MNRAS454,1202–1220(2015) doi:10.1093/mnras/stv2069 Does the obscured AGN fraction really depend on luminosity? S. Sazonov,1,2‹ E. Churazov1,3 and R. Krivonos1 1SpaceResearchInstitute,RussianAcademyofSciences,Profsoyuznaya84/32,117997Moscow,Russia 2MoscowInstituteofPhysicsandTechnology,InstitutskyPer.9,141700Dolgoprudny,Russia 3Max-Planck-Institutfu¨rAstrophysik,Karl-Schwarzschild-Str.1,D-85740GarchingbeiMu¨nchen,Germany Accepted2015September2.Received2015September2;inoriginalform2015July14 ABSTRACT D We use a sample of 151 local non-blazar active galactic nuclei (AGN) selected from the o w INTEGRAL all-sky hard X-ray survey to investigate if the observed declining trend of the nlo a fraction of obscured (i.e. showing X-ray absorption) AGN with increasing luminosity is d e d mostlyanintrinsicorselectioneffect.Usingatorus-obscurationmodel,wedemonstratethat fro in addition to negative bias, due to absorption in the torus, in finding obscured AGN in m h hardX-rayflux-limitedsurveys,thereisalsopositivebiasinfindingunobscuredAGN,dueto ttp Comptonreflectioninthetorus.Thesebiasescanbeevenstrongertakingintoaccountplausible ://m n intrinsiccollimationofhardX-rayemissionalongtheaxisoftheobscuringtorus.Giventhe ra s AGNluminosityfunction,whichsteepensathighluminosities,theseobservationalbiaseslead .ox fo to a decreasing observed fraction of obscured AGN with increasing luminosity even if this rd jo fractionhasnointrinsicluminositydependence.WefindthatifthecentralhardX-raysource u rn inAGNisisotropic,theintrinsic(i.e.correctedforbiases)obscuredAGNfractionstillshowsa als .o decliningtrendwithluminosity,althoughtheintrinsicobscuredfractionissignificantlylarger rg than the observed one: the actual fraction is larger than ∼85percent at L (cid:2) 1042.5erg s−1 at C/ (a1n7g–le60θkoefVa)n,aonbdscduercinregastoersutso,(cid:2)th6is0ipmeprlcieensttahtaLt θ(cid:3)(cid:2)103404◦eirnglso−w1.erInlutemrminsosoiftythAeGhNal,f-aonpdenθin(cid:3)g aliforn 45◦ inhigherluminosityones.If,however,theemissionfromthecentralsupermassiveblack ia In hisocleonissicsotelnlitmwaittehdaalsudmLi/ndo(cid:3)sit∝y-cinodseαp,etnhdeenintttroinrusischdaelfp-eonpdeennincegoafngthleeθob∼sc3u0re◦.dAGNfraction stitute o f T Keywords: galaxies:active–galaxies:nuclei–galaxies:Seyfert. e c h n o lo prompted by the fact that even hard X-ray ((cid:3)10 keV) surveys, gy 1 INTRODUCTION o which are usually flux (or signal-to-noise ratio) limited, should n J AlotofrecentstudiesbasedonX-rayandhardX-rayextragalac- be biased against detection of Compton-thick AGN, i.e. objects an u ticsurveyshavedemonstratedthatthefractionofX-rayabsorbed viewed through absorption column density NH (cid:3) 1024cm−2, let ary (hereafter referred to as obscured) active galactic nuclei (AGN) aloneX-raysurveysatenergiesbelow10keVwhichmustbebi- 3 1 decreases with increasing observed X-ray luminosity, at least at asedagainstevenCompton-thinobscuredsources.Duetothisde- , 2 0 (cid:3)1042ergs−1,bothinthelocal(z≈0)andhigh-redshiftUniverse tection bias, the observed fraction of obscured AGN is expected 16 (Ueda et al. 2003; Steffen et al. 2003; Hasinger 2004; Sazonov to be lower than the intrinsic fraction of such objects. Further- & Revnivtsev 2004 ; La Franca et al.2005 ; Sazonov et al.2007; more, this effect may depend on luminosity, somehow reflecting Hasinger2008;Beckmannetal.2009;Brightman&Nandra2011; the shape of the AGN luminosity function (LF). In fact, as dis- Burlonetal.2011;Uedaetal.2014;Airdetal.2015;Buchneretal. cussed by Lawrence & Elvis (2010), some mid-infrared selected, 2015; note also earlier evidence, Lawrence & Elvis 1982). This radio-selected and volume-limited AGN samples do not demon- mightindicatethattheopeningangleofthe(presumably)toroidal strateanyclearluminositydependenceoftheproportionoftype1 obscuringstructure–thekeyelementofAGNunificationschemes (i.e.containing broademissionlinesintheopticalspectrum)and –increaseswithAGNluminosity,forexampleduetofeedbackof type2AGN. thecentralsupermassiveblackhole(SMBH)ontheaccretionflow. Althoughtherehavebeenpreviousattempts(Uedaetal.2003; CouldtheobservedluminositydependenceoftheobscuredAGN DellaCeccaetal.2008;Maliziaetal.2009;Burlonetal.2011;Ueda fraction arise due to selection effects? This question has been etal.2014)totakeintoaccountdetectionbiaseswhenestimating occasionally raised before (e.g. Mayo & Lawrence 2013 ) and is thespacedensityofobscuredAGNbasedonhardX-raysurveys, they were, in our view, not fully self-consistent and/or used too smallsamplesofhardX-rayselectedAGN.Itisourgoalhereto (cid:5)E-mail:[email protected] improveonbothoftheseaspects. (cid:6)C 2015TheAuthors PublishedbyOxfordUniversityPressonbehalfoftheRoyalAstronomicalSociety ObscuredAGNfractionvs.luminosity 1203 Thepurposeofthepresentstudyisto(i)evaluatetheimpacton theobservedhardX-rayLFandobservedluminositydependenceof theobscuredAGNfractionofthenegativebiasforobscuredAGN discussedaboveandapositivebiasthatwedemonstratelikelyexists forunobscuredAGN,and(ii)reconstructtheintrinsicdependence of the fraction of obscured AGN on luminosity in the local Uni- verse.Ourtreatmentisbasedonarealistictorus-likeobscuration model and makes use of the INTEGRAL/IBIS 7 yr (2002–2009) hardX-raysurveyoftheextragalacticsky.Oursampleconsistsof ∼150local(z(cid:2)0.2)Seyfertgalaxiesandishighlycompleteand reliable.AlthoughtherearenowsignificantlylargerhardX-rayse- lectedsamplesoflocalAGN,basedonadditionalobservationsby INTEGRAL/IBISandespeciallybySwift/BAT,theycurrentlysuf- ferfromsignificantincompletenessasconcernsidentificationand absorptioncolumndensityinformation.Mostimportantly,oursam- D pleislargeenoughtocontainasignificantnumber,17,ofheavily o obscured(NH≥1024cm−2)AGN,forwhichweuseasmuchaspos- wnlo sibleN estimatesbasedonhigh-qualityhardX-rayspectraldata, a H d e inparticularfromtheNuSTARobservatory,whichhasrecentlybeen d systematicallyobservingAGNdiscoveredintheSwift/BATandIN- fro m TEGRAL/IBIShardX-raysurveys. h 0.3WaendadHo0pt=a7(cid:6)0kcomlds−d1arMkpmca−t3t.ercosmologicalmodelwith(cid:3)m = Ffuingcutrioen1o.fCfluuxmluimlatiitvientfhraecItNioTnEoGfRtAhLe7exytrrasguarlvaecyt.ic (|b| > 5◦) sky as a ttp://mn ra s .o 2 THE INTEGRAL AGN SAMPLE xfo rd Weusethecatalogueofsources(Krivonosetal.2010b)fromthe jo u INTEGRAL/IBIS7yrall-skyhardX-raysurvey(hereafter,theIN- rn a TEGRAL7yrsurvey;Krivonosetal.2010a).Tominimizepossible ls.o biasesinourstudyofthelocalAGNpopulationduetoremaining rg a/ unidentifiedINTEGRALsourcesandobjectswithmissingdistance t C asindde/roartioXn-rtahyeaGbasolarcpttiiconpliannfeorrmegaitoionn(,|bw|e<ex5c◦l)u.dTehferocmatatlhoeguceonis- aliforn composedofsourcesdetectedonthetime-averaged(2002Decem- ia ber – 2009 July) 17–60 keV map of the sky and is significance Ins limited (5σ). The corresponding flux limit varies over the sky: titu (fd9e0t p<er2c.e6nt×)of1t0h−e11ex(t<ra7ga×lact1i0c−(1|b1|e>rg5◦s)−s1kcym(−s2e)eFfoigr.510).percent te of T e c The main properties of the INTEGRAL 7 yr survey and of the h n corresponding catalogue of sources were described by Krivonos olo g etal.(2010a,b).Usingthiscatalogue,Sazonovetal.(2010)made y o preliminaryestimatesofthehardX-rayLFoflocalAGNandthe n J dependence of the obscured AGN fraction on luminosity. Subse- an u quent follow-up efforts by different teams have resulted in addi- ary tional identifications, classifications, distance measurements and 3 1 X-rayabsorptioncolumnestimatesformanyINTEGRALsources, , 2 0 which has significantly improved the quality of the catalogue, as Figure2. ObservedhardX-ray(17–60keV)luminosityversusredshiftfor 16 non-blazar AGN from the INTEGRAL 7 yr survey. Filled circles, empty detailedbelow. squaresandstarsdenoteunobscured,lightlyobscuredandheavilyobscured The final sample used here consists of 151 non-blazar (i.e. objects,respectively. Seyfert-like) AGN (see Table A1 in Appendix A), with blazars (15intotal)beingexcludedfromtheanalysis.Thesampleishighly complete,asthereareonlyfoursourcesat|b|>5◦fromtheINTE- out that some of our objects have blazar-like properties, i.e. their GRAL7yrcataloguethatremainunidentified.Moreover,allofour observed hard X-ray emission contains a significant contribution AGNshaveknowndistancesandreliableestimatesoftheirabsorp- fromarelativisticjet.Themostsuspiciousinthisrespectareobjects tioncolumnsbasedonX-rayspectroscopy.AsillustratedinFig.2, classifiedasbroad-line(i.e.presumablyorientedtowardsus)radio oursampleismostlylocal,with146outofthe151objectsbeing galaxies.TherearesixsuchAGNsinoursample:3C111,3C120, locatedatz<0.2,andspansaboutfivedecadesin(observed)lumi- Pic A, 3C 390.3, 4C +74.26 and S5 2116+81. All of them have nosity,fromLobs∼1041to∼1046ergs−1(hereafter,allluminosities Lobs > 1044 erg s−1 (but <1045 erg s−1), i.e. belong to the high- areinthe17–60keVenergyband,unlessspecifiedotherwise). luminositypartofthesample.However,thetotalnumberofobjects Wenotethatalthoughweusedthemostup-to-dateinformation with L > 1044 erg s−1 is much larger: 42. This suggests that obs from the NASA/IPAC Extragalactic Database (NED) and recent possibleincompletefilteringofthesamplefromblazarsisunlikely literaturetoremoveblazarsfromourAGNsample,wecannotrule tosignificantlyaffecttheresultsandconclusionsofthiswork. MNRAS454,1202–1220(2015) 1204 S.Sazonov,E.ChurazovandR.Krivonos Table1. HeavilyobscuredAGNfromtheINTEGRAL7yrsurvey. Object D Lobs NH ReferenceforNH (Mpc) (ergs−1) (cm−2) SWIFTJ0025.8+6818 52.0 3.2×1042 >1025 NuSTAR(Krivonos,inpreparation) NGC1068 12.3 3.8×1041 >1025 NuSTAR(Baueretal.2014) NGC1194 59.0 6.6×1042 ∼1024? XMM–Newton(below10keV;Greenhill,Tilak&Madejski2008) CGCG420-015 129.2 3.3×1043 >1025 NuSTAR(Krivonos,inpreparation) MRK3 58.6 2.8×1043 1024 Suzaku(Ikeda,Awaki&Terashima2009) IGRJ09253+6929 172.6 4.5×1043 >1024? LowX-ray/hardX-rayfluxratio(Swift/XRT+INTEGRAL/IBIS) NGC3081 28.6 4.4×1042 1024 Suzaku(Eguchietal.2011) NGC3281 46.3 1.2×1043 2×1024 BeppoSAX(Vignali&Comastri2002) ESO506-G027 109.5 5.8×1043 1024 Suzaku(Winteretal.2009b) NGC4939 34.7 2.3×1042 >1025? BeppoSAX(Maiolinoetal.1998),butvariedtoNH=1.5×1023cm−2 (XMM–Newton,below10keV;Guainazzietal.2005a) NGC4945 3.4 2.6×1041 4×1024 NuSTAR(Puccettietal.2014;Brightmanetal.2015),Suzaku(Yaqoob2012) D IGRJ14175−4641 348.3 1.6×1044 >1024? LowX-ray/hardX-rayfluxratio(Swift/XRT+INTEGRAL/IBIS) ow NGC5643 11.8 1.7×1041 >1025 NuSTAR(Krivonos,inpreparation) nlo NGC5728 24.8 3.2×1042 2×1024 NuSTAR(Krivonos,inpreparation) ad IGRJ14561−3738 107.7 1.6×1043 ∼1024 Chandra+INTEGRAL/IBIS(Sazonovetal.2008) ed ESO137-G034 33.0 1.8×1042 >1025 Suzaku(Comastrietal.2010) fro m NGC6240 107.3 5.8×1043 2.5×1024 NuSTAR(Krivonos,inpreparation) h ttp 2.1 Absorptioncolumns,heavilyobscuredAGN Intotal,oursampleconsistsof67unobscured(N <1022cm−2) ://m H n Fcoomr pthleetepuanrpdorseelsiaobfletihnifsosrtmudatyi,onitoinstihmepXo-rrtaanytatbosohrapvtieonmcaoxliummanllsy, aonbdsc8u4redob(NscHur≥ed10(N24Hcm≥−12)02o2ncems.