ebook img

The cosmic evolution of massive black holes in the Horizon-AGN simulation PDF

18 Pages·2016·9.67 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The cosmic evolution of massive black holes in the Horizon-AGN simulation

MNRAS460,2979–2996(2016) doi:10.1093/mnras/stw1123 AdvanceAccesspublication2016May12 The cosmic evolution of massive black holes in the Horizon-AGN simulation M. Volonteri,1‹ Y. Dubois,1 C. Pichon1 and J. Devriendt2,3 1Institutd’AstrophysiquedeParis,UPMCetCNRS,UMR7095,98bisbdArago,F-75014Paris,France 2Sub-departmentofAstrophysics,UniversityofOxford,KebleRoad,OxfordOX13RH,UK 3ObservatoiredeLyon,UMR5574,9avenueCharlesAndre´,SaintGenisLavalF-69561,France Accepted2016May10.Received2016May9;inoriginalform2016February4 ABSTRACT Weanalysethedemographicsofblackholes(BHs)inthelarge-volumecosmologicalhydro- dynamical simulation Horizon-AGN. This simulation statistically models how much gas is accretedontoBHs,tracestheenergydepositedintotheirenvironmentand,consequently,the back-reactionoftheambientmediumonBHgrowth.ThesyntheticBHsreproduceavariety ofobservationalconstraintssuchastheredshiftevolutionoftheBHmassdensityandthemass function.Strongself-regulationviaAGNfeedback,weaksupernovafeedback,andunresolved internalprocessesresultinatightBH–galaxymasscorrelation.Startingatz∼2,tidalstrip- ping creates a small population of BHs over-massive with respect to the halo. The fraction of galaxies hosting a central BH or an AGN increases with stellar mass. The AGN fraction agreesbetterwithmulti-wavelengthstudies,thansingle-wavelengthones,unlessobscuration istakenintoaccount.ThemostmassivehaloespresentBHmultiplicity,withadditionalBHs gainedbyongoingorpastmergers.Insomecases,bothacentralandanoff-centreAGNshine concurrently, producing a dual AGN. This dual AGN population dwindles with decreasing redshift,asfoundinobservations.SpecificaccretionrateandEddingtonratiodistributionsare ingoodagreementwithobservationalestimates.TheBHpopulationisdominatedinturnby fast,slow,andveryslowaccretors,withtransitionsoccurringatz=3andz=2,respectively. Keywords: methods:numerical–galaxies:active–galaxies:evolution. signpostfortheunderlyingBHpopulation(e.g.Marconietal.2004; 1 INTRODUCTION Merloni2004;Shankaretal.2004;Hopkinsetal.2006;Merloni Compellingevidenceformassiveblackholes(BHs)existsforthe & Heinz 2008; Draper & Ballantyne 2012). Through a wealth of nucleioftensofnearbygalaxies(e.g.Kormendy&Ho2013,and observationsweknowthatBHsarepartoftheevolutionofcosmic referencestherein).Scalingrelationshavebeenidentifiedbetween structures,buthowtheyformed,grewandinteractedwiththeirhost the BH masses and the large-scale properties of the host galaxy, galaxiesisstillunclear. such as mass, luminosity, and velocity dispersion, pointing to a The joint evolution of BHs and galaxies within the evolving co-evolution between BHs and galaxies (e.g. Heckman & Kauff- cosmoscanbestudied,theoretically,throughdifferenttechniques: mann2011,andreferencestherein).Activegalacticnuclei(AGN), analytical,semi-analyticalandnumerical.Werecallherethesemi- poweredbyaccretingBHs,havebeensuggestedtobethemain‘reg- nalpapersbye.g.Small&Blandford(1992),Silk&Rees(1998), ulators’oftheBHgrowthitself,andalsoofthestarformationrate Haehnelt, Natarajan & Rees (1998), Haiman & Menou (2000), intheirhostgalaxies,atleastinmassivesystems(e.g.Silk2011, Cavaliere & Vittorini (2000) that first suggested, through analyt- andreferencestherein).TherelationshipbetweengalaxiesandBHs ical arguments, how BHs evolve cosmologically. Later on, semi- hasbeensuggestedtoholdthekeytounderstandingmassivegalaxy analyticalmodelshavebeensuccessfulinmatchingtheevolution evolution. of, e.g. the luminosity function (LF) of quasars (Kauffmann & AGNarealso‘lighthouses’atearlycosmictimes:throughmulti- Haehnelt2000;Cattaneo2001;Volonteri,Haardt&Madau2003) wavelength campaigns we can track the AGN population and its andthescalingbetweenBHmassandgalaxyproperties(Haehnelt evolution (e.g. Hopkins, Richards & Hernquist 2007; Silverman & Kauffmann 2000; Monaco, Salucci & Danese 2000; Cattaneo etal.2008;Rossetal.2013,andreferencestherein),anduseitasa etal.2005)andshowedhowtheinclusionofAGNfeedbackim- provedthematchbetweenpredictedandobservedgalaxymassand LFs(Boweretal.2006;Crotonetal.2006;Monaco,Fontanot& (cid:2)E-mail:[email protected] Taffoni 2007; Hirschmann et al. 2012). We refer the reader to (cid:3)C 2016TheAuthors PublishedbyOxfordUniversityPressonbehalfoftheRoyalAstronomicalSociety Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 2980 M.Volonterietal. Fontanotetal.(2011)forareviewoftheresultsofsemi-analytical andhighlighthowBHsgothroughdifferentmodesofaccretionas modelspertainingtoBHsandAGN. afunctionofcosmictime.InSection7wesummarizeourresults More recently, hydrodynamical cosmological simulations have andconclude. alsoincludedBHevolution.Self-consistentsimulationscapturethe anisotropic gas accretion and how galaxies interact during merg- 2 THE HORIZON-AGN SIMULATION ers,whichsemi-analyticalmodelscannotdo.Manystudies(Sijacki etal.2007;DiMatteoetal.2008;Booth&Schaye2009;Dubois TheHorizon-AGNsimulationisdescribedindetailinotherpapers, etal.2010,2012;Hirschmannetal.2014;Sijackietal.2015)have e.g. Dubois et al. (2014). We recall here the information relevant shownthattheinclusionofBHsas‘sinkparticles’thatareallowed forBHphysics,andfortheanalysisweperformed. toaccretegasfromtheirsurroundings,andproduceinturnenergy,a The Horizon-AGN simulation is run with the Adaptive Mesh fractionofwhichisreleasedinthehosthalo,reproducestheglobal Refinement code RAMSES (Teyssier 2002). We adopt a standard evolutionoftheBH+galaxypopulation.Thiscanappearsurprising, (cid:3)CDM cosmology with total matter density (cid:4) = 0.272, dark m ascosmologicalsimulationsarefarfrombeingabletoresolvethe energydensity(cid:4)(cid:3)=0.728,amplitudeofthematterpowerspec- scales over which accretion on BHs take place. Nevertheless, the trum σ = 0.81, baryon density (cid:4) = 0.045, Hubble constant 8 b statisticalpropertiesofAGNandBHsaresufficientlywellrepro- H =70.4kms−1Mpc−1, n = 0.967 compatible with Wilkinson 0 s duced.ThismayperhapsreflectthatmostofthegrowthoftheBHs MicrowaveAnisotropyProbe-7cosmology(Komatsuetal.2011). andtheeffectsontheirsurroundingsofAGNfeedbackarelinked ThesizeoftheboxisL =100h−1Mpcwith10243 DMparti- box toshortandintensephases,aspreviouslynotedinsemi-analytical cles,whichgivesaDMmassresolutionofMDM,res=8×107M(cid:4). studies. Fromthelevel10coarsegrid,acellisrefined(orunrefined)upto In this paper we analyse Horizon-AGN1 (Dubois et al. 2014), an effective resolution of (cid:6)x = 1 proper kpc (level 17 at z = 0) a large-volume cosmological hydrodynamical simulation, with a whenthemassinacellismore(orless)thaneighttimesthatofthe physicalresolutionof∼1kpc,runwiththeadaptivemeshrefine- initialmassresolution.Thesimulationincludesprescriptions,de- ment(AMR)codeRAMSES(Teyssier2002).Comparedtomostex- scribedinmoredetailsinDuboisetal.(2014),forbackgroundUV isting studies (e.g. Di Matteo et al. 2008; Booth & Schaye 2009; heating,gascoolingincludingthecontributionfrommetalsreleased Sijackietal.2015),therearethreeaspectswheretheRAMSESim- bystellarfeedback,starformationfollowingaSchmidtlawwitha plementationdiffers(butseeBellovaryetal.2010;Tremmeletal. 1percentefficiency(Rasera&Teyssier2006),andfeedbackfrom 2015),Theseare:(i)BHsarenotplacedinhaloesonlybasedon stellarwindsandTypeIaandTypeIIsupernovae(SNae)assuming ahalomassthreshold,butonthepropertiesofgasingalaxies;(ii) a Salpeter initial mass function (Dubois & Teyssier 2008; Kimm BH are not anchored to the centre of the host or advected when 2012). farfromthecentre,(iii)BHfeedbackhasadualmode,thermalat BHsarecreatedincellswherethegasandstellardensityexceed highaccretionratesandthroughacollimatedjetatlowaccretion the threshold for star formation, n =0.1Hcm−3, and where the 0 rates. These more realistic physical implementations allow us to stellar velocity dispersion, in that cell, is larger than 100kms−1, push forward in the comparison between theoretical models and withaninitialseedmassof105M(cid:4).Inordertoavoidtheformation observations,andmakeadditionalpredictionsonthelinkbetween ofmultipleBHsinthesamegalaxy,BHsarenotallowedtoform BHsandtheirhosts. at distances smaller than 50 comoving kpc from any other BH Aseriesofpapers(Duboisetal.inpreparation;Sugataetal.in particle.BHformationisstoppedatz=1.5.Allgalaxieswithmass preparation)haveanalysedthegalaxypopulationinHorizon-AGN, >1010M(cid:4)atz=0havealreadyformedbythatpoint,i.e.theyhave andshowedagoodagreementwithseveralobservables(mass/LFs, atleastoneprogenitorwhereaBHwouldhavebeenseeded. morphological mix, galaxy sizes, etc.). In this paper we validate We use the same ‘canonical’ accretion and AGN feedback theBHandAGNpopulationandtheBH–galaxycross-correlated modelling employed in Dubois et al. (2012). The accretion properties (BH–galaxy scaling relations). A resolution study for rate on to BHs follows the Bondi–Hoyle–Lyttleton rate M˙ = BH thisimplementationhasbeendoneinDuboisetal.(2012),from0.5 4παG2M2 ρ¯/(c¯2+u¯2)3/2,whereM istheBHmass,ρ¯ istheav- BH s BH to4kpcvaryingthespatialresolution,anddarkmatter(DM)mass eragegasdensity,c¯ istheaveragesoundspeed,u¯istheaveragegas s resolutionaswell.BH–galaxyscalingrelationsweresimilarinslope velocityrelativetotheBHvelocity,andαisadimensionlessboost andnormalization.However,atlowmasses,theBHpopulationwas factorwithα=(ρ/ρ )2whenρ>ρ andα=1otherwise(Booth 0 0 affected,andthereadershouldkeepthisinmind;wealsostressin &Schaye2009)inordertoaccountforourinabilitytocapturethe thebodyofthepapertheinstanceswhereHorizon-AGNisaffected colderandhigherdensityregionsoftheISM.Theeffectiveaccre- byresolutionissues.Wethenidentifythelimitswhereobservables tionrateontoBHsiscappedattheEddingtonluminositywithan arewellreproducedandfurtherproposeadditionalpredictionsand assumedradiativeefficiencyof(cid:9) =0.1fortheShakura&Sunyaev r observationalcomparisons. (1973)accretionontoaSchwarzschildBH. The outline of the paper is as follows. In Section 2 we briefly InordertoavoidspuriousoscillationsoftheBHinthegravita- reviewtheBHandAGNimplementation.InSection3wevalidate tionalpotentialwellduetoexternalperturbationsandfiniteresolu- thesimulationagainstseveralobservationalconstraints.InSection4 tioneffects,weintroduceadragforcethatmimicsthedynamical weconsidertheconnectionbetweenBHs,galaxiesandDMhaloes, frictionexertedbythegasontoamassiveparticle.Thisdynamical and instances where tidal stripping breaks the BH–host relations. frictionisproportionaltoF =f 4παρ(GM /c¯ )2,wheref DF gas BH s gas In Section 5 we discuss galaxies that host no BH, or more than isafudgefactorwhosevalueisbetween0and2andisafunctionof oneBHorAGN.InSection6wefollowtheupsanddownsofBH themachnumberM=u¯/c¯ <1(Ostriker1999;Chapon,Mayer s growth over cosmic time, analysing the distribution of accretion &Teyssier2013),andwhereweintroducetheboostfactorαforthe rates(inEddingtonunits,orperunitstellarmass),thedutycycle, samereasonsthanstatedabove.SeealsodiscussionsinTremmel etal.(2015),Lupi,Haardt&Dotti(2015),Gaboretal.(2016)on the limitations of gravity solvers for BH dynamics in AMR and 1http://www.horizon-simulation.org/about.html SPHsimulations. MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 BHsinHorizon-AGN 2981 TheAGNfeedbackisacombinationoftwodifferentmodes,the at the redshifts of interest falls between those proposed by Ueda so-calledradiomodeoperatingwhenχ =M˙ /M˙ <0.01and et al. (2014) and Shankar et al. (2009), thus marking a ‘middle BH Edd thequasarmodeactiveotherwise.Thequasarmodecorrespondsto ground’.WealsoadoptthemostrecenthardX-rayLF(2–10keV; anisotropicinjectionofthermalenergyintothegaswithinasphere Buchneretal.2015),whichincludesacorrectionforobscured(1022 of radius (cid:6)x, at an energy deposition rate: E˙ =(cid:9)(cid:9)M˙ c2, < N < 1024cm−2, where N is the equivalent neutral hydrogen AGN f r BH H H where (cid:9) = 0.15 for the quasar mode is a free parameter chosen columndensity)andComptonthickAGN(N >1024cm−2)–but f H to reproduce the correlations between BHs and galaxies and the estimatingthecontributionoftheelusiveComptonthickAGNre- BHdensityinourlocalUniverse(seeDuboisetal.2012).Atlow mainschallenging.Fotopoulouetal.