-rayflux(duetophotoncountingstatistics),uncertaintiesin rtic theredshift(forsourceswithphotometricredshiftsornocounter- le aphotometricredshiftweaccountforthelargeuncertaintyinthe /4 parts),uncertaintiesintheX-rayspectralshape,andtheresulting 5 redshift. 1 uncertainty intheX-rayluminosityforanindividualsource.Our /2 InFig.3weplotthedistributionofredshiftestimatesforsources /1 methodologyisdescribedbelow. 8 inourChandrafieldswherewehaveaspectroscopicredshiftand 9 2 thosewhereweadoptthephotometricredshiftinformation.Gener- /1 7 allythesourceswithphotometricredshiftslieathigherredshifts,a 3.1 Probabilitydistributionfunctionforasinglesource 477 consequenceofspectroscopicfollow-upprogrammesbeingbiased 8 4 towardsopticallybrightsources.Wealsoshowtheintegratedcon- Forasinglesourceineitherourhard-orsoft-bandsamplewecan b tpr(izb)utiisonskferwomedthtoewpa(rzd)solfowalelrthreedpshhoifttos-.zTshoiusrcskese.wTihsediunetegtoratthede odneriovuertohbesperrovbeadbdilaittayfdoisrttrhibauttsioonurfcuenacltoionne,fowrhzi,chLXis,agnivdeNnHbybaps(ezd, y gues possibilitythatmanypotentialhigh-redshiftsourcescouldactually LX,NH|Di)whereDi indicatestheobserveddatafromsourceiin t on lieatlowerredshifts,whichisreflectedbytheirp(z)andmustbe oursample.Thisfunctionisnormalizedsuchthat 0 accountedforinmeasurementsoftheXLF. (cid:2) (cid:2) (cid:2) 9 M AsmallfractionofourX-raysources(<2percent)lackamulti- dz dlogLX dlogNHp(z,LX,NH|Di)=1. (5) arc h wavelengthcounterpart,precludingaphotometricredshiftestimate. 2 Weretainthesesourcesinouranalysis,ensuringcompletenessof Wecanrewritetheprobabilitydistributionfunctionas 02 3 oalulrowsaemdpreled,sbhuifttardanogpeta(0p<(z)zt<ha1t0is).cTohnisstarenflteinctlsoogu(r1la+ckz)ofoaveprrioourri p(z,LX,NH|Di)=p(z|di)p(LX,NH|z,Ti,bi), (6) knowledgeoftheredshift;theX-rayfluxinformationisretainedand where d indicates the multiwavelength data used to estimate the i thus a posteriori (after folding the constant p(z) through the final redshiftandT andb correspondtotheX-raydataforthissource: i i XLF)theremaybeapreferredredshiftsolution.Thelackofamul- thetotalobservedX-raycountsinthegivenbandandtheestimated tiwavelengthcounterpartcouldimplythatahigh-redshiftsolution background,respectively. should be given higher a priori preference for such sources, thus Ourknowledgeofzisbasedoneitheraspectroscopicredshift, ourapproachisconservative.Thereisalsoapossibilitythatthese in which case we assume p(z|d) is described by a δ-function at i X-ray sources lack counterparts as they are spurious detections, thespectroscopicvalue,oraphotometricredshift,whenp(z|d)is i correspondingtopositivefluctuationsinthebackgroundcountrate. givenbythep(z)fromourphotometricredshiftfittingdescribedin OuranalysisaccountsforthePoissonnatureoftheX-raydetection Section2.6above.Forthesmallfractionofsourceswherewewere andthusallowsforthispossibility. unabletoidentifyamultiwavelengthcounterpart–andthushaveno OtherestimatesofphotometricredshiftsareavailableforX-ray redshiftinformation–weadoptap(z)distributionwithaconstant sourcesinmanyofourfields(e.g.Bargeretal.2003;Cardamone densityinlog(1+z)over0<z<10. MNRAS451,1892–1927(2015) 1900 J.Airdetal. fixtheinclinationangleto30◦ asarepresentativevalueandallow theintensitytobesetbythenormalization,R,relativetothatex- pectedfromaslabsubtendingasolidangleof2π(whichisallowed tovarybetween0and2,spanningtheextremecasesofnoreflection uptoaneffective4πsolidanglecoverage).Weabsorbthereflec- tioncomponentbythesamecolumn