Mon.Not.R.Astron.Soc.000,1–15(2014) Printed23January2014 (MNLATEXstylefilev2.2) Morphologies of z∼0.7 AGN Host Galaxies in CANDELS: No trend of merger incidence with AGN luminosity C. Villforth1,2, F. Hamann1, D. J. Rosario3, P. Santini4, E. J. McGrath5, A. van der Wel 6, Y. Chang 6, Y. Guo7, T. Dahlen8, E. F. Bell9, C. J. Conselice10, D. Croton11, A. Dekel12, 4 1 S. M. Faber7, N. Grogin8, T. Hamilton13, P. F. Hopkins14,15, S. Juneau16, J. Kartaltepe17, 0 D. Kocevski18, A. Koekemoer8, D. Koo7, J. Lotz8, D. McIntosh19, M. Mozena7, R. Somerville20, 2 n V. Wild2 a 1DepartmentofAstronomy,UniversityofFlorida,32611Gainesville,FL,USA J 2SUPA,UniversityofSt.Andrews,SchoolofPhysicsandAstronomy,NorthHaugh,KY169SS,St.Andrews,Fife,Scotland 1 3Max-Planck-Institutfu¨rExtraterrestrischePhysik(MPE),Postfach1312,85741Garching,Germany 2 4INAF OsservatorioAstronomicodiRoma,viadiFrascati33,00040MontePorzioCatone,Italy 5DepartmentofPhysicsandAstronomy,ColbyCollege,Waterville,ME04901,USA ] 6MaxPlanckInstituteforAstronomy,Ko¨nigstuhl17,69117Heidelberg,Germany A 7UniversityofCaliforniaObservatories/LickObservatory,UniversityofCalifornia,SantaCruz,CA95064,USA G 8SpaceTelescopeScienceInstitute,3700SanMartinDr.,Baltimore,MD,21218,USA 9DepartmentofAstronomy,UniversityofMichigan,AnnArbor,MI,48104,USA . h 10UniversityofNottingham,SchoolofPhysicsandAstronomy,NottinghamNG72RD,UK p 11CentreforAstrophysicsandSupercomputing,SwinburneUniversityofTechnology,POBox218,Hawthorn,VIC3122,Australia - 12RacahInstituteofPhysics,TheHebrewUniversity,Jerusalem91904,Israel o 13DepartmentofNaturalSciences,ShawneeStateUniversity,Portsmouth,OH45662,USA tr 14TAPIR,Mailcode350-17,CaliforniaInstituteofTechnology,Pasadena,CA91125,USA s 15DepartmentofAstronomyandTheoreticalAstrophysicsCenter,UniversityofCaliforniaBerkeley,Berkeley,CA94720,USA a 16CEASaclay,IRFU/SAp,91191Gif-Sur-Yvette,France [ 17NationalOpticalAstronomyObservatory,950N.CherryAve.,Tucson,AZ85719,USA 1 18DepartmentofPhysicsandAstronomy,UniversityofKentucky,Lexington,KY40506,USA v 19DepartmentofPhysics&Astronomy,UniversityofMissouri-KansasCity,5110RockhillRoad,KansasCity,MO64110,USA 7 20DepartmentofPhysicsandAstronomy,RutgersUniversity,136FrelinghuysenRoad,Piscataway,NJ08854,USA 7 4 5 . 1 0 4 1 : v i X r a (cid:13)c 2014RAS 2 Villforthetal. ABSTRACT The processes that trigger Active Galactic Nuclei (AGN) remain poorly understood. While lower luminosity AGN may be triggered by minor disturbances to the host galaxy, stronger disturbances are likely required to trigger luminous AGN. Major wet mergers of galaxies areidealenvironmentsforAGNtriggeringsincetheyprovidelargegassuppliesandgalaxy scaletorques.Thereishoweverlittleobservationalevidenceforastrongconnectionbetween AGNandmajormergers.WeanalysethemorphologicalpropertiesofAGNhostgalaxiesasa functionofAGNandhostgalaxyluminosityandcomparethemtoacarefullymatchedsample of control galaxies. AGN are X-ray selected in the redshift range 0.5 < z < 0.8 and have luminosities41 (cid:46) log(L [erg/s]) (cid:46) 44.5.‘FakeAGN’aresimulatedinthecontrolgalaxies X by adding point sources with the magnitude of the matched AGN. We find that AGN host andcontrolgalaxieshavecomparableassymetries,Sersicindicesandellipticitiesatrestframe ∼950nm.AGNhostgalaxiesshowneitherhigheraverageasymmetriesnorhigherfractionsof verydisturbedobjects.ThereisnoincreaseintheprevalenceofmergersignatureswithAGN luminosity. At 95% confidence we find that major mergers are responsible for <6% of all AGNinoursampleaswellas<40%ofthehighestluminosityAGN(log(L [erg/s])∼43.5). X Major mergers therefore either play only a very minor role in the triggering of AGN in the luminosityrangestudiedortimedelaysaretoolongformergerfeaturestoremainvisible. Keywords: galaxies:active-quasars:general-galaxies:evolution-galaxies:interactions- galaxies:irregular 1 INTRODUCTION bancestotheirhostgalaxies.Majormergersofgalaxiesarethere- forethoughttodominatetriggeringatthehighestAGNluminosi- Supermassiveblackholes(SMBHs)arenowbelievedtobepresent ties. While there is no clear predicted break point, a transition is in the centres of most if not all massive galaxies (e.g. Kormendy likelyaroundlog(L [erg/s])=46,asarguedabove(seealsoHop- bol & Ho 2013, and references therein). While most SMBHs do not kinsetal.2013). accretelargeamountsofgas,asmallfractionofthemshowstrong signsofaccretion.TheseobjectsareknownasActiveGalacticNu- ThetopicofAGNtriggeringbecamerelevantforgalaxyevo- clei (AGN). The conditions under which SMBHs become active lution when it was discovered that super-massive black holes are remainpoorlyunderstood.AGNactivityrequiresa)theavailability notonlycommoninmassivegalaxiesbuttheirmassesalsocorre- ofeithergasorstarstofeedtheblackholeandb)aprocesstostrip latewellwiththepropertiesoftheirhostgalaxies(velocitydisper- saidmaterialofitsangularmomentum.Depositinglargeamounts sion(e.g.Gebhardtetal.2000;Ferrarese&Merritt2000;Tremaine ofgasinthecentresofgalaxiesmakesAGNactivityprobablesince etal.2002;Gu¨ltekinetal.2009,2011),stellarmass(e.g.Ha¨ring& itprovidesmaterial,aswellasanidealenvironmenttotransferan- Rix2004),centrallightconcentration(Grahametal.2001)aswell