A Homogenous Set of Globular Cluster Relative Distances and Reddenings

a r X i v :a s t r o -p h /9608082v 1 13 A u g 1996A&A manuscript no.(will be inserted by hand later)Send offprint requests to :M.Kissler–Patig (Bonn)⋆Based on data collected at the Las Campanas Observatory,Chile,run by the Carnegie InstitutionsKey words:globular cluster systems –globular clus-ters –elliptical galaxies –galaxies:individual:NGC 1399,NGC 1374,NGC 1379,NGC 1387,NGC 1427–galax-ies:clusters:individual:Fornax2Globular cluster systems in FornaxTable1.General data of our target galaxies,all members of the Fornax galaxy cluster,taken from Tully(1988),Poulain(1988), and Poulain&Nieto(1994)NGC1374033516-351335236.36-54.29E111.20 1.201105NGC1379033303-352626236.72-54.13E011.20 1.191239NGC1387033657-353023236.82-53.95S010.81 1.301091NGC1399033829-352658236.71-53.64E09.27 1.251294NGC1427034219-352336236.60-52.85E311.04 1.151416Galaxy Filter Obs.date Exposure time seeing not be considered in the following:several I exposures of the SWfield were corrupted,while the SEfield includes part of NGC1404(another member of Fornax),which makes the allocation of globular clusters to the one or the other galaxy confusing.2.2.The reductionAll reductions were done in IRAF.Bias frames were sub-tracted and skyflat–fields of different nights were aver-aged toflatten the images better than1%.The different long exposures were then combined with a sigma clipping algorithm,to remove the cosmetics from thefinal frames.Object search,photometry,and the determination of the completeness factors were done with the DAOPHOT II version in IRAF.For all galaxies we computed an isophotal model in eachfilter(using the STSDAS pack-age isophote)that we subtracted from ourfinal long ex-posure to obtain aflat background for the object search and photometry.The nights of the26,27and29were photometric and the calibration was done via typically15-30standard stars from the Landolt(1992)list,taken throughout the nights, by which our Bessell V colors were transformed to John-son V.In the middle of the night of the28th cirrus passed. Frames taken at that time were calibrated via aperture photometry on the galaxy published by Poulain(1988), and Poulain&Nieto(1994)and cross–checked with over-lapping frames in the case of NGC1399.All other calibra-tions were also inter–compared and found compatible with the aperture photometry values for the individual galax-ies.Table3shows ourfinal coefficients for the calibration equations:V inst=V+v1+v2·X V+v3·(V−I)I inst=I+i1+i2·X I+i3·(V−I)where instrumental magnitudes are normalized to1sec-ond and given with an offset of25mag.The completeness calculations were done by standard artificial star experiments.We added typically10000stars over many runs on a long exposure and repeated the re-duction steps starting with the objectfinding.The com-pleteness values are given in detail in Fig.1of Paper I. The completeness limit of50%is reached at V≃23.5 mag,I≃22.5mag for our four normal galaxies,0.5magGlobular cluster systems in Fornax3 Table3.Calibration coefficients for our different nights.Col-umn5lists the RMS of the difference between our standardmagnitudes and our calibrated ones26.9.1.718±.0180.100±.012−0.019±.0070.02027.9.1.671±.0090.133fixed−0.020±.0090.02428.9.1.713±.0180.101±.012−0.018±.0070.02029.9.1.668±.0090.133fixed−0.020±.0090.02426.9.2.077±.0120.038±.008−0.013±.0050.01227.9.2.065±.0050.047fixed−0.017±.0050.01128.9.2.069±.0160.041±.010−0.006±.0070.01329.9.2.066±.0070.047fixed−0.015±.0060.0154Globular cluster systems in FornaxTable4.Number of globular clusters around our target galaxies.Column1lists the name of the galaxy,column2the background corrected counts down to the turn–over of the GCLF together with the errors from the background correction and the error in the assumed turn–over.The measured turn–over value of the GCLF is shown in column3,column4and5list the uncovered area towards the center of the galaxies and the correction,column6lists the correction for the area beyond120′′from the center,column7and8give the total amount of globular clusters within120′′and around the galaxycounts from the GCLF in Vcounts from the GCLF in ITable5.Specific frequency for our target galaxies.Column1lists the name of the galaxy,column2the derived distancemodulus,column3the absolute magnitude in V,column4the total number of globular clusters N t,column5the specificfrequency S NNGC137431.0±.2−19.8±.2410±824.9±1.3NGC137931.1±.2−19.9±.2314±633.4±0.9NGC138731.0±.2−20.2±.2389±1103.2±1.1NGC142731.0±.2−20.0±.2510±875.1±1.3NGC139931.0±.2−21.7±.25940±57012.4±3.0Globular cluster systems in Fornax5Fig.2.Histogram of the globular cluster color distribution in (V −I )around NGC 1374,NGC 1379,NGC 1387,and NGC 1427.The solid line is the gaussian fit to the distribution,the dotted one shows the broadening from the errors in (V −I )alonethe intrinsic dispersion of the distributions must be less than 0.1mag.We performed the KMM test proposed by Ashman et al.(1994),as well as fits with multiple gaus-sians,in order to detect multi–modality in the distribu-tions,but in no case the hypothesis of an unimodal distri-bution could be rejected.While the width of the distribution is almost identical for these four galaxies,the median color slightly differs.We assumed E (B −V )=0.0towards Fornax (Burstein &Heiles 1982),and derived median (V −I )colors of the globular clusters shown in Table 6,together with a median metallicity.Deriving metallicities from broadband colors is com-plicated by second parameter effects:age and metallicity can hardly be disentangled.For similar ages and metallic-ities,the V −I color of galactic globular clusters spread over about 0.2mag.From the relatively red (V −I >0.8)mean colors,the quite narrow color distributions,the fact that we see no peculiarities in the luminosity functions of the globular clusters,and that there is no sign for any recent mergers,we are encouraged to assume that the globular clusters in these four galaxies are older then 10Gyr.With this assumption on age we can roughly convert our colors to metallicities using the empirical color–to–metallicity rela-tion found for the Milky Way globular clusters.We updated the color–metallicity relation given in Couture et al.(1990),using the Mc Master catalogue (Harris 1996)of Milky Way globular clusters.Figure 3shows the metallicity of all the galactic globular clusters with a reddening E (B −V )less than 0.4(60candidates)plotted against their (V −I )0color,that we corrected for extinction assuming E (V −I )=1.38×E (B −V )(Taylor 1986).The best linear fit gives the relation:(V −I )0=0.15(±0.02)[Fe/H]+1.13(±0.03)This is valid,strictly speaking,only in the range 0.7<(V −I )<1.1.A direct comparison to the colors of theTable 6.Median V −I colors of the globular clusters around NGC 1374,NGC 1379,NGC 1387,and NGC 1379;as well as derived median metallicitiesNGC 13741.10±0.03−0.2±0.3NGC 13791.17±0.030.3±0.3NGC 13871.20±0.060.5±0.4NGC14271.03±0.03−0.7±0.36Globular cluster systems in Fornax0.60.811.21.4-2.5-2-1.5-1-0.5Fig.3.Metallicity versus (V −I )color for the globular clusters in the Milky Way with E (B −V )<0.4.Dots are bulge clusters,circles are halo clusters,the line shows the best linearfitFig.4.Histogram over the (V −I )color of the globular clusters in the Milky Way.The dashed region marks the contribution of the halo clusters,the black ones that of the bulgeThis is comparable to the intrinsic dispersion of the color distributions of our galaxies,and corresponds to a range of metallicity from [F e/H ]=−2.4to 0.2dex.Thus,from the colors alone,accurate metallicities can-not be derived,but we can conclude that:–All our early-type galaxies have median colors of their globular clusters redder than that of the Milky Way.Assuming old globular cluster populations,this would mean average metallicities of the globular clusters ranging from slightly richer than the Milky Way halo (for NGC 1427)to about solar (for NGC 1379and NGC 1387).–There is no clear evidence for multiple populations,the range of metallicities for the globular clusters in eachgalaxy could span 2dex in [Fe/H]as in the Milky Way,but is concentrated around a median value.–The median color for globular clusters in all the galax-ies is bluer by about 0.1mag in (V −I )than the inte-grated galaxy light,as already noticed in all galaxies observed up to date,but all our galaxies possess also globular clusters as red as the galaxy itself.4.2.The globular cluster colors in NGC 1399For NGC 1399we considered all the globular clusters in the NE and NW fields obeying our selection criteria.Fig-ure 5shows the color distribution of the globular clusters with errors in V −I less than 0.1mag.Here the best gaus-sian fit returns a width of σ=0.22,three times as large as expected from the errors only.For the sample com-posed by all globular clusters in the NE and NW field the KMM test rejects the unimodal hypothesis (with a confi-dence over 95%)if we impose dispersions of the color dis-tributions comparable to the dispersions observed in our normal galaxies (σ≃0.12),and favors two populations centered on V −I =0.99and V −I =1.18(correspond-ing to [Fe/H]=−0.9and 0.3dex according to our relation in the previous section).We note that the globular clus-ters do not span a much wider range of colors than in the other galaxies,but rather populate all colors more homo-geneously,i.e.they are no globular clusters with peculiar colors in NGC 1399compared to the other galaxies.From Washington photometry Ostrov et al.(1993)also derived multiple peaks in the color distribution.They find two groups,at [Fe/H]=−1.5and −0.9,as well as a pos-sible third group near −0.2dex.The shift in the color to metallicity conversion between our results demonstrates how difficult metallicity estimates are from broad band colors alone.The results from Washington photometry are probably more reliable than from the less metal sensitive V,I colors.The main result here is the confirmation of at least two groups of globular clusters in the color distribu-tion.This would suggest that two or more distinct globular cluster enrichment or formation epochs/mechanisms hap-pened in NGC 1399.4.3.A composite color distributionAs an experiment we constructed an artificial color dis-tribution with all globular clusters found around NGC 1374,NGC 1379,NGC 1387,and NGC 1427with errors in V −I <0.1mag.This is equivalent to the sum of the four histograms shown in Sect.4.1.The resulting composite color distribution is shown in Fig.6.We performed the same test on it as for the other distributions:a single gauss fit returns a width of σ=0.31;the KMM test favors a double gauss fit with 99%confi-dence if we impose individual width of σ=0.12.This color distribution would probably be classified as bi–modal,inGlobular cluster systems in Fornax7 Fig.5.Color distributions for the globular clusters in the NE(upper panel)and NW(lower panel)fields around NGC1399.The solid line is a free gaussianfit and returns a dispersionof0.22,while the dotted line shows the broadening expectedfrom our errors in photometryaloneposite histogram of the(V−I)colors of all theglobular clusters with a V−I error less than0.1mag in NGC1374,NGC1379,NGC1387,and NGC1427.The distributionseems to be bi–modal but is based on four systemscase of real data,while it is composed of four globularcluster systems.We take it as a word of caution,thatbroad and multi–modal color distribution might hide amuch more complex history than a single merger eventbetween two galaxies,as Ashman&Zepf(1992)proposein afirst approximation.4.4.Color gradientsWe present the radial color distributions of the globularclusters in Fig.7.The globular clusters plotted are thesame as used for the color distributions in the sectionsabove,i.e.objects that match our criteria for point likesources have an error in V−I of less than0.1mag,andare closer than120′′(425′′in the case of NGC1399)tothe center of the parent galaxy.Color gradients inglobu-Fig.7.Color gradients for ourfive target galaxies.Plots ex-tend to120′′for NGC1374,NGC1379,NGC1387,and NGC1427;to425′′for NGC1399.Errorbars show the dispersionaround the medianlar cluster systems is a topic of much debate.Where theyare found,such gradients are small,typically of the or-der of0.1to0.3mag in common broad band indices overhundreds of arcseconds radius(e.g.in M87,Lee&Geisler1993,in M49and NGC4649,Couture et al.1991,in NGC3923Zepf et al.1995,or in NGC3311Secker et al.1995).For most galaxies where such a gradient is found,anotherstudy exists quoting a non–detection.The most extensive data to date are probably fromGeisler et al.(1996),who recently re–examined the glob-ular cluster system of M49with deep Washington pho-tometry,and clearly showed a color gradient.However,the gradient is rather due to a different mixture of twopopulations in the inner and outer parts than to a steadyincrease of metallicity to the center as interpreted by au-thors in the past.The gradient in M49is of the order of0.4∆[Fe/H]/∆logR.This would translate into0.06∆(V−I)/∆logR according to our relation of Sect.4.1,or∆(V−I)=0.12to0.16mag over the range studied in our cases.We are therefore clearly not sensitive enough to detectsuch gradients in our data,and can only exclude gradients8Globular cluster systems in Fornaxas large a0.15∆(V−I)/∆logR(or1.0∆[F e/H]/∆logR) for our galaxies.One would need very good photome-try a couple of magnitudes deeper in a sensitive system (e.g.Washington or B and I Johnson–Cousins,Geisler et al.1996)tofind gradients,if present,in the globular clus-ter systems studied here.Bridges et al.(1991)found a decrease of0.2mag in B−V from1′to3′of the center in NGC1399.Over this range,the increase might also exists in our data,but is smaller than the scatter and can in no case be extrapolated further out.An interesting characteristic of the globular clusters in NGC1399might be noticed:the large dispersion of colors derived in Sect.4.2exists at all radii,as in the case of M49 (Geisler et al.1996).5.Spatial distributions5.1.The angular distributionsWe looked for any anisotropic distribution of globular clus-ters around NGC1374,NGC1379,NGC1387,and NGC 1427.No deep image centered on NGC1399was avail-able.We computed the counts in the same rings as for the GCLFs(i.e.excluding the centers,and out to120′′). We devided the ring in16×22.5degrees segments around NGC1374,NGC1379,NGC1387,and NGC1427,and plotted the distributions moduloπ,(i.e.rotating the west-ern side by180degrees around the center to increase a possible excess along a given axis)in Fig.8.For NGC 1374and NGC1427(E1and E3galaxies respectively), we indicated the position angle of the galaxies with dot-ted lines.The amount of background contamination in a segment is shown as a solid line.All the distributions are compatible with the globular clusters being spherically distributed around the galaxy. In NGC1374a2σexcess of objects along the major axis of the galaxy is present.For NGC1427it can be ex-cluded that the globular cluster system is as elliptical as the galaxy.For NGC1379(E0),the distribution with an ellipticity of0.2±0.1and a position angle of70±10de-greesfits the data equally well as a spherical distribution.5.2.The radial distributions5.2.1.The“normal”galaxiesFor NGC1374,NGC1379,NGC1387,and NGC1427 we computed the surface density profile for all objects found around the galaxy down to V=24.0mag without any correction for completeness.Table7shows the densi-ties computed in increasing elliptical rings22.7′′(100pix) wide.The density profiles are plotted in Fig.9,the up-per panel showing the uncorrected distribution,the lower panel showing the distribution corrected for background contamination(see Sect.5.3)together with thearbitrarily Fig.8.The angular distributions of globular clusters around NGC1374,NGC1379,NGC1387,and NGC1427.The globu-lar clusters were counted in22.5degrees wide segments around the galaxy and taken moduloπshifted galaxy light profile(squares).The excess in NGC 1374at about150′′corresponds to the distance of NGC 1375,the smaller galaxy close in projection to NGC1374, which might contribute a few globular clusters.Table7.Density profiles for the normal galaxies.Column1 shows the semi–major axis at the center of the ring in arcsecs, columns2–5the density of objects per square arcmin for our four galaxies57′′45.4±6.137.9±5.333.4±5.059.4±8.0 80′′19.8±3.121.4±3.115.2±2.621.6±3.7 102′′12.4±2.112.1±2.010.2±1.817.3±2.8 125′′7.4±1.46.9±1.35.2±1.113.8±2.2 147′′10.6±1.55.7±1.15.3±1.07.8±1.5 170′′4.0±0.97.0±1.23.6±0.810.1±1.7 193′′6.4±1.04.1±0.811.4±1.9 216′′5.4±0.94.0±0.89.4±1.7 238′′3.5±0.76.2±1.06.2±1.4 261′′4.0±1.14.4±1.16.6±1.4Globular cluster systems in Fornax9 Fig.9.The radial distribution of objects in NGC1374,NGC1379,NGC1387,and NGC1427.The left and right panels respectively show the uncorrected density profile and the density profile corrected for background contamination together withthe arbitrarily shifted galaxy light profile(squares).Surface densities are given in number per square arcminutecomputed the fraction of the ring seen in thefield,andscaled the counts up to the total area.No correction forcompleteness of the counts was necessary,since for NGC1399the counts were almost complete down to the consid-ered magnitude.The background contamination was de-termined with a backgroundfield located about40arcmineast.The backgroundfield needed a small(<5%)correc-tion for completeness to be adjusted to thefields aroundNGC1399.The result gave361objects on59.0square ar-cmin in the backgroundfield down to V=24mag,or abackground density of6.1±0.3objects per square arcminas background value for the counts in bothfields aroundNGC1399.Table8shows the densities computed in in-10Globular cluster systems in Fornaxcreasing elliptical rings22.7′′(100pix)wide,plotted inFig.10.Table8.The density profile of globular clusters around NGC1399.Column one list the mean ring radii,column2and3show the density of objects found in the NE and NWfield57′′89.7±18.783.4±8.680′′49.7±10.134.5±5.8102′′43.8±7.940.2±5.7125′′38.5±6.443.1±5.4147′′32.0±5.326.8±4.0170′′24.5±4.225.2±3.6193′′20.6±3.619.7±3.0216′′14.2±2.814.0±2.4238′′22.3±3.312.4±2.2261′′15.3±2.611.4±2.0284′′16.0±2.513.5±2.1307′′12.1±2.110.2±1.8329′′12.2±2.011.5±1.8352′′10.7±1.811.3±1.8375′′11.2±1.88.1±1.4397′′13.2±1.98.2±1.4420′′9.3±1.68.0±1.4443′′8.8±1.58.8±1.4465′′9.7±1.58.4±1.7488′′6.7±1.36.0±1.7511′′6.5±1.59.5±2.4534′′5.9±1.66.9±2.3Galaxy name density slope gal.light slope bkg density 556′′4.8±1.68.6±3.0light profile due to the large cD envelope of the galaxy(e.g.Schombert1986).For the four normal galaxies,the values found are verysimilar to results of previous studies of globular clustersystems around normal early-type galaxies(e.g.Kissler–Patig et al.1996and references therein).6.DiscussionThe previous sections demonstrate that while the globularcluster systems of our faint early-type galaxies have verysimilar properties,the globular clusters in NGC1399,thecentral giant elliptical cD galaxy,are much more numerousand have a different color distribution as well as aflatterdensity profile.While it is true for NGC1399that globular clustersappear in a much larger number than in spirals,it is notfor our fainter galaxies.Harris&Harris(1996)compiledGlobular cluster systems in Fornax11all the globular cluster systems investigated to date.If we select from their list all the S0,Sa,and Sb galaxies (excluding the two outstanding galaxies with M V<−22), we get for the ten remaining galaxies an average number of globular clusters of345±185per galaxy,for an average luminosity of M V=−21.1.The three ellipticals and the S0galaxy that we investigated here have a mean of406±81 globular clusters and therefore do not have more globular clusters in absolute numbers than do these spirals.Comparing the specific frequencies of spirals to that of ellipticals is very difficult,if it makes sense at all.First because ideally it should relate the number of globular clusters to the mass of the galaxy by assuming a con-stant M/L ratio,which is a reasonable assumption when comparing ellipticals among each other but not when com-paring spirals with ellipticals.Second because even when reducing this discrepancy by normalizing the number of globular clusters to the spheroid luminosity,it is unclear which fraction of the globular clusters in spirals are associ-ated with the halo and the bulge,while elliptical galaxies are most probably bulge dominated.Thus it is not clear if we compare comparable values.However,as a comparison, the sample of spirals mentioned above has an average S of 1.3±0.8,and would S be computed for the spheroid lumi-nosities,it would increase by about1(e.g.Harris1991). The value for spirals does therefore not deviate that much from the values derived in Sect.3.2.Similarity seems to exist further in the color distribu-tion of the globular clusters in our faint galaxies and in spirals.They are slightly redder(i.e.probably more metal-rich),but show a similar dispersion around the median to that in the Milky Way,and cover a similar range of colors. Here again dominating bulge clusters,in contrast to the halo dominated Milky Way system,could possibly explain the small color differences.Finally we conclude that for our faint elliptical galax-ies there is no strong need to a different globular cluster formation or evolution scenario,as well as no need for any increase of the number of globular clusters during a hypothetic merger event.On the contrary,for NGC1399these conclusions are not true.NGC1399has far more globular clusters,and a much higher specific frequency than the spiral galaxies. The surface density profile is muchflatter and the globu-lar clusters cover rather homogeneously the full range of colors,and show signs of several populations.As pointed out by several authors before,the formation of the glob-ular cluster system in NGC1399must have undergone a different history,similar to other central giant ellipticals (e.g.Harris1991).Note that the globular cluster system of NGC1399confirms all the predictions that Ashman& Zepf(1992)made for a globular cluster system that ex-perienced a merger:it has a broad(multi–modal?)color distribution,aflat surface density profile,and an increased number of globular clusters.However,NGC1399is one of the galaxies with an outstanding specific frequency,even for a possible enrichment by a merger.While the forma-tion of a large number of globular clusters in coolingflows seems to be ruled out(Bridges et al.1996),it was specu-lated that NGC1399’s position in the center of the Fornax cluster favored the huge number of globular clusters also observed in other galaxies lying at the center of galaxy clusters(e.g.Harris1991).One possibility would be the increased number of merger events at early times,since we showed that the multi–modal color distribution does not exclude several components to have formed the globu-lar cluster system of NGC1399.Another hypothesis could be that the large number of globular clusters is related to the large number of dwarf galaxies whose density Hilker et al.(1995)reported to increase significantly towards the center of the Fornax cluster.One could speculate that ac-creted while still gaseous,the dwarf galaxies formed with high efficiency globular clusters in the dense environment of NGC1399.The high specific frequency would then be a consequence of a Searle&Zinn(1978)scenario combined with the dense environment of NGC1399.However,no more than speculations could be made to date to explain the high specific frequencies of central galaxies.Finally we note the constancy of the specific frequen-cies in all our faint galaxies.The mean for our faint early–type galaxies in Fornax is4.2with a dispersion of1.0.We can add NGC1404,another probable member of Fornax from a study of Richtler et al.(1992,however note their possible argument against a membership of the galaxy to the cluster),and assume a similar distance modulus of 31.0±0.2.We then get a absolute magnitude of−21.0±0.2, and a specific frequency of3.5±0.8.The effectiveness in globular cluster formation within the normal galaxies of the Fornax galaxy cluster must have been very similar and might hint to similar formation histories of the galaxies in the cluster.Acknowledgements.We wish to thank the staffof the Las Cam-panas observatory for the friendly atmosphere and their valu-able help during the observing run.Thanks also to Bill Har-ris for providing a electronic copy of his globular cluster sys-tem compilation,and later for his comments as referee that helped to improve the paper.MKP aknowledges a studentfel-lowship at the European Southern Observatory,SK and MH were supported by the DFG project Ri418/5-1,LI would like to acknowledge support from Proyecto FONDECYT# 1960414.This research made use of the NASA/IPAC extra-galactic database(NED)which is operated by the Jet Propul-sion Laboratory,Caltech,under contract with the National Aeronautics and Space Administration.ReferencesAguilar L.,Hut P.,Ostriker J.,1988,ApJ335,720Ashman K.M.,&Zepf S.E.,1992,ApJ384,50Ashman K.M.,Bird C.M.,Zepf S.E.,1994,AJ108,2348 Bridges T.J.,Hanes D.A.,Harris W.E.,1991,AJ101,469 Bridges T.J.,Carter D.,Harris W.E.,Pritchet C.J.,1996,MN-RAS in press12Globular cluster systems in Fornax Burstein D.,Heiles C.,1982,AJ87,1165Conti P.S.,Vacca W.D.,1994ApJ423,L97Couture J,Harris W.E.,Allwright J.W.B.,1990,ApJS73,671Couture J,Harris W.E.,Allwright J.W.B.,1991,ApJ372,97Geisler D.,Forte J.C.,1990,ApJ350,L5Geisler D.,Lee M.G.,Kim E.,1996,AJ111,1529Goudfrooij P.,Hansen,L.,Jorgensen H.E.,et al.,1994,A&AS104,179Hanes D.A.,Harris W.E.,1986,ApJ309,564Harris W.E.,1986,AJ91,822Harris W.E.,1991,ARA&A29,543Harris W.E.,1995,in“Stellar Populations”,IAU Symp.164,eds.P.C.van der Kruit&G.Gilmore,Dodrecht:Kluwer,p.85Harris W.E.,1996,electronically published catalog,Mc MasterUniversityHarris H.C.,&Harris W.E.,1996,Astrophysical Quantities,4thed.,in pressHilker M.,Kissler-Patig M.,Richtler T.,Infante L.,1995,AGAbstr.Series11,222Hilker M.,Kissler–Patig M.,1996,A&A in pressHoltzman J.A.,Faber S.M.,Shaya E.J.,et al.,1992,AJ103,691Kissler–Patig M.,Richtler T.,Hilker M.,1996A&A308,704Kohle S.,Kissler–Patig M.,Hilker M.,et al.,1995,A&A309,L37(Paper I)Landolt A.U.,1992,AJ104,340Lee M.G.,Geisler,D.,1993,AJ106,493Lutz D.,1991,A&A245,31Ostrov P.,Geisler D.,Forte J.C.,1993,AJ105,1762Poulain P.,1988,A&AS72,215Poulain P.,Nieto J.-L.,1994,A&AS103,573Richtler T.,Grebel E.K.,Domg¨o rgen H.,Hilker M.,Kissler M.,1992,A&A264,25Richtler T.,1995,in“Reviews of Modern Astronomy”,Vol.8,eds.G.Klare,Springer,p.163Schombert J.M.,1986,ApJS60,603Schweizer F.,Seitzer P.,1993,ApJ417,L29Searle L.,Zinn R.,1978,ApJ225,357Secker J.,Geisler D.,McLaughlin D.E.,Harris W.E.,1995,AJ109,1019Taylor B.J.,1986,ApJS60,577Tully R.B.,1988,Nearby Galaxies Catalog,Cambridge Uni-versity PressVan den Bergh S.,1990,in Dynamics and Interactions ofGalaxies,ed.R.Wielen(Berlin:Springer),492Wagner S.,Richtler T.,Hopp U.,1991,A&A241,399Whitmore B.C.,Schweizer F.,Leitherer C.,Borne K.,RobertC.,1993,AJ106,1354Whitmore B.C.,Schweizer F.,1995,AJ109,960Zepf S.E.,Ashman K.M.,Geisler D.,1995,ApJ443,570This article was processed by the author using Springer-VerlagL a T E X A&A stylefile L-AA version3.。

a rXiv:as tr o-ph/111293v115Nov21X-rays at Sharp Focus:Chandra Science Symposium ASP Conference Series,Vol.**VOLUME**,2002eds.S.Vrtilek,E.M.Schlegel,L.Kuhi Stellar X-ray Binary Populations in Elliptical Galaxies Raymond E.White III Department of Physics and Astronomy,University of Alabama,Tuscaloosa,AL 35487-0324Abstract.Chandra ’s high angular resolution can resolve emission from stellar X-ray binaries out of the diffuse X-ray emission from gaseous atmo-spheres within elliptical galaxies.Variations in the X-ray binary popula-tions (per unit galaxian optical luminosity)are correlated with variations in the specific frequency of globular clusters in ellipticals.This indicates that X-ray binaries are largely formed in globular clusters,rather than being a primordial field population.1.Introduction The X-ray emission from normal elliptical galaxies has two major components:soft (0.2−1keV)emission from diffuse gas and harder (5−6keV)emission from populations of accreting (low-mass)stellar X-ray binaries (LMXB).The hardness of the LMXB component (placing its emission outside the most re-sponsive parts of the ROSAT and ASCA bandpasses)and its spatial confusion with the softer gaseous component have made it difficult to constrain the global temperatures and luminosities of LMXB populations in elliptical galaxies.Chan-dra observations are now resolving out individual LMXBs in nearby ellipticals (Sarazin,Irwin &Bregman 2000,2001;Kraft et al.2000;Finoguenov &Jones 2001),making their composite spectral analysis much easier.Figure 1compares optical images of two ellipticals,NGC 1407and NGC 4552,to their X-ray imagesfrom the ROSAT PSPC and Chandra ACIS (White &Davis 2001).The PSPC images emphasize the diffuse gaseous atmospheres of these ellipticals,while the ACIS images are stretched to emphasize the discrete sources in each galaxy.Chandra imaging has clearly resolved out dozens of LMXBs from the diffuse gaseous emission in these ellipticals,as it has in several other ellipticals in the work cited above.These recent Chandra observations show that the hard X-ray emission from normal ellipticals is dominated by LMXBs,not the advection-dominated accre-tion flows (ADAFs)onto massive,central black holes,advocated by Allen,di Mateo &Fabian (2000).Meanwhile,as Chandra continues to observe more nearby ellipticals,there is a large database of long ASCA observations which has more to yield for ellipticals.12Raymond E.White IIIparison of optical and X-ray images(5′square)ofthe elliptical galaxies NGC1407and NGC4552.The ROSAT PSPCimages(middle)emphasize X-ray emission from the diffuse gaseousatmospheres in each galaxy,while the Chandra ACIS images(right)are stretched to emphasize emission from discrete sources(LMXBs).2.Constraining the LMXB Component in EllipticalsStrong spectral constraints on the hard stellar LMXB component in ellipticals can be made by simultaneously analyzing ASCA spectra from multiple ellipti-cals.Most ellipticals require both soft(gaseous)and hard(LMXB)components.I simultaneouslyfit ASCA GIS spectra of six ellipticals which provided individ-ually reasonable spectral constraints on their hard emission.In the jointfits, the temperature of the LMXB component was assumed to be the same for all galaxies;the temperatures of any soft gaseous components(if present)were al-lowed to vary individually.The resulting best-fit global spectrum of LMXBs is fit equally well by a bremsstrahlung spectrum with kT=6.3(5.2-7.9)keV or a power-law spectrum with photon index=1.83(1.72-1.93),where90%confi-dence limits are in parentheses.These are the tightest constraints to date on the spectral properties of the stellar LMXB component in ellipticals(White2001) and are consistent with the spectral character of many individual LMXBs in our Galaxy.Fluxes for the LMXB components in an additional six ellipticals which had poorer photon statistics were determined by adopting the bestfit LMXB temperature of6.3keV infits to their GIS spectra.Stellar X-ray Binary Populations in Elliptical Galaxies3 In comparing the X-rayfluxes deduced for the LMXB component in these ellipticals to the total optical magnitudes for these galaxies,Ifind there is a factor of4range in the X-ray to opticalflux ratio.Although this range is much smaller than that of the softer gaseous component(which has a factor of100range in X-ray/opticalflux ratio),it is still larger than expected,since the LMXB component is supposed to be directly proportional to the stellar component.What is the source of the variance in the X-ray/opticalflux ratiof LMXB/f opt?3.Globular Cluster Population Variations in EllipticalsElliptical galaxies exhibit a wide range of globular cluster populations for a given galaxian luminosity.Furthermore,in our galaxy,LMXBs are produced much more efficiently in globular clusters than in thefield:globular clusters contain∼20%of the known LMXBs,yet globular clusters contain<0.1%of the stars in our galaxy.Apparently,globular clusters make LMXBs>200times more efficiently than the stellarfield(Katz1975),presumably through tidal interactions between stars(Clark1975).It is therefore conceivable that nearly ALL stellar LMXBs are formed in globular clusters(Grindlay1985;Kulkarni 2000).LMXBs which are not now in globular clusters may have been ejected from globulars by supernova kicks immediately after the primary collapsed to a neutron star.In this case,we might expect the X-ray luminosity of the LMXB component to be correlated with the globular cluster population of a galaxy, regardless of whether all LMXBs are currently resident in globular clusters.To test this,I plot in Figure2the specific frequency of globular clusters (the number of globular clusters per unit galaxy luminosity,normalized by the luminosity corresponding to a visual absolute magnitude of M V=−15),versus the X-ray/opticalflux ratio of the LMXB populations to the opticalflux(mag-nitude)of the host galaxies.The globular cluster data are from the compilation of Kissler-Patig(1997),while the X-ray data are from the ASCA elliptical sam-ple described above(White2001).There appears to be a strong correlation, with the X-ray/opticalflux ratio directly proportional(within the errors)to the specific globular cluster frequency S gc:.f LMXB/f opt∝S1.3±0.3gcThis strongly suggests that LMXB populations are indeed controlled by globular cluster populations.4.ConclusionsVariations in the X-ray binary populations of elliptical galaxies(per unit galax-ian optical luminosity)are linearly correlated with variations in their specific globular cluster frequencies.This indicates that X-ray binaries are largely formed in globular clusters(Grindlay1985),rather than being a primordial field population.We predict that Chandra observations of central dominant galaxies with unusually large globular cluster populations will be found to have proportionally large numbers of LMXBs,as well.For a more detailed analysis, see White,Kulkarni&Sarazin(2001).4Raymond E.White IIIFigure2.The specific frequency of globular clusters plotted againstthe ratio of global LMXB X-rayfluxes to the optical magnitudes(fluxes)of the host ellipticals.ReferencesAllen,S.W.,di Mateo,T.,&Fabian,A.C.2000,MNRAS,311,493Clark,G.W.1975,ApJ,199,L143Finoguenov,A.&Jones,C.2001,ApJ,547,L107Grindlay,J.E.1985,in The Evolution of Galactic X-ray Binaries,eds.J.Truem-per,W.H.G.Lewin&W.Brinkmann(Dordrecht:Reidel),p.25Katz,J.E.1975,Nature,253,698Kissler-Patig1997,AA,319,83Kraft,R.P.et al.2000,ApJ,531,L9Kulkarni,S.2000,private communicationSarazin,C.L.,Irwin,J.&Bregman,J.N.2000,ApJ,544,L101Sarazin,C.L.,Irwin,J.&Bregman,J.N.2001,ApJ,in press(astro-ph/0104070) White,R.E.III,2001,preprintWhite,R.E.III&Davis,D.S.2001,in preparationWhite,R.E.III,Kulkarni,S.,&Sarazin,C.L.2001,preprint。

a r X i v :a s t r o -p h /9710041v 1 3 O c t 1997HST Imaging of the Globular Clusters in the Fornax Cluster:NGC 1379Rebecca A.W.ElsonInstitute of Astronomy,Madingley Road,Cambridge CB30HA,UKElectronic mail:elson@ Carl J.Grillmair Jet Propulsion Laboratory,4800Oak Grove Drive,Pasadena,CA 91109USA Electronic mail:carl@ Duncan A.Forbes School of Physics and Astronomy,University of Birmingham,Edgbaston,Birmingham B152TT,UK Electronic mail:forbes@ Mike Rabban Lick Observatory,University of California,Santa Cruz,CA 95064USA Electronic mail:mrabban@ Gerard.M.Williger Goddard Space Flight Center,Greenbelt,MD 20771USA Electronic mail:williger@Jean P.BrodieLick Observatory,University of California,Santa Cruz,CA 95064USAElectronic mail:brodie@ReceivedABSTRACTWe present B and I photometry for∼300globular cluster candidates in NGC1379,anE0galaxy in the Fornax Cluster.Our data are from both Hubble Space Telescope(HST)and ground-based observations.The HST photometry(B only)is essentially complete and free of foreground/background contamination to∼2mag fainter than the peak of the globular cluster luminosity function.Fitting a Gaussian to the luminosity function wefind B =24.95±0.30 andσB=1.55±0.21.We estimate the total number of globular clusters to be436±30.To a radius of70arcsec we derive a moderate specific frequency,S N=3.5±0.4.At radii r∼3−6 kpc the surface density profile of the globular cluster system is indistinguishable from that of the underlying galaxy light.At r∼<2.5kpc the profile of the globular cluster systemflattens, and at r∼<1kpc,the number density appears to decrease.The(B−I)colour distribution of the globular clusters(from ground-based data)is similar to that for Milky Way globulars,once corrected for background contamination.It shows no evidence for bimodality or for the presence of a population with[Fe/H]∼>−0.5.Unlike in the case of larger,centrally located cluster ellipticals,neither mergers nor a multiphase collapse are required to explain the formation of the NGC1379globular cluster system.We stress the importance of corrections for background contamination in ground-based samples of this kind:the area covered by a globular cluster system(with radius∼30kpc)at the distance of the Virgo or Fornax cluster contains∼>200background galaxies unresolved from the ground,with magnitudes comparable to brighter globular clusters at that distance.The colour distribution of these galaxies is strongly peaked slightly bluer than the peak of a typical globular cluster distribution.Such contamination can thus create the impression of skewed colour distributions,or even of bimodality,where none exists.Key words:galaxies:individual:NGC1379-globular clusters:general-galaxies:star clusters1.IntroductionAn accumulating body of observations suggests that the distribution of colours of globular clusters,and by inference of their metallicities,varies significantly from one galaxy to the next.Of particular interest is the colour bimodality which has now been observed in the globular cluster systems of several large elliptical galaxies,and suggests the presence of distinct metal rich and metal poor populations.The best example is the Virgo cD galaxy M87(NGC4486)(cf.Elson&Santiago1996).It has one population of globular clusters with colours similar to those of the Milky Way globulars,and one which is significantly redder,with inferred metallicities∼>solar.Other galaxies whose globular cluster systems show clear bimodality include M49(NGC4472),an E2galaxy in the Virgo Cluster with the same luminosity as M87(Geisler,Lee&Kim 1996),and NGC5846,a slightly less luminous E0galaxy at the centre of a small compact group(Forbes, Brodie&Huchra1997a).Two ideas have been invoked to explain the colour bimodalities.One is that elliptical galaxies are formed during mergers in which populations of metal rich(red)globulars are created and added to a‘native’population of metal poor(blue)clusters(cf.Ashman&Zepf1992).The other is that globular cluster populations with different mean metallicities form during a multiphase collapse of a single system(Forbes, Brodie&Grillmair1997b):metal poor globular clusters are formed early in the collapse,while metal rich ones form later,roughly contemporaneously with the stars.To understand fully the implications of the observed colour distributions for the origin of globular cluster systems,a much larger body of accurate data for systems surrounding galaxies of a variety of types and in a variety of environments is required.A question of particular importance,for example,is whether all elliptical galaxies have bimodal globular cluster systems,or whether such systems are restricted to large galaxies in rich environments.The data best suited to address this question are those acquired with the Hubble Space Telescope(HST).The resolution of HST allows even the crowded central regions of galaxies at the distance of Fornax and Virgo to be probed and allows most background galaxies to be eliminated. At these distances samples of globular clusters are complete and uncontaminated to well past the peak of the luminosity function.This paper and the others in this series are contributions to this growing database. Forbes et al.(1997a)and Grillmair et al.(1997)discuss the globular cluster systems of the Fornax cD galaxy NGC1399,its neighbour the E1galaxy NGC1404,and the peculiar galaxy NGC1316which may have undergone a recent merger.Here we present observations of the globular cluster system of NGC1379, a normal E0galaxy in the Fornax cluster.Hanes&Harris(1986)used photographic data to study the NGC1379globular cluster system toB=23.6(about a magnitude brighter than the peak of the luminosity function).They measured the profile of the outer part of the system(5<r<35kpc),and estimated the total population to number∼800. More recently the globular cluster systems offive galaxies in the Fornax cluster,including NGC1379,have been studied by Kohle et al.(1996)and Kissler-Patig et al.(1997a)using V and I−band photometry obtained with the100-inch telescope at Las Campanas.Their data are50%complete at B∼24,and cover a radial range∼3−10kpc.Our observations,obtained both from the ground at the Cerro Tololo Interamerican Observatory (CTIO),and from space using HST,are described in Section2.Section3presents the results,including a caveat concerning the need for accurate background corrections for ground-based data.Ourfindings are summarized in Section4.2.Observations and Data ReductionIn this section we describe the HST and CTIO images upon which our results are based,and the process for detecting,selecting,and determining magnitudes for the globular cluster candidates in each case.We also discuss completeness,and contamination from foreground stars and background galaxies.At an adopted distance of18.4Mpc(m−M=31.32;Madore et al.1996),1arcsec corresponds to89pc. Reddening in the direction of the Fornax cluster is assumed to be negligible(Bender,Burstein,&Faber 1992).2.1.HST ImagingFive images of NGC1379were obtained with HST on1996March11,using the F450W(∼B)filter. (Due to technical difficulties,the complementary I-band images were not acquired,and are anticipated in 1997.)Three images were taken at one pointing,and two more were offset by0.5arcsec.The total exposure time was5000seconds.The centre of NGC1379was positioned at the centre of the Planetary Camera (PC)chip to afford the greatest resolution in the most crowded regions.Details of the reductions are given by Grillmair et al.