−2) AGNs, including 17 heavily ras.oxfo N ,ofthestudiedAGN.Ourstartingsourceofsuchinformation Table 1 provides key information about our heavily obscured rd iei(nsStHaaaozllu.lor2nca0ops0vree7&sv,iw2oR0uhe1sev2rnpe)iaviapttssewerwvsaes2ol0nln0eac4sthe)o,senbsauItNrthywTeaEeRnGhXdaRTpvAEoeLsu(s/p3iIbd–Bla2Iet0Se(dskseetuehVrevT)NeasybHlel(eewSsAtasiz1muo)ra.nvtoeeyvs btwAhuGeattNoatdnh.oeFepo(stIroGstuheRrvacteJen1’Ns4oH5sf6pt>h1eec−st1r3e0u,72mt53h8cei)mrseo−rafe2rflt)eheefrcretorlieimoamnbalNdeinuoNiSmnHTgiAne1saRt0tiemoodbba(stjieeensrcvtowsarthaieorivcenihdse.eictnAhacseleerl ajournals.org/ Forunobscuredandlightlyobscured(N <1024cm−2)sources, planned to be observed by NuSTAR soon or have already been t C Xfo-rraeyvaslpueactitnrogsNcop.ySautchendeartgaiedsobeexloiswtf1o0rakHlelVofisouursusaolulyrcseusfafincdienint oobfswerrviteindgb.yHtohwisetveelre,scfoorpemboustttohfetdhaetsaeasroeuprcreosprtiheetareryexatistthsefatiimrlye aliforn amFduoorstpthtecNramsHeos=rteh0,eiarfentHdahreceoarnbeslsiiodarebprlteisoupncuhbclosiosluhumercdneNsisHunlveoasblsuscethusa,rnewdh1.i0c2h2cwme−a2d,owpte. rNAeHGlTiaN≥hbrle1aeer0ie2no4fcfocutmrhrmre−ea2nottib–loyjnescecftearsonTimdnaibcdollauethtde1ees.rdhraiantrhdoeurXrt-hsraaaymnmpfiliresmsoilfoynhesesaitnvadibliylcisaohtbiensdgcurterhepad-t ia Institute o 10HkoewVevbeerc,oambseorupntrioelniacbolleumfonressttriomnagtleysbabasseodrboendXs-oruarycdeas,tahbaevlionwg resentatives of this class: the quoted value NH ∼ 1024cm−2 for f Tec NraHy≥(a1b0o2v4ec1m0−k2e.VIn)ssupcehcctraossecso,pwye,pwrhefeenretvoeurspeorsessiublltes.fSropmechifiacrdalXly-, NenGceCo1f1N9H4c>om10e2s4fcrmom−2Xa-brasoyrdpatitoanbceololwum1n0skienVI,GwRheJ0re9a2s5t3h+e6p9re2s9- hnolog and IGR J14175−4641 is strongly suggested by very low y ourpreferencelistofinstrumentsisheadedbyNuSTAR–theunique o focusinghardX-raytelescope,followedbySuzakuandthenbyall ((cid:2)0.01) X-ray/hard X-ray flux ratios that we find for them from n Ja Swift/XRT and INTEGRAL/IBISdata. Note that we initially used n othercurrentlyoperatingorpreviouslyflownhardX-raymissions. u a canFdoirdafitvees,owfethcearhreieavdiloyutobosucruorewdn(NanHal≥ysi1s0o24fcpmu−bl2i)cloybjaevcatislaabnlde itSthWeinIsdFaeTmedeJ0pa0rr2og5vue.m8d+etn6ot8b1toe8,rseaugscahardolniakcneeolytwheehreaasnvoaiullyyrcsoeedbfsrNcouumrSeTdtAhAiRsGdsNaamt,a.panlIedn, ry 31, 2 NuSTARdata(Krivonos,inpreparation).Specifically,wefittedthe theanalysisbelow,weassumethatN =3×1024cm−2 forboth 016 spectrabyasumofastronglyabsorbedpower-lawcomponent(with H IGRJ09253+6929andIGRJ14175−4641. ahigh-energycutoff)andadisc-reflectioncontinuummodelledwith pexrav in XSPEC. The NuSTAR spectra of SWIFT J0025.8+6818, itseTlhfeasmaosrtedflieffictciuolnt-dcaosmeiinsattehdatsoofuNrcGeC(N493>9,1w0h25icchmm−2a)nidfuesritnedg CGCG 420-015 and NGC 5643 are consistent with being fully H BeppoSAX observations in 1997 (Maiolino et al. 1998), but was reflectiondominated(i.e.dominatedbyCompton-scatteredcontin- uum),andsoweprescribedN >1025cm−2tothem.Theothertwo foundtobeinaCompton-thinstate,withNH ∼1.5×1023cm−2, H byXMM–Newtonin2001(Guainazzietal.2005a).Wenevertheless objects, NGC 5728 and NGC 6240, along with strong reflection treatNGC4939asareflection-dominatedsourceinouranalysis, demonstrateasignificantcontributionfromtheprimarycomponent suppressed by intrinsic absorption at the level of N ∼ 2–2.5 × in part because the hard X-ray flux measured by INTEGRAL for 1024cm−2.MorephysicallymotivatedAGNtorusmodHelsconfirmed this source is similar to that measured by BeppoSAX but lower than the flux inferred from the XMM–Newton observation and so this qualitative result (see Krivonos, in preparation for details). INTEGRAL may have caught the source in a state similar to that Our derived spectral parameters for CGCG 420-015, NGC 5643, revealed by BeppoSAX. Generally, we adopt N = 1025cm−2 for NGC 5728 and NGC 6240 are consistent with pre-NuSTAR esti- H reflection-dominatedsources(thereareintotalsixsuchobjects)in mates for these objects (Severgnini et al. 2011 ; Matt et al.2013; ouranalysis,althoughinrealitythecolumndensityinsuchobjects Comastrietal.2010;Vignatietal.1999,respectively). maybeevenhigher,sayN ∼1026cm−2. H MNRAS454,1202–1220(2015) ObscuredAGNfractionvs.luminosity 1205 D o w n lo a d e d fro m h ttp ://m FGiRgAuLreA3G. NO.bUsenrovbesdcduirsetdrib(NutHio<no1f0X22-rcamy−ab2s),olripgthiotlnycoobluscmunresdfo(r1t0h2e2I≤NTNEH- Ffuingcutrioen4o.fOobbsseerrvveeddhfarradctXio-nrayofluombsincousreitdyf(oNrHth≥eIN10T2E2GcmRA−L2)7AyGrNsuravseya. nras.ox <inb1l0u2e4,cmma−g2e)ntaanadnhderaevdi,lyreosbpseccutirveedly(.NH≥1024cm−2)objectsareshown fordjo u rn a Wehavethusobtainedafairlylargeandhigh-quality(interms ls .o ofinformationonintrinsicobscuration)sampleofheavilyobscured rg AGN. The high completeness and reliability of this sample are a/ crucialforouranalysisbelow. t Ca lifo rn 3 OBSERVED PROPERTIES OF LOCAL AGN ia In s WefirstconsideranumberofobservedpropertiesofthelocalAGN titu populationusingourINTEGRALsample. te o Fig. 3 shows the observed distribution of absorption columns f T e for our objects, while Fig. 4 shows the observed dependence of ch n the obscured AGN fraction on hard X-ray luminosity. The latter o lo wasobtainedbycountingobscuredandunobscuredsourceswithin gy specifiedluminositybinsanddividingthefirstnumberbythesum on J ofthetwo.Onecanclearlyseeadecliningtrendoftheobscured a n u AGNfractionwithincreasingluminosity,whichiswellknownfrom a ry previousstudies. 