densityseenbytheprimary emission.Thisisagoodassumptionwhenthereflectionarisesfrom the accretion disc and also provides reasonable agreement with the shape and intensity of the reflection component based on so- phisticatedmodelsoftoroidalobscurers(e.g.Brightman&Nandra 2011a).AllcomponentsaresubjectedtoGalacticabsorption,with columndensitiesdeterminedfromDickey&Lockman(1990)via theHEASOFTNHtool.InXSPECterminology,ourmodelisdescribed by D (cid:3) ow wabs∗ (1−constant)∗zwabs∗cabs∗zpowerlw∗zhighect nlo Figure4. OurassumedX-rayspectralmodelforanAGN.Weassumethe (cid:4) a d intrinsicX-rayspectrumisapowerlawwithphotonindex(cid:5)=1.9±0.2 +constant∗zpowerlw+zwabs∗pexrav , (8) ed (dWtopeuenbrsaepiltlsyseoc,laifiantxlteleeo)rd.weTda,hftoeuNrnoHaabbsf=seroarvc5rbtei×edodns,1p(i0efns2ctc2toartucttmmh≈e−i2s2liapnfbeoesrrooctrfehbnisestid)gehobxtfyam(ttbhhpleeuleieinn(tdtrreoeirntdv–seiddncaaissnphhgoeewcddoellrliiunnlmaeew))n.. nwohteerethtahtemcoonrestpanhtysciocrarlelsypomnodtsivtaotetdhemsocdatetlesrecdoufrldacbtieona,dfoscpattte.dWtoe from https Inaddition,weallowforacomponentfromComptonreflectionfromcold, describetheX-rayspectrum,self-consistentlymodellingtheemis- ://a opticallythickmatter(suchasatorusoraccretiondisc)thatleadstothe sion, reflection, and absorption due to the accretion disc or torus ca d characteristic hump at high energies (green dotted line). The black line (e.g.Ross&Fabian2005;Brightman&Nandra2011a).However, e m indicatesthetotalobservedspectrum,whilethegreyregionindicatesthe Buchneretal.(2014)foundthatmoresimplisticmodelssuchasours ic 95percentconfidenceintervalonthespectrum,allowingfortherangeof aregenerallysufficienttoreproducetheobservedspectralshapeof .ou possiblespectralparameters(seeSection3.1,Table2). individual,distantAGNs,especiallyconsideringouranalysisuses p.c o broad-bandfluxesratherthanperformingadetailedX-rayspectral m analysis. /m n TheobservedX-raydatacanbedescribedbyaPoissonprocess. To describe our X-ray spectral model requires three additional ra s bTyhus,thelikelihoodofobservingTi countsfromasourceisgiven tphaerarmeleatteivrse–nothrmeaplhizoatotinoninodfexth,e(cid:5)r,etflheecsticoanttecroemdpforancetniot,nR,fs–cawtt,hainchd /artic le L(Ti |ci,bi)= (ciT+i!bi)e−(ci+bi), (7) wξ(ξ)e)i,ninzo,truLordXBu,caaeynedassiNa‘Hnn,uaiwnsaaenlyccsaein’s.pdFaeortaermramegtienivresent(hcseoeltelxeocpfteisvcpteeelcdytrcdaoeluspingatnraraamtteeed,tebcryis, /451/2/1 whereciistheX-raycountratefromsourceiintheobservedenergy andthuslinktheprobabilitydistributionfunctionforLX,NH,andξ 892 bandandwehaveassumedthattheexpectedbackgroundcountrate, tothePoissonlikelihoodgiveninequation(7)above.Thus, /1 7 tbhi,eiPsowiseslolndelitkeermlihinoeodd.gTihveenuninceerqtauianttiyonin(7c)i.iHsofuwlleyvedre,stcoricboendvebryt p(LX,NH,ξ |z,Ti,bi)∝L(Ti |ci(z,LX,NH,ξ),bi)π(ξ), (9) 4778 4 fromacountrate,c,toanestimateofL andN (givenz),wemust b assumeamodelfoirtheX-rayspectralXshapeHandfoldthismodel whereci(z,LX,NH,ξ)istheexpectedcountrateforasourcewith y g thrFoiugg.h4thsehoawppsroanpreiaxtaeminpslteruomfeonutralXre-rsapyonsspee.ctral model. We as- srepdeschtriaftl zp,arlaummeinteorssitξybLaXs,edabosnoroputironX-croalyumspnecNtrHa,lamndodaedl,diftoioldneadl uest o sume the intrinsic X-ray continuum is described by a power law throughtheappropriateinstrumentalresponseforsourcei. n 0 withphotonindex(cid:5)andahighenergycut-off(wefixthefolding Apriori(cid:5),th(cid:6)espectralparametersξforagivensourcearenotwell 9 M