(1997).Briefly,the images were reduced using the standard pipeline procedure.TheVISTA routine SNUC was used tofit and subtract the underlying galaxy.We ran DAOPHOT II/ALLSTAR (Stetson1987)separately on the sum of thefirst three images and the last two images,requiring that detections appear in both lists to qualify as real objects.We adopted a detection threshold of3σand measured magnitudes byfitting a point-spread function(PSF).Extended objects were eliminated by visual inspection.Count rates were converted to B magnitudes using the gain ratios and zeropoints given by Holtzman et al.(1995;1997,private communication).Photometry is available on request from CJG.Figure1shows a mosaic of the four WFPC2chips,with the galaxy subtracted.The total area of the field,excluding two60pixel wide unexposed borders on each chip,is4.8arcmin2.The scale of the Wide Field Camera(WFC)is0.0996arcsec pixel−1,and of the PC,0.0455arcsec pixel−1.At the distance of NGC1379,one WFC pixel corresponds to∼9pc and one PC pixel to∼4pc.A globular cluster with size typical of those in the Milky Way(core radius∼2pc,half-mass radius∼10pc,and tidal radius∼50pc) will thus appear essentially unresolved in our images.A total of∼300objects were detected and measured in ourfield.To determine the completeness of this sample,3000artificial PSFs(100at each of30magnitude levels)were added to the images,and the images were then processed in a manner identical to that for the original data.The completeness of the sample as a function of magnitude is shown in Figure2.The sample is∼100%complete to B=26for the WFC chips(80%of the sample),and to B=25.5for the PC chip.At B>26the completeness begins to drop rapidly.Photometric errors areδB≈0.10mag at B∼<25,rising to0.15mag at B=26.Next we consider the extent to which our sample may be contaminated by foreground stars and background galaxies.Few foreground stars are expected in an area of only4.8arcmin2at this Galactic latitude,and most background galaxies are resolved and thus easily distinguished from globular clusters. The main source of contamination is compact,spherical background galaxies.To determine the expected level of contamination in our sample,we observed a backgroundfield located∼1.4degrees south of the center of the Fornax cluster.The exposure time was5200seconds,so the limit of detection is comparable to that for the NGC1379sample.The image was processed and the sample selected in the same way as for the NGC1379field(see Grillmair et al.1997).Figure3shows a colour-magnitude diagram(CMD)for the84unresolved objects detected in the backgroundfield.The sample becomes incomplete at B>26.5,but as we shall see,this is∼1.5magnitudes fainter than the peak of the luminosity function,and so will not affect our results.At B∼<26.5the objects have a wide range of colours,with the majority concentrated around(B−I)∼1.The B luminosityfunction for the background sample is plotted in Fig.4,which also shows the luminosity function for the ∼300candidate globular clusters.The background luminosity function is tabulated in Table1.Since these background number counts are applicable to HST studies of any unresolved population at high latitude, we also include in Table1the I-band luminosity function for the backgroundfield.This is plotted as the solid histogram in Fig. 5.As a check on the consistency of the sample,and to assess the amplitude of spatialfluctuations in the background on this scale,we compared the luminosity function in Fig.5with the Medium Deep Survey(MDS)star count data from17high latitudefields obtained with HST in the V and I bands(Santiago,Gilmore&Elson1996).The dotted histogram in Fig.5shows the I-band luminosity function for the MDS data,normalized to an area of4.8arcmin2.The two distributions are in excellent agreement to I≈23.5which is the limiting magitude of the MDS star count data.Fainter than this stars can no longer be reliably distinguished from compact galaxies.The MDS stellar luminosity function in V, normalized to the same area,is also included in Table1.Finally,Fig.5also shows a luminosity function for stars and galaxies from the Canada-France Redshift Survey(CFRS)(Lilly et al.1995;S.Lilly1997,private communication).The magnitudes were measured in an aperture of diameter3arcsec.The I-band data cover an effective area of∼425arcmin2,and have again been normalized to an area of4.8arcmin2.These data illustrate the much larger degree of background contamination to be expected in the extreme case where no galaxies(with I∼>21)can be distinguished morphologically from unresolved objects.Background contamination in typical ground-based samples of globular clusters in distant galaxies will fall somewhere between the CFRS and the HST histograms.2.2.Ground–based ImagingBroadband B and I images of NGC1379were taken with the CTIO1.5m telescope,using a Tek2048 x2048array with a pixel scale of0.44arcsec pixel−1.Total exposure times were9000and3900seconds for the B and I images respectively.Although the seeing was only∼1.5arcsec,conditions remained photometric throughout the night of1995December24.Reduction was carried out in the standard way(i.e. bias and dark subtraction,flat–fielding and sky subtraction).After combining,the images were calibrated using aperture photometry from the catalogs of of Longo&de Vaucouleurs(1983)and de Vaucouleurs& Longo(1988).This procedure gave an rms precision of better than0.05mag.For the selection of globular cluster candidates in the CTIO images,we employed an iterative procedure using DAOPHOT II.We measured the background noise in both images and set the threshold for singlepixel detection at5σ,i.e.five times the noise due to the background.The other important detection parameters,SHARPness and ROUNDness(designed to weed out extended objects and cosmic rays),were initially given a large range.For each detected object we measured a3pixel radius aperture magnitude and applied an aperture correction based on a curve-of-growth analysis for a dozen isolated globular clusters. The rms error in the aperture correction is∼0.05mag.We then compared our B-band candidate list with the positions and B magnitudes of globular clusters detected with HST’s WFC.Our candidate list was matched to the HST list with the condition that the HST globular cluster lie within3CTIO pixels of our object.With this condition,28objects were matched. The average magnitude difference between the CTIO and HST B magnitude is0.03mag.Such excellent agreement is gratifying from a photometric standpoint,and reassures us that we have matched the data sets correctly.The matched clusters have SHARPness parameters0.3to0.7and ROUNDness−0.4to0.4. Assuming that these globular clusters are representative of all globular clusters in the CTIO image,we re–ran DAOPHOT II with the new restricted range in SHARPness and ROUNDness parameters.This resulted in the exclusion of about half the objects in the original sample,which are presumably background galaxies.The same parameters were used for both the B and I images.Figure6shows a CMD for365 objects selected in this way.A comparison of the CTIO object list and the HST list,within the area in common,will also indicate how complete our detection is as a function of magnitude.The completeness function estimated this way is shown in Fig.7.Our sample is∼100%complete at B∼21and∼50%complete at B∼23.5.The photometric errors are<0.1mag for all magnitudes brighter than our50%completeness limit.At this point it is necessary to investigate the level of contamination of our ground-based sample by foreground stars and background galaxies.Since thefield is much larger than that observed with HST(211 arcmin2compared to4.8arcmin2),and since the resolution is not sufficient to distinguish many background galaxies from unresolved sources,contamination from both stars and galaxies will be significant.Since no background comparisonfield was obtained,we rely instead on the CFRS data in the B and I-bands, discussed above,to estimate a correction for contamination.Figure8shows histograms of(B−I)for the full NGC1379sample in Fig.6,and for the CFRS sample of objects with19<B<23.Since selection effects in the two samples are different,we do not normalize the CFRS sample directly using the known area of the survey.Rather,we normalize it to match the tail of objects with(B−I)>2.5in Fig.8,whose colours are too red to be globular clusters.From this wecan infer the relative number of contaminants with(B−I)<2.5.The CFRS sample is strongly peaked at (B−I)∼1.0,which is just∼0.4mag bluer than the peak of the NGC1379sample.The NGC1379sample is lacking many of the bluest objects in the CFRS sample;this is probably because fainter galaxies are on average bluer,and the CFRS sample is much more complete than the NGC1379sample at the faintest magnitudes.Also,our process of excluding galaxies from the CTIO sample on the basis of ROUNDness would preferentially eliminate edge-on spirals,which are systematically bluer than ellipticals.The most important point illustrated by Fig.8is that in the ground-based data,much more so than in the HST data,failure to properly correct for foreground/background contamination may lead to a significant error in the deduced colour of the peak of the globular cluster colour distribution,and possibly to the erroneous impression of a bimodal colour distribution.With ground-based data obtained with better seeing it would of course be possible to exclude a greater proportion of background galaxies,so that any skewing of the colour distribution would be less severe.3.Properties of the NGC1379globular cluster systemThe principal properties of a globular cluster system which any theory of its origin and evolution must account for are the luminosity function(in particular,the absolute magnitude of its peak,and the width of the distribution),the colour distribution(in particular,the colour of the peak and the presence or absence of any bimodality),the radial distribution,and the total number of clusters,N tot,and therefore the specific frequency,defined as S N=N tot100.4(M V+15),where M V is the absolute magnitude of the galaxy.With B tot=12.07(Tully1988)and(B−V)=0.89(Faber et al.1989)we infer an apparent magnitude for NGC 1379of V tot=11.18.With distance modulus31.32,this implies M V=−20.16.This is the integrated magnitude within a radius of∼70arcsec.3.1.Luminosity functionThe luminosity function for the NGC1379globular cluster candidates derived from the HST datais shown in Fig. 4.This sample includes∼300unresolved objects and is essentially complete and uncontaminated well past the peak.The background-corrected luminosity function,that is,the differencebetween the solid and dashed histograms in Fig.4,is shown in Fig.9.Fitting a Gaussian function to the luminosity function in Fig.4,using a maximum-likelihood technique which is independent of binning, and takes into account completeness,background contamination,and photometric errors(Secker&Harris 1993),we derive a peak magnitude of B =24.95±0.30,and a widthσB=1.55±0.21.This Gaussian is shown as the solid curve in Fig.9.We know of no other direct measurements of B for this system to compare with ours.However,our values are the same,to within the errors,as the values measured for the globular cluster systems of two other Fornax galaxies,NGC1399and NGC1404(Grillmair et al.1997).Also,for four elliptical galaxies in the Virgo cluster Harris et al.(1991)find an average value B =24.77±bining this with the value(m−M)F ornax−(m−M)V irgo=0.08±0.09(Kohle et al.1996)implies B =24.85±0.22for NGC 1379.This is in good agreement with our value.We may either use the measured peak magnitude to infer a distance modulus for NGC1379,assuming a‘universal’value for the absolute magnitude of the peak,or we may adopt a distance modulus and infer an absolute magnitude.While the value of M V has been measured for many globular cluster systems, there are fewer B−band studies.Sandage&Tamman(1995)quote M B =−6.93±0.08for the Milky Way and M31globular clusters,while Ashman,Conti&Zepf(1995)give M B =−6.50for the Milky Way clusters.Adopting the Cepheid distance for NGC1379and our measured peak magnitude,implies M B =−6.37±0.36,which is in good agreement with the Ashman et al.value and somewhat fainter than the Sandage&Tamman value.As we shall see below,the(B−I)distribution of NGC1379globular clusters appears very similar to that for the Milky Way globular clusters.It should therefore be safe to assume that the(B−V)distribution is also similar.Adopting B−V =0.7for the Milky Way globulars(Harris1996),we may convert our value for the peak magnitude of the luminosity function, B =24.95±0.30,to V =24.25±0.30.This value is somewhat fainter than the value found by Kohle et al.(1996)of V =23.68±0.28,but their data reach just to the peak of the luminosity function,and the errors in their faintest bin are large.The sense of the difference is consistent with our result above thatfitting only the brighter part of the luminosity function results in a peak magnitude which is too bright.With the Cepheid distance modulus of31.32,our V value implies M V =−7.07±0.36,which is typical for elliptical galaxies of this absolute magnitude (Harris1991).Our value ofσB=1.55±0.21for the NGC1379globular cluster system is larger than the valuesσB=1.07and0.89quoted by Sandage&Tamman for the Milky Way and M31respectively,but is consistent with the valuesσB=1.37±0.07andσB=1.39±0.12found by Grillmair et al.(1997)for the globular cluster systems of NGC1399and NGC1404respectively.It is also consistent with the value σB=1.46±0.07for four elliptical galaxies in the Virgo cluster(Harris et al.1991).3.2.Radial profileHarris&Hanes(1987)compared the radial profile of the NGC1379globular cluster system with the surface brightness profile of the galaxy itself from Schombert(1986),over the radial range5−35kpc. They detected no difference between the two,although their uncertainties were large.Kissler-Patig et al. (1997a)found the same result from their ground-based data over the range3−10kpc.Radial surface density profiles for our HST sample of NGC1379globulars are shown in Figs.10a and b.Corrections have been made to compensate for the fraction of each annulus which falls outside thefield of view of the WFPC2.In Figure10a profiles are plotted for both the complete,uncontaminated sample with B<25.5, and for the objects with B>26.5(with no correction for completeness),which are expected to be primarily background galaxies.Indeed,the fainter sample shows almost no radial gradient.The dashed line is the background level(∼9.0objects per arcmin2)measured from the backgroundfield for B>26.5.The agreement is excellent.The background level for the brighter sample,again measured from the background field,is2.7objects per arcmin2,and is shown as the solid line.The difference between the radial profile and the background level is probably due to small numbers,but may indicate a small amount of residual contamination:thefive outermost points represent an average of<4objects each,and the total number of objects in the backgroundfield with B<25.5is just13.The radial profile of the brighter sample decreases smoothly from∼10−80arcsec(∼1−7kpc),at which point the globular cluster system is lost in the background.Figure10b shows the radial profile for the sample with B<25.5,with a background of2.7objects per arcmin2subtracted.Superposed is a curve representing the surface brightness profile of the underlying galaxy from Kissler-Patig et al.(1997a),scaled arbitrarily to match the profile of the globular cluster system.The two profiles agree well at∼35−70 arcsec(3–6kpc).There is no evidence that the surface brightness profile of the globular cluster system is shallower than that of the underlying galaxy light out at least to the limit of our data at r∼7kpc.The logarithmic slope of the profile in Fig.10b is−2.4at r∼>35arcsec.Inwards of r∼30arcsec(∼2.5kpc)the profile of the globular cluster systemflattens out.This core structure seems to be a common feature of the globular cluster systems of elliptical galaxies,and the radius at which theflattening occurs correlates with the galaxy luminosity(Forbes et al.1996).The mean surface density within∼10arcsec of the centre of NGC1379is∼200±60clusters per arcmin2.The core radius, where the surface density has fallen to half its central value,is r c∼23±6arcsec(2.0±0.5kpc).This is consistent with values for other galaxies with absolute magnitude comparable to that of NGC1379.Such a core structure is not present in the underlying galaxy light,which,while it changes slope slightly at r∼50 arcsec,rises with constant slope inwards to at least10arcsec(Schombert1986).Inwards of∼10arcsec(∼1kpc),the surface density of globular clusters appears to decrease.This radius corresponds to220pixels on the PC,and as can be seen from Fig.1,crowding even in these inner regions is not severe.Extrapolating a smoothly rising profile to the center of the galaxy would require6 clusters instead of2in the innermost bin,and14instead of12in the second bin.If the radial distribution really were steadily rising,this would imply that our data are only∼70%complete in these combined bins. Closer analysis of our completeness tests reveals that we are in fact97%complete for B<25.5in the region r<220pixels(the slight reduction over the PC-averaged completeness value being a consequence of the increased noise due to the integrated stellar light of the central regions of the galaxy).We conclude that the central dip in the cluster surface density distribution is not an artifact of our analysis,and may indeed indicate a quite substantial drop in the volume density of clusters near the nucleus of the galaxy.Radii less than1kpc from the nucleus are where we expect tidal stresses to begin to take a toll on the numbers of globular clusters(Lauer&Kormendy1986,Grillmair,Pritchet,&van den Bergh1986),either during their formation or subsequently through tidal stripping(Grillmair et al.1995),particularly if the clusters are on box orbits.We now turn to the ground-based data which cover a much greater area than the HST data.The radial distribution for the full sample from Fig.6is shown in Fig.11.The surface density for the CTIO sample has been increased by0.74in log N to match the HST sample,which is shown as the solid curve.At r∼>100arcsec,the profile of the CTIO sample is essentiallyflat,suggesting that the sample is composed overwhelmingly of foreground/background objects at these radii.Even a colour selection is unlikely to help disentangle the background contamination since,as shown in Fig.8,background galaxies have a similar colour distribution to the full NGC1379sample.We therefore conclude that,despite the larger spatial coverage of the ground-based sample,in the absence of a suitable background calibrationfield it does not contribute much to our knowledge of the radial structure of the NGC1379globular cluster system notcovered by our HST data.It is also unable to probe the innermost regions of the cluster system,due to crowding.3.3.Colour distributionIn the absence of an I-band image from HST,we attempt to extract a sample of globular clusters from the ground-based data which is as uncontaminated as possible,to investigate the(B−I)colour distribution.One approach is to select only objects with r<100arcsec,since these show a radial gradient in surface density(Fig.11).This gives a sub-sample of35objects.The background level inferred from the radial distribution of the CTIO sample at r>100arcsec is1.6arcmin−2,implying that14of the35 globular cluster candidates at r<100arcsec,or40%of this sample,are background galaxies or foreground stars.The colour distribution for the35objects at r<100arcsec,along with those for the full CTIO sample and for the CFRS sample,is shown as the dashed histogram in Fig.8.The r<100arcsec sample clearly peaks at a redder colour than the full sample,underlining the fact that the colour distribution of the uncorrected sample is misleading.Figure12shows the same histogram for the r<100arcsec sample,along with the histogram of residuals obtained by subtracting the normalized CFRS histogram from the full CTIO histogram in Fig.8.A histogram for95globular clusters in the Milky Way(Harris1996)is also shown.The colour distributions of the NGC1379sample with r<100arcsec,and with the CFRS sample subtracted,have a very similar peak colour,(B−I)≈1.6,which is indistinguishable from that for the globular cluster system of the Milky Way.Using the relation between(B−I)colour and metallicity from(Couture et al.1990)we infer from the peak colour a metallicity of[Fe/H]∼−1.5.Forbes et al.(1997b)plot a relation between mean metallicity of globular clusters and parent galaxy magnitude for11galaxies with−21<M V<23,with bimodal globular cluster colour distributions.Theyfind that,while in the metal rich populations([Fe/H]>−0.5)there is a strong correlation between mean globular cluster metallicity and galaxy magnitude,for the metal poor populations the scatter is much greater and the correlation much weaker.There is,however,a trend for less luminous galaxies to have globular clusters with a lower mean metallicity,and our results for NGC1379are consistent with this trend.There is a tail of bluer objects in the NGC1379samples which is probably comprised of residual。

