Table Of ContentAstronomy&Astrophysicsmanuscriptno.5942 (cid:13)c ESO20108
February5,2008
⋆
The VIMOS-VLT Deep Survey
∼
Color bimodality and the mix of galaxy populations up to z 2
7
0
0 P.Franzetti1,M.Scodeggio1,B.Garilli1,D.Vergani1,D.Maccagni1,L.Guzzo2,L.Tresse3,O.Ilbert4,F.
2
Lamareille5,T.Contini6,O.LeFe`vre3,G.Zamorani5,J.Brinchmann7,S.Charlot8,D.Bottini1,V.LeBrun3,J.P.
n Picat6,R.Scaramella9,10,G.Vettolani9,A.Zanichelli9,C.Adami3,S.Arnouts3,S.Bardelli5,M.Bolzonella5,A.
a
Cappi5,P.Ciliegi5,S.Foucaud12,I.Gavignaud13,A.Iovino2,H.J.McCracken8,14,B.Marano15,C.Marinoni16,A.
J
Mazure3,B.Meneux1,2,R.Merighi5,S.Paltani17,18,R.Pello`6,A.Pollo3,L.Pozzetti6,M.Radovich11,E.Zucca5,O.
9
1 Cucciati2,19,andC.J.Walcher3
2
1 INAF-IASFMilano,viaBassini15,I-20133,Milano,Italy
v
5 e-mail:[email protected]
7 2 INAF-OsservatorioAstronomicodiBrera,ViaBrera28,Milano,Italy
0 3 Laboratoired’AstrophysiquedeMarseille,UMR6110CNRS-Universite´deProvence,BP8,13376MarseilleCedex12,France
7 4 InstituteforAstronomy,2680WoodlawnDr.,UniversityofHawaii,Honolulu,Hawaii,96822
0 5 INAF-OsservatorioAstronomicodiBologna,ViaRanzani,1,I-40127,Bologna,Italy
6 6 Laboratoired’Astrophysiquedel’ObservatoireMidi-Pyre´ne´es(UMR5572),14,avenueE.Belin,F31400Toulouse,France
0 7 CentrodeAstrofisicadaUniversidadedoPorto,RuadasEstrelas,4150-762Porto,Portugal
h/ 8 Institutd’AstrophysiquedeParis,UMR7095,98bisBvdArago,75014Paris,France
p 9 INAF-IRA,ViaGobetti,101,I-40129,Bologna,Italy
- 10 INAF-OsservatorioAstronomicodiRoma,ViadiFrascati33,I-00040,MontePorzioCatone,Italy
o
11 INAF-OsservatorioAstronomicodiCapodimonte,ViaMoiariello16,I-80131,Napoli,Italy
r
t 12 SchoolofPhysics&Astronomy,UniversityofNottingham,UniversityPark,Nottingham,NG72RD,UK
s
13 AstrophysicalInstitutePotsdam,AnderSternwarte16,D-14482Potsdam,Germany
a
: 14 ObservatoiredeParis,LERMA,61Avenuedel’Observatoire,75014Paris,France
v 15 Universita`diBologna,DipartimentodiAstronomia,ViaRanzani,1,I-40127,Bologna,Italy
i
X 16 CentredePhysiqueThe´orique,UMR6207CNRS-Universite´deProvence,F-13288MarseilleFrance
17 IntegralScienceDataCentre,ch.d’E´cogia16,CH-1290Versoix
r
a 18 GenevaObservatory,ch.desMaillettes51,CH-1290Sauverny
19 Universita`diMilano-Bicocca,DipartimentodiFisica-PiazzadelleScienze,3,I-20126Milano,Italy
Received30June2006/Accepted11January2007
ABSTRACT
Aims. Inthispaperwediscussthemixofstar-formingandpassivegalaxiesuptoz∼2,basedonthefirstepochVIMOS-VLTDeepSurvey
(VVDS)data.
Methods. Wecomputerest-framemagnitudesandcolorsandanalysethecolor-magnituderelationandthecolor distributions.Wealsouse
the multi-band VVDS photometric data and spectral templates fitting to derive multi-color galaxy types. Using our spectroscopic dataset
weseparategalaxiesbasedon astar-formationactivityindicator derivedcombining theequivalent widthof the[OII]emission lineand the
strengthoftheD (4000)continuumbreak.
n
Results. Inagreementwithpreviousworkswefindthattheglobalgalaxyrest-framecolordistributionfollowsabimodaldistributionatz≤1,
andweestablishthat thisbimodalityholdsuptoat leastz=1.5.Thedetailsoftherest-framecolor distributiondepend however onredshift
andongalaxyluminosity,withfaintgalaxiesbeingbluerthantheluminousonesoverthewholeredshiftrangecoveredbyourdata,andwith
galaxiesbecomingbluerasredshiftincreases.Thislatterblueingtrenddoesnotdepend, toafirstapproximation, ongalaxyluminosity.The
comparisonbetweenthespectralclassificationandtherest-framecolorsshowsthatabout35-40%oftheredobjectsareinfactstarforming
galaxies.Henceweconcludethattheredsequencecannotbeusedtoeffectivelyisolateasampleofpurelypassivelyevolvingobjectswithin
acosmologicalsurvey.Weshowhowmulti-colorgalaxytypeshaveaslightlyhigherefficiencythanrest-framecolorinisolatingthepassive,
nonstar-forminggalaxieswithintheVVDSsample.Connectedtotheseresultsisalsothefindingthatthecolor-magnituderelationsderived
forthecolorandforthespectroscopicallyselectedearly-typegalaxieshaveremarkablysimilarproperties,withthecontaminatingstar-forming
galaxieswithintheredsequenceobjectsintroducingnosignificantoffsetintherestframecolors.Thereforetheaveragecoloroftheredobjects
doesnotappeartobeaverysensitiveindicatorformeasuringtheevolutionoftheearly-typegalaxypopulation.
1. Introduction ity has been foundto be a characteristic in the distribution of
manyotherobservableslikeH emission(Baloghetal.2004),
α
Early-typegalaxiesarethepreferredtargetforstudiesonhow
4000Åbreak(Kauffmannetal.2003b),starformationhistory
and when galaxies were formed, because of the simpler task
(Brinchmannetal. 2004), or clustering (Budava´rietal. 2003;
ofmodelinganoldstellarpopulationundergoingpassiveevo-
Meneuxetal.2006).
lutioncomparedtomodelingayoungpopulationcontinuously
Because of the age of their stellar population, early-type
modified by an irregular star formation history. Over the last
galaxies are expected to have very red optical colors over a
few years, mostly in response to the accumulating evidence
ratherlargeredshiftrange.Thisexpectationhasbeenconfirmed
infavorofapredominantlyoldstellarpopulationwithinearly
in clusters of galaxies, where the red-sequence for morpho-
type galaxies (see for a complete review Renzini 2006), the
logicallyselectedearly-typegalaxieshasbeenobservedupto
discussionhasfocusedontotwomainareas:thehistoryofstel-
z∼1.2 (Kodama&Arimoto 1997; Stanfordetal. 1998). As a
larformationandhowandwhenstarsassembledintogalaxies.
result the red color has often been used as a tracer to isolate
Still, a fundamentalpre-requisiteforsuch studiesis to isolate
samples of early-type galaxies. Most recently the widespread
fromcosmologicalsurveysasampleofearly-typegalaxiesrep-
acceptanceof color bimodalityhas resulted in the use of red-
resentativeofthetruegalaxypopulationatallepochsprobed.
sequencegalaxies(thosethatpopulatetheredpeakofthecolor
Commonly used galaxy classification schemes are based
bimodaldistribution)tostudytheevolutionofearly-typegalax-
purely on the morphological appearance of galaxies. It is
ies,implicitlyassumingthisredpopulationtobeentirelycom-
well known however that galaxies follow a number of scal-
posedofold,passivelyevolvingobjects.
ingrelationsinvolvingtheirstellar populationsandtheirmor-
However,whilemostoftheearly-typegalaxiesareindeed
phological, structural and photometric parameters. Compared
red,theyarepossiblynottheonlyredobjectsincludedinasur-
to their late-type counterparts, early-type galaxies have
veysample.Attemptsatquantifyingthecontaminationofnon
been found to be redder (color-morphological type relation,
early-type, passive objects within samples of red galaxies in-
Roberts&Haynes1994),moreluminous(color-magnitudere-
cludebothstudiesfocusedonExtremelyRedObjects(EROs),
lation, Visvanathan&Sandage 1977; Tullyetal. 1982), to be
and more general surveys which target the whole bulk of the
located in denser environments(morphology-densityrelation,
galaxy population. Among EROs, Cimattietal. (2002), using
Dressler1980),andtohaveastarformationhistorythattakes
spectroscopic data for 45 objects, find a roughly equal pro-
place over shorter time-scales (Sandage 1986; Gavazzietal.
portion of early-type, passive galaxies and of dusty starburst
2002).
objects. Among less extreme objects, in the local Universe
Cosmological survey studies, faced with the difficulty of
Stratevaetal. (2001) reported a significant fraction (20%) of
obtaining a morphological classification for all the galaxies
galaxiesmorphologicallylater than Sa in the red galaxy pop-
in their sample, have often tried to take advantage of those
ulation.Thisresultisconfirmedusingthedata forgalaxiesin
relations to define an alternate classification scheme. Galaxy
the Virgo Cluster and in the Coma Supercluster provided by
color, which is by far the easiest parameter to measure for
the GOLDMine database (Gavazzietal. 2003). At intermedi-
a full survey sample, has been most commonly used as a
ate redshift(upto z ∼ 1) variousstudiesare confirmingthese
substitute for morphologicalinformation. Lately this practice
results(seeforexampleKodamaetal.1999;Belletal.2004a),
has become even more commonplace, since the galaxy rest-
whilethenatureofredgalaxiesathigherredshiftislessclear.
frame color distributions have been found to be clearly bi-
However,thereareindicationsthatthefractionofmorpholog-
modal. This bimodality was well known to exist within clus-
ically late-type galaxies with red colors could be even higher
ters of galaxies, where the early- and late-type galaxies fol-
atz & 1(Stratevaetal.2001),inagreementwiththefindings
lowtworatherdistinctcolor-magnituderelations,asdiscussed
basedonEROsstudies.
byVisvanathan&Sandage(1977)andBoweretal.(1992)for
InthisworkweuseVIMOS-VLTDeepSurveyphotomet-
early-type galaxies and Tullyetal. (1982) and Gavazzietal.
ricandspectroscopicdata(VVDS,seeLeFe`vreetal.2005)to
(1996)forlate-typeones.Thatthesamebimodalitywaspresent
compare the galaxy population that can be isolated by using
in a general field sample was first noticed in the local uni-
either a red color selection criterion or a spectro-photometric
verse by Stratevaetal. (2001) using the SDSS galaxies sam-
classification. Although neither of these selection criteria can
ple,andafterwardsdiscussedbymanyotherauthors(see,asan
be considered entirely equivalent to a morphologicalclassifi-
example, Baldryetal. 2004a,b). Belletal. (2004b) (hereafter
cationinisolatingasampleofearly-typegalaxies,ourcompar-
B04)andWeineretal.(2005)detectedtherest-framecolorbi-
isondoesprovideanestimatefortheuncertaintiesinvolvedin
modalityalsoathigherredshiftsuptoz∼1,respectivelyinthe
selectingthoseobjectswithinacosmologicalsurveysample.
COMBO-17andDEEP2data.
Thepaperisorganizedasfollows.Insection2wedescribe
Color bimodality is interpreted as just one specific signa-
the VVDS galaxy sample, the data we use in this work and
ture introduced by the general processes controlling galaxy
we discuss the effect that the VVDS selection function could
formationandevolution(Mencietal.2005;Dekel&Birnboim
haveonourwork.Insection3weanalysetherest-framecolor
2006). Other similar signatures are observed, since bimodal-
distributions and we demonstrate that the color bimodality is
presentuptoatleastz=1.5.Thenwestudythecolorbimodal-
*BasedondataobtainedwiththeEuropeanSouthernObservatory
Very Large Telescope, Paranal, Chile, program 070.A-9007(A), and ityasafunctionofredshiftandofgalaxyluminosity.Insection
ondataobtainedattheCanada-France-HawaiiTelescope,operatedby 4 we focusonthe red-sequencegalaxiesstudyingtheir color-
theCNRSofFrance,CNRCinCanadaandtheUniversityofHawaii. magnituderelation. Insection 5, we analyse in detailthe red-
P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2 3
sequenceobjectsshowinghowthispopulationiscontaminated qualityflags2,3,4,9inLeFe`vreetal.2005)intheF02spectro-
byanonnegligiblefractionofstarformingobjects.Weusethe scopicsample,withtherestrictionofhavingz≤2.0,inorderto
strongbimodalityobservedinthe EW([OII])-D (4000)distri- ensurethatthe observedopticalandnear-infraredmagnitudes
n
butiontoisolateasampleofearly-typegalaxieslesscontami- providea reasonablyclose bracketing for the rest-frame opti-
nated by star-forminggalaxiesthan a simple color-magnitude calmagnitudeswe use in ouranalysis. A totalof6291galax-
selection.Finallyinsection6wediscussthecolorevolutionof ies areincludedin thissample with a measuredB, V, RandI
thissample. magnitude;hereafterwewillrefertothissampleas“complete
AllmagnitudesaregivenintheJohnson-Kron-Cousinsys- sample”.
tem; the adopted cosmologyis the standard Ω = 0.3, Ω =
m Λ
0.7andH =70km s−1 Mpc−1.
0
2.1.2. Rest-frameabsolutemagnitudesandcolors
2. Sampledescription
Rest-frame colors for all galaxies in the “complete sample”
2.1.Thedata
are computed from the absolute magnitude estimates derived
2.1.1. Observations asdescribedinAppendixAofIlbertetal.(2005).Foragiven
rest-frame photometric band and a given galaxy redshift, the
OursampleisselectedfromthefirstepochVVDS-Deepspec-
absolute magnitude is obtained from the apparent magnitude
troscopic sample within the VVDS-0226-04field (hereafter
measurementwithinthe observedphotometricbandthatmost
F02) (see LeFe`vreetal. 2005). This is derivedfrom a purely
closelymatchestherest-frameone.Ak-correctionisthenap-
magnitude limited sample (hereafter “photometric sample”),
plied to take into accountthe residual difference between the
including all objects in the magnitude range 17.5 ≤ I ≤
AB two photometric bands. The amplitude of the k-correction is
24.0fromacomplete,deepphotometricsurvey(LeFe`vreetal.
derivedbyselectingthebest-fittinggalaxyspectratemplateto
2004).
the galaxy B, V, R, I magnitude measurements. In this work
Thewhole1.2deg2 field hasbeenimagedin B,V,Rand I
we focus on the rest-frame U-V color, mainly because it al-
withthewide-field12KmosaiccameraattheCanada-France-
lows us to sample the amplitude of the 4000 Å break in the
HawaiiTelescope(CFHT),reachingthelimitingmagnitudesof
SpectralEnergyDistribution,whichisagoodindicatorofthe
B ∼26.5,V ∼26.2,R ∼25.9,I ∼25.0.Dataarecom-
AB AB AB AB galaxystellar populationage(Bruzual1983;Kauffmannetal.
plete and free from surface brightness selection effects down
2003a). U-V is also a color often used by previous works on
to I ≤ 24.0(seefordetailsMcCrackenetal.2003).U-band
AB thepropertiesofearly-typegalaxies,bothlocally(forexample
data,takenwiththeWide-FieldImagerattheESO/MPE2.2m
Visvanathan&Sandage 1977; Boweretal. 1992) and at red-
telescope, are available for a largefraction of the field, to the
shifts up to 1 (see B04). Another advantage provided by this
limitingmagnitudeofU ∼25.4(Radovichetal.2004).Inall
AB choiceofrest-framephotometricbandsisthattheyarebrack-
bands, apparentmagnitudeshave been measured using Kron-
etedbytheobservedbandsformostoftheredshiftrangecov-
like elliptical apertures (the same in all bands), with a mini-
eredbyoursample.AsdiscussedinIlbertetal.(2005),typical
mumKronradiusof1.2arcsec,andcorrectedfortheGalactic
k-correctionuncertaintiesareoftheorderof0.04magforthe
extinctionusingtheSchlegeldustmaps(Schlegeletal.1998).
U-band, and of 0.11 mag for the V-band. When these uncer-
ThemedianextinctioncorrectionintheI bandis∼ 0.05mag-
taintiesareaddedtotheapparentmagnitudemeasurementone
nitudes.
(approximately0.1magforthefaintestobject),theycontribute
Spectroscopic observations for about 23% of the objects
toatotalcolormeasurementuncertaintyofupto0.2mag.
included in the photometric magnitude limited sample have
being carried out with the VIsible Multi Object Spectrograph
(VIMOS, see LeFevreetal. 2003), on the UT3 unit telescope
2.1.3. Spectralindexes
oftheESOVeryLargeTelescope.Thespectroscopicsampleis
closetobeaperfectlyrandomsubsetoftheparentphotometric
sample, with only a small bias against large angular-size ob- Forallobjectsinthe“completesample”wehaveobtainedmea-
jects which can be easily corrected for (see Ilbertetal. 2005; surementsoftheequivalentwidthandlinefluxforallemission
Bottinietal.2005). linesandabsorptionfeaturesdetectedintheirspectra.Allthese
Observations have been reduced with the VIMOS measurementshavebeencarriedoutusingthePLATEFITsoft-
InteractivePipeline and GraphicalInterface software package ware package (Lamareille et al. 2007, in preparation), which
(VIPGI, see Scodeggioetal. 2005; Zanichellietal. 2005). A implements the spectral features measurementtechniques de-
redshiftmeasurementhasbeenderivedfor9036galaxieswith scribed by Tremontietal. (2004). In this work we use ex-
a success rate which is mildly dependent on both the galaxy clusively measurements of the [OII]λ3727 doubletequivalent
apparentmagnitudeandits redshift(see thediscussiononthe width and of the 4000 Å break (D4000), using the narrow
SpectroscopicSamplingRateandFigures2and3inIlbertetal. breakdefinitionproposedbyBaloghetal. (1999). Because of
2005).Themedianredshiftforthewholespectroscopicsample theVVDSobservationalset-up,thesemeasurementsarepossi-
isz=0.76(seeFigure25inLeFe`vreetal.2005). ble only for objects in the redshift range 0.45 < z < 1.2, i.e.
The sample we use for this work is composed of all the 4433objects(seeLeFe`vreetal.2005):hereafterwewillrefer
galaxieswithsecureidentificationandredshiftmeasurement(z tothissampleas“spectroscopicsub-sample”.
4 P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2
2.2.TheVVDSselectionfunction The assumption of a uniform luminosity and redshift dis-
tribution for the objects within each bin is not completely re-
Todiscussinameaningfulwaygalaxycolors,theirevolution, alistic. However within the two high-redshift bins, where the
and their relation to other physical propertiesof galaxies, we biascanbenon-negligible,thelargerfractionoffainterobjects
must consider possible biases against the inclusion of galax- (moreheavilybiasedagainst,becausetheyarelesslikelytobe
ies of a given(extreme)colorin our “completesample”.Any observableabovethesampleapparentmagnitudelimit)iscom-
such bias could be introduced either by the definition of the pensatedquiteeffectivelybyalargerfractionofobjectsinthe
magnitude limited “photometric sample” or by some bias in lowerhalfoftheredshiftbin(lessbiasedagainst,becausetheir
the efficiencyofthe redshiftmeasurementsusedto defineour smallerdistancemakesthemmorelikelytobeobservable).
“completesample”. Insection3we applythisestimateofthecolorbiasintro-
ducedbytheVVDSselectionfunctiontoshowthatwithinthe
adopted redshift bin limits (0.2,0.6,0.8,1.2,2.0) and V-band
2.2.1. Limitingmagnitude
absolute magnitude bin limits (-22, -21, -20, -19) the bias is
mostly neglibgible, and does not significantly affects the ob-
Within the VVDS, as in all magnitude limited surveys, the
servedcolordistributions.
rangeofluminositiescoveredbythesamplechangeswithred-
shift, with a globaltrend towardshigherluminositieswith in-
creasingz.Thefixedmagnitudelimithoweverisalsointroduc- 2.2.2. Redshiftmeasurementefficiency
ingsome colorbiasatthe faintluminosityendof the sample.
As the apparentmagnitudefor a galaxy in a given photomet- Anotherpossiblesourceofbiasforourdatacouldbethesys-
ricbanddependsonitsluminosity,redshiftandspectralenergy tematic lossof early-typegalaxieswithinour“completesam-
distribution (and therefore on its color), it is a expected that, ple”. This loss could be due either to the arguablylower effi-
for a given redshift and luminosity, galaxies with rest-frame ciencyinmeasuringspectroscopicredshiftsforearly-typewith
colorabovea certainvalueshouldbesystematicallyexcluded respect to late-type galaxies or to a possible bias against ob-
from the sample, simply because their apparent I-band mag- servations of early-type objects in multi-object spectroscopic
nitude becomes too faint, while bluer galaxies with the same surveysliketheVVDS.Thisbiasoriginatesfromthestronger
luminosity are still brighter than the limiting magnitude be- clusteringoftheseobjectswithrespecttolate-typeones,which
causeoftheirflatterspectrum(seeIlbertetal.2004).Toquan- cannotbematchedbecauseofMOSmaskdesignlimitations.
tify how this selection effect might bias our results we have Toquantifythispossiblebiaswehaveusedthe“photomet-
divided our ”complete sample” in four redshift and absolute rictype”classificationschemeproposedinZuccaetal.(2006);
magnitudebins andderivedin each bin a rest-framecolorvs. theavailableopticalphotometrydataforthewholephotomet-
apparentI-bandmagnituderelationusingsyntheticspectralen- ricmagnitudelimitedsamplehavebeenfittedwithlocalspec-
ergydistributions(SED). traltemplatestakenfromColemanetal.(1980),supplemented
with two starburst templates, to derive a photometrictype for
The SEDs were obtained using publicly available stel-
each galaxyin our sample. Here we are consideringthe E/S0
lar population synthesis models from Bruzual & Charlot
andtheearlyspiralphotometrictypeobjectstogetherasa“pho-
(Bruzual&Charlot 2003), which provide the time evolution
tometricearly-typepopulation”,andthelate-typespiral,irreg-
of a galaxy stellar spectrum as a function of galaxy age, star
ularandstarburstphotometrictypeobjectstogetherasa“pho-
formation history and stellar initial mass function. For this
tometriclate-typepopulation”.
work we have always used a Salpeter initial mass function
We have measured the percentage of photometric early-
(Salpeter 1955). For the star formation history (SFH) we fol-
type objects in the whole “photometric sample” and in our
low Gavazzietal. (2002) in usinga slightlymorerealistic set
spectroscopic “complete sample”. These values are listed in
ofSFHswithrespecttothecommonlyusedexponentiallyde-
table 1 for the whole redshift range and for two subsamples
creasingone.Tocoveraswidearangeofgalaxypropertiesas
at lowand highredshift.Inthis case we consistentlyused for
possible, we generated a wide array of synthetic SEDs (ages
both samples photometric redshifts, determined as described
from0to15Gyrandstarformationtime-scalesfrom0.1to25
inIlbertetal.(2006).Bycomparingthefractionofearly-type
Gyr).
galaxiesinthetwosamplesweestimatethatthemissingearly-
Becauseofthenonnegligiblerangesofredshiftandlumi-
type objectsin our spectroscopic”complete sample” are very
nosityspannedbyeachofourredshiftandabsolutemagnitude few(∼4%ofthetotalnumberofobjects).
bins we simulate a uniformdistributionof galaxiesinside the
bin,assigntoeachofthemoneSEDandusethemtoderivethe
3. Rest-framecolorbimodality
jointdistributionofapparentI-bandmagnitudeandrest-frame
U-Vcolor. Color bimodality has been observed for galaxy samples from
By imposing to this distribution the same apparent mag- manydifferentsurveys,andthe VVDSisno exceptionin this
nitudelimit used to select the VVDS sample we can measure respect. The distribution of the rest-frame U-V color against
thelikelihoodforagalaxywithagivencolortobeincludedin theabsoluteVmagnitudeforour“completesample”isshown
our “complete sample” which we term “color completeness”. in figure 1, with the total sample divided into four different
Whenever this completenessis too low (say . 50%) we con- redshiftintervals.Thesmallinsetineachpanelshowsthecor-
siderthecorrespondingcolorstronglybiasedagainstinthebin. responding color distribution, i.e. the projection of the color-
P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2 5
Fig.1.Therest-frameU-VcoloragainsttheabsoluteVmagnitudeforourspectroscopic“completesample”.Thefourpanelsshowgalaxies
indifferentredshiftintervals,asindicatedinthepanelsthemselves.Thesolidlineineachpanelshowsthebestfittingcolor-magnituderelation
fortheredsequenceobjectswithinthegivenredshiftinterval(seesection4).Theinsetineachpanelshowsthecolordistributionforallthe
objectsinthatredshiftinterval,i.e.theprojectionofthecolor-magnituderelationonthecoloraxis;noticethatweplotherepercentagesand
notabsolutenumbersofobjects.
magnitude relation on the U-V axis. A bimodal color distri- COMBO-17(B04)findings.Moreoverthedepthofoursample
bution is observedover all four redshift intervals. We remark isallowingustoextendthedetectionofsuchabimodalityupto
thatwithinourdatathebimodalityisvisibleirrespectiveofthe atleastz = 1.5(themeanredshiftforthegalaxiesinthehigh-
detailed choiceof filters that one coulduse to define the rest- estredshiftbin).Thisisingoodagreementwiththegalaxyfor-
frame color: we observe a rather similar bimodal distribution mation model discussed recently by Mencietal. (2005), who
usinganycombinationoftheU,B,V,R,andIfilters.Theob- predict a color bimodality to be clearly present starting from
servedbimodalcolordistributionisinagreementwithprevious z≈ 1.5(seetheirfigure4),asaresultoftheinterplaybetween
SDSS (Stratevaetal. 2001), DEEP (Weineretal. 2005) and
6 P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2
Fig.2.TheU-Vrest-framecolordistributionsforourspectroscopic“completesample”dividedintofourVmagabsolutemagnitudeandfour
redshiftbins;luminosityincreasesfromthebottomtothetop,redshiftincreasesfromlefttoright.Thehistogramsscaleisshownontheleft
sideoftheplot.ThesolidlineplottedinsomepanelshighlightsthecolorregionswhicharebiasedagainstbytheVVDSselectioncriteria(see
paragraph2.2.1fordetails);thescalefortheselines,i.e.thefractionalbiasagainstagivencolorvalueinthesample,isshownontherightside
oftheplot.
themerginghistoriesofforminggalaxiesandthefeedback/star differentfrom each other,has certainly an impacton how the
formationprocess. red and blue components in the color distribution presented
above are populated. Equally important in this respect is the
The insightthatthe bimodalrest-framecolordistributions
roleoftheenvironment,drivenbythedensity-morphologyre-
provideintogalaxyformationandevolutionprocessesishow-
lation,andtheworkofBaldryetal.(2004b)ontheSDSSsam-
ever limited, mainlybecause this bimodalityis observedonly
pleillustratesthegradualchangeintheproportionofblueand
when objects of all luminosities are considered at once. The
redgalaxiesasabivariatefunctionofgalaxyluminosityandlo-
factthatbothearly-andlate-typegalaxiesfollowsomeformof
color-magnituderelation, although the two relationsare quite
P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2 7
Table 1. The percentage of photometric early-type objects in
thewhole“photometricsample”(“Photo”columns)andinour
“completesample”(“Spec”columns).Valuesarelistedforthe
wholeredshiftrangeandfortworedshiftbins.