(2016),whoestimatetheLF accretion rates, on the other hand, the radio mode deposits the at [5–10 keV], find no evidence of a rapid decline of the LF up AGNfeedbackenergyintoabipolaroutflowwithajetvelocityof to redshift four, consistently with Buchner et al. (2015). For this 104kms−1 into a cylinder with a cross-section of radius (cid:6)x and comparisonweconvertbolometricluminositiestoX-rayusingthe height2(cid:6)xfollowingOmmaetal.(2004)(moredetailsaboutthe bolometriccorrectionfromHopkinsetal.(2007)forconsistency. jetimplementationaregiveninDuboisetal.2010).Theefficiency To give an idea of the current uncertainties, we also include, at oftheradiomodeislarger,with(cid:9) =1. z(cid:2)2,theconversiontobolometricLFofthemid-infraredselected f WeidentifyDMhaloesandsub-haloesusingHaloMaker,which AGN LF proposed by Lacy et al. (2015), where they suggest a uses AdaptaHOP (Aubert, Pichon & Colombi 2004; Tweed et al. largerpopulationoffaintAGNexists,whichcanbemissedatother 2009),astructurefinderbasedontheidentificationofsaddlepoints wavelengths. inthe(smoothed)densityfield.Atotalof20neighboursareused TheresultingLFsatdifferentredshiftsareshowninFig.2.The tocomputethelocaldensityofeachparticle,andwefixthedensity brightendoftheLFisgenerallyingoodagreementwithobserva- thresholdat178timestheaveragetotalmatterdensity.Theforce tions,whilethefaintendisoverestimatedwhenincludingBHsinall softening(minimumsizebelowwhichsubstructuresareconsidered haloesabovethenominalresolutionof8×1010M(cid:4),i.e.resolved irrelevant)is∼2kpc.OnlyDMhaloesidentifiedwithmorethan bymorethan1000DMparticles(dashedcurves).Theoverestimate 50particlesareconsidered.Galaxiesareidentifiedusingthesame ofthefaintendoftheLFisacommonproblemincosmologicalsim- methodandsameparametersbutusingthestellarparticledistribu- ulations;forinstanceSijackietal.(2015)findasimilarbehaviour tioninsteadoftheDMone.BecauseofAGNfeedback,DMdensity intheIllustrissimulations.Theysuggestthatagoodagreementis profilesflatteninthecentre,andmisidentificationofthehalocentre reachedwhenincludingonlyBHswithmass>5×107 M(cid:4),and canoccurwhen itispurelydetermined bythedensestparticle in discussothersourcesofpossiblediscrepancywithdata.Wesuggest thehalo(Peiranietal.,inpreparation):asmallcuspedsub-halocan that the main source of the overestimate is due to SN feedback. beidentifiedasthecentreofthemainmoremassivecoredhalo.We Duboisetal.(2015)findthatstrongSNfeedback(usinga‘delayed usetheshrinkingsphereapproachproposedbyPoweretal.(2003) cooling’prescriptionfromTeyssieretal.2013)inlow-massgalax- togetthecorrecthalocentre.Fromafirstguessofthecentreusing ies(orhaloes)isabletolimitBHgrowthbySN-drivenoutflows, thedensestparticle,werecursivelysearchthecentreofmasswithin whichdepletethecentralregionofthegalaxyofgas.Asthegalaxy aradius10percentsmallerthantheradiusatthepreviousiteration anditshaloaccumulatemass,theybecomeabletoconfinenuclear (startingfromthevirialradius).Weendthesearchwhenthesphere inflows and the BH can start to grow. If we included only BHs hasasizesmallerthan2kpc,andtakethedensestparticlewithin withhalomass>5×1011M(cid:4)(solidcurves),whereBHsgrowthis thatsphereasthenewhalocentre. unimpededbySNfeedback,theagreementgreatlyimproves.Sim- In Horizon-AGN BHs are not anchored in the centre of DM ilarresultsareobtainedwithacutingalaxymass>3×1010M(cid:4) haloes, therefore we have to assign BHs to their host haloes and atthethresholdwhere,intheirhigher-resolutionsimulationswith galaxies.WeassignaBHtoahaloifitiswithin10percentofthe delayed coolingSNfeedback, Duboisetal.(2015)findthatBHs virialradiusoftheDMhaloandtoahalo+galaxystructureiftheBH become unaffected by SN feedback. Using a minimum threshold isalsowithintwicetheeffectiveradiusofthemostmassivegalaxy BHmassof>2–3×107M(cid:4)leadstoanalogousresultsfortheLF. withinthathalo.IfmorethanoneBHmeetsthecriteria,themost A direct comparison of BH growth under the effect of SN feed- massiveBHisdefinedasthecentralBH,andremovedfromthelist. backinacosmologicalvolumeissubjectofapaperinpreparation Westartfrommainhaloesandproceedhierarchicallythroughsub- (Habouzitetal.,inpreparation),butwestressthatthemasslocked haloes.SomeBHsarenotassignedtoanyhalo,subhalo,orgalaxy. in BHs hosted in low-mass haloes, <5 × 1011M(cid:4), is relatively Thesearetreatedas‘off-centre’BHsanddescribedinSection5.In unimportantfromthecosmologicalpointofview,asshowninthe Fig.1weshowamapofgasdensitywithBHpositionsoverlaid. followingsection.Additionally,simulationsmayoverproducethe high-redshiftAGNLF,butgiventheratherpoorobservationalcon- straintsathigh-zanduncertaintiesinthespectralenergydistribu- 3 COMPARISON WITH OBSERVATIONS tion,thediscrepancymayeventuallybereducedordisappearwhen Ourfirststepistovalidateoursimulation,i.e.todemonstratethat betterandmorecalibratedhigh-zLFsbecomeavailable.Thereis itproducesapopulationofBHscompatiblewiththerealUniverse. alsoapossibleoverestimateofthehigh-luminosityendatz<1.5, Forthischeck,weconsiderfourmaindiagnostics:theLFofAGN, butthestatisticsarepoor,andthisisseenonlyinthebolometricLF, theBHmassfunction,theBHmassdensityversusredshiftandthe notinthehardX-rayLF. relationbetweenBHandgalaxyproperties. 3.1 AGNluminosityfunction 3.2 BHmassdensityandmassfunction WecomparethetheoreticalLF((cid:11))totheobservationaldetermina- The BH mass density and its evolution over cosmic time are im- tionofthebolometricLF,ofwhichdifferentestimatesexist(Hop- portantobservationaldiagnostics.Thewaytheyarecalculatedob- kinsetal.2007;Shankaretal.2009;Uedaetal.2014).Wechoose servationallyisworthadiscussionpriortothecomparisonwiththe herethefunctionalformproposedbyHopkinsetal.(2007),which theoreticalresultsfromHorizon-AGN. MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 2982 M.Volonterietal. Figure1. ThepositionsofBHs,shownasdiffractionspikes,inasliceof200×142×50h−1Mpcareoverlaidonagasdensityandmetallicity(top)ordensity andtemperature(bottom)mapinHorizon-AGNatz∼1.5. Atz=0thetotalBHmassdensity,ρ ,isestimatedstartingfrom fraction)andtheprobabilitydistributionfunctionforthevelocity BH theempiricalcorrelationsbetweenBHmassandgalaxyproperties dispersion(Shankar2013,andreferencestherein).Inprinciple,this (normally,thevelocitydispersion,thebulgemassandluminosity). BH mass density is driven by the correlations found on the local TheBHmassdensityiscalculatedthroughtheconvolutionofthe sampleofquiescentBHs,lessthan150BHsatthetimeofwriting, chosen correlation with the distribution function of galaxies as a then extrapolated to the global population, including AGN (see functionofthatproperty(Yu&Tremaine2002).Thismeans,for Reines&Volonteri2015,andreferencestherein,foradiscussionon instance,thegalaxymassfunctionandLFs(correctedforthebulge thecaseofactiveBHs).Theshape,normalizationandscatterinthe MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 BHsinHorizon-AGN 2983 Figure 3. Total mass density locked in BHs (dark red), and BH mass densityinactiveBHs(lightred),definedasthosewithbolometricluminosity >1043ergs−1 versusredshift.Triangles:BHsinhaloeswithmass>8× 1010M(cid:4),squares:BHsinhaloeswithmass>5×1011M(cid:4)(seetextfor details).ThegreenhatchedregionshowsthelimitsprovidedbySoltan’s argument,asproposedbyBuchneretal.(2015)usingaradiativeefficiency of0.1,andscalingtheradiativeefficiencybetween0.06and0.32(upper andlowerhatchedregions,respectively).Theverticalbluelineindicatesthe z=0BHmassdensityestimatedbyShankaretal.(2004),andcorrectingthe BH-to-bulgerelationassuggestedbyKormendy&Ho(2013).Thelimits atz>6arederivedfromstackedhigh-zgalaxies(Willott2011;Cowie, Barger&Hasinger2012;Fioreetal.2012;Treisteretal.2013)orfrom theintegratedX-raybackground(Salvaterraetal.2012);Comptonthick sourcesarenotincludedintheselimits. 2012;Bongiornoetal.2014),andthereforelargeuncertaintiesexist. Typically,Soltan’sargument(Soltan1982)anditsupdates(Fabian &Iwasawa1999;Elvis,Risaliti&Zamorani2002;Yu&Tremaine 2002;Hopkins,Richards&Hernquist2007;Merloni2016)areused insteadtoestimatetheBHmassdensity.TheLFofAGNisinte- gratedovertime,startingfromagivencosmictimet fromwhich max theLFisknownsufficientlywell,andrescaledbya(fixed)radiative efficiency,(cid:9),toobtainamassdensity,ρ (z)≤ρ (z),ofmass BH,acc BH accretedonBHsasafunctionofredshift: (cid:2) (cid:2) Freidgsuhrieft2s..TBheoldoamsheetrdiccu(brvluees)inanclduhdaeradllXB-Hrasyi(n2H–1o0rizkoeVn-,AreGdN)LinFhaatldoiefsfewreintht ρBH,acc(z)= 1(cid:9)−c2(cid:9) tt(z)dt dLL(cid:11)AGN(L,t). (1) mass>8×1010M(cid:4),whilethesolidcurvesincludeonlyBHsinhaloeswith max mass>5×1011M(cid:4),whereSNfeedbackshouldnotquenchBHgrowth In general, the population of BHs in Horizon-AGN is in good (Dubois et al. 2015). The dotted light blue curve is the bolometric LF agreement with a variety of observations based on Soltan’s argu- proposedbyHopkinsetal.(2007)(seealsoShankar,Weinberg&Miralda- ment,fromhighredshifttoz=0(Fig.3),butwecanalsoderive Escude´2009;Uedaetal.2014).TheorangehatchedregionisthehardX-ray additional information and constraints. The density of mass ac- LFbyBuchneretal.(2015)correctedforComptonthinandthickAGN.The cretedonBHsisobviouslyalowerlimittothetotalmassdensityin greendottedcurveistheLFderivedfromaspectroscopicsurveyofAGN BHs,whichincludesquiescentonesandtheinitialmassin‘seeds’ selectedfromSpitzerSpaceTelescopeimagingsurveys(Lacyetal.2015). (seeVolonteri2010;Volonterietal.2013;Merloni2016),plusob- scuredAGN(Merloni2016).ThisisexemplifiedinFig.3,where relationshipsplayanimportantrole.Forinstance,Kormendy&Ho wedistinguishthetotalBHmassdensity(darkred)andthemass (2013)suggestthattheratiobetweenBHandbulgemasspreviously densityinaccretingBHsintheAGNphase(bolometricluminosity usedinbenchmarkcalculations(e.g.Marconietal.2004;Shankar >1043ergs−1,lightred).Thelatterquantityisthetruemassdensity et al. 2004) should be increased, based on their updated sample, inaccretingBHsatagivenredshift,ratherthanthetotalluminos- leadingtoatotalmassdensityhigherbyafactorofalmost2. ityrescaledbytheradiativeefficiency((cid:9) =0.1inHorizon-AGN) Atz>0theshapeandnormalizationoftheBH–galaxycorrela- and the speed of light, integrated over time. This quantity allows tionsaresubjectofdebate(e.g.Pengetal.2006;Jahnkeetal.2009; us to immediately see the relevance of the bright AGN phases in Merloni 2010; Cisternas et al. 2011; Targett, Dunlop & McLure growing BHs. At high redshift, z > 2, most of the BH mass is MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 2984 M.Volonterietal. forbroad-linequasarsandAGN(Type1,Kellyetal.2010;Kelly & Shen 2013; Schulze et al. 2015), which are only the subset of quasars for which broad lines can be seen, i.e. with a favourable viewing angle and unobscured. Obscuration may be produced by the geometry of the torus, and depend only on viewing angle; in thiscaseitseffectcanbeaccountedforstatistically.Obscuration, however,mayalsobecausedbymaterialinthehostgalaxy,which ishardertoaccountfor.WeincludeinFig.4,atz=1andz=2,the estimatebySchulzeetal.(2015)atz=1.5basedonaconvolution oftheobservationalstellarmassfunction(Ilbertetal.2013)with theBH–bulgemassrelationofMcConnell&Ma(2013),without applyinganybulge-totalmassconversion.Wealsoincludethemass function proposed by Merloni & Heinz (2008) obtained by mod- ellingtheBHpopulationthroughacontinuityequationderivedfrom usingtheredshiftevolutionoftheLFandtheBHmassfunctionat z=0asconstraints.Theagreementbetweenthesimulationandthe observationsisgood;thedisagreementatlowBHmasses(<3× 107M(cid:4))atz=1–3issimilartowhatwefindfortheAGNLF.It isusefultocontrasttheBHmassfunctiontothestellarmassfunc- tioninhorizon-AGN(Kavirajetal.,inpreparation).Thesimulated galaxiesreproducewelltheobservationalstellarmassfunctionat z=0–5,withonlyanoverestimateof∼0.3dexforstellarmasses Figure 4. Mass function of BHs in Horizon-AGN at different redshifts. <3 × 1010M(cid:4). This overestimate is somewhat smaller than the Weindicatewiththehatchedregionstherangeproposedbyestimatesat z=0(Laueretal.2007b;Shankar2013,verticalanddiagonalhatching, onewefindintheBHmassfunctionatz=1–3atBHmasses<3× respectively),and,athighredshift,therangeproposedatz=1.5bySchulze 107M(cid:4). et al. (2015), shown as grey dashed curves at z = 1 and z = 2, and by FromFig.4wecanalsoaddressthequestionofdowntowhich Merloni&Heinz(2008)atz=1–5,shownwithhorizontalblackhatching. BH mass the simulation is reliable. The closer the BH mass is to the resolution, the more results are resolution dependent. One growninluminousAGN,whileatlowerredshiftthemassgrowth approachistoproceedlikeinthecaseofgalaxiesandderivethe isdrivenbyfaintAGN(bolometricluminosity<1043ergs−1).This minimalmassfromthecompletenessoftheBHmassfunction.The figurealsoshowsthatBHsinlow-masshaloes(<5×1011M(cid:4))are argumentisasfollows:thereisnoexplicitscalebreakinginourBH relativelyunimportantinthecosmicpictureasonecanseebycom- model,thereforethereisnoreasonforthenumberofBHsnottoin- paringtrianglesandsquares,whichincludeBHsinhaloeswithmass creasewiththenumberofhostgalaxies/haloes.Consequently,any >8×1010M(cid:4) or>5×1011M(cid:4),respectively.Finally,thetotal decreaseinthemassfunctionofBHatthelow-massendcouldbe massdensityinseedsis(cid:2)103M(cid:4)Mpc−3,similartowhatispre- attributedtolackofresolutioninthesimulation.Applyingthisar- dicted by analytical and semi-analytical models of BH formation gumenttheminimalBHmassis2×107M(cid:4).However,whilethere (Volonteri 2010), which means that accretion accounted for basi- isnoexplicitscalebreakinginourBHmodel,thereisanimplicit cally the whole mass density we find at z = 0, corresponding to threshold on the minimum gas density required for BH seeding, ∼8.3×105M(cid:4)Mpc−3. whichdisfavourslow-massgalaxies.Therefore,itdoesnotneces- Horizon-AGNreproduceswelltheBHmassdensity,butthisdoes sarilyfollowthatweshouldnottrustBHwithmasses<107M(cid:4), notnecessarilyensurethatitreproducestheBHmassfunction.The thisissayingthatthissampleisnotcompleteinthesimulationso BHmassdensityistheintegraloftheBHmassfunction,anditis ourconclusionsmightbebiasedinthatregime.Wefurtherdiscuss dominatedbyBHsatitsknee,thereforereproducingthemassden- therangeofvalidityofthesimulationinthenextsection. sitydoesnotgiveinformationonlow-andhigh-massBHs.InFig.4 weshowtheBHmassfunction(reddots),andattemptacomparison 3.3 CorrelationsbetweenBHsandgalaxiesatz∼0 withobservationalestimates,bothatz=0andathighredshift.The comparisoniscomplicatedbytryingtomatchthetheoreticalsample One of the most common ways to determine whether simulated toobservations.We referthe reader tothecomprehensive review BHsarerealisticistocomparetheirmassestopropertiesofthehost by Kelly & Merloni (2012) for a discussion on the difficulties in galaxiesthathavebeenshowntocorrelatewithBHmasses.Indeed, determiningtheBHmassfunctionobservationally,andtherelated inmostcasestheparametersofBHaccretionandfeedbackaretuned uncertainties. We discuss in the following the main problems of toreproducethesecorrelations(e.g.Sijackietal.2007;DiMatteo relevance for our comparison here. Regarding the comparison at etal.2008;Booth&Schaye2009;Duboisetal.2012).Inreality,a z = 0, we use the results from Lauer et al. (2007b) and Shankar ‘fair’comparisonisfarfromsimple.Ontheonehand,thesampleof (2013)(seealsoShankaretal.2004).Asnotedpreviouslyforthe galaxieswithdynamicalBHmassmeasurementsisnotanunbiased massdensity,theestimatesoftheBHmassfunctionhingeonacon- representation of the whole galaxy population. Most BH masses volutionofthecorrelationsbetweenBHandgalaxypropertieswith pertain to massive, dense elliptical galaxies, while disc galaxies thedistributionfunctionofgalaxiesasafunctionofthatproperty. andlow-velocitydispersiongalaxiesareunder-represented,aswell Thefunctionalformandtherangeofvalidityofthesecorrelations as low-mass galaxies (van den Bosch et al. 2015), with the rare are subject of debate, and therefore a source of uncertainty. For exception of megamasers (Greene et al. 2010). Additionally, the instance,theslopeoftheBHmass–velocitydispersioncorrelation definitionofthegalaxypropertyofinterest(e.g.bulgemass,velocity and the normalization of the BH–bulge mass one are still uncer- dispersion,galaxymass)isnotconsistentintheliterature,andthe tain.Athigherredshift,themassfunctionhasbeenestimatedonly analysesareheterogeneous. MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 BHsinHorizon-AGN 2985 themassesofsimulatedgalaxieshavebeenshiftedby−0.33dex (seethediscussioninReines&Volonteri2015).Atthehigh-mass end, M > 108 M(cid:4) and M > 1011 M(cid:4), simulations and ob- BH gal servationsareinrelativelygoodagreement,butlow-massBHsand low-mass galaxies are more problematic. Specifically, BHs with M <5×107 M(cid:4) inintermediate-massgalaxies(1010 M(cid:4) < BH M <1011 M(cid:4))arealmostabsent.AsdiscussedinSection3.1, gal thismaybecausedbythelackofstrongSNfeedbackinHorizon- AGN,whichallowsBHsinlow-massgalaxiestogrowmoreeasily. SNfeedbackwouldaffectonlyBHsinlow-massgalaxies.Oncea galaxyreachesamassof∼1010 M(cid:4)thentheBHcangrowunim- peded and self-regulation ensues, thus, for instance, in the simu- lationsbyDuboisetal.(2015)thehostgalaxypropertiesdidnot depend on SN feedback after z ∼ 3. In the presence of delayed coolingSNfeedbackthebulgegrowthwassuppressedatz>3.Af- terwards,whenthegalaxymassreachedafew×1010 M(cid:4)thepres- enceoftheAGNregulatedthebulgemassindependentlyoftheSN implementation.Inanalogywiththis,wearguethat,withsmaller BHs, causing reduced AGN feedback in small and intermediate- massgalaxies,thefractionofsimulatedbulge-dominatedgalaxies at stellar mass (cid:2) 1011 M(cid:4) would decrease. This would improve Figure5. BHversustotalstellarmassofthegalaxyfor10000random thematchwiththeobservedmorphologicalmixinthatmassrange gEadldaixnigestoantzra=tio0>(g0re.0y1)..DTahrekogrraenygsequlianreesm:BarHksswMiBthHL=bol2<×101404−e3rMggsa−l.1Tanhde (3D.2u,bboeilsowetBaHl.,minaspsreesp2ar×ati1o0n7).MA(cid:4)dd(ictioonnsaelqlyu,enastlyn,ogteadlaxinymSeacstsio(cid:2)n blacklineisthebest-fittinglinearcorrelation.Thedarkergreenpentagons 1010M(cid:4))resolutioneffectsbecomeimportant. arelow-luminosityAGNfromReines&Volonteri(2015),andthelighter greentrianglesarethequiescentBHsfromtables2and3inKormendy&Ho In Fig. 6 we turn to BH mass versus bulge mass. This is one (2013)withstellarmassesrevisedbyReines&Volonteri.Thestellarmasses of the ‘classical’ scalings that seem to have small scatter in the ofHorizon-AGNgalaxieshavebeenshiftedby−0.33dextoaccountforthis observations(<0.3dex),atleastforclassicalbulges,andwhichis correction.Ifweselectforlow-luminosityAGN,theirBHsoccupyaregion widelyextrapolated,alsoathigh-redshift,andalsowhenbulge/disc inthelower-rightcorneroftheBH–galaxydistribution,reminiscentofthe decompositionsarenotavailable,tostudytheco-evolutionofBHs regionwhereinobservationswetendtofindlow-luminosityAGN.However, and galaxies. We start by highlighting that indeed the measure- thescatterinthesimulatedsampleismuchlessthanintheobservations(see ment of a univocal bulge mass is tricky. In our simulations we textfordetails). candefineeitherakinematicalbulgemass,wherebulgeparticles arethosewithatangentialvelocitycomponent(incylindricalco- Westartwiththegalaxypropertythatiseasiertodefineandmea- ordinates with the spin of the galaxy as the axis of symmetry) sureinbothsimulationsandobservations,thetotalstellarmassof smallerthanthenon-tangentialvelocitycomponent,orperforma thegalaxy,althoughwhetherthetotalstellarmassisagoodpredictor bulge/discdecomposition,witheithertwoSe´rsicprofiles,onewith oftheBHmass,andatightcorrelationcanbedefined,isamatterof n = 1 for the disc, and one with n = 4 for the bulge, or two debate(Kormendy&Ho2013;La¨skeretal.2014).Recently,Reines Se´rsicprofiles,bothwithn=1,oronewithn=1,andonewith &Volonteri(2015)re-analysedinasmuchconsistentawayaspos- n=[1,4],keepinginmindthatourphysicalresolution,1kpc,lim- siblethesampleoflocalBHswithdynamicalmassmeasurements itsourabilitytoperformthedecompositioninsmallorlow-mass (Kormendy&Ho2013),andalsoincludeBHsinAGNwithmass galaxies. measuredthroughreverberationmapping(Bentz&Katz2015),as The difficulties of defining a bulge mass is apparent in Fig. 6 wellasasampleofbroad-lineAGNswithsingle-epochBHmass andtheyarealsodiscussedinappendixAofDuboisetal.(2012) estimatesselectedfromtheNASA-SloanAtlas,whichisbasedon for simulations and by La¨sker et al. (2014) for observations (see the Sloan Digital Sky Survey Data Release 8 spectroscopic cata- alsoLaskeretal.2015).Wenotethatthebulgemassesaretaken logue(Aiharaetal.2011).TheyfoundthattherelationbetweenBH from Kormendy & Ho (2013), therefore, when the galaxy mass andstellarmassdefinedbylocalmoderate-luminosityAGNinrel- corresponds to the bulge mass, they are larger by 0.33 dex than ativelylow-massgalaxies(pentagonsinFig.5)hasanormalization the stellar masses shown in Fig. 5, because of the difference in thatislowerbyapproximatelyanorderofmagnitudecomparedto zero-point described by Reines & Volonteri (2015). Despite the theBH–bulgemassrelation,andtotherelationobtainedonmas- intrinsic difficulties, the match between Horizon-AGN BHs and sive,bulge-dominatedgalaxies(trianglesinFig.5),whichappears the observational sample is good, where the comparison can be alsotobesteeper. performedmeaningfully(thiswastobeexpected,aswereproduce Wedefineasgalaxymassthetotalstellarmass.Thestellarmass correctlytheBHmassfunctionthatisobservationallydetermined is the sum of all star particle masses within a region meeting a throughthiscorrelation,aswellasthestellarmassfunctions;see densitythresholdcriterion,setat178timestheaveragetotalmat- Kavirajetal.,inpreparation).Thelogarithmicslopesforthevarious ter density. Only galaxies identified with more than 50 particles fits,includingonlyobjectswithM >109M(cid:4),varyfrom1.05 bulge are considered. (See Section 2 for details.) Fig. 5 reports 10 000 forthekinematicalbulgemass,0.75forthetwoSe´rsicprofileswith randomBHs+galaxiesfromHorizon-AGNatz=0(squares).We n=1andn=4,1.0forboththetwoSe´rsicprofileswithn=1, notethattheBHparametersinRAMSEShadbeenoriginallychosen andn=1andn=[1,4]. toreproduceolderdataandtrends(Duboisetal.2012),therefore,to In summary, the tighter BH–bulge mass relation is well re- makethecomparisonwiththeobservationalsampleself-consistent, produced in Horizon-AGN, while the broader BH–stellar mass MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 2986 M.Volonterietal. the simulations to reproduce a correlation that is tight only for a specificclassofgalaxies,thereistheriskoffailingtoobtainthe fullviewoftheBH/galaxypopulation.Onewaytoseethisisthat relativelylowmassBHsinintermediate-massgalaxiesthatarenot bulge-dominated,high-velocitydispersiongalaxies,arenotpresent inthesimulation. Alternatively,wecouldseethisasaproblemwithgalaxyprop- erties.Galaxieswithstellarmassessuchthattheyshouldtypically host ∼107 M(cid:4) BHs instead harbour BHs with mass 10 times or morelarger.Atthesametimethesimulationreproduces,however, the observed BH mass versus bulge mass. The simulation may therefore over-predict bulge masses and bulge-to-disc ratios. We have compared our simulated galaxies to distributions of bulge- to-total ratios, Krajnovic´ et al. (2013) for early-type galaxies and Gadotti(2009)fordiscgalaxies,andBlucketal.(2014)foramor- phologicalmix.Simulatedgalaxiesunder-estimatethefractionof purebulges,butover-estimatebulge-to-totalratiosforgalaxieswith mass1010(cid:2)M (cid:2)1011M(cid:4).Westress,however,thatthisshould gal beconsideredaqualitativecomparison,asobservationalbulge–disc decompositionssufferfromthesameproblemsthatweencountered indefininguniquelyabulgemass,plusissuesrelatedtoinclination, surfacebrightnesslimits,spatialresolutionandsignal-to-noiseratio (seee.g.thediscussioninSimardetal.2011). Althoughmergerhistoriesandenvironmentaleffectsinducesome scatter in mock BH–galaxy relations, internal galactic processes, such as gas turbulence, variation in the compactness, which are notcapturedbecauseoflimitedresolution,mayalsoinfluencethe scatter.Thescattergivenbydifferentmergerhistoriesarisingfrom cosmological initial conditions appears insufficient to explain the observational scatter, especially in M at fixed M , therefore BH gal these unresolved internal processes are likely to be an important contributiontothescatter.Asimilarbehaviourcanbeseeninother cosmologicalhydrodynamicalsimulations(figs4and10inSijacki etal.2015;Schayeetal.2015,respectively). We end by studying the BH mass versus stellar velocity dis- persion, σ, in Fig. 7. This is also a benchmark scaling, and it is consideredtobeoneofthebestpredictorsforBHmasses,i.e.to haveasmallintrinsicscatter(<0.3dex).Thisisperhapsthemost problematicforHorizon-AGN.Ifwefitasinglelogarithmicslope totheBHmassversusstellarvelocitydispersion,wefindavalue of4.02forgalaxieswithσ >60kms−1,comparedto4.38found byKormendy&Hoonaculledsubsampleofthegalaxiesshown inFig.7.Themainreasonisthatvelocitydispersionsinthesimu- Figure6. BHversusbulgemassfor10000randomgalaxiesatz=0(grey). lationareunderestimatedatMbulge <1011M(cid:4) (σ <200kms−1), Thebulgemassismeasuredeitherkinematicallyorthroughabulge/disc andoverestimatedabove,whencomparingtotheobservedFaber– decompositionwithtwoSe´rsicprofiles,eitherbothwithn=1,oronewith Jackson relation (Bernardi et al. 2003). The dearth of BHs with n=1,andonewithn=4,oronewithn=1,andonewithn=[1,4].We M <5×107 M(cid:4) inintermediate-massgalaxieshighlightedin BH reportalsoallBHsfromtables2and3inKormendy&Ho(2013),using thediscussionofFig.5alsocontributes,butgiventhegoodagree- bluesymbolsforthesubsetusedforfittingtheBHmassversusbulgemass mentintheM versusM relation,itisaminorcontribution, BH bulge relation. for this specific comparison. One contribution to the discrepancy isthatgalaxysizesinthesimulationarelargerthanforobserved scaling(s) are in worse agreement. The light green triangles in galaxies with mass (cid:2) 1011M(cid:4). In fact, the galaxy sizes are in Fig. 5 are almost entirely the same points that appear in Fig. 6. goodagreementwithobservedlow-masslate-typegalaxies,butare If we had restricted our analysis to this subsample, composed of larger, by a factor of ∼ 3–4, than observed low-mass early-type galaxies with dynamically measured BH mass, which tend to be galaxies(Duboisetal.,inpreparation).Sincethecompletesample denseandbulgedominated,wewouldhaveagoodmatchbetween isamixoflate-andearly-typegalaxies,thereapp√earstobeanover- simulationandobservationsalsoinFig.5,i.e.inBH–galaxymass. allmismatchatσ <200kms−1ofafactorof∼ 2inthevelocity Infact,oursimulationandallothersimilarsimulationshavebeen dispersion (roughly corresponding to an average overestimate of tunedtomatchtheBH–stellar(bulge)masscorrelationforthesub- the galaxy size of a factor of 2). The velocity dispersion is also sampleofgalaxiesthatworkswell(greentrianglesinFig.5,blue sensitivetotheDMhalocentralprofile,whichdependsnotonlyon symbolsinFig.6).Suchgalaxies,however,areabiasedsubsam- DMhalomassbutalsoonasubtleinterplaywiththebaryonsand ple of the real universe (see also Shankar et al. 2016). By tuning AGNfeedback.Indeed,thecentralDMdensitiesinHorizon-AGN MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 BHsinHorizon-AGN 2987 Figure 7. BH mass versus bulge velocity dispersion at z = 0 for non- interactinggalaxies(grey).Only10000randomgalaxiesareshown.We reportalsoallBHsfromtables2and3inKormendy&Ho(2013),using bluesymbolsforthesubsetusedbytheauthorsforestablishingtheBHmass versusbulgevelocitydispersionrelation.Low-massearly-typegalaxiesin Horizon-AGNtendtohavelargersizethanrealones,thereforethevelocity dispersionisunderestimatedforthesegalaxies. are lower than for the Horizon-noAGN simulation, which differs only in that AGN are not included in the latter. We defer a more detailed study quantifying the exact contribution of this effect as a function of galaxy mass and assembly history to another paper (Peiranietal.,inpreparation). 4 BHS IN HALOES AND SUBHALOES: THE EFFECT OF STRIPPING SomeBHsinlocalgalaxiesappeartohavemassesmuchlargerthan expectedonthebasisofthestellarorbulgemass(e.g.NGC4486B, NGC1277,M60-UCD1,NGC4342,NGC4291),andapossibility isthatthehostgalaxywasoriginallymoremassiveandlostsomeof itsmassbecauseofstripping(Volonteri,Haardt&Gu¨ltekin2008b; Sethetal.2014),althoughthisisunlikelyatleastforsomecases (NGC4342,NGC4291;Bogda´netal.2012). Figure8. BHmassversushalomass(top)andversusgalaxymass(bottom) Weinvestigateherehowstrippinginsub-haloesaffectsthecon- atz=0.Red:mainhaloes;orange:sub-haloes.Contoursarebasedon10 equallyspacedlogarithmiclevels,droppingthefivelowestlevelsforclarity. nectionwithDMhaloesandstellarmasses.InthetoppanelofFig.8 AtlatecosmictimesapopulationofBHsinstrippedsub-haloesemerges, weshowhowsub-haloestendtohostlargerBHs(atfixedDMhalo wheretheBHisover-massiveatagivenhalomass.Inreality,thehalois mass)comparedtohaloesatz=0.Thisisaneffectthat,although under-massive,havinglostpartofitsmass.Stripping,however,isseldom presentalsoearlier,buildsupwithcosmictime.Onlyveryseldom, effectiveallthewaytothestellardistributionofgalaxies.Wehighlightwith however,strippingiseffectiveallthewayintothestellardistribution greensquaresthefewgalaxieswheretheBHmassis>0.01×Mgal.All andcreatesover-massiveBHs(atfixedstellarmass).Normally,the suchgalaxiesresideinsub-haloes. ratiobetweentheBHandgalaxymassisbetween0.002and0.005 forellipticalgalaxiesandbulges(Marconi&Hunt2003;Ha¨ring& >0.01×M ,andthisfractionincreasesonlybyafactorof3ifwe gal Rix2004;Kormendy&Ho2013),andcanbe20–60timeslower restrictthesampletogalaxieslocatedwithinsub-haloes.Athigher formoreactiveormorediscygalaxies(Greeneetal.2010;Reines redshift,over-massiveBHsare0.10percentofthefullpopulation &Volonteri2015).WedefineaBHtobeover-massiveifitsmass atz=3andz=2,and0.07percentatz=1.Restrictingthesample is>0.01×M . tosubhaloes,thepercentageschangeto0.08,0.11and0.17atthe gal The few subhaloes with over-massive BHs are shown as green threeredshifts.Atearlytimesmostover-massiveBHsarehostedin squares. Although we note that there are no main haloes hosting normalhaloes:BHsandgalaxiesarestillgrowing–andnotalways over-massive BHs based on the definition given above, the cases exactlyatthesamepace. wheretheBHisover-massivebecauseofstrippingarefew.Only This is short by an order of magnitude compared with obser- 0.2percentoftheentiregalaxypopulationhaveaBHwithmass vations, where out of ∼ 135 galaxies with dynamical BH mass MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018 2988 M.Volonterietal. measurements,approximatelyeightgalaxieshavebeensuggested to be over-massive at 3 − σ above the BH–bulge mass correla- tion(Ferre´-Mateuetal.2015,butseeSavorgnan&Graham2016). Bogda´n et al. (2012) proposed that over-massive BHs where the hosthasnotbeenstrippedcanbeexplainedbynon-standardpaths of BH–galaxy co-evolution (see also Fabian et al. 2013; Trujillo etal.2014;Ferre´-Mateuetal.2015).Someoftheproposedpaths includeasequenceofaccretioneventsdominatedbylowangular momentum gas, boosting BH growth, and causing star formation tooccurinacompactconfiguration.OncetheBHhasgrownsuffi- ciently,AGNfeedbackpreventedadditionalBHandgalaxygrowth. Ina similarvein, thesegalaxies andtheirBHssimplydonotex- perienceanygrowthbymergersafterz∼3,remainingstructurally untouchedandwithoutfurthermassandsizeincrease,butrecord- ingatimewhen,statistically,theratiobetweenBHandbulgemass waslargerthantoday.Keepinginmindthatscatterinthesimulated BH–galaxyrelationsislessthaninobservations,wehavelooked forBHsthatare‘almost’over-massive,>0.004×M ,andindeed gal foundsomecaseswheretheBHisnothostedinasub-halo,andthe galaxygrowthhasnearlystoppedatz∼2–3.Thesecasesremain veryfew,at10percentofthe‘almost’over-massiveBHs,whichin Figure 9. Occupation fraction of BHs as a function of halo mass at turnrepresentthe2.5percentofthefullpopulation. z = 0 (top) and z = 3 (bottom). We separate main haloes (red, thin Insummary,averysmallfractionofBHsappearover-massive curves)andsub-haloes(orange,thickcurves).Solidlinesrepresentcentral withrespecttothestellarmassoftheirhostbecausethelatterhas BHs, i.e. BHs located within 10 per cent of the virial radius. The occu- beenstripped,beingasub-haloofalargerhalo.However,allthe pationfractionofcentralBHsreachesunityathalomasses∼1010M(cid:4)at BHs with M > 0.01 × M we identified in our simulation at z=3and∼1011M(cid:4) atz=0,thustracingthemassgrowthofhaloes. BH gal z = 0 are in sub-haloes (see also Barber et al. 2016). If we are Theoccupationfractionofsubhaloes,atagivenmass,ishigherthanthat moregenerousinourdefinitionofover-massive,>0.004×M , ofhaloesbecauseofstripping:theinitialhalomasswaslargerthanatthe gal thenalargerpopulationemerges,includingalsosomecaseswhere presenttime,thereforetheOFtracestheinitialhaloproperties(cf.Fig.8). Dashedlinesrepresentoff-centreBHswhicharelocatedwithinthevirial tidal stripping is not the culprit. To put in context the fraction of radiusofagivenhaloandarenotassociatedwithanysub-halo.Tensof over-massiveBHsinthesimulationandinobservations,itisalso off-centreBHscanbefoundinthemostmassivehaloes. importanttoremarkthatmany(ifnotmost)galaxiesforwhichdy- namicalBHmassmeasurementsexistarelocatedingalaxyclusters wheretidalstrippingandgalaxy–galaxyinteractionsareexpected theexpectationthatlow-massgalaxieshavemoretroubleforming tobemore important(especially inthe centre ofgalaxy clusters) andkeepingBHs,andfeedingthem. thaninasimulationspanningawiderangeofenvironmentssuchas The OF of BHs as a function of halo mass is shown at two theonepresentedhere. differentredshifts(z=0andz=3)inFig.9.Allhaloeswithmass >1011M(cid:4)atz=0and>3×1010M(cid:4)atz=3hostacentralBH. TheOFishigheratlowhalomassforsub-haloesforreasonssimilar 5 THE OCCUPATION FRACTION OF tothosediscussedintheprevioussection.Beforethehalobecame CENTRAL AND OFF-CENTRE BHS a substructure, it was more massive, and this is the information InmassivegalaxieswhereaBHhasbeenlookedfor,onewasfound. recordedbytheOF. However, the situation differs for low-mass galaxies. Faint AGN Therearealsohaloesthathostoff-centreBHs.Tofindoff-centre arenowbeingfoundinmanylow-massgalaxies(Reines,Greene& BHs, we first assign all the central BHs, within 0.1 × R , and vir Geha2013;Lemonsetal.2015;Milleretal.2015;Paggietal.2015; remove them from the list. We then consider the BHs which do Trump et al. 2015; Mezcua et al. 2016), but there are only upper notbelongascentralstoanyhalo,andassignanoff-centreBHto limits for BHs in NGC 205 (Valluri et al. 2005), M33 (Gebhardt ahaloifitiswithinR .Westartfromsubstructuresandproceed vir et al. 2001), Fornax (Jardel & Gebhardt 2012). The fraction of thentomainhaloes.Theseoff-centreBHsaregeneratedbymergers haloes or galaxies hosting BHs as a function of galaxy mass and duringthehierarchicalassemblyofhaloes.Someofthemarethe properties(BHoccupationfraction,OF)isanimportantdiagnostic resultofarecentmerger,andtheyareontheirwaytojoiningand tolearnaboutthefirstBHs,inawaysimilartonear-fieldcosmology coalescingwiththecentralBH,andmayhaveveryhighluminosity, and the first galaxies (Volonteri, Lodato & Natarajan 2008a; van throughmerger-drivengasinflows.Inhaloeswithmasses3×1012 Wassenhoveetal.2010).ThefractionofgalaxieshostingAGNasa < M < 3 × 1013M(cid:4) these merger-driven off-centre AGN are h functionofgalaxymassandpropertiesisalower-limittotheBHOF, locatedwithin10kpcfromthehalocentre,andhaveluminosities andadditionallyprovidesinformationonAGNfuelling.InHorizon- ∼1043 − 1047 erg s−1. Other off-centre BHs are stranded in the AGNBHseedinghasbeenperformedinarelativelysimpleway,not outerpartsofhaloesandhaveverylowluminosities.Inhaloeswith directlybasedonmodelsofBHformationintheearlyUniverse,and masses3×1012 <M <3×1013M(cid:4) mostoff-centreBHsare h ithasbeenstoppedatz=1.5.ThechoiceofhaltingBHformation located at ∼100 kpc from the halo centre and have luminosities atthisredshiftdoesnotaffectourconclusions,asallwell-resolved of∼1036ergs−1.TheluminositiesarelowbecausetheseBHsare galaxiesattheendofthesimulationshaveatleastoneprogenitor surrounded by hot, low-density gas. By z = 0 the most massive atz=1.5.AlthoughwecannotderiveconstraintsonBHformation haloescanhosttensoftheseBHs,asshownwithdashedlinesin scenarios,wecantestqualitativelythetrendswithgalaxymass,i.e. Fig.9. MNRAS460,2979–2996(2016) Downloaded from https://academic.oup.com/mnras/article-abstract/460/3/2979/2609412 by Said Business School user on 30 January 2018

Description:
The cosmic evolution of massive black holes in the Horizon-AGN simulation observations we know that BHs are part of the evolution of cosmic and they are also discussed in appendix A of Dubois et al. (2012) for simulations and by Läsker et al. (2014) for observations (see also Lasker et al. 2015)
See more

The list of books you might like

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