gularmomentum. as absolute magnitudes in some bands (McLure & Dunlop 2002; Differentprocessescouldprovidesuchafavourableenviron- Marconi&Hunt2003)).Theoreticalmodelspositthatmajormerg- ers of gas-rich galaxies trigger both starbursts and AGN, and the ment for AGN triggering: major and minor mergers of galaxies AGNsubsequentlyshutsdownthestarformationbydepositingen- (e.g.Silk&Rees1998;Hopkinsetal.2008;DiMatteoetal.2008; ergyintotheISM(commonlytermedAGNfeedback)andthereby Springeletal.2005),bars(e.g.Shlosmanetal.1989),closepas- establishestheM−σrelation(Hopkinsetal.2008;Somervilleetal. sages of galaxies disturbing the gravitational potential (e.g. Hop- 2008;King2003;Silk&Rees1998;DiMatteoetal.2005;Springel kinsetal.2008),coolingofgasfromthehothalo(e.g.Crotonetal. etal.2005).Althoughsomeauthorshavepointedoutthatrepeated 2006;Popeetal.2012;Fabian2012),masslossfromstellarwinds mergersofgalaxiescontainingblackholeseedsexplainthecorrela- (e.g. Davies et al. 2012), cold flows in combination with violent tionsaswell(Peng2007;Jahnke&Maccio2011;Angle´s-Alca´zar diskinstabilities(Dekeletal.2009;Bournaudetal.2011),galaxy etal.2013). scaletorques(Angle´s-Alca´zaretal.2013)aswellasaccretionof smallamountsofgasfromthehalo(King&Pringle2007). Despiteitshightheoreticalappeal,observationalevidencefor AGNofdifferentluminositiesrequirevastlydifferentamounts aconnectionbetweenmergersandAGNremainsmixed.Forcer- of accretion material. Given typical AGN lifetimes of 108yr (see tainsamplesofAGN,ratesofrecentmergersareextremelyhigh. e.g.Martini&Weinberg2001;Yu&Tremaine2002;Martini2004, This is true for local quasars that also show large FIR luminosi- andreferencestherein,althoughawiderangeofAGNlifetimesis ties-similartoULIRGs-aswellaspeculiarAGNsamplessuch possible)alowluminosityAGNoflog(L [erg/s]) = 42requires as red quasars (Canalizo & Stockton 2001; Urrutia et al. 2008). bol aslittleas2×104M ,whilealuminousAGNoflog(L [erg/s])= However,whenhostgalaxiesofAGNarecomparedtogalaxiesof (cid:12) bol 46requiresasmuchas2×108M .Thiscorrespondstoabout1% similarmasstheyshowcomparableincidencesofdisturbancesin- (cid:12) ofthetotalmassintypicalmassivegalaxies(Catinellaetal.2010; dicative of recent mergers (Kocevski et al. 2012; Cisternas et al. Saintongeetal.2011).Strippingsuchasubstantialfractionofthe 2011;Dunlopetal.2003;Boehmetal.2013;Groginetal.2005; gasmassinagalaxyoflargepartsofitsangularmomentuminthe Gaboretal.2009;Pierceetal.2010,2007).However,hostgalaxies shortlifetimeoftheAGNischallenging.Itishenceexpectedthat ofmoderatelyluminousAGNandradio-selectedAGNshowweak whileawiderangeofprocessescanleadtothetriggeringoflow mergerfeatureswithhighersurfacebrightnessesthanfoundincon- luminosityAGN,highluminosityAGNrequiresubstantialdistur- trol galaxies (Bennert et al. 2008; Ramos Almeida et al. 2012). (cid:13)c 2014RAS,MNRAS000,1–15 CANDELS:hostgalaxiesofz∼0.7AGN 3 Treister et al. (2012) studied the incidence of merger features as paper,weuseabsorptioncorrectedrest-frame0.5-8keVluminosi- afunctionofAGNluminosityandfoundthatthehighestluminos- tiesinerg/sfromXueetal.(2011).X-raysourcesarematchedto ityAGNhavehigherincidencesofmergers.However,thesample theH-bandusinga1arcsecondaperture. wasnotuniformlyselectedandnocontrolsampleswereused.Itis Fromthe4MsCDFSSample,westudyall76objectscovered thereforeuncleariftheseresultswillholdinwell-controlledstud- inCANDELSintheredshiftrange0.5<z<0.8.Additionally,we ies. rejectobjectswithsoftX-rayspectra(Γ>1,whereΓistheeffec- In this study, we examine the incidence of disturbances in- tivephotonindex)thatlackapointsourcedetectionintheF160W dicative of major mergers in AGN hosts as a function of AGN datasincethesesourcesarelikelystarbursts.Wecautionthatex- luminositycomparedtoacarefullymatchedcontrolsampleusing tremeComptonthicksourcescanbepotentiallyrejectedusingthis CANDELS (Koekemoer et al. 2011; Grogin et al. 2011) data in methodandthismethodisthereforenot100%effectiveinidenti- GOODS-S.Oneparticulargoalofthisstudyistotestthehypoth- fyingstarbursts(Juneauetal.2011).Therejectionaffectssources esisthatmergersdominatetriggeringinmoreluminousAGN.We with detections in only a single Chandra band and very low lu- selectasampleofAGNspanningawiderangeinluminositiesusing minosities (see Fig. 1), leaving a sample of 60 AGN in the field. theirX-rayemissiononly.ThemostluminousAGNinoursample TheX-rayluminositydistributionsoftheAGNaswellasthere- requireaccretingmaterialaround108M assumingstandardquasar jected starburst sources are shown in Fig. 1 (right). Note that the (cid:12) lifetimes.Weaimtoanswerthequestionofwhethermergertrig- redshiftrangechosenforthisstudyincludesaclusterofgalaxiesat gering becomes more important with increasing AGN luminosity aredshiftof∼0.75(Salimbenietal.2009;Castellanoetal.2011). and,ifso,atwhatAGNluminositymergersbecomethedominant Theclusterhasa M ∼ 3×1014M andavelocitydispersionof 200 (cid:12) mechanism.Theincidenceofmajormergerswillbeassessedus- σ∼630km/s. ingquantitativemorphologicalmeasuresthatdeterminethelevelof disturbanceinthehostgalaxy(Conseliceetal.2000).Whilemany 2.2 MatchedSample studiesrelyonhumanclassifiers(e.g.Kocevskietal.2012;Cister- nas et al. 2011; Treister et al. 2012), using quantitative measures Forthecontrolsample,weusecataloguesbyDahlenetal.