energyat300keV,althoughthishasanegligibleimpactonthelower known;π ξ denotesthepriordistributionthatweadoptforour arc h energies we observe). The observed spectrum is attenuated along spectralparameters, whichdescribes therangeofpossiblevalues 2 0 theline-of-sightbytheinterveningcolumndensity,N .Weinclude andthusencapsulatestheuncertaintyinX-rayspectralshape.We 2 H 3 bothphotoelectricabsorption(usingthewabsmodelinXSPEC)and assumethephotonindex,(cid:5),isdrawnfromaGaussiandistribution Comptonscattering(viacabs),whichsuppressesthecontinuumfur- withameanof1.9andstandarddeviationof0.2(correspondingto therforCompton-thickcolumndensities.Afractionofthispower theobserveddistributionofintrinsicphotonindicesinX-rayspec- law,f ,isallowedtoemergeasanunabsorbedcomponentthatis tralstudiesofnearbyAGNs,e.g.Nandraetal.2007).Weassume scatt thoughttobescatteredintothelineofsightbyionizedgasinthe thescatteredfraction,f ,isdrawnfromalognormaldistribution scatt vicinityoftheAGN.WealsoincludeacontributionduetoComp- withmeanof2percentandscatterof0.8dexbasedontheobserved tonreflectionfromcold,opticallythickmatter,whichgivesriseto distributionofpartialcoveringfactorsofsourcesintheSwift/BAT acharacteristic‘hump’intheX-rayspectrumat∼30keV.Sucha samplefromWinteretal.(2009).Forthereflectionstrength,R,we reflectioncomponentisexpectedduetoreprocessingoftheprimary assumeauniformdistributionintherange0<R<2.Wethusallow X-rayemissionbyasurrounding,dustytorusoranaccretiondisc. foralargeuncertaintyinthestrengthofthereflectioncomponent, Weadoptthepexravmodel(Magdziarz&Zdziarski1995),which which is reasonable to encapsulate uncertainties in the geometry isbasedonMonteCarlosimulationsofComptonreflectionfroma of the accretion disc and/or torus with our simplified modelling cold,opticallythickslabofmaterial.Weassumetheincidentpower of the reflection. Table 2 summarizes this prior information. The lawhasthesameshapeasthedirectlytransmittedcomponent.We colouredlinesinFig.4showeachcomponentevaluatedattheprior MNRAS451,1892–1927(2015) TheXLFsofunabsorbedandabsorbedAGNs 1901 Table2. Priorsonthespectralparametersfor ofN ,butweareabletoplaceconstraintsontherangeofpossi- H anindividualX-rayAGN,ξ. blevaluesofL andhowthisdepends onN .Forthehard-band X H example,wecanconstrainL towithin∼0.25dex,providedtheab- X Parameter Priortype Priorspecification sorptioncolumnisN (cid:2)1023cm−2.Ifthecolumndensityishigher, H thenthesameobservedcountsmustcorrespondtoahighervalue (cid:5) Gaussian 1.9±0.2a logfscatt Lognormal −1.73±0.8b aopfpLaXr.enFtofroarsloowurecrecdoeltuemctneddeinnstihteiesso(fNtba(cid:3)nd1,0a2b2socmrp−ti2o)n.Aeftfeccotlsumarne R Constant 0–2 H densitiesN (cid:3)2×1023cm−2(atz=1.0)theobservedfluxinthe Notes. aBased on observed distribution from H 0.5–2keVbandisdominatedbythescatteredcomponentonly;the Nandraetal.(2007). intrinsicluminosityispoorlyconstrainedbutmustbeafactor∼30 bBased on observed distribution from Winter greaterthanifthesourcehadalowerN . etal.(2009). H Wenotethattheeffectofabsorptionontheobservedcountsinthe mean of the spectral parameters and the black line corresponds hardorsoftenergybandswillvarysignificantlywithredshift.For to the total observed spectrum for these values. The grey region example,atz∼3,wheretheobserved0.5–2keVenergybandprobes D indicatesthe95percentconfidenceintervalonthetotalobserved rest-frameenergies∼2–8keV,theinferredLX isonlyaffectedfor ow spectrum,adoptingourpriordistributionsforthespectralparame- column densities (cid:3) 1023 cm−2. These redshift-dependent