Table Of Content

The GlobularCluster System ofNGC1399: III.VLTSpectroscopy andDatabase10 1 1 1 2 3 7 6 B.Dirsch ,T.Richtler ,D.Geisler ,K.Gebhardt ,M.Hilker ,M.V.Alonso ,J.C.Forte ,E.K. 8 4 5 4 5 Grebel ,L.Infante ,S.Larsen ,D.Minniti ,M.Rejkuba [email protected] 4 ABSTRACT 0 0 2 Radialvelocitiesof468globularclustersaroundNGC1399,thecentralgalaxyintheFornax n cluster, havebeenobtainedwithFORS2andtheMaskExchangeUnit(MXU)attheESOVery a J Large Telescope. This is the largest sample of globular cluster velocities around any galaxy 2 obtained so far. The mean velocity uncertainty is 50km/sec. This data sample is accurate and 1 large enough to be used in studies of the mass distribution of NGC1399 and the properties of 1 its globular cluster system. Here we describe the observations, the reduction procedure, and v discuss the uncertainties of the resulting velocities. The complete sample of cluster velocities 8 1 which is used in adynamical study of NGC 1399 is tabulated. Asubsample is compared with 2 previously published values. 1 0 4 Subjectheadings: galaxies: individual(NGC1399)—galaxies: ellipticalandlenticular, cD— 0 / galaxies: star clusters —galaxies: kinematics anddynamics —galaxies: halos —cosmology: h p DarkMatter - o r t s a 1. Introduction : v i In the context of understanding the distribution of dark matter in galaxies, early-type galaxies in the X centers of galaxy clusters play an important role. In these places, the highest dark matter densities are r a 1UniversidaddeConcepcioń,DepartamentodeF´ısica,Casilla160-C,Concepcioń,Chile 2DepartmentofAstronomy,UniversityofTexasatAustin,TX78712,Austin,USA 3SternwartederUniversita¨tBonn,AufdemHu¨gel71,53121Bonn,Germany 4DepartamentodeAstronom´ıayAstrof´ısica,P.UniversidadCato´lica,VicunãMackenna4860,Santiago22,Chile 5EuropeanSouthernObservatory,Karl-Schwarzschild-Str.2,D-85748Garching,Germany 6UniversidadNacionaldeLaPlata,FacultadCienciasAstronomicasyGeofisicas,PaseodelBosqueS/N,1900-LaPlata,UNLP andCONICET,Argentina 7ObservatorioAstrono´micodeCo´rdoba,Laprida854,Co´rdoba,5000,CONICET,Argentina CNRSUMR5572,ObservatoireMidi-Pyreńeés,14AvenueE.Belin,31400Toulouse,France 8Max-PlanckInstitutfu¨rAstronomie,Ko¨nigstuhl17,D-69117,Heidelberg,Germany 10BasedonobservationscollectedattheEuropeanSouthernObservatory,CerroParanal,Chile;ESOprogram66.B-0393. –2– found and accordingly, the transition between luminous and dark matter dominated regions can be best investigated. Intheinnerregionsofelliptical galaxies thiscanbedonebylongslitspectroscopy measuring integratedspectrallineprofiles(seeKronawitteretal. ((2000))andGerhardetal. ((2001))foralargesample ofgalaxies). In the outer region of an elliptical galaxy – in an area where dark matter begins to dominate – this approachishardlyfeasibleduetothelowsurfacebrightness. Globularclusters(GCs)andplanetarynebulae (PNe), which can betraced tolarge radii, arethe best tools tostudy the dynamics atlarger radii. However, the distribution of radial velocities of particles is degenerate with respect to the galaxy mass profile and the orbital structure of the dynamical probes (e.g. Merritt & Tremblay (1993)). If no other constraints on the velocity dispersion tensor and massdistribution are available, then radial velocities for several hundred or more clusters are required to break this degeneracy (e.g., Merritt (1993)). Therefore, if we want to lift this degeneracy, we are limited to those nearby large elliptical galaxies with sufficiently populous cluster systems. M87 is such a candidate in the northern hemisphere. Cohen & Ryzhov (1997) observed about 200 radial velocities, using the Keck telescope. This sample has been improved and enlarged to about 350 velocities (Coˆte´ etal. (2001),Hanesetal. (2001)). AnotherpromisingtargetisthecentralgalaxyoftheFornaxcluster,NGC1399. It’spopulousglobular