a r X i v :a s t r o -p h /0006182v 2 15 J u n 2000A&A manuscript no.(will be inserted by hand later)ASTRONOMYANDASTROPHY SICSKey words:Stars:early-type –Stars:fundamental pa-rameters –Stars:horizontal-branch –globular clusters:individual:NGC 67521.IntroductionThe colour-magnitude diagrams of metal-poor globular clusters show a large variety of horizontal-branch (HB)morphologies,including “gaps”along the blue HB and long “blue tails”that extend towards higher effective tem-2S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixing process which dredges up Al will also dredge up he-lium.Possible dredge-up mechanisms include rotation-ally induced mixing(Sweigart&Mengel1979,Zahn1992,Charbonnel1995)and hydrogen shell instabil-ities(Von Rudloffet al.1988,Fujimoto et al.1999).Such dredge-up would increase the helium abundancein the red giant’s hydrogen envelope and thereby in-crease the luminosity(and the mass loss)along theRGB(Sweigart1997a,1997b).The progeny of thesestars on the horizontal branch would then have lessmassive hydrogen envelopes than unmixed stars.Asthe temperature of an HB star increases with decreas-ing mass of the hydrogen envelope,“mixed”HB starswould be hotter than their canonical counterparts.Thehelium enrichment would also lead to an increased hy-drogen burning rate and thus to higher luminosities(compared to canonical HB stars of the same tem-perature).The luminosities of stars hotter than about20,000K are not affected by this mixing process be-cause these stars have only inert hydrogen shells.Inthis framework the low gravities of hot HB stars wouldnecessarily be connected to abundance anomalies ob-served on the RGB,thereby explaining both of thesepuzzles at once.Radiative levitation of heavy elements:Caloi(1999)and Grundahl et al.(1999)suggested that thelow surface gravities of the HBB stars are related toa stellar atmospheres effect caused by the radiativelevitation of heavy elements.Such an enrichment in the metal abundance would change the temperature structure of the stellar atmosphere and thereby affect theflux distribution and the line profiles(Leone& Manfr`e1997).This scenario would also account for the fact that there is no evidence for deep mixing amongstfield red giants(e.g.Hanson et al.1998, Carretta et al.1999)even thoughfield HBB stars show the same low surface gravities as globular cluster stars(Saffer et al.1997,Mitchell et al.1998).Behr et al.(1999,2000b)have recently reported slightly super-solar iron abundances for HBB stars in M13 and M15,in agreement with the radiative levitation scenario.NGC6752is an ideal test case for these scenarios: Its distance modulus is very well determined from both white dwarfs(Renzini et al.1996)and HIPPARCOS par-allaxes(Reid1997),and thus any mass discrepancies can-not be explained by a wrong distance modulus.Spectro-scopic analyses of the faint blue stars in NGC6752showed them to be subdwarf B(sdB)stars.As mentioned above, their mean mass agrees well with the canonical value of 0.5M⊙.However,almost no stars in the sparsely popu-lated region above the sdB star region have been anal-ysed.If these stars show low surface gravities and canoni-cal masses,then the combination of deep mixing and the long distance scale(for the other globular clusters)would Fig.1.The colour-magnitude diagram of NGC6752(Buo-nanno et al.1986).Stars analysed in this paper(including 10stars discussed by Moehler et al.1997b)are marked by open squares(some of which overlap due to almost identical photometric data).resolve the discrepancies described above.If they show low surface gravities and low masses,diffusion may indeed play a rˆo le when analysing these stars for effective temperature and surface gravity.Then the low surface gravities found for HBB stars could be artifacts from the use of inap-propriate model atmospheres for the analyses.We there-fore decided to observe stars in this region of the colour-magnitude diagram and to derive their atmospheric pa-rameters.First results,which strongly support radiative levitation of heavy elements as the explanation,have been discussed by Moehler et al.(1999a).Here we describe the observations and their reductions,provide the detailed re-sults of the spectroscopic analyses(temperatures,surfaces gravities,helium and partly iron and magnesium abun-dances,masses),and discuss the consequences of ourfind-ings in more detail.2.ObservationsWe selected our targets from the photographic photom-etry of Buonanno et al.(1986,see Table1and Fig.1).S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixing3For our observations we used the ESO1.52m telescope with the Boller&Chivens spectrograph and CCD#39 (2048×2048pixels,(15µm)2pixel size,read-out noise 5.4e−,conversion factor1.2e−/count).We used grat-ing#33(65˚A mm−1)to cover a wavelength of3300˚A–5300˚bined with a slit width of2′′we thus achieved a spectral resolution of2.6˚A.The spectra were obtained on July22-25,1998.For calibration purposes we observed each night ten bias frames and ten domeflat-fields with a mean exposure level of about10,000counts each.Be-fore and after each science observation we took HeAr spectra for wavelength calibration purposes.We observed dark frames of3600and1800sec duration to measure the dark current of the CCD.Asflux standard stars we used LTT7987and EG274.We also analyse data that were obtained as backup targets at the NTT during observing runs dedicated to other programs(60.E-0145,61.E-0361).The observational set-up and the data reduction are described in Moehler et al.(2000,1999b).These data have a much lower resolution of5.4˚A.3.Data ReductionWefirst averaged the bias andflatfield frames separately for each night.As we could not detect any significant change in the mean bias level we computed the median of the bias frames of the four nights and found that the bias level showed a gradient across the image,increasing from the lower left corner to the upper right corner by about1%.Wefitted the bias with a linear approximation along both axes and used thisfit as a bias for the fur-ther reduction.As no overscan was recorded we could not adjust the bias level.Bias frames taken during the night, however,revealed no significant change in the mean bias level.The mean dark current determined from long dark frames showed no structure and turned out to be negligi-ble(3±3e−/hr/pixel).We determined the spectral energy distribution of the flatfield lamp by averaging the meanflatfields of each night along the spatial axis.These one-dimensional“flat field spectra”were then heavily smoothed and used af-terwards to normalize the domeflats along the dispersion axis.The normalizedflatfields of thefirst three nights were combined.For the fourth night we used only theflat field obtained during that night as we detected a slight variation in the fringe patterns of theflatfields from the first three nights compared to that of the fourth(below 5%).For the wavelength calibration wefitted3rd-order poly-nomials to the dispersion relations of the HeAr spec-tra which resulted in mean residuals of≤0.1˚A.We re-binned the frames two-dimensionally to constant wave-length steps.Before the skyfit the frames were smoothed along the spatial axis to erase cosmic ray hits in the back-ground.To determine the sky background we had tofind regions without any stellar spectra,which were sometimes not close to the place of the object’s spectrum.Neverthe-less theflatfield correction and wavelength calibration turned out to be good enough that a linearfit to the spa-tial distribution of the sky light allowed the sky back-ground at the object’s position to be reproduced with suf-ficient accuracy.This means in our case that after thefit-ted sky background was subtracted from the unsmoothed frame we do not see any absorption lines caused by the pre-dominantly red stars of the clusters.The sky-subtracted spectra were extracted using Horne’s(1986)algorithm as implemented in MIDAS(Munich Image Data Analysis System).Finally the spectra were corrected for atmospheric extinction using the extinction coefficients for La Silla (T¨u g1977)as implemented in MIDAS.The data for the flux standard stars were taken from Hamuy et al.(1992) and the response curves werefitted by splines.Theflux-calibration is helpful for the later normalization of the spectra as it takes out all large-scale sensitivity variations of the instrumental setup.Absolute photometric accuracy is not an issue here.4.Atmospheric ParametersTo derive effective temperatures,surface gravities and he-lium abundances wefitted the observed Balmer and he-lium lines with stellar model atmospheres.Beforehand we corrected the spectra for radial velocity shifts,de-rived from the positions of the Balmer and helium lines. The resulting heliocentric velocities are listed in Table1. The error of the velocities(as estimated from the scatter of the velocities derived from individual lines)is about 40km s−1.The spectra were then normalized by eye and are plotted in Figs.2and3.To establish the bestfit we used the routines developed by Bergeron et al.(1992)and Saffer et al.(1994),which employ aχ2test.Theσnecessary for the calculation of χ2is estimated from the noise in the continuum regions of the spectra.Thefit program normalizes model spectra and observed spectra using the same points for the continuum definition.We computed model atmospheres using ATLAS9(Ku-rucz1991)and used Lemke’s version1of the LIN-FOR program(developed originally by Holweger,Stef-fen,and Steenbock at Kiel University)to compute a grid of theoretical spectra which include the Balmer lines Hαto H22and He i lines.The grid covered the range7,000K≤T eff≤35,000K, 2.5≤log g≤ 6.0,−3.0≤log n He1For a description see http://a400.sternwarte.uni-erlangen.de/∼ai26/linfit/linfor.html4S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixingFig.2.Normalized spectra of the programme stars that were observed at the ESO1.52m telescope.The part shortward of3900˚A was normalized by taking the highestflux point as continuum value.The He i linesλλ4026˚A,4388˚A, 4471˚A,4922˚A,and the Mg ii line4481˚A are marked(if visible in the spectrum).S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixing5 Fig.3.Normalized spectra of the programme stars that were observed at the NTT during1997and1998.See Fig.2 for details.4921˚A.The errors given are r.m.s.errors derived from the χ2fit(see Moehler et al.1999b for more details).These errors are obtained under the assumption that the only er-ror source is statistical noise(derived from the continuum of the spectrum).However,errors in the normalization of the spectrum,imperfections offlatfield/sky background correction,variations in the resolution(e.g.due to seeing variations when using a rather large slit width)and other effects may produce systematic rather than statistic er-rors,which are not well represented by the error obtained from thefit routine.Systematic errors can only be quanti-fied by comparing truly independent analyses of the same stars.As this is not possible here we use our experience with the analysis of similar stars and estimate the true errors to be about10%in T effand0.15dex in log g(cf. Moehler et al.1997b,1998).Two stars show B−V colours that are significantly redder than expected from their ef-fective temperatures(B2697:B−V=+0m.08,T eff=15,700K;B3006:B−V=−0m.10,T eff=30,000K),pos-sibly indicating that the colours are affected by binarity or photometric blending with a cool star.While the spectra look quite normal,we will not include these stars in any statistical discussion below.To increase our data sample we reanalysed the NTT spectra described and analysed by Moehler et al.(1997b).We did not reanalyse the EFOSC1 data published in the same paper as they are of worse quality.Wefind that the atmospheric parameters deter-mined by line profilefitting agree rather well with those published by Moehler et al.(1997b).The temperatures and gravities obtained from these metal-poor atmospheres are compared with the values pre-dicted by canonical HB tracks in Fig.4(top panel).These tracks,which were computed for a main sequence mass of 0.805M⊙,an initial helium abundance Y of0.23and a scaled-solar metallicity[M/H]of−1.54,define the locus of canonical HB models which lose varying amounts of mass6S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixingTable2.Physical parameters,helium abundances,and masses for the target stars in NGC6752as derived using metal-poor modelatmospheres.We used the photometry of Buonanno et al.(1986)to derive the masses.T efflog g log n He[K][cm s−2][M⊙] 65217300±520 4.31±0.09−2.46±0.160.50 115215800±460 4.14±0.09−2.89±0.310.50 173811100±260 3.78±0.12−1.14±0.360.73 325312000±270 3.73±0.07−2.18±0.380.65 340815500±460 4.14±0.09−2.22±0.190.41 342413200±290 3.84±0.07−2.05±0.240.43 346125800±1300 5.15±0.16−2.32±0.240.68 373612200±260 3.68±0.07−2.24±0.540.46 442415400±530 3.96±0.09−2.21±0.240.39 482217300±580 4.38±0.09−2.63±0.220.5694419700±570 4.49±0.09−2.04±0.100.39 178020000±820 4.61±0.12−2.38±0.220.48 269715400±610 4.11±0.10−2.07±0.280.49 274718600±700 4.63±0.12−1.57±0.120.47 300610400±120 3.81±0.17−1.83±1.35 1.09 3140113700±470 3.75±0.10−1.85±0.310.45 3699ESO NTT observations in199329000±520 5.41±0.07≤−30.38 91617400±630 4.10±0.10−2.17±0.160.26 162833400±390 5.78±0.07−1.94±0.090.45 239531300±510 5.55±0.09≤−30.59 397530700±920 5.61±0.12≤−30.54 45481This star is omitted from further analysis as it lies ina temperature range that is difficult to analyse andnot of great interest for our discussion.Fig.4.a-c.Temperatures and gravities of the programme stars in NGC6752.a determined using model atmospheres with cluster metallicity([M/H]=−1.5),b adopting a solar metallicity([M/H]=0)for the model atmospheres,c adopting a super-solar metallicity([M/H]=+0.5)for the model atmospheres(see Sect.4.1for details).The dashed lines mark the locus of the HB evolutionary tracks for [M/H]=−1.54,as computed with helium mixing for the indicated values of the Reimers mass-loss parameterηR (see Sect.4.2for details).The solid lines mark the locus of canonical HB tracks for[M/H]=−1.54.These loci define the region within which the HB models spend99 percent of their HB lifetime.Representative error bars are plotted.during the RGB phase.According to the Reimers mass-loss formulation the value of the mass-loss parameterηR would vary from≈0.4at the red end of the observed HB in NGC6752to≈0.7for the sdB stars,given the present composition parameters.One can see from Fig.4(top panel)that the HBB stars in NGC6752show the same effect as seen in other globular clusters,namely,an offset from the zero-age horizontal branch(ZAHB)towards lower surface gravities over the temperature range4.05<log T eff<4.30(11,200K<T eff<20,000K).At lower or higher temperatures the gravities agree with the locus of the canonical HB tracks.S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixing7 4.1.Radiative levitation of heavy elementsAs described in Moehler et al.(1999a,see also Fig.5),we found evidence for iron enrichment in the spectra ofthe HBB stars obtained at the ESO 1.52m telescope,whereas the magnesium abundance appeared consistentwith the cluster magnesium abundance.The actual ironabundances derived for these stars byfitting the iron linesin the ESO1.52m spectra are listed in Table4.The meaniron abundance turns out to be[Fe/H]=+0.12±0.40(internal errors only,logǫF e=7.58)for stars hotter thanabout11,500K–in good agreement with thefindingsof Behr et al.(1999,2000b)for HBB/HBA(horizontalbranch A type)stars in M13and M15and Glaspey et al.(1989)for two HBB/HBA stars in NGC6752.This ironabundance is a factor of50greater than that of the clus-ter,but still a factor of3smaller than that required toexplain the Str¨o mgren u-jump discussed by Grundahl etal.(1999,logǫF e=8.1).The mean magnesium abundancefor the same stars is[Mg/H]=−1.13±0.29(internal er-rors only),corresponding to[Mg/Fe]=+0.4for[Fe/H]=−1.54.This value agrees well with the abundance[Mg/Fe]=+0.4found by Norris&da Costa(1995b)for red giantsin NGC6752.The abundances are plotted versus temperature inFig.5.The trend of decreasing helium abundance withincreasing temperature seen in the ESO1.52m data(andalso reported by Behr et al.1999for HB stars in M13)isnot supported towards higher temperatures by the NTTdata.This could be due to the lower resolution of theNTT data which may tend to overestimate abundances(Glaspey et al.1989).As iron is very important for the temperature strat-ification of stellar atmospheres we tried to take the in-creased iron abundance into account by computing modelatmospheres for[M/H]=0.Indeed a backwarming ef-fect of2–4%on the temperature structure was found inthe formation region of the Balmer lines,when compar-ing solar composition models with the metal-poor models.,log g,and log n HeWe then repeated thefit to derive T2We determined this temperature by comparing the(u−y)0value,at which the stars return to the ZAHB((u−y)0≈+0.4)to theoretical colours from Kurucz(1992)for[M/H]=+0.5,which is the metallicity required to explain the u-jump.As-suming log g=4.0this comparison results in T eff≈15,000K.8S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus heliummixingFig.6.a-d This plot shows the differences in effec-tive temperature(a,c)and surface gravity(b,d)derived fromfits with model atmospheres of different metallicity (solar−metal-poor[a,b],metal-rich−metal-poor[c,d]).It is obvious that an increase in the metallicity of the model atmospheres usually decreases the resulting temperatures and increases the resulting surface gravities.stars do not return to the ZAHB until a temperature of about20,000K.We next repeated the Balmer line profilefits by in-creasing the metal abundance of the model atmospheres to[M/H]=+0.5(see Fig.4,bottom panel,and Table5), which did not significantly change the resulting values for T effand log g.In particular,note that especially the“de-viant”stars(now between15,300K and19,000K)remain offset from the canonical ZAHB.4.2.Helium mixingAs outlined in Sect.1,helium mixing during the RGB phase may also be able to explain the low gravities of the HBB stars.Under this scenario the mixing currents within the radiative zone below the base of the convective enve-lope of a red giant star are assumed to penetrate into the top of the hydrogen shell where helium is being produced by the hydrogen burning reactions.Ordinarily one would expect the gradient in the mean molecular weightµto prevent any penetration of the mixing currents into the shell.If,however,the timescale for mixing were shorter than the timescale for nuclear burning,then the helium being produced at the top of the shell might be mixed outward into the envelope before aµgradient is estab-lished.Under these circumstances aµgradient would not inhibit deep mixing simply because such a gradient would not exist within the mixed region.Since deep mixing is presumably driven by rotation, one would expect a more rapidly rotating red giant to show a larger increase in the envelope helium abundance.Table3.Physical parameters,helium abundances,and masses for the target stars in NGC6752as derived using solar metallicity model atmospheres.T efflog g log n He[K][cm s−2][M⊙]65216300±460 4.31±0.07−2.45±0.160.51 115215000±360 4.15±0.07−2.89±0.310.52 173811400±170 3.96±0.07−1.51±0.280.98 325312100±220 3.86±0.07−2.44±0.380.80 340814900±370 4.17±0.07−2.25±0.190.43 342413000±210 3.92±0.05−2.22±0.240.49 346124900±1250 5.14±0.16−2.32±0.240.65 373612300±200 3.81±0.05−2.49±0.550.57 442414900±410 3.99±0.09−2.26±0.240.41 482216300±520 4.38±0.09−2.61±0.220.5794418500±570 4.45±0.09−2.02±0.100.36 178018800±790 4.58±0.10−2.36±0.220.46 269714700±490 4.13±0.10−2.10±0.260.50 274717500±600 4.61±0.10−1.54±0.100.46 300610800±310 4.01±0.14−2.33±1.97 1.52 325321800±1050 4.60±0.12−2.30±0.100.3249129400±480 5.60±0.07−1.70±0.050.46 150920600±620 4.78±0.09−2.52±0.120.43 216221000±750 5.06±0.10−1.78±0.090.53 391520400±520 4.92±0.07−2.02±0.120.62 400920700±1490 5.06±0.19−2.00±0.170.62S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixing9Table4.Helium,iron,and magnesium abundances of the HBB stars observed with the ESO1.52m telescope(ex-cept B3655,which has a too noisy spectrum).[Fe/H]and [Mg/H]are derived using solar abundances of logǫFe,⊙= 7.46and logǫMg,⊙=7.53.The physical parameters and the helium abundances are taken from Table3.n H[Fe/H][Mg/H][K][cm s−2]the HB has not,however,been confirmed by the recent observations of M13by Behr et al.(2000a).These obser-vations show that HB stars in M13hotter than11,000K are,in fact,rotating slowly with v sin i<10km s−1in contrast to the cooler HB stars where rotational velocities as high as40km s−1are found(see also Peterson et al. 1995).There are a couple of possible explanations for this apparent discrepancy.One possibility is that the greater mass loss suffered by the HBB stars might carry away so much angular momentum that the surface layers are spun down even though the core is still rotating rapidly.Alter-natively Sills&Pinsonneault(2000)have suggested that the observed gravitational settling of helium in HBB stars might set up aµgradient in the outer layers which in-hibits the transfer of angular momentum from the rapidly rotating interior to the surface.Thus the surface rotational velocities may not necessarily be indicative of the interior rotation.In order to explore the consequences of helium mix-ing for the HBB stars quantitatively,we evolved a set of 13sequences up the RGB to the heliumflash for varying amounts of helium mixing using the approach of Sweigart (1997a,1997b).As in the case of the canonical models dis-cussed previously,all of these mixed sequences had an ini-tial helium abundance Y of0.23and a scaled-solar metal-licity[M/H]of−1.54.The main-sequence mass was taken to be0.805M⊙,corresponding to an age at the tip of the Table5.Physical parameters,helium abundances,and masses for the target stars in NGC6752as derived using metal-rich model atmospheres.T efflog g log n He[K][cm s−2][M⊙]65216300±430 4.36±0.07−2.55±0.140.54115215100±370 4.20±0.07−2.98±0.260.55173811600±180 4.08±0.07−1.74±0.24 1.20325312300±210 3.94±0.07−2.65±0.360.89340815000±350 4.23±0.07−2.40±0.190.47342413200±200 3.99±0.05−2.45±0.260.53346125000±170 5.16±0.16−2.31±0.240.64373612500±200 3.88±0.05−2.70±0.500.62442415000±400 4.05±0.09−2.43±0.240.44482216300±480 4.42±0.09−2.72±0.210.6094418400±580 4.48±0.09−2.08±0.100.37178018700±810 4.60±0.10−2.41±0.220.46269714800±490 4.20±0.10−2.25±0.280.55274717300±570 4.65±0.10−1.61±0.100.49300611100±290 4.14±0.10−2.54±1.97 1.87325321800±030 4.61±0.12−2.31±0.100.3149129300±460 5.59±0.05−1.70±0.050.43150920700±720 4.81±0.09−2.55±0.120.43216220900±810 5.07±0.09−1.80±0.090.52391520300±580 4.94±0.07−2.06±0.120.62400920400±590 5.06±0.19−2.03±0.170.6010S.Moehler et al.:Hot HB stars in globular clusters-V.Radiative levitation versus helium mixingunmixed model would lie near the red end of the observed blue HB in NGC6752.Both the mixing and mass loss were turned offonce the models reached the core Heflash at the tip of the RGB,and the subsequent evolution was then followed through the heliumflash to the end of the HB phase using standard techniques.We did not investigate the changes in the surface abun-dances of CNO,Na and Al caused by the helium mixing, since such a study was beyond the scope of the present pa-per.Rather,our objective was to determine how the mix-ing affected those quantities which impact on the HB evo-lution,i.e.,envelope helium abundance and mass.We do note that the mixing in the more deeply mixed RGB mod-els would have penetrated into regions of substantial Na and Al production according to the calculations of Cav-allo et al.(1996,1998).However,the resulting changes in the surface Na and Al abundances will depend on the as-sumed initial Ne and Mg isotopic abundances and on the adopted nuclear reaction rates,which in some cases are quite uncertain.The locus of the above helium-mixed sequences in the log g-log T effplane is indicated by the dashed lines in the top panel of Fig.4.The red end of the mixed ZAHB in this panel,located at log T eff=3.93,is set by the canonical, unmixed sequence for the present set of model parameters. Since mixing increases the RGB mass loss,a mixed HB model will have a higher effective temperature than the corresponding canonical model.At the same time mixing increases the envelope helium abundance in the HB model, which,in turn,increases both the hydrogen-burning and surface luminosities.The net effect is to shift the mixed locus in Fig.4towards lower gravities with increasing T effcompared to the canonical locus,until a maximum offset is reached for15,500K<T eff<19,000K.At higher tem-peratures the mixed locus shifts back towards the canoni-cal locus,as the contribution of the hydrogen shell to the surface luminosity declines due to the decreasing envelope mass.The predicted locus along the extreme HB(EHB) does not depend strongly on the extent of the mixing,since the luminosities and gravities of the EHB stars are primar-ily determined by the mass of the helium core,which is nearly the same for the mixed and canonical models.Over-all the variation of log g with T effalong the mixed locus in the top panel of Fig.4mimics the observed variation.The results presented in Sect.4.1demonstrate that radiative levitation of heavy elements can account for a considerable fraction of the gravity offset along the HBB, especially for temperatures cooler than15,100K.Conse-quently the amount of helium mixing required to explain the remaining offset between15,300K and19,000K is much less than the amount required to explain the off-sets found without accounting for radiative levitation(top panel of Fig.4).In order to compare the gravities pre-dicted by the helium-mixing scenario with those derived from the metal-enhanced atmospheres,we computed a sec-ond set of mixed sequences using the same approach as above but with a larger value of the mass-loss parameter ηR,i.e.,ηR=0.45.The red end of the mixed ZAHB for these sequences is located at log T eff=4.01and is there-fore hotter than the red end of the mixed ZAHB for the sequences withηR=0.40.The HB stars cooler than this temperature in NGC6752would then be identified with unmixed stars which lost less mass along the RGB.By increasing the mass loss efficiency we reduce the amount of mixing needed to populate the temperature range15,300K<T eff<19,000K and therefore the size of the resulting gravity offset.The locus of the mixed se-quences withηR=0.45is indicated by the dashed lines in the central panel of Fig.4.The gravity offsets along this mixed locus seem to provide a reasonablefit to the gravi-ties given by the model atmospheres with solar metallicity.Finally we computed a third set of mixed sequences with the mass-loss parameter increased further toηR= 0.50for comparison with the gravities obtained from the atmospheres with super-solar metallicity in the bottom panel of Fig.4.As expected,these mixed sequences show a smaller gravity offset in the temperature range15,500K <T eff<19,000K.Moreover,the red end of the mixed ZAHB shifts blueward to log T eff=4.08.5.MassesWe calculated masses for the programme stars in NGC6752from their values of T effand log g using the equation:log M。

a r X i v :a s t r o -p h /9906247v 1 15 J u n 1999A&A manuscript no.(will be inserted by hand later)ASTRONOMYANDASTROPHY SICS1.IntroductionGlobular star clusters (GCs)are among the oldest stellar systems in the Universe and provide a powerful tracer of2S.Holland et al.:Globular clusters in NGC5128Sharples1988).Recently G.Harris et al.(1998)used HST WFPC2images to construct a color–magnitude diagram for C44,a GC in the halo of NGC5128.They found that this GC was an old,intermediate-metallicity object simi-lar to the GCs in the Milky Way.G.Harris et al.(1992, hereafter referred to as HG92)used Washington CMT1T2 photometry to derive metallicities for62of confirmed GCs in NGC5128and found a mean iron abundance of [Fe/H]=−0.8±0.2,which suggest that the NGC5128 GC system is∼3times more metal rich than the Milky Way GC system.They found no evidence for any GCs having metallicities significantly greater than those found in the Milky Way GCs.Such metal-rich GCs might be ex-pected if some of the NGC5128GCs had formed recently in a gas-rich merger event.HG92do,however,suggest that several blue GCs in NGC5128may be analogues of the intermediate-age GCs found in the Magellanic Clouds. On the other hand,Zepf&Ashman(1993)suggest that the metallicity distribution of the NGC5128GCs is bi-modal,with the high-metallicity peak at[Fe/H]=+0.25 due to GCs formed in a merger.Hui et al.(1995)analyzed the kinematics of the NGC5128GC system and found that the metal-rich GCs are part of a dynamically sepa-rate system from the metal-poor GCs.Numerical simula-tions suggest that the merger event occurred between160 (Quillen et al.1993)and500(D/5Mpc)Myr ago,where D is the distance to NGC5128in Mpc(Tubbs1980). This suggests that any GCs that formed in this particular merger should be quite young and,therefore,rather blue (∼0.4<V−I<0.6;see Sect.5).Minniti et al.(1996)and Alonso&Minniti(1997, hereafter referred to as AM97)used HST Wide-Field/Planetary-Camera1(WF/PC-1)images,taken be-fore the corrective optics package was installed in1993, to search for GCs in the inner regions of NGC5128. They identified125GC candidates,young associations, and open cluster candidates in the inner three kpc of NGC5128.They also used ground-based RK photometry to estimate metallicities for47GC candidates.Schreier et al.(1996)found74compact sources along the northern edge of the NGC5128dust lane using HST WF/PC-1im-ages.They estimate that most of these sources are young stars(spectral class A or earlier)but note that some are resolved and may be GCs.Identifying GC candidates in the inner regions of NGC 5128is difficult since there is nonuniform extinction,con-tamination from foreground stars and background galax-ies,and confusion with open clusters and blue,star-forming knots in NGC5128.GC candidates can not be identified based solely on their colors since the large amount of uneven reddening makes it very difficult to determine the dereddened color of an object.A better approach is a scheme to identify GC candidates based solely on their structural parameters.All known GCs in the Local Group can be reasonably wellfit by Michie–King models(Michie1963;King1966),although∼20%Table1.Log of the observations.Fieldα(J2000)δ(J2000)Filter ExposureS.Holland et al.:Globular clusters in NGC51283cleus of NGC5128.Adjacentfields overlap by∼0.′5givinga total effective area of∼25⊓⊔′for the survey.2.2.Data ReductionsWe combined the exposures for eachfield by taking the average of the three images in eachfilter(four images for the F555W exposures of Field2).No re-registration of the images was performed since the shifts between the images were typically less than0.1pixel(0.′′01on the WFC and 0.′′005on the PC).We estimate that combining the images in this way may result in the sizes of the GC candidates be-ing systematically overestimated by no more than∼0.′′02. We prefer to introduce this simple systematic offset than deal with the poorly-understood systematic uncertainties that arise from interpolatingflux across fractional-pixel shifts.2.2.1.Identifying Globular Cluster CandidatesAt the distance of NGC5128(d=3.6±0.2Mpc),the mean King core-and tidal-radii of the Milky Way GCs would appear to be r t=2.′′59±0.′′20,respec-tively.Therefore,any GCs in NGC5128will appear to be semi-stellar and be strongly affected by the point spread function(PSF)of the WFPC2.After some experimenta-tion,we adopted the following procedure for identifying GC candidates.We wish to stress that this procedure is quite strict and will probably result in the rejection of some legitimate GC candidates.However,we prefer to re-ject real GCs rather than have our sample contaminated with stars or background galaxies.In order to increase the signal-to-noise ratio(S/N)of the GC candidates–a particularly important point for the faint(V<20)GC candidates–we combined the F555W and F814W images for eachfield to getfinding images. The dust lane introduces variations in the background on spatial scales of∼1′′,comparable to the expected sizes of the GC candidates in NGC5128.To reduce the effects of the uneven background light,large-scale spatial varia-tions in the background were removed by running a ring medianfilter(Secker1995)over thefinding image,sub-tracting the resulting smoothed background,and adding back the mean background value.The medianfilter radius was set to1′′,which is∼3.5times the expected full-width at half maximum(FWHM)of a typical GC candidate. This choice offilter radius ensures that the cores of the GC candidates will not be altered by the medianfilter and that any background structure larger than a typical GC candidate will be removed.Since the most extended Milky Way GCs have tidal radii that are significantly greater than2.5times their FWHM,and extended halos have been detected around several Galactic and extra-Galactic GCs(Grillmair et al.1995;1996;Holland et al.1997), this approach will alter the distribution of light in the outer regions of most of the GC candidates.However,this is not important since thefinding images are used only to construct a preliminary list of GC candidates.A more rigorous set of criteria,based on the structures of the GC candidates as determined from the original images,will be applied to the preliminary list to obtain afinal list of GC candidates in the central regions of NGC5128.Thefirst step in our identification procedure was to run the daophot ii(Stetson1987;1994)find routine on the background-subtracted images to identify GC can-didates.Thefinding thresholds were set to6σsky for the PC images and10σsky for the WFC images.Tests with artificial GCs suggested that any detections below these thresholds would be rejected at some point in our identifi-cation process.Daophot ii find has an algorithm for re-jecting non-stellar objects based on two parameters called “sharpness”and“round”.This algorithm was turned offsince images of GCs can have different shapes and concen-trations from images of stars.Next,the daophot ii photometry routine was used to obtain aperture photometry for each of these detec-tions.The photometry was performed separately on each of the combined F555W and F814W frames,not on the combinedfinding frame.An aperture radius of0.′′2was used since most Galactic GCs,if moved to the distance of NGC5128,would appear to have core radii smaller than this.Therefore,the signal within the aperture will be dominated by the light from the object and not from the background.Candidate objects with S/N<5within the photometry aperture were discarded since the signal was not strong enough to determine reliable shape parameters. The sky brightness was determined in an annulus with an inner radius of∼0.′′9and an outer radius of∼1.′′1.This annulus was chosen to be far enough from the center of the GC candidate that the light in the annulus will be dominated by the background,yet near enough to the GC candidate that the light in the annulus will be a reasonable approximation of the mean background at the location of the object.For large GC candidates this annulus will be inside the tidal radius of the object so our estimate of the background will be contaminated.However,the values determined at this stage are only preliminary estimates, which will be improved upon later in the identification process when Michie–King models arefit to the GC can-didates.The lists of GC candidates in each of the F555W and F814W images were matched using the daomatch and daomaster software.Only objects that appeared in both the F555W and F814W images,and whose centers matched to within0.′′05(∼1.1pixel on the PC images and∼0.5pix on the WFC images),were considered to be real GC candidates.Distinguishing bonafide GCs from stars and back-ground galaxies is challenging.The colors of the objects can not be used since we are interested in studying the color distribution of GCs in NGC5128and do not wish to bias our sample.To make matters worse,the presence of dust in NGC5128will add a significant amount of scatter4S.Holland et al.:Globular clusters in NGC5128to the intrinsic color distribution,and may cause legit-imate GCs to be rejected if a color-based identificationscheme is used.The solution is to identify GC candidatesby their structural parameters,although the best choiceof structural parameters is not obvious.At the distanceof NGC5128a typical Galactic GC would appear to havean intrinsic FWHM of∼0.′′25,or approximately twicethe FWHM of the WFPC2PSFs.Therefore,the observedFWHM,concentration,and ellipticity of a GC candidatecan be heavily influenced by the PSF.Since the PSF variesstrongly with position on the WFPC2CCDs,the poten-tial for confusion between stellar images and concentratedGC candidates is great if the PSF is not removed,in someway,from the data.Therefore,the observed shape of anobject can not be directly used to classify it as a star,GCcandidate,or galaxy.After some experimentation with adding and recover-ing artificial GCs and artificial stars,we found that thefollowing procedure was reasonably reliable for identify-ing GC candidates.For each GC candidate we took allthe pixel values within1′′of the center of the object andsubtracted an estimate of the local background(the centerand background were determined by the daophot ii pho-tometry algorithm).A one-dimensional Moffatian(Mof-fat1969),M(r eff)=M(0) 1+ r eff1Image Reduction and Analysis Facility(IRAF),a softwaresystem distributed by the National Optical Astronomy Obser-vatories(NOAO).S.Holland et al.:Globular clusters in NGC51285Fig.2.Thisfigure shows the best-fitting Moffatianαand βparameters for each object(small circles)on the F555W images.Simulated data(see Fig.1)suggest that objects that lie inside the wedge formed by the solid lines are ex-tended objects.Therefore we consider any objects that lie inside the wedge to to be GC candidates.The solid squares show the locations of the GC candidates from Table2. 2.2.2.Fitting Michie–King ModelsWefit a two-dimensional,PSF-convolved,single-mass Michie–King model to each of the403GC candidate us-ing software developed by Holland(1997).This software assumes that the surface brightness profile along the ef-fective radius axis of a GC candidate with an ellipticity ofǫand a position angle ofθ0has a King profile with a concentration of c and a core radius of r c.It then builds a two-dimensional model based on this surface brightness profile,ǫ,andθ0.The two-dimensional model is convolved with the appropriate PSF for the location on the CCD and a chi-square minimization is performed between the PSF-convolved model and the original data image.The soft-ware uses CERN’s minuit function minimization package tofit simultaneously the concentration,core radius,to-talflux in the object,ellipticity,position angle,and mean background.Objects located within32pixels of the edge of a CCD(=3.′′2for the WFC and1.′′6for the PC)were notfit to avoid the edges of the CCD biasing thefits.Once a bestfit had been determined,the King tidal radius,r t, and the half-mass radius,r h,of the model were computed.Separatefits were made to the F555W images and the F814W images and an object was considered to be GC Fig.3.Thisfigure shows the best-fitting Moffatianαand βparameters for each object(small circles)on the F814W images.Simulated data(see Fig.1)suggest that objects that lie inside the wedge formed by the solid lines are ex-tended objects.Therefore we consider any objects that lie inside the wedge to to be GC candidates.The solid squares show the locations of the GC candidates from Table2. candidate only if a Michie–King model could befit in both colors.We were able tofit Michie–King models to98of the 403potential GC candidates.Mean structural parameters were calculated for these object by taking the mean of the values found in eachfilter.Four objects(#8,#113, #128,and#129)(see Tables2,3,and4)were identified on multiplefields.In these cases we computed the mean of the structural parameters measured in eachfield.We elected to separate GC candidates from back-ground galaxies based on theirfitted ellipticities and half-mass radii(see Fig.4).Half-mass radii are preferred to tidal radii or core radii because Fokker–Planck models of spherical stellar systems show that half-mass radii re-main reasonably constant over periods of several Gyr (e.g.Cohn1979;Takahashi1997),making it a unique length scale for GCs.The mass interior to the half-mass radius tends to undergo a gravo-thermal collapse and become concentrated at the center of the GC over time(i.e.core-collapse),which results in the core radius shrinking.Meanwhile,the mass exterior to the half-mass radius tends to expand outwards,causing the tidal ra-dius to grow.Since we are interested infinding young, intermediate-age,and old GCs in NGC5128,it is useful to have a selection criterion that does not depend on the age6S.Holland et al.:Globular clusters in NGC5128Fig.4.The ellipticity vs.half-mass radius of the best-fitting single-mass Michie–King model for each object where a Michie–King model was successfully fit.Objects with r h >10′′(∼175pc)have not been plotted.The solid box in the lower left of the plot shows the region occupied by Galactic GCs.Based on this plot we have assumed that any object with r h <2′′(∼35pc)and ǫ<0.4(the dashed box)is a GC candidate in the NGC 5128system.of the GC candidate.Galactic GCs have half-mass radii of approximately 1.3<r h <31.9pc (W.Harris 1996),which corresponds to 0.