3 1 WenextcalculatedtheobservedhardX-rayLF,φ(Lobs)(number , 20 ofobjectsperMpc3 perlogLobs),oflocalAGN:bothinbinnedand 16 analyticform(seeFig.5).TheanalyticLFmodelusedthroughout Figure5. Observed(intheINTEGRAL7yrsurvey)hardX-rayluminosity thisstudyisabrokenpowerlaw: functionoflocalAGN(filledcircles)fittedbyabrokenpowerlaw(black dN A solidline).Thebest-fittingparametersaregiveninTable2.Forcomparison, dloAgGLN = (L/L∗)γ1+(L/L∗)γ2. (1) theLFbasedontheSwift/BATsurvey(Ajelloetal.2012)isshownbythe magentadashedline. ThebinnedLFwasconstructedusingthestandard1/V method, max whereasthebest-fittingvalues(andtheiruncertainties)ofthechar- 7yrsurveyforagivenL ,whichcanbecalculatedfromthesky acteristicluminosity,L∗,andofthetwoslopes,γ1 andγ2,ofthe coveragecurve(seeFig.1o)b.sThenormalizationoftheanalyticmodel analyticmodel(seeTable2)werefoundusingamaximumlikeli- isderivedfromtheactualnumberofobjectsinthesample. hoodestimator(similarlytoSazonovetal.2007): ComparingthisnewlydeterminedobservedhardX-rayLFwith L=−2(cid:2)ln(cid:3) φ(Lφ(L)oVbs,i)V(mLax(L)doblso,ig)L , (2) oveurrsuosld1r5e1suolbtj(eScatsz)onoofvAeGtNal.de2t0e0c7te)dbwasiethd oINnTaEGsmRaAlLle,rwseetfi(6n6d i obs max obs obs good agreement between the two, but the constraints on the LF whereL aretheobservedluminositiesofAGNinoursample,and parametershavenowsignificantlyimproved.Wecanalsocompare obs,i V (L )isthevolumeoftheUniverseprobedbytheINTEGRAL theINTEGRALLFwiththatderivedfromastilllarger(361objects) max obs MNRAS454,1202–1220(2015) 1206 S.Sazonov,E.ChurazovandR.Krivonos Table2. FitsofdifferenthardX-rayluminosityfunctionsbyabrokenpowerlaw. AGN NAGN logL∗ γ1 γ2 Aa, Num.density Lum.density class 10−5Mpc−3 (logL=40.5–46.5) (logL=40.5–46.5) 10−4Mpc−3 1039ergs−1Mpc−3 ObservedLF All 151 43.74±0.19 0.93±0.10 2.33±0.15 1.122 54(41÷82) 1.57±0.20 Unobscured 67 43.98±0.32 0.83±0.18 2.37±0.28 0.243 10(7÷25) 0.45±0.08 Obscured 84 43.65±0.21 0.99±0.21 2.48±0.22 0.839 48(37÷78) 1.14±0.18 IntrinsicLF,isotropicemission,θ=30◦ All 151 43.69±0.18 0.89±0.10 2.36±0.16 1.806 61(47÷96) 1.97±0.25 Unobscured 67 43.87±0.31 0.84±0.18 2.39±0.28 0.201 7.0(5.1÷16.7) 0.30±0.05 Obscured 84 43.62±0.21 0.93±0.12 2.45±0.21 1.581 59(46÷101) 1.63±0.26 IntrinsicLF,isotropicemission,θ=45◦ All 151 43.71±0.19 0.89±0.10 2.36±0.16 1.494 52(41÷82) 1.71±0.21 Unobscured 67 43.90±0.31 0.84±0.18 2.39±0.27 0.203 7.5(5.4÷18.2) 0.32±0.06 Do Obscured 84 43.62±0.21 0.92±0.12 2.44±0.21 1.698 59(46÷105) 1.70±0.26 wn IntrinsicLF,cosine-lawemission,θ=30◦ loa d All 151 43.69±0.17 0.90±0.10 2.43±0.17 1.593 57(44÷87) 1.79±0.21 ed Unobscured 67 43.70±0.30 0.86±0.17 2.42±0.28 0.188 5.3(3.9÷11.6) 0.19±0.03 fro Obscured 84 43.68±0.21 0.93±0.12 2.45±0.21 1.565 66(51÷114) 1.88±0.30 hm Note.aThenormalizationAisgivenwithoutanerrorbecausethisparameterisstronglycorrelatedwiththeothers. ttp://m n ra s .o x fo rd jo u rn a ls .o rg a/ t C a lifo rn ia In s Figure7. Torusmodel. titu te o f T observer,norforanyintrinsicanisotropyoftheemissiongenerated e c h bythecentralsourceinAGN.ThisobservedLFisexpectedtobeaf- n o fectedbyabsorptionbias:anobscuredAGNwillbeinferredtohave lo g alowerluminosity,Lobs =fobs×4πD2(here,fobsisthemeasured y o n hardX-rayfluxandDisthedistancetothesource),thanitsintrinsic J a (i.e. emitted by the central source) luminosity, Lintr, and a source nua likethiscanbefoundinaflux-limitedhardX-raysurveywithina ry 3 F10ig22urcem6−.2,ObbluseervfielldedhacridrcXle-sr)ayanludmoibnsocsuitryedfu(nNcHtio≥ns1o0f22uncomb−s2c,urreedd(eNmHp<ty sXm-raalylearbvsoolrupmtieono:fVthmeaxU(Lnoibvs)e/rsVemtahx(aLninittr)w≈ou(lLdobbse/Linintrt)h3e/2a.bHseenrec,ethoef 1, 201 squares)AGN,fittedbybrokenpowerlaws(bluedottedandreddashed approximationsymbolreflectsthefactthatAGNobscurationmay 6 lines,respectively).Thebest-fittingparametersaregiveninTable2. alsoaffecttheshapeofthemeasuredX-rayspectrumandthusthe number of photons recorded by a given detector with its specific sampleof(mostly)localAGNfoundinnearlythesameenergyband energyresponse.Ontheotherhand,asdiscussedbelow,unobscured (15–55keV)intheSwift/BATsurvey(Ajelloetal.2012).Ascanbe AGNareexpectedtohavehigherobservedluminositiesthantheir seeninFig.5,thetwoLFsareingoodagreementwitheachother. intrinsic angular-averaged luminosities and can thus be detected Finally,wecalculatedseparatelytheobservedLFsofunobscured within a larger V . We can correct for both of these biases and max andobscuredAGN(seeFig.6).ItcanbeseenthattheseLFsare obtainanintrinsichardX-rayLFoflocalAGN.Tothisend,weuse differentinshape,asisverifiedbythebest-fittingparametersofthe aphysicallymotivatedobscurationmodeldescribedbelow. correspondinganalyticfits(seeTable2). 