Mon.Not.R.Astron.Soc.000,1–21(2007) Printed25May2007 (MNLATEXstylefilev2.2) Environmental dependence of AGN activity in the supercluster A901/2 R. Gilmour1,2? , M. E. Gray3, O. Almaini3, P. Best1, C. Wolf4, K. Meisenheimer5, 6 5 C.Papovich and E. Bell 1ScottishUniversitiesPhysicsAlliance,InstituteforAstronomy,RoyalObservatory,BlackfordHill,Edinburgh,EH93HJ,UK 2EuropeanSouthernObservatory,AlonsodeCordova3107,Vitacura,Casilla19001,Santiago19,Chile 3SchoolofPhysicsandAstronomy,UniversityofNottingham,UniversityPark,Nottingham,NG72RD,UK 4DepartmentofPhysics,DenysWilkinsonBldg.,UniversityofOxford,KebleRoad,Oxford,OX13RH,UK 5Max–Plank–Institutfu¨rAstronomie,Ko¨nigstuhl17,D-69117Heidelberg,Germany 6StewardObservatory,TheUniversityofArizona,933NorthCherryAvenue,Tuscon,AZ85721,USA Accepted.Received;inoriginalform ABSTRACT WepresentXMMdataforthesuperclusterA901/2,atz ∼0.17,whichiscombinedwith deepimagingand17-bandphotometricredshifts(fromtheCOMBO-17survey),2dFspectra andSpitzer24µmdata,toidentifyAGNinthesupercluster.The90ksecXMMimagecontains 139 point sources, of which 11 are identified as supercluster AGN with LX(0.5−7.5keV) > 1.7×1041erg cm−2s−1. The host galaxies have M < −20 and only 2 of 8 sources with R spectra could have been identified as AGN by the detected optical emission lines. Using a largesampleof795superclustergalaxieswedefinecontrolsamplesofmassivegalaxieswith nodetectedAGN.ThelocalenvironmentsoftheAGNandcontrolsamplesdifferat>98per ∼ cent significance. The AGN host galaxies lie predominantlyin areas of moderate projected galaxydensityandwithmorelocalbluegalaxiesthanthecontrolsample,withtheexception ofoneverybrightTypeIAGNverynearthecentreofacluster.Theseenvironmentsaresimilar to,butnotlimitedto,clusteroutskirtsandbluegroups.Despitethelargenumberofpotential hostgalaxies,noAGNarefoundinregionswiththehighestgalaxydensity(excludingsome cluster coreswhere emission from the ICM obscuresmoderateluminosity AGN). AGN are also absent from the areas with lowest galaxy denstiy. We conclude that the prevalence of clusterAGNislinkedtotheirenvironment. Keywords: Galaxies:clusters:individual:A901/2-Galaxies:active 1 INTRODUCTION etal. (2003)findthesameresultforstar-formationrate.Wolfetal. (2005)findthatdustystar-forminggalaxiesaregenerallyfoundin Thepropertiesandevolutionofgalaxiesareknowntobestrongly moderatedensityenvironments. linked to their external environment. In particular, populations of galaxiesinclustersarestrikinglydifferenttothoseinthefield,as Variousprocesseshavebeensuggestedtoaccountfortherapid shownbythemorphology–densityrelation(Oemler1974,Dressler transformationofclustergalaxies,fromlocaleffectssuchasmerg- 1980)anddramaticchangesinstar-formationrates(e.g.Lewisetal. ers(e.g.Mihos&Hernquist1996)andrepeatedclosegravitational 2002 and Go´mez et al. 2003). It appears that galaxies change encounters(e.g.Mooreetal. 1998) tothelargescaleeffectslike significantlyastheyjoindenserenvironmentssuchasgroupsand the tidal field (e.g. Byrd & Valtonen 1990) and the intra-cluster clusters. medium, viaram-pressure stripping (e.g. Abadi et al. (1999)) or Itisincreasinglyevident thatmanyofthechangesingalaxy ‘strangulation’(e.g.Larsonetal. 1980). propertiesbetweenclustercoresandthefieldaretriggeredininter- The processes which affect galaxy properties may also, di- mediatedensityenvironments,andthatadistinctionbetweenfield rectlyorindirectly,affecttheaccretionontothecentralblackhole and cluster populations is overly simplistic. For example, Wake foundinmost,ifnotall,galaxieswithastellarbulge(e.gMagor- et al. (2005) find that galaxy colour isa function of local rather rianetal. 1998).Bothlocalandlarge-scaleprocesseswhichmay thanextendedgalaxydensity,andLewisetal. (2002)andGo´mez affectclustergalaxiesalsohavethepotentialtoaffectthedistribu- tionofgasinthegalaxies,andhencemaytriggerorsuppressAGN ? [email protected] activity.Recentlylargeandmoderatesizedsurveyshavebegunto (cid:13)c 2007RAS 2 R. Gilmouret al. shed light on the local and extended environments of AGN, and Giventhatmanygalaxytransformationsoccurinintermediate produceobservationalevidenceforsomeoftheseprocesses. densityenvironments, itmaybethatAGNarealsoalteredbythe The first evidence of a suppression of AGN in the cores of hostgalaxyenvironment.TounderstandthelinksbetweenAGNac- galaxy clusters was found in the optical survey of Dressler et al. tivityandtheirextendedenvironmentitisdesirabletolookbeyond (1984), although a large source of bias was suggested by Edge galaxyclustersasasingleentity,andinsteaddeterminetheeffect (1994).Furtheropticalsurveys(suchasColdwelletal. 2002and oflocal(∼100kpc)andlargescale(∼1Mpc)environment,from Kauffmannetal. 2004)havealsofoundadeficitofluminousAGN thefieldthroughgroupsandclusteroutskirtstotheclustercores. indenseregions.HoweverMilleretal. (2003)findthatthefraction Superclustersareidealtestbedsforsuchastudyastheyconsist ofluminousgalaxieswithAGNisindependent ofgalaxydensity, ofalargenumberofgalaxiesinarangeofenvironments,butatthe a conclusion also drawn from the auto-correlations of AGN and sameepoch.ThecorrelationsbetweenenvironmentandAGNprop- galaxies presented by Wake et al. (2005). Surprisingly, some of ertiescanthereforebestudiedinonefield,withoutcomplications these studies use the same datasets, but draw contrasting conclu- duetogalaxyorAGNevolution.Forexample,theAGNpopulation sions,probablyduetodifferentAGNselectiontechniques. ingalaxygroupscanbecomparedtothatinclusteroutskirtsofsim- The picture is further complicated when detections in other ilarlocalgalaxydensitytodistinguishbetweenlocalandlargescale wavebandsareconsidered.Incontrasttotheopticalresults,many environments.Inadditionsuperclusterscontainbothdisturbedand radio studies (e.g. Best et al. 2002, Miller & Owen 2003, Barr relaxedregions,whichmayaffectAGNindifferentways. etal. 2003,Reddy&Yun2004,andBest2004)showanincrease This paper presents the results of investigations into AGN in radio-loud AGN activity in galaxy clusters, at a range of red- inthesuperclusterA901/2,whichhasextensivemulti-wavelength shiftsand in both relaxed and merging systems. Best (2004) find imagingandspectroscopy, summarisedinSection2.Thedatare- thatthemajorityofradio-loudAGNinthedensestregionsarenot ductionandidentificationofthesuperclusterAGNaredescribedin emission-linesources,andsomaybemissedbyopticalstudies. Sections3and4.ThepropertiesoftheAGNandtheirhostgalax- However, radio-loud AGN arenot representative of AGN as iesareinvestigated,alongwiththeenvironmentsinwhichtheyare awhole,andopticalstudiesarepronetoselectioneffects:studies found, and the environments of the AGN hosts are compared to of X-ray emission are an alternative method to remove some se- othersuperclustergalaxiesinSection5.Theresultsaresummarised lectioneffects, andtodetectalargerpopulationof AGN.Indeed, in Section 6. Details of individual AGN candidates are given in Martini et al. (2006) find that only four of at least 35 X-ray de- theAppendix.Throughout thepaperthecosmologicalparameters tectedAGNinasampleofclustershaveopticalspectralsignatures Ωm,ΩΛ andH0aresetto0.3,0.7and70kms−1Mpc−1,andall ofAGNactivity.ThemajorityofstudiesintheX-rayhavefocused COMBO-17absolutemagnitudesareconvertedtothiscosmology. ongalaxyclusters,whichhavetheadvantageofalargenumberof galaxiesandhighdensity,butarecomplicatedbytheX-rayemis- sionfromtheintra-clustermedium.Thistendstomaskanydetec- 2 THESUPERCLUSTERA901/2 tionsofAGNintheverycentreofthecluster,inparticularinthe 2.1 Opticaldata centralgalaxy,ifoneexists,whichisoftenamassiveellipticaland radioloud(e.g.Peresetal. 