Type All 0.2<z<0.8 0.8<z<2.0
Photo Spec Photo Spec Photo Spec
Early 29.1 26.6 30.3 28.5 29.0 24.5
Late 70.9 73.4 69.7 71.5 71.0 75.5
caldensity.Cucciatietal.(2006)haveobtainedsimilarresults
ontheVVDSdata.
Theinfluencethatluminosityhasontheproportionofred
andbluegalaxiesisparticularlyimportantinourstudy,because
of the large redshift interval covered by our sample, and the
variation in the range of luminosities that are spanned at dif-
ferentredshiftsbyamagnitude-limitedsampleliketheVVDS
one.Infigure2thereforewefurthersubdividedour“complete
sample”,bysplittingitwithineachredshiftbinshowedinfig-
ure1intofourV-bandabsolutemagnitudebins.Inagreement
Fig.3.Colorevolutionoftheredsequenceasmeasuredbythevalue
with the results of Baldryetal. (2004b), the global bimodal oftheinterceptoftheCMRatM =−20.Smallopencirclesandthe
V
colordistributionisreplacedmostlybyunimodaldistributions, dottedlinearefromB04;largefilledcirclesandthesolidlinerepresent
whosepropertiesdependbothonredshiftandongalaxylumi- ourVVDSdataandthebestlinearfittothem.Theopensquarepoint
nosity. It is evident that a global color-magnitude relation is at z = 0 is the U −V CMR zero point at z = 0 computed by B04
presentoverthewhole0 < z < 2redshiftinterval,withgalax- fromSDSSdata.Thethicksolidcurveshowsthecolor evolutionof
iesbecomingredderastheirluminosityincreases.Atthesame thetemplatewhich,inour grid,betterapproximates aSingleStellar
timewealsonoticehow,withinafixedluminositybin,galaxies Population(asingleburst0.1Gylongatz=5,followedbypurepassive
becomebluerwithincreasingredshift,andthiseffectispresent evolution), and it is not a fit to the observed evolution of the color
of the red sequence. Error bars on our points account only for the
atallluminositiescoveredbyoursample.
statisticaluncertaintiesintheCMRzero-pointdetermination
Asdiscussedinsection2.2.1,thistrendtowardsbluercol-
ors could be, in principle, partly due to a sample selection
Table2.Thered-sequenceCMRinterceptanddispersion
bias, introduced by the fixed magnitude limit used to define
the VVDS sample. To demonstrate that in practice this is not
thecase,withineachredshift-absolutemagnitudepanelwein- Redshift Ngalaxies (U−V)MV=−20 σ
dicate with a thick solid line our estimate of the color com-
0.2−0.4 82 1.31 0.31
pleteness.Wheneverthiscompletenessislow(.50%,seedis- 0.4−0.6 123 1.22 0.25
cussion in paragraph2.2.1) the correspondingcolor is clearly 0.6−0.8 259 1.20 0.29
biased againstin oursample,and thereforethe colordistribu- 0.8−1.0 229 1.18 0.21
tionplottedforthegivenredshiftandluminositybinmightnot 1.0−1.2 133 1.16 0.18
be representative of the whole galaxy population within the 1.2−1.4 44 1.05 0.25
bin boundaries. From the figure we can clearly see that the 1.4−2.0 17 1.00 0.08
color completeness is good, with the partial exception of the
1.2 < z < 2.0 and −21 < V < −20 bin, demonstrating
abs
how the visible trend towards bluer color at high redshift for
allluminositiesisnotartificiallycreatedbytheVVDSsample our attention on the galaxies in the read peak of the bimodal
definition. distribution,underthetemporaryassumptionthattheseredob-
These results are in agreementwith the findings of previ- jects can be identified with the early-type galaxy population
ous redshift surveyslike the CFRS (Lillyetal. 1995) and the withinoursample.Thepropertiesofthecolor-magnituderela-
Hawaii Deep Fields Survey (Cowieetal. 1996), but it is the tion(CMR)ofearly-typegalaxies(scatter,slope,andevolution
firsttimethatthesefindingsareextendedsignificantlyoverthe withredshift)haveoftenbeenusedtoconstrainformationand
z≈1limit. evolutionscenariosforthisgalaxypopulation(seeforexample
Boweretal. 1992; Kodama&Arimoto 1997; Bernardietal.
2003). These studies were mostly based on cluster samples
4. Thecolor-magnituderelationoftheredgalaxies
ofmorphologicallydefinedearly-typegalaxies,anditwasnot
Whileintheprevioussectionweanalyzedthecolordistribution entirely obvious how a similar kind of analysis could be per-
forthewholespectroscopic“completesample”,herewefocus formed on a sample of color-selected objects in a large high
8 P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2
Table3.TheD (4000)andrest-frameEW([OII])valuesforthe 5. Thenatureoftheredpopulation:separating
n
Virgotemplatesplottedinfigure4 star-formingfrompassivelyevolvinggalaxies
5.1.Spectralproperties
Label Type D (4000) EW([OII])(Å)
n
1 dE 1.42 >−0.5 Early-typegalaxiesaredominatedbyanoldstellarpopulation
2 E/S0 1.73 >−0.5 undergoing an almost purely passive evolution, and this last
3 Sa 1.52 >−5.0 propertyiswhatmakesthemanattractivetargetforgalaxyevo-
4 Sb 1.49 >−8.0 lutionstudies.Therefore,inselectingearly-typegalaxiesfrom
5 Sc 1.18 −12.7 a global survey sample we should consider how to efficiently
6 Sd 1.05 −30.3 select passively evolving objects, and not just red galaxies or
7 Irr 1.03 −15.7 galaxiesthataremorphologicallyclassifiedasellipticalofS0.
8 BCD 1.09 −24.8
As a first step towards this goal, we start by quantifying
how the contamination from late-type, star forming galaxies
isaffectingthepropertiesoftheredcolor-selectedpopulation,
andhowthiseffectischangingwithredshift.WeusetheVVDS
spectroscopicinformationto separate our“spectroscopicsub-
sample”intotwoclassesofold,mostlikelypassivelyevolving
redshiftsurvey lackingthe necessary high resolution imaging
andofyoung,star–formingobjects.Weremindthatthissample
to perform the appropriate morphological classification. B04 isrestrictedtotheredshiftrange0.45<z<1.2(seeparagraph
have demonstrated that with a large sample and good photo-
2.1);alltheanalysisdoneinthefollowingsectionsistherefore
metricredshiftestimateslikethoseprovidedbytheCOMBO- limited to this redshiftrange. A more detailed analysis of the
17 survey, such kind of analysis is indeed possible. Here we spectralpropertiesofthissamplewillbepresentedinVergani
followtheirproceduretostudytheredshiftevolutionofthered
etal.2007(inpreparation).
CMRintheVVDSsample.
In figure4 (top panel)the relationbetween the equivalent
Following B04, we have used our data to estimate the widthofthe[OII]λ3727line(EW[OII])andthe4000Åbreak
zero-point of the CMR, keeping its slope fixed to the value (D4000) is shown for the whole “spectroscopic sub-sample”.
whichhasbeendeterminedforlocalgalaxyclusters(-0.08,see BluesymbolsareobjectswithU−V <1.0,whileredsymbols
Boweretal.1992).Withineachredshiftbinwefirstcorrected areobjectswithU−V >1.0.Itisclearlyvisiblefromtheplot
theindividualcolormeasurementstoeliminatethefixedCMR howobjectsfollowabimodalL-shapeddistribution,withmost
slope, and then we used the bi-weight estimator (Beersetal. ofthestar-formingobjectsthathavealargeEW[OII]havinga
1990) to computethe meancolor for all red galaxies,defined negligibleD4000,andvice-versamostoftheobjectsthathave
(asinB04)asobjectswithrest-framecolorU-V>1.0.Thesolid astrongD4000showingnosignificant[OII]emissionintheir
lineplottedineachpaneloffigure1showsthecorresponding spectra. The bottom panels of figure 4, obtained dividing the
fitfortheredsequenceinthatredshiftrange. “spectroscopicsub-sample”infourredshitbins,showthatthis
ThechangeinthevalueofthefittedCMRinterceptatM = bimodalityisessentiallyindependentfromredshift.Thelarger
V
−20asafunctionofredshiftisshowninFigure3,andtheplot- dispersionintheD4000measurementsofstar-formingobjects
tedvaluesarealsosummarizedinTable2.Theevolutionofthe inthehigherredshiftbinismostlyduetothehigheruncertain-
red population is quite visible; from z ∼ 0.3 to z ∼ 1.7 red tiesinmeasuringthisfeaturefordistantandfaintobjects.
galaxiesbecomeblueronaverageby∼0.3mag.Thesolidline These distributions suggest a natural sub-division of the
isalinearfittotheevolution:U −V = 1.355−0.206×z.The galaxy population into two classes. We define spectral early-
resultsofB04ontheCOMBO-17dataareplottedforcompari- type objects (ET) galaxies in the vertical arm of the two pa-
son;alsotheU−VCMRzeropointatz=0determinedbyB04 rametersdistribution,i.e.thosethatdonotshowanydetectable
fromSDSSdataisshown.Ourresultsqualitativelyconfirmthe signofstarformationactivity,andwhichcanbeexpectedtoun-
evolutionclaimedbyB04andextendtheirmeasurementupto dergofurtherevolutiononlyviapassiveevolutionoftheirstel-
z∼2.Howeverwefindthattheamountofthisevolution,quan- larpopulation.Conversely,wedefinespectrallate-typeobjects
tifiedbytheslopeoftherelation,issomewhatmilderthanthat (LT)galaxiesinthehorizontalarm,stillundergoingavigorous
reported by B04, and this result is bringing the whole evolu- star formationactivity.Moreelaboratedspectralclassification
tionary trend in better agreementwith the z = 0 SDSS-based schemeshavebeenproposedinthepast,likethefour-foldsub-
point.Thethicklineplottedinfigure3shows,purelyasarefer- division discussed by Mignolietal. (2005), but we prefer to
ence,thecolorevolutionofthesyntheticSEDwithinthemod- use here a simpler two-fold subdivision, that mirrors the one
elsgriddiscussedinsection2.2.1whichbetterapproximatesa obtainedusingtherest-framecolors.
SingleStellarPopulation(asingleburst0.1Gylongatz=5,fol- To a first approximation the definition of the separat-
lowedbypurepassiveevolution).AsalreadynoticedbyB04, ing boundary between LT and ET galaxies in the D4000
there is a general good agreementbetween the observed evo- vs. EW[OII] parameters space could be a vertical line at
lutionoftheCMRandtheexpectationsfora purelypassively EW[OII]=10Å , the approximatedetection limit for the [OII]
evolvingpopulation.Ourresultsshow thatthisagreementex- line in our data. This definition, however, would lead us to
tendsatleastuptoz∼2. include many objects with very low D4000 and a real, albeit
P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2 9
Fig.4.TheD (4000)distributionasafunctionoftherestframeEW[OII]forthewhole“spectroscopicsub-sample”(toppanel)andforthe
n
samesamplesubdividedintothefourindicatedredshiftbins(bottompanels).BluesymbolsareobjectswithU−V <1.0,whileredsymbols
areobjectswithU−V >1.0.Only2σorhigherconfidenceD (4000)measurementsareplotted.ForEW[OII]measurementswithaconfidence
n
levellowerthan2σupperlimitsareplottedTheerrorbarsintheupperleftcorneroftheplotshowthetypicalmeasurementuncertaintiesfor
thehigherS/Nspectra(left)andthelowerS/Nspectra(right)sub-samples.Thesolidlinedividesthesampleinthepassivepopulation(ET)
andstar-formingpopulation(LT).ForcomparisonthelargenumberedcirclesaretheD (4000)/EW[OII]valuesfortheVirgoclustertemplates
n
fromGavazzietal.(2002)aslistedinTable3
undetected, [OII] emission in the ET sample, which is obvi- checkedvariousdefinitionsoftheET-LTseparatingboundary
ously contrary to the natural definition of early-type galaxies line that would keep constant the ratio of ET to LT galaxies.
asobjectswithanevolvedstellarpopulation(largeD4000;see We havebuilt a compositespectrumof the ET populationfor
Kauffmannetal.2003b)andnocurrentstarformationactivity each one of the definitions so that the higher signal-to-noise
(no [OII] emission line). To alleviate this problem, we have ratio thus achieved would allow a robust measurementof the
10 P.Franzettietal.:TheVVDS-Colorbimodalityandthemixofgalaxypopulationsuptoz∼2
Fig.5. a) U-V rest-frame color distribution; the shaded histogram is for the LT population, while the heavy line histogram is for the ET
population. Theinset shows thesamehistograms drawn only for thehigher S/N objects. b)U-V vsV color-magnitude relation; filleddots
are ET objects, tinydots areLT ones. Both plots include only the objects withredshift within the interval 0.6-0.8 from the “spectroscopic
sub-sample”
[OII] emission at fainter intensities. Then we have selected tant physical properties like the age of the stellar population
the subdivision that minimizes the equivalent width of the andtheamountofstarformationactivitytakingplaceineach
[OII] line in the composite spectrum of the resulting ET galaxy;unlikecolors,itisminimallyaffectedbytheunknown
population. This optimization procedure results in a slightly amountof reddeninginside each galaxy.One possible limita-
tiltedboundarylinedescribedbytherelation: tionaffectingourspectralclassificationisthefactthatanum-
berofobjectswithrelativelyyoungstellar population,aswit-
D (4000)+EW([OII])/15.0=0.7 nessed by the small value of their D4000, is included within
n
theETpopulation.Howeverweprefernottoexcludetheseob-
According to this definition, ET galaxies have jectsbysomemodificationoftheET-LTseparationboundary,
D (4000) + EW([OII])/15.0 above 0.7, while LT galax- because any such a modification would make our classifica-
n
ies are below that value. Purely for comparison, in figure 4 tion scheme much more vulnerable to progenitor bias effects
we have also plotted as large numbered circles the D (4000) (vanDokkum&Franx2001).Withthecurrentschemeassoon
n
and EW([OII]) values measured on Virgo cluster templates asagalaxyhascompletedthebulkofitsstarformationactivity
from Gavazzietal. (2002). Table 3 recaps the legend for the and is starting the purely passive evolution phase it becomes
circles. It should not be surprising that our ET-LT boundary anETobject,anditremainssuchatallsubsequenttimes,asits
is including early spirals within the ET population, and only stellarpopulationagesandtheD4000amplitudeinitsspectrum
Sc and later types within the LT population. This is a rather increases.Instead,anysignificantstarformationactivitywould
general property of classification schemes based on galaxy moveanobjecttowardstheleftinthefigure4diagram,effec-
color or spectral properties. For example, the earliest spiral tivelyremovingitfromtheET population.Thisminimization
galaxySEDpresentedbyColemanetal.(1980)isthattypical of progenitor bias effects on our classification scheme is the
of Sbc galaxies, and this same SED was used by Lillyetal. main reason we can use a redshift-independent classification
(1995)toseparateredandbluegalaxiesintheirCFRSgalaxy schemeinouranalysis.
sample. Similarly, the earliest spiral galaxy color-color track
usedbyAdelbergeretal.(2004)indefiningthecolorselection
5.2.Thecontaminationeffect
criteriaforisolatingstar-forminggalaxiesintheredshiftrange
1<z<3isthattypicalofanSbgalaxy.
Havingobtainedanearly-vs.late-typegalaxyseparationbased
We consider this spectral classification as a better sub- onspectralproperties,weanalysewhatisthecolordistribution
stitute for a true morphological classification with respect to of these two categories, and compare this spectral classifica-
color for a number of reasons: it is based on directly mea- tionwiththe“red-peak”colorone.Figure5ashowstheglobal
surable quantitiesand avoids the uncertaintiesinvolvedin es- U-V rest-frame color distribution for our “spectroscopic sub-
timating rest-frame colors; it is directly based on two impor- sample”intheredshiftintervalwherethepeakoftheN(z)dis-