4 INTRINSIC PROPERTIES OF LOCAL AGN 4.1 TorusmodelandAGNspectra TheobservedLFjustdiscussedhasnotbeencorrectedforanyef- We have a developed a Monte Carlo code for modelling AGN fectsassociatedwithabsorptionorscatteringofhardX-raysemitted X-ray spectra modified by reprocessing in a toroidal structure of bytheAGNcentralsourceonthewaybetweenthesourceandthe gas. The adopted geometry (see Fig. 7) is similar to that used in MNRAS454,1202–1220(2015) ObscuredAGNfractionvs.luminosity 1207 otherexistingmodels,e.g.Ikedaetal.(2009),Murphy&Yaqoob (2009)andBrightman&Nandra(2011).Thekeyassumptionsof ourmodelare: (i) Thegeometricalshapeisthatofaringtorus. (ii) Thegasishomogeneous,cold,neutralandofnormalcosmic chemicalcomposition. (iii) The X-ray spectrum emitted by the central source is a powerlawwithanexponentialcutoff,dN/dE∝E−(cid:10)e−E/Ecut,with (cid:10)=1.8andE =200keV. cut (iv) The central (point-like) source is either isotropic, dL /d(cid:3) = const – hereafter, Model A, or emitting according intr toLambert’slaw,dL /d(cid:3)∝cosα,whereαistheviewingangle intr withrespecttotheaxisofthetorus–hereafter,ModelB. TheintroductionofModelBisanimportantaspectofthepresent D studyandismotivatedbythecommonbeliefthatthehardX-ray o w emission observed from AGN is produced by Comptonization of n lo softeremissionfromanaccretiondiscaroundanSMBHinahot a d e coronalyingabovethedisc.Ifsuchacoronahasquasi-planarge- d ometry, the hard X-ray flux it produces will be collimated along fro m theaxisofthedisc/coronaroughlyasF∝μ(theexactlawbeing h dependentonthephotonenergyandtheopticaldepthofthecorona; ttp Pozdnyakov,Sobol&Sunyaev1983;Sunyaev&Titarchuk1985), ://m n whereμisthecosineoftheanglebetweentheoutgoingdirection ra s andtheaxisofthedisc/corona.Becausetheobscuringtorusinturn .ox islikelycoalignedwiththeaccretiondisc,theemergenthardX-ray ford radiationwillbecollimatedalongtheaxisofthetorus.Inreality,a jo u significantfractionofthecoronalemissionisreflectedbytheunder- rn a lyingaccretiondisc,butthisalsooccurspreferentiallyalongtheaxis ls.o ofthedisc/torus(Magdziarz&Zdziarski1995).Sincethereisstill rg a/ significantuncertaintyintheoverallphysicalpicture,weintroduce t C taosgimetpalen,iedneeargoyf-ihnodwepsetnrodnegnltycoinlltirminastiicocnoflalicmtoartidoLninotrf/dh(cid:3)ard∝Xco-rsaαy aliforn emissioncanaffectobservedpropertiesoflocalAGN.Wenotethat ia possibleeffectsofanisotropicemissionontheobservedLFandob- Ins scuredfractionofAGNhavebeenpreviouslydiscussedbyZhang titu te (2005)andLiuetal.(2014). o Apart from the two alternatives for the angular dependence of f T e c intrinsic emission (Model A or Model B), our model has three h n freeparameters:(i)theequatorialcolumndensity,NH,eq (thetotal olo numberofHatomspercm2alonganequatoriallineofsightbetween gy o thecentralsourceandtheobserver,(ii)half-openingangleofthe n J torus,θ and(iii)theviewinganglerelativetotheaxisofthetorus, an u α(seeFig.7). ary X-rayphotonsemittedbythecentralsourcecanscattermultiple 3 1 timeswithinthetorusbeforetheyeithergetphotoabsorbedinthe , 2 0 gasorescapefromthesystem.Ourradiativetransfercalculations Figure 8. Top: examples of simulated AGN spectra for Model A, half- 16 arebasedonamethoddevelopedbyChurazovetal.(2008).Thegas opening torus angle θ = 30◦ and equatorial column density NH,eq = inthetorusisassumedtobeneutral,withtherelativeabundances 1025cm−2, for various viewing angles. The dashed curve shows the in- ofallelementsasinthesolarphotosphere.Thefollowingprocesses trinsic(angular-averaged)spectrum.TheshadedareaindicatestheINTE- areincludedinthesimulations:photoelectricabsorption,Rayleigh GRAL/IBISenergybandusedforAGNselectioninthiswork.Bottom:The same,butforModelB. and Compton scattering and fluorescence. Photoelectric absorp- tion is calculated using the data and approximations of Verner & Yakovlev(1995)andVerneretal.(1996).Forfluorescence,weuse Fig.8showsexamplesofemergentAGNspectrasimulatedusing theenergiesandyieldsfromKaastra&Mewe(1993).Comptonand ourmodeloftheobscuringtorus.Asexpected,forobscuredAGN Rayleigh scattering are modelled using differential cross-sections ( α >θ),theobservedhardX-rayfluxcanbestronglyattenuated provided by the GLECS package (Kippen 2004 ) of theGEANT code relativetotheemittedfluxandforhighabsorptioncolumns(NH(cid:10) (Agostinelli et al. 2003). Namely, the Livermore Evaluated Pho- 1024cm−2)thespectrumcanbecomereflectiondominated,asthe ton Data Library (see Cullen, Perkins & Rathkopf 1990) and the observerwillmostlyseeemissionreflectedfromtheinnerwallsof Klein–Nishinaformulaforfreeelectronsareusedtocalculatetotal thetorusratherthanemissionfromthecentralsourcetransmitted cross-sectionsandtheangulardistributionofscatteredphotonsfor through the torus. In this last case, there also appear strong iron eachelement. Kα and Kβ fluorescent lines at 6.4 and 7.06 keV. These spectral MNRAS454,1202–1220(2015) 1208 S.Sazonov,E.ChurazovandR.Krivonos propertiesandtrendsforheavilyobscuredAGNareofcoursewell known. WefurtherseefromFig.8thatthespectraobservedfromdirec- tionsα<θ,correspondingtounobscuredAGN,alsodifferfromthe intrinsic spectrum. Namely, they have an excess due to Compton reflectionofhardX-raysfromthetorusinthedirectionoftheob- server.Thishumpislocatedapproximatelywithintheenergyband of 17–60 keV that we use for detecting AGN in the INTEGRAL survey.ItisobviousthatthisComptonreflectioncomponentshould biasobservedluminositiesofunobscuredAGNhigherinthisand