(2010), enables us to detect more subtle levels of disturbance (Conselice includingphotometricdataoverawiderangeofwavelengths(Gi- etal.2000;Lotzetal.2010a,b)aswellasallowmoredetailedsta- avaliscoetal.2004;Dahlenetal.2010,see)aswellasacompila- tisticalanalysisoftheresults. tionofspectroscopicredshifts(Cristianietal.2000;Croometal. The paper is organized as follows: data reduction and anal- 2001;Dickinsonetal.2004;LeFe´vreetal.2004;Stanwayetal. ysis as well as sample selection are presented in Section 2. The 2004;Strolgeretal.2004;Szokolyetal.2004;vanderWeletal. resultsarepresentedinSection3,followedbydiscussioninSec- 2004;Dohertyetal.2005;Mignolietal.2005;Rocheetal.2006; tion 4 and summary and conclusion in Section 5. Supplemental Ravikumaretal.2007;Popessoetal.2009;Vanzellaetal.2008). information about morphological measures used and simulations TheAGNhostgalaxysampleisdominatedbymassivegalax- performed are presented in Appendix A. The cosmology used is ies, while the full galaxy sample in the same redshift range gen- H0 =70kms−1Mpc−1,ΩΛ =0.7,Ωm =0.3.Throughoutthepaper, erallycontainsmanymorelowermassgalaxies.Thisismostlyan weuseABmagnitudes. effectofdetectionprobabilitysincelowmassgalaxieshavelower mass black holes which makes detection at equal Eddington rate lesslikely(Airdetal.2012).ThehistogramsofAGNhostgalaxies andtheparentcontrolsampleareshowninFig.2.Fromthisparent 2 DATA&ANALYSIS controlsample,wecreateacontrolsamplebymatchingbetween5 Forthestudy,weusetheF160W/HbandimagingdatainGOODS- and25galaxiestoeachAGN(Fig.4). ControlgalaxiesarematchedinabsoluteF160W(host)galaxy South(Groginetal.2011).Thedatareductionisdescribedinde- magnitude and stellar mass from Santini et al. (2012) as well as tailinKoekemoeretal.(2011).Throughoutthepaper,whenciting redshift.ThestellarmassesarederivedbyfittingtheSEDsofthe magnitudes, we refer to the observed frame F160W band, which correspondstoarest-framewavelengthof∼950nmintheredshift AGNwithamixtureofgalaxyandQSOtemplates,foramorede- taileddescriptionoftheprocess,wereferthereadertoSantinietal. rangestudied. (2009)andSantinietal.(2012).Thestellarmassistracedrather wellbytheHbandmagnitude(Fig.3)withascatterof∼0.25dex instellarmass.Someofthisscatterinstellarmassisexplainedby 2.1 AGNSample differencesinstarformationhistories(i.e.starburstage,reddening). ThesampleisselectedfromtheChandraDeepFieldSouth(CDFS) Initialmatchingisperformedbyselectinggalaxieswith∆z= 4Msdata(Xueetal.2011).X-rayselectionprovidesaminimally 0.05and∆m = 0.1,stellarmassesarematchedwithin10%.Ifno biasedsampleoverawiderangeofluminosities.Itisleastaffected sufficient number of matches are found, the criteria are relaxed. byobscurationandatthegreatsensitivityofthedataused,com- As can be seen in Figure 4, most galaxies are matched isotropi- pletedowntolowAGNluminosities.Itshouldhoweverbenoted callyinredshiftandmagnitude.Atlowergalaxymagnitudes,AGN that X-ray selection, even with the great depth of the 4Ms data hostscanbematchedtocontrolgalaxieswithrelativelylargemag- willmissthemostCompton-thickAGN.However,thesesystems nitude differences (∼ 0.5∆mag). Additionally, two host galaxies areexpectedtoberare(seee.g.Juneauetal.2011,andreferences are amongst the most luminous and massive galaxies in the field therein).Figure1(left)showsthefull4Mssampleaswellasthe andredshiftrange(seeFig.4).TheseAGNhostsarematchednon- wavelengthrangechosenforthisstudy.Thechosenredshiftrange isotropically to control galaxies of lower mass. While this is not coversamaximumamountofdynamicalrangeinX-rayluminosity optimal, we simply lack appropriate optimally matched galaxies whilenotcoveringtoolargearedshiftrangetohavesignificantcos- for these AGN. However, we find that the morphological proper- mologicalevolutionwithinthesample((cid:46)2Gyrincosmictime). ties(inparticularasymmetryandellipticity)donotshowastrong Surface brightness dimming effects are minimal. Throughout the trend with absolute galaxy magnitude. Additionally, a large per- (cid:13)c 2014RAS,MNRAS000,1–15 4 Villforthetal. log(L ) Bolometric 43 44 45 46 46 AGN 47 20 Starbursts (Rejected) 45 KDE (AGN) 46 KDE (Starbursts) 44 s ct15 log(L)Xray4432 4445log(L)Bolometric ber of obje10 43 m 41 Nu 5 42 40 CDFS 4 Ms This Sample 41 39 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 41 42 43 44 Redshift log(LXray) Figure1.Basicpropertiesofthesample.Leftpanel:AbsorptioncorrectedX-rayandbolometricluminositiesofX-raysourcesintheCDFS4MsCatalogue (greydots),thesampleusedinthispaper(redcircles).Rightpanel:Histogramsandkerneldensityestimators(KDE)ofbolometricluminositiesintheredshift rangeforallX-ray-sourcesused,objectsrejectedasstarburstsareshownascross-hatched.Inbothpanels,thedottedredlinesshowtheredshiftrangeused,the dashedredlineshowsabolometricluminosityoflog(Lbol[erg/s])=45.5.ThisistheluminosityabovewhichmergersarethoughttodominateAGNtriggering (Somerville&Primack1999;Somervilleetal.2008;Hopkinsetal.2013). centageofobjectsarecloselymatched.Ourresultsholdwhenomit- tingthetwoAGNwiththemostanisotropicmatching.TheAGN 0.8 withanisotropicmatchingdonotbelongtothebinwiththemost AGN Host Galaxies luminousAGN. 