effects nlo a ters.Wenotethatthelargeuncertaintiesinthescatteredfraction, are fully accounted for in our calculation of the expected counts, d e ffsocrattm,oledaedrattoellyaragneduhnecaevritlayinatbiseosribnedthseousrpceecstraunmdtahtussolfetaedrsetnoelragrigees ci(Fz,oLrtXh,eNanHa,lξy)s,isusinintghiosupraXpe-rr,awyesptreecatrtaelamchodsoeul.rceinourhard-or d fro m uncertaintiesintheintrinsicLX. soft-bandsamplesasindependentdetections.Thus,theonlyX-ray h ToobtaintheprobabilitydistributionfunctionforL andN only dataweuseforanindividualsourceisthedetectedcounts(andthe ttp X H s (foragivenzandourobservedX-raydata),wemustmarginalize expectedbackground),whichasdescribedabovedoesnotallowus ://a overthenuisancespectra(cid:2)lparameters.Thus, toconstrainthevalueofNH foranindividualsource.Instead,our cad approach(seeSections3.2below)involvesdeterminingtheoverall e m p(LX,NH|z,Ti,bi)= dξ p(LX,NH,ξ |z,Ti,bi). (10) distributionofX-rayluminositiesandabsorptioncolumndensities ic .o (given by the X-ray luminosity and absorption-distribution func- u InFig.5weshowanexampleofp(LX,NH|z,Ti,bi)forasource tion–XLAF)thatcorrectlydescribesbothourhard-andsoft-band p.c o detected in the hard band (left) and a source detected in the soft samples.Usingthismethod,certainvaluesofL andN maybe m band (right) with Chandra. In both cases we fix z = 1.0 and as- disfavoured a posteriori (after performing our fuXll analyHsis of the /m n s2ou0fmb<ei =Tloig=5N.0H30(ccotmuont−at2ls).o<Wbsee2r6vr.eesTdtrhciecotushtnhatesdiwpnoigtshsainabdnleecxocponeltuocmuternds ibdnaedcnikscigatrtieeosutnhtdoe fsoahvvoeowruaslrlehdsigaahmpeprolsLet)Xe.rsiFoooruirrbceeaxssaesmdhopoulneld,thhbeiegsrhhaaerpreerv)o.afClutohensevXoerfLseFLlXy(,thmloaawtygeerbnveearldaulielsys- ras/article rangeofpossiblevaluesofLXandNHforeachsource.Detectionin of NH may be disfavoured if our overall XLAF requires a large /45 asingle,broadenergybanddoesnotplaceconstraintsonthevalue fractionofabsorbedsources. 1 /2 /1 8 9 2 /1 7 4 7 7 8 4 b y g u e s t o n 0 9 M a rc h 2 0 2 3 Figure5. Exampleconstraintsontherest-frame2–10keVluminosity,LX,andtheline-of-sightabsorptioncolumndensity,NH,forsourcesdetectedinthe observedhardband(2–7keV,left)andsoftband(0.5–2keV,right)inatypicaldeepChandraobservation.Inbothcases,weassume30totalcountsdetected intheobservedenergybandwithanexpectedbackgroundof5counts,typicalofoursampleintheAEGIS-XDsurveyregion.Wefixtheredshiftatz=1.0. ThegreyshadingindicatestheprobabilitydistributionfunctionforLXandNH,allowingforthePoissonuncertaintyfromtheobservedcountsandassuming theX-rayspectralmodelshowninFig.4,marginalizedovertheuncertaintyinthe‘nuisance’spectralparameters((cid:5),fscatt,andR).Thesolidanddashedlines indicatethe68.3and95.4percent(i.e.1σ and2σ equivalent)confidenceintervalsonthejointparameterspace.Detectioninasinglebanddoesnotconstrain thevalueofNHbutdoesplaceconstraintsontherangeofpossiblevaluesofLX.AtlowvaluesofNH,theX-rayluminosityisconstrainedtowithin∼0.25 dex.However,forhigherNHvaluestheX-rayluminositymustbecorrespondinglyhighertoproducethesamenumberofobservedcounts.Absorptionaffects theX-rayspectrum(andthustheconstraintsonLX)forlowercolumndensitiesinthesoftband(NH(cid:2)1022cm−2)thanthehardband(NH(cid:2)1023cm−2) atz=1.0. MNRAS451,1892–1927(2015)

Description:
J. Aird,1,2‹ A. L. Coil,3 A. Georgakakis,4,5 K. Nandra,4 G. Barro6 and P. G. Pérez-González7. 1Institute of Astronomy, University of Cambridge,
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.