′′07<r h <1.′′83at the distance of NGC 5128.There is no evidence that the radius of a Galactic GC depends on its mass (van den Bergh et al.1991).There-fore,we have assumed that only objects with r h ≤2′′(∼35pc at the distance of NGC 5128)were GC candi-dates.It is possible that some of the objects in Fig.4that have high ellipticities and low half-mass radii are double clusters.However,Innanen et al.1983have shown that a binary GC could not survive a single Galactic orbit in the Milky Way so it is unlikely that there are any old,or intermediate-age double GCs in NGC 5128.It is possi-ble that very young multiple GCs that formed within the last ∼100−200Gyr could have survived to the present day,but we are unable to differentiate between them and background galaxies.The most elliptical Galactic GC is M19with ǫ=0.27(White &Shawl 1987)and the most elliptical GC known is NGC 2193in the Large Magellanic Cloud (LMC)which has ǫ=0.33(Geisler &Hodge 1980).Geisler &Hodge (1980)modelled the distribution of observed el-lipticities for 25GCs in the LMC and found that it was unlikely that the largest true ellipticity exceeded ǫ=0.4.The LMC contains both dynamically young and dynami-Fig.5.A finding chart for Field 1.The GC candidates are circled with their identification numbers (see Table 2)printed near each object.Fig.6.A finding chart for Field 2.Fig.7.A finding chart for Field 3.Fig.8.A finding chart for Field 4.cally old GCs,so the largest ellipticity seen in the LMC is a reasonable estimate of the largest ellipticity that we can expect to see in NGC 5128.Therefore,only objects with ǫ≤0.4were considered to be GC candidates.The final step was to examine visually the WFPC2images of each GC candidate to ensure that the Michie–King model fits looked realistic.We found that ∼20%of the objects were either located on diffraction spikes from saturated stars,or exhibited unusually large resid-uals when the best-fitting Michie–King models were sub-tracted.These spurious identifications were discarded.Fig.4shows the measured half-mass radii and ellip-ticities for the surviving objects in NGC 5128and Ta-ble 2shows the final list of GC candidates that we find in the central regions of NGC 5128.The second and third columns show the J2000coordinates of the objects as determined using the IRAF/STSDAS (v2.0.1)task sts-das.toolbox.imgtools.xy2rd .Column 4is the ob-served (projected)distance of the GC candidate from the center of NGC 5128in arcminutes.The center of NGC 5128was taken to be αJ2000=13h 25m 27.s 3,δJ2000=−43◦01′09′′(Johnston et al.1995).Columns 5and 6give the field (from Table 1)and CCD that the object was found on.Columns 7and 8give the X and Y coordinates (in pixels)on the CCD.Column 9lists the identification number of the object in Table 1of Minniti et al.(1996).Tables 3and 4lists the coordinates for the 61extended objects with r h >2′and ǫ>0.4.Some of these objects may be GCs in NGC 5128while others may be background galaxies with structures similar to those of Michie–King models.Six of these objects have been previously identi-fied as GCs by Minniti et al.(1996)and Sharples (1988).Figs.5through 10show the locations of the 21GC can-didates on the F814W-band WFPC2images.Only objects that pass all of the criteria described above are marked on these figures.Objects (such as #15)were only marked on the fields that they were identified as GC candidates in.In most of the cases where a GC candidate is present in multiple fields,but only identified in one field,the GC candidate was located very near the edge of one of the CCDs.Spatial variations in the PSF are largest near the edges of the CCDs so the Michie-King model fits are less reliable near the edges of the CCDs.S.Holland et al.:Globular clusters in NGC51287 Table2.The GC candidates in the central regions of NGC5128.IDαJ2000δJ2000D Field CCD X Y Other8S.Holland et al.:Globular clusters in NGC5128Table3.Extended objects with r h>2′′andǫ>0.4in the central regions of NGC5128IDαJ2000δJ2000D Field CCD X Y OtherS.Holland et al.:Globular clusters in NGC51289 Table4.Extended objects with r h>2′′andǫ>0.4in the central regions of NGC5128,continued.IDαJ2000δJ2000D Field CCD X Y Other∆W0)and standard error in the mean ofthesefive values for each GC candidate.This gave us anestimate of the systematic uncertainty in the value of W0that we derived for each GC candidate.Finally,we com-puted the mean,standard error in the mean,and medianof the individual10S.Holland et al.:Globular clusters in NGC5128Table5.The best-fitting structural parameters for the NGC5128GC candidates.The values are the means of the structural parameters derived from the F555W and F814W images.ID W0±σr c±σr h±σr t±σc±σǫ±σθ0±σχ2ν∆r c0.′′0100.′′0030.′′004∆r t0.′′1540.′′0510.′′060∆ǫ0.0240.0040.022n n.(3) The mean core radius for the21NGC5128GC candidates isr c=0.′′11±0.′′01(se).A Kolmogorov–Smirnov(KS)test shows that we can reject the hypothesis that the two samples are drawn from the same distribution at the46%confidence level.Therefore, there is no evidence that the core radii of the GC can-didates in NGC5128are distributed differently from the core radii of the Milky Way GCs.The most noticeable difference between the core radii of the NGC5128GCcandidates and the core radii of the Milky Way GCs in Fig.11is the lack of a tail extend-ing to large core radii in the NGC5128data.This may be an artifact of the small number(21)of NGC5128GC candidates in our sample.The mean of a distribution is sensitive to the presence of tails and outliers,but the me-dian is much more robust against outliers.Therefore we computed the median core radius for each data set.Both the NGC5128GC candidates and the Milky Way GCs have median core radii of[r c]=0.′′07(=1.22pc,or∼0.7 pixels on the WF CCDs and1.4pixels on the PC CCD). The similarity in the median core radii suggests that theFig.11.Thisfigure compares the distribution of core radii for GC candidates in NGC5128with the distribution of core radii for selected GCs(see Sect.3.1)in the Milky Way.The vertical axis is the fraction of the total number of GCs and the error bars show the Poisson uncertainties in each bin.“typical”core radius of a GC candidate in NGC5128is similar to that of the Milky Way GCs.There may be systematic biases in the core radii that we have derived for the GC candidates in NGC5128.Fit-ting Michie–King models to objects with core radii that are similar to the pixel scales of the images requires that the centers of the objects be accurately known.Small er-rors in determining the center of a candidate GC,and small systematic errors introduced by integrating Michie–King model profiles over the area of a pixel,may be suf-ficient to bias thefitted core radii towards smaller values. In addition to pixelation effects,the similarity between the core radii and the FWHMs of the PSFs may also be biasing ourfits toward smaller core radii.3.2.Tidal RadiiThe tidal radius of a GC is affected by the gravitational potential of its parent galaxy(e.g.,Innanen et al.1983; Heggie&Ramamani1995).In order to compare the tidal radii of GC candidates in NGC5128with those of GCs in the Milky Way it is necessary to correct for the tidal fields of both galaxies.Thefirst step is to normalize the tidal radius of each GC candidate by its mass,M cl,to getr t/M1/3cl .If we assume that the gravitational potentials ofthe Milky Way and NGC5128can be approximated by aspherical logarithmic potential of the formΦ=V2rot ln(R2+R2s)+ln(C),(4)where R is the galactocentric distance,R s is a scale length,and C is a constant,then we can compare the mean valueof the normalized tidal radii of the GC candidates usingr tV rot 2/3GM1/3cl= V rot,MW g(e) 1/3 R2/3pM1/3cl MW,(6)where the subscript MW denotes the value for the MilkyWay.Eq.6assumes that the shape of the galactic potentialis the same in both galaxies,but allows the total mass,as parameterized by V rot,of each galaxy to vary.It alsorequires a knowledge of the distribution of GC orbits ineach galaxy,as parameterized by g(e)and R p.We as-sumed a rotation velocity of V rot,MW=220km s−1for theMilky Way and V rot=245km s−1for NGC5128(Huiet al.1995).The only information available on the distri-bution of orbits for GC candidates in NGC5128is theprojected radial distances of the GC candidates from thecenter of NGC5128,so we have assumed that the NGC5128GC system is dynamically similar to the Milky WayGC system.This involves two assumptions about the na-ture of the GC orbits.First,we assume that the mean ec-centricity of the NGC5128GC orbits is the same as thatfor the Milky Way GCs(Fig.12.Thisfigurecompares the distribution of normal-ized tidal radii(see Sect.3.2)for GC candidates in NGC 5128with the distribution of normalized tidal radii(af-ter correcting for the difference in mass between the two galaxies)for GCs in the Milky Way.The vertical axis is the fraction of the total number of GCs and the error bars show the Poisson uncertainties in each bin.5128GC candidates from the observed distance from the center of NGC5128usingR p=1−e8D.(7)Fig.12shows the distribution of the normalized tidal radii for the NGC5128GC candidates and the Milky Way GCs.The cluster masses were computed from their total V-band luminosities assuming a mass-to-light ratio of two.Our sample of GC candidates has r t/M1/3cl=0.36±0.05(se)pc/M1/3⊙(N=21).The mean normalized tidalradius for73selected Milky Way GCs is r t/M1/3clMW=0.69±0.08(se)pc/M1/3⊙.The multiplicative factor in Eq.6is0.703,which yields a corrected r t/M1/3clMW of0.49±0.06(se)pc/M1/3⊙.A KS test says that we can reject the hypothesis that the two samples are drawn from the same distribution at the74%confidence level.Therefore,there is insufficient evidence to state that the distribution of the tidal radii of the NGC5128GC candidates differs from that of the Galactic GCs if the difference in the masses of the two galaxies is taken into account.However,we wish to stress that this calculation assumes that the distribution of GC orbits are statistically similar for both galaxies.Fig.13.Thisfigure compares the distribution of half-mass radii for GC candidates in NGC5128with the distribution of half-mass radii for a subset of GCs in the Milky Way. The vertical axis is the fraction of the total number of GCs and the error bars show the Poisson uncertainties in each bin.3.3.Half-Mass RadiiThe distribution of half-mass radii for the NGC5128GC candidates is shown in Fig.13along with the same dis-tribution for our subsample of73Milky Way GCs.The half-mass radii for the Milky Way GCs were determined by computing a Michie–King model(with concentrations and core radii taken from W.Harris1996)for each Milky Way GC.This allowed us to make a direct comparison between the half-mass radii of the best-fitting single-mass Michie–King models for the NGC5128GC candidates and the half-mass radii of the best-fitting single-mass Michie–King models for the Milky Way GCs.The mean half-mass radius for the21NGC5128GC candidates isr h=0.′′67±0.′′05(se).However, a KS test says that we can reject the hypothesis that the two samples are drawn from the same distribution at only the74%confidence level,so there is no evidence that the distribution of half-mass radii in our sample of NGC5128 GC candidates is different from that of the GCs in the Milky Way.3.4.EllipticitiesThe distribution of ellipticities for the NGC5128GC can-didates is shown in Fig.14.The21NGC5128objects haveFig.14.The upper panel shows the distribution of ellip-ticities for the GC candidates in NGC5128.The lower panel shows the distribution of ellipticities for the Milky Way’s GCs(from White&Shawl1987).ǫ=0.07±0.01(se)for the 73Milky Way GCs.A KS test says that we can reject the hypothesis that the two samples are drawn from the same distribution at the99.7%confidence level.Therefore,we conclude that the NGC5128GC candidates may have a different distribution of ellipticities from the Milky Way GCs.The NGC5128GC candidates appear to be system-atically more elliptical than the Milky Way GCs.There appears to be a lack of objects with low ellipticities and an excess of GC candidates withǫ∼0.3.The lack of GC can-didates withǫ≤0.05is probably due to the elliptical PSF not being fully removed from the data.Another possible source of ellipticity is the stochastic distribution of bright stars near the center of the object.Geisler&Hodge(1980) found that the random placement of stars with respect to the adopted center of a GC can introduce a systematic error in the observed ellipticity of+0.045±0.015for GCs which are intrinsically spherical.This effect acts to make nearly spherical GCs appear to be more elliptical than they actually are.They also found that this systematic error decreases as the intrinsic ellipticity of the GCs in-creases.This would explain the lack of nearly circular GC candidates in NGC5128,relative to the Milky Way.AM97identified125GC candidates in the inner2.′8×2.′8of NGC5128.Table7lists the ellipticities from those GC candidates in common between the two studies.The Table7.A comparison of our ellipticities with those of AM97.ID AM97Ourǫ±σAM97ǫ。

a r X i v :a s t r o -p h /9707060v 1 4 J u l 1997METAL ABUNDANCES OF ONE HUNDRED HIPPARCOS DW ARFSR.G.Gratton 1,E.Carretta 2,G.Clementini 2,C.Sneden 31Osservatorio Astronomico di Padova,Vicolo dell’Osservatorio 5,35122Padova,ITALY2Osservatorio Astronomico di Bologna,ITALY3Department of Astronomy,The University of Texas at AustinABSTRACTAbundances for Fe,O,and the α−elements (Mg,Si,Ca,and Ti)have been derived from high resolution spectra of a sample of about one hundred dwarfs with high precision parallaxes measured by HIPPAR-COS.The stars have metal abundances in the range −2.5<[Fe/H]<0.2.The observational data set con-sists of high dispersion (20,000<R <70,000),high S/N (>200)spectra collected at the Asiago and McDonald Observatories.The abundance analysis followed the same precepts used by Gratton et al.(1997a)for ∼300field stars and for giants in 24glob-ular clusters (Carretta &Gratton 1997),and includes corrections for departures from LTE in the formation of O lines.Our main results are:1.the equilibrium of ionization of Fe is well satisfied in late F –early K dwarfs2.O and α−elements are overabundant by ∼0.3dex This large homogeneous data set was used in the derivation of accurate ages for globular clusters (See paper by Gratton et al.at this same Meeting).Key words:Stars:chemical abundances -Stars:ba-sic parameters1.INTRODUCTIONHIPPARCOS has provided parallaxes with accura-cies of ∼1mas for several hundreds dwarfs.We had access to data for about 100dwarfs with metal abun-dances in the range −2.5<[Fe/H]<0.2and have used them in a thorough revision of the ages of the old-est globular clusters derived by Main Sequence (MS)fitting technique.A crucial step in the derivation of ages via this method is the assumption that the nearby subdwarfs have the same chemical composi-tion of the globular cluster main sequence stars.This assumption was verified through a careful abundance analysis of the vast majority of nearby dwarfs with HIPPARCOS parallaxes available to us.Our data set and the HIPPARCOS parallaxes were also used to test whether an appreciable Fe overion-ization occurred in the atmosphere of late F –early K dwarfs (Bikmaev et al.1990;Magain &Zhao 1996).This was done by comparing abundances provided by neutral and singly ionized lines,once the surface gravity of each program star had be derived from its mass,temperature and luminosity rather then from the equilibrium of ionization of Fe.Finally,our abundances are fully consistent with those presented by Gratton et al.(1997a)for about 300field dwarfs.A large,homogenous data base of high accuracy (errors ∼0.07dex)abundances com-puted with the Kurucz (1993)model atmospheres is now available and can be used to recalibrate photo-metric and low S/N spectroscopic abundances.2.BASIC DATA FOR SUBDWARFS Average V magnitudes and colors (Johnson B −V and V −K ,and Str¨o mgren b −y ,m 1and c 1)for the programme stars were obtained from a careful discussion of the literature data.We used also the Tycho V magnitudes and B −V colors,after cor-recting them for the very small systematic difference with ground-based data.Absolute magnitudes M V were derived combining ap-parent V magnitudes and Hipparcos parallaxes.No Lutz-Kelker corrections were applied.Lutz-Kelker corrections (Lutz &Kelker 1973)take into account that stars with parallaxes measured too high are more likely to be included in a sample if the sample selection criteria are based on the parallaxes them-selves.Since our sample was selected before the HIP-PARCOS parallaxes were known;Lutz-Kelker correc-tions should not be applied when the whole sample is considered,as we do when comparing the abundances obtained from Fe I and Fe II lines.Multiple high precision radial velocity observations exist for a large fraction of our objects (80out of 99).Twenty stars in the sample are known and four are suspected spectroscopic binaries.Two further stars display very broad lines in our spectra,possibly due to fast rotation.They were discarded.A few other stars display some IR excess,which also may be a signature of binarity.No evidence for binarity dis-turbing the present analysis exists for the remaining stars.Sixty-eight out of the99stars of our sample are in-cluded in Carney et al.(1994)catalogue.Reddening estimates are given for58of them.All but two have zero values.We have thus assumed a zero reddening for all the programme stars.3.OBSERVATIONS AND REDUCTIONSHigh dispersion spectra for about two thirds of the programme stars were acquired using the2D-coud`e spectrograph of the2.7m telescope at McDonald Ob-servatory and the REOSC echelle spectrograph at the 1.8m telescope at Cima Ekar(Asiago).McDonald spectra have a resolution R=70,000,S/N∼200, and spectral coverage from about4,000to9,000˚A; they are available for21stars(most with[Fe/H]<−0.8).Cima Ekar telescope provided spectra with resolution R=15,000,S/N∼200,and two spectral ranges(4,500<λ<7,000and5,500<λ<8,000˚A) for65stars.Equivalent widths EW s of the lines were measured by means of a gaussianfitting routine applied to the core of the lines;appropriate average corrections were included to take into account the contribution of the damping wings.Only lines with log EW/λ<−4.7 were used in thefinal analysis(corrections to the EW s for these lines are≤7m˚A,that is well be-low10per cent).The large overlap between the two samples(14stars)allowed us to tie the Asiago EW s to the McDonald ones.External checks on our EW s are possible with Ed-vardsson et al(1993:hereinafter E93)and Tomkin et al.(1992:hereinafter TLLS).Comparisons per-formed using McDonald EW s alone show that they have errors of±4m˚A.From the r.m.s.scatter,σ, between Asiago and McDonald EW s,we estimate that the former have errors of±6.7m˚A.When Asi-ago and McDonald EW s are considered together,we find average residuals(us-others)of−0.2±1.0m˚A (39lines,σ=6.1m˚A)and+0.8±1.0m˚A(36lines,σ=5.9m˚A)with E93and TLLS,respectively.4.ANALYSIS4.1.Atmospheric ParametersThe abundance derivation followed precepts very similar to the reanalysis of∼300field and∼150 globular cluster stars described in Gratton et al. (1997a)and Carretta&Gratton(1997).The same line parameters were adopted.The effective tem-peratures were derived from B−V,b−y,and V−K colours using the iterative procedure outlined in Gratton et al.(1997a).Atmospheric parameters are derived as follows:1.we assume as input values log g=4.5and themetal abundance derived from the uvby photom-etry using the calibration of Schuster&Nissen (1989)2.T effis then derived from the colours,using theempirical calibration of Gratton et al.(1997a) for population I stars(assumed to be valid for [Fe/H]=0),and the abundance dependence given by Kurucz(1993)models3.afirst iteration value of log g is then derived fromthe absolute bolometric magnitude(derived from the apparent V magnitude,parallaxes from Hip-parcos,and bolometric corrections BC from Ku-rucz1993),and masses obtained by interpolation in T effand[A/H]within the Bertelli et al.(1997) isochrones4.steps2and3are iterated until a consistent setof values is obtained for T eff,log g,and[A/H] 5.the EW s are then analyzed,providing new val-ues for v t and[A/H](assumed to be equal to [Fe/H]obtained from neutral lines)6.the procedure is iterated until a new consistentset of parameters is obtained4.2.Error analysisRandom errors in T eff(±45K)were obtained by com-paring temperatures derived from different colours. Systematic errors may be larger;the T eff-scale used in this paper is discussed in detail in Gratton et al. (1997a).We assume that systematic errors in the adopted T eff’s are≤100K.Random errors in the gravities(±0.09dex)are esti-mated from the errors in the masses(1.2per cent), M V’s(0.18mag),and in the T eff’s(0.8per cent), neglecting the small contribution due to BC’s.Sys-tematic errors(±0.04dex)are mainly due to errors in the T effscale and in the solar M V value.Random errors in the microturbulent velocities can be estimated from the residuals around thefitting re-lation in T effand log g.We obtain values of0.47and 0.17km s−1for the Asiago and McDonald spectra, respectively.Random errors in the EW s and the line parameters significantly affect the abundances when few lines are measured for a given specie.Errors should scale as σ/√Figure1.Run of the difference between the abundances derived from neutral and singly ionized Fe lines as a func-tion of temperature(panel a)and overall metal abundance (panel b).Open squares are abundances obtained from the Asiago spectra;filled squares are abundances obtained from the McDonald spectraMcDonald spectra,respectively.Systematic errors (∼0.08dex)are mainly due to the T effscale.parison with other abundancesOn average,differences(Asiago−McDonald)in the Fe abundances are−0.01±0.02dex(12stars,σ= 0.07dex).Analogous differences for the[O/Fe] and[α/Fe]ratios are+0.02±0.08dex(5stars,σ=0.17dex),and+0.01±0.03dex(12stars,σ=0.10dex).E93measured abundances for∼200dwarfs;six stars are in common with our sample.Abundance residu-als(our analysis−E93)are+0.08±0.03,−0.02±0.03, and+0.02±0.02dex for[Fe/H],[O/Fe],and[α/Fe], respectively.Residual differences are mainly due to our use of a higher temperature scale(our T eff’s are larger by63±12K).We have six stars in common with TLLS,which used a restricted wave-length range.Average differences(ours−TLLS)are: +0.34±0.04and−0.31±0.07dex for[Fe/H]and [O/Fe],respectively.They are due to different as-sumption in the analysis:(i)our temperature scale is higher;(ii)TLLS used a different solar model; (iii)our non-LTE corrections to the O abundances are slightly larger.Finally,Gratton et al.(1997a) made a homogenous reanalysis of the original EW s for∼300metal-poorfield stars.On average,the present Fe abundances are larger by0.02±0.02dex (11stars,σ=0.06dex).Since the same analysis procedure is adopted,these differences are entirely due to random errors in the EW s and in the adopted colours.In the following,we assume that Gratton et al.abundances are on the same scale of the present analysis.4.4.Fe abundancesSince gravities are derived from masses and luminosi-ties rather than from the equilibrium of ionization for Fe,we may test if predictions based on LTE are sat-isfied for the program stars.In Figure1we plot the difference between abun-dances of Fe obtained from neutral and singly ion-ized lines against effective temperature and metal abundance.Different symbols refer to results ob-tained from McDonald and Asiago spectra,respec-tively.McDonald spectra have a higher weight be-cause the higher resolution allowed us to measure a larger number of Fe II lines(10∼20),and errors in the EW s are smaller;very few Fe II lines could be measured in the crowded spectra of cool and/or metal-rich stars observed from Asiago.Average dif-ferences between abundances given by Fe I and II lines are0.025±0.020(21stars,σ=0.093dex)for the Mc Donald spectra,and−0.063±0.019(52stars,σ=0.140dex)for the Asiago spectra.The scatter obtained for McDonald spectra agrees quite well with the expected random error of0.085dex.The average value is consistent with LTE if the adopted T effscale is too high by∼20K,well within the quoted error bar of±100K.The lower mean difference obtained for the Asiago spectra is due to a few cool metal-rich stars which have very crowded spectra.Very few Fe II lines could be measured in these spectra and the line-to-line comparison with the superior McDonald data suggests that even these lines may be affected by blends.We conclude that the equilibrium of ionization for Fe is well satisfied in the late F–K dwarfs of any metallicity in our sample.This result depends on the adopted temperature scale.Our empirical result agrees very well with the ex-tensive statistical equilibrium calculations for Fe by Gratton et al.(1997b).In that paper,the uncertain collisional cross sections were normalized in order to reproduce the observations of the RR Lyraes,where overionization is expected to be much larger than in late F–K dwarfs.The lower limit to collisional cross sections given by the absence of detectable overion-ization in RR Lyrae spectra(Clementini et al.1995) implies that LTE is a very good approximation for the formation of Fe lines in dwarfs.4.5.O andα−element abundancesO abundances were derived from the permitted IR triplet,and include non-LTE corrections computed for each line in each star following the precepts of Gratton et al.(1997b).Wefind that O and the other α−elements are overabundant in stars with[Fe/H]<−0.5(see Figure2):[O/Fe]=0.38±0.13[α/Fe]=0.26±0.08,(error bars are the r.m.s.scatter of individual val-ues around the mean).The moderate O excess de-rived from the IR permitted lines is a consequence of the rather high temperature scale adopted.When this adoption is made,abundances from permitted OI lines agree with those determined from the forbidden [OI]and the OH lines.The present abundances agree very well with those derived in Gratton et al.(1997c).Note also that the overabundance of O andα−elements found for thefield subdwarfs is similar to the excesses foundFigure2.Runs of the overabundances of O(panel a)and α−elements(panel b)as a function of[Fe/H]for the pro-gramme subdwarfs.Filled squares are abundances from McDonald spectra;open squares are abundances from Asi-ago spectrafor globular cluster giants(apart from those stars af-fected by the O-Na anticorrelation,see Kraft1994).5.CALIBRATION OF PHOTOMETRICABUNDANCESOnce combined with the abundances obtained by Gratton et al.(1997a),the sample of late F to early K-typefield stars with homogenous and accu-rate high dispersion abundances adds up to nearly 400stars.Schuster&Nissen(1989)have shown that rather accurate metal abundances for late F to early K-type can be obtained using Str¨o mgren uvby pho-tometry(available for a considerable fraction of the HIPPARCOS stars).Furthermore,the extensive bi-nary search by Carney et al.(1994)has provided a large number of metal abundances derived from an empirical calibration of the cross correlation dips for metal-poor dwarfs.We have recalibrated these abundance scales.Schus-ter&Nissen(1989)abundances onlydiffers for a zero-point offset(see panel a of Figure3);the mean difference is:[Fe/H]us =[Fe/H]SN+(0.102±0.012),(1)based on152stars(the r.m.s.scatter for a single star is0.151dex).In the case of Carney et al.(1994,panel b of Fig-ure3),a small linear term is also required.The best parison between the abundances obtained from high dispersion spectra(present analysis or Gratton et al.1997),and those provided by the original calibration of Schuster&Nissen(1989,panel a)and Carney et al. (1994,panel b)fit line(66stars)is:[Fe/H]us=(0.94±0.03)[Fe/H]C94+(0.18±0.17),(2) The offsets between the high dispersion abundances and those provided by Schuster&Nissen(1989)and Carney et al.(1994)are mainly due to different as-sumptions about the solar abundances in the high dispersion analyses originally used in the calibrations of Schuster&Nissen(1989)and Carney et al.(1994).REFERENCESBertelli,P.,Girardi,L.,Bressan,A.,Chiosi,C.,&Nasi,E.1997,in preparationBikmaev,I.F.,Bobritskij,S.S.,El’kin,V.G.,Lyashko,D.A.,Mashonkina,L.I.,&Sakhibullin,N.A.1990,inIAU Symp.145,Evolution of Stars:the Photospheric Abundance Connection,G.Michaud ed.Carney,B.W.,Latham,D.W.,Laird,J.B.,&Aguilar, L.A.1994,AJ,107,2240Carretta,E.,&Gratton,R.G.1997,A&AS,121,95 Clementini,G.,Carretta,E.,Gratton,R.G.,Merighi, R.,Mould,J.R.,&McCarthy,J.K.1995,AJ,110, 2319Edvardsson,B.,Andersen,J.,Gustafsson,B.,Lambert,D.L.,Nissen,P.E.,&Tomkin,J.1993,A&A,275,101Gratton,R.G.,Carretta,E.,&Castelli,F.1997a,A&A, in pressGratton,R.G.,Carretta,E.,Gustafsson,B.,&Eriksson, K.1997b,submitted to A&AGratton,R.G.,Carretta,E.,Matteucci,F.,&Sneden,C.1997d in preparationKing,J.R.1993,AJ,106,1206Kraft,R.P.1994,PASP,106,553Kurucz,R.L.1993,CD-ROM13and CD-ROM18 Lutz,T.E.,Kelker,D.H.1973,PASP,85,573 Magain,P.,Zhao,G.1996,A&A,305,245Schuster,W.J.,&Nissen,P.E.1989,A&A,221,65 Tomkin,J.,Lemke,M.,Lambert,D.L.,&Sneden,C.1992,AJ,104,1568。

a r X i v :a s t r o -p h /0103501v 1 29 M a r 2001Astronomy &Astrophysics manuscript no.(will be inserted by hand later)The Globular Cluster System of NGC 1316(Fornax A)M.G´o mez 1,2,T.Richtler 3,L.Infante 1,and G.Drenkhahn 41Departamento de Astronom ´ıa y Astrof ´ısica,P.Universidad Cat´o lica de Chile.Casilla 306,Santiago,Chilee-mail:mgomez@astro.puc.cl,linfante@astro.puc.cl 2Sternwarte der Universit¨a t Bonn,Auf dem H¨u gel 71,D-53121Bonn,Germany 3Departamento de F ´ısica,Universidad de Concepci´o n.Casilla 4009,Concepci´o n,Chile e-mail:tom@coma.cfm.udec.cl 4Max-Planck-Institut f¨u r Astrophysik,Postfach 1317,D-85741Garching bei M¨u nchen,Germany e-mail:georg@mpa-garching.mpg.deReceived .../Accepted ...Abstract.We have studied the Globular Cluster System of the merger galaxy NGC 1316in Fornax,using CCD BV I photometry.A clear bimodality is not detected from the broadband colours.However,dividing the sample into red (presumably metal-rich)and blue (metal-poor)subpopulations at B −I =1.75,we find that they follow strikingly different angular distributions.The red clusters show a strong correlation with the galaxy elongation,but the blue ones are circularly distributed.No systematic difference is seen in their radial profile and both are equally concentrated.We derive an astonishingly low Specific Frequency for NGC 1316of only S N =0.9,which confirms with a larger field a previous finding by Grillmair et al.(1999).Assuming a “normal”S N of ∼4for early-type galaxies,we use stellar population synthesis models to estimate in 2Gyr the age of this galaxy,if an intermediate-age population were to explain the low S N we observe.This value agrees with the luminosity-weighted mean age of NGC 1316derived by Kuntschner &Davies (1998)and Mackie &Fabbiano (1998).By fitting t 5functions to the Globular Cluster Luminosity Function (GCLF),we derived the following turnover magnitudes:B =24.69±0.15,V =23.87±0.20and I =22.72±0.14.They confirm that NGC 1316,in spite of its outlying location,is at the same distance as the core of the Fornax cluster.Key words.galaxies:distances and redshifts –galaxies:elliptical and lenticular,cD –galaxies:individual:NGC 1316–galaxies:interactions1.IntroductionThe analysis of globular cluster systems (GCSs)in ellipti-cal galaxies can have different motivations.One of them is to investigate the variety of GCS morphologies in relation to their host galaxy properties in order to gain insight into the formation of cluster systems (see Ashman &Zepf 1997and Harris 2000for reviews).On the other hand,GCSs have been successfully em-ployed as distance indicators (Whitmore et al.1995,Harris 2000).This is particularly interesting if the host galaxy is simultaneously host for a type Ia supernova whose ab-solute luminosity can accordingly be determined,as has been the case for SN 1992A in NGC 1380(Della Valle et al.1998)and SN 1994D in NGC 4526(Drenkhahn &Richtler 1999).However,it can happen that both aspects are equally interesting as with the target of the present contribution,NGC 1316.2M.G´o mez et al.:The Globular Cluster System of NGC1316(Fornax A)during a merger event.But despite the strong evidence for a previous merger in NGC1316,the only indication for cluster formation is that Grillmair et al.(1999,here-after Gr99),could not see a turnover in the GCLF at the expected magnitude.They interpreted thisfinding as an indication for an enhanced formation of many less massive clusters,perhaps in connection with the merger.However, their HST study was restricted to the innermost region of the galaxy.In contrast to other galaxy mergers,where,presum-ably caused by the high star-formation rate(Larsen& Richtler1999,2000),the specific frequency of GCs in-creases,Gr99found an unusually small total number of clusters relative to the luminosity of NGC1316(M V∼−22.8,adopting a distance modulus to Fornax ofµ= 31.35,Richtler et al.2000).There is also evidence from stellar population synthesis of integrated spectra that NGC1316hosts younger popu-lations(Kuntschner&Davies1998,Kuntschner2000)and thus the question arises,whether the surprisingly low spe-cific frequency is caused by a high luminosity rather than by a small number of clusters.Goudfrooij et al.(2000, hereafter Go00)obtained spectra of27globular clusters and reported that the3brightest clusters have an age of about3Gyr,indicating a high star-formation activity 3Gyr ago,presumably caused by the merger event.Thesefindings and the hope for a good distance via the GCLF were the main motivation to do the present study of the GCS of NGC1316in a larger area than that of the HST study.As we will show,this galaxy resembles in many aspects the“old”merger galaxy NGC5018,whose GCS has been investigated by Hilker&Kissler-Patig(1996).The paper is organized as follows:in Sect.2we dis-cuss the observations,the reduction procedure and the selection of cluster candidates.The photometric and mor-phological properties of the GCS are discussed in Sect.3. Sect.4contains ourfindings concerning the Specific Frequency.We conclude this work with a general discus-sion in Sect.5.2.Observations and ReductionThe B,V and I images were obtained at the3.6m tele-scope at La Silla during the nights29th and30th of December,1997(dark moon),using the ESO Faint Object Spectrograph and Camera,EFOSC2.Thefield of view was 5.′6×5.′6with a scale of0.′′32/pixel.During thefirst night, short-and long-exposures in eachfilter were centred on the galaxy.In the second night,a backgroundfield located about5′away from the centre of NGC1316was observed, overlapping by1′the observations of thefirst night.In addition,severalfields containing standard stars from the Landolt catalog(Landolt1992)were acquired in eachfil-ter,as well as some short exposures of NGC1316.Fig.1 shows the combined frames from both nights and Table1 summarises the observations.Table1.Summary of the observations.Dec.29,1997B4×6001.′′1Dec.29,1997V5×3001.′′0Dec.29,1997I6×3001.′′0Dec.30,1997B4×6001.′′3Dec.30,1997V3×6001.′′3Dec.30,1997I3×6001.′′2M.G´o mez et al.:The Globular Cluster System of NGC1316(Fornax A)3 Table2.General parameters of the target galaxy,from de Vaucouleurs et al.(1991)and Poulain(1988).NGC131603h22m41.s6−37◦06′10′′240.◦16−56.◦69(R’)SAB(s)08.53±0.080.861793±12filter A j1A j2rms of thefitture radii between22.′′9and86.′′6are: ∆V =0.013,∆B−V =0.007and ∆V−I =0.006mag,well below therms of thefit(see Table3).We then defined5local stan-dard stars in thefield of NGC1316to set the photometryof both nights in a consistent way.2.3.Selection criteriaSeveral criteria have been applied to select cluster can-didates,according to colours,magnitude,photometric er-rors,stellarity index and projected position around thegalaxy.We assume that the clusters are similar to theMilky Way clusters.Adopting an absolute turnover mag-nitude(TOM)of V=−7.60for the galactic clusters(Drenkhahn&Richtler1999,Ferrarese et al.2000)andµ=31.35(Richtler et al.2000),the TOM of NGC1316isexpected to be V∼23.7mag,and the brightest clustersabout V∼20.Although no reddening corrections are normally ap-plied when looking towards Fornax(Burstein&Heiles1982),we cannot restrict the colours of the clusters inNGC1316to match exactly the galactic ones.One pointis the smaller sample of galactic clusters.Besides,we mustallow for significant photometric errors of faint cluster can-didates.Fig.2shows a colour-magnitude diagram for allobjects detected simultaneously in B,V and I in bothnights(before the selection),together with the cut-offval-ues adopted as criterium for this colour.As can be seen,the majority of objects have colours around V−I=1.0.Objects bluer than0.5mag are very probably foregroundstars.A fraction of the data points redder than V−I=1.6are background galaxies.We also tested in our images the robustness of the“stellarity index”computed by SExtractor.This indexranges from0.0(galaxy)to1.0(star)and varies for thesame object by about0.2when classifying under differentseeing conditions,except for the brightest objects,whichare clearly classified.By visual inspection of the images,we are quite confident that bright galaxies are always givenindices near to zero.However,this classification becomesprogressively more difficult with fainter sources.4M.G´o mez et al.:The Globular Cluster System of NGC 1316(Fornax A)−0.50.51 1.522.5V−I161820222426V Fig.2.A colour-magnitude diagram for all objects de-tected in B ,V and I in both nights before the selection criteria (911points).The dashedlines indicates the selection criterium in the V −I colour (see Sect.2.3.)V0.00.20.40.60.81.0s t e l l a r i t y i n d e xFig.3.The stellarity index computed by SExtractor as function of the V magnitude.The dashed lined indicates the cut value adopted in the selection criteria to reject background galaxies.Fig.3shows the “stellarity index”for the 911objects in our sample (before the selection criteria)as a function of the magnitude.Two groups of objects having indices of ∼0.0and ∼1.0can be seen,but it is apparent that faint clusters cannot be unambiguously distinguished from background galaxies due to the uncertainty of the stellar-ity index in the case of faint sources.Very similar results were obtained with our “artificial stars”(see Sect.2.5),where indices down to 0.2were measured for faint objects that are constructed using the PSF model and,therefore,are expected to have a stellarity index of ∼1.0.Guided by our experience with artificial stars,we de-fined the cut-offvalue for the stellarity index to be 0.35.Remaining galaxies will be statistically subtracted be-cause the same criteria are applied to the background field (see Sect.2.4).Finally,we rejected objects with photometric errorlarger than 0.15mag.To summarise,the following criteria were applied to our sample of 911sources detected in B ,V and I :i)V >19.5ii)0.4<B −V <1.4iii)0.3<V −I <1.8iv)stellarity index >0.35v)error(V ),error(B −V ),error(V −I )<0.15375objects met this set of criteria and are our globular cluster candidates.2.4.Background correctionAfter the selection,there might still be some contamina-tion by foreground stars and background galaxies in our sample.To subtract them statistically,one needs to ob-serve a nearby field,where no clusters are expected,and to apply the same detection and selection criteria as with the galaxy frame.However,the analysis of our background field still shows a concentration towards the galaxy centre,which means that only part of the field can be considered as background.We used the radial profile of the GC surface densi-ties (see Sect.3.3)to select all objects on the flat part of the profile,i.e.,where the number of globular cluster can-didates per area unit exhibits no gradient.Fig.11(top)shows that this occurs at r ≈300′′,where r is the distance from the optical centre.Thus,all objects with galactocen-tric distances larger than 300′′and matching the above criteria,constitute our background sample.We constructed a semi-empirical luminosity function of the background clusters with a technique described by Secker &Harris (1993),where a Gaussian is set over each data point,centred at the corresponding observed magni-tude.The sum of all Gaussians gives a good representation of the background,without introducing an artificial undu-lation or loss of information due to the binning process.Fig.4shows the histogram of the background objects,us-ing a bin size of 0.4mag,and the adopted semi-empirical function,which we use as background in the calculation of the luminosity function (see Sect.3.4)and the specific frequency.Admittedly,the numbers are small.The dip at V =23may be simply a result of bad statistics.On the other hand,as Table 6shows,the background counts are small compared to the clusters counts.Therefore,errors of the order of the statistically expected uncertainty do not significantly influence our results and are accounted for in the uncertainties of the total counts in Table 6.pleteness correctionTo correct statistically for completeness,one needs to de-termine which fraction of objects are actually detected by the photometry routines.These ‘artificial stars ex-periments’were performed using the task addstar inM.G´o mez et al.:The Globular Cluster System of NGC 1316(Fornax A)52021222324V05101520N Fig.4.The