similar(e.g.Swift/BAT)hardX-raysurveys.Anyintrinsiccollima- tionofemissionalongtheaxisoftheobscuringtoruswillmakethis positivebiasevenstronger(seethespectrumforModelBandα≈ 0inthelowerpanelofFig.8).Thisimportantaspectisfrequently overlooked in AGN population studies, even though a reflection D componentiswellknowntobepresentinthehardX-rayspectraof o w unobscuredAGN. n lo a d e d 4.2 AGNdetectionbias fro m Toquantifybiasesaffectingdetectionofunobscuredandobscured h ttp aAnGdNMiondethleBI,NthTeErGatRioA,LRs(NurHv,eeqy,,θw,eα)sh=owLobins/LFiingtr.,9o,fftohreMobosdeervleAd ://mn tointrinsicluminosityinthe17–60keVenergybandasafunction ras ofN ,foratorushalf-openingangleθ =30◦andseveralnarrow .ox rangHes,eqoftheviewingangleα.OnecanseethatRisalwayslarger ford Mtahfaoendweul1nA0it2y4a,ncimd.e.α−L2≈oabns0d>,RreLmiinnatcri,rnefsaosaretusanfporpobrmsocxu1irmteodat∼AelGy2Ntahs.iNsFHloe,revqeeixlnatchmreepraelseae,fstfeotror. journals.o rg Thistrendcanbeeasilyunderstood:theamplitudeofComptonre- a/ flectionisexpectedtobeproportionaltothetorusopticaldepth,τ, t C cinastehe(τo(cid:10)ptic1a)l.lAysthriengarredgsimobes(cτur(cid:11)ed1A)GaNnd(cαo>nstθa),nRtidnetchreeaospespowsiitthe aliforn –iNnHdc,rueeqea∼stoin3tghe×NrHe1,efl0qe2a4cntcedmdi−nc2ocrmaetapnsoiennaegrn-αte)q,(uaaasptaocrrotiuafllrdodmbireeacetxliopocenacsltefmodardxMuimeoutdomelthaBet ia Institute increasingattenuationofthetransmittedcomponent.Themostob- of T viousandimportantdifferenceofModelBwithrespecttoModelA e c isthattheobservedhardX-rayfluxisanisotropicevenintheab- hn o senceofanobscuringtorus(i.e.forNH=0)–justduetotheinitial log collimationofemission. y o We can proceed further and ask the question: what would be n J a the average observed/intrinsic flux ratio for the local populations nu a ofunobscuredandobscuredAGNif(i)AGNtoriwererandomly ry orientedwithrespecttotheobserver,whichisanaturalassumption, 31 and (ii) all the tori had the same half-opening angle θ (this, of Figure9. Top:calculatedratioofobservedtointrinsic(angular-averaged) , 20 1 course, permits the physical size of the torus to vary from one luminosityinthe17–60keVenergybandforatorushalf-openingangle 6 objecttoanotherand,e.g.todependonluminosity).Tothisend, θ=30◦,fordifferentviewingangles(α),asafunctionofthetoruscolumn wejustneedtoaveragethedependencesshowninFig.9overthe densityforModelA.Bottom:thesame,butforModelB. viewingangleαfortheunobscuredandobscureddirections: (cid:3) byNH,eq ∼1.5×1025cm−2 andstaysatapproximatelythislevel Runobsc(NH,eq,θ)= c1osθR(N1H−,eqc,oθs,θα)dcosα, (3) fAoGrNhigmhoernoctoolnuimcanlldyednescitrieeass.eTshferoamver1agtoe∼ra0ti.o2RasobsNcHf,oeqrionbcsrceuarseeds from(cid:11)1024cm−2to∼1.5×1025cm−2andstaysatthislevelthere- Robsc(NH,eq,θ)= (cid:3)0cosθR(NHc,eoqs,θθ,α)dcosα . (4) aqofubtaseclri.utarIentidvtehdleyircesacimstieoilnoasfr,aibsucmtostohirneeec-ploarnowtnraoesumtnibctteeitdnw:geisetoniustrhcpeer,eustnheonebtssaictluurreaeatiddoynanaidst The result is shown in Fig. 10 as a function of N for torus N = 0 and increases further, due to Compton reflection, with H,eq H half-opening angles θ = 30◦, 45◦ and 60◦. One can see that increasingN . H for Model A, R (the average observed/intrinsic flux ratio ItisobviousfromFig.10thatahardX-raysurvey,liketheones unobsc for unobscured AGN) reaches a maximum of ∼1.5–2, depend- performed by INTEGRAL and Swift, will find unobscured AGN ing on θ, at N ∼ 5 × 1024cm−2, then declines to ∼1.4–1.7 more easily than even lightly obscured objects, let along heavily H,eq MNRAS454,1202–1220(2015) ObscuredAGNfractionvs.luminosity 1209 torus.GiventhefairlysmallnumberofobscuredAGN,especially of heavily obscured ones, in our sample, we are bound to make somesimplifyingassumptions.Forexample,wemayassumethat the intrinsic N distribution does not depend on luminosity. In H,eq this case, the intrinsic N distribution can be estimated simply H,eq bydividingtheobservedonebyRo3b/s2c(NH,eq,θ)(andnormalizing theresultingdependencesothatitsintegraloverN equalsunity), H with the bias factor R = L /L having been discussed in obsc obs intr Section4.2. In doing this exercise, we assumed that NH,eq=(4/π)NH≈ 1.27N for our obscured AGN. This is because we do not know H the orientation of our objects apart from the fact that some of themareunobscuredandhenceα <θ,whileothersareobscured and hence α > θ. For our assumed torus geometry (see Fig. 7), theline-of-sightcolumndensitydependsontheviewingangleas D follows: o (cid:4) w (cid:5) (cid:6) n cosα 2 lo NH(α)=NH,eq 1− cosθ , (5) aded sothatthemeanN overallobscureddirectionsis fro H m NH,obsc= (cid:3)0cosθNcHo(sαθ)dcosα = π4NH,eq. (6) http://m n Hence,thecoefficientintheconversionofN toN above.Note ra that the NH values adopted from the literaHture foHr,eqsome of our s.oxfo Compton-thick AGN may already have been ascribed the mean- rd ingofanequatorialratherthanline-of-sightcolumndensitybythe jou correspondingauthors.However,consideringoursampleofheav- rna ilyobscuredsourcesasawhole,theinformationitcontainsonthe ls.o absorptioncolumnsisveryheterogeneous,asitisbasedonvarious arg/ spectralmodelsusedbyvariousauthors.Fortunately,atypicalex- t C ∼pe2c0tepdedricfefnerte(nsceeebeeqtuwaetieonnN6Ha,ebqoavned)NanHdfohrasobnsecgulrigedibAleGimNpiascotnolyn aliforn ourresults. ia In TheresultingintrinsicN distributionispresentedinFig.11. s It is only weakly dependeHn,etqon both the assumed half-opening titu te angle θ of the obscuring torus and the assumed emission model o (ModelAorModelB).Thisdistributioncanberoughlydescribed f Te as log-uniform between NH,eq = 1022 and 1026cm−2, although chn o the upper boundary is, of course, fairly uncertain. A similar re- lo g sultwaspreviouslyobtainedusingAGNfromtheSwift/BAThard y o X-raysurvey(Burlonetal.2011;Uedaetal.2014).Moreover,the n J a intrinsicN distributionshowninFig.11issimilartotheonein- n H u a ferredforopticallyselectedSeyfert2galaxies(Risaliti,Maiolino& ry Salvati1999). 