Towards an Understanding of the Over-abundance around the Central -Giant Elliptical NGC 1399 I

bIarkus Kissier-Patig '92 Uco/Lick observatory, University of California, Santa Cruz, CA 9,5064. USA Electronic mails: mkissler8ucblick.org

Carl J. Grillmajr SIRTF Science Center, Mail Stop 100-22, California Instibute of Technology, Pasadena, CA 9112'5, USA Electronicmail: [email protected]

Georges Meylan European Southern Observatory, Iiarl-Schwarzschild-Strasse 2, 0-85748 Garching bei i\.lunchen, Germany Electronic mail: [email protected]

Jean P. Brodie UCO/Lick observatory, University of California, Santa Cru:, CA 95064, USA Electronic mail: [email protected]

Dante Minniti Lawrence Livermore National Laboratory, Livermore, CA 94550, USA Departamento de Astronomia y Astrofisica, P. Universidad Catdlica, Casilla 104, Santiago 22, Chile Electronic mail: [email protected]

Paul Goudfrooij Space Telescope Science Institute, 9700 San ,\[artin Drive, Baltimore, MD 21218. lJSA Electronic mail: [email protected]

ABSTRACT We investigate the kinematics of a combined sample of 74 globular clusters around NGCl 1399. Their high , increasing with radius, supports their associ- ation with the gravitational potentialof the ga.laxy cluster rather than with thatof NGC

'Feodor Lynerl Fellow of the Alexander von Hunlholdt Foundation 2Current address: ESO, Karl-Schwanschild-Str. 2, 85748 Garching, Germany. Email: rnkissler'Qeso.org 3Afiliated with the Astrophysics Division, Space Science Department, European Space Agency, ESTEC, Postbus 299, NL-2200 AG Noordwijk, The Netherlands

1 1399 itself. We find no evidence for rotatiod'fnihe fdIsample, although some indication for rotation in the outer regions. The ,data*donoballow us to detect differences between the kinematics of the blue and red sub-populations of globular clusters. A comparison between the globular cluster systems of NGC 1399 and those of NGC 1404 and NGC 1380 indicates that the globular cluster's in all three are likely to have formed via similar mechanisms and at-similar epochs. The only property which distinguishes the NGC 1399 globular cluster system from these others is that it is ten times more abundant. We summarize the evidence for associating these excess globulars with the cluster rather than with NGC 1399 itself, and suggest that the over-abundance can be explainedby tidal strippink of neighboring galaxies and subsequent accumulation of globulars in the gravitational pdtential of the . 'L" Subject headings: globularclusters: general , galaxies:elliptical and lenticular, cD, galaxies: halos , galaxies: kinematicsand dynamics, galaxies: formation, galaxies: evo- lu t lu ion I

I J'

I I 1. Introduction pectetl to be 011 orbit,sradial innerin the regions and to ~110~solne systemic rotation far out. The kinemat- Extragalactic: globular clusters have in recent years ics of globular clusters that mighthave formed during establishecl tllernseives as potential tracers of the for- d plerger Was not aciclressed in their study; it \vould mation and evolution of galaxies (see Ashmarl & Zepf depend 011 the kinelnatics of the in-falling gas from 1998 for a recent review). Thenumber of detailed dli(:tl theyformed. Other simulations (e.g. bIuzzio photometric studies has rapidly increaed, and these 1987) studied the accretion and stripping of globular studies reveal a number of interesting connections be- clusters in galaxyclusters, but noclear predictions tween globular cluster systems and their host galax- for the kinematics of accreted globular clusters were ies. With the recent commissioning of l0m-class tele- formulated. scopes, measuring absorption line indices of a large Tlie primary goal of the present paper is to corn- numberof individual globular clusters has become bine all existing kinematic data on the globular ~1~s- feasible (Kissler-Patig et al. 1998; Cohen, Blakeslee, Fer system of NGC 1399 to better constrain itsorigin. & Ryzhov 1998). Colors, magnitudes, total numbers, A secondary aim is to use the photometric properties spatial distributions, radial density profiles, ages, and of globular clusters in the brightest galaxies , can now be obtained and used to con- .to further constrain formation scenarios. The sample strain the formation history of globular clusters and , of radial velocities for globular clusters around NGC their host galaxies. , 1399 is compiled and presented in Sect. 2. These are Anotheressential source of information fordis- ,;used to investigatethe kinematic properties of the criminating between different formation scenarios is full sample, as well as sub-samplesselected on the the kinematics of globular cluster systems, as deter- basis of radius and color in Sect. 3. In Section 4 we mined from the measured radial velocities of individ- compare the kinematics of the globular clusters with ual clusters. For example,the studies of M87 and those of the , cluster galaxies and X-ray gas, and NGC' 1399 found a significantly higher velocity dis- derive the mass-to-lightratio in theouter environs persion for theglobular clusters than for thestars of the galaxy. In Sect. 5 we compare the properties (Huchra and Brodie 1987; Mould et al. 1990; Brodie of the globular clusters in NGC 1399 with those of andHuchra 1991; Grillmair et al. 1994;Cohen & the globular cluster systems of the next two brightest Ryzhov1997; Minniti et al. 1998;Kissler-Patig et early-type galaxies in Fornax, NGC 1380 and KGC al. 1998). Mould etal. (1990) demonstrated that, 1404. We then discuss the implications for different in M87, this was consistent with the the surface den- formation scenarios. A summary and our conclusions sity distribution of globular clusters being more ex- are given in Sect. 6. tended than the surface density distribution of stars. Grillmairet al. (1994)suggested thatthe globular 2. The data clusters around NGC 1399 were reacting to the grav- itational potential of the as a whole Our sample is based on the compilations of Grill- ratherthan just that of thehost galaxy. In NGC nlair (1992), Grillmairet al. (1994), Minnitiet al. 5128,Harris et al. (1988)and Huiet al. (1995) re- (1998),and IGissler-Patig et al. (1998). Briefly, the ported rotation in the globular cluster system, though data from Grilllnair et al. were obtained at the Anglo- only for the metal-rich clusters. This is contrary to Australian Telescope with the Low-Dispersion Survey the findings in NGC 4472 (Sharples et al. 1998) and Spectrographand the Image PhotonCounting Sys- M87 (Iiissler-Patig & Gebhardt1998) in which the tem in the wavelength range 3800-4800 A with 2: 13 metal-poorglobular clusters seem to dominate the resolution. Minniti et al. obtained their data with the rotation. New Technology Telescope at the European Southern Unfortunat,ely, few models exist to compare with Observatory, using the ESO Multi-i\.lode Instrument the (still sparse) data. Predictions for the kinematic in the wavelength range 6000-9000 .-\ with a resolu- signature in globularcluster systems after spiral- tion of 7.q X. Iiissler-Patig et al. observed with the spiral mergershave been presented by Hernquist & LOW Reso1ution Imaging Spectrograph on the Keck 1 Bolte(1992). They studied the kinematics of glob- Telescope; their spectra covered a wavelength range rlIar clusters already present in the progenitorsand from about 4000 .-I to 6100 A with 5.6 A resolution. follllcl that, in the merger product these clusters areex- We refer the reader to the original papers for a more I I: debailed description of the observations, the data re- duction, and the velocity measurements. Hereafter, we will refer to the respective sanlples as the AAT sample(Grillmair 1092, C;rillniair et 3.2. Spatial distribution of the velocities al. 1994), the NTT sample (5Iinniti et al. 1991)), and t,he Keck sample(Iiissler-Patig et aI. 1098).The In Figure 3 we plot the cluster velocities against combinedsample of 74 globularclusters is listed projected radius from the center of NGC 1399. The in Table 1. For each globularcluster we give the mean velocity does not change significantly with ra- ID number(taken from the papers with theprefix dius (see also Table 3). However, we note that within aat/ntt/keck added respectively), the equatorial coor- 4’ the velocities seem to cluster around the mean ve- dinates (B1950), the heliocentric (with locity of the stellar component of NGC 1399, whereas the weighted mean when multiple measurementswere beyond 5’ the globular clusters seem to have either available), as well as the availablephotometric infor- . significantly higher or lower velocities than the mean. mation. For 52 globularclusters V-I colorsaccu- This is better illustrated in Fig. 4, where we have plot- rate to 0.035 mag where obtained from the work of ted histograms of the velocities in concentric annuli. Kissler-Patig et al. (1997a). The B, magnitude ,and Using the ROSTAT package (Beers et al. 1990, Bird & Bj-R color were taken from Grillmair (1992). We in- Beers 1993), we tested the samples for normality and cluded in our combined sample all objects with radial unimodality. While the globular clusters in the inner velocities greater than 500 km s-l and less than 2500 ring are consistent at the 98% confidence level with km s-l (i.e. within 3a of the mean). A list of posi- a normal distribution, the statistics are inconclusive tions, velocities and colors of the 74 globular clusters in the middle ring and inconsistent with normality in in electronic form is available from the first author. the outer ring at the 90% confidence level. However, Figure 1 shows thepositions with respect to NGC the number statistics are rather small, and the dis- 1399 of all globularclusters observed; the symbols tribution in the outermost annulus is only consistent are proportional to the difference between the globu- with multi-modality at the 80% confidence level. lar cluster velocity and the systemic velocity of NGC It seems reasonable to question whether beyond 6’ 1399. The two pairs of larger circles (continuous and one is starting to sample globular clusters belonging dashed lines) show 1 and 5 reff for the stellar light dis- to NGC 1404, which lies only 9’ in projection from tribution of NGC 1399 (centered) and NGC 1404 (in NGC 1399. The answercomes from Fig. 1: there is the SE), respectively. In Table 2 we list the adopted no obvious sign of any concentration of high veloc- properties of the three brightest early-type galaxies ity clusters in the direction of NGC 1404. The second in the center of Fornax: NGC 1399, KGC 1404, and velocity peak is therefore probably not due to contam- KGC 1380. ination of the sample by globular clusters belonging toNGC 1404. The recently derived density profile 3. Kinematic properties of the sample for the globular clusters around NGC 1404 (Forbes et 3.1. Velocity distribution of the full sample al. 1998) indeed shows that contaminationofour sam- 1: ple by NGC 1404 clusters is likely to be small; half- Figure 2 shows a histogram of theglobular cluster way between NGC 1399 and NGC 1404, the surface velocities. As already noted by Minniti et al. (1998), density of globular cluster belonging to NGC 1399 is the velocity distribution exhibits two. peaks roughly = 40 times higher than that of NGC 1404. Even at 5 centered on thesystemic velocity of NGC1399, al- r&(NGC1404) (22.5’ ) from NGc 1404 the contribu- though a single Gaussian cannot be rejected at better tion of t,he two galaxies is still equal. This makes it than the 95% confidence level according to a 1’ or unlikely for KGC 1404 clusters t.o strongly influence a Kolmogorov-Smirnov test.The nleanvelocity of our sample. However, we cannot exclude the possibil- the combined sample of globular clusters is 1429 & 4.5 ity that some of the globular clusters in our sample km s“, similar to the velocity derived for the stellar may have been stripped from NC;C 1404 or NGC 1380 conlponent (see Table 2). No rotation’is detected in at some point in the past (see Sect. 5.3). the combined sample. The maximum rptation along any axis is found to be 74 f 1Oi km s-?; A possible cause for the double peak in the velocity distribution

4 3.3. Rotation of annularsub-samples were computecl using a masinluln likelihood disper- sion estimator (seePryor k hIcylan 1993). The dis- We,have searched for evidence of rotation in var- persions of the different samples around a fixed mean ious annular sub-samples. No significant rotation is (taken to be the mean of the full sample, i.e., 1429 found in t,he inner regions. However, in the outer re- km s-’) are also given. We find thatthe velocity gions (>.5’) we fild a masimum rotation amplitude dispersion measured for the combined sample agrees of 1.53 3~ 93 krn S“ along a position angle 120 f 40 with those derived inclivitlually for the X.\T, NTT, degrees (east of north), roughly along the isophotal and Keck samples. major axis of NGC 1399 (z100 degrees, extrapolat- ing from the measurements ofcioudfrooij et al. 1994). More interesting is the fact that the velocity dis- Figure 5 shows the velocities of globular clusters with persion increases with radius. This behavior is clearly projectedradii greater than 5’ versustheir position shown by a locally-weighted, scatter-plot smoothing angles. The velocity profile predicted by the rotation fit to the velocity deviation squared (LOWESS, see amplitude computed above is shown superimposed. Cleveland ,& McGill 1954, as implemented by Geb- We ran Monte Carlo simulations (1000 realizations, hardt et al. 1994) and shown as a solid line in Fig. 6, computing random data points with the same veloc- ontop of thedata for theglobular clusters. Also ity dispersion, velocity errors and position angles as shown in Fig. 6 are the measured velocity dispersions spanned by our sample) to estimate the significance of the stars, the planetary nebulae, and the Fornax of this result: the hypothesis of no rotation for this cluster galaxies. sample is ruled out at the 95% confidence level. 4. Dynamics Ostrovet al. (1993),Iiissler-Patig et al. (1997a) and Forbes et al. (1998) found the color distribution 4.1. Are the Stellarand Globular Cluster of NGC 1399 globular clusters to exhibit two distinct Measurements Inconsistent? peaks. We split the sample into red and blue glob- ular clusters (at V-I = 1.05, following Kissler-Patig The difference between the line-of-sight velocity et al. 1997a). The small number statistics do not al- dispersions of the globular clusters and the starscould low us to see any difference in the kinematics of the be explained by very different density profiles of the blue and red samples. We only note that our sam- two components, or by different degrees of anisotropy. ple is dominated by red clusters inside 5’ (mean V-I Wagner et al. (1991), Bridges et al. (1991), Iiissler- = 1.16 f 0.02, where the error is the standard error Patig et al. (1997a)and Forbes et al. (1998) found of the mean), while blue and red clusters are about little or no difference in the surface density profiles equal in number beyond 5’ from the center of NGC of the stars and the globular clusters, accounting for 1399 (mean V-I = 1.04 f 0.04).

D ..

4.2. Can M/L be Constant? clerlying theoretical assrrrnpt,ions. E.g., adopting ra- dial instead of isotropic orbitsfor the globular clusters The spectroscopic measurements of the integrated increases the nlass estinlate by a fact.or of two. stellar lightalmost reach the range in radius where al. theinnermost globular clusters are measured. To- Other mass estimators (Heisler et 198.5) applied tothe same sample give the following results: the gether with the data for the planetary nebulae, there is n consistent indication of a marked up-turn in the virial mass 1LfL.T = 8.0 x loL?.\f,? implies !IfvT/L - velocity dispersion between I’ and 3’ from NGC! 1399. 167, themedian mass ill.\[ = 6.6 x 10” iLI(3 im- This is seen independently in each of the stellar, plan- plies M.vr/L - 138, andthe average mass Jl~v= etary , and globular cluster data. Regardless of 7.8 x IO” LC~,implies M/L - 162. The virial, pro- whether the orbit shapes are changing or the Fornax jected, and average masses share the same sensitivity cluster potential is becoming dominant, this suggests to interlopers, i.e., globular clusters from the intra- that all luminous components are similarly affected. cluster medium or from the nearby galaxies. Elimi- This is not a trivial result; whereas we find only little nating from the sample even a single such object can evidence for rotation in the globular cluster system, decrease the M/L values by as much as 30% (hIin- Arnaboldi et al. (1994) claim that the systemof plan- nitiet al. 1998). ‘These MIL values, obtainedfrom etary nebulae exhibits a rotation amplitude of 250 the globular cluster system of NGC 1399, are similar - to the corresponding values obtained with the same km s-l with a PA N -35O, i.e. counter-rotating with respect tothe globular clusters. If real,this might mass estimators by Huchra 9i Brodie (1987) from the constitute an important clue concerning the manner globular cluster system of M87. in which the cD envelope of NGC 1399 was built up. Contrary to thevelocity dispersion determinations, Following Grillmairet al. (1994), we determine which suffer only from the observational errors on the whether or not the increasein the measured veloc- radial velocities, the mass-to-luminosity estimates ac- itydispersion at largerradii must be attributed to cumulate the errors associated with many various ob- an increase in MIL by assumingthe extreme case served parameters, such as the total luminosity within that all of our observed globular clusters are on per- a given radius, the distance modulus and the radial fectly circular orbits (which for a given potential field velocity of the host galaxy, not to mention the dif- strength will producethe largest observed line-of- ferenttheoretical assumptions. Nevertheless, all our sight velocity dispersion). Assuming a spherical dis- M/L determinations (50 - 200) are larger than that tribution of mass in hydrostatic equilibrium, and that of a typical old stellar population (1 < hl/L < IO), the surface density of globular clusters and the stel- supporting the existence of a substantial amount of lar light both follow the surface brightness parame- dark matter. We also confirm that MIL is not con- terization of Killeen & Bicknell (1988), we use Equa- stant with radius. tions (2) and (3) of Grillmair et al. (1994) to solve in These values are significantly larger than t.he value a least-squares sense for a radially-invariant MIL ra- M/L = 17 determined by Bicknell et al. (1989), from tio. Sampling the velocity dispersion profile over the the stellar velocity dispersion measurements at radii same distribution of projectedradii as in our com- <86”. There is therefore strong evidence for a change bined sample, we find that our observations are best in the characterof the gravitational potential at about fit using MILE = 505 15 Ma/La, where the quoted 2’ - 3’ from the center of NGC 1399. We conclude, error reflects only the uncertainty in velocity disper- as suggested by Grillmair et al. (1994), that most of sion computed for the full sample in Table 3. the globular clusters in our sample are orbiting in a In a way similar to hlinniti et al. (1998) we use the gravitational potential which is more closely associ- projected mass estimator (Heisler et al. 1985), under ated with the galaxy cluster rat,her than with NGC the hypothesis of a spherical distribution of matter 1399 it.self. As t,he velocity dispersion profile mea- andisotropy of the velocity dispersion. For the 74 sured for planetary nebulae at large radii is very sim- ilar to that of the globular clusters, we must conclude globularclusters, we obtain Mp = 1.0 x MB. This massestimate, along with L = 4.8 x 10’O L, that the stars in t.he outer cD envelope are similarly (Grillmair etal. 1994, scaled to our adopted distance), associated with the potential of the Fornax cluster as gives Alp/L - 208. The error in this quantity is dam- a whole (see also Arnabolcli et al. 1996, Mendez et inatecl by the small sntnple size and the different un- al. 1997). This would be consistentwith the recent detection of red giant branch stars in the intergalactic

6 region of the Virgo cluster (Ferguson, Tanvir, & von Persio11 of fl N 450 km s“. The temperature profile, tiippel 1998). converted into a velocity dispersion profile, is sIlowl1 Recent radial velocity measurements by Cohen & asthe dashed line in Fig. 6. Once againthe agree- R.yzhov (1997) of globular clusters in h.187 show simi- ment with the globular cluster measurement is excel- lar velocity dispersions for stars and globular clusters lent.Ikebe et al. (1996) showed that the X-ray gs within 100’’ of the cent.er of h.18’7, but a steadily ris- profile could be modeled as the sum of two compo- ing velocity dispersion from there outward. These au- netlts, perhaps responding to the potentials of NC;C thors are similarly driven to conclude that the mass- 1399 and the Fornas cluster, respectively. The poten- luminosity ratio increases substantially with radius. tial of the galaxy evidently falls off steeply and only Thissimilarity may not be surprising as M87 and clearly dominates over the cluster potential in the in- NGC 1399 both occupy the center of their respective ner 2’ to 3’. Beyond 5’ or 6’, thepotential of the cluster potentials. cluster dominates.

4.3. A comparisonwith theFornax galaxy 5. Discussion cluster Inorder to better understand the origin of the 4.3.1. Thegalaxies surrounding NGC 1399 NGC 1399 globular cluster system, with its high ve- locitydispersion and specific frequency (number of An examination of the velocity distribution of 68 globular clusters per unit luminosity, Harris van Fornax galaxies studied by Ferguson (1989) using the & den Bergh 1981), we pose the question whether the statistical tests described in Sect. 3 reveals multiple properties of these globular clusters are peculiar in peaks of marginal significance similar to those appar- any way. We briefly compare the globular cluster sys- ent in Figure 2. This motivated us to look for sub- tem of NGC 1399 with those of the nextbrightest structure within the Fornaxcluster. We applied a early-type galaxies in Fornax, NGC 1380 and NGC Dressler 9i Shectman (1988) test on the entire sample 1404. We then discuss our findings within the frame- of galaxies; with the exception of a group of galax- work of different scenarios that could explain the glob- ies surroundingNGC 1316 (Fornax A, located at a ular cluster over-abundance. projected distance of N 1.2 Mpc SW of NGC 1399), we found no evidence for sub-structure. The Fornax Notethat NGC 1399 is traditionally associated cluster appears to be very homogeneous. This is in with high-S,v galaxies, although Ostrov et al. (1998) good accord with the general view that Fornax is a recently suggested that it has only a moderate over- very compact, relatively relaxed cluster. Leaving the abundance. The new value is a consequence of tak- group around NGC 1316 out of the velocity distri- ingproper account of the light in the extended cD bution considerably diminishes the statistical signifi- halo. The S,V value for NGC 1399,galaxy and cD cance of the double peak (to < 50%). envelope is still about a factor of two higher than for other galaxies in Fornax. In particular,the SN of The mean velocity of the 57 remaining galaxies in the cD envelope reaches values a factor of 3 higher Fornax is 1459 f 40 km s-l, consistent with the sys- than the mean of thebrightest surrounding ellipti- temic velocity of NGC 1399. This confirms that NGC cals, and when theglobular cluders are associated 1399 sits at the center of the cluster potential well. with the “galaxy” component alone, the SN reaches The velocity dispersion of the galaxy clusterrises from values of a factor .!I to 6 higher than the Fornax mean. = 276 km s-l in the outer parts (several degrees from the cluster center) to u = 413 km s-l within 24’ In the following we will distinguish between NGC of NGC: 1399 (den Hartog 9i Katgert 1997). There is 1399 including the cD envelope, the “galaxy” compo- very good agreement (Figure 6) between the velocity nentalone (to which we associate the light of a de dispersions of the galaxies and the globular clusters. Vaucouleurs fit to the central region), and the cD en- velope. Mc, and s,~for the different components are 4.3.2. The X-ray gus listed in Table 2. Joneset al. (1997) find thatthe hot X-ray gas arouncl NGC: 1399 has a temperature of 1.30 f 0.0.5 ke\‘ (clerived within an annulus extending from 2’ to 18’). This is energetically equivalent bo a velocity dis-

7 5.1. Photornetric anti chemicalproperties of macle of the globular clusters surroundingthese other the globular clusters surrounding NGC two galasies, a consiclerahle amount of photometric 1309 work has been published (NC;C 1380: Iiissler-Patig et al. 1097b; NC;C 1404: Hanes k Harris 1986, Richtler The globular cluster syst.em of NGC 1399 has been et al. 1002, Forbes et al. 1998, Grillmair et ai. 1999). estensively studiecl photometrically (Hanes S: Harris Qualitatively,the globular cluster systems of NGC 1986, Geisler k Forte 1990, \Vagner, Riclltler ,E: Hopp 1399, NCiC 140.1 and NC;C 1380 look very similar. 1991, Bridges, hies S: Harris 1991, Ostrov, Geisler All threesystems exhibit broad color distributions 9; Forte 1993, Kissler-Patig et al. 1997a,Forbes et with two (or possibly three) peaks (Ostrov etal. 1993, al. 1998,Ostrov et al. 1998,GriIlmair et all 1999). Iiissler-Patig et al. 1997a,b, Forbes et al. 1998, Grill- The moststriking result of these investigations is mairet al. 1999). Theratio of red to blue globu- the high number of globular clusters surrounding the larclusters is roughlythe same inall three cases, galaxy,and the over-abundance in thenumber of and between 1 and 2 when integrated over the whole globular clusters per unit light when compared with system. Furthermore, the two main sub-populations the other “normal” cluster ellipticals (see Table 2). peak at similar metallicities (as derived from broad- Anotherinteresting finding of themore recent band colors) in the three systems. The blue popula- studies is thatthe globular cluster color distribu- tions are distributed around a typical for tionhas at leasttwo significant peaks, suggesting the Galaxy or hi31 halo globular clusters,i.e., -1.5 < two cr more sub-populations. The photometric stud- [Fe/H] < -1.0 dex. The red populations seem to have ies found that thesepeaks are separated by about metallicities slightly higher than disk/bulge globular 1dex in [Fe/H] and, taking into account the lumi- clusters in theGalaxy (i.e., -0.5 < [Fe/H] < -0.1 nosityfunction, thatthe majority of theglobular des, cf. Minniti 1995, although the former could be clustersare as old (within a few Gyr) as the old- slightly over-estimated from the broad-band colors, estglobular clusters in the Galaxy. Thenature of see Kissler-Patig et al. 1998). these sub-populations was explored spectroscopically Finally, the globular cluster luminosity functions by Kissler-Patig etal. (1998). The absorption line of all three systems look similar and are well repre- indices for the two major sub-populations of glob- sented by a Gaussian with a 1.2f 0.2 mag dispersion ularclusters were found tobe very similarto the and peak at about the same magnitude to within the ones observed in globularclusters around M31 and errors (see Table 4). Assuming that the underlying the Galaxy, suggesting similar formation epochs and globular cluster mas distributions are invariant, this mechanisms. Thus the majority of globular clusters result suggests that the globular clusters in the three around NGC 1399 appear to be “normal” oldglob- galasies have similar old ages. ularclusters. A small fraction of very red globular clusters are significantly more metal rich, and could The only significant difference between the glob- perhaps have formed in a later merger. ular cluster system of NGC 1399 and those of NGC 1404 and NGC 1380 is the muchhigher number of 5.2. Comparison withthe globular cluster clustersaround NGC 1399. Quantitatively, NGC systems of NGC 1380 and NGC 1404 1399 is surrounded by 5800 f 300 globular clusters, while NGC 1404 and NGC 1380 host only 800 f 100 Do NGC 1380 and NGC 1404 have globular clus- and 560 k 30 globular clusters respectively (see Ta- ter systems comparable to that of NGC 1399? These ble 2). In terms of specific frequency, NGC 1399 has two galaxies are the next brightest early-type galax- S’,V= 4.1 k0.6 (S,V= 11 f 1 for the “galaxy” compo- ies in Fornax (neglecting currently -forming NGC nent) NGC 1404 has s,~= 2.3 f 0.3, and NGC 1380 1316). NGC 1380 and KC;C 1404 are 1.4 and1.5 has S.v = 1.5 f 0.2. Using the distance in Table 2 magnit,udes fainter, respectively, than the central cD and the magnitudes and colors of the galasies from galasy, and only 0.3 and 0.4 magnitudes fainter if the the RC3, the four smaller early-type galasies in For- light of the cD envelope is neglected (see Table 2). nas NC:C 1387. NGC! 1374, NGC‘ 1379, IL’GC 1427 These galaxies are projected only 38’ (208 kpc) and (iiissler-Patig et al. 1997a) have specific frequencies 9’ (48 kpc) from NGC 1399 and are presumably well between ‘2.1 and 3.2 with a mean of 2.6 f 0.6 (where within the embrace of the Fornas cluster potential. the error is thedispersion around the mean). NGC W‘hile no spectroscopic observations have yet been 1399, or ratherthe center of the Fornas cluster, is

8 *

outstanclillg in its relative overabundance of globular crease in SN of thecentral galaxy. Finally, moti- clusters, while NGC 1404 and NCC 1380 have cluster vat(xl the constantlulninosity of Brightest C:lus- populations close to the mean expected value. ter Galaxies (BCGs), Blakeslee et a]. propose a ne!\: In summary,the globular clusters around NGC scenario wherein globular clusters fornletlearly and 1:199 are qualitatively very similar to the ones sur- their numbers scaled with the available 111a.w in the rounding NGC 1380 and NGC 1404, and appear “nor- galaxy cluster, whereas t,he luminosity of the BCGs mal” when compared to the globular clusters of the is relativelyindependent of thecluster and is Galaxy. However, the central galaxy hosts a factor 10 truncated at a given luminosity. The upshot is that more globular clusters than its neighbors and has a abnormally high Slv values would beinterpreted as specific frequency 2 to 3 times higher than expected. a luminosity deficiency rather than a globular ~1~s- Based on all the evidence summarized above, we are ter over-abundance (see also Harris et al. (1998) for led to two conclusions: i) the excess globular clusters further support to this idea). do not have unusual properties, i.e., they are likely Based on the observations presented above, we con- to haveformed in the same kind of process and at cur that excessive numbers of globularclusters are a similar time as the globular clusters in NGC 1404 likely to be linked to the properties (X-ray temper- and NGC 1380, and ii) the excess number of globular ature; velocity dispersion) of the galaxy cluster and clusters is most likely linked to the special location of the special location of high-S,v galaxies. NGC 1399 in the center of the cluster. 5.3.2. Constraints from theproperties of theexcess 5.3. Constraints on scenariosexplaining the globular cluste.rs high specific frequencies around central Any scenario that purports toexplain the excess of giant ellipticals globular clusters around NGC 1399 must be consis- 5.3.1. Currentscenarios tent with the finding thattheir properties are very similarto those of globularclusters in NGC 1404 Variousscenarios toexplain the high S,v values andNGC 1380. As discussed above, severalsub- aroundcentral galaxies have beencomprehensively populations have been identified in NGC 1399. The discussed by Blakeslee et al. (1997).These authors two major ones, making up over 90% of the system, found several correlations of specific frequency with are old and have mean metallicities of [Fe/H]- -1.3 the galaxy cluster properties, includingX-ray temper- and 2: -0.6 des respectively. The ratio of these two ature, velocity dispersion, and galaxy distance from populations changes slowly with radius, varying from the center of the cluster. In summary, ab initio sce- about2:l (red/blue) in thecenter of thegalaxy to narios (Harris 1991 and references therein) that asso- about 1:2 in the cD halo, with an overall average close ciate theexcess numbers of globular clusters to super- to 1:1 (see Iiissler-Patiget al. 1997a and Forbes et efficient globularcluster formation in these galax- al. 1998). In principal,the excessglobular clusters ies fail to explain the correlation of SN, cluster X- could be made up by metal-rich or metal-poor glob- ray properties and velocity dispersion. Mergers as a ular clusters only. Both cases would be inconsistent source of theincrease in SN (e.g.Ashman si Zepf with NGC 1404 and NGC 1380having also metal- 1992) also fail to explain these correlations and have poor and metal-rich clusters, but showingno over- difficulty explaining the ratio of blue and red glob- abundance. This suggests that the excess of globular ular clusters if the red globulars are thought to be clusters(especially around the “galaxy” component) responsible for the high SN values (Forbes, Brodie si must be made up of both old, metal-poor and old, Grilltnair 1997). metal-rich clusters. Scenarios which involve biased formation of glob- This complicat.es scenarios wherein all the excess ular clusters in deep potential wells (West 1993) and clustersare formed at a very early stage, before intra-cluster globular clusters (West et al. 1995) are the galaxies fully assembled (i.e., West et al. 1993; largely unconstrained, and Blakeslee et al. cannot rule Blakeslee et al. 1997; Harriset al. 1998). Such sce- the111 out,.Tidal stripping can easily explain all of narios would require that NGC 1399 have a specific their observed correlations;the problem, as noted frequency about a factor of two higher than normal, by Blakeslee et al., is that the stripping simulations but that all of the extra clusters be metal-poor. A (sutnnmrizccl in Muzzio 1987) yield too slow an in- mechanisnl that can explain at least some irwrease in

9 I

the number of red globular clust8ers seems required. (25 f 9). Miller et al. (1998) recentlyenlarged the sample of dE galaxies with studied globular cluster 5.3.:’. Corrsttnitrts fromthe properties of BCCs systems by 24 and extentled Durrel et, al.’s results in the sense that their dE,N galaxies with Mv < -1.5.5 Recent work by Arag6rt-Salamanca etal. (1997) also have S,v values around 10 or above. However, putsfurthcr constraints on scenarios which require brighter ancl non-nucleated dEs appear to have more an escess nunlher of globular clusters to have formecl normal specific frequencies. Simulations by Cot6et at very early times. These authors see no (or nega- al. (1998) indicate that the accretion/stripping of a tive) passive evolution in the 1;-band Hubble diagram large number of dwarf galaxies couldexplain thecolor of BCGs. Furthermore, they detect no young stellar distributions seen around giant ellipticals, assuming populations.They conclude that the total mass of that every individual galaxy forms globular clusters stars in BCGs has grown by a factor of 2 (4) since with a mean metallicity proportional to itssize. How- ”- 1 for qo 0 (qo 0.5), and cannot be ex- - = = ever, the presence of a large number ofsuch nucleated plained bynew stars forming in mergers or cooling dwarf galaxies remnants at the current epoch around flows. BCGs must therefore have accreted or annexed NGC 1399 can be ruled out (Hilker 1998 (chapter 5), mass in the form of old stars (and presumably old Hilker et al. 1998a,b and references therein). Hilker globular clusters) since z 1; either by cannibalism = (1998)further showed in Monte-Carlo simulations - accreting gas-poor galaxies - or by stripping of that dwarfgalaxies (or theirhalos) cannot explain material from other galaxies. all the excess of globular clusters around NGC 1399, 5.3.4. Constraintsfrom the high specific frequency even if one assumes an extremely steep faint-end of the galaxy luminosity function in the past and effi- Irrespective of the high total number of globular ciently accrete all the dwarf galaxies “missing” in to- clusters, the high specific frequency, when compared days luminosity function onto NGC 1399. Therefore, to the neighbor galaxies, also needs to be taken into while accretion of dwarf galasies may have played a account. If NGC 1399 gainedmass simply by dis- role in the enrichment of the globular cluster system sipationless mergers of typical galaxies surrounding of the central galaxy, it cannot be its only cause. it,then the specific frequency would decrease (by Another source of high specific frequency mate- about 30% for each doubling of the mass for NGC rialmay begalactic “halos”. Extended globular 1399), sinceno neighboring galaxy has a SN value clustersystems often fall off more slowly with ra- much higher than 3. Dissipationless merging therefore diusthan the stellar light profile,leading to grad- seems unlikely to be the main causefor the growth of uallyincreasing specific frequenciesin the outskirts the central galaxy since I = I, unless the specific fre- of thedistributions. Recent studieshave identified quency of NGC 1399 was even higher at early times. distinct “halo” globular cluster populations in early- LVe would then need an esplanation for the original, type galaxies (Iiissler-Patig et al. 199Tb; Lee, Kim ,P: even higher, specific frequency. It seems to us more Ceisler 1998) comprised of clusters following a some- reasonable to suppose that the mass gained by NCC whatshallower density profile. &I87 is a clearcase, 1.799 consisted of accreted material with an intrinsi- for which McLaughlin et al. (1994) computed an in- cally high specific frequency of globular clusters that crease of SN from values around 10 inside 1 reff to 2.5 matches the S,v z 6 of the cD halo (see table 2). at >5 rev. Unfortunately, such studies do not yet ex- Potentialaccretion candidates withhigh specific ist for isolated field galaxies, but “halo” material can frequencies include faint dwarf galaxies and galactic be considered a plausible candidat,e for the buildup halos.Durrel et al. (1996) compiled a table of SN of the envelope of NCiC 1399 (see also Kissler-Patig values for local group dwarfsand severaldwarf el- 1997a andForbes. Brodie, k Grillmair1997). \Ye lipticals in Virgo. They find thatdE galaxieswith note in passing that. in the galaxy formation models < -15.5 tendto have a “normal” specific fre- of the Searle k Zinn (1978) type, halos are built up quency (4.2 f 1.8), whereas their two faintest dE,N of fragments resembling faint dwarf galaxies. galaxies have very high SN values (1.5 f 8 ancl 12 zt 9 respectively). At least two other, local group, low- 5.4. Globularcluster stripping luminosity dwarf galaxies are also known to exhibit Several authors have investigated the importance very high values: Fornas (29 f 6) andSagittarius of stripping and harassment for the fortnation of cen-

10 tral cD galaxies (see e.g.,Dressler 1984 for an early re- the globular clusters seen in tllese galaxies today. If view). Following these ideas, Harris (1986) suggested we assumethat the next seven brightestearly-type that the central regions of galaxy clusters could have galaxies (NGC 1XK, NC;C ~:wJ, NC;C ~1.51,IC 1963, accumulated globular clusters stripped from their par- NC;C.: I:ldO..\, h‘(::(;! 1981, NGC 1389, take11 from Fer- ent galaxies. White (1987) suggested that these clus- gusoll 1989), all with total magnitudes brighter than ters could have built up the over-abundant globular Jfb. N -19, suffered the same losses as the other non- clustersystems around cD galaxies. Muzzio (198i) centralgalaxies, the initial specific frequency coulc] summarized the results ofvarious simulations of glob- have been as low as 3. Note thatthe largest losses ular cluster stripping in galaxy clusters. Some of his were suffered by the brightestgalaxies. Including conclusions were: i) stripping of globular clusterswas contributions from dwarf galaxies could lower the re- most efficient before the galaxies and clusterwere viri- quired initial mean value of S,Vstill further. alized, ii) the initial extent of the halos is not criti- Even if all globular clusters were associated with cal, i,ii) thelocation in thecluster and the amount t,he “galaxy”component, and todays S,V value for of dark matter determine the number of encounters NGC 1399 was takento be >lo, a similar calcula- and the amount of stripping a galaxy suffers, and iu) tion would lead to an initial S,Vof N 4 for all Fornax the brightest galaxies tend to gain globular clusters, galaxies. but capture most of these from galaxies only some- SN values of 3 to 4 are not unusual among early- what fainter than themselves. Intermediate luminos- type, cluster ellipticals. The mean for all early-type ity galaxies are the oneswhich lose their globular clus- galaxies with -19.0 < ~Mv< -21.4 in various compi- ters; faint or dwarf galaxies are hardly affected. Un- lations is S,v = 3.7f3.1 (Harris 1991), S,v = 4.1f2.9 fortunately, the exact propertiesof the resulting glob- (Ashman & Zepf 1998) and S,V= 4.6 k 2.7 (Kissler- ular cluster systems depend too much on the initial Patig 1997b),where the quoted error is the dispersion conditions and the details of the simulations (which aboutthe mean. Howeuer, these numbers need to had rather poor resolution) to allow detailed predic- be confirmed for a sample of isolated field ellipticals tions. We discuss the plausibility of the “stripping” withabsolute magnitude -19 < MV < -22 before scenario in more detail below. we can say with confidence that this is likely to be representative of cluster ellipticals prior to any inter- 5.4.1. Do thenumbers work out? actions. Nonetheless, thepoint we wish to stress is We first determinewhether the total number of that the excess population around the central galaxy globularclusters seenin the early-typegalaxies in in Fornax does not require the production of a large Fornax requires that an additional formation process number of newglobular clusters. .A simple redistri- be invoked, or if it can be explained by a simple re- bution,separating globular clustersfrom their par- distribution of existing clusters. In other words, do ent galaxies through tidal encounters. is sufficient to the galaxies in the cluster form a reservoir of globu- account for theobservations. NGC 1399 (or per- lar clusters large enough to feed the central galaxy? haps more correctly the central potential of the For- To answer this question, we assume t,hat all galaxies nax cluster) would have gained about 6.5% of its cur- had the same number of globular clusters per unit lu- rent light (assuming for the galaxy without envelope minosity to begin with, and that the excess around izlv = -21.i6, see Table 2), and 7.5% of its current NGC 1399 must be accounted for by the losses suf- globular clusters through stripping and harassment. fered by othergalaxies. What, then, was the initial NGC 1380 and NGC 1404 would have lost 30% to 50% “universal” specific frequency in Fornax? of their globular clusters, and all other intermediate- Table 5 summarizes the result. It shows the num- size galaxies would have lost between 0% and 30% of ber of globular clusters observedtoday in each of their initial globular cluster populat.ions. brightest early-type galaxies in the center of the For- In the above, we neglected the possible exist.ence of [lax cluster, the number of globular clusters initially a population of intra-cluster globular clusters (West resident in each galaxy for an assumed L‘universal’’ et al. 199.5, Blakeslee 1996) which was never associ- specificfrequency, arid the number which had to be ated wit,h a clustergalaxy. If indeed suchglobular gainecl or lost. Assuming an initial S,v = 3.2 for NGC clusters formed at early tirnes,then the numbers of 1:3$)9, NC;C 1404, NC;C 1380, NC;C lL374, NGC 1379, globulars removed from cluster galaxies needs to be NC;(3‘ 1387, ancl NGC! 1427,one can account for all even less;, and the initial specific frequency could be ?

ever1 lower than the values we have estimated above. sensible, cletailed simolat,ions will be needed to check However, little is yet known concerning such a popu- thissuggestion. X potentialremaining problem, M lation, making any estimate of its importance uncer- alreadypointed out by Blakeslee et al. 1997, is the tain. inefficiency of stripping once galaxy clusters are viri- hlergers, both gas-rich and gas-poor, must presum- alizecl. This apparent problem would however be re- ably have occurrecl during the evolution of the Fornax solveti if a significant fraction of the tidal stripping cluster. However, rather than increasing the specific occurred during the early phases of galasy formation, frequency as has been often suggested (e.g., Zepf &Z when the halos of galaxies first collapsed and the old, Ashman 1993), such mergers could just as easily con- metal-poor globular cluster population formed. How- serve S,v (e.g., van den Bergh 1995). ever, tidal effects would continue to pull clusters from their host galaxies until the present day, although at 5.4.2. Are the other properties of the globular cluster reduced efficiency. This, as well as dissipative post- systems compatible with stripping? collapse merger events, would allow for the addition of more metal-rich globulars to the mis. All the con- Iiissler-Patig et al. (1997a) showed that the com- straints discussed in the above sections would remain bined color distribution of globular clusters in the in- satisfied. termediate luminosity galaxies surrounding the cen- tral galaxy (NGC 1374, NGC 1379, NGC 1387, and 6. Summary and conclusions NGC 1427) is also consistent with the observed color distribution in NGC 1399, and that thisis compatible We have combined data from three separate spec- with the stripping scenario. troscopic investigations to study thekinematics of 74 Unfortunately, as noted above, few simulations of globular clusters around NGC 1399. The velocity dis- the stripping of globular clusters in galaxy clusters persion of the globular clusters increases with radius, exist. hluzzio (1987) summarized a number of simu- rising from a value not unlike that for the outermost lations but was unfortunately not able to make clear stellar measurements at 2 Teffr to values almost twice predictions. Globularcluster systems are expected as high at >5 r,~.The outer velocity dispersion mea- to become extended in response to the tidal stresses surements are in good. agreement with the temper- imposed by neighboring galaxies. However, it is not ature of the X-ray gasand the velocitydispersion clearwhether by this time we shouldexpect to see of galaxies in theFornax cluster. R‘e conclude, as surface density profiles which are tidally truncated,or alreadysuggested by Grillmair et al. (1994), that a whether they should appear extended. Tidal encoun- large fraction of the globular clusters which we asso- ters will naturally lead to the growthof tidal tails and ciate with NGC 1399 should rather be attributed to to the possible destruction of numerous globular clus- the whole of the Fornax cluster. No significant dif- ters(Gnedin CP: Ostriker 199’7, Combeset al. 1998). ference in the kinematics could be found between the The onset of such tidal tails may be detectable in the blue and red globularcluster sub-populations, but form of a “break” in the surface density profile (Grill- there is some evidence for rotation in the outer (> 5’) mair et al. 1995; 1999). regions. X qualitative comparison of the globularcluster 5.4.3. A possible scenario systems of NGC 1399 and neighboring, next-brightest Currentlyavailable information on theglobular early-typegalaxies NGC 1404 and IVGC 1380 indi- cluster system of NGC 1399 seems to favor the view catesthat these systemsare indistinguishable from thatits high specific frequencymainly results from one another, ancl that there is no reason to suppose tidalstripping of relativelyhigh-Sv material from that they formed at significantly different epochs or neighboringgalaxies. If this is thecase for other via a different sequence of events. The NGC 1399 nearby central cluster galaxies, it would explain the globular cluster systenl is clistinguishecl only in being correlations ofS,v with galaxy cluster properties (Blakeslee a factor of10 more abundantthan the cluster sys- etal. 1997). In the case of NGC 1399, stripping is tems of either of t,he other two galasies, and having a consistent with all known properties of the globular specific frequency 2 to 3 times higher. The excess is clust,er system, inclucling the total number of globu- best understood if a significant fraction of the globu- lar clusters in t,he cluster galasies. LVtlile this seems lar clusters is indeed associated with the light of the

12 I

cD envelope. By associationthis would meanthat Blakeslee, J.P., 8; Tonry,J.L., Metzger, M.R. 1997, the cD envelope around NGC 1399 should rather also AJ 114, 482 be associated with the Fornax cluster than with the Buts, R., & Williams, r;.r,. 199.7, AJ 109, 543 galaxy itself. Lire review different scenarios to explain the high c!kvehlld, w., k h[c(;iil, R. 1984, J. Am. Stat. AS- SOC., 79, 807 specific frequencies aroundthe central galaxies an(] examinethe consequences of variousexisting con- Cohen, J.G., 8; Ryzhov, A. 1997, ApJ 486, 2:30 straints for each. We come to theconclusion that tidal Cohen, J.G.,Blakeslee, J.P., 8; Ryzhov, A., 1998, ApJ stripping of globular clusters from neighboring galax- 496, 808 ies in the early history of the galaxy cluster and the Combes F., Leon S., Rileylan G., 1998, A&A, submit- consequent buildup of the cD envelope in the Fornax ted cluster potential well is the most likely explanation. Cot6, P., Marzke, R.O., 8; West, M.J. 1998, ApJ 501, We would like to thank Ann Zabludoff for her im- 5.54 proved code of the Dressler 8; Shectman test as well as DellaValle, M., Kissler-Patig, M., Danziger, J., Storm, useful comments, and Karl Gebhardtfor his LOWESS J., 9; Richtler, T. 1998, MNRAS in press code. We arealso thankful to Steve Zepf andBill Dressler, A., 1984, ARA&A 22, 185 IvIathews for interesting discussions. MKP gratefully Dressler, A., 8; Shectman, S.A., 1988, AJ 95, 985 acknowledges the support of the Alexander von Hum- boldt Foundation. Part of this research was funded by Durrel,P.R., Harris, W.E., Geisler, D., 8; Pudritz, the faculty research funds of the University of Califor- R.E. 1996, AJ 112, 972 nia at Santa Cruz, the HST grant G0.06554.01-95A, de Hartog, R., 8; Katgert, P., 1996, MNRAS 279, 349 and the US. Department of Energy by Lawrence Liv- Elson, R., 8; Santiago, B. 1996a, MNRAS 278, 617 ermore National Laboratory under Contract W-7405- Eng-48. AtJ 0 j%T k-T%"T * Elson, R., 9c Santiago, B. 1996b, MNRAS 280, 9T1 dod6%4GT d Ferguson, H.C. 1989, AJ 98, 367 REFERENCES 0.s ul Ferguson, H.C., Tanvir, N.R., 8; von Hippel, T. 1998, Arag6n-Salarnanca, A,, Baugh, C.M., si Kauffrnann, Nature 391, 461 G. 1998, MNRAS in press Forbes, D., Brodie, J., Grillmair, C.J., 1997, AJ 113, Arnaboldi, &I., Freeman,K.C:, Hui, X., Capaccioli, 1652 M., Ford, H. 1994, ESO Messenger 76, 40 Forbes, D.A., Grillrnair, C.J., Williger, G.M., Elson, Arnaboldi, M., et al. 1996, ApJ 472, 145 R.A.W., si Brodie, J.P. 1998, hINRAS 293, 325

Xshrnan, K.M., si Zepf, S.E. 1992, ApJ 429, 557 Frans, AI., Illingworth, G., Heckman, T. 1989, AJ 98, 538 Ashman, I<.&l., 9; Zepf, S.E. 1998, Globular Cluster Systems, Cambridge University Press Gebhardt, I<., Pryor, C., Williams, T.B., 9; Hesser, J.E. 1994, AJ 107, 2067 Beers, T.C., Flynn, I<., si Gebhardt, I<. 1990, AJ 100, 32 Geisler, D., si Forte, J.C., 1990, ApJ, 350, L5 Bicknell, G.V., Carter,D., Killeen, N.E., 8; Bruce, Geisler, D., Lee, M.G., Kim, E. 1996 AJ 111, 1529 T.E. 1989, ApJ 336, 639 Gnedin O.Y., Ostriker J.P., 1997, ApJ, 474, 223 Binney, J., ,P: Marnon, G. 1982, MNRXS 200, 361 Goudfrooij, P., Hansen, L., Jorgensen, H.E., Norgaard- Bird, C.M., k Beers, T.C. 1993, AJ 105, 1596 Nielsen, H.U.. deJong, T., van denHoek, L.B., Bridges, T.J., Hanes, D.A., si Harris, \V.E., 1991, AJ 1904, t1k.U 104, 179 101, 469 Grillmair, C..J. 1992,PhD Thesis, Australian Na- tional University Brotlie, ,J.P., si Huchra, J.P. 1991, ApJ 379, 157 Grillmair, C!..J., Freeman, I<.C.,Bicknell, G.V., et Blakeslee, ,J.P., ,P: Tonry, J.L. 1996, ApJ 465, 19L 1994 ApJ 42'2, L9

13 t

Grillmair, C.J., Freeman, K.C., Irwin, M., & Quinn, Lee, bI.G., Kim- E., k Geisler, D., 1998, A.111.5, 94T P..J., 199.5, AJ 109, 25.53 McLaughlin,D., Harris, W.E., & HanesD.A., ApJ Grillmair, C. J., Forbes,D. X., Brodie, J. P., S; Elson, 422. 486 R. 1999, A.J, in press htendez, R.H., Ciuerrero, hI..-l., Freeman, I<.C., Arn- Crillmair, C.J., Quinn, P.J.. Freeman, K.C, Salmon, aboldi, hf., Iiudritzki, R.P., Hopp, U., Capaccioli. J., Messina, P., k Prince. T. 1999, in preparation M., S; Ford, H., 1997. XpJ 491, L23 Hanes, D.A., & Harris, W.E. 1986 ApJ 309, 564 Miller, B.W.,Ferguson, H.C., Lotz, J., Stiavelli. Harris, W.E. 1986, AJ 91, 822 M., & Whitmore, B.C., 1998, in DwarfGalaxies andCosmology, theProceedings of the XVIIIth Harris, W.E. 1991, ARAA 29, 543 Moriond Astrophysics Meeting, eds. T. X. Thuiin, Harris,H.C., Harris, G.L.H, Hesser, J.E. 1988, in C. Balkowski, V. Cayatte, & J. Tran Thanh Van, Globular Cluster Systems in Galaxies, IAU Symp. Editions Fronti&res, Gif-sur-Yvette 126, eds. J.Grindlay and A.G. Davis Philip, p.205 Minniti, D. 1995, AJ 109, 1663 Harris,W.E., Harris, G.L.H., & Mclaughlin,D.E. Minniti, D., Kissler-Patig, hl., Goudfrooij, P., Mey- 1998, AJ 105,1801 lan, G. 1998, AJ 115, 121 Harris, W.E., & van den Bergh, S., 1981, AJ 86, 1627 hIould, J.R., Oke, J.B., & Nemec, J.M. 1987, AJ 93, Heisler, J., Tremaine, S., & Bahcall, J. N. 1985, ApJ 53 298,8 Mould, J.R., Oke, J.B., de Zeeuw, P.T., & Nemec, Hernquist,L., & Bolte, M. 1992, in The Globular J.M. 1990, AJ 99, 1823 Cluster Galaxy Connection ASP Conf. Series Vol. Muzzio J.C., 1987, PASP 99, 245 48, eds. GSmith & J.P.Brodie, p.788 Ostrov, P., Geisler, D. and Forte, J.C. 1993 AJ 105, Hiker, hl. 1998, PhD Thesis, Sternwarte Bonn 1762 Hiker, hl., Kissler-Patig,bf., Richtler, T.,Infante, Ostrov, P., Forte, J.C., and Geisler, D. 1998 AJ De- L., 8; Quintana, H. 1998a, A&AS in press cember Hiker, hI., Infante, L., Vieira, G., Kissler-Patig, hi., Pryor, C., & hleylan, G., 1993, in Structure and Dy- & Richtler, T. 1998b, AkAS in press namics of Globular Clusters ASP Conf. Series Vol. Huchra, J.P., si Brodie, J.P. 1987 AJ 93, 779 50, eds. S.G.Djorgovski & G.Meylan, p.357 Hui, X., Ford, H.C., Freeman, K.C., k Dopita, bl.A., Richtler, T., Grebel, EX., Domgorgen,H., Hiker, 1995, ApJ 449, 592 M., & Kissler, M. 1992, ASiA 264, 25 Ikebe, Y., et al. 1996, Nature 379, 427 Searle, L., k Zinn, R. 1978, ApJ 22.5,3.57 Jones, C., Stern, C., Forman, W., Breen, J., David, Sharples,R.M., Zepf, S.E., Bridges, T.J., Hanes, L., S; Tucker, W. 1997, ApJ 482, 143 D.A.,Carter, D., Ashman, K.M., & Geisler, D.. 1998, AJ 10.5, 2337 Iiilleen, N.E.B., S; Bicknell, G.V. 1988, ApJ, 32.5, 16.5 van den Bergh, S. 1995, ApJ 450, 27 Iiissler-Patig bf., 1997a PhD thesis, Sternwarte Bonn Vaucouleurs, G. de, Vaucouleurs, A. de, Corwin, Kissler-Patig hf., 1997b A&4 319, 83 H.G.,et al. 1991, ThirdReference Catalogue of Iiissler-Patig, bf., Kohle, S., Hiker, hl., Richtler, T., Bright Galaxies, Springer:New York Infante, L., Quintana, H. 1997a, AsiA 319, 470 Wagner, S., Richtler, T., &k Hopp, U., 1991, .4&A 241, Iiissler-Pat.ig RI., Richtler T., Storm J., Della Valle 399 bI., 1997b, AkA 327, 503 West, M.J. 1993, MNRXS 265, 755 Iiissler-Patig, hf., Brodie, J.P., Schroder, L.L., Forbes, West, M.J., C!ot,t;, P., Jones C., Forman,W., & D.A., Grillmair, C.J., Huchra, J.P. 1998, AJ 11.5, Marzke, R.O., 1995, ApJ 4.53, L77 105 White, R.E. 111 1987, IlINRAS 227, 185 Kissler-Patig, hf., k Gebhardt, Ii., 1998, AJ Nov 98 Winsall, M.L., S; Freeman, 1i.C. 1993, AkA 268, 443

14 Zepf, S.E., 8i Ashrnan, K.M. 1993, MNRAS 264, 6 11

This 2-columrl preprintwas prepared with the AAS D'QX macros v4.0.

1.5 I

ID RA( 10.50) DEC( 1950) Vhelio V V-I Bj B, - R other Z)helio f0.02f0.035 f N 0.2 f 21 0.3 aat 1 3 3.5 47.3 -35 40 21 1121 f 150 ...... 21 .i 1.31 aat 4 3 3.5 49.2 -3.5 36 44 2478 f 150 ...... 22.3 1.04 aat 5 3 3.5 50.1 -3.5 38 48 1624 f 150 ...... 21.3 1.32 aat 6 3 35 51.4 -3.5 37 52 1186 f 150 ...... 21.9 1.13 aat 7 3 3.5 52.0 -3.5 32 44 1385 f 150 ...... 21.5 0.98 aat 8 3 3.5 54.8 -35 38 09 1152 f 150 ...... 22.4 0.84 aat 10 3 3.5 58.4 -35 36 03 1068 f 150 ...... 21.9 0.91 aat 13 3 36 03.6 -35 31 58 1922 f 150 21.15 0.97 21.9 1.05 aat 15 3 36 05.2 -35 39 54 1355 f 150 ...... 21.9 1.05 aat 16 3 36 05.9 -35 34 20 1766 f 150 21.27 1.17 22.1 1.16 aat 17 3 36 07.8 -35 34 44 1784 f 150 21.57 0.92 22.3 0.89 aat 20 3 36 17.8 -35 38 42 1836 f 150 ...... 21.6 1.33 aat 21 3 36 18.2 -35 40 52 2085 f 150 ...... 21.9 1.27 aat 25 3 36 22.7 -35 38 12 2182 f 150 ...... 22.4 1.19 aat 26 3 36 23.5 -3.5 37 25 1646 f 150 20.80 1.06 21.8 1.54 aat 27 3 36 14.7 4533 54 1921 f 150 2 1.85 0.96 22.3 0.81 aat 30 3 36 24.2 -35 42 07 18.59 f 150 ...... 21.8 1.0.5 aat 31 3 36 24.7 -3.5 33 24 1236 f 150 20.59 1.01 21.4 1.23 aat 33 3 36 30.1 -3.5 39 10 1350 f 150 ...... 21.6 1.48 aat 34 3 36 35.1 -3.5 31 15 1701 f 150 20.78 1.01 21.6 1.17 aat 36 3 36 41.3 -3.5 41 56 1038 f 150 21.97 1.43 22.3 1.17 aat 38 3 36 43.0 -35 33 16 574 f 150 20.85 1.18 21.7 1.40 aat 39 3 36 44.5 -35 38 31 1639 f 150 21.19 1.27 22.0 1.32 aat 40 3 36 44.4 -3.5 36 49 1539 f 150 20.78 1.21 21.8 1.54 aat 41 3 36 45.7 -3.5 38 53 571 5 150 20.93 1.22 21.7 1.31 aat 42 3 36 45.8 -3.5 37 33 1504 f 150 20.74 1.42 21.5 1.69 aat 43 3 36 46.0 -35 32 26 1623 f 150 21.23 1.13 21.9 1.17 aat 48 3 36 52.4 -3.5 34 33 88.5 f 150 21.71 1.16 22.4 1.19 aat 49 3 36 46.9 -35 35 44 2026 f 150 2 1.30 1.28 22.2 1.55 aat 54 3 36 51.9 -3.5 33 32 1941 f 150 21.04 0.97 21.9 1.29 aat 55 3 36 54.0 -35 37 27 1821 f 150 20.94 1.oo 21.6 1.01 aat 56 3 36 54.1 -3.5 31 25 1206 f 150 21.39 1.15 21.7 1.29 aat 57 3 36 55.4 -35 31 51 1742 f 150 20.91 1.12 22.2 1.26 aat 59 3 36 59.9 -3.5 41 21 1862 f 1.50 21.12 0.75 21.5 1.24 aat 62 3 37 01.2 -3.5 40 49 794 f 150 21.25 0.89 21.2 1.35 aat 66 3 37 10.1 -35 36 36 84.5 f 1.50 21.17 1.12 21.8 0.98 aat 67 3 37 10.8 -3.5 38 42 1343f 150 21.61 0.83 22.3 1.09 aat 68 3 37 14.3 -3.5 37 11 1166 f 1.50 ...... 21.6 1.18 aat 69 3 37 15.1 -3.5 33 46 1938 f 150 ...... 22.4 0.97 aat 71 3 37 21.5 -35 39 41 1843 f 1.50 ...... 22.4 1.06 ntt 201 3 36 44.2 -35 35 40 1061 f 13.5 21.17 1.24 ...... ntt 203 3 37 02.9 -3.5 34 40 994 f 073 20.69 1.08 ...... ntt 208 3 36 55.0 -3.5 31 31 127.5 f 091 20.91 1.12 ...... ntt 407 3 36 00.8 -35 35 02 2 107 & 159 20.19 1.01 ...... 16 ID RA( 1950) DEC( 1950) Uhelio I CI - r Bj Bj - R other L'hclio f0.02f0.035 f 2 0.2 f zz 0.3

ntt 410 3 36 17.7 -3.5 33 45 1190 f 094 19.83 1.27 ...... ntt 414 3 36 15.5 -35 32 40 1565 f 105 19.56 1.09 ...... ntt 101 3 36 55.4 -35 44 05 1270 f 118 ...... ntt 109 3 36 58.9 -35 41 46 1426 f 120 2k.24 1.27 ...... 1249 f 103 , aat 581801 f ntt 113 3 36 40.7 -3.5 40 46 1440 f 138 21.15 1.26 ...... ntt 119 3 37 01.7 -35 38 18 1327 f 121 21.16 1.19 ...... 1349 f 105, aat 631282 f 1 ntt 122 3 37 03.5 -35 37 43 1731 f 092 20.77 1.06 ...... ntt 123 3 36 54.1 -35 37 32 1307 f 164 20.93 1.00 ...... ntt 124 3 36 43.0 -35 37 14 1142 f 189 21.18 1.25 ...... ntt 125 3 36 41.0 -35 37 00 1772 f 142 21.02 1.17 ...... ntt 126 3 36 44.7 -35 36 52 723 f 207 20.76 1.22 ...... ntt 127 3 36 43.5 -35 36 32 1811 f 095 21.06 1.16 ......

keck 1 3 36 13.8-35 3924.8 732 f 032 ...... 21.8 ... keck 2 3 36 14.2 -3.5 38 51.2 1094 f 034 ...... 22.4 1.19 keck 3 3 36 09.4 -3.5 37 32.4 1571 f 031 ...... 22.3 1.02 keck 5 3 36 13.2 -35 37 37.8 1775 f 066 ...... 21.8 1.17 keck 6 3 36 17.8 -35 37 50.2 1386 f 031 ...... 22.3 keck 7 3 36 16.7 -35 37 01.7 1448 f 103 21.01 1.23 21.7 1.31 1376 f 84,aat 28 1677 f 13 keck 9 3 36 19.2 -3.5 36 28.7 1155 f 042 21.04 1.25 21.8 1.33 1150 f 31,aat 29 1280 f 1.5 keck 10 336 21.5-35 36 04.4 843 f 045 20.55 1.05 21.41.50 815 f 30, ntt 406917 f 55, keck 11 3 36 25.0 -3.5 36 28.3 1338 f 033 21.34 1.17 22.2 1.50 keck 12 3 36 20.3 -3.53.5 15.3 1734 f 042 21.97 0.94 22.4 1.061736 f 31, aat 231701 f 1.5 keck 13 3 36 23.4 -3.53.5 37.2 1247 f 030 21.51 0.91 22.2 ..' keck 14 3 36 24.5 -35 3.i 36.8 1260 f 066 21.17 1.37 22.1 1.6.5 keck 15 3 36 23.2 -35 34 39.3 1523 f 030 21.26 1.04 22.0 1.26 keck 17 3 36 26.3 -35 34 20.7 866 f 071 21.55 1.14 22.4 1.33 keck 18 3 36 31.4 -35 35 05.9 1688 f 042 21.32 1.08 22.31.50 keck19 3 36 34.5 -35 34 52.2 1150 f 059 21.41 1.29 22.3 '.. keck 20 3 36 28.5 -3.5 33 17.0 1374 f 126 21.63 1.13 22.30.88 keck 21 3 36 35.7 -35 34 24.6 1154 f 117 21.15 1.09 22.01.37 1194 f 98, aat 35 1062 f 1.5

The AAT data were taken from Srillmair 1992, the NTT data from Minniti et al. 1998, and the Keck data from Kissler-Patig et al. 1998. We used the original ID numbers, preceded with aat/ntt/keck respectively. The weighted mean velocity was computed when multiple measurements were available,the original measurements and cross references are given in the last column. C' and C' - I were taken from Iiissler-Patig et al. 1997a, Bj and Bj - I? I i were takenfromGrillmair 1992.

17 ‘I

TABLE2 PROPERTIESOF THE GALAXIES NGC 1399, NGC 1404, AND NGC 1380

NGC 1399 NGC 1404 NGC 1380 Reference

RA (1950) 03 36 34.0 03 36 57.0 03 34 31.0 RC3 DEC (1950) -35 36 42 -35 45 18 -35 08 24 RC3 1 236.71 236.95 235.93 RC3 b -53.64 -53.56 -54.06 RC3 Type (T) -5.0 -5.0 -2.0 RC3

1447 f 12 1929 f 14 1841 f 15 RC3 31.35 f 0.16 31.35 f 0.16 31.35f 0.16 Della Valle et al. 1998“

8.48 st 0.08’ 10.00 f 0.13 9.93 f 0.10 Ostrov et al. 1998, RC3 -22.87 f 0.18 -21.3.5 f 0.20 -21.42 f 0..19 using (rn - M)above 0.96 f 0.01 0.97 Zt 0.01 0.94 f 0.01 RC3 1.23 f 0.005 1.23 f 0.008 1.18 f 0.004 Buta & Williams 1995 44.7“ 28.5” ... Goudfrooij et al. 1994

5800 f 300 800 f 100 560 f 30 KP97a,F97,R92,KP97b 4.1 f 0.6 2.3 f 0.3 1.5 f 0.2 derived from this table 6.0 f 1.2 ... Ostrov et al. 1998’ 11 f 1‘ .. ...

RC3: de Vaucouleurs et al. (1991); IiP97a,b: Kissler-Patig et al. 1997a,b; F97: Forbes et al. 1998; R92: Richtler et al. 1992 aThe distance modulus was derived for NGC 1380 and is assumed to be the same for NGC 1399 and NGC 1404. It corresponds to a distance of 18.6 Mpc bNote that for NGC 1399, this includes the cD envelope. A rough estimate of the “galaxy” component can be obtained using the older value extrapolated from a de Vaucouleurs fit to the inner regions: = 9.59 f 0.08 and ilfv = -21.76 f 0.19, leading to Siv = 11 f 1 if all globular clusters were associated with the galaxy component. ‘Using our assumed distance modulus, computed for a annulus with 150”< radius < 240”.

18 TABLE3 bIEAN VELOCITY AND VELOCITY DISPERSION FOR VARIOUS SAMPLES

Mean velocity StandardDeviation Velocity dispersion(meanfixed)Comment

1518 f 91 388 f 54 ... AAT sample a 1353 f 79 338 f 56 NTT sample 1293 f 71 302 f 51 Keck sample

1429 f 45 373 f 35 ... Full sample

1484 f 128 256 f 86 263 f 92 6 globular clusters within 2’ 1393 f 84 355 f 63 357 f 64 20 globular clusters within 3’ 1378 f 64 371 f 46 375 f 47 41 globular clusters within 5’ 1498 f 68 362 f 52 368 f 54 33 globular clusters outside 5’ 1421 f 83 371 f 63 372 f 63 23 globular clusters outside 6’ 1515 f 134 399 f 101 408 & 107 10 globular clusters outside 8’

1611 f 8i 313 f 69 362 f 104 16 blue globular clusters (V-I< 1.05) 1322 zt 58 323 f 4.5 341 f 51 36 red globular clusters (V-I> 1.05) 1356 f 90 274 f 71 284 f 77 11 red globular clusters outside 5 arcmin 1643 f 154 381 zt 117 437 f 174 7 blue globular clusters outside 5 arcmin

Taken from a Grillmairet al. (1994), Minniti et al. (1997), Kissler-Patig et al. (1998) The dispersion was calculated around the fixed mean of the full sample (1429 km s-l)

19 TABLE4 PEAK MAGNITUDES OF THE GLOBULAR CLUSTER LUMINOSITY FUNCTION

~ ~~~~~ Galaxy rnv (peak) Reference NGC 140424.1 f0.2 Richtler et al. 1992 23.92 f0.20 Blakeslee si Tonry 1996 24.01 f0.20" Grillmair et al. 1999 NGC 138024.05 f0.25' Blakeslee & Tonry 1996 23.68 fO.11 Della Valle et al. 1998 NGC 139923.90 f0.09 Kohle et al. 1996 23.83 f0.15 Blakeslee & Tonry 1996 23.85 f0.30 Bridges et al. 1991 23.45 f0.16' Geisler & Forte 1990 24.0 f0.2 Madjesky & Bender 1990 23.73 f0.08" Grillmair et al. 1999 23.71 f0.12d Ostrov et al. 1998

" Measured on V data, obtained by transforming B measurements into V using (B- I) colors. But see com- ments in Della Valle et al. 1998. Converted by Geisler si Forte using V - TI = 0.45,Converted from m~,(peak) assuming (V - T~)o= 0.47.

20 TABLE5 TOTALNUMBER OF GLOBULARCLUSTERS IN FORNAXGALAXIES

~~~ ~~~ ~~ ~ ~ Galaxypresent NGC initial NGC # gained

Initial SN = 3.2 NGC 1399 5800 4527 1273 NGC 1380 560 1195 -634 NGC 1404 800 1113 -313 NGC 1427 510 510 0 NGC 1379 310 472 -162 NGC 1387 390 594 -204 NGC 1374 410 410 0 total 8780 8821 -4 1

Assumed distance to Fornax: (m- M)= 31.35. Galaxy luminosities taken from the RC3 to compute the S,V. The present number of globular clusters are taken from Table 2 and Kissler-Patig et al. 1997a. Column 3 shows what must have been the initial number of globular clusters for the as- sumed initial s,~,Column 4 shows the difference between the initial nunlber and the number of globular clusters presently observed.

21 750

500

250

a

-250

-500

-750 -500 -250 0 250 500 750 RA [arcsec]

Fig. 1.- Positions of the 74 globular clusters with respect to NGC 1399 and NGC 1404. For both galasies we indicate 1 and 5 reR. The symbols represent the globular cluster velocities: open symbols are approaching. filled symbols receding, with the size proportional to the difference between the globular cluster velocity and the mean systemic velocity of NCX 1399.

22 L L L nU 1 0 500 1000 1500 2000 2500 3000 velocity [km/s]

Fig. 2.- Histogram of 74 globular cluster velocity distribution. The long dashed line indicates the systemicvelocity of NGC 1399, the short dashed lines the velocities of NGC 1380 and NGC 1404. The relative contributions of the AAT sample (crossed regions), the NTT sample (narrow hatched regions) and the Iieck sample (hatched regions) are also shown. 3000

2500

2000 n v) \ E

U 1500 3 .4 0 0 4 Q) 3 1000

500

0

Fig. 3.- The radial velocity of the 74 globular clusters is plotted against their radius from the center of NGC 1399. The long dashed line indicates the systemic velocity of NGC 1399, the short dashed lines the velocities of NGC 1380 and NGC 1404. 10

VI 3 s5 0 0

0 0. 500 1000 15002500 2000 3000 velocity [km/s]

10

(II 3 5 ‘5 0 0

0 0 500 1000 1500 2000 2500 3000 velocity [km/s]

10

v) 3 55 0 0

0 0 500 1000 1500 2000 2500 3000 velocity [km/s]

Fig. 4.- Same as Fig. 2, except that the full sample was divided into three radial bins: 0’-3’ (top), 3’4’(middle), and 6”-11’ (lower panel), and no distinction was made between the AAT, NTT, and Keck samples.

25 2500

2000

1000

500

I1 IIIII II I1 IIIIIIOI -180 -135 -90 -45 0 45 90 135 180 P.A. [degrees]

Fig. 5.- The velocities of the 33 globular clusters at a distance greater than 5' from NGC 1399 are plotted against their position angle. The solid line shows the best fitted rotation (amplitude: 153593 km s-l, PA:120&40 degrees the major axis of the NGC 1399 isophotes). The long dashed line shows the systemic velocity of NSC 1399.

26 I

I I I I I Ill I I I I I Ill I I I I I Ill I I I I1Ill I 11 1 10 100 1000 104 radius[arcsec]

Fig. 6.- Velocity dispersion versus radius for various components. Triangles show the velocity dispersion of the stellar light taken from Franx et al. (1989, solid symbols), Bicknell et al. (1989, open symbols), and Winsall & Freeman(1993, diamond); the stars show the velocity dispersion of planetary nebulae at two radiitaken from Arnaboldi et al. (1994); the filled circles mark the velocity dispersion of our different radial sub-samples, the open circles mark the velocity dispersion of the red and blue sub-samples, the pentagon marks the velocity dispersion of the entire sample (data from Table 3); the squares show the velocity dispersion of Fornax galaxies, taken from Hartog 9i Katgert(1996, filled symbols) , and Ferguson(1989, open symbols).The solidline represents the LOWESS fit to the globular cluster and galaxy data. The dashed line shows the velocity dispersion profile derived from the temperature of the X-ray gas (Jones et al. 1997).