cluster system (GCS)has been investigated by Hanes&Harris ((1986)), Bridges et al. ((1991)), Wagner et al. ((1991)), Ostrovetal. ((1993)), Kissler-Patig etal. ((1997)),Grillmairetal. ((1999))and mostrecently by Dirsch et al. ((2003), PaperI) the latter covering a field of 36′ ×36′. The GCS has been the target of three spectroscopic studies. Grillmair et al. ((1994)) obtained low-dispersion spectra of 47 GCs within a ′ radialdistanceof9 aroundNGC1399withtheLowDispersionSurveySpectrographontheAAT.Atypical uncertainty of their velocities is 150km/sec. Another contribution comes from Minniti etal. ((1998))who usedtheESOMulti-ModeInstrumentattheESO3.5mNewTechnologyTelescopetomeasurevelocitiesof 19GCs. Theirvelocityuncertainties arealsolarge: themeanuncertainty is128km/sec. Thevelocitieswith the smallest errors available so far are obtained by Kissler-Patig et al. ((1998)) with the Low Resolution Imaging Spectrograph at the Keck telescope. They determined the velocities of 21 GCs with an average uncertainty of 45km/sec. Kissler-Patig et al. ((1999)) compiled the velocity determinations from these threestudies, ending upwith74GCvelocities. Beside theGCs, planetary nebulae (PNe)have been used fordynamical studies of NGC1399 byArn- aboldietal. ((1994))andNapolitanoetal. ((2002)). Bothstudiesusedthesamedatabaseof37PNvelocities intheinner 4′ ofNGC1399. Atsmaller radial distances (≤ 97′′)thegalaxy light itself hasbeen employed toconstraindynamicalmodelsbySagliaetal. ((2000)),Kronawitteretal. ((2000)),Gerhardetal. ((2001)). In order to greatly improve our knowledge of its GCSkinematics and galaxy halo dynamics, weused ′ theFORS2attheVLTtoobtainspectraofglobularclustersouttoaradialdistanceof8 whichcorresponds toapproximately 45kpc. Adistance modulus ofm−M = 31.4 (19Mpc)isused (see Dirsch etal. (2003) anddiscussion therein). The observations aim at a significant improvement in the number and quality of dynamical probes. Theselected cluster candidates haveluminosities brighter thanV = 22.5mag. Anidealtelescope toobtain –3– the required spectra of such faint objects in the southern hemisphere is the ESO Very Large Telescope. A spectroscopic resolution of approximately 3A˚ is desirable for our task and thus FORS2 with the MXU (Mask eXchange Unit) is an appropriate instrument since it allows the simultaneous observation of some 100objects. In this article we present the data analysis and error discussion of our measurements. The complete dataset of 502 cluster velocities is presented which is the basis of the dynamical analysis published in PaperII. These papers are part of a bigger effort to study the dynamics and stellar content of NGC 1399, whichalsoincludes newphotometric widefieldobservations published inPaperI. 2. Spectroscopicmaskpreparation The preparation of the slit masks we used for observing NGC1399 at the VLT is an important step in the overall investigation. Hence, we describe this first step in detail, before discussing the observations themselves. Ourcandidates GCshavebeenselectedonthebasisofnewwide-fieldphotometry inWashing- tonCandKron-Cousins Rofthecluster system ofNGC1399, obtained attheCTIO4mMOSAICsystem. Foradescription ofthephotometric observations thereaderisreferredtoPaperI. Forthepointsourceselectionweusedacolorcriterion(0.9 < (C−T1) < 2.4)andalsoamorpholog- icalcriterion basedontheSExtractorclass(Bertin&Arnout(1996);onlyobjectswithclasslargerthan0.8 areretained, basedonartificialstarexperimentsandcomparison withtheVLTpreimages. Thismorpholog- icalselectionlimitedoursampletoobjectsbrighterthanR=23.3magwhichallowsustoobtaintherequired signal-to-noise inthespectra. For accurate slit placement on the MXU masks, preimaging is required. Service mode observations withFORS2attheVLTinOctober2000provideduswithfourFORS2fields,eachofthem7′×7′,arranged inaquadratic configuration (seeFig.1). Eachfieldhasbeenexposed for30secinJohnson V.Weidentified the selected clusters in these images and placed the slits accordingly. Since the VLT images had superior ′′ seeing(around0.6),comparedtoourMOSAICobservations,weusedthemtocheckthestar-likeappearance oftheselected objects. Thisleftuswithabout500globular clustercandidates. Amajorpointduringthepreparation ofthemasks,whichwasdonewiththeESOFORSInstrumental Mask Simulator (FIMS 1), were the choices of the slit width and particularly the slit lengths. We decided ′′ foraslitwidthof1 ,whichturnedouttobeappropriate forthenormallysub-arcsecond seeingconditions. The slit length was a more complicated issue. The spectra of our fainter targets are sky-dominated and the sky subtraction is best done by measuring object and sky in the same slit. However, the severe ′′ disadvantage with slits of the appropriate length, 5 or longer, is that they would cover up many objects whichotherwise wouldhavebeen observable withsmallerlength, contrary toourobjective ofobserving as manytargetsaspossible. Moreover,itconsiderablylowerstheflexibilityofchoosingslitpositionsaccording 1http://http.hq.eso.org/observing/p2pp/OSS/FIMS/FIMS-tool.html –4– totherandomlocationoftargetswithahighsurfacenumberdensityinamask. Therefore,wedecidedtoset skyslitsindependently fromtheobjectslitsandtosubtracttheskyafterthewavelengthcalibration, trusting initsaccuracy. For most slits, the size then was 1′′ × 2′′, being 1′′ × 5′′ only for the few bright objects (normally foreground stars), whichwereneeded toaccurately position themask. Inthismanner, wecould fillamask with up to 120 slits, typically 40 cluster candidates, 40 sky slits, and many miscellaneous objects, among themotherpointsourcesnotmatchingourstrictselectioncriteria, orbackground galaxies. The color-magnitude diagram of the final sample for which we obtained velocities is shown in Fig.7 together with its color distribution that is compared to the color distribution of the total MOSAIC sample within the same radial range. This figure illustrates that the final sample is not subject to color selection effects. 3. Theobservations Theobservationshavebeenperformedduring30.11.–2.12.2000attheEuropeanSouthernObservatory atCerroParanalwiththeVeryLargeTelescopefacility. ThetelescopewasUT2(Kueyen)andtheinstrument thefocalreducerFORS2,equippedwiththeMXU.ThedetectorwasaSiTESI-424AbacksidethinnedCCD withapixelsizeof24×24microns. Thetotalfieldofviewis6.′83×6.′83. Theobservationsaresummarized in Table1. The four VLT fields in which the masks were placed are shown overlayed on a DSS image of NGC1399inFig.1. Weusedthe600Bgrismwithoutfilter,whichprovidesaresolution ofaround2.5A˚.Thespectralrange covered wasabout 2000 A˚.However, duetothe position ofagiven slitinthe mask, thelimiting short- and long wavelengths varied considerably, ranging from 3500 A˚ to6500 A˚.In mostcases, the grism efficiency degraded the signal-to-noise short-wards of 3800 A˚ drastically, so this region could not be used. Mask 89 has been observed using the 300V grism with a lower spectral resolution (110A˚/mm). This has been done to see the effect of using a certain grism. The analysis showed no difference in the quality of the derived velocities forthetwogrisms. Weexposed 13masks(exposure timeswereeither 45minor2×45min). Theseeing wasalwayssub- ′′ ′′ arcsecond andranged from0.6to0.9. Flat-fielding wasdonewithstandard lampflats. Forthewavelength calibration, aHeHgCdlampwasexposedduringservice day-timecalibration. Thetotalnumberofspectraobtainedis1462. Thissampleiscomposedof531skyspectra(seeTable2), 512 spectra of cluster candidates, 190 spectra of point sources of unknown nature at the time of mask preparation (stars, unresolved galaxies or globular clusters), 176 galaxies, and 53 “bright” objects, mainly stars which were needed to adjust the mask. Since some objects (about 80) have been observed in two different masks in order to assess the velocity uncertainties, the total number of objects is smaller by this numberthanthenumberofspectra. Therearetworeasonswhywefoundclustersinthe(randomlyselected) pointsourcesample: firstly,theMOSAICdataisnotditheredandthusforaconsiderablefractionofclusters –5– nophotometry wasobtained. Secondly, ouremployed morphological selection criteria wereverystrict and thecompleteness offaintclusteridentification wasratherlow(seealsoPaperI). Table 1 lists the relevant data of the observations, starting with the number of the respective FORS field,themasknumber,thecentercoordinates(J2000),theexposuretime(90minmeansthattwoexposures of45minhavebeenstacked). 4. Thedatareduction The basic reduction (bias subtraction, flatfielding, trimming) has been done with standard procedures withinIRAF.Forthelaterextraction ofthespectratheapextract package hasbeenused. Regardingtheremovalofcosmicrays,wefoundaftersomeexperimentingthatthetaskFILTER/COSMICs under MIDAS gave the most satisfactory results. A few artefacts remained in spite of that, best visible in thefaintspectra. In the flat-fielded image the spectra have a separation of only a few pixels and are curved along the dispersion axis which has to be fitted before they can be extracted. The curvature is strongest at the frame edges, where the deviation from the center to the edge isapproximately 6pixels. Wetraced each spectrum along the dispersion axis on the flatfield image of the mask because of their much clearer signal. We then kept the tracing parameters and optimized the size of the apertures with the goal to minimize the sky contributionintheobjectspectrawhichhasbeendoneonthesciencespectra. Thefinalsizesoftheapertures ′′ ′′ dependontheseeingandvariesbetween1 and1.4. Onsomemasksafewspectraoverlapwhichhavebeen excluded fromfurtheranalysis. Weemployed theIRAFtask identify tocalibrate both object andskyspectra. Typicallyaround 18He, HgandCd-lineswerekeptfortheline-list. Thecalibration uncertainty is±0.04A˚. The bright OI skyline at 5567A˚ which is present in all spectra can be used to correct zero point differences betweenthemasks. Differences upto±0.8A˚ havebeenfound. Thereasonforthesesystematic differences isthe uncertainty inthe maskplacement which iscited intheMXUmanual 2 tobe13microns. This is approximately half a pixel or 0.6A˚ using the 600B grism and thus in good agreement with our measuredzeropointdifferences. Within a mask the position of the OI sky line was constant to within 0.04A˚ consistent with the calibration uncertainty. An exception is mask #89 (with the grism 300V), where we found a systematic linear increase/decrease of the OI line position from 5566.3A˚ to 5567.91A˚. We corrected for this with a linear interpolation inthewavelengthcalibration. Thesystematicbehavior inmask#89isprobably duetoaslight rotation ofthemask. For the sky subtraction we used spectra that were observed through slits placed in empty sky areas. 2http://www.eso.org/instruments/fors/userman/ –6– Foreach object adjacent sky spectra within a certain radial distance from NGC1399 have been selected to attempt to obtain the best sky spectrum. The radial width of these annuli had been chosen to be smaller ′ ′ ′ near to NGC1399 and wider further away (1.2 for distances smaller than 2.5 and 1.7 for larger distances ). Typically 3 to 11 sky spectra were averaged and subtracted from the object spectra. The quality of the subtraction wasjudgedbyexaminingtheresiduals aroundthebrightskylinesinthered. This procedure resulted in good sky subtracted spectra as long as wavelengths longer than 4000A˚ are considered. For shorter wavelengths the uncertainties in sensitivity (“flat-field uncertainties”) between sky andobjectspectramaketheskysubtraction difficult. However,duetothelowsignal-to-noise ofthespectra blue-wards of 4000A˚ this uncertainty is anyway not important. Some example spectra around T1=21 are showninFig.2 5. Velocities 5.1. Velocitydetermination Wedetermined velocitieswithtwodifferenttechniques: firstbyusingthecorrelation withanobjectof knownvelocityandsecondbymeasuringtheredshift ofanobjectdirectlywithidentified absorption lines. The line measurements were performed with the help of the rvidlines task within IRAF. Typically, around15featureshavebeenfittedforthevelocitydeterminations. Forthecorrelation technique weusedthe taskfxcor implemented inIRAF(thetechnique isdescribed byTonry&Davis(1979)). WedidnotutilizeanyFourierfilteringbutwerathersmoothedverynoisyspectra withamedianfilter(3pixels). Thecontinuumhasbeensubtractedbyfxcorwithaspline-fittothespectraand a2−σclippingalgorithmaroundthefitline. Therangethathasbeenproventobebestsuitedforthevelocity determinationis4500A˚ to5500A˚.However,forfainterobjectsweadjustedtherangeindividuallytofindthe mostsignificant correlation peak. Astemplate, weusedahighS/N-spectrum (S/Nabout 40)ofNGC1396, asmallgalaxywithlowintrinsicvelocitydispersionandaspectrumsimilartothatofacluster. Thisobjectis onmask#82, whichaccordingly served asthereference mask. Zero-point differences withtheother masks had been accounted for by using the position of the sky OI-line. We determined NGC1396’s velocity by measuring lines and obtained 815±8km/sec asthe heliocentric velocity. Thisvalue isconsiderably lower than most older values from the literature: daCosta et al. ((1998)) found 856±37km/sec and in the RC3 catalogue 894±29km/sec is given (de Vaucouleurs et al. (1991)). However, Drinkwater et al. ((2001b)) quote 808±8km/s. Themeanvelocity ofallclusters is1438±15 km/s. AsforNGC1399 itself, onefinds 9 measurements of its radial velocity with quoted uncertainties consulting the NASA/IPAC Extragalactic Database. After skipping two of them (one has a discrepant value and the other a large uncertainty of 200 km/s), the weighted mean value is 1442±9 km/s. Our mean radial velocity of the entire cluster sample is 1438±15km/s,soweareconfident ofourabsolute velocitycalibration. –7– 5.2. Velocityuncertainties The differences between cluster velocities that were derived with individual line and with correlation measurements (shown in Fig. 3) can be used to study the velocity uncertainties. A dependence of the differences on the brightness of the objects is expected and can be observed. However, deviations as large as 500km/sec are more than 3-σ larger than what is expected from the measurement uncertainties. This indicates that at least for some measurements systematic errors dominate over statistical errors. We will showlaterthaterrorsinthelinemeasurements ofnoisyspectraareresponsible forthisbehavior. Themean valueofthedifferencebetweenlineandcorrelationmeasurementsis−6.2±0.3km/sec. Toderivethemean weexcludedclusters withabsolute differences largerthan500km/sec. Inthiscasemedianandmeanagree. Thequestion remainswhetherthisisasignificant deviation. Theabsolute scaleofthecorrelation velocities are based on the heliocentric velocity of NGC1396, for which we give an error of 8km/sec. Taking this into account the deviation from zero of the differences is −6± 8km/sec and hence well inside the given uncertainty. For mask#80 the spectra extraction and subsequent velocity determination has been done indepen- dently by two of us (TR and BD) and the differences for the line measurements are plotted in Fig. 4. The standard deviation between the two measurements is σ = 32km/sec while we expect 37km/sec from the linemeasurement errors. Thescatter reflectstheuncertainty resulting from independent treatment (tracing, aperture definition, wavelengthcalibration, lineidentification) ofthedata. Thebestestimationoftheuncertainties isacomparisonofmeasurementsthatareobtainedondifferent masks, which has been done for 31 point sources (29 clusters and two stars). The differences between the velocities are plotted in Fig.5, in dependence on brightness and color. The scatter of the velocities measured viacorrelation is57km/secandinverygood agreement withtheexpectation from theindividual uncertainties (52km/sec), whichshowsthatthegivenuncertainties arecorrect. Thescatter ofthevelocities based on direct line measurements is 79km/sec and much larger than expected from the individual errors (40km/sec). Thisindicates thattheerrorsofthelinemeasurements aredominated bysystematic errorsand are not of a statistical nature. The reason for the systematic errors is most probably line misidentification in noisy spectra. Line misidentification in noisy spectra also accounts for small velocity uncertainties at relativelyfaintclusters. Summarizing, we suggest to use the correlation velocities for any purpose, however, we provide the line measurements for a consistency check. For this reason we do not attempt to derive a more refined uncertainty estimationforthelinevelocities. 5.3. Thefinalsample Thewhole dataset istabulated inTable3. Thefirstcolumn identifies thecluster: weused anidentifier consisting of two 2-digit numbers, the first indicating the mask and the second the aperture number on the mask which was assigned during the extraction process. Column two and three are right ascension and –8– declination of the clusters (J2000). The positions are based on the USNO2.0 catalogue 3 (Urban et al. (1998)). The color and magnitude information in columns 4 and 5 are taken from Dirsch et al. ((2003)). The uncertainties do not include the photometric calibration errors that are 0.03 in C-T1 and 0.02 in T1. The sixth column gives the velocity determined with the correlation measurement and the seventh column the velocity determined using direct line measurements. The uncertainties given are those returned by the respective packages used. The last column is reserved for comments. An identifier for a different cluster indicatesthatthisclusterhasbeenobservedindependentlyontwomasksandcanbeidentifiedviaitsnumber ontheothermask. Thespatialdistribution ofthewholeclustersampletabulatedinTab.3isshowninFig.6. Colorandmagnitudeinformationaremissingforsomeclusters. Thereasonistheincompletecoverage of the field due to the undithered gaps in the MOSAIC image or that they are located within saturated regions caused by bright nearby objects. These candidates had been selected ”by eye” after all object slits on the mask had been set as described above. The velocity given is the mean value of the two correlation measurements and the error asimple mean error. Wequote the mean velocity only for the firstappearance oftheclusterinthelist. Thecolor and magnitude dependence of thecorrelation velocities are shown inFig.8. Thedynamical interpretation isgiveninPaperII. Starsin our sample for which velocities have been determined are tabulated inTable4. Insome cases no correlation velocities are given, only line velocities. The reason is that our template is not particularly suitedtobecorrelated withlate-typestars,whichalsoexplainsthesystematically largeruncertainties inthe correlation velocities. 6. Comparisonwiththeliterature Globularclusters aroundNGC1399havebeenspectroscopically observed earlier. Thelargest sample, containing 47 globular cluster velocities, has been obtained by Grillmair et al. ((1994)) using the Anglo- AustralianTelescopewiththeLow-DispersionSurveySpectrographwith≈ 13A˚ resolution, whichresulted inavelocityuncertaintyofapproximately150km/sec. Bettervelocitieswithuncertainties ofapproximately 100km/sechavebeenmeasured for18clusters byMinnitietal. ((1998))withtheESOMulti-Mode Instru- ment at the ESO New Technology Telescope with a resolution of 7.5A˚. The observations done by Kissler- Patigetal. ((1998))withtheLowResolutionImagingSpectrographatKeck(resolutionof5.6A˚)yieldedthe bestvelocities sofarfor, 18globulars withuncertainties around 35km/sec. Someofthese globular clusters wereobservedinourrunaswellandinTab.5ourvelocities arecompared withthevelocities determined in the earlier studies. In addition we compiled stars that were observed in our run and are also present in the TwoDegreeField(2dF)Survey(Drinkwateretal. (2001a)). The mean difference and standard deviation between our velocities and those of Kissler-Patig et al. 3http://tdc-www.harvard.edu/software/catalogs/ua2.html –9– ((1998))(wealwayssubtract thereference velocity from ours)is+101km/secand58km/sec, respectively. Fromthepublished uncertainties wewouldexpectastandard deviation ofapproximately 60km/sec, which agreeswell. FromtheGrillmairetal. ((1994))datawefindforthemeandifferenceandstandarddeviation- 293km/secand153km/sec,respectively. Thislargedifferenceisdrivenmainlybyafewextremelydeviating objects,probablycausedbythelowS/NofGrillmairetal.’sspectra. However,alsointhiscasetheobserved and expected standard deviations are in good agreement. Regarding the stars in common with the 2dF survey, wefindadifference of-56km/secwithastandard deviation of83km/sec, whilewewouldexpect a standard deviation of61km/sec. Thereason fortheslightdiscrepancy probably isthatweusedlineinstead ofcorrelationvelocitiesforthestarsandwehaveshownearlierthattheerrorsforthevelocitiesobtainedvia linemeasurementsisunderestimated. Thesemeasurementsarehenceingoodagreementwithourvelocities. Acknowledgments BD, TR, DG, LI and DM gratefully acknowledge support from the Chilean Center for Astrophysics FONDAP No. 15010003. BD gratefully acknowledges financial support of the Alexander-von-Humboldt FoundationviaaFeodorLynenStipendium. Wethankouranonymousrefereeforher/hishelpfulcomments. REFERENCES ArnaboldiM.,FreemanK.C.,HuiX.etal.,1994,ESOMessenger 76,44 Bertin,E.,Arnouts, S.,1996,A&AS117,393 Bridges,T.J.,Hanes,D.A.,Harris,W.E.,1991,AJ101,469 CohenJ.G.,RyzhovA.,1997,ApJ486,230 Coˆte´ P.,McLaughlinD.E.,HanesD.A.etal.,2001,ApJ559,828 daCostaL.N.,WillmerC.N.A.,PellegriniP.S.etal.,1998, AJ116,1 de Vaucouleurs G., de Vaucouleurs A. Corwin Jr., H.G. et al, 1991, Third Reference Catalogue of Bright Galaxies,Springer-Verlag DirschB.,RichtlerT.,GeislerD.etal.,2003,AJ125,1908,PaperI Drinkwater M., EngelC.,Phillipps S.etal., 2001 Anglo-Australian Observatory -Newsletter (ISSN0728- 5833),No.97 DrinkwaterM.J.,GreggM.D.,HolmanB.A.,BrownM.J.I.,2001,MNRAS326,1076 DrinkwaterM.J.,JonesJ.B.,GreggM.D.,Phillipps S.,2000, PASA17,227 GerhardO.,KronawitterA.,SagliaR.P.,BenderR.,2001,AJ121,1936 –10– GrillmairC.J.,FreemanK.C.,BicknellG.V.etal.,1994,AJ423,L9 Grillmair,C.J.,Forbes,D.A.,Brodie,J.P.,Elson,R.A.W.,1999, AJ117,167 Hanes,D.A.,Harris,W.E.,1986, ApJ309,564 HanesD.A.,Coˆte´ P.,BridgesT.J.etal.,2001ApJ559,812 Kissler-Patig, M.,Kohle,S.,Hilker,M.etal.,1997,A&A319,470 Kissler-PatigM.,BrodieJ.P.,SchroderL.L.etal.,1998,AJ115,105 Kissler-PatigM.,GrillmairC.J.,MeylanG.etal.,1999,AJ117,1206 KronawitterA.,SagliaR.P.,GerhardO.,BenderR.,2000,A&AS144,53 MerrittD.,1993,ApJ413,79 MerrittD.,TremblayB.1993,AJ106,2229 MinnitiD.,Kissler-Patig M.,GoudfrooijP.,MeylanG.,1998,AJ115,121 Napolitano N.R.,ArnaboldiM.,Capaccioli M.,2002,A&A383,791 Ostrov,P.G.,Geisler,D.,Forte,J.C.,1993,AJ105,1762 RichtlerT,DirschB.,GebhardtK.etal.,2003,submitted toAJ,PaperII SagliaR.P.,KronawitterA.,GerhardO.,BenderR.,2000,AJ119,153 TonryJ.,DavisM.,1979,AJ84,1511 UrbanS.E.,CorbinT.E.,WycoffG.L.etal.,1998, AJ115,1212 Wagner,S.,Richtler, T.,Hopp,U.,1991,A&A241,399 ThispreprintwaspreparedwiththeAASLATEXmacrosv5.2.