1998,Bestetal. 2006).Theexclusion ThesuperclusterconsistingofAbell901andAbell902(A901/2), of such galaxies may in fact be an advantage as they are in very firstidentifiedbyAbell(1958),isidealforastudyoftheeffectof different environments fromtheother cluster galaxies, frequently environmentonAGNduetothelowredshift(∼ 0.17)andwealth lyinginthecentreofthepotentialwell.Thisunusualenvironment ofopticaldataavailable.ItisoneofthefieldsintheCOMBO-17 probablyhasaverydifferenteffectonAGNactivitycomparedto survey(ClassifyingObjectsbyMedium-BandObservations in17 theotherclustermembers, anditisthereforepreferabletodistin- Filters, Wolf et al. 2003), and in addition 2dF spectra are avail- guishbetweencentralandnormalgalaxieswhenevaluatingtheen- ablefor282superclustergalaxies,fromobservationswiththetwo vironmentsofAGN. degreefield(2dF)spectrographontheAnglo–Australiantelescope. Statisticalstudiesofpointsourcesinthefieldsofgalaxyclus- TheCOMBO-17surveyusedtheWideFieldImager(WFI)at tershavefoundnumerousclustersthathavemoreluminouspoint theMPG/ESO2.2mtelescopetoobtainimagesofa0.56×0.55 sourcesthanexpectedfromanon-clusterfield(e.g.Henry&Briel degree field with a pixel size of 0.23800. Images were taken in 5 1991,Lazzatietal. 1998,Molnaretal. 2002,Cappietal. 2001, broadand12narrowbandfiltersandmatchedtoasetoftemplate Johnsonetal. 2003,Cappellutietal. 2005andRuderman&Ebel- spectratodeterminephotometricredshifts(z ).Reliablephoto- phot ing2005andGilmouretal.inprep.)andwhicharethereforelikely metricredshiftswerefoundforthe∼18000objectswithm <24, R tocontainAGN.Othersstudiesfindclusterswhichappeartohave with errors of order σ /(1 + z) < 0.01 (which is comparable z noexcessofsources(e.g.Molnaretal. (2002),Kimetal. 2004a to the velocity dispersion of the supercluster) at m < 20, and R and2004b),andthereforenodetectable(generallymoderatelumi- σ /(1 + z) < 0.02 for m < 23 (Wolf et al. (2005), here- z R nosity)AGN. after WGM05). The accuracy of the photometric redshifts when Martinietal. (2006)confirmspectroscopicallythateightlow- comparedtotheavailablespectroscopicredshiftsisshownbyWolf redshiftclusterseachcontainbetween2and10X-raysourceswith et al. to be good, such that it is possible to select a magnitude- L > 1041ergs−1,themajorityofwhichareAGNwithoutopti- limitedsupercluster samplewithminimal contamination fromin- X calemissionlines.Thiscorrespondsto5percentofgalaxieswith terlopersandonlyafewpercentlossoftruesuperclustergalaxies. M < −20 hosting AGN with L > 1041erg s−1, which is a Acutof0.155<z <0.185gives795galaxieswithtotal R X phot far higher AGN fraction than previously determined fromoptical absoluteVbandmagnitude<−17,whichareusedinWGM05and surveys. Recent results (Martini et al. 2007) show that the radial Laneetal. (2007).Thislargesamplemakesitpossibletodetermine distributionof thefainter AGN inthissample followsthat of the very accurately the distribution and properties of the galaxies in clusterpopulation, butthemoreluminousAGNarefoundprefer- A901/2(Grayetal. 2004). entially in the central regions. This result is in agreement with a TheopticallyidentifiedstructureofA901/2isshowninFig- recentstatisticalsurveyof18clusters(Branchesietal.2007). ure 1. A901 consists of two dark matter halos of comparable (cid:13)c 2007RAS,MNRAS000,1–21 EnvironmentaldependenceofAGNactivityin thesuperclusterA901/2 3 from Bell and Papovich, private communication) will help in the identificationofsomeAGN,asshowninSection4.2. 2.3 X-raydata TheA901/2superclusterregionhadpreviouslybeenobservedfor ∼ 0.4 ksecas part of theROSATAllSkySurvey(Ebeling et al. 1996) and in addition with the ROSAT High Resolution Imager for ∼ 12ksec. Schindler (2000) found seven sources in the field, twoofwhichcoincidewithA901aandA901b.Thebrightemission coincidentwithA901awasfoundtobeapointsource. In this paper a new, deep (90ksec) XMM-Newton image of thesupercluster,obtainedin2003,ispresented.Bycombiningthe deep X-ray image and the optical data a sample of AGN in the superclusterareselectedandanalysed. 3 X-RAYDATAREDUCTION Figure 1. Galaxy number density in A901 (grey-scale) and supercluster 3.1 Datareduction darkmatterdensityfrom2DweaklensinganalysisbyGrayetal. (2002) (contours).Galaxies areselected intherange0.155 < zphot < 0.185. A90ksecXMMimageofA901/2wastakenon6th/7thMay2003 Theopticallyidentifiedclustersandgrouparemarked.Theimageis∼300 usingthethreeEPICcameras(MOS1,MOS2,PN)andathinfilter. squareandthetopleftcornerisNorth-East. The level 1 data were taken from the supplied pipeline products, andreducedwithSASv5.4andthecalibrationfiles.Thedatawere filteredforbadpixels,thestandardgoodpatternsof0–12andXM- size,A901aandA901b,eachwithamassivegalaxyinthecentre. MEA EM or XMMEA EP and energy between 0.5 and 7.5 keV. A901acontainsfarmoregalaxiesandisfarmoreconcentratedthan Removing times when the count rate was > 0.2 counts s−1 for A901b.Atailofsmaller,bluergalaxiesextendssouthofA901bto- MOS1andMOS2detectorsand>0.67countss−1forPNresulted wardsA902,whichisamoreopticallydiffusecluster.Thereisalso inanexposuretimeof∼ 67ksecforMOSand∼ 61ksecforPN, agroupofgalaxiesinthesouth-westcorner,andopticaldataand andremovedallepisodesofsignificantflaring. 3Dweaklensinganalysis(Tayloretal. 2004)hasidentifiedaclus- teratredshift∼0.5almostdirectlybehindA902.Itisclearthatthe superclustercontainsawiderangeofenvironmentswithdiffering 3.2 Sourcedetection ratiosofdarktoopticallyvisiblematter.Theeffectoftheseenvi- ronmentsonthegalaxystar-formationratehasbeeninvestigatedby SourcesweredetectedusingWAVDETECT(Freemanetal. 2002) Grayetal. (2004),whofoundthattheproportionofgalaxiesthat on600x600pixelunvignettedfull-bandimagesforeachdetector. arestar-formingisastrong function of local dark-matter density, Theimagesandthecorrespondingexposuremapshadapixelsize withfarlessstar-formationingalaxiesindenserregions.Inaddi- of4.100.Amaskwascreatedforeachdetector,whichremovedareas tion,theWGM05studyfound apopulationof dustystar-forming withlessthan25percentofthemaximumexposureoranexposure galaxieswhichpreferentiallyexistinmediumgalaxy-densityenvi- mapgradientofover0.4forMOSor0.03forPN.Threeareasof ronments,avoidingboththefieldandtheclustercores. streakingwereremovedbyhandinthePNmask. 2dF spectra are available for 282 of the brightest galaxies 102sourcesweredetectedintheMOS1image,96inMOS2 in the supercluster, and cover the wavelength range 3900–6000A and128inPN.Thetotalnumberofuniquesourcesdetected,with- (Grayetal.inprep.).TheCOMBO-17SEDsand2dFspectraare outapplyinganycutonsourcesignificance,was150(ofwhich64 notsufficienttocompileasampleofAGNasmanyAGNareop- were detected in all three images, 33 intwo and 53 in one). The ticallyobscured.Incomparison,X-raysamplesarefarmorecom- vast majority of those missed inone or twoimages wereoutside plete(seeforexampleMartinietal. 2006andSzokolyetal. 2004) the fieldof view of those detectors, or only detected in the more butsufferfromconfusionwithheavilystar-forminggalaxies.Com- sensitivePNimage. bininganX-raysourcelistwiththe2dFspectraandCOMBO-17 Apointsourcecataloguewasconstructedforeachdetectorby data can help identify supercluster X-ray sources and distinguish removingalldetectionsofextendedsuperclusteremission.Asthe between X-ray emission from AGN and that from other sources sizeandshapeofthePSFisnotwelldefinedinXMMtwomethods suchasstar-formationandpopulationsoflowmassX-raybinaries. wereusedtodeterminewhichwerepointsources: Comparing the positions of AGN hosts with the other identified • The FWHM was found for each object. As the sources be- supercluster galaxies will determine whether AGN activity isen- comeincreasinglyellipticaltowardstheedgeoftheimage,itwas hancedorsuppressedinarangeofenvironments. requiredthatthesemi-minoraxishadaFWHMof<3pixels.This includesallbrighton-axispointsources,whichhaveaFWHMof 2.2pixels,andallowssomemarginoferrorforthefaintersources. 2.2 Infra-reddata Thismethodwasonlyusefulformoderatetobrightsources. This field is currently being surveyed using MIPS (Multi-band • The catalogue was compared to the results for this field ImagingPhotometerforSpitzer)onSpitzer.Anearlyreleasecat- from the XCS survey (private communication, see M. Davidson, alogueofthe24micronsourcesinthisfield(1/7ofthefinaldata, inprep.). Thissurvey uses asophisticated wavelet reconstruction (cid:13)c 2007RAS,MNRAS000,1–21 4 R. Gilmouret al. methodtofindextended emissioninXMMimages.Duetoprob- intwoimages;howeverithasaverylargepositionalerror,andfor lemswiththerawdatasetthismethodcouldonlyusetheMOSdata consistencyandaccuracywasnotincludedinthecatalogue. intheNEquarteroftheimage,whereasalldataareusedintherest Thefinallistof139significantpointsources,includingposi- of the image. Detections of extended emission are therefore less tionalerrors,isgiveninTable2andshowninFigure2. accurateintheNEquarter. Theresultsofthesemethodsarebroadlyinagreement,within the errors described, and identified eight areas of possible super- 4 FINDINGTHESUPERCLUSTERACTIVEGALACTIC cluster emission, shown in Figure 2. Of these, the smallest three NUCLEI arelikelytobeartifactsastheyalloccurnearthechipboundaryof 4.1 MatchingX-rayandOpticalcatalogues oneimageonly.Allofthesesourceswereremovedfromthecata- logue.AnalysisoftheextendedemissionwillbecoveredinGray TheCOMBO-17catalogueconsistsof63776objectsdetectedus- etal.,inprep. ingSExtractorontheR-bandimage(Wolfetal. 2003).Thesewere A further consideration is the brightest source in the field, matchedwiththeXMMpointsourcestoidentifytheX-raysources whichhastheFWHMofapointsource.Asthissourceissobright in the supercluster. Some saturated stars and fainter objects near (similar influx to theX-ray emission from A901b) and liesvery diffraction spikes are not included in the COMBO-17 catalogue, closetothecentreofA901a,itcouldbeconcentratedclusteremis- sotheareasaroundeachX-raysourcewereexaminedmanuallyfor sionoracoolingflow.Thesescenarioswereruledoutbyanalysis missingobjects,andfoursuchopticalobjectswhichcouldpossibly ofthespectrum,whichisapowerlawratherthanthermalandthe matchanX-raysourcewereaddedtothecatalogue. fact that the X-ray emission is centred on a galaxy which is not A maximum-likelihood technique was used to match the X- thebrightestclustergalaxyandwhichhasradioemission(fromthe ray sources to the COMBO-17 optical catalogue. Matching was NVSS,Condonetal. 1998).Itisthereforeconcludedthatthisob- performed by comparing the value LR (a measure of the as- i,j jectisanAGN. sociation between two sources, i and j, see Equation 1) with the distributionofthisvalueforX-raysourcesplacedrandomlywithin the field. Because LR depends on the error on the source co- i,j 3.3 Pointsourceproperties ordinates, which varies significantly in the X-ray sample, the ex- pecteddistributionofLR wascalculatedforeachX-raysource, Thereality,positionandpositionalerrorofthesourceswerefound i,j j,byrandomlyplacing14000X-raysourceswitherrorσ overthe bycomparingthesourcesfoundineachofthethreedetectors.For j opticalcatalogue.Theresulting(normalised)distributionN(LR) sources that were detected in more than one image the position j givestheprobabilityofobtainingeachlikelihoodratiobychance was defined as the midpoint of the two closest positions for that (ifsourcejhadnoopticalcounterpart),asshownforonesourcein source.(Inthemostcommoncaseofsourcesdetectedinallthree Figure3. imagesthisremovederrorsduetoonedetectionbeingnearachip Thelikelihoodratiowasdefined(followingMannetal. 1997 boundary.)Forsinglydetectedsourcesthegivenpositionwasused. andTaylorE.L.etal. 2005,whouseamethoddescribedindetail TheWAVDETECT1-sigmapositionalerrorswerefoundtobe inSutherland&Saunders1992)as generallylessthanonearcsecondevenforthefaintestsources.In comparison,theseparationsbetweendetectionsofthesamesource e−ri2,j/2σj2 in different images were on average 3 arcseconds, which is just LRi,j = σ2N(<m ) (1) less than one pixel. This is a random error, rather than astromet- j i ric,duetothedifficultyoffindingthecentreoffaintobjectswith whereσ isthepositionalerroronX-raysourcej,r thedistance j i,j a large pixel size compared to the PSF, and is a better measure- tooptical object i fromX-ray source j, and N(< m ) the num- i mentofthe‘true’erroronthestatedsourceposition.TheWAVDE- ber of optical objects brighter than object i in the r band image. TECTerrorsaredependent onthesourcesizeandcounts,andare This takes account of angular separation and optical magnitude, correlated with the distance between detections