Feldmeier et al. ApJ 575779

22 Pages·2002·2.23 MB·English

TheAstrophysicalJournal,575:779–800,2002August20
#2002.TheAmericanAstronomicalSociety.Allrightsreserved.PrintedinU.S.A.
DEEP CCD SURFACE PHOTOMETRY OF GALAXY CLUSTERS. I. METHODS AND INITIAL STUDIES
OF INTRACLUSTER STARLIGHT
JohnJ.Feldmeier,1 J.ChristopherMihos,2 HeatherL.Morrison,1,2 StevenA.Rodney, and PaulHarding
DepartmentofAstronomy,CaseWesternReserveUniversity,10900EuclidAvenue,Cleveland,OH44106;[email protected],
[email protected],[email protected],[email protected],[email protected]
Received2002January3;accepted2002April29
ABSTRACT
Wereporttheinitialresultsofadeepimagingsurveyofgalaxyclusters.Theprimarygoalsofthissurvey
are to quantify the amount of intracluster light as a function of cluster properties and to quantify the fre-
quencyoftidaldebris.Weoutlinethetechniquesneededtoperformsuchasurvey,andwereportfindingsfor
the first two galaxy clusters in the survey: A1413 and MKW 7. These clusters vary greatly in richness and
structure. We show that our surface photometry reliably reaches to a surface brightness of l ¼26:5 mag
v
arcsec(cid:1)2.Wefindthatbothclustersshowclearexcessesoverabest-fittingr1/4profile:thiswasexpectedfor
A1413butnotforMKW7.Bothclustersalsoshowevidenceoftidaldebrisintheformofplumesandarclike
structures,butnolongtidalarcsweredetected.WealsofindthatthecentralcDgalaxyinA1413isflattened
atlargeradii,withanellipticityof(cid:2)0.8,thelargestmeasuredellipticityofanycDgalaxytodate.
Subjectheadings:galaxies:clusters:general — galaxies:clusters:individual(A1413,MKW7) —
galaxies:interactions — galaxies:kinematicsanddynamics
1. INTRODUCTION darkmatterinclustergalaxies.Simulationshaveshownthat
thestructureofdarkmatterhalosingalaxiesplaysacentral
Theconceptofintraclusterstarlightwasfirstproposedby
roleintheformationandevolutionoftidaldebris(Dubin-
Zwicky (1951), who claimed to detect excess light between
ski,Mihos,&Hernquist1996,1999).Ifclustergalaxydark
thegalaxiesof the Coma Cluster. Follow-up photographic
halos are tidally truncated at small radii (e.g., Whitmore,
searchesforintraclusterluminosityinComaandotherrich
Forbes,&Rubin1988),strippedmaterialcanbemoreeasily
clusters (e.g., Welch & Sastry 1971; Melnick, White, &
unbound from the galaxies and end up being distributed
Hoessel 1977; see V´ıchez-Go´mez 1999 and Feldmeier 2000
smoothlythroughoutthecluster.Conversely,ifclustergal-
forreviews)producedmixedresults,anditwasnotuntilthe
axy halos survive, some tidally stripped material may
adventofCCDsthatmorepreciseestimatesoftheamount
remain bound to these galaxies, leaving them embedded in
of intracluster starlight were made (e.g., Uson, Boughn, &
verylowsurfacebrightness‘‘cocoons.’’TheICLmayactas
Kuhn1991;V´ılchez-Go´mez,Pello´,&Sanahuja1994;Bern-
asensitiveprobeofthemechanicsoftidalstripping,thedis-
stein et al. 1995; Gonzalez et al. 2000). These observations
tributionofdarkmatteraroundgalaxies,andclusterevolu-
are extremely difficult to perform and interpret because of
tioningeneral.
thelowsurfacebrightnessofthephenomenon:typically,the
Recently, much progress has been made in the study of
surfacebrightnessofintraclusterlightislessthan1%ofthe
intracluster starlight on numerous fronts. Individual intra-
brightnessofthenightsky.Measurementsofthisluminos-
cluster stars, namely, planetary nebulae detected from the
itymustalsocontendwiththeproblemspresentedbyscat-
groundandredgiantsdetectedusingtheHubbleSpaceTele-
teredlightfromnearbybrightobjectsandthecontribution
scope(HST),havebeendiscoveredintheVirgoandFornax
ofdiscretesources.
Clusters (Arnaboldi et al. 1996; Theuns & Warren 1997;
Despite these difficulties, intracluster light (ICL) is of
Me´ndez et al. 1997; Ciardullo et al. 1998; Feldmeier, Ciar-
potentially great interest to studies of galaxy and galaxy
dullo, & Jacoby 1998; Ferguson, Tanvir, & von Hippel
cluster evolution. The dynamical evolution of cluster gal-
1998; Feldmeier 2000; Durrell et al. 2002). Although some
axies is complex, involving poorly understood processes
oftheintraclusterplanetarycandidateswerelaterfoundto
such as galactic encounters, dark matter, cluster accretion,
be background objects (Kudritzki et al. 2000; Ciardullo et
andtidalstripping(seeDressler1984).TheICLprovidesa
al.2002),mostarebonafideintraclusterplanetarynebulae
direct way to study these different mechanisms. Various
(Freeman et al. 2000; Ciardullo et al. 2002). There is also
studieshavesuggestedthatanywherebetween10%and70%
some evidence for intracluster supernovae, although the
ofacluster’stotalluminositymaybecontainedintheICL
resultsherearemoretentative,andtheratemaybesignifi-
(Richstone & Malumuth 1983; Miller 1983), with a strong
cantly smaller than that seen in galaxies (Smith 1981; Gal-
dependenceonthedynamicalstateofthecluster.Theprop-
Yam&Maoz2000;N.D.Tysonetal.2002,inpreparation).
ertiesoftheICLmayalsobesensitivetothedistributionof
Theseindividualintraclusterstarsgivethepromiseofstudy-
ingindetailthekinematics,metallicity,andageoftheintra-
clusterstellarpopulationinnearbygalaxyclusters.
