Mon.Not.R.Astron.Soc.000,1–7(2008) Printed2February2008 (MNLATEXstylefilev2.2) Why are AGN found in High Mass Galaxies? Lan Wang1,2⋆, Guinevere Kauffmann2 1Department of Astronomy, Peking University,Beijing 100871, China 2Max–Planck–Institut fu¨r Astrophysik, Karl–Schwarzschild–Str. 1, D-85748 Garching, Germany 8 0 0 Accepted 2008??????.Received2008??????;inoriginalform2008?????? 2 n a ABSTRACT J We use semi-analytic models implemented in the Millennium Simulation to analyze 3 the merging histories of dark matter haloes and of the galaxies that reside in them. 2 We assumethat supermassiveblackholes onlyexistingalaxiesthathaveexperienced atleastonemajormerger.Onlyafewpercentofgalaxieswithstellarmasseslessthan h] M∗ <1010M⊙arepredictedtohaveexperiencedamajormergerandtocontainablack p hole. The fraction of galaxies with black holes increases very steeply at larger stellar - masses. This agrees well with the observed strong mass dependence of the fraction of o nearbygalaxiesthatcontaineitherlow-luminosity(LINER-type)orhigher-luminosity r t (Seyfert or composite-type) AGN. We then investigate when the major mergers that s a first create the black holes are predicted to occur. High mass galaxies are predicted [ to have formed their black holes at very early epochs. The majority of low mass galaxiesneverexperienceamajormergerandhencedonotcontainablackhole,buta 1 significant fraction of the supermassive black holes that do exist in low mass galaxies v are predicted to have formed recently. 0 3 Key words: galaxies: interactions – galaxies: haloes – galaxies: nuclei 5 3 . 1 0 8 1 INTRODUCTION during major mergers of galaxies (Kauffmann & Haehnelt 0 2000; Wyithe& Loeb 2003; Croton et al. 2006). By studying active galactic nuclei (AGN), we learn about : At low and moderate redshifts, there is no conclusive v thephysicalmechanismsthattriggeraccretionontothecen- observational evidence that mergers play a significant role i tralsupermassiveblackholesofgalaxies.Whenablackhole X intriggeringAGNactivityingalaxies.InthelocalUniverse, accretes, it increases in mass. By studying populations of r AGNat low and at high redshifts, we hopetoinfer thehis- Li et al. (2006) have shown that narrow line AGN do not a havemore close companions than matched samples of inac- tory of how black holes build up their mass. tivegalaxies. Evenatintermediateredshifts(z∼0.4−1.3), It has been established that supermassive black holes moderate luminosity AGN hosts do not have morphologies mostoccuringalaxieswithbulges(Kormendy& Richstone indicativeofanongoingmergerorinteraction(Hasan2007). 1995), and that the mass of the black hole correlates with The conclusion seems to be that although major mergers the luminosity and the stellar velocity dispersion of the mayberesponsibleforAGNactivityinsomegalaxies,other hostbulge(Magorrian et al.1998;Ferrarese & Merritt2000; fueling mechanisms are likely to be most important in the Gebhardt et al. 2000). This indicates that the formation lowredshiftUniverse.Ithasalsobeenestablishedthathigh of galaxies and supermassive black holes are likely to be mass black holes havelargely stopped growing at early cos- closely linked. In the local Universe, the fraction of bulge- mic epochs, whereas low mass black holes are still accret- dominated galaxies hosting AGN decreases at lower stel- ing at significant rates today (Heckman et al. 2004). X-ray lar masses (Hoet al. 1997; Kauffmann et al. 2003). In or- observations show that very high-luminosity AGN activity der to form a black hole, it is necessary for gas to lose peakedatearlycosmicepochs(z∼2),whilelow-luminosity angular momentum and sink to the centre of the galaxy AGN activity peaks at lower redshifts (Steffen et al. 2003; (Haehnelt & Rees1993;Volonteri et al.2003).Thegravita- Barger et al. 2005; Hasinger et al. 2005). tional torques that operate during galaxy-galaxy mergers It has been postulated that this so-called “anti- are known to be a very effective mechanism for concentrat- hierarchical” growth of supermassive black holes can be inggasatthecentersofgalaxies(Mihos & Hernquist1996). explained if there are two modes of accretion onto black Models for AGN evolution have often assumed that black holes that have very different efficiencies (Merloni 2004; holes areformed andfuelled, and AGNactivity istriggered Mueller & Hasinger2007).Theearlyformationformationof “new” black holes may result in very luminous quasar-like ⋆ Email:[email protected] events. To form a supermassive black hole, a more violent 2 L.Wang, G.Kauffmann process such as a major merger may be required to funnel canrepresenttheFOFhalo,whilesubhaloreferstosubstruc- a large amount of gas into the central region of the galaxy. ture other than the main one. Halo merger happens when Subsequentaccretionofgasontoalreadyexistingblackholes twoFOFgroup mergeintoonegroupandoneof thehaloes may be an inefficient process and produce lower luminosity becomes a subhalo of thelarger structure. AGN(Haehnelt & Rees1993;Duschl& Strittmatter2002). The substructure merger trees form the basic in- The history of accretion after the black hole is formed may put to the semi-analytic model used to associate galax- not necessarily be tightly linked to the dynamical history ies with haloes/subhaloes (DeLucia & Blaizot 2007). The of the galaxy, but may be controlled by the accretion and semi-analytic galaxy catalogue we are using in this study feedbackprocessesoccurringinthevicinityoftheblackhole is publicly available. A description of the publicly available itself. catalogues, and a link to the database can be found at the Inthiswork,weusethecombinationoftheMillennium webpage: http://www.mpa-garching.mpg.de/millennium/. Simulationandsemi-analyticmodelsofgalaxyformationto Once a halo appears in the simulation, a (central) galaxy study the fraction of galaxies that have undergone major beginstoformwithinit.Thecentralgalaxyislocatedatthe mergers as a function of mass and cosmic epoch. We inves- position ofthemostboundparticleofthehalo.Asthesim- tigate whether this can be related to the demographics of ulation evolves, thehalo may merge with a larger structure black holes in the local Universe and to the apparent dis- and become a subhalo. The central galaxy then becomes a appearanceofthemostluminousquasaractivityinmassive satellitegalaxyinthelargerstructure.Thegalaxy’sposition galaxies at late times. andvelocityarespecifiedbythepositionandvelocityofthe In Sec. 2, we briefly introduce the simulation we use most bound particle of its host halo/subhalo. Even if the and explain how galaxy mergers are tracked in the simula- subhalohostingthegalaxyistidallydisrupted,theposition tion. In Sec. 3, we show that if we assume that black holes and velocity of the galaxy is still traced through this most onlyformwhengalaxiesundergomajormergingevents,then boundparticle.Galaxiesthusonlydisappearfromthesimu- mostpresent-daylowmassgalaxiesarepredictednotnotto lationiftheymergewithanothergalaxy.Thetimetakenfor containblackholesandhencewill nothostAGN.InSec.4, a galaxy without subhalo to merge with the central object we use the simulations to predict when galaxies of different isgivenbythetimetakenfordynamicalfrictiontoerodeits masseshaveunderdonetheirfirstmajormerger.Conclusions orbit,causingit tospiral intothecentreand merge.This is and discussions are presented in thefinal section. calculated using the standard Chandrasekhar formula. All theinformationabouttheformationandmerginghistoryof galaxies is stored. Byanalyzingthesehaloandgalaxymergertrees,weare 2 SIMULATION AND MERGER TREES able to track when two haloes merge together and whether The Millennium Simulation(Springel et al. 2005) is used in the galaxies within them also merge into a single object by this work to study the merging histories of dark matter thepresentday.Inthisstudy,wefocus onmergersbetween haloes. The merging histories of galaxies can be inferred satellite and central galaxies, and exclude mergers between whenthesimulation iscombinedwithsemi-analyticmodels two satellites. These events are rare (Springel et al. 2001) thatfollowgascooling,starformation,supernovaandAGN andneglectingthemshouldnotaffectourconclusionsabout feedbackandotherphysicalprocessesthatregulatehowthe theincidence and fueling of black holes in galaxies. baryonscondense into galaxies. TheMillenniumSimulationfollowsN =21603particles ofmass8.6×108h−1M⊙fromredshiftz=127tothepresent 3 HALO AND GALAXY MERGERS day, within a comoving box of 500h−1Mpc on a side. The cosmological parametersvaluesinthesimulationareconsis- Inthisstudy,weassumethatblackholesformwhenagalaxy tent with the determinations from a combined analysis of undergoesamajor mergingevent.Galaxies thathavenever the2dFGRS(Colless et al.2001)andfirstyearWMAPdata experienced a major merger do not have a black hole. We (Spergel et al. 2003). A flat ΛCDM cosmology is assumed define major mergers as events in which the mass ratio of with Ωm = 0.25, Ωb = 0.045, h = 0.73, ΩΛ = 0.75, n = 1, thetwoprogenitorsisgreaterthan0.3.Forhalomerger,the and σ8 =0.9. mass ratio is the virial mass ratio of two progenitor haloes. Full particle data are stored at 64 output times. For Forgalaxymerger,itisthestellarmassratiooftwoprogen- each output, haloes are identified using a friends-of-friends itor galaxies. When we track mergers in the simulation, we (FOF) group-finder. Substructures (or subhaloes) within a includemajormergersthatoccurinallbranchesofthetree, FOF halo are located using the SUBFIND algorithm of not just the“main branch”. Springel et al.(2001).Theself-boundpartoftheFOFgroup Sincegalaxiesresideindarkmatterhaloesandareable itselfalsoappearinthesubstructurelist.Thismainsubhalo tomergeonlyoncetheirhosthaloeshavecoalesced,webegin typicallycontains90percentofthemassoftheFOFgroup. byanalyzingthemerginghistoriesofthedarkmatterhaloes After finding all substructures in all the output snapshots, themselves. In the left panel of Fig. 1, we plot the average subhalo merging trees are built that describe in detail how numberofmajormergersapresentdaydarkmatterhalohas thesesystemsmergeandgrowastheuniverseevolves.Since experiencedoveritslifetimeasafunctionofhalomass.Note structures merge hierarchically in CDM universes, for any that in this analysis we track mergers down to an effective given subhalo, therecan be several progenitors, but in gen- resolution limit of 20 particles, which corresponds toahalo eraleachsubhaloonlyhasonedescendant.Mergertreesare ofmass1.7×1010h−1M⊙.Weseethatthenumberofmajor thus constructed by defining a unique descendant for each mergers(abovetheresolutionlimit)experiencedbyahalois subhalo. Werefer below halo tothemain substructurethat a strongly increasing function of mass; haloes with present- AGN in High Mass Galaxies 3 Figure 1. Left panel: the average number of major mergers that a dark matter halo of given mass has experienced over its lifetime. Rightpanel:thefractionofhaloesofgivenmassthathavehadatleastonemajormerger. Figure 2. Left panel: The relation between the stellar mass of the central galaxy and the the mass of its host dark matter halo as predictedbythesemi-analyticmodelsofDeLucia&Blaizot(2007).Theerrorbarsindicatethe95percentilerangeinstellarmassata given value of Mhalo . Middle panel: The solid lineshows the fraction of dark matter haloes that have experienced at least one major merger as afunction of the stellarmass of the central galaxy. The dotted lineshows the fraction ofcentral galaxies ofgiven mass that havehadatleastonemajormerger.Rightpanel:Thesolidlineshowstheaveragenumberofmajormergersexperiencedbyadarkmatter haloasafunction ofthestellarmassofitscentral galaxy. Thedashedlineshows theaverage numberofmajormergersexperienced by thecentral galaxyitself. day masses of 1012M⊙ have typically experienced only one haveplotted the relation between the stellar mass of a cen- one major merger, whereas the progenitors of present-day tral galaxy and the mass of its host halo in the left panel haloeswithmassesof1015M⊙ havemergedwitheachother of Fig. 2, as predicted by the semi-analytic models we use close to 100 times. in this study (DeLucia & Blaizot 2007). This mean rela- tion can be used to transform between central galaxy mass In the right panel, we show the fraction of haloes that and halo mass in an approximate way (this conversion ne- have had at least one major merger during their lifetime, glects scatter between the two quantities and the fact that asafunction of halomass. Thefraction of haloes thathave some galaxies are actually satellite systems). If thefraction had major mergers also increases rapidly with halo mass. Almost all haloes more massive than 1013h−1M⊙ have had ofgalaxieswithmajormergersfollowed therelationderived for theirhost haloes, this would yield thesolid curvein the atleast onemajormergerandthisfraction dropstoaround 20 % for haloes with masses of around 1011h−1M⊙. middle panel of Fig. 2. Why are the merging histories of galaxies and their host haloes so different? We now investigate the fraction of galaxies that have had major mergers. The results are shown as a dotted line Once two dark matter haloes merge, the galaxies in- in the middle panel of Fig. 2. Rather than rising steeply side them will merge together over a timescale that is de- as a function of mass, thegalaxy major merger fraction re- terminedbydynamicalfriction.Uponinvestigation,wefind mains close to zero up to a stellar mass of 1010.5M⊙ and that nearly all galaxies that have experienced major merg- then rises sharply. This is somewhat surprising in view of ers are located in dark matter haloes that have also expe- the behaviour of the same quantity for dark matter haloes, rienced a major merger. There are almost no galaxy major plotted in the right-hand panel of Fig. 1. For reference, we mergers that have occurred in a halo that has only experi- 4 L.Wang, G.Kauffmann itor haloes often correspond to minor mergers between the progenitor galaxies. Howcanweunderstandthis?Duringtheperiodoftime betweenthemergerofthetwohaloesandthemergerofthe galaxies within them,thestellar mass ofthesmaller “satel- lite” galaxy remains about the same because ongoing star formation isquenchedwhenthegassurroundingthegalaxy is shock–heated and no longer cools onto the satellite. The centralgalaxy,however,willcontinuetoincreaseinmassas aresultofcoolingandstarformation.Thestellarmassratio of two galaxies therefore becomes smaller as a function of time. This is illustrated in Fig. 4. For every merging event that occurs over the history of a galaxy, we record stel- lar mass ratio information at the time when the progenitor haloes merge and at the time when thegalaxies themselves merge together. For simplicity, we keep information for one randomlychosenmergingeventinthehistoryofeachgalaxy. In the left panel of Fig. 4, we plot the average time that elapses between the time when the two haloes merged and the time when the galaxies themselves merged. Results are shownasafunctionofgalaxystellarmassandthetheerror barsindicate68percentilerangeinthedistributionofdelay times.Ascanbeseen,thetypicaldelaytimeisaround2Gyr, Figure3. Thethicksolidline(HMM)showsthefractionofcen- but individual time delays can range between 1 and 5 Gyr. tralgalaxieswhoseprogenitorhaloeshavehadatleastonemajor The delay times are typically shorter for the progenitors of merger. The other lines split this sample of central galaxies ac- more massive galaxies. cordingtothehistoryofthecentralgalaxyitself.Thedottedline In the right panel of Fig. 4, we plot the average stellar (GnoM) shows the contribution from central galaxies that have mass ratios of the galaxies at the time when their haloes not experienced a mergers of any kind. The dashed line (GmM) merge (solid curve) and at the time when the two galaxies shows the contribution from central galaxies that have experi- themselvesmerge(dashedline).Noticethatthestellarmass encedonlyminormergers.Thethinsolidline(GMM)showsthe contribution from central galaxies that have experienced major ratio can sometimes be larger than 1; this happens when majors. thegalaxy insidethesmaller halo ismore massive than the galaxy in the larger halo. As we expect, the mass ratio of galaxiesatthetimewhenthegalaxiesmergeissmallerthan that it is at the time when the haloes merge. This effect is enced a minor merger (∼ 0.15 percent). However, the con- somewhat larger for the mergers that give rise to the most verse is not true; we find that a substantial fraction of halo massive galaxies at the present day. major mergers give rise to galaxy minor mergers. This is illustrated in the right-hand panel of Fig. 2. The solid line 3.1 Comparison with Observations showsthenumberofmajormergersexperiencedbythepro- genitorhaloesofapresent-daycentralgalaxyasafunctionof In this section we have seen that the fraction of galaxies itsmass.Thedashedlineshowsthenumberofmajormerg- that haveexperienced one or more major merging eventsis ersexperiencedbytheirprogenitorgalaxies.Ascanbeseen, predicted to very close to zero at stellar masses less than the number of major mergers experienced by the progeni- ∼1010M⊙,butaverysteeplyrisingfraction ofstellar mass tor galaxies is an order of magnitude smaller. Notice that forM∗ >1010M⊙.Wenowcomparethispredictionwiththe the number of galaxy mergers is less than 1 for galaxies up fractionofSloanDigitalSkySurveygalaxiesthatcontainan to∼1011h−1M⊙,andincrease steeplyfor massivegalaxies. AGN.Werestrict theSDSSsample toredshifts z <0.06 so This is in nice agreement with what is shown in Fig.9 of that we are still able to detect AGN with weak line emis- DeLucia et al. (2006),which shows thenumberof effective sion(LINERs).AsshownbyKauffmann et al.(2003),weak- progenitors as a function of the stellar mass for elliptical linedAGNbecomeprogressivelymoredifficulttoidentifyat galaxies. higher redshifts using SDSS spectra. This is because these In Fig. 3, we again plot the fraction of central galaxies spectra are obtained through 3 arcsecond diameter fibre of a given mass whose progenitor haloes have had a ma- apertures and the contribution from the stellar population jor mergers (thick solid line). The thin solid line shows the of the host galaxy becomes increasingly dominant in more fractionwhoseprogenitorgalaxieshavehadamajormerger. distant galaxies. Thedashedlineshowsthefractionofsuchgalaxiesthathave Theresultsof thecomparison areshown in Fig. 5.The had minor mergers and the dotted line is the fraction that black curve shows the fraction of galaxies in the Millen- have had no merger of any kind. The main conclusion that niumSimulationofgivenstellarmassthathavehadatleast can be gleaned from this plot is that the reason why the one major merger. The red histogram shows thefraction of thick solid and thin solid curves differ in shape, is because galaxies in the SDSS survey that are classified as AGN. As at lower stellar masses, major mergers between the progen- can beseen, both fractions rise steeply from values close to AGN in High Mass Galaxies 5 Figure4. Somecharacteristicsofthemergingeventswiththehigheststellarmassratiosthattakeplaceduringthehistoryofagalaxy: Left:theaveragetimedelaybetweenthetimethattheprogenitorhaloesmergeandthetimethatthecentralgalaxiesmerge.Right:the stellarmassratioofthegalaxiesatthetimethatthehaloesmerge(solidline)andatthetimewhenthecentralgalaxiesmerge(dashed line).Allresultsareplottedasafunctionofthestellarmassofthecentralgalaxyanderrorbarsshowthe68percentiledispersionaround themeanvalue. 0.4. Compared with the solid line where we use 0.3 as the mass ratio threshold to define a major merger, the increas- ingtrendsareaboutthesamefordifferentthresholdsinthe range from 0.2 to 0.4. 4 FIRST BLACK HOLES Inthissection,weanalyzewhenthefirstmajormergerthat producestheblack hole in thegalaxy is predicted tooccur. In Fig. 6, we plot the distribution of the times of the first major merging events for galaxies with different present- day stellar masses. The vertical dashed lines indicate the median values of the distributions. The dotted bar in each panel indicates the fraction of galaxies in each stellar mass binthathavenotexperiencedamajormergerandarehence not included in thedistribution of merging times. As can be seen, massive galaxies experience their first major merging event at earlier epochs than less massive galaxies.Almostnonewblackholesforminmassivegalaxies atthepresentday.Thedistributionofblackholeformation timesinlowmassgalaxies ismuchflatter.Ifthebulkofthe black hole mass is built up in a short period following the firstmajormerger,thiswouldexplainwhypresent-daymas- Figure5. Thesolidcurveshowsthefractionofgalaxiesofgiven siveblackholeshavestoppedgrowing,whilelowmassblack stellar mass that are predicted to have experienced at least one holes are still growing at a significant rate (Heckman et al. major merger. The red histogram shows the fraction of SDSS 2004). galaxies with z < 0.06 that are classified as AGN. The dotted We now assume that the black holes formed from the anddashedlinesshowtheresultsfromsimulationwhenthemass ratio threshold for defining major merger is changed to 0.2 and firstmajormergersofgalaxiescanshineandbeobservedfor 0.4. 107 years.Bycountingthenumbersofsucheventsatdiffer- ent redshifts, we can compute the evolution in the number density of newly formed black holes. The result is plotted as diamonds in Fig. 7. The comoving number density of zero at M∗ < 1010h−1M⊙ to nearly unity at stellar masses such events peaks at redshift of z ∼ 2−3, consistent with greater than 1011h−1M⊙. In Fig. 5, the dotted and dashed the observed peak in the number density of bright quasars lines show the results from simulation when the mass ratio (Richardset al. 2006). The decrease in the number density threshold for defining major merger is changed to 0.2 and of newly formed black holes to high redshifts is less pro- 6 L.Wang, G.Kauffmann Figure 6. Distributionoftimewhenagalaxyexperiences itsfirstmajormergerforgalaxiesindifferentstellarmassbins(solidlines). Galaxy stellar massis plotted inunits of h−1M⊙. Thevertical dashed linesshow the medianvalue of the distributions.In each panel, thedotted barshowsthefractionofgalaxiesthathavenever experiencedamajormerger. nounced than that found by Fan et al. (2001), who show thattheluminousquasardensitydecreasesbyafactorof∼6 from redshift 3.5 to 5. Note that we have not attempted to model thepredicted luminosity of thequasars in this work, so a direct comparison with the observations is not possi- ble. As we have discussed, it is well possible that processes otherthan mergers contributeto thelow-luminosity quasar population. 5 CONCLUSIONS We analyze the merger histories of dark matter haloes and galaxiesintheMillenniumSimulationanduseourresultsto trytounderstandthedemographicsofblackholesinnearby galaxies. Black holes are assumed to form only if a major merger occurs. Although a significant fraction of low mass (< 1010M⊙) galaxies have experienced minor mergers, less thanafewpercentarepredictedtohaveexperiencedamajor merger. If our assumption that a major merger is required in order to form a black hole is correct, the majority of low mass galaxies are predicted not to contain black holes at the present day. This is one possible explanation of the Figure 7. The number density of galaxies experiencing their observed lack of AGN in low mass galaxies (Ho et al. 1997; firstmajormerger(diamonds)isplottedasafunctionofredshift. Eachmergerisassumedtobevisiblefor107 years. Kauffmann et al. 2003). We also investigate when galaxies of different stellar AGN in High Mass Galaxies 7 masses are predicted to have formed their first black holes. Hasan P., 2007, ArXive-prints, 707 Highmassgalaxies form theirfirstblackholesat veryearly Hasinger G., Miyaji T., Schmidt M., 2005, A&A,441, 417 epochs. The distribution of formation times is almost flat Heckman T. M., Kauffmann G., Brinchmann J., Charlot as a function of lookback time for low mass galaxies. This S.,Tremonti C., White S.D. M., 2004, ApJ, 613, 109 means that if a low mass galaxy has a black hole, there HoL.C.,FilippenkoA.V.,SargentW.L.W.,1997,ApJS, is a significant probability that it formed in the last few 112, 315 Gigyears. We also compute the number density of newly Hopkins P. F., Hernquist L., Cox T. J., Di Matteo T., formedblackholesasafunctionofredshift.Wefindthatthe Robertson B., Springel V., 2006, ApJS,163, 1 peaknumberdensityoccursatz∼2−3,ingoodagreement Kauffmann G., Haehnelt M., 2000, MNRAS,311, 576 withtheobservedpeakinthequasarspacedensity.Morede- KauffmannG.,HeckmanT.M.,TremontiC.,Brinchmann tailed predictions forhow AGNofdifferentluminosities are J.,CharlotS.,WhiteS.D.M.,RidgwayS.E.,Brinkmann expected to evolve requires a more detailed physical model J., Fukugita M., Hall P. B., Ivezi´c Zˇ., Richards G. T., for how the black holes accrete gas over the history of the SchneiderD. P., 2003, MNRAS,346, 1055 Universe. In addition, in certain wavebands AGN activity Kormendy J., RichstoneD., 1995, ARA&A,33, 581 might be obscured by gas and dust surrounding black hole Li C., Kauffmann G., WangL., White S.D.M., Heckman (Hopkinset al. 2006). More detailed consideration of these T. M., Jing Y. P., 2006, MNRAS,373, 457 issues will form thebasis for future work. MagorrianJ.,TremaineS.,RichstoneD.,BenderR.,Bower G.,DresslerA.,FaberS.M.,GebhardtK.,GreenR.,Grill- mair C., KormendyJ., Lauer T., 1998, AJ, 115, 2285 Merloni A., 2004, MNRAS,353, 1035 ACKNOWLEDGEMENTS Mihos J. C., Hernquist L., 1996, ApJ, 464, 641 We are grateful to Zuhui Fan, Gabriella De Lucia, Roderik Mueller A., Hasinger G., 2007, ArXiv e-prints,708 Overzier and Qi Guo for their detailed comments and sug- Richards G. T., Strauss M. A., Fan X., Hall P. B., Jester gestionsonourpaper.LanWangwouldliketoacknowledge S.,SchneiderD.P.,VandenBerkD.E., StoughtonC., et thesupports from NSFCundergrants 10373001, 10533010, al., 2006, AJ, 131, 2766 and 10773001, and 973 Program (No. 2007CB815401). The Spergel D. N., Verde L., Peiris H. V., Komatsu E., Nolta simulation used in this paper was carried out as part of M. R., Bennett C. L., Halpern M., Hinshaw G., et al., theprogrammeoftheVirgoConsortiumontheRegattasu- 2003, ApJS,148, 175 percomputerof theComputing Centre of theMax–Planck– Springel V., White S. D. M., Jenkins A., Frenk C. S., Society in Garching. Yoshida N., Gao L., Navarro J., Thacker R., et al., 2005, Nature,435, 629 ThispaperhasbeentypesetfromaTEX/LATEXfileprepared Springel V., White S. D. M., Tormen G., Kauffmann G., by theauthor. 2001, MNRAS,328, 726 SteffenA.T.,BargerA.J.,CowieL.L.,MushotzkyR.F., YangY., 2003, ApJ, 596, L23 Volonteri M., Haardt F., Madau P., 2003, ApJ, 582, 559 REFERENCES WyitheJ. S.B., Loeb A., 2003, ApJ, 595, 614 BargerA.J.,CowieL.L.,MushotzkyR.F.,YangY.,Wang W.-H.,Steffen A. T., Capak P., 2005, AJ, 129, 578 CollessM.,DaltonG.,MaddoxS.,SutherlandW.,Norberg P., Cole S., Bland-Hawthorn J., Bridges T., et al., 2001, MNRAS,328, 1039 Croton D. J., Springel V., White S. D. M., De Lucia G., Frenk C. S., Gao L., Jenkins A., Kauffmann G., Navarro J. F., YoshidaN., 2006, MNRAS,365, 11 De Lucia G., Blaizot J., 2007, MNRAS,375, 2 DeLuciaG.,SpringelV.,WhiteS.D.M.,CrotonD.,Kauff- mann G., 2006, MNRAS,366, 499 Duschl W. J., Strittmatter P. A., 2002, in Green R. F., Khachikian E. Y., Sanders D. B., eds, IAU Colloq. 184: AGN Surveys Vol. 284 of Astronomical Society of the Pacific Conference Series, The Formation and Feeding of Massive Black Holes in theEarly Universe. pp 343–+ FanX.,StraussM.A.,SchneiderD.P.,GunnJ.E.,Lupton R.H.,BeckerR.H.,DavisM.,NewmanJ.A.,etal.,2001, AJ, 121, 54 Ferrarese L., Merritt D., 2000, ApJ, 539, L9 Gebhardt K., Bender R., Bower G., Dressler A., Faber S.M.,FilippenkoA.V.,GreenR.,GrillmairC.,HoL.C., Kormendy J., Lauer T. R., Magorrian J., Pinkney J., RichstoneD., Tremaine S.,2000, ApJ, 539, L13 Haehnelt M. G., ReesM. J., 1993, MNRAS,263, 168