in different im- butmadenodistinctionbetweenobjectclassificationorphotomet- ages.Thisdistanceisonaverage7timestheWAVDETECTerrorfor ricredshift.Theerrorsonthepositionsoftheopticalobjectswere low-significancesources(onlychangingto6timeshigherforhigh- small enough to neglect compared to those in the X-ray, and the significancesources).Therefore,althoughtheWAVDETECTerrors astrometricerrorswerealsofoundtobenegligibleastheminimum are unphysically low they can be used to estimate the true error, errorontheX-raypositionof0.5pixelsissignificantlylargerthan especiallyforsinglydetectedsources.Forthisreasontheerroron theastrometricerroronthisimage. thesourcepositionwasgivenbythelargerofthefollowingthree This method treats optical quasars and galaxies of the same measurements:thedistancebetweenthetwoclosestdetections(if fluxinthesameway,anddoesnotaccountforthefactthatquasars theyexisted),7timesthestatedWAVDETECTerror,or0.5pixel(an are rarer and more likely to be X-ray sources. In addition, the erroroflessthanhalfapixelwasdefinedasunphysical). methoddoesnotdistinguishbetweenthebrightestgalaxies,which Asafinalstage,allsourcesthatwerenotdetectedatsignifi- arerareand quitelikelytobeX-raysources, and starsof asimi- cance> 3inatleastoneimagewereremoved,wherethesignifi- larmagnitude.TheseissuesareimportantwhenafaintQSOisthe canceisgivenby C ,forsourcecountsC andbackgroundcounts onlymatchandisassignedtoolowaprobability.Thiswillleadto σB B. (Thismeans that all sources arenot random fluctuations to at incompletenessintheX-raymatching,butwillnotaffectthesuper- least3σsignificance,butinmostcasesisoverlyconservativeasit clustersample.Inaddition,theprobabilitieswillnotbeaccurateif doesn’ttakeintoaccountthespatialdistributionofphotonswithin morethanonepossiblematchisidentified,andoneofthematches thesourcearea).Thiscutremoved11sources,andalsoagreeswell isaQSOorbrightstar.Thiscanaffectthesuperclustersample,and with the reality determined by eye. At least one of these sources inthesecasestheopticalclassifications,locationsanderrorsofall is real (at ∼09:56:24 -10:01:52, probably matching a z =2.2 thepossiblematcheswereexaminedindetailtodeterminethetrue phot quasarintheCOMBO-17catalogue),asitwasmarginallydetected sourceoftheX-rays. (cid:13)c 2007RAS,MNRAS000,1–21 EnvironmentaldependenceofAGNactivityin thesuperclusterA901/2 5 Figure2.Theidentifiedsourcesinthesuperclusterfield.Sourcesmarkedwitharectanglearepossibleextendedemission.Pointsourceswithsignificance C >3aremarkedwithanellipse,andthosedrawninboldarepossiblesuperclustermembers.TheemissionintheregionofA901aisaverybrightpoint σB sourceratherthanextendedemission.Thesourcesareplottedoveracombined,vignettingweightedimage,andsomesourcesarenotvisiblebyeyeduetothe imagecombinationandscale.Theimageis∼300indiameterandthetopleftcornerisNorth-East. In addition this technique compares likelihoods to the aver- ageoverthefieldandignoresanyclustering,whichgivesslightly higher likelihood values for X-ray sources in the line of sight to cluster centres, andslightlylower for those inareaswithfew su- perclustergalaxies.Thisisveryunlikelytochangeanyresults,es- peciallyasthecatalogueisdominatedatallopticalmagnitudesby non-supercluster objects(> 80percentofopticalobjectsarenot inthesuperclusteratm <20,and>93percentatm <24). R R Thestandardmethodofcalculatingthereliabilityofamatch requiresalargesampleofsources,soinsteadthereliabilityforeach X-ray–optical pairi,j wasdefinedastheprobability ofnot ob- tainingLR randomly, i,j ΣN(LR >LR ) R =1− j i,j . (2) i,j 14000 Most X-ray sources have more than one potential optical counterpart,aswellasasignificantprobabilityofhavingnomatch, Figure3.TheexpecteddistributionoflikelihoodratiosforoneX-raysource suchthatthereliabilitiessumto> 1.Theprobabilitythatoptical basedon14000trials(solidline),cumulativeexpecteddistribution(dashed objectiisthetruecounterparttoj,(P )andtheprobabilitythat i,j line) and thehistogram ofactual likelihood ratios forthe optical objects thereisnocounterpart (P ),givenasetofpossiblecounter- none,j nearthissource(thevastmajorityofoptical objects havelog(likelihood) parts,k,arecalculatedfollowingRutledgeetal. (2000), <1).Onepossiblematchisidentified,withareliabilityof∼0.8. P = Ri,jQMk6=i(1−Rk,j) (3) i,j S Aliberalcatalogueofpossibleandsecurematcheswascon- P = QMk=1(1−Rk,j) (4) structed.Thethresholdsusedweredeliberatelylooseasmanyfac- none,j S torscouldincreasethelikelihoodofamatchonacasebycasebasis. whereSisanormalisationfactorsothattheprobabilitiessumto1, Theseincludematchingwithotherwavelengths,visualinspection andM isthenumberofpossibleopticalcounterpartstotheX-ray oftheshapeandcentringoftheX-raypointsource,andcharacter- source. isticsoftheopticalcounterpart,forexampleifitisafaintquasar. (cid:13)c 2007RAS,MNRAS000,1–21 6 R. Gilmouret al. Inaddition,someoftheX-raysourceshavesuchlargeerrorsthat identifyopticalmatchesinpreviouslyambiguouscases.For97X- thechanceofarandomassociationwillalwaysbehigh,evenifthe raysourcestheadditionofthe24µmdataconfirmedtheresultof matchappearstobeexcellent.Thereforeallmatcheswhichfulfill thepreviousmatching,including49caseswhereneitherthe24µm thecriteriadescribedbelowareincludedinTable2,andtheaddi- data nor the optical catalogue gave a good match. This includes tionalfactorswhichmayaffecttheprobabilityofamatcharelisted caseswherethe24µmdataconfirmtheexistenceoftwopossible inthe‘Notes’ column. Thepossible supercluster AGN areevalu- matches, and cases where two 24µm sources correspond well to atedonacasebycasebasis. oneX-rayandoneopticalobject(andappeartobemultipledetec- Forsourceswithonlyonelikelyopticalcounterpart,amatch tionsofnearbygalaxies).Forafurther14X-raysourcesthe24µm isdefinedasP > 0.8,whichisconfirmedbyvisualinspection. matching identifies at least one of the optical counterparts but at i,j Forsourceswithmorethanonepossiblecounterparttheconditions a lower probability (between 0.65 and 0.8). Four X-ray sources for object i to be a unique counterpart given a set of options, k, hadasecure 24µmmatchwithnooptical counterpart, and20X- werePkPk,j > 0.8andPi,j/Pk6=iPk,j > 4.Thisresultedin raysourceshadanopticalcounterpart butnolikely24µmmatch. 66secureidentificationsoutof139sources,ofwhichfewerthan6 ForfourX-raysourcesthe24µmdatachangedtheassignedoptical areexpectedtoberandomassociations.Forsourceswithmultiple match, by eliminatingor confirming possible optical objects, and possible counterparts (PkPk,j > 0.8 and Pi,j/Pk6=iPk,j 6 thesesourcesareflaggedinTable2. 