31 , 2 0 1 6 Figure10. Top:calculatedratioofobservedtointrinsic(angular-averaged) 4.4 Intrinsicluminosityfunction luminosityinthe17–60keVenergybandaveragedseparatelyoverallun- obscuredandobscureddirections(α<θandα>θ,respectively)forthree We now calculate the intrinsic hard X-ray LF of unobscured and valuesofthetorushalf-openingangle,asafunctionofthetoruscolumn obscuredAGN,φ(L )≡dN/dlogL .AsfortheobservedLFs intr intr density,forModelA.Bottom:thesame,butforModelB. discussedinSection3,weusebothbinnedandanalyticrepresenta- tions. obscuredones.Ourgoalnowistocorrecttheobservedstatistical ForthebinnedLFs,theprocedureisasfollows: propertiesoflocalAGNforthisobviousbias. (i) First, based on the observed luminosity L and estimated obs,i torus column density N of each source in the sample (Ta- 4.3 Intrinsicdistributionoftoruscolumndensities H,eq,i bleA1),wedetermineitsintrinsichardX-rayluminosityaseither Wecanfirstestimatetheintrinsicdistributionofthecolumnden- L = L /R (N , θ) (for unobscured sources) or L intr,i obs,i unobsc H,eq,i intr,i sities,N ,ofAGNtorii.Tothisend,weneedtocorrecttheob- = L /R (N , θ) (for obscured sources), where the ratios H,eq obs,i obsc H,eq,i servedN distribution(Fig.3)forabsorptionbias,excludingfrom R and R are calculated as discussed above (from equa- H unobsc obsc the consideration the first, N < 1022cm−2, bin since it pertains tions3and4,seeFig.10),assumingsome(thesameforallobjects) H tounobscuredAGNforwhichourlineofsightdoesnotcrossthe torushalf-openingangleθ andusingModelAorModelB.Here MNRAS454,1202–1220(2015) 1210 S.Sazonov,E.ChurazovandR.Krivonos Aswassaidbefore,theN valuesforourobscuredobjectsare H,eq estimated from their measured N columns as N = 1.27N . H H,eq H However,wecannotdeterminesimilarlythetoruscolumndensities forourunobscuredAGN.1Therefore,wesimplyassumethatN H,eq =1024cm−2 fortheseobjects,sincethisisapproximatelytheme- dianvalueoftheinferredintrinsicabsorptioncolumndistribution forobscuredAGN(seeFig.11). (ii) Secondly, we calculate for each source the volume of the Universe,V (L ),overwhichAGNwithsuchobservedlumi- max,i obs,i nositycanbedetectedintheINTEGRALsurvey.Sincethedetection limitforagivenhardX-rayinstrument(IBISinourcase)isactu- allydeterminedbyphotoncounts,itshoulddependontheobserved X-rayspectralshape,whichfortheproblemathandisaffectedby absorptionandreflectioninthetorus(seeexamplesofAGNspectra inFig.8).WecorrectV forthiseffect,butthiscorrectionproves max,i D tobenegligible(asisthek-correctionduetocosmologicalredshift). o w Asaresult,weobtainessentiallythesameVmax,iforoursourcesas nlo weusedinconstructingtheobservedLFinSection3. a d (iii) The final step consists of summing up the 1/V contri- ed butionsoftheindividualsources,i.e.addingthe1/V max,fioreach fro max,i m AGNofagivenclass(unobscuredorobscured)tothespacedensity h ofsuchobjectswithinaluminositybincontainingL (ratherthan ttp L )forthissource. intr,i ://m obs,i n ra ToobtainanalyticformsoftheintrinsicLFs,weusethesamebroken s.o x power-lawmodelasforourobservedLFsbutadifferentlikelihood fo eLst=im−at2or(cid:2):i ln(cid:3) (cid:3) φφ((LLiinnttrr),iV)m(cid:3)axV(mLaixnt(rL,NintHr,i,e,qN)dH,leoqg)dLliontgrdNloHg,eqNH,eq. rdjournals.org (7) a/ t C HouerreobLjienctrt,isaarsewtheeussaemdebeefsotriemtaotecsonosftrtuhcetitnhteribnisnicnelduminitnroinssitiicesLFosf aliforn o(ib.es.cucraelcduAlaGteNdafrnodmasLsuobms,iinugsitnhgattNhe act=ua1l0N24Hc,ime−st2imfoartethsefournothbe- ia Ins scuredones),butV (L ,N )iHs,enqowthevolumeoverwhich titu max intr H,eq te AGNwithgivenintrinsicluminosityL andtoruscolumndensity o NvoHl,ueqmceasn,wbeedaegtaeicnteudsienththeeαIN-aTvEerGagReAdiLntqrsuuarnvteiyti.eTsoRcunaolbcsuc(laθt,eNthHe,esqe) f Techn (infittingtheintrinsicLFofunobscuredAGN)andRobsc(θ,NH,eq) olog (in fitting the intrinsic LF of obscured AGN). The integrals over y o dlogNH,eq in equation (7) are computed from 1022 to 1026cm−2, n J a i.e.weassumethattheintrinsicdistributionoftoruscolumndensi- n u a tiesislog-uniformoverthisrange,assuggestedbytheresultofour ry preceedinganalysisshowninFig.11. 