PDF (1.698 MB)

12 Pages·2008·1.7 MB·English

A&A491,681–692(2008) Astronomy
DOI:10.1051/0004-6361:200809845 &
(cid:2)c ESO2008 Astrophysics
Photometric redshifts as a tool for studying the Coma
(cid:2)
cluster galaxy populations
C.Adami1,O.Ilbert1,2,R.Pelló3,J.C.Cuillandre4,F.Durret5,A.Mazure1,J.P.Picat3,andM.P.Ulmer1,6
1 LAM,OAMP,UniversitéAix-Marseille&CNRS,Pôledel’Étoile,SitedeChâteauGombert,38rueFrédéricJoliot-Curie,
13388Marseille13Cedex,France
e-mail:[email protected]
2 InstituteforAstronomy,2680WoodlawnDr.,UniversityofHawaii,Honolulu,Hawaii96822,USA
3 Laboratoired’AstrophysiquedeToulouse-Tarbes,UniversitédeToulouse,CNRS,14Av.ÉdouardBelin,31400Toulouse,France
4 Canada-France-HawaiiTelescopeCorporation,Kamuela,HI96743,USA
5 Institutd’AstrophysiquedeParis,CNRS,UMR7095,UniversitéPierreetMarieCurie,98bisBdArago,75014Paris,France
6 DepartmentPhysics&Astronomy,NorthwesternUniversity,Evanston,IL60208-2900,USA
Received26March2008/Accepted13September2008
ABSTRACT
Aims.Weapplyphotometricredshifttechniques toaninvestigationoftheComaclustergalaxyluminosityfunction(GLF)atfaint
magnitudes,inparticularintheu∗bandwherebasicallynostudiesarepresentlyavailableatthesemagnitudes.
Methods.Clustermemberswereselectedbasedonprobabilitydistributionfunctionfromphotometricredshiftcalculationsappliedto
deepu∗,B,V,R,Iimagescoveringaregionofalmost1deg2(completenesslimitR∼24).Intheareacoveredonlybytheu∗image,
theGLFwasalsoderivedafterastatisticalbackgroundsubtraction.
Results.Global and local GLFs in the B, V, R, and I bands obtained with photometric redshift selection are consistent with our
previousresultsbasedonastatisticalbackgroundsubtraction.
TheGLFintheu∗bandshowsanincreaseinthefaintendslopetowardstheouterregionsofthecluster.
Theanalysisofthemulticolortypespatialdistributionrevealsthatlatetypegalaxiesaredistributedinclumpsintheclusteroutskirts,
whereX-raysubstructuresarealsodetectedandwheretheGLFintheu∗bandissteeper.
Conclusions.WecanreproducetheGLFscomputedwithclassicalstatisticalsubtractionmethodsbyapplyingaphotometricredshift
technique. Theu∗ GLFslopeissteeperintheclusteroutskirts,varyingfromα ∼ −1intheclustercentertoα ∼ −2inthecluster
periphery.Theconcentrationsoffaintlatetypegalaxiesintheclusteroutskirtscouldexplaintheseverysteepslopes,assumingashort
burstofstarformationinthesegalaxieswhenenteringthecluster.
Keywords.galaxies:clusters:individual:Coma–galaxies:luminosityfunction,massfunction
1. Introduction contiguous coverage of the spectral regions of interest, in par-
ticular around the strong spectral features, such as the 4000 Å
Ever-growing optical imaging surveys of galaxies for which
break or the Lyman α break. This technique has proven to be
completespectroscopicfollow-upisimpossibleduetotelescope
a veryvaluabletoolforseveralcosmologicalpurposes,suchas
limitations has triggered the development of photometric red-
derivingfieldgalaxycorrelationfunctionsorgalaxyluminosity
shift techniques (e.g. Bolzonella et al. 2000; or Ilbert et al.
functions(seee.g.Ilbertetal.2006b;Meneuxetal.2006).Ithas
2006a, and references therein). Based on the comparison of
alsobeenappliedinthestudyofdistantclustersofgalaxies(see
multi-band photometry with synthetic spectral templates, this
e.g.Whiteetal.2005,andreferencestherein).
technique can be viewed as very low-resolution spectroscopy.
Inthepresentpaper,weapplythephotometricredshifttech-
Thequalityofphotometricredshiftsdependsonthewavelength
nique to study the galaxy population of the rich Coma cluster.
range covered by the photometric survey. High-quality photo-
Due to its proximity (z ∼ 0.023), Coma covers a large extent
metric redshifts and related quantities require a complete and
overthesky(ontheorderof1deg2),requiringawidefieldcam-
(cid:3) BasedonobservationsobtainedwithMegaPrime/MegaCam,ajoint era;dataintheU bandareneededtodeterminewithasufficient
project of CFHT and CEA/DAPNIA, at the Canada-France-Hawaii accuracytheprobabilitythatagalaxybelongsornottotheclus-
Telescope(CFHT)whichisoperatedbytheNationalResearchCouncil ter,basedinparticularonthe4000Åbreak.
(NRC)ofCanada,theInstitutNationaldesSciencesdel’Universofthe As a continuation of our Coma photometric survey (e.g.
Centre National de la Recherche Scientifique (CNRS) of France, and Adamietal.2006a)whichalreadyincludesdeep(R∼24)wide
theUniversityofHawaii.Thisworkisalsopartlybasedondataprod- field (42 × 52 arcmin2) B, V, R and I images, we recently
uctsproducedatTERAPIXandtheCanadianAstronomyDataCentre acquired a Megacam u∗ band image of comparable depth and
aspartoftheCanada-France-Hawaii TelescopeLegacySurvey,acol-
field of view. In addition to our spectroscopic redshift catalog,
laborativeprojectofNRCandCNRS.AlsobasedondatafromW.M.
thisallowsustocomputephotometricredshiftsandProbability
KeckObservatorywhichisoperatedasascientificpartnershipbetween
theCaliforniaInstituteofTechnology,theUniversityofCalifornia,and Distribution Function (PDF hereafter) along the Coma clus-
NASA.Itwasmadepossiblebythegenerousfinancial support of the ter line of sight down to the dwarf galaxy regime. One of the
W.M.KeckFoundation. main goals of the paper is to qualify this photometric redshift
ArticlepublishedbyEDPSciences
682 C.Adamietal.:Comaclustergalaxypopulations
techniqueappliedtothegalaxyluminosityfunction(GLFhere-
195.300 195.000 194.700
after)determination.Forthis,wewillcomparethepresentpho-
tometricredshifttechniquewiththegalaxyluminosityfunctions 28.400
computed with the same data set but with a statistical back-
groundremoval(Adamietal.2007a,b). 28.300
We will also compute luminosity functions in the u∗ band
based on statistical background subtraction in the area where 28.200
onlytheu∗bandisavailable.
Wedescribeournewphotometricdataandphotometricred- 28.100
shift and PDF estimates in Sect. 2. We compute in Sect. 3 the
ComaGLFusingthePDFandinSect.4theComau∗ GLFap- 28.000
plying statistical subtraction. The spatial distributions of mul-
ticolor galaxy types are discussed in Sect. 5. Conclusions are 27.900
drawninSect.6.
Inthispaperwe assumeH =70 kms−1 Mpc−1, Ω = 0.3, 27.800
0 m
ΩΛ=0.7,adistancetoComaof100Mpc,adistancemodulus=
35.00,andascaleof0.46kpcarcsec−1. 27.700
27.600
2. Photometricdataandphotometricredshifts
2.1.B,V,RandIbandimagingdata
Fig.1.Megacamu∗bandweightimage(thedarkerthecolor,theshorter
These data are fully described in Adami et al. (2006a) and we theexposuretime).Thecoordinatesindegreesareindicated.
give here only the salient points. Coma was observed with a
2 field mosaic of the CFH12K camera in 4 bands(B,V, R and
I). Altogether,our data cover a 52 × 42 arcmin2 field over the the galaxy positions (assuming an initial 3 arcsec identifica-
sky (the CFH12K f.o.v. in the following) centered on the two tion distance) in order to find the best identification. We could
dominant cluster galaxies (NGC 4874 and NGC 4889) with a also have applied SExtractor in double-image mode, but this
completenesslevelinRclosetoR ∼ 24.Theseeingconditions wouldhaverequiredto degradealltheMegacamimagestothe
were allcloseto 1 arcsec.Thesedataare availableathttp:// CFH12Kcharacteristics.
cencosw.oamp.fr.We have already published several papers We note that the seeing values slightly differed from one
basedonthesedata(Adamietal.2005a,b,2006a,b,2007a,b). bandtoanother,andthiscouldproduceoffsetsintheestimated
magnitudevalues.We didnotattempttocorrectforthisseeing
effect before computing the photometric redshifts. However, it
∗
2.2.Newu bandimagingdata
wastakenintoaccountwhenfittingtheobservedmagnitudesto
Newu∗ banddata(seeFig.3)includingthepreviousfieldwere the synthethic SEDs by the photometric redshift codes for ob-
obtainedbetween2006and2007withtheCFHMegacamcam- jectsofourspectroscopiccatalog(seefollowing).Thesecorrec-
era.Megacamisamosaicof36individualCCDs,givingafield tionsprovedtobesmall.
ofview(f.o.v.)of1deg2.Theaverageseeingwas1.1arcsec.The
totalexposuretimewas9.66h,obtainedbycombining58indi-
2.3.Zeropointvariationsacrossthefieldofview
vidualspatiallyditheredexposuresof10mineach.Weshowin
Fig. 1 the weight map that was generated during the reduction Besidesthepossibleseeingeffects,itisimportanttoassesspos-
procedure. This gives a good idea of the exposure time varia- sible magnitude zero point variations across our field of view.
tions across the field of view (the darker the color, the shorter Given the u∗ data treatment, the u∗ zero points are stable over
the exposure time). The sigma of the distribution of the expo- the whole field of view at better than 0.05 mag (a calibration
sure times per pixel is ∼30%of the mean value.However,this atbetterthan∼4%isachievedbytheTerapixElixir codefrom
affects only the gaps between individual CCDs, that represent thefieldcentertothefieldedges).RegardingtheCFH12Kdata,
lessthan10%ofthetotalarea. we alreadyestimated inAdamietal.(2006a)thatcorrectionin
The data reduction was performed with the Terapix tools magnitudedueto smallairmass variationswouldhave been,at
(http://terapix.iap.fr/) and the standard procedure ap- maximum,of0.033magforR,0.020magforV,and0.015mag
plied for example to the CFHTLS fields (McCracken et al. for B, much less than the estimated errors on the magnitudes.
2008). A ∼1 deg2 image was produced with a zero point set Inthe samepaper(Fig.16)we alsoconsideredstellartracksto
to 30.0. We extracted an object catalog from this image with estimate the additional shifts that must be applied to the zero
the SExtractor package (Bertin & Arnouts 1996) with a detec- points to match the empirical stellar library of Pickles (1998).
tion threshold of 0.4 and a minimum number of pixels (above The mean (over the whole f.o.v.) corrections were negligible.
this threshold) of 3. We derived total Kron AB magnitudes However,this doesnotexcludesome possible small zeropoint
(Kron 1980) to compute photometric redshifts. We corrected variationsaccrossthefieldofviewthatcouldremainundetected
each detected object for galactic extinction using the Schlegel due to the width of the observedstellar tracks (on the order of
maps. This catalog was then cross-correlated with the B, V, 0.2mag).Inordertotestthispoint,weredidthesamefigureas
R, and I catalog already available. The cross-correlation was Fig.16ofAdamietal.(2006a),separatingthenorthernandthe
made by applying a technique similar to that of Adami et al. southernarea(asdefinedinAdamietal.2007a).Figure2shows
(2006a) and optimized to take into account small astrometric that we nearlyhave the same agreementbetweenthe empirical
differences between CFH12K and Megacam data. We defined stellar library of Pickles (1998) and the observed stellar tracks
small regions 2 arcmin wide in which we iteratively adjusted forboththenorthernandsouthernfields.Weestimatedthemean
C.Adamietal.:Comaclustergalaxypopulations 683
of 1.5, this u∗ completeness level is fully consistent with the
previouslyestimatedRcompletenesslevel(R∼24).
2.5.Photometricredshifttechniques
We have computedphotometric redshiftswith two of the most
well-knownpackages: Hyperz (e.g. Bolzonella et al. 2000and
1)and LePhare(e.g.Ilbertetal.2006aand2 (authors:Arnouts
&Ilbert)),mainlytocheckthatthesetwosoftwaresgiveconsis-
tentresults.Thesetoolsarefullydescribedinthequotedpapers.
In a few words, Hyperz (Bolzonella et al. 2000) and LePhare
(Arnouts & Ilbert) are both based on a template-fitting proce-
dure. However, the setting of the two photo-z codes presents
two main differences:Hyperz includes a large set of templates
fromBruzual&Charlot(2003)withdifferentstarformationhis-
tories and differentages. LePhare includes a limited library of
9 templatesfromPolletta et al. (2007) and 3 star-formingtem-
platesfromBruzual&Charlot(2003).InHyperz,thezero-point
calibration has been done by comparing the stellar locus in a
color−colordiagramwiththeexpectedstarcolors.In LePhare,
the zero-point calibration has been done with an iterative pro-
cedure described in Sect. 4.1 of Ilbert et al. (2006a), by com-
paring the predicted magnitudesfrom the best fit template and
the observed magnitudes. In order to compute these possible
small magnitude zero point shifts in LePhare, we trained the
photometricredshiftestimateswiththespectroscopiccatalogde-
scribedinAdamietal.(2005a).Thiscatalogwassupplemented
with20archiveredshiftsfromtheKeckLRISmultispectrograph
Fig.2.R−IversusB−V forstars(asdefinedinAdamietal.(2006a). (Secker et al. 1998), increasing the total number of redshifts
Dots:observedstars,openred(greyinblackandwhiteversion)circles: available in the CFH12K f.o.v. by ∼5%. These galaxies have
empiricalstellarlibraryofPickles(1998).Theupperandlowerfigures magnitudesbetweenR=16.5and21.5andredshiftslowerthan
correspondtothenorthandsouthCFH12Kfieldsrespectively.
0.5(onlyoneoftheseobjectsispartoftheComacluster).
The additional zero point shifts applied to our photome-
try to compute optimal photometric redshifts with Hyperz and
LePhare were quite small. Hyperz requiredmanuallyimposed
magnitudeshift between the northernand southernfields to be
shiftsoflessthan0.05maginRandI andnoshiftinu∗, Band
smallerthan0.05mag.
V.LePharerequiredshiftsof0.028±0.02inu∗,−0.141±0.09
inB,−0.062±0.03inV,0.±0.04inRand0.12±0.05inI(er-
∗
2.4.Completenessoftheu image rorbarscomingfromthedispersionoftheshiftsoverthewhole
spectroscopicsample).
ThegoodefficiencyofMegacamintheblueallowedustoreach
A set of fit parameterswas then produced(see also Adami
asimilardepthinu∗asintheB(∼24.75forpointlikeobjectsat
et al. 2008): internalextinction,age, multicolortype (based on
the 90% level, see Adamiet al. 2006a), V (∼24),R (∼24),and
a color classification in a 5 mag space), redshift, and PDF es-
I (∼23.25) images. These new data were required to compute
timates. However, we mainly consider here the (photometric)
photometricredshiftsinordertohaveabluephotometricband,
redshift, the template multicolor type, and the PDF estimates.
encompassingthe4000Åbreak. Types 1 to 4 are the Elliptical, Sbc, Scd and Irr from Coleman
ThepercentageofRdetectionswithanassociatedu∗ detec- et al. (1980) respectively and types 5 are starburst 1 templates
tion is between 87% and 92%. This leads to estimate the 90% fromKinneyetal.(1996).
completeness level of the u∗ image to R ∼ 24. The ∼10% of Wealsocomputedtheintegratedprobabilityforeachgalaxy
galaxiesdetectedinRandnotinu∗are,atleastpartially,objects to be at a redshift lower than a given value z (P hereafter).
lim
thatarenotformingstarsandwhichthereforedonotappearin This is the area of the PDF enclosed in the consideredredshift
u∗eveninadeepexposure. interval:[0.,z ].
lim
In order to assess the completeness level of our u∗ image
in another way, we directly compared our u∗ counts with
the u∗ counts of the CFHTLS D1, D3 and D4 fields (T0004 2.6.Advantagesofthephotometricredshifttechnique
release). These fields have long enough exposure times (see We could argue that a simple color−magnituderelation (CMR
http://terapix.iap.fr/cplt/table-syn-T0004.html) hereafter) could also provide a good way to discriminate be-
tobecomparablewithourdata.WedidnotconsidertheD2field tweenclustermembersandnonmembers(Bivianoetal.1996).
thatonlyhasatotalexposuretimeof1.3hinu∗.Figure4shows However, besides the simple fact that five bands appear intu-
that the Coma countsdominatethe D1, D3, and D4 countsfor itively better than two to characterize the spectral energy dis-
u∗ ≤ 25, due to the presence of the highly populated Coma tributions of galaxies, the use of a CMR first requires the
cluster in the field. Our data become incomplete at fainter
magnitudes, leading to estimate the u∗ completeness level to 1 http://webast.obs-mip.fr/hyperz/
∼25.5. Given that the u∗−R color in our data is on the order 2 http://www.oamp.fr/people/arnouts/LE_PHARE.html
684 C.Adamietal.:Comaclustergalaxypopulations
195.300 195.000 194.700
28.400
G4
28.300
28.200
28.100
28.000
G5
G3 27.900
27.800
G2
27.700
G1
27.600
Fig.3.Megacamu∗ bandimageoverplottedwiththe5detectedgroups(circlescorrespondto300kpcindiameteratz = 0.1)andtheNeumann
etal.(2003)X-raysubstructures.WecorrectedforasmalloverallshiftbetweentheXMMandtheMegacamastrometry.