4)allopticalobjectswithP > 0.15wereincludedaspossible ThefinallistofX-raysourceswithuniqueandmultiplecoun- k,j matches.Thisresultedin17sourceswithtwopossiblecounterparts terparts is given in Table 2, which lists the X-ray IDs, positions, and3sourceswiththreeoptions. positional error and count rates, and the possible COMBO-17 matches, optical position and photometric redshift. Theprobabil- itiesand reliabilitiesof theoptical matches aregiven to allow an 4.2 Matchingwiththe24-microndata evaluation of the accuracy and uniqueness of each match. The TheSpitzer24microncatalogueof1194sourcesintheX-rayfield 24µmfluxofanymatchingSpitzersourcesandthecombinedprob- ofviewwasusedtoimprovetheX-raytoopticalmatching.These abilityoftheX-ray–SpitzermatchandtheSpitzer–Opticalmatch dataareuseful asAGNoftenhaveinfra-redemission,butalsoin bothbeingtruearealsogiven. a purely statistical sense there are far fewer Spitzer sources than opticalobjectsinthefieldofview,sotheprobabilityofarandom 4.3 CriteriaforidentifyingsuperclusterAGN association is far lower. As the 24µm sources have smaller posi- tionalerrorsthantheX-raysources,andarerarerthantheoptical To identify the AGN in the supercluster it is necessary to use objects, amatch between an X-ray and 24µm source can signifi- thephotometric and, if possible, spectroscopic redshifts from the cantlyimprovetheaccuracyoftheopticalmatch. COMBO-17 survey. A cut of 0.155 < z < 0.185 was used phot Asimplelikelihoodratiotest,followingthemethodfortheop- toensurethatallAGNassociatedwiththesuperclusterwereiden- ticalcatalogue,produced81uniquematchesbetweentheX-rayand tified(seeWGM05fordetailsoftheredshiftcut).Thisrangealso 24µmsources,and20X-raysourceswithmorethanonepossible allowsfortheerrorsinthephotometricredshifts,whichmayalso 24µmcounterpart.ThesameprobabilitycutsastheX-raymatch- beaffectedbytheAGNemission.Itisfoundthatadding/subtract- ingwereapplied,andthelowsurfacedensityof24µmandX-ray ingtheCOMBO-17redshift errorforeachX-rayemittingsource sourcesmeanthatveryfewfalsematchesareexpected.Examina- does not reveal any more possible supercluster X-ray sources. In tionoftheopticalimagesforthefewX-raysourceswithmorethan addition some galaxies have bimodal photometric redshift distri- onepossible24µmcounterpartshowsthatinmostcasesthe24µm butions,sothesecondchoiceredshiftswerecheckedandnoextra ‘pointsources’arelikelytobemultipledetectionsofnearbygalax- superclusterX-raysourceswerefound. ies,whicharesignificantlylargerthanthe24µmPSF. The presence of an AGN may cause the template fitting in The24µm sources werethen matched withtheoptical cata- theCOMBO-17surveytogivewrongphotometricredshifts,asthe logue.Theerrorsforthe24µmcatalogueweretakenas1.200,which COMBO-17templatesdonotincludeSeyfert-likespectrawithboth is half a pixel. This will underestimate the errors on faint 24µm AGN and host galaxy contributions. To check for missed super- sources,andgiveamoreconservativecatalogue.Theadvantageof cluster X-ray sources we examined all optical counterparts with takingthesameerrorsforallsourcesisareductionincomputing 21>m >17.75(betweenthefaintestsuperclusterX-raysource R timeasonlyoneexpecteddistributionneedstobecalculated.Inad- and the brightest supercluster galaxy), and B −R < 2.3 (on or dition,asthesamplewaslargeenough,atruereliabilitywascalcu- bluerthanthesuperclusterredsequence)whichwereclassifiedas latedbycomparingthelikelihoodratiodistributionforthesample galaxiesaccordingtotheirtemplatespectra.1.Thephotometricdata (Ntrue(LR))withthatof10randomcatalogues(Nrandom(LR)) forthesegalaxiesweremanuallycomparedtospectraltemplatesat followingthemethodofTaylorE.L.etal. (2005).Thereliability thesuperclusterredshift.Twoopticalcounterparts(COMBOcata- isdefinedasafunctionoflikelihoodratio, loguenumbers12953and41435, matchingX-raysources#3and N (LR)−N (LR) #135) were found to fit well with templates at z ∼ 0.16 despite R(LR)= true random . (5) having different photometric redshifts in the COMBO catalogue. N (LR) true ThesearediscussedindetailintheAppendix. ProbabilitiesforeachpossiblematchwerecalculatedusingEqua- X-rayemissionoftheluminositiesseeninthissamplecould tions 3 and 4, and the criteria for unique and multiple matches be caused by a large population of low mass X-ray binaries usedintheX-raymatchingwereapplied.The24µmsourcesthat (LMXBs),hot coronae of massive galaxies, or highlevelsof star matched the X-ray catalogue were examined by eye to identify those which had good optical matches, but were rejected due to underestimatedpositionalerrorsinthe24µmdata. 1 Photometricredshiftsofobjectsclassedashighredshiftquasarsareac- The combined probability of the X-ray – 24µm – optical curateasthechanceofagalaxyatmR <24beingmistakenforaquasar match was used to identify the 24µm counterparts and to help isverysmall(Wolfetal. 2004). (cid:13)c 2007RAS,MNRAS000,1–21 EnvironmentaldependenceofAGNactivityin thesuperclusterA901/2 7 formation,aswellasAGNactivity.EmissionfromLMXBsisruled • X-ray Hardness Ratio – A good indication of the spec- out followingthe method of Martini et al. (2006), who compare tral properties and absorption of X-ray sources is given by the theirobservationstothetightrelationbetweenB-bandgalaxylumi- luminosity hardness ratio, HR = H−S, where H = L H+S 2−8keV nosityandthetotalX-raybroadbandluminosityfromallLMXBs and S = L . Sources with HR > 0.8 are unlikely be 0.5−2keV (Kim&Fabbiano2004).EvenwithoutcorrectingforthewiderX- star-formingduetotheverylargeamountsofabsorptionrequired raybandusedinthisrelation(0.3−8keVcomparedto0.5−7.5keV (Mainieriet al. 2002) (unlesstheemissionisdominatedbyhard fortheA901/2sources)theX-rayemissionfromthepossiblesu- X-raybinaries),andsourceswithHR > −0.2aremorelikelyto perclusterX-raysourcesisatleastafactorof6,andmedianfactor beAGNthanstar-forming(Szokolyetal. 2004).Hardnessratios of 32 higher than the Kim & Fabbiano average relation. This is werecalculatedfromthebackgroundsubtractedimageswhichwill significantlyhigherthanthescatterintheirobservations.Emission be described in detail in the paper on extended emission in this fromhotcoronaeisalsohighlyunlikelytobethecauseoftheX- field. rayemission,astheupperlimitontherelationshipbetweenB-band • OpticalLineRatios– Lineratiosinopticalspectracandis- and X-ray luminosity for such sources is very similar to that for tinguishbetweenstar-forminggalaxiesandAGN.Asthe2dFspec- LMXBs(Sunetal. 