0.7 All Galaxies in Redshift Range Objectsforwhichstellarmassesarenotavailablearematched inHbandgalaxymagnitudeinstead(seeSection2.3foradescrip- 0.6 tionoffitsdescribedtoderiveAGNhostmagnitudes).ForAGNlo- catedintheclusteratz∼0.75,werejectallcontrolgalaxiesthatare 0.5 outside the clusters redshift range. Since the cluster environment can have a strong influence on galaxy morphological parameters 0.4 (Dressler1980;Adamsetal.2012),matchingAGNhoststocon- trolgalaxieslocatedinverydifferentenvironmentscouldintroduce 0.3 biases.Theresultsofourstudyholdwhenmatchingisperformed purelyinHbandmagnitude. 0.2 0.1 2.3 HostGalaxyFits 0.0 GalaxyfitsareperformedusingGalfit(Pengetal.2002).Empiri- 17 18 19 20 21 22 23 24 M F160W calPSFswerederivedfromstackingtheimagesofseveralisolated andunsaturatedstarsinthefield.Inordertoprovideamoreaccu- Figure2.HistogramsshowingthedistributionsoftheAGNHostgalaxies rate description of the central region, we replaced the inner-most (red)aswellasallgalaxiesintheredshiftrangeofthestudy(hatched). pixels (within a radius of 3 pixels from the center) with a simu- Notethatsincethehistogramsarebothnormalizedtointegratetoone,this latedPSFgeneratedwiththeTinyTimpackage(Krist1995).The histogramdoesnotrepresentthetotalnumberofgalaxiesinbothsamples. TinyTimPSFwasditheredanddrizzledinthesamemannerasthe observations,andnormalizedsuchthatthetotalfluxofthenewly constructed hybrid PSF model is the same as that of the stacked star. We found this hybrid PSF accurately reproduced the growth cases, the fits diverged for a single Sersic component or resulted curvesofstarsoutto3”.FurtherdetailsonthePSFmodelscanbe in strong residuals indicative of the presence of a component not foundinvanderWeletal.(2012). accountedforinthefits.Sinceweusethepoint-sourcesubtracted FortheAGNhostgalaxies,weuseamixtureofpointsource imagesratherthantheresidualimagesforfurtheranalysis,details andSersiccomponent(withbothellipticityandradiusleftasafree inthehostgalaxyfitsthemselvesdonotstronglyaffectourresults. parameter). If necessary, a second Sersic component is added. In The image resolution of 0.1” corresponds to about 0.5 kpc most cases, a fit with a point source and single Sersic yielded a intheredshiftrangeofoursample.Acompactbulgeorstarburst good fit with minimal residuals. The goodness of fit was judged mightthereforebefitasapointsource.ForallAGNthatrequire both based on the χ2 values given by Galfit and visual inspec- pointsourcefits,wethereforecheckthecolourofthecentralpoint tions of the residuals. It was only necessary to add a second ser- sourceusingACSimages.Coloursofcentralpointsourcesareblue, sic component in four cases out of the 60 AGN studied. In these consistentwithAGN.Starburstshoweverhavesimilarcolours.We (cid:13)c 2014RAS,MNRAS000,1–15 CANDELS:hostgalaxiesofz∼0.7AGN 5 addedtoitsmatchedcontrolgalaxies.The’fakeAGN’arethenfit withapointsourceandSersicmodel.Amoredetaileddiscussion 13 oftheinfluencesofresidualsonmorphologicalparameterscanbe foundinAppendixA. 12 ) ] 11 2.4 QuantitativeMorphologyMeasures:Asymmetry fl M [ Forquantitativemorphologymeasurements,weuseAsymmetryA Mstellar10 wclhasicshifiicsaptiaorntosfysthteemCA(CSo(nCseolmicpea2c0tn0e3s)s.,TAessytsmomneotruyr,Sdmatoaosthhonwesesd) ( thattheconcentrationCwasmostsensitivetoPSFresidualssince og 9 itworksbymeasuringhowcentrallyconcentratedgalaxiesare,in l agreement with previous studies (Grogin et al. 2005). Due to the 8 limited resolution of our data, Smoothness S is not used. Asym- Control Sample metryisfoundtotracemajorandminormergerswell(Lotzetal. AGN Host Galaxies 2010a,b).Additionally,wewillusetheellipticityfromthegalfit 716 18 20 22 24 26host galaxy fits. Relaxed post-merger systems could show lower M ellipticitiesthangalaxiesofsimilarmassthathavenotundergone F160W amergerrecently.Adetaileddescriptionofthemorphologymea- Figure3.ComparisonbetweenstellarmassesandabsoluteHbandmagni- suresandpossibleerrorsduetoPSFresidualscanbefoundinAp- tudeforfullcontrolsample(greydots,allgalaxiesingivenredshiftrange) pendixAbelow.Hereweprovideabriefoverview. andAGNhostgalaxies(redfilledcircles). AsymmetryAisdefinedas: (cid:115) (cid:80)1 ×(I −I )2 A≡ 2 (cid:80)0 180 (1) I2 26 0 whereI isthefluxineachpixelandI isthefluxineachpixel 0 180 24 rotatedby180deg(Conselice2003).DifferentfromConseliceetal. (2000)andConselice(2003),wedonotsubtractbackgroundasym- metry.However,asshowninConseliceetal.(2000),thiswillhave y) 22 littleeffectforthetypicallybrightgalaxiesusedinthisstudy.For x a thepurposeofthisstudy,weuseSextractorsegmentationmapsto l Ga avoidincludingnoisefromthebackgroundintothemeasurement. ( 20 Thesemapsdeterminetheregionoverwhichthegalaxyisdetected. W F160 Inthefollowing,wewilldiscussdifferentinfluencesonthemea- M 18 suredasymmetry.CentringisperformedfollowingConseliceetal. (2000). We have ensured that the algorithm generally reaches a Control Galaxies well-definedminimum.VisualinspectionofallAGNhostsanda Matched Control Galaxies 16 AGN Host Galaxies: Stellar Mass matched randomlychosensubsetofthecontrolsampleisperformedtomake AGN Host Galaxies: Magnitude matched surethecentralpointdeterminedbythealgorithmdeterminesthe centerofthegalaxycorrectly.Threegalaxieswithdifferentlevels 0.50 0.55 0.60 0.65 0.70 0.75 0.80 ofasymmetryareshowninFig.5asareferenceforthereader. Redshift Central pixels are in some cases affected by PSF residuals. Figure4.MatchingperformedforallAGN,theblackdotsshowthefull Whilesimulating’fakeAGN’ensuresthatthisisalsothecasefor controlsample.TheredstarsmarktheAGN,thegreendotsshowgalaxies the control sample, we still do not wish these pixels to dominate thathavebeenmatchedascontrolgalaxies,thegreenlinesconnectthem theoverallasymmetrymeasurement.Hence,thecentralareaofall totheAGNtheyhavebeenmatchedto.Notetheclusterataredshiftof objects is masked using a circular aperture with a radius of two z∼0.75. pixels.Theexactsizeofthemaskdoesnotchangetheoverallresult andvisualinspectionsshowthatfortheAGNmagnitudescommon inoursample,theseaperturesizescoverthecorruptedpixelswhile find that contributions of central point sources in low-luminosity notmaskinguncontaminatedareasofthehostgalaxy.Moredetails AGNareweak.ThereisthepossibilitythatAGNcontributionsare canbefoundinAppendixA. overestimated at low X-ray luminosity, however, our results hold In order to keep asymmetry measures comparable indepen- whenomittingAGNwithlog(L <42)andthereforethisdoesnot dent of optical AGN magnitude, we use the same mask size for X affecttheoverallconclusionsofthisstudy. allobjects,independentofAGNmagnitude.Whilethisalsomasks AllControlgalaxiesarefitwithasingleSersiccomponentand the central areas of galaxies not affected by PSF residuals, it en- nopointsourcetodeterminetheirSersicindicesaswellaselliptici- suresthatthesamephysicalareasareusedforthecalculationofthe ties.TomimicresidualsfromPSFfittingperformedforAGNhosts, asymmetryforallsources.Ideally,themasksizewouldbeadjusted wecreatefakeAGNforthemorphologicalanalysis.Pointsources tothesizeofthegalaxy.However,sinceAGNhostsandcontrolare withmagnitudesmatchedtothecorrespondingAGNmagnitudeare carefullymatched,thisshouldnotresultinabiasesmeasurement. (cid:13)c 2014RAS,MNRAS000,1–15 6 Villforthetal. High Asymmetry (A>0.1) Moderate Asymmetry (A~0.05) Low Asymmetry (A~0.01) Figure5.ThreeexampleimagesshowingdifferentlevelsofasymmetryinthreeAGNhostgalaxies.Allimagesareonalogarithmicscaleand4.2”×4.2”in size. 2.5 Humanclassifiers We compare the distribution of Sersic indices for the AGN andcontrolsample(Fig.6).Twosampletestsshownostatistically We use results from human classifiers for comparison and con- significantdifferencebetweenAGNhostgalaxiesandthecontrol sistencychecks.ThegalaxieswereclassifiedbyCANDELSteam sample.Asexpected,Sersicindicesarehigherformoreluminous membersintheH/F160Wband.Bluerbandsarealsoinspectedto hostgalaxies,thereishowevernostrongtrendwithX-rayluminos- facilitate the classification (Kartaltepe et al. 2012, , Kartaltepe et ity. al.inprep).TheboundariesofthegalaxiesaredefinedusedSex- tractorsegmasks.Inafirststep,theclassifierisaskedtodecide Next, we analyse differences between the asymmetries of AGN host and control galaxies (Fig. 7). The overall distribution between four main morphological classes: spheroid, disk, irregu- lar/peculiarandcompact/unresolved.Additionally,thehumanclas- of the asymmetries for the two samples show no significant dif- ferences. We also find no trend with absolute galaxy magnitude. sifier is asked to decide if the galaxy meets any of the following We perform two- sample Kolmogorov-Smirnov (KS) and Mann- interaction classes: merger (a clearly interacting system with dis- Whitney U (MW) tests between the AGN and control galaxies. turbances), interaction (interaction of two distinct objects within Thesetestsarecalculatedforboththefullsampleandsub-samples or beyond the segmentation mask) or non-interaction companion binned in both X-ray luminosity and absolute galaxy magnitude. (galaxyhasacompanionthatisnotclearlydisturbed).Thisclassi- The results are shown in Table 1. The two-sample Kolmogorov- ficationschemeisthesameasusedinKocevskietal.(2012). Smirnov test yields p=0.38 and the Mann-Whitney U test yields p=0.09.Wehencefailtorejectthenullhypothesisthatthetwosam- 2.6 Noteonconfidenceintervalsused plesaredrawnfromthesameparentpopulation.Whenbinnedin X-rayandgalaxyluminosity,theKolmogorov-SmirnovandMann- BinomialprobabilitiesandtheirconfidenceintervalsfortheAGN WhitneyUtestfindnosignificantdifferencesbetweenthesamples. samplearederivedusingabetastatistic(Cameron2011).Asstated Weonlyfindp(cid:54)0.05inasinglebin.However,giventhemultiple inSection2.2,severalcontrolgalaxiesarematchedtoeachAGN tests(28)performed,thisisconsistentwiththeexpectednumberof to increase statistical power. Due to the fact that the number of falsepositives(1.4). matchesdiffersbetweenAGN,confidenceintervalsforthecontrol Whilethestatisticaltestsrevealnosignificantdifferencebe- sample cannot be derived straightforwardly using the beta statis- tweentheasymmetries,wedonotethatbyeyethedistributionof tics since it does not account for weighting. We therefore use a asymmetriesappearstobeslightlymoreskewedwithalargerhigh jackknife method in which we randomly choose a subset of five asymmetrytailintheAGNhostgalaxieswhencomparedtocon- matchedcontrolgalaxiesforeachAGNandthenderivetheresult- trol. To determine if these differences are quantifiable, we calcu- ingvalueforthebinomialprobability.Thisrandommatchingisre- latethemomentsofthedistributionsofasymmetriesfortheAGN peated100timesandtheconfidenceintervalsarederivedfromthe andcontrolgalaxies.Ajackknifemethodisusedtodeterminethe finaldistribution.Jackknifemethodsarealsousedtodeterminethe typicalscatterinthemomentsderivedforthecontrolsample.We expectationvaluesanderrorindistributionmomentsforthecontrol calculatethefirstfourmoments(mean,standarddeviationσ,skew sample. γ andkurtosisκ).Forthefullsample,wefindthatboththeskew andkurtosisarehigherfortheAGNsamplecomparedtocontrol. Bothlieabovethethirdquartileofthecontrolsampledistribution. 3 RESULTS:MORPHOLOGIESOFAGNHOSTS ThisimpliesthatthedistributionfortheAGNhostshasatailto- Wenowcomparethemorphologicalproperties(Sersicindices,el- wards larger values of the asymmetry A and a higher peak with lipticitiesandasymmetries)oftheAGNhostgalaxiestothoseof morepowerinthetails.However,whendividingthesampleinto thecontrolgalaxies.Theaimistodetermineifthedataareconsis- sub-binsineitherAGNorhostgalaxyluminosity,thesamplesizes tentwiththenullhypothesisthatAGNhostsaredrawnrandomly aretoosmalltodetermineifthereisatrendinthemoments.The from the sample of their control galaxies. Note that when plot- fullcalculatedpropertiesarelistedinAppendixB. tingX-rayluminosity,weplottheX-rayluminosityofthematched AnotherwaytocomparetheasymmetriesofAGNhostsand AGNforthecontrolgalaxies. controlgalaxiesistocomparethevaluesforeachsingleAGNto (cid:13)c 2014RAS,MNRAS000,1–15 CANDELS:hostgalaxiesofz∼0.7AGN 7 itssampleofcontrolgalaxies.Wethereforeanalysethepercentiles Table 1. Results from 2 sample Kolmogorov-Smirnov and Mann- ofscoresofeachAGNhostgalaxyasymmetrywithrespecttoits Whitney-Utests comparingasymmetries ofAGN hostsand matched matchedcontrolgalaxies.ForeachAGN,wecalculatethecumu- control samples. Measure: morphological measure used for compari- lative density function of the asymmetries of its control sample. son;secondcolumns:AGNpropertyusedforbinning;p(KS):p-value Fromthisdistribution,wethencalculatethepercentileatscorefor forKolmogorov-Smirnovtests.p(MW):p-valueforMann-Whitney-U eachAGN.Forexample,anAGNwithanasymmetryequaltothe test. medianasymmetryofitscontrolsamplewillhaveapercentileat BinProperty BinMean p(KS) p(MW) scoreof50%,whileanAGNhavingasymmetryhigherthanallits matchedcontrolgalaxieswillhaveapercentileatscoreof100%. log(LXray)(All) 42.27 0.34 0.09 If the AGN hosts were drawn randomly from the sample of its log(LXray) 41.18 0.85 0.21 matched hosts, the distribution of percentiles at score for the full log(LXray) 41.59 0.18 0.07 AGNsampleshouldbeflatsinceeachAGNisequallylikelytobe log(LXray) 41.91 0.56 0.27 sampledateachpercentileatscore.Weshowthekerneldensityes- log(LXray) 42.33 0.66 0.40 timator(KDE)ofthepercentilesatscoreforthefullsampleinFig. log(LXray) 43.04 0.10 0.12 8.Thedistributionisnotflat,showingexcessatlowpercentilesat log(LXray) 43.60 0.85 0.25 score as well as a small excess at very high percentiles at score. This is consistent with the eye ball assessment that the asymme- MHost(All) -22.53 0.38 0.09 MHost -24.02 0.76 0.43 triesappearonaveragesomewhatlowerintheAGNcomparedto MHost -23.10 0.37 0.23 control while showing a slight high-asymmetry excess. However, MHost -22.75 0.85 0.38 duetosmallnumberstatistics,thedifferencesarenotstatistically MHost -22.32 0.37 0.18 significant(p=0.098). MHost -21.94 0.18 0.05 An additional factor in asymmetry levels might be obscura- MHost -21.06 0.85 0.43 tion. While it is widely acknowledged that obscuration in AGN inthelocalUniverseismostlyduetoadustytorusandtherefore AllCluster – 0.0049 0.0098 AllField – 0.66 0.41 primarily a function of AGN orientation (Antonucci 1993; Urry &Padovani1995),someveryyoungAGNmightbeinanearlier ObscuredvsUnobscured – 0.31 0.17 buried phase in which the obscuration is due to dust in the host galaxy and not the torus. We note that obscuration in the optical and the X-ray is not necessarily tracing the same obscuring ma- terial.Suchasub-sampleofyoungobscuredAGNmightbemore 0.020 closelyconnectedtomergers.TheasymmetrydistributionsofAGN with X-ray effective photon indices Γ < 1 (X-ray obscured) and Γ > 1 (X-ray unobscured) are compared in Fig. 9. There are no differencesbetweenthetwoAGNsub-sampleasymmetries(Table 0.015 1),indicatingthathigherlevelsofobscurationdonotleadtocom- parativelylargerasymmetries. ) As mentioned earlier, the redshift range studied contains a (x0.010 cluster of galaxies at z≈0.75 (Salimbeni et al. 2009; Castellano p et al. 2011). We separately compare the host galaxies and their matched control galaxies in both the cluster and field (Fig. 10). 0.005 Wefindthat-as forthefullsample-theAGNhostsin thefield areconsistentwithbeingdrawnrandomlyfromthecontrolgalax- ies. However, in the cluster, we do find a statistically significant differencebetweentheAGNandcontrolgalaxies(P<0.01)with 0.0000 20 40 60 80 100 AGNhostshavinglowermeanasymmetriesbuthigherskewinthe Percentile at Score asymmetrydistributionwhencomparedtocontrolgalaxies. Figure8.PercentiledistributionofAGNasymmetrieswithrespecttotheir While asymmetry traces levels of disturbance in the host controlsample.ForeachAGN,wecalculatethepercentileatscorewith galaxy, more relaxed mergers will not be identified by this index respecttoitsmatchedcontrolgalaxies.TheKDEisreflectedoffthebound- (for example mergers between high ellipticity disk galaxies will aries.IftheAGNhostasymmetriesweredrawnrandomlyfromthesame generallyreducetheellipticityinthemergedsystem).Wethusad- distributionasthematchedcontrolgalaxiesasymmetries,thedistribution ditionallycomparetheellipticitiesfromthegalfitgalaxyfitsbe- shouldbeflat(indicatedbythedashedredline).Thedifferenceishowever tweenAGNhostsandmatchedcontrols.Theresultsareshownin oflowstatisticalsignificance(p=0.098). Fig. 11. There are no statistically significant differences between AGNhostgalaxiesandcontrolgalaxies. sonablevalueabovewhichallgalaxiesshowcleardisturbance(see alsoFig.5).Theseratesarecomparedtodifferentmeasuresfrom 3.1 Comparisonbetweenquantitativemeasuresandhuman thehumanclassifierresultsdescribedinSection2.5inFigure12. classifiers In particular, we compare two different classifications, ir- In addition to the distributions of asymmetries, we also compare regularityandmerger.Irregularityencompassesallobjectsshow- the probabilities of objects having high (A>0.1) asymmetries be- ing some disturbance or irregularity, even if they show a well- tweenAGNhostgalaxiesandcontrolsamples.Thecut-offA>0.1 pronounceddiskorspheroidcomponent.Theclassifiersareasked issomewhatarbitrary,butvisualinspectionshowsthistobearea- toclassifyobjectsasirregulariftheyseeasymmetricfeatures,not (cid:13)c 2014RAS,MNRAS000,1–15 8 Villforthetal. 6 6 All AGN All AGN 5 Matched Control 5 Matched Control ex4 ex4 d d n n I I 3 3 c c i i s s r r e2 e2 S S 1 1 0 0 41 42 43 44 18 19 20 21 22 23 24 25 26 log(LXray[erg/s]) MF160W Figure6.SersicindexasafunctionofAGNLuminosity(left)andAGNHostmagnitude(right)forAGN(redcircles)andcontrol(greydots).Forcontrol galaxies,theX-rayluminosityofthematchedAGNisplotted.ProjectedhistogramsareshownfortheAGN(red)andcontrolsample(hashed).Thereareno statisticallysignificantdifferencesbetweenthetwosamples. 0 0 1 1 A) A) g( g( o o l l 2 2 Matched Control Matched Control All AGN All AGN 3 3 41 42 43 44 18 19 20 21 22 23 24 25 26 log(LXray[erg/s]) MF160W Figure7.AsymmetryasafunctionofXrayluminosity(left)andAGNhostgalaxymagnitude(right)forAGN(redcircles)andcontrol(greydots).Forcontrol galaxies,theX-rayluminosityofthematchedAGNisplotted.ProjectedhistogramsareshownfortheAGN(red)andcontrolsample(hashed).Thereareno statisticallysignificantdifferencesbetweenthetwosamples. takingintoaccountiftheobjectappearstobeinamergerornot. matchedcontrol.Wealsofindthatusingacut-offA>0.1leadsto ThisclassificationisthereforecomparabletotheasymmetryA.Ad- similaroverallasymmetry/irregularityratesasclassificationbyhu- ditionally,weusethemergerclassification.Classifiersareaskedto mans. identifyanyobjectsappearingtoundergointeractionorshowinga Wefindtheratesofhuman-classifiedmergersaresignificantly nearbycompanion.Assuch,thiscategoryismorepronetosubjec- higherintheAGNhostswhencomparedtomatchedcontrol.The tiveinterpretationsbytheclassifiersandisexplicitlynotcompara- excessratesforAGNcomparedtocontrolseeninthemergerrates bletotheasymmetry. areduetocompanionsshowingnosignsofmerger-nottrain-wreck mergersorevenmergersshowingsomesignsofinteraction.Asub- Asexpectedfromthesimilaritiesintheoveralldistributions, stantialfractionoftheseobjectsisassociatedwiththeclusterenvi- theratesofobjectswithhighasymmetries(A>0.1)arenothigher ronmentatz∼0.75. intheAGNhostscomparedtocontrol.Whenstudyingtheratesas a function of X-ray luminosity, no differences between the AGN hosts and control are found. As a function of galaxy magnitude, 4 DISCUSSION thereisamildexcessinhighasymmetryratesatmoderategalaxy masses;withAGNhostshavingslightlyhigherasymmetries.How- We have compared the morphologies of X-ray selected AGN to ever,thisisnotstatisticallysignificant.Similarly,wefindthatthe thoseofmatchedgalaxiesofthesamemass.TheAGNstudiedare rates of objects classified as irregular by human classifiers show atredshiftz=0.5-0.8andhaveX-rayluminosities(log(L [erg/s])≈ X no statistically significant difference between the AGN hosts and 41−44.5).Assumingstandardbolometriccorrections(Runnoeetal. (cid:13)c 2014RAS,MNRAS000,1–15 CANDELS:hostgalaxiesofz∼0.7AGN 9 0 0 1 1 A) A) g( g( o o l l 2 2 Control Control Cluster AGN Field AGN 3 3 41 42 43 44 45 41 42 43 44 45 log(L [erg/s]) log(L [erg/s]) Xray Xray Figure10.Asymmetryinthecluster(left)andfield(right)asafunctionofXrayluminosityforAGN(redcircles)andcontrol(greydots),showingdifferences betweentheclusterandfield.Forcontrolgalaxies,theX-rayluminosityofthematchedAGNisplotted.ProjectedhistogramsareshownfortheAGN(red)and controlsample(hashed).Therearenostatisticallysignificantdifferencesbetweenthetwosamplesinthefield,however,inthecluster,thedifferencebetween AGNhostsandcontrolisstatisticallysignificant(p<0.01). 0.9 0.9 0.8 0.8 0.7 0.7 y y t t i0.6 i0.6 c c i i t0.5 t0.5 p p li0.4 li0.4 l l E0.3 E0.3 0.2 0.2 All AGN All AGN 0.1 0.1 Matched Control Matched Control 0.0 0.0 41 42 43 44 18 19 20 21 22 23 24 25 26 log(LXray[erg/s]) MF160W Figure11.EllipticityasafunctionofAGNLuminosity(left)andAGNHostmagnitude(right)forAGN(redcircles)andcontrol(greydots).Forcontrol galaxies,theX-rayluminosityofthematchedAGNisplotted.ProjectedhistogramsareshownfortheAGN(red)andcontrolsample(hashed).Thereareno statisticallysignificantdifferencesbetweenthetwosamples. 2012;Nemmen&Brotherton2010),theAGNluminositiesspana asymmetriesaswellasahigherskew,indicativeofastrongerdis- rangeofaboutfourordersofmagnitude. tributiontailathighasymmetries.Additionally,morphologicalin- spections by human classifiers show that AGN hosts have higher The host galaxies of moderately luminous X-ray selected ratesofcompanions,althoughtheseshownosignsofinteraction. AGN at low redshift show no strong differences in morphologi- ThisdifferenceisduetoAGNfoundintheclusterenvironment. calproperties(asymmetries,Sersicindices,ellipticities)compared togalaxiesmatchedinmassorabsoluteHbandmagnitude.There Beforecomparingourresultstothoseofotherauthors,wewill isnostatisticallysignificantdifferencebetweenAGNhostgalaxies brieflyaddressthesurfacebrightnesslimitanditsinfluenceonthe andcontrolgalaxiesintheratesofveryhigh(A>0.1)asymmetry types of merger features detected. We reach surface brightnesses galaxies or the rates of galaxies classified as irregular by human around26mag/arcsec2inourstudy.Generally,thebrightestmerger classifiers.Theasymmetrydistributionsandhighasymmetryrates features have surface brightnesses as high as 22 mag/arcsec2, alsodonotshowgreaterdifferencesbetweenAGNhostsandcon- whilemostfaintertidaltailshavesurfacebrightnessesdownto25 trolasafunctionofeitherabsolutegalaxymagnitudeorAGNX- mag/arcsec2 (Elmegreen et al. 2007). Merger features are found rayluminosity.Therearenodifferencesbetweentheasymmetries even in local red and dead elliptical galaxies when depths of 28 ofX-rayobscuredandunobscuredAGN.Theonlystatisticallysig- mag/arcsec2 are reached (van Dokkum 2005). However, at these nificantdifferencefoundbetweenAGNandcontrolgalaxiesisthat extremelyfaintsurfacebrightnesslevels,mergerfeaturesarealso AGN host galaxies located in a cluster environment show lower likely to trace disruption of small satellites (e.g. Johnston et al. (cid:13)c 2014RAS,MNRAS000,1–15 10 Villforthetal. 100 100 Control Control AGN Asymmetry AGN Asymmetry 75 75 % 50 % 50 25 25 0 0 100 100 Control Control AGN Irregularity AGN Irregularity 75 75 % 50 % 50 25 25 0 0 100 100 Control Control AGN Merger AGN Merger 75 75 % 50 % 50 25 25 0 0 41.0 41.5 42.0 42.5 43.0 43.5 20.5 21.0 21.5 22.0 22.5 23.0 23.5 24.0 24.5 log(L ) M Xray F160W Figure12.RatesofdisturbancesinAGNhostgalaxiescomparedtocontrol.TheleftpanelshowtheratesasafunctionofX-rayluminosity,rightpanelasa functionofhostgalaxyabsolutemagnitude.Thetoppanelisfromquantitativeasymmetrymeasurements,the2bottompanelrowsarefromhumanclassifiers, inparticular,weshowirregularity(allgalaxiesshowingsomeirregularfeatures)andmergers(galaxieseithershowingclearsignsofinteractionorhaving close-byneighbours).Over-plottedare68.75%(1σ)confidenceintervals. 2001).Giventhelimitsofourdata,wearethereforelikelytodetect tically significant results differences (Dunlop et al. 2003; Grogin ongoing mergers, as well as tidal tails, while the data is not sen- etal.2005;Boehmetal.2013;Cisternasetal.2011).Dunlopetal. sitive enough to detect disruption of smaller satellites. However, (2003) studied samples of radio-loud and radio-quiet quasars at sinceweareprimarilyinterestedintheincidenceofrecentmajor z<0.25 with luminosities comparable to those in our study. Ko- mergers(mergerswithgalaxymassratios(cid:62)0.1lessthan∼1Gyr cevskietal.(2012)analysedhumanclassificationsofhostgalax- inthepast),thislimitisidealforthesciencegoalofourstudy.The ies of X-ray selected AGN at z≈2 with slightly higher luminosi- spatialresolutionofourstudyis∼0.5kpc.Wemightthereforemiss tiesthanthosestudiedinthispaper.Cisternasetal.(2011)exam- asymmetricfeaturesonverysmallscales. inedthehostsofz∼1X-rayselectedAGN.Allusedhumanclassi- fiers to determine the incidence of merger features and found no We will start by comparing our results to previous studies differences when compared to control. Grogin et al. (2005) and of AGN host galaxy morphologies compared to control samples Boehmetal.(2013)bothusedquantitativemorphologymeasures of quiescent galaxies. A number of studies have found no statis- (cid:13)c 2014RAS,MNRAS000,1–15