1VisitingAstronomer,KittPeakNationalObservatory,NationalOpti- Anotherareaofprogressistheadventofmodernnumeri-
calAstronomyObservatory,whichisoperatedbytheAssociationofUni- calstudiesofthedynamicalevolutionofgalaxiesinclusters.
versities for Research in Astronomy (AURA), Inc., under cooperative
High-resolutionN-bodysimulationsnowhavetheabilityto
agreementwiththeNationalScienceFoundation.
followhundredsofclustergalaxiesinteractingwithinacos-
2CottrellScholarofResearchCorporationandNSFCAREERfellow;
alsoDepartmentofPhysics,CaseWesternReserveUniversity. mological context (Moore et al. 1996; Dubinski 1998).
779
780 FELDMEIER ET AL. Vol. 575
Thesehigh-qualitysimulationsfinallyallowtestablepredic- CCDimaginghasshownatleastonecasewhereacDgalaxy
tions of the production and properties of intracluster star- mayinfactbewellcharacterizedbyapurer1/4law(Gonza-
light(Moore,Lake,&Katz1998;J.Dubinski,C.Murali,& lezetal.2000).Inlightofthisresult,revisitingthequestion
R. Ouyed, 2000, unpublished preprint; Korchagin, Tsu- of the structure of cD galaxies using deep CCD imaging
chiya,&Miyama2001).Whencombinedwithearliertheo- may shed new light on the origin of the cD envelopes and
retical studies (Gallagher & Ostriker 1972; Merritt 1983; theevolutionofgalaxyclusters.
Richstone & Malumuth 1983; Miller 1983; Merritt 1984) Toaddressthesequestions,wehavebegundeepimaging
there is now a growing theoretical framework in which to of a sample of galaxy clusters to quantify the structure of
interpretobservationsoftheICL. ICL. Although quantitative surface photometry several
Recently,anotheraspectofintraclusterstarlighthasbeen magnitudesbelowskyisanextremelychallengingtask,over
discovered:tidaldebrisarcs.Thesefeaturesarelarge((cid:3)100 thepastdecadethenecessaryCCDimagingtechniqueshave
kpc),lowsurfacebrightness(l (cid:3)26magarcsec(cid:1)2)arclike beendevelopedandcarriedoutonbothgalaxies(e.g.,Mor-
v
structures found in nearby galaxy clusters and are not due rison,Boroson,&Harding1994;Morrisonetal.1997;Fry
togravitationallensing. These arcs have been found in the et al. 1999; Zheng et al. 1999) and galaxy clusters (e.g.,
Coma and Centaurus Clusters (Trentham & Mobasher Tyson, Kochanski, & dell’Antonio 1998; Gonzalez et al.
1998;Gregg&West1998;Calcan´eo-Rolda´netal.2000).It 2000). Using these techniques, our plan is to image galaxy
has been proposed that these arcs are due to tidal interac- clustersthatdifferinrichness,concentration,andsubstruc-
tionsbetweengalaxiesandthecluster’sgravitationalpoten- turetoquantifyhowtheICLchangesasafunctionofenvi-
tial (Moore et al. 1996). Since several of these debris arcs ronment. In conjunction with increasingly sophisticated
were found by chance, it is plausible that they might be models of cluster galaxy evolution, such observations can
presentinothergalaxy clusters. The scientific potentialfor provideconstraintsontheevolutionofclustersandcluster
thesearcsisexciting.Byobservingthemorphology,and— galaxies, the formation of the ICL, and the distribution of
inthefuture—kinematicsofthesestellarstreams,muchcan darkmatterinclustergalaxies.Inthispaper,wedescribein
belearnedabouttheorbitsoftheinfallinggalaxiesandthe detailourimagingtechniquesandshowresultsfromthefirst
gravitationalpotentialofthegalaxycluster(i.e.,Calcan´eo- twoclusterssurveyed.
Rolda´netal.2000).
Although the presence of intracluster stars has been
2. SELECTION CRITERIA
clearlydemonstrated,thereislittleinformationonhowthe
amountanddistributionofintraclusterstarlightvarieswith OurprogramisaimedatstudyingtheICLinclusterspos-
thepropertiesoftheclusteritinhabits.Wedonotyethavea sessing a variety of structural properties in order to probe
globalpictureoftheprevalenceoftheICLingalaxyclusters the relationship between the ICL and cluster environment.
oroftheinformationitcontainsaboutthedynamicalstate OurinitialsamplewillprimarilyfocusonAbell(Abell,Cor-
of clusters, both of which are crucial in developing more win, & Olowin 1989) clusters of distance class 5–6
advancedmodelsofclusterevolution.Studiesofindividual (z(cid:2)0:1 0:175) with differing richness and Bautz-Morgan
intracluster stars are invaluable in nearby clusters but are classifications.Thelowerendoftheredshiftrangeischosen
flux-limitedandsocannotprobetheevolutionofintraclus- suchthattheinner(cid:3)0.75Mpcoftheclusterfitsonthefield
ter light to higher redshift. Finally, there is little data on of view of moderate-size CCD detectors, allowing us to
how common tidal debris arcs might be in galaxy clusters. studytheclusterasawholewithoutmosaicingandpermit-
Currently,themajorityoftidaldebrisarcsdiscoveredhave tingareasonableamountofskyattheouteredgeofthefield
been found in the Coma Cluster (Trentham & Mobasher for sky subtraction. The upper limit is set so that ð1þzÞ4
1998; Gregg & West 1998; Calcan´eo-Rolda´n et al. 2000). surface brightness dimming is not prohibitive and also to
The Coma Cluster is well known to be unusually rich (see prevent the angular size of any tidal arcs from being too
Dressler 1984), and it is possible that it might contain an small. For comparison purposes, we also observe nearby
unusuallylargenumberoftidaldebrisarcs. poor galaxy clusters from the MKW/AWM catalogs of
Finally,anotherinterestingfacetoftheICLinclustersis poor clusters that appear to contain cD galaxies (Morgan,
the nature of cD envelopes. First identified in deep photo- Kayser, & White 1975; Albert, White, & Morgan 1977).
graphic imaging of clusters, cD galaxies are characterized These clusters will provide a significantly different cluster
by an excess of diffuse light (compared to an r1/4 law) at environmentintermsofdensityanddynamicalinteraction.
large radius. The origin of cD envelopes remains unclear: Scattered starlight is a crucial source of systematic error
are they formed in the initial stages of cluster collapse, or in our program. Therefore, we must screen our candidate
later, asgalaxies continue tofallintheclusterandbecome clusters carefully, making sure there are no bright stars in
tidallystripped?ThedetailedlightdistributionofcDenve- the CCD field or up to several degrees nearby. Because of
lopesmayholdcluestotheanswer.Thestatisticalmechan- the complex spatial nature of the scattered light distribu-
icsofviolentrelaxationnaturallyproducesr1/4-likeprofiles tion,wedoourscreeningbymanualinspectionoftheorigi-
(Lynden-Bell1966);ifcDenvelopesformduringclustercol- nal Palomar Observatory Sky Survey (POSS) plates in the
lapse, they should show such a profile. On the other hand, area around the target cluster. Approximately half of our
continuedaccretionandstrippingofinfallinggalaxiesneed candidateclustersarerejectedbythisprocess.Oncetheclus-
not produce an r1/4 profile, as the distribution of stripped terpassesboththecatalogcriteriaandscatteredlighttests,
lightwillbemoresensitivetotheorbitalenergyandangular itisincludedasapotentialtarget.
momentumoftheinfallinggalaxies.Whilethecharacteriza-
tionofcDenvelopesasanexcessoflightoveranr1/4profile
3. CLUSTER PROPERTIES
wouldseemtoargueforastrippingorigin,moststudiesof
cD galaxies have used older photographic data, with very For the first targets of our survey, we chose two galaxy
uncertain flat-fielding characteristics. Newer work using clusters with greatly differing properties. A1413
No. 2, 2002 DEEP CCD SURFACE PHOTOMETRY OF GALAXY CLUSTERS. I. 781
((cid:1)¼11h55m2295, (cid:2)¼þ23(cid:4)2201800 [J2000.0]) is a galaxy
cluster of richness class 3 (richer than 95% of the original
Abell catalog), with a Bautz-Morgan type of I (cD domi-
nated;Leir&vandenBergh1977)andaRood-Sastrytype
ofcD(Struble&Rood1987).ItscentralcDgalaxyhasbeen
studiedwithphotographicsurfacephotometryatlargeradii
andCCDsurfacephotometryatsmallerradii(Oemler1976;
Schombert 1986, 1988; Schneider, Gunn, & Hoessel 1983;
Porter,Schneider,&Hoessel1991)numeroustimes,allow-
ingustocompareourresultswithothers.Theseearlierstud-
ies imply that the properties of A1413’s cD halo are
extremelyimpressive:Oemler(1976)foundthatthecDhalo
ofA1413mightextendasfaras240(2.4h(cid:1)1Mpc,H ¼100
0
h km s(cid:1)1 Mpc(cid:1)1) away from the center of the cluster, and
Morgan & Lesh (1965) indicated that A1413 might be the
largest ofallcDgalaxies. Schombert (1988) studied A1413
indetailandfoundthatthecDhaloextendedto(cid:3)500kpc
h(cid:1)1,withalargeexcessoverthebest-fittingdeVaucouleurs Fig.1.—SpectrumofthenightskyatKittPeakNationalObservatory
r1/4profileoftheinnerregions.A1413hasarelativelyhigh takenfromthedataofMassey&Foltz(2000).Overlaidoverthespectrum
X-raytemperature(8keV;White2000)andaSunyaev-Zel- is the filter transmission curve of the Washington M filter used in these
dovichdecrement(Graingeetal.1996),confirmingthatthe observations(KP1581),shownasthesolidline.Forcomparison,astand-
ardHarrisVfilter(KP1542)isalsoshownasthedashedline.TheWashing-
clusterisindeedmassive.Therefore,A1413isarepresenta-
tonMfiltercontainsfewerstrongsky-emissionlinescomparedtotheV
tive example of a rich cluster, albeit containing an abnor- filter,mostnotablyOi(cid:3)5577.
mallylargecDgalaxy.
MKW7(WBL514[Whiteetal.1999];(cid:1)¼14h34m0099,
(cid:2)¼þ03(cid:4)4605200 [J2000.0]) is a poor galaxy cluster whose tributed with elliptical symmetry about the center of the
brightestclustergalaxywasfoundtobecD-likeinappear- clusterpotential.’’
ance from inspection of the POSS (Morgan, Kayser, & Basedonsurfacephotometryaloneitisdifficulttodisen-
White1975;seeTonry1987andSchombert1992fordiscus- tangle luminosity associated with a cD envelope from that
sionsofcDclassification).Ithasarichnessclassof(cid:1)1(Bah- oftheextendedICL,andinfactsuchadistinctionmaynot
call1980) andingalaxycountsisoverafactorof8poorer be well motivated from a physical point of view. For the
than A1413. Photographic surface photometry has been purposes of our work, we will simply use the term ‘‘intra-
madeofMKW7’sbrightestclustergalaxy(Morbey&Mor- cluster light’’ to refer to the diffuse light in galaxy clusters
ris 1983), but these measurements are complicated by the andaddressissuesrelatedtocDenvelopes,diffusearcs,and
presence of a m (cid:2)11:9 star within 2100 of the galaxy extendedICLinthecontextofdiffuselightasawhole.
v
nucleus.VandenBergh(1977)notesthatthebrightestclus-
ter galaxy is embedded in a bright but asymmetrical enve-
lope. From galaxy density maps and redshift information,
5. OBSERVATIONS AND REDUCTIONS
Beers et al. (1995) argue that MKW 7 is gravitationally
boundtoanotherpoorcluster,MKW8,whichiswithin1.5 5.1. Observations
h(cid:1)1 Mpc. MKW 7 has been detected in X-rays multiple
ThedataforA1413andMKW7wereobtainedovertwo
times(e.g.,Priceetal.1991),butnogastemperaturehasyet
photometric nights during a four-night run in 2000 April,
been determined. In comparison to A1413, MKW 7 is a
using the 2.1 m telescope at Kitt Peak National Observa-
poorer,lessdynamicallyevolvedcluster.
tory.3Theimagesweretakenusinga2048(cid:5)2048Tektronix
CCD(T2KA).Withthissetup,thefieldofviewis10.4arc-
4. A NOTE ON NOMENCLATURE min2,witheachpixelimaging0>305ofsky.Thegainwasset
atthedefaultvalueof3.6e(cid:1)ADU(cid:1)1,andthereadoutnoise
We note that the term ‘‘intracluster starlight’’ has been
was 4 e(cid:1) (1.1 ADU). All exposures were made through a
appliedinmanydifferentwaysintheliterature.Adynami-
Washington M filter, which is similar to Johnson V but is
cal definition might be stars that are unbound from any
slightly bluer in mean wavelength and therefore contains
individualclustergalaxyyetboundtotheclusterasawhole.
fewerairglowemissionlines(seeFig.1).Theseairglowlines,
Fromanobservationalpointofview,ofcourse,thisdefini-
produced in the upper atmosphere by a variety of mecha-
tionisinaccessiblewithoutknowingthedetailedkinematics
nisms (Roach & Gordon 1973), are a significant source of
of the ICL and the total mass distribution in the cluster.
sky background and are well known to be variable over
Withdeepimaging,thedefinitionofICLcanonlybemade
timescalesofminutes(e.g.,Pilachowskietal.1989;Krisciu-
basedonthesurfacebrightnessdistributionwithintheclus-
nas1997;Morrisonetal.1997).Therefore,toreducethesky
ter.Indeed,itisdebatablewhethercDenvelopesshouldbe
background and to simplify the process of sky subtraction
consideredasICL—istheenvelopeafeatureofthecDgal-
andflat-fielding, wechosetheWashingtonMfilter forour
axyitself,orareboththecDandtheenvelopesimplymate-
observations. We transformed these observations to John-
rialthathascollectedatthebottomoftheclusterpotential
well? Uson etal. (1991) succinctly summarize the situation
in their observations of Abell 2029: ‘‘Whether this diffuse
3KittPeakNationalObservatoryisadivisionofNOAO,whichisoper-
lightiscalledthecDenvelopeordiffuseintergalacticlightis
atedbyAURA,undercooperativeagreementwiththeNationalScience
amatterofsemantics;itisadiffusecomponentwhichisdis- Foundation.
782 FELDMEIER ET AL. Vol. 575
sonV(x5.4),andunlessotherwisestated,allsurfacebright-
nessesaregiveninVmagnitudes.
We began the telescope run by preparing the detector,
telescope,andfiltertoreducetheamountofscatteredlight,
whichsetstheultimatelimittooursurfacephotometry.We
first placed a black cardboard mask over the detector’s
Dewar window in order to reduce scattered light from the
mounting hardware surrounding the CCD. These metallic
components are highly reflective, and a clear difference in
the amount of scattered light is readily apparent. We next
took pinhole images of the telescope pupil to search for
othersourcesofscatteredlight(Grundahl&Sørensen1996)
andbaffledanysuchareaswithblackcloth.
Anaccurateflat-fieldiscriticaltothesuccessofourpro-
gram.Asmentionedinx1,weareinterestedinrecoveringa
signal that is less than 1% of the sky background. Our flat
fieldmustbeatleastafewtimesmoreaccuratethanthis1%
Fig.3.—Differencebetweensurfacebrightnessofdatacorrectedforthe
valueforourdatatobemeaningful.Forthisreason,dome
nonlinearityeffectfoundbyMochesjkaetal.(2001)fortheT2KAdetector
flatfieldscannotbeusedbecauseofpossiblescatteredlight,
anduncorrecteddataovertherangeofsurfacebrightnessesapplicableto
differingpupililluminations,andintrinsiccolordifferences. oursurvey.Notethattheeffectisnowheregreaterthan0.012magandthat
For similar reasons, twilight flats are also inadequate for theeffectisrelativelyconstantovertherangeofsurfacebrightnessesthat
ourpurposes.Therefore,dark-skyflatsareanecessity,and ICLwouldbepresent(lV ¼24 30).
weperformedtheobservations inthemanner describedby
Morrisonetal.(1997).Halfofthetimewasusedobserving
large, a clear nonlinearity is present in the data. We fitted
the galaxy clusters, and the other half was used to obtain
theresidualsinmagnitudewithaleast-squareslinearmodel
dark-skyflats.Thedark-skyflatimagesweretakenatprede- and found a residual slope of 0:0095(cid:6)0:0008 mag mag(cid:1)1.
terminedareasawayfrombrightstarsatapproximatelythe
This is consistent with the measurements of Mochejska et
samehourangleanddeclinationastheclusterimages.Over
al.(2001)fortheirtestphotometryofNGC7789(seetheir
thecourseoftheobservingrun,atotalofnineimageswere
Fig.3).Sincethenonlinearbehaviorisidenticaltothatseen
taken of A1413, 12 images were taken of MKW 7, and 23
in the data presented in Mochejska et al. (2001), the data
dark-skyflatswereobtained.Foreachoftheclusterandsky
wastakenwiththesameinstrumentonly5–6monthsapart,
images,theexposuretimewas900s.
andtheMochesjskaetal.databetterconstraintheeffect,we
adoptidenticalcorrectionsfornonlinearity:
5.2. CorrectionforNonlinearity " #
I (cid:1) I (cid:2)2
After our run, we were made aware of the presence of I ¼I c þc i þc i ; ð1Þ
e i 1 2 3
32767 32767
nonlinearity in the T2KA detector by K. Stanek (reported
inMochejskaetal.2001).Figure2showsthecomparisonof
where I is the measured intensity and I is the corrected
stellar photometry derived from a 60 s test exposure of i e
intensity in ADU. The constants derived by Mochejska et
MKW 7 and a median-combined exposure of 900 s, both
al.(2001)foragainof3.6e(cid:1)ADU(cid:1)1are
reduced in the standard manner. Although the scatter is
c ¼0:983282; c ¼(cid:1)0:0765595; c ¼0:0252555:
1 2 3
ð2Þ
FollowingMochejskaetal.(2001),weusedtheirlincortask
withinIRAF4tocorrectthedatafornonlinearityafterover-
scanremovalandbiassubtraction.
Fortworeasonsthenonlinearityislessofaproblemfor
ourprojectthanitmightfirstseem.First,sinceourskyflats
haveexactlythesameexposuretimeasourdata,theyhave
the identical nonlinearity inherent inthe exposures, so any
difficulty in flat-fielding is canceled out to first order. Sec-
ond, at low surface brightness levels, the error caused by
anynonlinearityissignificantbutrelativelysmall.Figure3
shows the difference in magnitudes between the corrected
andnoncorrectedsky-subtractedsurfacebrightness.None-
theless, any error in the correction for nonlinearity adds a
sourceoferrortooursurfacebrightnessestimates.Todeter-
Fig.2.—ComparisonofT2KAphotometryobtainedfroma60stest
exposure anda median-combined900s exposure of MKW7.Although
thereislargescatter(primarilyfromphotonnoiseinthe60sexposure), 4IRAFisdistributedbytheNationalOpticalAstronomyObservatory,
thereisaclearsystematicresidualwithinstrumentalmagnitudeidenticalto which are operated by the Association of Universities for Research in
thatseenbyMochesjkaetal.(2001).Thelinedenotestheleast-squareslin- Astronomy,Inc.,undercooperativeagreementwiththeNationalScience
earfitthroughthedata. Foundation.
No. 2, 2002 DEEP CCD SURFACE PHOTOMETRY OF GALAXY CLUSTERS. I. 783
mine the amount of such error, we obtained the linearity orderpolynomialfit.Apreliminaryskyvaluewasfoundfor
datafromMochejskaetal.(2001)(kindlyprovidedtousby each cluster image by finding the mode of two regions on
B.Mochejska)andreplicatedthelinearityfit.Wefoundthe eachchipwellawayfromthecenteroftheclusterandaver-
identical constants with the following 1 (cid:4) errors on the aging the results. This sky value was then subtracted from
parameters: each image. The median sky value was 886.0 ADU pixel(cid:1)1
forA1413and932.7ADUpixel(cid:1)1forMKW7.Afterapply-
(cid:4)c1 ¼0:0052; (cid:4)c2 ¼0:012; (cid:4)c3 ¼0:0057: ð3Þ ingthephotometriczeropointin(x5.4)below,thesecorre-
spond to l ¼21:11 and l ¼21:05 mag arcsec(cid:1)2,
Thisuncertaintyisaddedtoourfinalerrormodel(seex6.4). v v
respectively. Since the source of sky brightness is mostly
withintheEarth’satmosphere,weremoveour0.17magair
5.3. Flat-Fielding
mass(cid:1)1 extinction correction and find that the average
Afteroverscanremovalandbiassubtraction(doneinthe brightness of the night sky at zenith was approximately
standardmanner),weconstructeda‘‘master’’skyflatfrom l ¼21:25magarcsec(cid:1)2,inreasonableagreementwiththe
v
the dark-sky images taken. First, each individual sky flat solar maximum value of 21:287(cid:6)0:048 of Krisciunas
wasvisuallyinspectedtoensurethatnobrightstarsorscat- (1997).
teredlightpatternswerepresentintheimage.Thisisneces- Withtheoverscan,biassubtraction,flat-fielding,andsky
sary becausein some of our exposuresof candidate galaxy subtraction complete, we then combined the images
clusters at the telescope, we found that a gridlike scattered together, using a 2 (cid:4) clipped median as before and scaling
light pattern appeared. This pattern was rotated 45(cid:4) from for air mass. The final images for A1413 and MKW 7 are
the CCD axes and typically covered an area of 370(cid:5)350 displayedinFigures5and6.Themeasuredseeing(FWHM)
pixels. When the pattern did appear, its surface brightness forthefinalcombinedimageswas1>22forA1413and1>37
varied, but it could be as large as l (cid:2)25:2 mag arcsec(cid:1)2. forMKW7.
v
Threeofthedark-skyimageswerefoundtohaveunaccept-
ably large scattered light patterns and were removed from 5.4. PhotometricZeroPoint
the list, leaving 20 dark-sky images to construct the dark-
TheLandoltstarfieldsSA98,SA107,andSA110(Land-
skyflat.Threeotherdark-skyimagesalsocontainthescat-
olt 1992) were observed, giving us a total of 37 well-
teredlightpattern,buttheamplitudeofthepatternwasso
observedstandardstarsoverarangeofcolorandairmass.
smallinthesecases(theirpresencewasbarelyvisibleonthe
For the purposes of our analysis, we converted our Wash-
image)thattheywereleftinthesample.Noscatteredlight
ington M exposures to V-band magnitudes. This transfor-
patternswereseeninanyimageofA1413andMKW7.
mationisstraightforwardbecausealloftheWashingtonM
To construct the best possible master sky flat, we com-
standard stars used in these observations are also Landolt
binedtheindividualskyframesusingaveryaccuratedeter-
(1992) V standards. A photometric zero point of
mination of their modes. We do this using the iterative
V ¼21:09(cid:6)0:04 mag arcsec(cid:1)2 [corresponding to 1 ADU
proceduredescribedinMorrisonetal.(1997).Wefirstpre- s(cid:1)1 pixel(cid:1)1 and assuming a (B(cid:1)V) color of 1.0] was deter-
scaledtheimagesbytheirmode,foundfromtheimstattask
mined. For a 900 s exposure this yields V ¼28:48 mag
withinIRAF.Prescalingisimportantbecauseevenwithour
arcsec(cid:1)2,correspondingto1ADUpixel(cid:1)1atunitairmass.
relativelynarrowMfilter,themodalvalueoftheskyimages
Asourexposuresweretakenonlyinonefilterandwedonot
variesfrommaximumtominimumbyupto22%.Thenwe
knowtheexactcoloroftheintraclusterlight,wecannotadd
combinedtheindividual,modal-dividedskyimagestomake
acolorcorrectiontermtoourtargetphotometry,butfrom
apreliminaryflat-fieldframe,usingIRAF’simcombinetask,
thestandardstarobservationsweestimateitsmagnitudeas
withtheCCDCLIPalgorithm,settoremovepixelsthatdif-
less than 0.1 mag over the entire likely color range of our
feredfromthemedianbymorethan2(cid:4).
target objects [0:8(cid:7)ðB(cid:1)VÞ(cid:7)1:3]. The color term is rea-
Each of the individual sky frames were then divided by
sonably well fitted as a linear function of (B(cid:1)V), with a
this preliminary flat-field frame to reduce the width of the slopeof0.2magmag(cid:1)1of(B(cid:1)V)color.
distribution of modal sky values, making rejection of out-
liersduetofaintstarsandstellarwingsmoreaccurate.The
flat-fieldedskyframeswerethenaveragedinto50(cid:5)50pixel 6. ANALYSIS AND RESULTS
bins, and a plane was then fitted to the binned-up images
WeadoptanapproximateangularsizedistancetoA1413
usingtheIRAFtaskimsurfit.Thisstepisnecessarybecause
and MKW 7 of 465 and 111 Mpc, respectively, assuming
each individual sky frame has noticeable sky variations
redshifts of z¼0:1427 for A1413 (Struble & Rood 1999)
across the image due to a number of atmospheric effects,
and z¼0:0290 for MKW 7 (Beers et al. 1995), a Hubble
suchasairglow(Roach&Gordon1973;Wild1997;Zheng
constant of H ¼75 km s(cid:1)1 Mpc(cid:1)1, and a cosmology of
et al. 1999). Figure 4 shows the binned-up images for each 0
(cid:1) ¼0:3,(cid:1) ¼0:7.Atthesesmallredshifts,thesedistances
skyframeaftertheyhavebeenflat-fielded.Clearsystematic m (cid:2)
depend little on (cid:1). Given these assumed distances, 100 sub-
skyvariationscanbeeasilyseeninthedata.Aftertheindi-
tends (cid:2)2.3 kpc in A1413 and (cid:2)0.52 kpc in MKW 7. The
vidual planes were fitted and normalized, the sky frames
correspondingluminositydistancemoduli,ignoringanyK-
were each divided by their normalized plane. The modes
corrections,are39.0forA1413and35.4forMKW7.
were then recalculated using our own software, and the
entireprocedurewasrepeatedusingtheimprovedflat-field
6.1. Masking
frame. The procedure was repeated until the calculated
modeshadconverged(about15timesinthiscase). In order to reach the faintest possible surface brightness
The galaxy cluster images were then flat-fielded by this levelsofthecDgalaxy+intraclusterlight,wemustmaskout
final flat and registered using stars common to all frames all other sources—both stars and galaxies—in the frame.
and the IRAF tasks geomap and geotrans, using a second- Webeginbycreatingabinarymaskimageinwhich1indi-
784 FELDMEIER ET AL. Vol. 575
Fig.4.—Imagesofthe20blankskyframesusedinconstructingthemasterskyflat,afterbeingflat-fieldedandaveragedinto50(cid:5)50pixelbins.Thesmall-
scalevariationsareduetobrightstarsandgalaxiesineachindividualframe,butthelarge-scalevariationsareduetochangesintheskyillumination.These
large-scalefeaturesarepartiallyremovedbytheplanenormalization.Thegridlikescatteredlightpatterncanalsobebarelyseeninthreeoftheframes(bottom
row,firstandfourthcolumnandsecondrowfrombottom,fifthcolumn).Thesegridlikepatternsarealmostinvisibleintheunbinnedimages.
cates a good pixel and zero indicates a bad pixel. This has Nextwemustmaskoutallofthegalaxiesineachcluster,
theadvantageofallowingustovisuallycompareourmask excluding the central cD. Unresolved background galaxies
image at any point in the construction process by simply have been treated as point sources and have already been
multiplyingthemaskbythedataimageanddisplayingthe maskedbytheDAOPHOTprocedureabove,butmanyre-
results. solved sources remain in both clusters. We chose to mask
We first begin by masking out the stars in each image. outthegalaxiesusingthesegmentationimagefromtheSEx-
Sinceweareconcernedwithverylowsurfacebrightness,we tractor software package (V2.2.1; Bertin & Arnouts 1996).
mustdeterminethepoint-spreadfunction(PSF)outtovery Again,sincewemustmaskdowntoverylowsurfacebright-
large radii. Using the DAOPHOT (Stetson 1987) package, ness levels, the SExtractor detection parameters are set for
wedetectedallofthestarsintheframedowntoasignal-to- faint surface brightness levels. After experimentation, we
noise ratio of 3 and used a subset of bright stars to first adopted a minimum detection threshold of 4 pixels that
determinethePSFouttoaradiusof20pixels.Wethenused were 0.6 (cid:4) above the local sky background. This corre-
thispreliminaryPSFtomaskoutallofthestarsandsmall sponds to 3.3 ADU in A1413 and 2.7 ADU in MKW 7.
galaxies around two bright saturated stars in our A1413 AssumingGaussianstatistics,theprobabilityofafalseSEx-
data.Saturatedstarshavemuchhighersignal-to-noiseratio tractordetectionattheselow-lightlevelsis5:6(cid:5)10(cid:1)3 per4
in the wings on the PSF, which are our primary concern. pixel block. This is uncomfortably high and allows for the
Othersources,suchasresolvedgalaxiesandstellardiffrac- possibilityof‘‘overmasking’’ourdata,thatis,maskingout
tion spikes, were removed manually. Then the unmasked noise spikes instead of real objects. This would alter the
pixelsfromthetwosaturatedstarswereaveraged inradial noisepropertiesofourdataandleadtosystematicerrorsin
annuli and joined to the preliminary PSF (which measures oursurfacephotometry.However,weneedthelowthresh-
theinnercoreofthestarmoreaccurately).Thefinalradial old to ensure that the low surface brightness outer regions
profileisdisplayedinFigure7.Usingthislarge-radiusPSF oflarge,luminousgalaxiesarebeingproperlymasked.
and the list of stars found by DAOPHOT, we masked all WedealwiththeovermaskingproblembyrunningSEx-
stellarsourcesintheframeouttoaradiuswherethemagni- tractor without deblending the various detections, that is,
tude-scaledPSFwas1ADUabovetheskyvalue. not assigning faint isolated objects as part of a much
No. 2, 2002 DEEP CCD SURFACE PHOTOMETRY OF GALAXY CLUSTERS. I. 785
Fig.5.—Ourfinal,median-combinedimageforA1413.Northisattheleftofthisimage,andeastisatthebottom.Theimageis10.2arcmin2,corresponding
toalineardistanceof1.5MpcatouradopteddistancetoA1413.Notetheincrediblerichnessofthiscluster:almosteverybrightobjectintheframewiththe
exceptionofthetwosaturatedstarsatthebottomisagalaxy.
brighter object. Then we removed all sources whose total are in good agreement with one another. We set the cutoff
magnitude was fainter than a cutoff value. We found the magnitude to 23.8 for A1413 and 23.0 for MKW 7 and
cutoffvalueintwodifferentways.First,wecreatedtheraw removedallsourcesfromthesegmentationimagethatwere
galaxybrightnessdistributionforbothclustersbyselecting belowthisvalue.
allobjectswithastellarityindexlessthan0.5,wherethestel- Thereisoneotherchangethatwemustmaketotheseg-
larity index defines the likelihood that a source is or is not mentationimage.Afterinspectionofthecorrectedsegmen-
extendedthroughmeasurementsofimagemomentsbySEx- tation mask multiplied by the data, we occasionally found
tractor’sneuralnetwork(Bertin&Arnouts1996).Wethen smallgroupsofpixelsthatwerecompletelysurroundedbya
noted where the raw galaxy brightness distribution slope large number of masked pixels. These ‘‘islands’’ of
rapidly increased. This will indicate the onset of the noise unmasked pixels are due to SExtractor treating this small
spikes. Second, we ran SExtractor on the mathematical area as a separate object within the larger source. The
inverseofeachclusterimage: islands were corrected in the segmentation image by an
automated process. We masked each individual pixel that
I ðx;yÞ¼(cid:1)I ðx;yÞ; ð4Þ was surrounded by N already masked pixels, and we
inverse image
repeated this process M times. By experimentation, we
where x and y are the pixel coordinates of the image and found that N ¼6 and M ¼20 filled in the majority of the
Iðx;yÞisthefluxinADUateachpoint.Wethenfoundthe island-like structures with minimal changes to any other
brightness distribution of negative noise spikes, which region.Finally,theimagesmultipliedbythemaskwerevis-
should provide an accurate measure of the cutoff value, ually inspected, and any regions that needed any further
assuming that the noise is symmetrically distributed. The masking were masked using imedit. These regions were
results of both of these tests are displayed in Figure 8 and mostlylarge-scaleislandsofunmaskedpixelsthatwerenot
786 FELDMEIER ET AL. Vol. 575
Fig.6.—Ourfinal,median-combinedimageforMKW7.Northisattheleftofthisimage,andeastisatthebottom.Theimageis10.2arcmin2,correspond-
ingtoalineardistanceof330kpcatouradopteddistancetoMKW7.Thebright,saturatedstarsuperposednearMKW7snucleuscannotbeseeninthis
image’sgrayscale.
removed by our automated procedure. Less than 2% and
0.5%ofthepixelsinA1413andMKW7,respectively,were
removed manually. The fraction of the images that was
masked at this point is 43.4% for A1413 and 52.36% for
MKW7.Figure9showsasubregionoftheMKW7image
that contains stars and galaxies through each step of the
maskingprocess.
6.2. FinalSkySubtraction,Masking,andLarge-Scale
Flat-FieldingErrors
Accurateskysubtractioniscrucialtodeterminethetrue
amountofintraclusterstarlightineachclusterandisoneof
thedominantsourcesoferrorinouranalysis.Wenowfind
a more accurate sky level for each cluster by using the
maskedimage.Wefirstbinuptheentireimageintosquares
of49(cid:5)49pixels.Foreachbin,wecalculatearobustaver-
Fig. 7.—Surface brightness profile of a saturated star on our A1413 age (Morrison et al. 1994), ignoring all masked pixels; the
image,averagedazimuthally.Ascanbeclearlyseen,theprofileextendsto
resultsaredisplayedinFigures10and11,respectively.Itis
verylargeradii.Thesolidhorizontallineissetatthesurfacebrightness
limitof1ADUpixel(cid:1)1abovetheskyvalue,correspondingtoasurface importanttonotethattheentiregray-scalerangedisplayed
brightnessof28.48magarcsec(cid:1)2. inFigures 10and 11 is(cid:6)5 ADU from the sky level, which
No. 2, 2002 DEEP CCD SURFACE PHOTOMETRY OF GALAXY CLUSTERS. I. 787
Fig.8.—BrightnessdistributionofallobjectsintheA1413(left)andMKW7(right)fieldsfoundbySExtractorthathaveastellarityindexlessthan0.5,
shownasthefilleddots.Notethatthisisnottheclusterluminosityfunction,asnobackgroundsubtractionhasbeendone,andtheblendingparameterofSEx-
tractorhasbeenturnedoff.Thebrightnessdistributioniscomparedtothebrightnessdistributionofnegativenoisespikes,shownastheopendiamonds.As
canbeclearlyseen,thebrightnessdistributionsteepensatthesamemagnitude(m(cid:2)23)thatthenegativenoisespikesbecomepresentingreatnumbers.This
denoteswheretheovermaskingproblembegins.
corresponds to a surface brightness of l ¼26:7 mag masked,exceptforafewpixelsthataresignificantlybelow
v
arcsec(cid:1)2,or5.5magbelowtheskylevel. the median value of the total number of pixels in the bin.
Several distinct features are apparent in these binned Thisistheoriginofthelowerflux‘‘ring’’seenaroundeach
images. First, the bright central portion of the cD is com- cD galaxy and is simply an artifact of masking the cD. As
pletelymasked.Then,asthedistancefromthecDincreases, we move even farther outward, we find a region where the
there is an annulus of bins where almost all pixels are cDhaloisstilldetectable,butithasdroppedbelowthesur-
Fig.9.—TypicalregionofMKW7shownthroughallstagesofthemaskingprocess.Fromlefttorightandtoptobottom,thesubimagesare(1)theoriginal
image,(2)theimagemultipliedbythestellarmaskfoundthroughDAOFIND,(3)thepreviousimagemultipliedbythemaskfromSExtractor,(4)theprevious
mask,withthe‘‘overmasking’’correctionapplied,(5)thepreviousimage,withthesurroundingpixelscorrectionmade,and(6)thepreviousimageafterman-
ualmasking.Thegrayscaleinalloftheimagesis10ADUaboveandbelowtheskyvalue,andthestandarddeviationoftheskybackgroundis5.8ADU
pixel(cid:1)1.Themaskingprocedureremovesthevastmajorityofstellarandgalaxylightintheimage.
788 FELDMEIER ET AL. Vol. 575
Fig.10.—Binned-upimageofA1413,withagray-scalestretchoffiveADUaboveandbelowtheskyvalue.Northisagaintotheleft,andeastisatthebot-
tomofthisimage.AsthebinsincreaseinradiusfromthecentralcDgalaxy,theirfluxessystematicallychange,asdiscussedinthetext.Atthefarrightofthis
image,alarge-scaleflat-fieldingerrorof1ADUisclearlyvisible.
facebrightnessatwhichSExtractormasksindividualpixels ess will remove any ICL that covers the entire image. We
(l (cid:3)27:3 mag arcsec(cid:1)2). Finally, at the edges, the flux thencreatedahistogramofskyvaluesin49(cid:5)49pixelbins
v
comestoamoreorlessconstantvalue. wellawayfromthecenterofeachcluster.Wealsorequired
However,inthecaseofA1413,thereisanadditionallow that the bins contain at least 200 unmasked pixels to be
surfacebrightnessfeaturestretchingalongtherightsideof includedinthehistogram.Thereare653suchbinsinA1413
the frame at an amplitude of (cid:3)1 ADU ((cid:3)l ¼28:5 mag and746binsinMKW7.Thesehistogramsaredisplayedin
v
arcsec(cid:1)2). This feature is almost certainly instrumental in Figure12.Thewidthofthehistogramsprovidesameasure
nature, corresponding to the vignetting of the southern of our uncertainties due to large-scale flat-fielding errors
regionoftheT2KAchipbythe2.1mguidecamera(Massey and the faint outer wings of stars and galaxies that remain
etal.2000)5.Thisflat-fieldingresidualissmall,butsinceitis unmasked, even after the involved procedure above. We
systematic in nature, we chose to mask all pixels in the findthatthelarge-scaleflat-fieldingerrorforbothimageis
A1413 image with x>1300. For MKW 7, we see no evi- conservatively 1 ADU,6 which corresponds to an uncer-
dence for this effect, and so we do not mask further inthis taintyof0.11%.
case.
To better determine our sky values and to measure our
6.3. ConstructingtheSurfaceBrightnessProfile
large-scale flat-fielding errors, we fit and subtract a plane
fromeachmasked,binnedclusterimage,usingtheimsurfit WenowunmasktheregionaroundthecentralcDgalaxy
task in IRAF. We took care to use regions on each image and rebuild the mask, leaving the cD+intracluster light
that are well away from the central cD. The mean correc- intact.Weproceedasfollows:wefirstusetheellipsetaskin
tions from this step are small: less than 0.5 ADU for both IRAF/STSDAS (Busko 1996), based on the algorithms of
clustersonaverage.However,weemphasizethatthisproc- Jedrzejewski (1987), to obtain an approximate geometrical
5This document is available at http://www.noao.edu/kpno/
kpno.html. 6Equivalentto1.3(cid:4)ifthedistributionisGaussian.