2007)andtheA901/2sourceshavefarhigher traarenotfluxcalibratedonlythe[OIII]andHβlineswereused,as X-raytoB-bandluminosityratios. theyareclosetogetherinwavelengthsoconsideringthelineequiv- To distinguish between X-ray emission from high levels of alentwidths,andassumingaflatcontinuumspectrum,willnotin- starformationandthatfromAGN,variousmethodswereusedde- troduceoverlylargeerrors.Mostofthe2dFspectraoftheoptical pendingontheinformationavailableforeachsource.Usedalone counterpartshaveveryfaintornolinessoonlyupperlimitscanbe mostofthesemethodscannotdistinguishabsolutelybetweenstar- measured. Lamareille et al. (2004) compare the classification of forminggalaxiesandthosewithAGN,butcombiningtheavailable emissionlinegalaxiesusingthetraditionalBaldwinetal. (2003) datacangiveareliableindicator ofthenatureoftheX-rayemis- diagnosticmethod,andamethodbasedononlyblueemissionlines sion. (λ > 3700). Their results show that if [OIII]λ5007/Hβ > 5.5 ∼ then the object is almost certainly an AGN. However if the ratio • X-ray Luminosity – Star forming galaxies generally have is< 5.5 theobject could stillcontain someAGN activity, and if low X-ray luminosities. A source in the local universe with no lines are visible it may be an obscured AGN, so this method L0.5−8keV >3×1042ergs−1isextremelyunlikelybepurelystar canconfirmthepresenceofanAGNbutcannotruleoutanyAGN forming (Bauer et al. 2004), and any source with L0.5−8keV ∼> activity. 1×1041ergs−1islikelytobeanAGN(seeFigure7ofBaueretal. –mostsourceswithL > 1041 whicharenotnearthefluxlimit X are securely identified as AGN by their column density or hard- 4.4 DetailsofsuperclusterAGNcandidates ness ratio). Luminosities were calculated using aperture photom- Afterapplyingtheredshiftcut,thecandidatesforsuperclusterAGN etry on images with the mean background subtracted. The back- arereduced toelevenoptical matcheswithsecure photometricor ground subtraction process followed the method of Arnaud et al. spectroscopicsuperclusterredshifts,twomatcheswithrevisedpho- (2002),andwillbedescribedintheforthcomingpaperontheex- tometricredshiftsandthreelesssecurematcheswithconfirmedsu- tended emission (Gray et al. in prep.). The redshift of the super- perclustergalaxies.ElevenofthecandidatesuperclusterAGNhost clusterAGN(∼0.17)meansthattheobserved0.5–7.5keVcounts galaxieswereobservedwith2dF,andfivedonothavespectra. canbeconvertedintoanemitted0.58–8.7keVluminosity.Asthis The details of the candidates are given in the Appendix and studydoesnotrequireveryaccurateluminosities,andastheerrors inTable1,includingtheirX-rayandopticalproperties,imagesof onthecountratesarelarge,thisistakenasanapproximationtothe thehostgalaxiesandspectrawhereapplicable.Thespectralenergy 0.5–8keVluminosity.Thetrueluminositiesmaybeslightlyhigher, distribution(SED)ofthesuperclustergalaxiesisgivenineachcase butthisdependsontheX-rayspectrum. aseither‘oldred’,‘dustystar-forming’or‘blue-cloud’,asdefined • [OII] Star-Formation Rates – If the X-ray emission is intheWGM05samplefromphotometricfitstogalaxytemplates. purelyduetostar-formation,withnoAGNpresent,thenthestar- The morphologies of the host galaxies which are included in the formation rate (SFR) can be estimated from the soft band X-ray sampleofLaneetal. (2007)arealsogiven. luminosityas ThecandidateswereassignedtobesuperclusterAGN,possi- SFR(M(cid:12)/yr)=2.2×10−47L0.5−2keV(W) (6) blesuperclusterAGN,orrejectedoutright.ThesuperclusterAGN aresources #20, #24, #34, #37, #71, #79, #81, #104, #105, #135 (Ranallietal. 2003).IfthereisnoAGNandminimalabsorption and#139.ThepossibleAGNis#3.Thefinalsamplethereforecon- thenthismustbeneartotheSFRderivedfromthe[OII]λ3727line sistsof 11 likely supercluster AGN and 1 possible member. Two flux(Hopkinsetal. 2003); sources,#3and#135,showsignificantvariabilityinthebroadband photometry. Thesearealsothetwomost luminous sourcesinthe L SFR(M(cid:12)/yr)= 2.97×[O1I0I3]3W (7) X-ray,andduetotherapidvariabilitymustbeopticalTypeIAGN. AlloftheAGNhostsappeartobemassivegalaxiesandsome whereL canbeestimatedforthoseobjectswith2dFspectra appear to be morphologically disturbed. The morphologies as- [OII] fromtheequivalentwidthofthelineandtheCOMBO-17magni- signedtotheAGNhostsareindistinguishablefromtheparentpop- tude in the rest-frame Johnson U band. (This method assumes a ulationofluminousclustermembers.Detailedstudyofthesmaller flatspectrumintheUband,butwillgiveanestimateofthefluxat scalemorphologies of theAGNhostsisbeyond thescopeofthis 3727A˚ towithinatleastafactoroftwo,astheUbandmagnitude work,andwillbeleftuntilrecentlyobtainedHubbleSpaceTele- hasminimalcontaminationfromfluxabovethe4000A˚ break). scopeimagesofthisfieldareanalysed. Thefractionof ‘oldred’, • X-ray/OpticalFluxRatio– Sourceswithf /f > ‘dustystar-forming’ and‘blue-cloud’ host galaxies(asdefined in 0.5−8keV R 1areverylikelytobeAGN,andthosewithf /f > 0.1 theWGM05samplefromphotometricfitstogalaxytemplates,and 0.5−8keV R arelikelytobeAGN(seeBaueretal. 2004andreferencestherein). excludingtheverybrightAGNasitissignificantlycontaminated (cid:13)c 2007RAS,MNRAS000,1–21 8 R. Gilmouret al. 2.5 secure AGN 10-2 possible AGN 2.0 n) o gt B-R colour 1.5 X) / L(X,eddin 10-3 L( 10-4 1.0 secure AGN AGN in excluded area possible AGN 0.5 10-5 -24 -23 -22 -21 -20 -19 -18 -23.5 -23.0 -22.5 -22.0 -21.5 -21.0 -20.5 -20.0 R band magnitude MR Figure4.Thecolour–magnitude plot forall supercluster galaxies (small Figure5.Approximate accretion efficiency (fraction ofEddingtonX-ray dotswitherrorbars)andAGNhosts.Totalabsolutemagnitudesareplotted, luminosity)asafunctionofhostgalaxyabsolutemagnitude.Thesolidline witherrorsinthecolourandfluxmeasurement.Theerrorsdonotinclude shows the approximate minimum detectable accretion efficiency at each uncertainties intheredshifts, whichincrease theerrorbarsto±0.1.The point. The X-ray bright source #135 is excluded from this plot, and has areaoftheAGNsymbolsisproportionaltolog(LX(0.5–8keV)).TheX-ray anaccretionefficiencyof∼4percentatMR=−23. source#135ismarkedwithastarasitisnotinthecontrolareadescribed inSection5.2. Thecolour–magnitudediagramforthesuperclusterisshown withAGNopticallight)are0.5,0.2and0.3,comparedtofractions inFigure4,withthesuperclusterAGNhostgalaxiesindicated(it intheintermediatedensitygalaxypopulation of ∼ 0.54, ∼ 0.26 is important here to note that the R band magnitudes of the host and∼0.20,sotheAGNhostsareagainindistinguishablefromthe galaxiesarenotsignificantlychangedbythepresenceofanAGN, parentpopulation. ascanbeseendirectlyfromthe2dFspectra).AlloftheAGNand ItisworthnotingthatallofthesuperclusterAGNcandidates possibleAGNlieinbright(M <−20)galaxies.TheX-raylumi- R areclassedasgalaxiesintheCOMBO-17surveyasthephotomet- nosityoftheAGN,asindicatedbythesizeofthesymbolsinFig- ricmethodisnotsensitivetolowluminositySeyfert-likeAGNand ure4,showsthatthelackofAGNinfaintergalaxiesisnotdueto obviouslymissesopticallyobscuredAGN.Inaddition,ofthe8su- fainterX-rayAGNbeingfoundinopticallyfaintergalaxies,which perclusterAGNwithopticalspectraonly5haveemissionlines,and wouldleadtoAGNinlowerluminositygalaxiesfallingbelowthe onlytwohave[OIII]/Hβratioswhichwouldleadtoaclassification