31 Fig. 12 shows the resulting intrinsic LFs for unobscured and , 2 0 obscured AGN, calculated assuming θ = 30◦ for Model A and 16 Figure11. Reconstructedintrinsicdistributionofcolumndensitiesofob- scuringtoriinlocalAGN,calculatedassumingθ=30◦andeitherModelA ModelB.Weseethatintheformercase,theshapesoftheintrinsic LFs of unobscured and obscured AGN are clearly different from (top)orModelB(bottom).Thedottedlinecorrespondstoalog-uniform distribution. eachother,althoughtoalesserdegreethatitwasfortheobserved LFs (Fig. 6) from which they derive. However, for Model B the intrinsicLFsofunobscuredandobscuredAGNarenotsignificantly again we use the average ratios Runobsc and Robsc rather than the differentinshapefromeachother.Theseconclusionsareverified viewing-angledependentR(NH,eq,θ,α)fromwhichtheyderivefor bythebest-fittingparametersobtainedfortheseLFs(seeTable2). the lack of knowledge of the orientation of our objects. Strictly NotethatthederivedintrinsicLFs(bothbinnedandanalyticones) speaking,thisprocedureisnotfullycorrect,becauseforgivenN H,eq and θ, a hard X-ray flux-limited survey will preferentially find objects with smaller viewing angles α within the corresponding groupsofα<θ andα>θ,asisclearfromFig.9.However,this 1Inprinciple,onecouldtrytoestimateNH,eqforunobscuredAGNfromthe contributionofthereflectioncomponenttotheobservedspectrum,butthat mayberegardedasanext-ordercorrectiontothebiasconsidered requireshigh-qualityhardX-raydata,whichisnotalwaysavailable,and here and does not significantly affect our results, as we verify in ismodeldependent.Inparticular,theresultwilldependontheunknown Section5. openingangleθ. MNRAS454,1202–1220(2015) ObscuredAGNfractionvs.luminosity 1211 (ii)eachobscuredAGNfromtheINTEGRALsamplemovesitsown distancetotheright-handsideoftheplot,thisshiftbeingsmallfor lightly obscured objects (N < 1024cm−2) but substantial (up to H logL /L ∼1forN (cid:3)1025cm−2)forheavilyobscuredones. intr obs H Finally,wecancalculatetheintrinsicLFoftheentirelocalAGN population, by summing up the contributions of unobscured and obscuredsources.Inobtainingtheanalyticfitinthiscase,wedefine V (L ,N )asfollows: max intr H,eq Vmax=Vmax,unobsc(1−cosθ)+Vmax,obsccosθ, (8) whereV (L ,N )andV (L ,N )arethecorre- max,unobsc intr H,eq max,obsc intr H,eq spondingvolumesforunobscuredandobscuredAGN.Theresulting LFisshowninFig.13anditsbest-fittingparametersarepresented inTable2. OnecanseethatthetotalintrinsicLFisnotverydifferentfrom D thetotalobservedLF.Thismeansthatthetwoeffectsobservedin o w Fig. 12, namely the shift of the LF of unobscured AGN to lower n lo luminosities and the shift of the LF of obscured AGN to higher ad e luminosities almost compensate each other, with this conclusion d beingonlyweaklysensitivetotheassumedtorusopeningangleand fro m angulardependenceofintrinsicemission. h ttp ://m 4.5 TotalAGNspacedensity n ra s Integration of the total intrinsic and observed LFs over luminos- .ox itysuggeststhatthecumulativehardX-rayluminositydensityof ford local AGN may be underestimated by the observed LF by ∼10– jo u 30percent,althoughthisincreaseisstatisticallyinsignificant(see rn a Table2).Specifically,theintrinsicluminositydensityofAGNwith ls.o L >1040.5ergs−1isfoundtobe∼1.8×1039ergs−1Mpc−3(17– rg intr a/ 60keV),withtheexactvalueslightlydependingonourassumptions t C (seFeoTrable2o)u.r assumed intrinsic AGN spectrum aliforn (dN/dE∝E−1.8e−E/200keV), the ratio of luminosities in the ia 2–10 and 17–60 keV energy bands is about unity. Therefore, the Ins luminosity density of AGN with L > 1040.5 erg s−1 may be titu estimatedat∼1.8×1039ergs−1Mpicn−tr3alsointhestandardX-ray te o band(2–10keV).Wemaycomparethisvaluewithapredictionfor f T e z=0basedonaredshift-dependentintrinsicLFderivedbyUeda ch n o etal.(2014)usingalargeheterogenoussampleofAGNcompiled lo g from various surveys. Integration of this LF over the luminosity y rangefrom1040.5to1046.5ergs−1gives∼8×1038ergs−1Mpc−3 on J (2–10 keV), which is a factor of ∼2 smaller than the above an u estimate.Inreality,theLintr(17–60keV)/Lintr(2–10keV)ratiomay ary wellbe∼1.5ratherthan∼1duetotheexpectedpresenceinAGN 3 1 spectra of a Compton reflection component associated with the , 2 0 Figure12. Top:reconstructedintrinsichardX-rayLFsofunobscured(blue accretion disc. In fact, this component was already discussed in 16 filledcircles)andobscured(redemptysquares)AGN,fittedbyabroken Section4.1asoneofthereasonswhyhardX-rayemissionmaybe power law (the best-fitting parameters are given in Table 2; blue dotted intrinsicallycollimatedinAGNandisimplicitlytakenintoaccount lineandreddashedline,respectively).ModelAisadopted,withθ =30◦. in our anisotropic Model B. Taking this spectral component into Bottom:thesame,butforModelB. account, we can lower our estimate of the luminosity density to ∼1.2×1039ergs−1Mpc−3(2–10keV),whichisstillhigherthan are only weakly sensitive to the torus half-opening angle θ that the Ueda et al. (2014) result by a factor of ∼1.5. The remaining wasassumedinconstructingthem,andnearlythesameresultsare differencemayberelatedtothedifferentproceduresusedinthese obtainedforθ=30◦andθ=45◦.Thisisduetotheweaksensitivity works to construct the intrinsic LFs and to the larger and more oftheR andR factorstoθ (seeFig.10). completesampleoflocalheavilyobscuredAGNusedinourstudy. unobsc obsc The transformation of the observed binned LFs to the intrinsic onescanbeunderstoodasfollows:(i)allunobscuredAGNmaking 4.6 IntrinsicdependenceofobscuredAGNfraction uptheLFshiftbythesameamount,logL /L ∼0.1and∼0.3for obs intr onluminosity ModelAandModelB,respectively,totheleftalongtheluminosity axis(sincewehaveassumedthesameequatorialopticaldepthof SimilarlytotheobservedLF,theobserveddependenceofthefrac- thetorus,N =1024cm−2,forallofourunobscuredobjects),and tionofobscuredAGNonluminosity(Fig.4)mustbeaffectedby H,eq MNRAS454,1202–1220(2015)
Description: