Table Of Content

MNRAS000,1–5(2016) Preprint20January2016 CompiledusingMNRASLATEXstylefilev3.0 The formation of massive, quiescent galaxies at cosmic noon Robert Feldmann1(cid:63), Philip F. Hopkins2, Eliot Quataert1, Claude-André Faucher-Giguère3, and Duˇsan Kereˇs4 1Department of Astronomy, University of California, Berkeley, CA 94720-3411, USA 2TAPIR 350-17, California Institute of Technology, Pasadena, CA 91125, USA 3Department of Physics and Astronomy and CIERA, Northwestern University, Evanston, IL 60208, USA 6 4Center for Astrophysics and Space Sciences, University of California, San Diego, CA 92093, USA 1 0 2 AcceptedXXX.ReceivedYYY;inoriginalformZZZ n a J ABSTRACT 8 The cosmic noon (z ∼ 1.5−3) marked a period of vigorous star formation for most 1 galaxies.However,aboutathirdofthemoremassivegalaxiesatthosetimeswerequi- escent in the sense that their observed stellar populations are inconsistent with rapid ] A star formation. The reduced star formation activity is often attributed to gaseous outflows driven by feedback from supermassive black holes, but the impact of black G holefeedbackon galaxies intheyoung Universeisnotyet definitivelyestablished.We . analyze the origin of quiescent galaxies with the help of ultra-high resolution, cosmo- h p logical simulations that include feedback from stars but do not model the uncertain - consequencesofblackholefeedback.Weshowthatdarkmatterhaloswithspecificac- o cretion rates below ∼0.25−0.4 Gyr−1 preferentially host galaxies with reduced star r t formation rates and red broad-band colors. The fraction of such halos in large dark s matter only simulations matches the observed fraction of massive quiescent galaxies a (∼1010−1011 M ). This strongly suggests that halo accretion rate is the key param- [ (cid:12) eter determining which massive galaxies at z ∼ 1.5−3 become quiescent. Empirical 1 models that connect galaxy and halo evolution, such as halo occupation distribution v or abundance matching models, assume a tight link between galaxy properties and 4 the masses of their parent halos. These models will benefit from adding the specific 0 accretion rate of halos as a second model parameter. 7 4 Key words: galaxies: formation – galaxies: evolution – galaxies: high-redshift 0 . 1 0 6 1 INTRODUCTION choicetotestthisproposedpicture,butuntilnownoexist- 1 : ingsimulationproducedbothagalaxysampleoftheneces- v There is mounting evidence that star formation in galaxies sary size for statistical analysis and properly resolved and i is tied to the accretion of gas from intergalactic distances X modeled the relevant physical processes that take place in (e.g., Dekel et al. 2009; Cresci et al. 2010; Lilly et al. 2013; galaxies at z ∼ 1.5−3. In addition, cosmological, hydro- r SańchezAlmeidaetal.2014;Martinetal.2014;Rodriguez- a dynamical simulations have a long history of struggling to Puebla et al. 2015; Brisbin et al. 2015; Narayanan et al. reproduce key properties of observed galaxies such as their 2015).Hence,reducedgasaccretionontogalaxiescouldpo- typicalstellarmassesandstarformationrates(Scannapieco tentiallyberesponsibleforthereducedstarformationrates et al. 2012). The present study, MassiveFIRE, overcomes (SFRs)ofquiescentgalaxies(Feldmann&Mayer2015).The thesechallengesbyadoptingtheaccuratephysicalmodeling reducedsupplyofgastogalaxiesandhaloswouldalsomake of the Feedback In Realistic Environments (FIRE) project it easier for additional processes, e.g., black hole feedback (Hopkins et al. 2014) and by applying it, for the first time, (e.g., Ciotti & Ostriker 1997; Di Matteo et al. 2005; Si- to a population of massive galaxies. jacki & Springel 2006; Teyssier et al. 2011; Kormendy & Wehavesimulatedtheevolutionof35massivegalaxies Ho 2013), to fully suppress any remaining star formation for the first 4 billion years after the Big Bang (until red- activityinquiescentgalaxies(Dekel&Birnboim2006;Cat- shiftz≥1.67).Thegalaxysampleisextractedfrom17dis- taneo et al. 2006). Numerical simulations are the tool of tinct sub-regions, each containing at least one dark matter (DM)halowithamassintherange3×1012−3×1013 M , (cid:12) (cid:63) E-mail:[email protected](RF) embedded in a representative volume of the Universe. The (cid:13)c 2016TheAuthors 2 R. Feldmann et al. sub-regions sample the full range of cosmological assembly 1012 MassiveFIRE histories of halos harboring massive galaxies, i.e., galaxies z=2 lwaittihonsstealrlaerrmunaswseitshlatrhgeerhythdarondaybnoaumtic1s01a0ndMg(cid:12)r.avOituyrssoimlvuer- 1011 zz==59 GIZMO(Hopkins2015)atultra-highspatial(∼10pc)and High resolution Medium resolution mass resolution (m = 3.3×104 M at high resolution, 10 gas (cid:12) 10 Centrals 2.7×105 M(cid:12) at medium resolution) in P-SPH mode. The Satellites high numerical resolution allows us to model reliably many 9 olafrtmheedrieulmevaonftgparlaoxceiesss.esSttehlalatrtfaekeedbpalcakceprinoctehsseesinstuecrhsteals- M]sun10 energy and momentum injection from supernovae, stellar M[star108 winds, photo-heating, and radiation pressure interact in a non-linearmanner(Hopkinsetal.2012)andareallincluded 7 inoursimulationswithlittlerelianceontunableparameters. 10 Feedback from supermassive black holes (SMBH) is not in- cluded. We will present the setup and methodology of our 6 Moster et al. 2013 simulations in more detail in Feldmann et al. in prep. 10 z=2 z=5 We showed in previous work that the computational z=9 approach of this paper reproduces well the integral prop- 105 9 10 11 12 13 14 erties of lower mass galaxies (M∗ (cid:46) 3×1010 M ) since 10 10 10 10 10 10 (cid:12) M [M ] cosmicnoon.Forinstance,wereportedontheSHMR(Hop- vir sun kins et al. 2014), on the stellar mass – metallicity relation Figure1.Stellar-to-halo-massrelation(SHMR)ofMassiveFIRE (Ma et al. 2015), and on the properties of galactic outflows galaxies at z = 2 (green circles), z = 5 (purple circles), and drivenbystellarfeedback(Muratovetal.2015)findinggood z = 9 (cyan circles). Central (satellite) galaxies at each redshift agreementwithavailableobservations.Wealsoshowedthat areshownbyfilled(empty)circles,andlarge(small)circlesdenote theH coveringfractionsin(cid:46)1012M halosatcosmicnoon galaxiessimulatedathigh(medium)numericalresolution.Dotted I (cid:12) match observations (Faucher-Giguère et al. 2015). We will linesshowanempiricalestimate(Mosteretal.2013)oftheSHMR reportcorrespondingpropertiesofMassiveFIREgalaxiesin anditsextrapolationtohighredshiftsandlowstellarmasses.The 1−σ scatter of individual galaxies above and below the mean upcoming work. relation is about 0.2 dex (Reddick et al. 2013) (shaded region). SatellitegalaxiestendtolietotheleftoftherelationastheirDM 2 RESULTS AND DISCUSSION halosareoftentidallystripped.MassiveFIREgalaxieshavestellar massesinfairagreementwiththeempiricallyderivedSHMR. Figure1comparesthestellarmassesofMassiveFIREgalax- ies with those of actual galaxies that reside in DM halos of similarmass.StellarmassesofMassiveFIREgalaxiesagree, servations (Lee et al. 2013). In contrast, quiescent galaxies to within a factor of ∼ 2, with the empirically inferred es- often have an early type morphology with a more compact timateofthestellar-to-halo-massrelation(SHMR)(Moster stellar distribution (van der Wel et al. 2014) and contain et al. 2013) at z = 2 and its extrapolation to higher red- only low levels of dust and cold gas (bottom row of Fig. 2). shifts. We note, however, that the exact functional form of The specific SFR for both star forming and quiescent theSHMRrelationdifferssomewhatamongstudies(Moster galaxies in the MassiveFIRE sample is shown in Figure 3. et al. 2013; Behroozi et al. 2013; Garrison-Kimmel et al. WemeasurethespecificSFRin5kpcradiitoroughlymimic 2014). aperture based flux measurements (Whitaker et al. 2011; Fig. 2 shows four typical example galaxies from Mas- Schreiber et al. 2015) and to minimize potential contribu- siveFIRE. Each panel displays a composite image (in rest- tionsfromlowmasssatellitegalaxies.SpecificSFRschange frame U, V, and J broad-band filters) of the dust repro- typically by less than 0.1 dex if measured within a radius cessed star light. Following conventional practice (Wuyts of 0.1 R instead. Star forming galaxies at cosmic noon vir et al. 2007; Williams et al. 2009; Whitaker et al. 2011) we have high specific SFRs of the order of ∼ 1 Gyr−1, while classify galaxies as quiescent based on their U-V and V-J quiescent galaxies form stars at significantly lower specific broadbandcolors;specifically,ifU−V >1.2,V −J <1.4, rates.WenotethatinmostcasestheSFRsofgalaxiesclas- and U −V >0.88×(V −J)+0.59. Our sample contains 8 sified as quiescent remain low for extended periods of time quiescentand15starformingcentralgalaxies(galaxiesthat (>3×108 yr). The specific SFRs of MassiveFIRE galaxies dominate the central potential well of their host DM halo), are in good agreement with observations (Schreiber et al. and 4 quiescent and 8 star forming satellites (galaxies or- 2015; Brammer et al. 2011). bitingacentralgalaxy).Starforminggalaxieshaveyounger We fit the growth history of the (cold) baryonic mass stellarpopulationsthanquiescentgalaxies,resultinginbluer (M = M +M +M ) of each galaxy with a modi- bar HI H2 ∗ intrinsiccolors,althoughintersperseddustlanesextinctand fied exponential ∝ (1+z)βe−γz over an extended redshift reddenthelightalongparticularlinesofsight(seetoppan- rangestartingfromz=7downtoeitherthefinalsimulation els of Fig. 2). Fortunately, the color-color classification is snapshotortothelastsnapshotatwhichthegalaxyisstilla relativelyinsensitivetotheamountofdustreddening.Most central, whichever comes first. Gas and stars within 10% of star forming galaxies in our sample have a late type mor- the virial radius from the center of a galaxy are considered phology with either large stellar and gas disks (half mass part of that galaxy. Our results do not change qualitatively radii > 3 kpc) or irregular shapes, in agreement with ob- ifweusea50%largerorsmallerradiusinstead.Wesimilarly MNRAS000,1–5(2016) Massive, quiescent galaxies at cosmic noon 3 -8 10 -9 10 MassiveFIRE Star forming -1yr]10-10 Quiescent R [ Centrals F Satellites S s -11 High res. 10 Medium res. z=2 z=1.7 Schreiber et al. 2015 -12 10 z=2 z=1.7 10 11 10 10 M [M ] star sun Figure 2.Color-compositeimagesoffourcentralgalaxiesinthe Figure3.SpecificSFRwithinthecentral5kpcofMassiveFIRE MassiveFIRE sample as they would appear in rest-frame U, V, galaxies as function of stellar mass. SFRs are averaged over the and J bands ∼ 4 billion years after the Big Bang (z = 1.67). past100Myr.Starformingandquiescentgalaxiesareshownby Galaxiesareshownface-onwitheachimagespanning30kpcon blue and red symbols respectively (the classification is based on each side. (Top left) a star forming, disk galaxy, (top right) a rest-frameU,V,andJbroadbandfluxes).Linesdenotetheloca- starforming,irregulargalaxy,(bottomrow)twoexamplesofqui- tion of the star forming sequence inferred from rest-frame ultra- escent, early type galaxies. Star forming and quiescent galaxies violetandinfraredobservations(Schreiberetal.2015).The1−σ differintheircolors,morphologies,andlevelsofdustextinction. scatter of individual galaxies above and below the star forming sequenceatz∼2isabout0.3dex(Daddietal.2007;Whitaker etal.2012;Schreiberetal.2015)(shadedregion).MassiveFIRE fit the growth of the DM mass, M , contained within the galaxies classified as star forming have specific SFRs consistent DM withtheobservedstarformingsequence.Starformationinquies- virialradiusofthehalossurroundingthesegalaxies.InFig.4 centgalaxies,however,proceedsatmuchlowerratesthaninstar we plot dlnM /dt anddlnM /dt for bothstar forming bar DM forminggalaxiesofcomparablestellarmass. and quiescent galaxies in MassiveFIRE, linking the growth ofDMhalostothegrowthofgalaxiesresidingatthecenters of those halos. The figure demonstrates that most galaxies In agreement with this analysis, dlnM /dt ∼ 0.4 bar at cosmic noon grow on the same timescale as the DM ha- Gyr−1largelyseparatesstarformingfromquiescentgalaxies los they live in, complementing previous work that showed intheMassiveFIREsample,asshowninFig.4.Asgalaxies that baryonic masses and DM masses of halos assemble on andhalosgrowonverysimilartimescales(seeFig.4),wecan similar timescales (Faucher-Giguère et al. 2011). This is a re-interpret this result in terms of the specific growth rates non-trivial result as galaxies contain only a small fraction, of DM halos, i.e., dlnM /dt (cid:46) 0.4 Gyr−1 is a necessary DM less than a fifth, of the baryons in halos (Papastergis et al. condition for a halo to host a quiescent galaxy at its center 2012). atcosmicnoon.Weproposethatthemajorityofmoderately At cosmic noon, galaxies with declining SFRs are on massive, quiescent galaxies in the young Universe form via their way to becoming quiescent. Hence, we may introduce thismechanism,i.e.,theyresideinthesub-setofhalosthat an alternative definition of “quiescence” that is not based accretegasfromthecosmicwebatsuchlowratesthatthey on broad-band colors, but on the star formation history of cannotmaintainSFRscharacteristicoftypicalstarforming galaxies. In particular, by manipulating the standard equa- galaxies (Schreiber et al. 2015). As discussed more below, tions for one-zone galaxy models including inflow, outflow, numerous halos undergoing such“cosmological starvation” star formation, and gas build up (e.g., Lilly et al. 2013; (Feldmann&Mayer2015)shouldexistgiventhevariations Feldmann 2015) the condition of a declining SFR can be inthegravity-drivencollapsehistoriesofDMhalos(McBride shown to be equivalent to dlnM /dt < X ≡ [1−R+ et al. 2009). bar crit dt /dt]/[t +sSFR−1]. Here, t =(M +M )/SFR 35% of the central galaxies and 34% of all galaxies in dep dep dep HI H2 is the gas depletion time, R is the return fraction of gas oursamplearequiescent.Thesenumberscomparefavorably fromevolvedstellarpopulations,andsSFR(t)=SFR/M = with observations of quiescent fractions of 25−50% over ∗ AsSFR (M (t),t)isthespecificSFR.sSFR isthespe- a broad stellar mass range (Tomczak et al. 2014; Muzzin MS ∗ MS cific SFR of galaxies of the same mass on the star forming et al. 2013); see Fig. 5. We note that the cumulative frac- sequenceandAcanbederivedfromthecriticalitycondition tion of quiescent galaxies in MassiveFIRE is lower at larger dSFR/dt=0.Uponinsertingvaluesappropriateforgalaxies stellar masses, while the observed fraction remains remains intheM ∼1010−1011M range,wefindthatsuchgalax- relatively flat. This difference could point towards missing ∗ (cid:12) iesshouldbereducingtheirstarformationactivity,andthus physics in our simulations, such as black hole feedback, or becoming quiescent, when dlnM /dt(cid:46)0.25−0.4 Gyr−1. it could be an artifact related to the low number of galax- bar MNRAS000,1–5(2016) 4 R. Feldmann et al. 1.5 log M [M ] 10 star sun Star forming 10 10.5 10.8 11 11.1 11.2 ] Quiescent 1 −1 1.25 Tomczak et al. 2014 z~2 yr Centrals 0.9 Muzzin et al. 2013 G [ 1 Satellites 0.8 MassiveFIRE (hydro) simulations dt N-body simulation + analytic model / )vir0.75 log10 Mstar 0.7 R 10 0.1 10.5 M) 0.6 (< 0.5 11 (>nt0.5 MH+H+starI20.25 on growth 11.5 fquiesce00..34 Xcrit x1.1 dln 0 ger bary 0.2 xx01..90 on 0.1 −0.25 str stronger DM growth 0 −0.25 0 0.25 0.5 0.75 1 1.25 1.5 12 12.5 13 13.5 14 dlnMDM(<Rvir)/dt [ Gyr−1] log10Mvir[Msun] Figure 4. Comparison between the growth rate of baryonic Figure5.Fractionofquiescent,centralgalaxiesresidinginhalos masses (stars, HI, and H2) of galaxies and the DM masses of aboveagivenmass.ThefractionspredictedbyMassiveFIREare their parent halos. Red circles and blue squares show quiescent shownbyfilledcircles.Errorbarsindicate1−σ standarddevia- andstarforminggalaxiesinMassiveFIRE,respectively.Theclas- tionsbasedonabinomialdistributionwiththesamesamplesize sification is based on rest-frame U-V and V-J colors (Whitaker and quiescent fraction as in MassiveFIRE. For the largest mass etal.2011)appropriateforz∼2.Filledandemptysymbolsde- binweassumea40%quiescentfractiontocomputetheerrorbar. note galaxies thatarecentralsorsatellitesbythefinalsnapshot Oursimulationspredictthataboutathirdofmassivegalaxiesat ofthesimulation(z=1.7−2).Symbolsizesreflectstellarmasses. cosmic noon are quiescent. Solid lines showthe fraction of halos Forcentralgalaxies,growthratesandcolorsarecomputedatthe withspecificgrowthratesbelowthecriticalvaluerequiredforqui- finalsnapshotofeachsimulation.Forgalaxiesthatbecomesatel- escentgalaxies(fromFig.4),0.9−1.1×Xcrit∼0.25−0.4Gyr−1, lites by z ∼ 2, we compute growth rates and colors in the last basedontheMillenniumN-bodysimulation(Springeletal.2005; snapshotbeforetheyentertheirhosthalo.Thesolidlinemarksa McBride et al. 2009). These theoretical estimates agree reason- 1:1 relationship and is not a fit. Galaxies residing at the centers ablywellwiththeobservedquiescentfractionderivedfromstel- of fast growing halos (dlnMDM/dt(cid:38)0.4 Gyr−1) are essentially larmassfunctionsofquiescentandstarforminggalaxiesoverthe alwaysstronglystarforming.Incontrast,slowlygrowing(oreven z=1.5−2.5range(dashed(Tomczaketal.2014)anddot-dashed shrinking)halostypicallyharborquiescentgalaxies. (Muzzinetal.2013)linesandshadedregions). Nelsonetal.2013).Bylimitingthesupplyofgastogalaxies ies (two) in our highest halo mass bin. Hence, whether cos- and halos, cosmological starvation makes it much easier for mological starvation is an effective quenching mechanisms additionalprocesses,e.g.,feedbackfromblackholes,tofully for galaxies residing in the most massive halos at z = 2 (M >1013 M ,n<10−5 Mpc−3)remainstobestudied counteracthotgascoolingandtoheatorejectanyremaining halo (cid:12) coolgas.Cosmologicalstarvationthusenablestheformation in future work. ofquiescentgalaxieswithredbroad-bandcolorsandreduced Fig. 5 also plots the fraction of halos with low specific SFRs at cosmic noon. However, it may be a necessary but growth rates based on a large-volume cosmological N-body not a sufficient condition for completely shutting-down star simulation (Springel et al. 2005; McBride et al. 2009). The formation in such galaxies. fractionofhaloswithdlnM /dt<0.9−1.1X matches DM crit The different accretion histories of quiescent and star fairly well the observed quiescent fraction. The former de- forminggalaxiesinhalosofthesamemasshaveanumberof clines