detectionthreshold. Infact,ifanycorrelationexistsitislikelyto asanAGNusingopticaldataalone.Thishighlightsagaintheneed be theopposite –there isan 84 per cent chance that more X-ray forX-raystudiestoinvestigatetheAGNpopulation. luminousAGNarefoundinopticallyfaintergalaxies,accordingto aSpearmanranktest. To find the range of accretion rates covered by this sample, L /L wascalculated for eachAGN, whereitwasas- 5 ANALYSISOFAGNPROPERTIESAND X X,eddington sumedthat10percentofthebolometricluminosityisemittedin ENVIRONMENTS the0.5–8keV band. Therelationlog(MBH/M(cid:12)) = −0.5MR − ThesuperclusterA901/2isverydiverseandcontainsawiderange 2.96ofMcLure&Dunlop(2002)wasusedtocalculatetheblack ofenvironmentswhichmayhaveaneffectonAGNactivity.Eleven hole mass from the rest frame R-band absolute magnitude (M ) R X-raysourcesinthisfieldarelikelytobesuperclusterAGN.Using given in the COMBO-17 catalogue (as derived from the galaxy theCOMBO-17datasetthepropertiesofthehostgalaxiesofthese template,correctedtothecosmologyofthispaper).Theresulting AGNcanbefound,andthenumberofgalaxieshostingAGNcan plot (Figure5, excluding thebright source #135), although crude becalculated.Thisinformationcanbeusedtoconstructacontrol duetotheapproximationsmade,appearsatfirstglancetoshowa sampleofgalaxieswhichappearsimilartoAGNhosts,butdonot correlationbetweenM andL /L (whichwouldnot R X X,eddington havesignificantX-rayemission.Bylookingatnearbygalaxydistri- bechangedbyalteringmanyoftheassumptionsabove,forexample butions,theenvironmentsoftheAGNhostscanthenbecompared the10percentemittedintheX-ray,astheywouldonlychangethe tothose of thecontrol sample to determine whether environment scaleoftheplot).However,therearemoregalaxiesatM ∼−21 R andAGNactivityarelinked. thanM ∼−23andforthefaintergalaxieslowerefficiencyAGN R mayfallbelowthedetectionlimit.Tocalculatewhetherthelackof moreefficientAGNinmoreluminousgalaxiesisduetothesmaller 5.1 PropertiesoftheAGNhosts samplesizeoraphysicaleffectitisnecessarytoknowthenumber TheWGM05superclustersamplecontains795galaxies,whereall ofpossibleAGNhostsasafunctionofMR. galaxies with 0.155 < z < 0.185 and absolute Johnson V phot magnitude < −17 were identified as supercluster members. The 5.2 ThefractionofgalaxiescontainingadetectedAGN largeredshift range (the same asapplied for the AGN inSection 4.3)allowsfortheerrorsinthephotometricredshifts,andthemag- Todetermine theproportion of galaxieswithAGN and theprop- nitudecutremovesfaintobjects,whichhavefarlessaccuratepho- ertiesoftheAGNhostsitisnecessary todefineacontrolareain tometricredshifts. which AGN could have been detected, and compare the galaxies (cid:13)c 2007RAS,MNRAS000,1–21 EnvironmentaldependenceofAGNactivityin thesuperclusterA901/2 9 0.4 tiniet al. (2006), whofounda∼5percent X-raydetectedAGN fractionforgalaxieswithM < −20inlowredshiftclustersata R 1.2) fluxlimitofLX = 1041ergs−1.Figure6alsoshowsanincrease g(L)>4X0.3 isnivAeGgaNlaxfriaecst,ibount(tahbeosvmeatlhlesalmumplienosisziteymliemanits)tihnatinthcreeearsrionrgblyarmsaarse- o N (l verylarge.InthissuperclusterAGNareonlyfoundingalaxieswith G M < −20.4.Asstatedpreviously, theluminositiesoftheAGN A R with 0.2 showthatthisisnotduetothefluxlimitofthesample.Rather,it es appearsthatfainterAGNaremorelikelytobefoundinmorelumi- xi a nousgalaxies,brighterAGNinmoderatelyluminousgalaxies,and al g of 0.1 noAGNintheleastluminousgalaxies. op. Returning to Figure 5, it is now clear that the lack of AGN Pr withL /L >∼3×10−4andM <−22.5isprob- X X,eddington R ablyduetothelackof brightgalaxies,ratherthanatendencyfor 0.0 -23 -22 -21 -20 -19 more luminous host galaxiesto havelower efficiencies: there are MR only19possibleAGNhostgalaxiesabovethisluminosity,andas ∼3percentoffainter(−22.5 <M <−21.5)galaxieshaveac- Figure6.ProportionofgalaxieshostingAGNasafunctionofR-bandab- R cretionefficienciesabove3×10−4 itisnotsurprisingtofindno solutemagnitude,with1σerrorbars.Thedashedlinesshowtheresultsif bright galaxieswithAGN efficienciesabove thislevel. Similarly, thepossiblesuperclusterAGNisincluded.Thedottedlineinthefinalbin showsthe95percentconfidencelimitforthefirstemptybin.Againsource theonesuperclusterAGNwithefficiency>10−3onlycorresponds #135andgalaxieswhereAGNactivitycouldnotbeeasilydetectedarenot to∼0.5percentofthegalaxiesinthat0.5magnitudebin,explain- included. ingthelackofbrightergalaxieswithsimilarefficiencies.However theremustbeasignificantdecreaseintheefficiencyofanyAGNin galaxieswithM >−20asthenumberofsuperclustergalaxiesis R withandwithoutanAGNinthisarea.Toselectthecontrolareathe verylargeyetnoAGNareobservedabovetheX-rayfluxlimit. COMBO-17 catalogue wascut toremove objectswithin16000 of thetopandsidesand30000ofthebottomoftheimage.Thiscuten- 5.3 Definingacontrolsample suresthereturnedareais97percentcoveredbytheX-rayimage, andalsoensuresthattheedgeofthecataloguedoesnotaffectprop- To compare the AGN environments and properties, control sam- ertiessuchaslocaldensity.Inadditionareaswherethesensitivity pleswerecreated,consistingofgalaxiessimilartotheAGNhosts, topointsourcesdecreasedsignificantlyduetoextendedsourcesor whereAGNactivitycouldhavebeendetectedbutwasnotfound. very bright X-ray emission, wereremoved fromthecontrol area: Whereasarandomlyselectedcontrolsamplewouldcontainmany three areas were bright enough to obscure moderate luminosity faintgalaxies,itisinsteadpreferabletodefineacontrolsamplewith AGN:a3000circlearoundtheextendedemissiontothenorth-west asimilardistributionofgalaxypropertiesastheAGNhosts.Any of A901a (marked with a rectangle in Figure 2), 6700 around the difference betweenthe AGNhosts’ environments andthe control AGNinA901a(#135)and8300aroundA901b.Intheseregionsthe sampleenvironmentswouldthereforebeduetoanenvironmental noise level is at least 50 per cent higher than for the rest of the effectonAGN. field.Attheedgeoftheexcludedregionsthefaintestdetectedsu- 100controlsamplesweremade,eachconsistingof65ofthe perclusterAGN(1.8×1042erg/sec)wouldbedetectableatlowsig- 183 supercluster galaxies with aperture magnitude m < 20 ap,R nificance,butthesensitivitytowardsthecentreof#135andA901b whichlieinthecontrol areadescribed inSection5.2andarenot decreasestoalevelwhereonlyAGNbrighterthan∼ 1044erg/sec AGNhosts.Thesesampleswereselectedatrandomsuchthatthere wouldbedetectable.Nosourcesaredetectedintheseareas(except areequalnumbersofgalaxiesineach0.5aperturemagnitudebin for#135itself),evenatlowsignificance.Thesmallchangesinde- toreplicatethedistributionofAGNhostmagnitudes.Thismethod tection sensitivity due to emission from A902 and the SWgroup ignores the possible increase in the number of AGN in brighter andthechangesinPSFwerenotincludedastheyonlyaffectvery