existenceofawelldefinedredsequence(RShereafter).Thisis earlytypegalaxycoresshowninAdamietal.2006b).Thesim-
notalwaysverifiedeveniftheRSinComaisverywellknownat pleextrapolationofthebrightComaclustergalaxyRSatR>21
brightmagnitudes,thankstothelargepercentageofbrightellip- wouldmissthesepeculiarobjectsasclustermembers.
ticalgalaxiesinthiscluster(e.g.Adamietal.1998).Inthiscase,
the photometricredshiftand CMR techniquesgive comparable Third,thephotometricredshifttechniqueoffersamuchmore
results. direct way to characterize the spectral energy distributions of
the galaxies. The use of the CMR RS can at best discriminate
Second, this RS is only poorly known beyond the present betweenearlyandlatetypegalaxiesfornottoofaintmagnitudes.
spectroscopic limit (typically R ∼ 21 for the Coma cluster). It On theotherhand,the photometricredshifttechnique(alsosee
may show a reddeningat R > 21, as suggested by e.g. Adami Sect. 2.8) allows the separation of galaxies in 5 spectral types
etal.(2000).Moreover,atR > 21peculiargalaxytypesappear, andgivesadirectestimateoftheagesandinternalextinctionsof
directlyresultingfromthedisruptionsoflargergalaxies.Anex- thegalaxies.
amplearethetidaldwarfgalaxies(Bournaudetal.2003).These
peculiargalaxies(notonlytidaldwarfgalaxies)sometimeshave For all these reasons, we decided to apply the photometric
ellipticallikespectraltypes,butareatypicallyred(e.g.depleted redshifttechniqueinthispaper.
C.Adamietal.:Comaclustergalaxypopulations 685
agivengalaxy.Thevalueofz willbechosentominimizethe
lim
number of intruders (galaxies at spectroscopic redshift greater
thanz thathaveaphotometricredshiftlowerthanz )andof
lim lim
lost galaxies (galaxies at spectroscopic redshift lower than z
lim
withaphotometricredshiftgreaterthanz ).Thisminimization
lim
wasdoneonthebasisofthespectroscopiccatalogandtheresults
are shownin Fig.6.Setting z to a highvaluewouldproduce
lim
nointruderorlostgalaxies,butthediscriminationpowerofsuch
a high value would be low. We therefore selected z = 0.20,
lim
which is a good compromize between intruder and lost galaxy
percentagesandthediscriminationpowerofthemethod.Wesee
inFig.6thatthepercentagesofintruderorlostgalaxiesdonot
decrease very strongly for z ≥ 0.2. With this limit and consid-
eringourspectroscopiccatalog,weestimatethatwelose(inthe
Fig.4. Coma line of sight u∗ counts (solid line) compared to three magnituderangecoveredbythespectroscopiccatalog)lessthan
deeper u∗ fields from the CFHTLS:D1 (short dashed line), D3 (long 10% galaxies and we include about 15% intruders. These per-
dashedline),andD4(short/longdashedline). centages will be taken into accountwhen computingthe GLFs
inthenextsection.
2.7.Results
2.8.Uncertaintiesinthegalaxydiscriminationbased
First,welimitedthecatalogtotheobjectsdetectedinallu∗, B, onthePDF
V,R,and I bands.Thisexcludedthesouth-westareaforwhich
With this limit of z = 0.2 and considering our spectroscopic
BandV dataarenotavailable(seeAdamietal.2006a).
catalog,weestimatethatweloselessthan10%galaxiesandwe
The two sets of estimates (from LePhare and Hyperz)
include about15% intrudergalaxiesif we limit our analysisto
are most of the time in good agreement. As usual, we lim-
thespectroscopiccatalogmagnitudelimit.
ited the sample to unmasked regions (this leads to remove all
Given the fact that our spectroscopic catalog is by far not
objects at less than twice the radius of objects brighter than
R = 18) to avoid uncertain magnitude estimates, and we plot- as deep as our photometric catalog, we need to quantify the
accuracyofthe photometricredshiftsbeyondthe spectroscopic
tedforbothmethodsthephotometricversusspectroscopicred-
limit.Ilbertetal.(2006a)haveshownthatthe1σerrorbarsare
shifts.Ourspectroscopiccatalogincludes172spectroscopicred-
shifts in non masked areas, among which 103 are at z ≤ 0.2. representative of a measurement at 68% confidence level, and
Figure 5 shows the results for LePhare and Hyperz. As both we can therefore quantify this accuracy based on the 1σ error
bars.Figure7showsthe fractionofphotometricredshiftsfrom
methods provide very similar results, we merged the two esti-
mates (LePhare and Hyperz) keeping the value with the best LePhare(resultswouldbesimilarwithHyperz)witha1σerror
reducedχ2(afterhavinghomogenizedthetwosetsofχ2). barsmallerthan0.2×(1+z)(0.2beingtypicallythewidthofthe
lowredshiftintervalwewanttocharacterize).Thisfigureshows
Beyond the general agreement between photometric and
that we still have more than half of the galaxies with a photo-
spectroscopicredshifts,weclearlyseeadegeneracyattheComa
metric redshiftestimate lower than 0.2× (1+z) for R brighter
cluster redshift. For some galaxies which have spectroscopic
than22.Wealsohaveagoodagreementbetweenthedecreaseof
redshifts inside the cluster, LePhare and Hyperz producepho-
thefractionofphotometricredshiftswitha1σerrorbarsmaller
tometricredshiftsnotonlyattheclusterredshiftbutspreadover
than0.2×(1+z)andthenumberofavailablespectroscopicred-
the interval z = 0, z ∼ 0.2. Moreover,there is a clear system-
shifts(shownasthehistograminFig.7).Thesepercentageswill
aticoffsetontheorderof0.05inredshift(mainlyvisibleinthe
betakenintoaccountwhencomputingtheGLFuncertaintiesin
LePhare results) between z = 0.1 and 0.4. This bias can have
thenextsection.
two origins. First, the magnitude estimates can be partially bi-
ased because of observational effects (small seeing variations
betweenthe differentphotometricbands,peculiarproblemsfor 2.9.Uncertaintiesinthegalaxycolor-typeestimate
agivenbandatagivenskylocation,etc.).Second,clustergalax-
ies are in general redder than their equivalents in the field due In order to determine the reliability of color-types assigned to
to environmental effects, while photometric redshift estimates galaxies in the Coma cluster, we have carried out a series of
are computedwith syntheticgalaxytemplatesmainlybased on simulations with Hyperz and related software. Synthetic cata-
fieldgalaxies.Thiscanleadtoamis-interpretationofthespec- logs contain 105 galaxiesin total at the redshiftof Coma, with
tral energy distribution, confusing the intrinsic red color and R-bandmagnitudesrangingbetweenR=19and24,spanningall
the reddening of galaxies due to redshift. This shows the need the basic spectraltypes,i.e. ∼4000galaxiespermagnitudeand
forthedefinitionofsyntheticclustergalaxyspectroscopictem- typebin.Photometricerrorsinthedifferentfilterswereassigned
plates. This also implies that we cannot use directly the pho- followingaGaussiandistributionwithvariancescaledtoappar-
tometric redshift estimates to efficiently discriminate between entmagnitude,i.e.σ(m) (cid:7) 2.5log[1+1/(S/N)],whereS/N is
clustermembersandfieldgalaxies. the signal to noise ratio corresponding to the apparent magni-
Moreover, certain galaxies may have a very extended PDF tudeminourcatalog.Hyperzsettingsusedtofitthesesynthetic
and consideringjustthe value givingthe maximumprobability catalogsarethesameasforrealcatalogs.
could producea wrongphotometricredshift. We will therefore Table1summarizestheresultsobtainedfromthesesimula-
classify galaxies as being at a greater or lower redshift than a tions as a functionof R-bandmagnitudeand photometrictype.
limitingvaluez ,basedontheprobabilityPquotedaboveand UptoR≤23(weassumedalimitofR=22.5inthefollowing),
lim
computedusingtheareabelowthePDFforzlowerthanz for photometrictypesarecorrectlyretrievedbySEDfitting,witha
lim
686 C.Adamietal.:Comaclustergalaxypopulations
Fig.5.LePhare(left)andHyperZ(right)photometricredshiftsintheCFH12Kf.o.v.andoutsidethemaskedregionsversusspectroscopicredshifts.
Thetwodashedlinesshowa±0.2uncertaintyenveloppearoundtheperfectrelation(continuousline).Thelargenumberofdotsbelowz ∼0.2
phot
butoutsideofComaisexplainedbythegroupsdetectedinSect.3.2.
Table1.PercentageofsimulatedgalaxiesintheComaclusterwithpho-
tometrictypescorrectlyassignedasafunctionofR-bandmagnitude.
R E/S0 Sbc Scd Im SB
mag % % % % %
19–20 100.0 99.9 99.9 98.9 98.8
20–21 100.0 99.9 99.9 98.7 99.2
21–22 100.0 99.7 99.8 97.9 97.5
22–23 100.0 98.3 98.2 93.3 90.3
23–24 98.0 88.4 84.7 76.2 77.3
quotedarethereforelowerlimitsonthetrueuncertaintiesonthe
Fig.6.Variationofthepercentageofintruders(fullblackline)andlost galaxycolor-typeestimates(seealsoSect.5).
galaxies(green(greyinblackandwhiteversion)dottedline)asafunc-
tionofz .ErrorbarsarePoissonianuncertainties.
lim
3. Thegalaxyluminosityfunctioncomputation
Aspreviouslystated,photometricredshifttechniquesallowone
tomakeanoptimalsubtractionofbackgroundgalaxiesforared-
shiftlimitof0.2.However,westillneedtoremovefromthesam-
plethegalaxiesatz≤0.2whichdonotbelongtotheComaclus-
ter. Thesegalaxiescanbefieldgalaxiesorgalaxiesincludedin
groupsofgalaxiesnotrelatedtoComa.Wechosetosubtractsta-
tisticallythesetwocontributionsbyconsideringfieldandgroup
luminosityfunctionstaken in the literature.In thisway,we are
still applying a statistical subtraction, but photometric redshift
techniqueshelptocutdownverysignificantlytheredshiftrange
thatincludesbackgroundgalaxies.
3.1.Fieldcontribution
Fig.7.Fractionofphotometricredshifts(asafunctionofRmagnitude)
fromLePharewitha1σerrorbarsmallerthan0.2×(1+z).Solidline:
We estimated the field contribution to be subtracted from the
photometricredshiftslowerthan0.2,dashedline:photometricredshifts
VVDSfieldluminosityfunctionofIlbertetal.(2005),whoob-
greaterthan0.2.Thehistogram(inarbitraryunits)showsthenumberof
tained a very large redshift catalog complete to I(AB) ∼ 24
availablespectroscopicredshiftsasafunctionofRmagnitude.
(close to R ∼ 24 considering the passband of our filters). This
is close to our own magnitude limit and ensures that the field
luminosity function we subtract is really constrained over our
percentageof failuresusuallybelow∼3%, andupto ∼10%for wholemagnituderange,withoutrequiringanyextrapolation.
the bluest galaxies. As expected, early-type galaxies are better TheIlbertetal.(2005)luminosityfunctionisalsocomputed
identifiedthanlatetypes,evenforthefaintestsourcesinourcat- infivebandsfromU toI,allowingafieldsubtractionadaptedto
alog, butthe differenceis small. At least ∼75%of galaxiesare eachphotometricband.
stillcorrectlyclassifiedinthefaintestmagnitudebin. Thesubtractionofthefieldcontributionwassimplydoneby
Of course, since we fit the templates used to generate the computingthecosmologicalco-movingvolumeincludedinour
catalogs, this must help recovering the color-types; the errors fieldofviewatredshiftlowerthan0.2andusingtheφ∗,α,and
C.Adamietal.:Comaclustergalaxypopulations 687
Table 2. Characteristics of the background groups detected by the 4. Galaxyluminosityfunctions(GLF)
Serna-Gerbal method, based on spectroscopic redshifts: coordinates,
meanredshiftandtotalmass(Adamietal.2005). 4.1.Galaxyluminosityfunctionsbasedonphotometric
redshifts
Group α δ Meanz Mass(M(cid:8))
G1 195.04 27.61 0.149 7.34×1012 We computedtheGLFinthesubregionspreviouslyconsidered
G2 194.90 27.71 0.138 5.25×1011 by Adami et al. (2007a). These ∼10(cid:9) × 10(cid:9) regions cover the
G3 194.92 27.87 0.144 1.25×1013 whole cluster area and represent a good compromize between
G4 194.94 28.27 0.097 1.46×1011 thespatialresolutionandthenumberofgalaxiesincludedinthe
G5 194.70 27.89 0.133 4.15×1012 individual GLFs. Figure 8 shows the GLFs together with their
1σ error bars (red symbols) overplotted on the previous GLFs
computedbyAdamietal.(2007a:blacksymbols)withstatistical
M∗ parameters of the Schechter function given by Ilbert et al.
backgroundsubtractions.
(2005).
Thegoalofthissectionismainlytocompareourresultswith
thoseofAdamietal.(2007a).Asthesetwostudiesarebasedon
3.2.Groupcontribution thesamedataset,thisisagoodwaytotestthereliabilityofthe
two methods (statistical field subtraction and photometric red-
We first located in our field of view the groups unrelated with
shifttechnique).Theagreementisgenerallyquitegoodbetween
theComaclusterandatredshiftlowerthan0.2byapplyingthe
thetwoestimates.Afewregions,however,showsignificantdis-
Serna-Gerbalmethod(Serna&Gerbal1996)totheredshiftcat-
crepancies:fields5,6,8,9,16,and20.Thesefieldshavediffer-
alog(Adamietal.2005a).Briefly,thismethodisabletodetect
entluminosityfunctions(derivedwiththetwotechniques)over
dynamicallylinked galaxies(whatwe calla groupof galaxies)
morethan15%ofthemagnitudeintervalR =[19.25,22.5],the
basedonthepositions,magnitudesandredshiftsofthegalaxies
largestdifferencesoccuringforfields16 and6.Itisinteresting
(seeTable2).ThismethodwasalreadyappliedbyAdamietal.
to note that among these fields, three contain very bright stars
(2005a)to studythe Comacluster,andfivebackgroundgroups
atz≤0.2weredetected. (fields5,8,and9)andone(field8)theclusterdominantgalaxy
NGC 4874. These objects have a very extended light halo that
The group luminosity functions were computed from the
possiblyaffectsthephotometry(eveniftheyweremaskedinside
SDSS groupluminosity functionestimates of Zandivarezet al.
twicetheirradius)andthereforethephotometricredshiftcompu-
(2006). Since these authors have data in the five SDSS bands,
tations,orthefaintobjectdetectioninthecaseofthestatistical
wetransformedtheirmagnitudestoourownsystemapplyingthe
backgroundremovaltechniqueofAdamietal.2007a).Wealso
relationsgivenbyFukugitaetal.(1995).Thenumberofgalax-
ies brighter than R = 17.75 (where our spectroscopic catalog knowthatatleastoneofthesefields(field9)containsasignif-
icantpopulationofveryfaintandveryblueComagalaxies(the
isnearlycomplete)wasestimatedfromourspectroscopiccata-
log.Thisallowedustocomputetheφ∗normalizationofthefive field aroundNGC4858/4860,Adamietal. 2007b). Thesefaint
blue knots, similar to those discussed by Cortese et al. (2007),
detectedgroups(seeFig.3).
are perhaps not well represented in our galaxy synthetic tem-
Inordertosubtractthisgroupcontribution,weassumedthat
plates;thiscouldproduceanincorrectvalueforphotometricred-
thewholegroupgalaxycontributionwasenclosedina300kpc
shifts,outsidetheComaclusterrange.
diametercircle,atypicalgroupsize.
We note here that we do not know the background group Inordertohaveathirdestimate(beyondAdamietal.2007a,
GLF per spectral type and this prevents us from computing a andthepresentpaperdeterminations),wealsocomparedourre-
ComaclusterGLF per spectraltypewiththephotometricred- sults for one of these possiblybiased fields (the NGC 4874re-
shifttechnique.Wewillsimplystudythegeneralmulticolortype gion) with a study dedicated to the computationof the GLF in
spatialdistributioninSect.5. theComaclustercenter(Seckeretal.1996,basedonKeckim-
ages).Figure9showsthattheseauthorspredictgalaxynumber
densitiesconsistentatthe2σlevelwiththephotometricredshift
3.3.GLFuncertainties
technique estimates in the range where these counts are lower
We tookintoaccountseveraluncertaintiesontheComacluster than the onescomputedin Adamietal. (2007a). So the counts
GLF(computedin0.5magbins,asinAdamietal.2007a): ofAdamietal.(2007a)couldalsobeoverestimated(despitethe
great care that was devoted to the statistical field subtraction).
– thePoissonnoiseineachbin;
It is beyond the scope of this paper to present a full compari-
– theuncertaintyonthefieldandgroupluminosityfunctions,
sonwithliteratureresults.ThiswasalreadydoneinAdamietal.
computed by generating 1000 field and group luminosity
(2007a:Sect.7.4).
functionswithparametersGaussianlyvaryinginsidetheer-
We also note that we still see the discrimination between
rorrangequotedintheliterature;
the north-northeast and the south-southwest parts of the clus-
– the uncertainty due to the photometric redshift computa-
ter (Fig. 10) that was detected by Adami et al. (2007a). The
tions. As previously shown, the number of galaxies below
z = 0.2canbeoverestimatedby15%orunderestimatedby north-northeastregionismorepopulatedinthefaintmagnitude
regime than the south-southwest regions. However, this trend
10% within the magnitude range where the spectroscopic
is not significant when applying the photometric redshift tech-
catalog is contributing. For fainter magnitudes, we consid-
nique.Wealsoseeinthesouthernclusterpartasimilarturn-over
eredFig.7 toinfera valueoftheoverestimationandofthe
astheonealreadyshowninRin Adamietal. (2007a),butitis
underestimation.
poorly significant due to the large error bars. These error bars
Thesumofthethreeuncertaintiesisassumedtobethetotalerror could only be reducedby using a deeper spectroscopiccatalog
andisclosetoa3σerror.Theuncertaintyduetothephotometric tobetterconstrainthephotometricredshiftestimatesofthefaint
redshiftcomputationsisclearlythedominatingsourceoferror. galaxies.
688 C.Adamietal.:Comaclustergalaxypopulations
Fig.8.Rband GLFsfordifferent regionsintheComacluster.Northistopandeast isleft.Theemptysubgraphs correspond toareaswhere B
andV datawerenotavailable.Blackcontinuousandshortdashedlines(alongwitherrorbars)aretheGLFstatisticalestimatesofAdamietal.
(2007a).Red(greyinblackandwhiteversion)long-dashedlines(alongwitherrorbars)arethepresentestimates.
Fig.9.RbandGLFfortheNGC4874fieldcomputedwithphotometric Fig.10.RbandGLFforthenorthern(black)andsouthern(reddashed
redshifts(reddashedline–greyinblackandwhiteversion).Theslope line–greyinblackandwhiteversion)partsasdefinedinAdamietal.
fromSeckeretal.(1996)isoverplotted(blackfullline). (2007a).
4.2.Galaxyluminosityfunctionsintheu*bandbased
Thecomparisonfieldscoveratotalareaofmorethan3deg2,
onstatisticalbackgroundsubtraction
thusreducingthecosmicvariance(seeAdamietal.2007a).The
The Megacam u∗ band image is more extended than the resultingGLFsaresimilartotheu∗bandGLFscomputedinthe
CFH12K f.o.v. We typically have a 7 arcmin strip all around CFH12K f.o.v.basedonphotometricredshifts.We sub-divided
theCFH12Kf.o.v.whichisonlycoveredinu∗.Inordertocom- theMegacamf.o.v.in7×6subfieldsandcomputedtheslopeα
puteu∗ GLFsinthisexternalarea,weappliedthesamestatisti- of the u∗ GLF (computedapplyinga statistical subtractionand
calfieldgalaxysubtractiontechniqueasinAdamietal.(2007a modelledbyaSchechterfunction)intheu∗=[22,24]magrange
andb).Themaingoalofthiscomputationwastoinvestigatepo- foreachzone.The42αslopesaregiveninTable3andwereused
tentialvariationsoftheGLFfaintendslopeintheu∗band,which to generate Fig. 11, where the contoursof GLF slope α within
is sensitive to recent star formationbursts, and to compareour theMegacamf.o.v.aredisplayed.
resultstothoseofDonasetal.(1995). Weclearlyseeasteepeningoftheu∗GLFslopebetweenthe
ThecomparisonfieldswerethoseamongthedeepCFHTLS centerandoutskirts,changingfrom−1to−2.2.Theregionswith
fields (D1, D3, and D4) that are deep enough to allow us to thesteepestu∗ GLFscorrespondtothosewhereX-raysubtruc-
make a statistical subtraction from our Coma data. In order to turesaredetectedandarelocatedat∼25(cid:9)fromtheclustercenter.
limitthesecatalogstothesamedepth,weselectedonlygalaxies Thisis exactlywhereDonasetal. (1995)detectedanenhance-
brighterthanu∗ = 24intotalmagnitudeandbrighterthan26.3 ment in the median UV flux and in the fraction of bright blue
insurfacebrightness. starforminggalaxiesconsideredasclustermembers.
C.Adamietal.:Comaclustergalaxypopulations 689
Table3.Slope(andassociated1-σuncertainty)oftheu∗Schechterlu- our analysis to the search of galaxy concentrations outside the
minosityfunctionbetweenu∗=22and24asafunctionofcoordinates. maskedareas.
ThesevalueswereusedtogenerateFig.11. In order to estimate the multicolor type spatial distribution
insidetheComacluster,wemustadresstheproblemofz ≤ 0.2
α δ Slope Err.slope non Coma member galaxies (field or loose group objects) that
12.9639 27.5911 –1.36 0.22 cannot be discriminated by the photometric redshift technique
12.9639 27.7339 –2.22 0.61 alone.
12.9639 27.8767 –1.52 0.18
Ifweconsiderthefieldgalaxyluminosityfunctioncomputed
12.9639 28.0194 –1.59 0.28
by Ilbertet al. (2006b) fromsimilar data,we find thatthe field
12.9639 28.1622 –1.18 0.17
contribution represents about 15% of the Coma cluster galax-
12.9639 28.3050 –1.96 0.21
12.9639 28.4478 –2.20 0.27 ies down to R = 24. Among these 15%, about 1/4th are bulge
12.9761 27.5911 –1.48 0.20 galaxiesand3/4arediskgalaxies.However,thiscontributionis
12.9761 27.7339 –1.71 0.21 spread over the whole field of view and will act as a homoge-
12.9761 27.8767 –1.61 0.09 neousbackgroundcontributionofgalaxies.Itwillthereforenot
12.9761 28.0194 –1.84 0.56 modify the relative variation of the estimated multicolor types
12.9761 28.1622 –1.84 0.56 inside the Coma cluster. We could argue that dense regions of
12.9761 28.3050 –1.43 0.41
filaments (without being real massive structures such as clus-
12.9761 28.4478 –1.96 0.40 ters or groups) at z ≤ 0.2 could provide different galaxy type
12.9883 27.5911 –2.07 0.74
counts across the field of view because they probably contain
12.9883 27.7339 –0.34 0.57
moreearlytypegalaxies.However,evenconsideringthelargest
12.9883 27.8767 –1.67 0.31
12.9883 28.0194 –0.13 0.65 knowncosmicbubblesizes,atleastadozenofsuchbubblesare
12.9883 28.1622 –0.98 0.28 superimposedbetweenComaandz = 0.2.Thesumofallthese
12.9883 28.3050 –1.81 0.23 bubbles therefore homogenizes the distribution of field galaxy
12.9883 28.4478 –0.08 1.51 morphologicaltypesatz≤0.2.
13.0006 27.5911 –1.41 0.49 Similarly, we estimated that the contribution of galaxies in
13.0006 27.7339 –0.32 0.63 loosegroupsrepresents∼5%oftheComaclustergalaxiesdown
13.0006 27.8767 –1.66 0.33 to R = 24 over the whole field of view. However, these con-
13.0006 28.0194 –1.62 0.24
tributions are concentrated in precise locations, so locally, the
13.0006 28.1622 –0.85 0.18
contribution can be much higher. Assuming the group GLF of
13.0006 28.3050 –1.08 0.26
Zandivarezet al. (2006) and the galaxyspectral type estimates
13.0006 28.4478 –1.99 0.37
13.0128 27.5911 –0.73 0.16 inlowmassgroups(≤1013.5M(cid:8))ofDominguezetal.(2002),we
13.0128 27.7339 –1.18 0.73 estimatethatalongthelinesofsighttoloosegroups,galaxiesin
13.0128 27.8767 –1.25 0.22 groupscanrepresentmorethan85%ofallz≤0.2galaxies.This
13.0128 28.0194 –1.06 0.13 contributionisontheorderof50%forearlytypesgalaxiesand
13.0128 28.1622 –1.96 0.41 can reach 100%for other galaxytypes. As we cannotestimate
13.0128 28.3050 –1.51 0.18 preciselythegalaxytypecontributionsforthe5loosegroupsde-
13.0128 28.4478 –2.62 0.53
tectedbehindComa,thisclearlysuggeststhatwehavetoremove
13.0250 27.5911 –1.88 0.37
theseareas.
13.0250 27.7339 –1.52 0.39
Maps of galaxy multicolor type density were then gener-
13.0250 27.8767 –2.12 0.78
atedfromasimplecountincelltechnique,avoidingthemasked
13.0250 28.0194 –2.30 0.43
13.0250 28.1622 –2.12 0.28 regions and considering the whole sample of galaxies down
13.0250 28.3050 –1.43 0.18 R = 24. The original cell size was 1(cid:9) and we applied an addi-
13.0250 28.4478 –1.63 0.36 tionnalsmoothingof3×3pixelsinFig.12.Weonlytookinto
accountcellscompletelyincludedin ourfield ofview,so there
are no edge effects. This figure shows that early type galaxies
are spread over the whole field of view with the possible ex-
5. Multicolortypespatialdistribution ception of the north-west regions. Early and late Spirals show
localiseddensityenhancementsintheclusteroutskirts.Wenote
Thegoalofthissectionistoinvestigatethespatialdistributionof however that the late (and perhaps early) spiral concentration
galaxiesaccordingtotheirmulticolortype.Westresshereonce located north of the cluster center could be well explained by
againthatourmulticolortypesarebasedonacolorclassification the loose G4 group. Later types (irregulars and starbursts) are
ina5magspaceandarenotrealmorphologicaltypes.Realmor- alsodistributedpreferentiallyintheclusteroutskirts.Inparticu-
phologicaltypesarealreadynotwelldeterminedatR≥ 18(our lar,starburstgalaxiesaremainlylocatedatthenorthwest,where
starting magnitude) and completely unknown at R ≥ 22, wait- X-raysubstructuresarealso detectedasaninfallinggroup(see
ingfortheresultsoftheComaclusterHSTimagingsurvey(e.g. Adami et al. 2005a) and where a concentration of early mul-
Carteretal.2008).Ifweconsiderthemorphologicaltypescom- ticolor types is also detected. This is also the place where the
piledinBivianoetal.(1996)for8galaxiesfainterthanR = 18 u∗GLFisthesteepestofthefield(seeprevioussection).
for which we computed photometric redshifts, we find that all The first thing to note in Fig. 12 is that we confirm that
areclassifiedasellipticalgalaxiesandallhaveamulticolortype the Coma cluster is mainly populated by early type galaxies
assignedtoDelta burstsorellipticalgalaxies,fromtheBruzual (∼E+S0),asexpectedforsucharichstructure:about80%ofthe
&Charlot(2003)evolutionarysyntheticSEDs.Thisagreement galaxy population is made of ellipticals. Spirals representonly
clearlyneeds,however,tobeconfirmed. 15%ofthefaintgalaxies.
Duetomaskedregions,wepostponethegalaxydensitypro- Second, as expected and observed for bright galaxies (e.g.
filepermulticolortypedeterminationtoafutureworkandlimit Whitmore et al. 1993), there are later type galaxy clumps in
690 C.Adamietal.:Comaclustergalaxypopulations
Fig.11.Upperpanel:u∗bandGLFsfordifferentregionsintheComaclustercomputedwithastatisticalsubtraction(continuousblacklineswith
errorbars)andthephotometricredshifttechnique(red(greyinblackandwhiteversion)dotslinkedwithdashedlinesanderrorbars).Northis
top and east is left. TheCFH12K field of view isshown as the central red (grey in black and white version) rectangle. Theempty subgraphs
correspondtoareaswherethefieldcountsweregreaterthantheclustercounts.X-raysubstructuresareshownasthickcontinuouscontours.Lower
panel:contoursoftheGLFslopewithintheMegacamf.o.v.(thincontinuouslines:slopesbetween−1and−1.2,thindottedlines:between−1.2
and−1.4,thinlongdashedlines:between−1.4and−1.6,thinlongdot-dashedlines:between−1.6and−2.2).Shadedareascorrespondtozones
wheretheslopeisnotsignificantlydifferentfrom0atmorethana2σlevel.X-raysubstructuresfromNeumannetal.(2003)areshownassmooth
thickblackcontours.
theclusteroutskirts.Thisbehavioressentiallyknownforbright Third, we also see a clear concentration of starburst ob-
galaxies(R≤∼18fortheComacluster)isnowextended6mag jects in a place whereearlyspiralmulticolortypesare also de-
deeper,andforafinerandmoreobjectivetypeseparation. tectedandwhereX-raysubstructuresarepresent.Thesestarburst