slightly towards the largest stellar masses, indicat- observational consequences. First, as quiescent galaxies are ing that additional physics besides cosmological starvation assembled earlier, they will be surrounded by more evolved islikelyinvolvedinshuttingdownstarformationinthemost massive galaxies (M > 1011 M ). Fig. 4 shows that there satellite populations. In particular, orbital decay and tidal ∗ (cid:12) stripping(Zentneretal.2005)shouldreducethenumberof is overlap between star forming and quiescent galaxies at satellites of a given stellar mass, and the longer exposure intermediatespecificgrowthrates.Wefindthatthefraction (Feldmann et al. 2011) to the hot atmospheres of massive of slowly accreting halos still matches the observed fraction galaxiesmayexplaintheincreasedfractionofsatelliteswith ofquiescentgalaxiesevenifasizablefractionofsuchhalos, low SFRs (Weinmann et al. 2006). Second, we expect that e.g., a third, host star forming galaxies. thedominanthalosofover-denseenvironmentsshouldhave large accretion rates and, thus, should host vigorously star 3 CONCLUSIONS forminggalaxies.Incontrast,quiescentgalaxiesatthosered- The star formation activity of massive galaxies in a young shiftsshouldpreferentiallyresideinaverageorbelowaverage Universeisultimatelyfueledbytheaccretionofintergalactic environments.Thisideaiscorroboratedbyourfindingthat gas (Kereˇs et al. 2005; Dekel et al. 2009; Davé et al. 2010; 50% (25%) of the quiescent central galaxies vs 13% (73%) MNRAS000,1–5(2016) Massive, quiescent galaxies at cosmic noon 5 of the star forming central galaxies in our sample reside in FeldmannR.,CarolloC.M.,MayerL.,2011,Astrophys. J.,736, the lower (upper) quartile of the local environment density. 88 Third, the clustering of halos of a given mass depends on Garrison-Kimmel S., Boylan-Kolchin M., Bullock J. S., Lee K., their formation time, the so-called assembly bias (Wechsler 2014,Mon. Not. R. Astron. Soc.,438,2578 Hearin A. P., Watson D. F., 2013, Mon. Not. R. Astron. Soc., et al. 2006). As massive DM halos are less clustered if they 435,1313 collapsed earlier, we predict that massive, quiescent galax- HopkinsP.F.,2015,Mon. Not. R. Astron. Soc.,450,53 ies at cosmic noon have a lower clustering amplitude than HopkinsP.F.,QuataertE.,MurrayN.,2012,Mon. Not. R. As- starforminggalaxiesresidingwithinhalosofthesamemass. tron. Soc.,421,3522 Finally, we speculate that age-matching (Hearin & Watson Hopkins P. F., Kereˇs D., Onõrbe J., Faucher-Giguère C.-A., 2013), an empirical correlation between galaxy colors and Quataert E., Murray N., Bullock J. S., 2014, Mon. Not. R. halo formation time, has its physical origin in cosmological Astron. Soc.,445,581 starvation. KereˇsD.,KatzN.,WeinbergD.H.,DavéR.,2005,Mon.Not.R. Astron. Soc.,363,2 ACKNOWLEDGEMENTS KormendyJ.,HoL.C.,2013,Ann. Rev.Astron.Astrophys.,51, 511 RF was supported in part by NASA through Hubble Fel- LeeB.,etal.,2013,Astrophys. J.,774,47 lowshipgrantHF2-51304.001-AawardedbytheSpaceTele- LillyS.J.,CarolloC.M.,PipinoA.,RenziniA.,PengY.,2013, scope Science Institute, which is operated by the Associ- Astrophys. J.,772,119 ation of Universities for Research in Astronomy, Inc., for Ma X., Hopkins P. F., Faucher-Giguere C.-A., Zolman N., NASA, under contract NAS 5-26555, in part by the Theo- Muratov A. L., Keres D., Quataert E., 2015, preprint, retical Astrophysics Center at UC Berkeley, and by NASA (arXiv:1504.02097) ATP grant 12-ATP-120183. Support for PFH was provided MartinD.C.,ChangD.,MatuszewskiM.,MorrisseyP.,Rahman S.,MooreA.,SteidelC.C.,2014,Astrophys. J.,786,106 by an Alfred P. Sloan Research Fellowship, NASA ATP McBrideJ.,FakhouriO.,MaC.-P.,2009,Mon. Not. R. Astron. Grant NNX14AH35G, and NSF Collaborative Research Soc.,398,1858 Grant#1411920andCAREERgrant#1455342.CAFGwas Moster B. P., Naab T., White S. D. M., 2013, Mon. Not. R. supported by NSF through grants AST-1412836 and AST- Astron. Soc.,428,3121 1517491, by NASA through grant NNX15AB22G, and by MuratovA.L.,KereˇsD.,Faucher-GiguèreC.-A.,HopkinsP.F., Northwestern University funds. DK was supported by NSF Quataert E., Murray N., 2015, Mon. Not. R. Astron. Soc., grantAST-1412153.EQwassupportedbyNASAATPgrant 454,2691 12-ATP-120183, a Simons Investigator award from the Si- MuzzinA.,etal.,2013,Astrophys. J.,777,18 mons Foundation, and the David and Lucile Packard Foun- NarayananD.,etal.,2015,Nature,525,496 dation.Simulationswererunwithresourcesprovidedbythe Nelson D., Vogelsberger M., Genel S., Sijacki D., Kereˇs D., Springel V., Hernquist L., 2013, Mon. Not. R. Astron. Soc., NASA High-End Computing (HEC) Program through the 429,3353 NASA Advanced Supercomputing (NAS) Division at Ames Papastergis E., Cattaneo A., Huang S., Giovanelli R., Haynes Research Center, proposal SMD-14-5492. Additional com- M.P.,2012,Astrophys. J.,759,138 puting support was provided by HEC allocation SMD-14- ReddickR.M.,WechslerR.H.,TinkerJ.L.,BehrooziP.S.,2013, 5189 and NSF XSEDE allocation TG-AST120025. Astrophys. J.,771,30 Rodriguez-Puebla A., Primack J. R., Behroozi P., Faber S. M., REFERENCES 2015,preprint, (arXiv:1508.04842) Behroozi P. S., Wechsler R. H., Conroy C., 2013, Astrophys. J., Sańchez Almeida J., Elmegreen B. G., Munõz-Tun˜ón C., 770,57 ElmegreenD.M.,2014,Astron. Astrophys. Rev.,22,71 BrammerG.B.,etal.,2011,Astrophys. J.,739,24 ScannapiecoC.,etal.,2012,Mon.Not.R.Astron.Soc.,423,1726 Brisbin D., Ferkinhoff C., Nikola T., Parshley S., Stacey G. J., SchreiberC.,etal.,2015,Astron. Astrophys.,575,A74 Spoon H., Hailey-Dunsheath S., Verma A., 2015, Astrophys. SijackiD.,SpringelV.,2006,Mon.Not.R.Astron.Soc.,366,397 J.,799,13 SpringelV.,etal.,2005,Nature,435,629 Cattaneo A., Dekel A., Devriendt J., Guiderdoni B., Blaizot J., Teyssier R., Moore B., Martizzi D., Dubois Y., Mayer L., 2011, 2006,Mon. Not. R. Astron. Soc.,370,1651 Mon. Not. R. Astron. Soc.,414,195 CiottiL.,OstrikerJ.P.,1997,Astrophys. J. Lett.,487,L105 TomczakA.R.,etal.,2014,Astrophys. J.,783,85 Cresci G., Mannucci F., Maiolino R., Marconi A., Gnerucci A., Wechsler R. H., Zentner A. R., Bullock J. S., Kravtsov A. V., MagriniL.,2010,Nature,467,811 AllgoodB.,2006,Astrophys. J.,652,71 DaddiE.,etal.,2007,Astrophys. J.,670,156 WeinmannS.M.,vandenBoschF.C.,YangX.,MoH.J.,2006, Davé R., Finlator K., Oppenheimer B. D., Fardal M., Katz N., Mon. Not. R. Astron. Soc.,366,2 KereˇsD.,WeinbergD.H.,2010,Mon. Not. R. Astron. Soc., WhitakerK.E.,etal.,2011,Astrophys. J.,735,86 404,1355 WhitakerK.E.,vanDokkumP.G.,BrammerG.,FranxM.,2012, DekelA.,BirnboimY.,2006,Mon. Not. R. Astron. Soc.,368,2 Astrophys. J. Lett.,754,L29 DekelA.,etal.,2009,Nature,457,451 Williams R. J., Quadri R. F., Franx M., van Dokkum P., Labbé DiMatteoT.,SpringelV.,HernquistL.,2005,Nature,433,604 I.,2009,Astrophys. J.,691,1879 Faucher-Giguère C.-A., KereˇsD., MaC.-P., 2011, Mon. Not. R. WuytsS.,etal.,2007,Astrophys. J.,655,51 Astron. Soc.,417,2982 ZentnerA.R.,BerlindA.A.,BullockJ.S.,KravtsovA.V.,Wech- Faucher-GiguèreC.-A.,HopkinsP.F.,KereˇsD.,MuratovA.L., slerR.H.,2005,Astrophys. J.,624,505 Quataert E., Murray N., 2015, Mon. Not. R. Astron. Soc., vanderWelA.,etal.,2014,Astrophys. J.,788,28 449,987 FeldmannR.,2015,Mon. Not. R. Astron. Soc.,449,3274 ThispaperhasbeentypesetfromaTEX/LATEXfilepreparedby Feldmann R., Mayer L., 2015, Mon. Not. R. Astron. Soc., 446, theauthor. 1939 MNRAS000,1–5(2016)