galaxies(Figure4)butthisislesssignificantwhenaperturemag- marginaldetections,suchthatthefaintestsuperclusterAGNwould nitudesareused,andduetothesmallnumberofgalaxiesselected stillbedetected at low significance, and asmall change insensi- shouldnotaffectanyresultssignificantly. tivity is not significant in a sample of this size. The applied cuts The 100 samples are identical at m < 18.5, due to the ap,R alsoremovetheAGNinA901a(#135)fromthesample,whichis small number of available galaxies, but at 18.5 < m < 20 ap,R necessaryasitobscurespossibleAGNactivityfromallofthesur- different sets of galaxies were chosen, although the samples still roundinglargegalaxiessowillbiasthesample.Inadditionitshigh haveconsiderableoverlapwitheachother.Eachofthe100control accretionefficiencyandluminosityandopticalvariabilityshowthat sampleswascomparedtotheAGNsampleandthemedianstatis- it,andtheonepossiblesuperclusterAGN,aretheonlyX-rayType- tictakentoreducetheerrors.AKolmogorov–Smirnov(K–S)test IAGNinthesample.Thecontrolsamplecontains604supercluster (toidentifychangesinthemean)andKuipertest(aK–Stestusing galaxies,as149wereremovedintheedgecutand42wereinre- Kuiper’sstatisticwhichisbetteratidentifyingchangesinspread) gionswhereAGNcouldnotbedetected. confirmthatthecontrolsamplesandAGNhostsaredrawnfromthe ThenumberofAGNhostsinthesuperclustercanbedirectly sameR-bandmagnitudedistribution(atatleast56percentconfi- compared to the number of possible host galaxies in the control denceforthesuperclusterAGN). samplearea.ThenumberofAGNperpossiblehostgalaxyisshown Thecolour–magnitudediagraminFigure4alsoshowsthatat inFigure6 for arange of host luminosities. Thetotal fractionof leastthreeAGNhostsaresignificantlybluerthantheredsequence. galaxies with M < −20 hosting AGN with L > 1041.2erg K–SandKuiper’stestsonthedeviationfromtheredsequencegive R X s−1is10/253,or∼4percent.ThisissimilartotheresultsofMar- 0.59 and 0.48 probabilities that the supercluster AGN hosts and (cid:13)c 2007RAS,MNRAS000,1–21 10 R. Gilmouret al. -10.0 C E D -10.2 -149.2 RA -149.0 Figure8.Regionswithaclearenvironmentalcategoryinthesupercluster. Regionsareidentifiedbyeye,usingdeepimagesandgalaxypropertiesas Figure 7. Distribution of Σ10 for the control samples (solid line, mean wellasthegalaxypositionsshown.Galaxiesinthe80percentcomplete valueofall100samplesisgiven)andsuperclusterAGNhostgalaxies(filled superclustersamplearemarkedwithasmallcross,andforcomparisonthe histogram).Thecontrolsampledoesnotincludegalaxiesinthreeareasof AGNarealsomarkedbysquares(superclusterAGN)anddiamonds(pos- high X-ray emission, where moderate luminosity AGN could not be de- siblesuperclusterAGN).Thefigureisaround300 by240,withnorth-east tected. TheAGN hostgalaxies havefarlowervalues ofΣ10,andanar- tothetopleft.Thedashedcircleinthetopright-handcorneris1.50inra- rowerspreadofvalues,thanothersimilargalaxies.Thedifferencebetween dius,thesizeusedtodeterminethelocalenvironmentalpropertiesofeach the control samples andAGN (taking the median value) issignificant to galaxy. >0.98usingKuiperstestand>0.99usingaK–Stest. 5.5 ThetypesofenvironmentsthatcontainAGN Thesuperclustercontainsawiderangeofenvironmentswhichare controlgalaxiesaredrawnfromthesamedistribution.Martinietal. evidentbyeye,suchasclusters,groupsandfilamentsofgalaxies. (2002)foundapropensityforAGNhoststobebluerthansimilar InordertobetterclassifytheenvironmentsofAGN,itisdesirable galaxiesintheclusterA2104,butthissampleissosmallthatsuch tofindamethodofdistinguishingbetweentheseenvironmentsin aneffectcannotbeconfirmedhere. termsoftheirproperties.Asafirststep,areaswhichfallintoaclear environmental categorywereselectedbyeye,asshowninFigure 8.Althougharbitrarilydefined,thepropertiesofgalaxiesinthese areascanhelptoworkoutwhatpropertiesdistinguishdifferenten- vironments, andhence todefinetheenvironments ofother super- 5.4 ThelocalgalaxydensitiesofAGNhostgalaxies clustergalaxiesandtheAGNhosts. TheenvironmentsoftheAGNwereevaluatedbycomparingthelo- Theregionsweredefinedasfollows:clusterandgroupregions caldensitiesoftheAGNhostgalaxiestothoseofthecontrolsam- arecirclesof10 radius,centredonthebrightestclustergalaxy.As ples. Thesurface massdensityfromthe weak-lensing analysisin A902doesnothaveaclearcentre,aradiusof1.50isusedtoinclude Grayetal. (2002)wasnotusedasthesampleofAGNissmalland the whole cluster region. For the rich cluster A901a and the SW theylieinareas(outsidecluster cores)wheretheerrorsarequite groupasecondregionoutto30 radiusisconsidered,markingthe large,such thatany resultswould haveverylow significance. In- clusterandgroupedgeenvironments. Thefilamentregion,witha stead,theprojectedgalaxydensitywasused,followingthemethod largenumberofbluegalaxies,ismarkedwitharectangle. for the WGM05 sample, where Σ is defined as the number of Therearemanypossiblewaystodefinegalaxyenvironments, 10 superclustergalaxiesper(Mpch−1)2 withinacirclewitharadius suchasmeanlocalluminosity,galaxycolourordistancefromthe given by the average distance to the 9th and 10th nearest neigh- restframeU-Vcolour-magnitudemainsequence(asusedbyGray bours.Thesuperclustersampleusesonlythebrightestgalaxiesas etal. 2004),butnotallofthesewerefoundtobeusefulindistin- definedinSection5.1. guishingbetweenthepredefinedareasinFigure8.Itwasfoundthat The distributionsof Σ for thecontrol samples and thesu- thegalaxiesindifferentenvironmentscouldnotbeseparatedwell 10 perclusterAGNareshowninFigure7.TheAGNhostgalaxieslie purelyintermsoflocalprojecteddensity,andthattwootherfactors predominantlyinareasofmoderatedensitycomparedtothecontrol are also important: firstly the number of less luminous galaxies, sample,withasignificanceof98.4percentand99.0percentfrom and secondly the colour of the local galaxies. The distance from Kuipers and theK–S test respectively. It isparticularly clear that the colour-magnitude relation was investigated as a measure of AGNhostsavoidthedensestregionsofthesupercluster,evenac- colour,buttheenvironmentswerebetterseparatedusingthemea- countingforthefactthattwooftheclustercoresareremovedfrom sured colour, as the latter combines colour with a distinction be- thisstudy.Theexception,ofcourse,istheveryluminousAGNin tweenbrightandfaintgalaxies.Themeanlocalcolourwasusedas thecoreofA901a,whichwasexcluded,alongwiththesurround- thismeasureissensitivetothepresenceofafewveryblueoractive inggalaxies,astheydonot lieinthecontrol area. Thissource is galaxies,wherasthemediancolourisdominatedbyredgalaxieson differentinbothpropertiesandpositionfromtheotherAGNinthe thecolour-magnitudemainsequence(Figure4),andismoreamea- supercluster. sureoflocalmagnitude.Toincludefaintergalaxiesthesupercluster (cid:13)c 2007RAS,MNRAS000,1–21
Description: