On the Evolution of the Bi-Modal Distribution of Galaxies in the Plane of Specific Star Formation Rate versus Stellar Mass

0.43 MB·English

by  Bogdan C. Ciambur

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed23January2013 (MNLATEXstylefilev2.2)
On the Evolution of the Bi-Modal Distribution of Galaxies
in the Plane of Specific Star Formation Rate versus Stellar
Mass
3
1 B. C. Ciambur 1⋆, G. Kauffmann 1, S. Wuyts 2
0
2 1 Max Planck Institute for Astrophysics, Karl Schwarzschild Str. 1 85748 Garching, Germany
2 Max Planck Institute for Extraterrestrial Physics , Giessenbach Str. 1 85741 Garching, Germany
n
a
J
23January2013
1
2

ABSTRACT
]
A
G Wehavecomparedtheobserveddistributionofgalaxiesintheplaneofspecificstar
formation rate versus stellar mass with the predictions of the Garching semi-analytic
.
h modelatredshifts 0,1 and2.The goalis to test whether the implementationof radio
p mode AGN feedback, which is responsible for terminating the formation of stars in
- highmassgalaxies,providesanadequatematchto currenthigh-redshiftobservations.
o
Thefractionofquenchedgalaxiesasafunctionofstellarmassinthemodelsisingood
r
t agreementwithdataatz=0andz=1.Byz=2,therearetoofewquenchedgalaxieswith
s
a low stellar masses in the models. At z=2, the population of galaxies with no ongoing
[ starformationisclearlyseparatedfromthe‘mainsequence’ofstar-forminggalaxiesin

thedata.Thisisnotfoundinthemodels,becausez=2galaxieswithstellarmassesless
v1 than∼1011M⊙ arepredictedtohostblackholeswithrelativelylowmasses(lessthan
5 108M⊙). The current implementation of radio mode feedback from such black holes
2 reduces the cooling rates from the surrounding halo, but does not generate sufficient
0 energytostopstarformationentirely.Wesuggestthatthemodelsmaybebroughtinto
5 betteragreementwiththedataifblackholegrowthistriggeredbydiscinstabilitiesin
. additionto majormergers,andiffeedbackmechanismsassociatedwiththe formation
1
of galactic bulges act to quench star formation in galaxies.
0
3
Key words: galaxies: evolution – galaxies:active – galaxies: high-redshift.
1
:
v
i
X
1 INTRODUCTION cupy a different, but well-defined region of the SFR - mass
r
a plane. These objects can be understood from the point of
Deep galaxy surveys have revealed that in star-forming
viewoftheirhistories:theyarelikelytheremnantsofmajor
galaxies, starformation rateiscorrelated with stellar mass,
mergersbetweenstarforminggalaxies.Suchmergersdestroy
giving rise to a relation referred to as the ‘main sequence
disksand producebulges, and lead to starbursts if the pro-
of star-forming galaxies’ (Daddiet al. (2007); Noeske et al.
genitors are gas-rich. The merger may drive accretion onto
(2007)). This relation displays time evolution – galaxies of
a central supermassive black hole, thustriggering an active
fixed mass form stars at progressively lower rates at later
galactic nucleus (AGN). The effect of the consumption of
times. The scatter around the relation is quite small (∼
gas into stars during the merger and feedback from super-
0.3 dex) and exhibits little redshift dependence. The pic-
novaeandAGN‘quench’thesemassivesystems,shuttingoff
ture which emerges from this is that the star formation
theirstar formation and thusreddeningtheir colours.
histories of most star-forming galaxies have been relatively
quiescent.Longer periodsofcontinuousstarformation may In this paper, we compare the distribution of galax-
be interspersed with short periods of elevated star for- ies in the specific star formation rate - stellar mass plane
mation (starbursts) resulting from interactions or mergers from three different observational galaxy catalogues with
(Kauffmann et al. (2006)). The bursts are thought to con- predictions from the Guo et al. (2011) semi-analytic model
tributeto thescatter around themain sequence. of galaxy formation and evolution. We do this for three
Incontrast,massivegalaxiesthatare‘redanddead’oc- redshift intervals to analyse the evolution of this relation
and to understand the key processes which lead to it. Al-
thoughsatellite galaxies inclusters mayhavetheirstarfor-
⋆ Email:[email protected] mation inhibited due to the removal of their gas reservoir
(cid:13)c 0000RAS
2 B. C. Ciambur, G. Kauffmann, S. Wuyts
by ram-pressure stripping or tidal interactions with their
Table 1. Number of objects in each field. In the case of the
neighbours(calledsatellitequenching),thisprocesspertains GOODS-SandUDS,thefinalnumberofgalaxiesafterthestellar
to a relatively small fraction of the more massive galaxies. masscutisdisplayedalongwiththeinitial,totalnumber.
Inthesemi-analytic models,themaindriverofthequench-
ing of very massive galaxies is AGN ‘radio mode’ feedback
Redshift SDSS UDS GOODS-S
(Croton et al. (2006), hereafter C06). This mechanism in-
jects energy into the gas surrounding the galaxy, prevent- ∼0 10456 - -
ing it from cooling, condensing onto the galaxy and subse- ∼1 - 975/8033 748/5467
quentlyformingstars.Theefficiencyofthistypeoffeedback ∼2 - 1000/6649 554/3763
is dependent on the amount of gas in a hot phase within
the halo (this is higher for more massive haloes), as well
as on the mass of the accreting black hole, which in turn
depends on the host galaxy’s merger history. Because both
such as stellar masses and SFRs. We refer the reader to
haloes and black holes form hierarchically through merg-
theirpaperforfurtherdetails. W11constructedSFR-stel-
ers in the models, it is therefore important to check that
larmassplanesandstudiedthevariationofstructuralprop-
sufficient massive haloes and black holes have had time to
ertiesofgalaxiesacrosstheplane,findingforexampleaclear
growandproducetheobservedproportionofmassive,quies-
separation between the two galaxy populations in terms of
cent galaxies at high redshifts (e.g. McCarthy et al. (2004);
Sersic index.
Toft et al. (2009); Williams et al. (2010)). Such a compari-
In our work, we make use of the W11 data with a cut
soncanthereforebeausefultooltoprobethisaspectofthe in stellar mass at M⋆ =1010M⊙,abovewhich thetwo deep
model.
samplesarecomplete.Thenumberofgalaxiesineachofthe
redshift binsis shown in Table 1.
2 DATA
2.1 Observational Datasets
2.2 Semi-analytic model
The galaxy samples in this work were drawn from several
surveys and three redshift bins were considered. The shal- TheGarching semi-analytic model simulates galaxies start-
lowest sample, with galaxies in a redshift interval of 0.025 ing from themerger trees of dark matter haloes of the Mil-
< z < 0.05 and referred throughout as z ∼ 0 , was drawn lennium Run (Springel et al. (2005)). The latter is a cos-
from data as used in Wang et al. (2011) - a catalogue con- mological, N-body,dark matter only simulation with 21603
structed by matching galaxies with stellar masses greater particles in a box 500 Mpc/h in size. The cosmological pa-
than 1010M⊙ from the sixth data release (DR6) of the rameters used were based on a combined analysis of the
Sloan Digital Sky Survey (SDSS) (Yorket al. (2000)) with 2dFGRS (Colless et al. (2001)) and the first-year release
thefourthdatareleaseoftheGALEXsurvey(Martin et al. WMAP data (Spergel et al. (2003)) and are as follows: Ωm
(2005)). = 0.25, Ωb = 0.045, ΩΛ = 0.75, σ8 = 0.9, n = 1 and H0
The deeper samples were based on those used by =73kms−1Mpc−1.Thesemi-analyticmodelpopulatesthe
Wuytset al.(2011),hereafterreferredtoasW11.Theywere haloes with galaxies, which evolve according to the merger
drawnfromtheGOODS-SandUDScatalogues,asdescribed histories of their parent haloes as well as according to ana-
intheirpaper.Eachofthesetwocataloguescontainedgalax- lytic‘recipes’governingthebaryonicphysics.Theseinclude
ies separated in two bins of redshift, namely 0.5 < z < 1.5 gas cooling, star formation, supernova and AGN feedback,
and 1.5 < z < 2.5. They are referred throughout as z ∼ 1 chemical enrichment, and the effects of galaxy interactions
and z ∼ 2, respectively. The UDS catalogue was based on on star formation and bulge formation.
datafromseveraldifferentsurveys(CosmicAssemblyNear- The version of the model used in this analysis is that
infrared Deep Extragalactic Legacy Survey (CANDELS), of Guo et al. (2011). The galaxy parameters used in this
Grogin et al. (2011); Koekemoer et al. (2011); Spitzer Ex- work were retrieved from thepublicly available Millennium
tendedDeepSurvey(SEDS),PI:G.Fazio; UltraDeepSur- database (Lemson & Virgo Consortium (2006)). Two out-
vey(UDS),PI:O.Almaini;SpUDS-aSpitzerpubliclegacy puts of this model are available on the database: one as-
programontheUDS,PI:J.Dunlop)andground-baseddata, sociated with the Millennium simulation (described above)
on an area of ∼ 200 arcmin2, which constituted the UDS and a second with Millennium II (Boylan-Kolchin et al.
area covered by CANDELS (see Galametz et al. (in prep.) (2009)),whichisahigh-resolutionversionoftheformerwith
for details on the photometric catalogue and W11 for the the same cosmological parameters and particle numberbut
derivedstellar population properties). TheGOODS-Scata- withinasimulationbox100Mpc/hinsize.Hereweusedthe
loguewasconstructed inasimilar way,onanarea from the outputsresulting from theMillennium merger trees, dueto
GOODS-Southfield(GreatObservatoriesOriginsDeepSur- the fact that this simulation contains a larger number of
vey,a148arcmin2 areacentredontheChandraDeepField massivegalaxies.Inordertoassuregoodresolutiontowards
South) covered by CANDELS-Wideand CANDELS-Deep. thelower mass end,a cut was madeon thenumberof dark
Wherepossible,W11estimatedthestarformationrates matter particles per halo, i.e. Np ≥ 300. This did not af-
using combined ultraviolet and mid-infrared or far-infrared fectanyresultsforgalaxieswithstellarmassesgreaterthan
photometry. For objects lacking an infrared detection, they 1010M⊙.Cataloguesfromthreesimulationsnapshotsclosest
used stellar population synthesis model fits to the UV-to- inredshifttothethreeobservationaldatasetswereretrieved,
8 µm spectral energy distributions to estimate parameters corresponding to z=0, 0.99 and 2.07.
(cid:13)c 0000RAS,MNRAS000,000–000
The Galaxy Bi-Modality Evolution in the sSFR - M Plane 3
Figure 1. Conditional density distributions of specific star formation rate (sSFR) versus stellar mass, as calculated in equation 1.
Resultsareshownforredshift2(left),redshift1(middle)andredshift0(right).Forthetwohigherredshiftbinsthepanelsareordered,
fromtoptobottom:Guoetal.model,UDSdata,GOODS-Sdata.Forthez ∼0panelsontheright-handside,thetopplotisbasedon
themodelwhilethebottom onSDSSdata.
Figure 2.HistogramsofsSFRatfixedstellarmass.Forclarity,onlyfiveoutofthetenmassbinsareplottedineachpanel.AsinFig.
1, forthetwo higher redshiftbins thepanels areordered, fromtop tobottom: Guoet al.model,UDSdata, GOODS-Sdata. Forthe z
∼0panelsontheright-handside,thetopplotisbasedonthemodelwhilethebottom onSDSSdata.
(cid:13)c 0000RAS,MNRAS000,000–000
4 B. C. Ciambur, G. Kauffmann, S. Wuyts
3 ANALYSIS AND RESULTS
3.1 sSFR - Mass planes
We first converted the log(SFR) - log(M⋆) density plots in
W11intolog(sSFR)-log(M⋆)densityplots,i.e.starforma-
tionrateperunitmassorspecificstarformationrate(sSFR)
asafunctionofstellarmass.Thespecificstarformationrate
is a measure of the time taken by a galaxy to form its stel-
lar mass at its current star formation rate. This quantity
allowsforaclearerseparationbetweenthestar-formingand
quiescent populations, as will be seen later.
Next,all massbinswere normalised tounity,usingthe
equation
N(x,y)
N′(x,y)= (1)
N(x,y′)dy′
R
whereN(x,y)representsthecountatposition(x,y),withx
≡M⋆ andy ≡sSFR.N′ isthenormalisedcount,thequan-
tity plotted in Fig. 1.
Thisallowedforaneasiercomparisonbetweenthedataand
the model, removing the need to scale by the volume of
the survey, which, in some cases was unknown. It also al-
lowsustovisualizethebi-modalityofthepopulationonthe
densityplotsmoreeasily (seeKauffmann et al.(2003)).We
Figure 3. The fraction of quenched galaxies as a function of
note that model galaxies with sSFR=0 were randomly re-
stellar mass. The red line with no symbols, which corresponds
assigned avaluecorrespondingtothelocusofthequenched
tothemodel,includessSFR=0galaxiesfromthesimulationin
populationinthedata.Thiswasdoneinordertoaccountfor
thequenchedpopulation.Theerrorbarsonthedatapointswere
the fact that the observations always have a lower thresh-
computedusingbootstrapresampling.
old in sSFR, set by the range of models used in the SED
fitting.Theeffect ofthis was negligible at high redshift but
more important at z ∼ 0, where AGN feedback completely
halts star formation in a significant number of galaxies in tions,and whethermodeland observationsevolvethesame
thesimulations. way as a function of redshift.
Fig.1showsthatthesemi-analyticmodelisinqualita- AseparationbetweenthetwopeaksinsSFRwasmade
tive agreement with the data in that there is a clear main visuallyandaGaussianwasfittedtothestar-formingpeak;
sequence of star forming galaxies and another separated the Gaussian was subtracted from the total distribution
population of massive, red and dead galaxies. A more de- (whichwasbyconstructionnormalisedtounity),leavingthe
tailedcomparisonrevealsthatathighredshiftsthemainse- remaindertobeinterpretedasthequenchedfraction.Figure
quenceappearstolieatlowersSFRcomparedwiththedata. 3 shows the quenched fraction for both data and model at
This discrepancy is well-documented in the literature (e.g. various stellar masses and redshifts. The error bars on the
Damen et al. (2009)). Further, in the models the quenched data were computed by bootstrap-resampling the original
peakappearstobelocatedveryclosetothemainstarform- distribution(500bootstrap subsetswereused).Overall,the
ing sequence at early times and then travels down in time modelagreesqualitativelywiththeobservations.Atredshift
towards lower starformation rates. Intheobservations, the ∼0,themodelslightlyunderpredictsthequenchedfraction
quenched galaxies are well separated from the star-forming at high masses whereas at higher redshift, the model un-
ones even at z ∼ 2. derpredicts the fraction at low masses and overpredicts it
at high masses. This, taken in conjunction with the infor-
mation in Figure 2, leads to conclude that the bi-modality
persists in the data over a broad mass range, whereas the
3.2 Quenched fractions as functions of stellar modelonlytendstobebi-modaloveramassintervalwhich
mass narrows progressively with increasing redshift (at z ∼2 the
transition is quitesharp at M⋆ =1011−11.2M⊙). Outsideof
In order to quantify the evolution of the quenched popu-
thisinterval, a single mode is preferred.
lation with redshift, the sSFR-M⋆ plane was divided into
slices of constant stellar mass and histograms of the sSFR
distribution at fixed mass were constructed (Fig. 2). These
3.3 Peak separation and its evolution
histograms can easily bedecomposed intotwo components:
onepeakcorrespondingtothestar-forminggalaxies,andthe Althoughthequenchedfractions indicateaqualitativesim-
remainder corresponding to the quenched population. The ilarity between the model and the observations, they are
objective of this exercise is to compute the fraction of the notsensitivetotheexactshapesofthedistributions,orthe
population in the quenched state, per stellar mass bin, and separationbetweenthestar-formingmainsequenceandqui-
ascertain whether the model fraction matches the observa- escent peaks, which is visibly different between model and
(cid:13)c 0000RAS,MNRAS000,000–000
The Galaxy Bi-Modality Evolution in the sSFR - M Plane 5
thequenchedfraction increases significantly towards higher
masses).Incontrasttothis,inthemodelsthequenchedpop-
ulationbecomesprogressivelymoreseparatedfromthestar-
formingpopulationtowardslowerredshifts.Inaddition,the
quenchedpopulation ofhighredshift galaxiesin themodels
isconfinedtogalaxieswithhigherstellarmassesthanatlow
redshifts.
Let us now consider the implementation of the AGN
‘radio-mode’ feedback mechanism in the models. This type
of black hole activity results from accretion of hot gas from
a static halo around the host galaxy, described in detail by
C06:
M˙BH =κ108MMB⊙Hh−1(cid:18)200Vkvmirs−1(cid:19)3f0h.o1t (2)
Figure 4. The separation between the star-forming main se-
quenceandthequenchedpeakinunitsoflogsSFR,asafunction where MBH is the mass of the black hole, Vvir is the virial
of redshift.As inFig. 3, the error bars werecalculated by boot- velocityoftheparentdarkmatterhalo(orsubhalo)andfhot
strapre-sampling. isthefraction ofthetotalhalo/subhalomassintheformof
hot gas. Here κ is a free parameter which controls the effi-
ciencyofaccretion,anditistakentobe1.5×10−5M⊙yr−1.
data in Figure 1 at higher redshifts. The evolution of the The black hole luminosity is given by:
mean peak separation (averaged over all stellar masses) is
shown in Fig. 4. It is noteworthy that the general trend
in the data is opposite to that in the model. In the obser- LBH =ηM˙BHc2 (3)
vations, the star-forming main sequence and the quenched where η = 0.1 is the efficiency with which mass produces
populationarewellseparatedathighredshiftandgradually energyneartheeventhorizon.Theeffectofthisaccretionis
approacheachotherwithtime.Thisapproachismainlydue assumed to be injection of heat into the surrounding envi-
tothedecreaseinstarformation rateonthemain sequence ronment, preventing gas to cool, condense onto the galaxy
at lower redshifts. In the model, there is a slight increase and form stars. The effective cooling rate of gas is then af-
in the separation of the peaks with time. This is caused by fected in the following way:
the fact that residual star formation rates in the quenched
populationarelargerathighredshiftsthanatlowredshifts.
M˙cool,eff =M˙cool− 1LVBH2 (4)
2 vir
4 DISCUSSION AND CONCLUSIONS This feedback mechanism is assumed to be quiescent
and continual and is most effective at late times and for
We havecompared the distribution of galaxies in the plane high black hole masses. In equation 2, fhot has little effect
of specific star formation rate versus stellar mass with the ontheaccretion rateandisapproximatelyconstantforVvir
predictions of the Garching semi-analytic model, for three ≥150kms−1 atredshift0,asshowninC06.(Wenotethat
redshift bins. Ourmain findingsare listed below. because this threshold in Vvir is actually at a fixed value of
i.) There is qualitative agreement between models and Mvir, it evolves as (1+z)21) . This means that gas cooling
datainthatthereisawell-separatedmainsequenceofstar- suppression,and hencequenching,dependsmostly on MBH
forming galaxies at low stellar masses and a quenchedpop- andVvir,whichinturndependonthehistoryofthegalaxy’s
ulation at higher masses at all three redshifts in the semi- parent halo.
analytic models. The top three rows of Fig. 5 illustrate the average
ii.) The fraction of quenched galaxies as a function of Mbulge/M⋆ratio,MBH andVvirformodelgalaxiesasafunc-
stellarmassinthemodelsisingoodagreementwithdataat tion of position in the sSFR - M⋆ plane. The bottom row
z=0.Athigherredshifts,therearetoofewquenchedgalaxies representsthedistributionofMBHandVvir,colour-codedby
with low stellar masses in themodels. sSFR. At a given redshift, the quiescent galaxies are more
iii.) The evolution in the separation between the star- bulge-dominated,hostmoremassiveblackholes1 andreside
formingandquenchedpopulationsisnotreproducedbythe in haloes with larger virial velocities than the star-forming
models. In the data, the separation decreases towards low ones. The trends of the latter two parameters with redshift
redshift, whereas the the models it remains roughly con- goinoppositedirections:atafixedlocationinthesSFR-M⋆
stant. plane,blackholemassesgrowintime,butthevirialvelocity
Perusal of Figure 1 shows that in the observations, ofthehalodecreasesingeneral.Thebottomrowshowsthat
the quenched population gains more and more members
as time progresses, but it remains clearly separated from
the star forming population even at early times. Moreover, 1 Thereisanexceptiontothisifthegalaxyinquestionhasunder-
it is clear that in the observations, the quenched popu- gonearecentgas-richmerger.Inthiscasethereisalsoastarburst
lation extends across the entire stellar mass range from associated withthe mergerwhichboosts the starformationrate
1010M⊙ to ∼ 2 × 1011M⊙ at all redshifts (even though ofthegalaxy.
(cid:13)c 0000RAS,MNRAS000,000–000
6 B. C. Ciambur, G. Kauffmann, S. Wuyts
Figure5.AveragegalaxypropertiesacrossthesSFR-M⋆ planes,andthedependence ofquenchingonblackholemassandhalovirial
velocity.TheupperthreerowscorrespondtothesSFR-M⋆ planeswithconditionaldensityofgalaxiesfromtheobservational datasets
(asinFig.1)overplottedastheblackemptycontours.Thefilledcontoursmapthedistributionofaveragebulgemasstototalmassratio
(upper row), black hole mass (second row) and virial velocity (third row). The bottom row shows the distribution of MBH and Vvir,
colour-codedbysSFR.Everyrowhasfixedcontour levelstoillustratetheredshiftevolutionofthequantities inquestion.
at z=2, the dynamic range in specific star formation rate drivenbygalaxymergers,particularlymajormergers.There
across the Vvir versus MBH plane is much smaller than at are however other viable black hole feeding mechanisms,
z=0.Itcanbeconcludedthat,withoutadditionalquench- such as disc instabilities or deviations from axisymmetry
ing mechanisms, a stronger dependence of the feedback on (Bournaud et al. (2011)), that could play a role, especially
Vvir and/or a weaker dependenceon MBH (with the appro- at higher redshifts, where star-forming galaxies are more
priatere-adjustmentof κ)at z=2might improvethepeak gas-rich and so more likely to undergo such processes. In
separation discrepancy. Favouring Vvir over MBH in equa- theGuoetalmodels,lowmassbulgesareproducedbydisk
tion2wouldleadtostrongerfeedbackatearlytimes,where instabilities, but black hole growth does not occur during
halovirial velocities are high andblack holemasses arelow this process. The black hole mass vs. stellar mass relation
compared to theirpresent values. predicted by the model, for z ∼ 2, is plotted in Figure 6
Imposing an ad hoc redshift dependence on the equa- togetherwithobservationaldatatakenfromSimmons et al.
tionsthatgovernso-calledradio-AGNfeedbackiscertainly (2012).Atlowstellarmasses,blackholesareslightlyunder-
not ideal. Another possibility is to modify the black hole massive by about ∼ 0.5 dex on average compared to obser-
growth model. As can be seen from Figure 6, the z ∼ 2 vations,thoughtowardsthehighmassendtheagreementis
quenchedpopulationofgalaxieswithstellarmasseslessthan good.Inaddition,asseenfromthecountoursinFigure6,at
∼ 1011M⊙, should arise from feedback by black holes with allstellarmasseslessthan2×1011M⊙,themodelspredicta
massesof∼106−107M⊙.Blackholegrowthinthemodelis tailofblackholesthatscattertomassesmorethanafactor
(cid:13)c 0000RAS,MNRAS000,000–000
The Galaxy Bi-Modality Evolution in the sSFR - M Plane 7
of10belowthemean.Thisisnotseeninthedata.Asecular
black hole growth mechanism associated with disk instabil-
ities may help boost the average mass of this population,
resulting in more efficient quenching.
In addition, quenching mechanisms other than radio-
AGN feedback may be required. In the local Universe, the
frequencyoflarge-scale FRI-typeradiojetsingalaxieswith
low mass black holes is extremely low (Best et al. (2005)).
Almostallactivegalacticnucleiwithblackholesofthismass
are Seyfert galaxies or quasars, where much of the energy
liberated during black hole accretion emerges at ultraviolet
oropticalwavelengths.Itmaythusbenecessarytoconsider
so-called ‘quasarmode’feedback.Inthemodels,thequasar
accretion mode occurs during gas-rich mergers, when pro-
genitorblackholesform throughtherapidaccretion ofcold
gas.Thereisastarburstresultingfromthemerger,whichis
visible is thesecond panel of Figure 5 as aband of galaxies
withhighspecificstarformationratesandblackholemasses, Figure 6. The relation between galaxy central black hole mass
andthemergerremnantformsagalacticbulge.Themerger andstellarmassatredshift∼2.Thefilledcontoursrepresentthe
consumes most of the gas, and supernovae eject gas from model predictions while the green data points are observational
resultscorrespondingtogalaxiesinaredshiftintervalof1.5≤z≤
the galaxy and from the halo. In the current models, feed-
2.5.
back from this black hole accretion mode is not explicitly
modelled. If feedback from black hole formation is violent
enoughtoshutdown starformation in agalaxy entirely for
a few Gyr, this may result in more realistic SFRs in lower Damen M., F¨orster Schreiber N. M., Franx M., Labb´e I.,
masses galaxies at z ∼ 2. Indeed, examination of the top ToftS.,vanDokkumP.G.,WuytsS.,2009,ApJ,705,617
panel of Figure 6 shows that models and data move into Dubois Y., Devriendt J., Slyz A., Teyssier R., 2012, MN-
better agreement if galaxies with bulge mass-to-total mass RAS,420, 2662
ratios greater than ∼ 0.5 have been able to eject all their Grogin N.A., et al. 2011, ApJS,197, 35
gas. GuoQ.,WhiteS.,Boylan-KolchinM.,DeLuciaG.,Kauff-
Dubois et al. (2012), for example, have implemented a mann G., Lemson G., Li C., Springel V., Weinmann S.,
sub-grid model of black hole growth and feedback through 2011, MNRAS,413, 101
quasarandradiomodeinhydrodynamicalcosmologicalsim- KauffmannG.,HeckmanT.M.,DeLuciaG.,Brinchmann
ulations.Theyhaveshownthatquasarsareincreasinglyim- J., Charlot S., Tremonti C., White S. D. M., Brinkmann
portant at high redshift, where cold gas accretion triggers J., 2006, MNRAS,367, 1394
the quasar-mode of feedback. Gradually the cold gas reser- KauffmannG.,HeckmanT.M.,TremontiC.,Brinchmann
voir is depleted by star formation, shock-heating of mas- J.,CharlotS.,WhiteS.D.M.,RidgwayS.E.,Brinkmann
sivestructuresandfeedback.Thus,theradio-modebecomes J., Fukugita M., Hall P. B., Ivezi´c Zˇ., Richards G. T.,
dominantatlatertimes.Theseauthorsdidnot,however,in- SchneiderD. P., 2003, MNRAS,346, 1055
vestigatewhethertheirsimulations couldreproducetheob- Koekemoer A. M., et al. 2011, ApJS,197, 36
served evolution of the bi-modal galaxy population. This is LemsonG.,VirgoConsortiumt.,2006,ArXivAstrophysics
animportanttestofallgalaxymodellingeffortsthatinclude e-prints
feedback from supermassive black holes. Martin D. C., et al. 2005, ApJL, 619, L1
McCarthy P. J., Le Borgne D., Crampton D., Chen H.-
W., Abraham R. G., Glazebrook K., Savaglio S., Carl-
bergR.G.,MarzkeR.O.,RothK.,JørgensenI.,HookI.,
REFERENCES
Murowinski R., Juneau S., 2004, ApJL, 614, L9
Best P. N., Kauffmann G., Heckman T. M., Brinchmann Noeske K.G., et al. 2007, ApJL, 660, L43
J., Charlot S., Ivezi´c Zˇ., White S. D. M., 2005, MNRAS, Simmons B. D., Urry C. M., Schawinski K., Cardamone
362, 25 C., Glikman E., 2012, ApJ, 761, 75
Bournaud F., Dekel A., Teyssier R., Cacciato M., Daddi Spergel D. N., Verde L., Peiris H. V., Komatsu E., Nolta
E., Juneau S., ShankarF., 2011, ApJL, 741, L33 M.R.,BennettC.L.,HalpernM.,HinshawG.,JarosikN.,
Boylan-Kolchin M., Springel V., White S. D. M., Jenkins Kogut A.,Limon M., Meyer S.S.,Page L., TuckerG. S.,
A., Lemson G., 2009, MNRAS,398, 1150 WeilandJ.L.,WollackE.,WrightE.L.,2003,ApJS,148,
Colless M., et al. 2001, MNRAS,328, 1039 175
Croton D. J., Springel V., White S. D. M., De Lucia G., Springel V., White S. D. M., Jenkins A., Frenk C. S.,
Frenk C. S., Gao L., Jenkins A., Kauffmann G., Navarro Yoshida N., Gao L., Navarro J., Thacker R., Croton D.,
J. F., YoshidaN., 2006, MNRAS,365, 11 Helly J., Peacock J. A., Cole S., Thomas P., Couchman
Daddi E., Dickinson M., Morrison G., Chary R., Cimatti H., Evrard A., Colberg J., Pearce F., 2005, Nature, 435,
A., Elbaz D., Frayer D., Renzini A., Pope A., Alexander 629
D. M., Bauer F. E., Giavalisco M., Huynh M., Kurk J., ToftS.,FranxM.,vanDokkumP.,F¨orsterSchreiberN.M.,
Mignoli M., 2007, ApJ, 670, 156 LabbeI., WuytsS.,Marchesini D., 2009, ApJ,705, 255
(cid:13)c 0000RAS,MNRAS000,000–000
8 B. C. Ciambur, G. Kauffmann, S. Wuyts
WangJ.,KauffmannG.,OverzierR.,CatinellaB.,Schimi-
novich D., Heckman T. M., Moran S. M., Haynes M. P.,
Giovanelli R.,Kong X., 2011, MNRAS,412, 1081
Williams R. J., Quadri R. F., Franx M., van Dokkum P.,
Toft S., Kriek M., Labb´e I., 2010, ApJ, 713, 738
WuytsS., et al. 2011, ApJ, 742, 96
York D. G., et al. 2000, AJ, 120, 1579
(cid:13)c 0000RAS,MNRAS000,000–000