semi-empirical luminosity function for the background field (dashed line).A histogram with a bin size of 0.4mag is over-plotted for comparison.DAOPHOT.In one step,100stars were added in the sci-ence frames from V =20to V =25in steps of 0.1mag,distributed randomly to preserve the aspect of the image and not to introduce crowding as an additional param-eter,but with the same coordinates in the B ,V and I frames every loop.The colours of the stars were forced to be constant B −V =0.7and V −I =1.0,close to the mean colours of the globular clusters in NGC 1316(see Sect.3.1).The object detection,photometry,classi-fication and selection criteria were applied in exactly the same way as for the globular cluster candidates.To get sufficiently good statistics,we repeated the whole proce-dure 10times,using different random positions in each of them and averaging the results.In all,50000stars were added.However,the completeness is not only a function of the magnitude.Due to the remaining noise after galaxy subtraction,the probability of detecting an object near the centre of the galaxy is smaller than in the outer parts.We divided our sample of artificial stars into elliptical rings,in the same way as we did in deriving the radial profile (see Sect.3.3).The results of the completeness tests are summarised in Fig.5.It can be seen that the 50%limit goes deeper with increasing galactocentric distance.3.Photometric and morphological properties 3.1.Colour distributionIn this section we discuss the colours of our GC candidates.No interstellar reddening is assumed (Burstein &Heiles 1982)and therefore,only internal reddening might affect the colours.The histograms in Fig.6show the colour distribution of the cluster candidates.A fit of a Gaussian function with free dispersions returned the values listed in Table 4and is overplotted for comparison.As can be seen,the dispersion in the B −V and V −I histograms is completely explained by the photometric er-rors alone.This is not the case for the B −I histogram,due probably to the greater metallicity sensitivity of thisV0.00.20.40.60.81.0c o m p l e t e n e s spleteness factors in four elliptical annuli.The probability of detection strongly decreases near the centre of NGC 1316.Table 4.Fit of a Gaussian function to the colour his-tograms of the GC candidates.B −V 0.80±0.020.13±0.04V −I 0.95±0.010.17±0.02B −I 1.77±0.020.35±0.036M.G´o mez et al.:The Globular Cluster System of NGC 1316(Fornax A)0.00.51.0 1.52.0B−V0204060N0.00.51.0 1.52.0V−I0204060N0.81.31.82.3 2.8B−I010203040N Fig.6.Colour histograms of the cluster candidates.The bin size is 0.05mag in each colour.A Gaussian function has been fit to the histograms (dashed line,see text).The dot-dashed line is a Gaussian with σ=0.15.The long-dashed line at B −I =1.75(lower panel)was set to divide the sample in red and blue clusters.the clusters are on average ∼0.2mag bluer than the un-derlying galaxy light.This is a property commonly found also in normal early-type galaxies,where the blue and presumably metal-poor clusters indicate the existence of−3.0−2.0−1.00.0 1.0[Fe/H]1020304050NFig.7.Metallicity histogram of the GCS of NGC 1316,from its B −I broadband colour.The galactic system (dashed line)is overplotted for comparison.The long-dashed line near [Fe /H]=−1.0divides the sample in metal-rich and metal-poor clusters,using the calibration from Couture et al (1990)and B −I =1.75(see text and Fig.6.)a faint metal-poor stellar (halo?)population,as the case of NGC 1380suggests (Kissler-Patig et.al 1997).However,the non-existence of a colour gradient is consistent with the finding that blue and red clusters have similar surface density profiles (Sect.3.3).Differential reddening caused by the irregular dust structure might affect the width of the colour distribution,but apparently does not produce any colour gradient.3.2.Angular distributionFor all objects in our sample of clusters,a transformation from cartesian (x,y )coordinates to polar (r,θ)was done,with origin in the optical centre of NGC 1316.An offset of 50◦in θwas applied to match the PA of the galaxy quoted by RC3.In this way,θ=0represents the direction of the semi-major axis (sma)of the galaxy,a ,and θ=90◦the direction of b ,the semi-minor axis.To analyse the angular distribution,we rejected ob-jects inside a radius of 150pixels (corresponding to 4.3kpc with µ=31.35),where the completeness is significantly lower (see Fig.5)and some clusters appear over ripples and dust structures.Objects outside of 450pixels were also rejected,as one needs equally-sized sectors to do this analysis,and r =450is roughly the radius of the largest circle fully covered by our frame (see dotted line in Fig.10).Only candidates brighter than V =23.8(the 50%completeness level in the entire frame)were considered.The data were then binned in θ.Several bin sizes from 18◦to 30◦were tested,and Fig.9shows the histogram for a bin size of 22.◦5,which corresponds to dividing the sample into 16sectors.The bins were taken modulo πfor a better statistics,that is,we assume that the distribution of clusters is symmetric along the semi-mayor axis of the galaxy.Due to the substructures present in NGC 1316,andM.G´o mez et al.:The Globular Cluster System of NGC 1316(Fornax A)71.21.4 1.61.82.0 2.2 2.4log r (arcsec)1.02.03.0B −I1.02.03.0B −IFig.8.The B −I colour gradient along the projected galactic radius.Top:the cluster candidates and a least-square fit(solid line).The crosses indicate the B −I colour of the galaxy.Bottom:the mean colour in several rings,with the error bars indicating the σof the mean at that ring.090180270360θ [deg]10203040N Fig.9.The angular distribution of globular clusters for a radius from 150to 450pixels (48′′to 144′′)down to V =23.8.The bin size is 22.◦5and the data were taken mod π.The histogram from 0◦to 180◦is repeated from 180◦to 360◦for a better visualisation.The dashed line indicates the best fit of a double-cosine function.from the fact that it is a merger galaxy,one could expect some systematic differences in the azimuthal distribution of the clusters between both halves.Although this is not observed at any bin size,the small number of counts does not allow us to address this question clearly.The histogram shown in Fig.9demonstrates the strong correlation of the globular clusters with the galaxy light.By fitting isophotes to the galaxy,we obtained ellipticities ranging from 0.27to 0.32and position angles from 49◦to53◦,between semi-major axes of 150to 450pixels (thesame used with the clusters).From the least-square fit to this histogram (with fixed period π),we derived PA =63◦±9◦and an ellipticity of 0.38±0.06.There are,however,some problems that cannot be eas-ily resolved with our ground-based data.As indicated,the detection of cluster candidates near the centre of the galaxy is quite poor.This is,unfortunately,the most in-teresting region to search for young clusters which might be related to a merger event.3.3.Radial profileTo derive the radial surface density of GCs,we divide our sample into elliptical annuli,as shown in Fig.10.The annuli have a width of 100pixels along the major axis (a )and start from a =50pixels.Due to the small number of objects in the periphery of NGC 1316,rings beyond a =900were given a width of 300pixels.The ellipticity,position angle and centre of these rings were taken from the fit of the galaxy light (see Sect.3.2)and fixed for all annuli.In particular,the ellipticity e ,defined as 1−(b/a ),where a and b are the semi-major and semi-minor axis respectively,was set to 0.3and the PA to 50◦.We then counted the number of globular cluster can-didates per unit area in each ring,down to V =23.8,and corrected them with the corresponding completeness function.The results are presented in Table 5.The first column lists the mean semi-major axis of the ring.Column 2gives the raw number of clusters down to V =23.8,without correction for completeness.Column 3lists the corrected data with their errors.Column 4lists the visible area of the rings,in ⊓⊔′.Column 5gives the number of candidates per unit area.Column 6,7and 8are used to compute the Specific Frequency (see Sect.4).Fig.11shows the radial profile of the clusters’surface density,before and after the subtraction of the background counts.A fit of a power-law ρ(r )=A ·r α,where ρis the surface density and r the projected distance along the semi-major axis,gives αgcs =−2.04±0.20and αgal =−2.03±0.02for the clusters and the galaxy light,respectively.We are well aware that a King profile may be more adequate than a power function,but our purpose is to compare our result with previous work in other GCSs in Fornax,which quote only power functions.The similarity between both slopes indicates that the GCS of NGC 1316is not more extended than the galaxy light.Moreover,αGCS =−2.04is in good agreement with other GCS of “normal”early-type galaxies in Fornax (Kissler-Patig et al.1997).We divided the system at B −I =1.75mag into blue (presumably metal-poor)and red (metal-rich)clusters,and searched for systematic differences in the morpholog-ical properties between both subgroups.Again,only clus-8M.G´o mez et al.:The Globular Cluster System of NGC1316(Fornax A)Table5.This table gives the result of the radial profile of the cluster candidate surface density.Thefirst column lists the centre of each annulus(in pixels).Then follows the raw number counts down to V=23.8,before and after the correction for completeness.The fourth column gives the visible area of the annulus in⊓⊔′.Column5lists the mean density of GC per⊓⊔′.Columns6and7give the number of clusters down to the TOM in V,before and after the correction for completeness.The applied geometrical corrections are listed in column8.Finally,the number of the clusters in each annulus,after doubling the counts around the TOM.Note that the corrected and total number of clusters for the innermost annulus(sma=100)are NOT derived using the completeness correction.Instead,they were estimated by extrapolating the radial profile towards the centre(see text in Sect.4).1002058.6±18.1 1.25146.86±14.4720200.2±37.5 1.400±7520053100.3±17.9 2.50240.09±7.1553100.7±18.0 1.201±3630068104.4±15.2 3.75127.84±4.0569109.4±15.8 1.219±324005465.4±10.7 5.00613.06±2.135668.9±11.0 1.134±225004858.2±10.0 6.2349.34±1.605162.0±10.40.997124±216003742.9±8.5 6.1596.96±1.223744.5±8.80.821108±217001922.4±6.2 5.1454.35±1.211922.8±6.30.58878±219001923.2±6.49.4402.45±0.68————12001112.7±4.67.1591.77±0.64————150045.3±3.2 2.8801.84±1.11————M.G´o mez et al.:The Globular Cluster System of NGC 1316(Fornax A)9Fig.10.The elliptical annuli used in the calculation of the globular cluster density and specific frequency.The open circles indicate the position of all GC candidates brighter than V =23.8.Objects outside the annulus defined by r =150and r =450pixels (dotted line)were rejected for the analysis of the angular distribution (see Sect.3.2).Objects outside the dashed ellipse are considered as background.The scale and orientation is the same as in the Fig.1.There are two additional columns,namely the number of blue and red clusters.They were calculated in the same way and will be discussed in Sect.5.Bin sizes of 0.3,0.4and 0.5mag were tested,and the results were always in good agreement with each other.Finally,we have chosen 0.5mag as our bin size from the appearance of the fit histogram,and the absence of undu-lations which are present for the other cases.We note,however,that our derived TOM is close to the limit of the observations,and the last bins are strongly affected by the completeness correction.Nevertheless,the similarity of the results using different bin sizes and cen-ters,and the robustness of the fit against skipping the last bin,is encouraging.100110100G C /a r c m i n2100r (arcsec)110100G C /a r c m i n2Fig.11.Top:the radial profile of the density of glob-ular cluster candidates.The dashed line indicates thevalue adopted for the background,corresponding to 2.0objects /⊓⊔′.Bottom:the radial profile after subtrac-tion of the background counts.The crosses represent the profile of the galaxy light,arbitrarily shifted.1.61.82.02.2 2.4log r (arcsec)−0.50.00.51.01.52.0l o g (G C /a r c m i n2)Fig.12.The radial profile of the GC surface density for the red (triangles)and blue population (squares).No sys-tematic difference is seen and both are equally concen-trated.Fig.14shows the three luminosity functions in B ,V ,I .For fitting the LF we chose t 5functions,which are of the form:t 5(m )=85πσt1+(m −m 0)210M.G´o mez et al.:The Globular Cluster System of NGC1316(Fornax A)Table6.The counts(in bins of0.5mag)used in the determination of the GCLF.Given are the V-magnitudes(or B,I,respectively)of the bin centers.Then follow the raw counts of the four elliptical annuli N i(see text)together with the corresponding completeness factors f i.The background counts are then listed as defined in Sect.2.4.The completeness factors of the fourth annulus have been used for the background as well.The total number of clusters is the sum of the four annuli minus the background,normalised to the same area(see text).Also given is the number of clusters for the blue and the red population separately,where the separating colour was B−I=1.75.19.500.7110.981 1.002 1.000.04.0±2.01.0±1.03.0±1.720.010.7110.983 1.001 1.000.06.4±2.72.4±1.74.0±2.020.510.6700.984 1.002 1.000.26.9±3.13.9±2.63.0±1.721.020.6260.983 1.003 1.00 1.012.6±5.01.4±3.311.2±3.821.520.5830.94120.9970.99 2.518.8±7.07.8±4.611.3±5.222.020.5650.84140.97110.98 4.622.2±8.814.4±6.98.0±5.422.570.40110.68320.95140.95 4.269.8±12.536.6±9.033.3±8.623.030.35180.54390.93140.90 1.794.2±12.944.3±8.749.9±9.523.520.1480.28520.71190.79 3.7127.2±19.965.5±15.661.7±12.124.000.0720.20320.2280.25 3.8145.1±38.848.8±24.896.5±28.124.500.0000.0040.0310.030.5———B N B1f1N B2f2N B3f3N B4f4N B bkg.N B total N B blue N B red18.500.7110.980 1.002 1.000.03.0±1.70.0±0.03.0±1.719.000.7110.984 1.001 1.000.06.0±2.51.0±1.05.0±2.219.530.6710.983 1.000 1.000.08.5±3.34.0±2.34.5±2.320.000.6350.983 1.005 1.000.511.7±4.13.2±2.58.5±3.320.540.5840.94110.9950.99 2.819.8±7.49.5±4.910.3±5.521.020.5660.84160.97100.98 3.826.7±8.59.3±5.417.4±6.621.540.4090.68350.95170.95 5.163.1±12.133.0±9.030.1±8.022.050.35180.54370.93120.90 1.994.9±13.539.0±9.255.9±9.822.510.1490.28460.71150.79 2.3115.0±17.962.0±13.953.4±10.923.010.0700.20380.22150.25 4.3199.4±45.4130.6±36.468.8±24.623.500.0210.0030.0300.03 1.4———。

a rXiv:as tr o-ph/1712v12J ul21Astrophysical Ages and Time Scales ASP Conference Series,Vol.TBD,2001T.von Hippel,N.Manset,C.Simpson Galaxy Deconstruction:Clues from Globular Clusters Michael J.West Dept.of Physics &Astronomy,University of Hawaii,Hilo,HI 96720Abstract.The present-day globular cluster populations of galaxies re-flect the cumulative effects of billions of years of galaxy evolution via such processes as mergers,tidal stripping,accretion,and in some cases the partial or even complete destruction of other galaxies.If large galaxies have grown by consuming their smaller neighbors,or by accreting ma-terial stripped from other galaxies,then their observed globular cluster systems are an amalgamation of the globular cluster systems of their pro-genitors.Careful analysis of the globular cluster populations of galaxies can thus allow astronomers to reconstruct their dynamical histories.1.Introduction The origin of galaxies is one of the great outstanding problems in modern astro-physics.How and when did galaxies form?How have they evolved over time?How does environment influence their properties?One way to unravel the secrets of galaxy formation is by studying their globular cluster populations.Most galaxies possess globular cluster systems of various richness,ranging from dwarf galaxies with only a handful of globulars,to supergiant elliptical galaxies with tens of thousands of globulars surroundingthem (see Harris 1991or van den Bergh 2000for reviews).Because globular clusters are among the oldest stellar ensembles in the universe,they can provide important clues about the formation of their parent galaxies.The earliest studies of globular clusters were,by necessity,limited to our own Galaxy and its nearest neighbors.Over the past few decades,however,there has been tremendous progress in our understanding of globular cluster systems of other galaxies.One of the most important recent discoveries in the study of extragalactic globular cluster systems is that most large galaxies appear to possess two or more chemically distinct globular cluster populations (e.g.,Gebhardt &Kissler-Patig 1999;Forbes &Forte 2000;Kundu &Whitmore 2001).Some examples are shown in Figure 1,where two peaks are seen in the distribution of globular cluster metallicities for clusters associated with four large elliptical galaxies.A number of different theories have been proposed to explain the origin of these bimodal globular cluster metallicity distributions.An obvious way to1Galaxy Deconstruction2Figure1.The observed metallicity distribution of globular clustersystems associated with four giant elliptical galaxies.Note the presenceof two distinct peaks in most cases.The majority of large ellipticalgalaxies studied to date exhibit bimodal globular cluster metallicitydistributions.generate two or more chemically distinct globular cluster populations in galaxies would be through two or more bursts of globular cluster formation.This might occur,for example,if mergers of gas-rich galaxies trigger the formation of new globulars(Schweizer1987;Ashman&Zepf1992),resulting in the birth of mul-tiple generations of globular clusters.Similarly,one might envision a multiphase galaxy collapse model in which the metal-poor globular clusters formed during the initial collapse of a protogalactic gas cloud,and the metal-rich globulars formed some time later(Forbes,Brodie&Grillmair1997;Larsen et al.2001). In both of these scenarios,the metal-poor globular clusters surrounding galaxies such as those shown in Figure1would be their original population that formedGalaxy Deconstruction3 from low-metallicity gas at early epochs,and the metal-rich globulars would have formed more recently from gas that was enriched by stellar evolution.Alternatively,bimodal or multimodal globular cluster metallicity distribu-tions could also arise quite naturally from galaxy mergers and/or accretion of globulars stripped from other galaxies without needing to invoke the formation of multiple generations of globulars(Cˆo t´e,Marzke,&West1998;Cˆo t´e,Marzke, West&Minniti2000).Motivation for this model came from a simple fact:for those elliptical galaxies that exhibit a bimodal globular cluster metallicity dis-tribution,the metallicity of the metal-rich peak shows a clear correlation with parent galaxy luminosity,in the sense that the most luminous galaxies have the most metal-rich globulars(Forbes,Brodie&Grillmair1997;Forbes&Forte 2001).However,no such correlation is seen for the metal-poor peak,it appears to be largely independent of parent galaxy luminosity.To my collaborators and I,this suggests that the metal-rich globular clusters are innate to large ellipticals, and the metal-poor ones were added later either through mergers or accretion.2.Globular clusters as diagnostics of galaxy mergersThere is no doubt that galaxy mergers have occurred frequently throughout the history of the universe.Figure2shows an image of a supergiant elliptical galaxy in which the partially digested remains of several smaller galaxies are still clearly visible.Many large galaxies today may have grown to their present sizes by devouring smaller companions.If so,what becomes of the globular cluster populations of the galaxies that were consumed?There is also evidence of ongoing galaxy destruction in rich clusters,and countless galaxies may have met their demise over a Hubble time(e.g.,Gregg &West1998;Calcaneo-Rodin et al.2000).An example is shown in Figure3. Because they are dense stellar systems,globular clusters are likely to survive the disruption of their parent galaxy,and will accumulate over time in the cores of rich galaxy clusters.The ongoing destruction of the moderate-sized elliptical galaxy shown in Figure3,for example,will likely strew several hundred globulars into intergalactic space.Some of these may eventually be incorporated into other galaxies,a sort of recycling on cosmic scales(Muzzio1987).In particular,giant elliptical galaxies at the centers of rich galaxy clusters,which are observed to have enormously rich globular cluster populations,may have inherited myriad intergalactic globular clusters(West et al.1995).If large galaxies have grown by consuming smaller neighbors or by accreting material torn from other galaxies,then their present-day globular cluster systems are an amalgamation of the globular cluster systems of their victims.Cˆo t´e et al. (1998)and Cˆo t´e et al.(2000)showed that the growth of large galaxies through mergers or accretion will invariably be accompanied by the capture of metal-poor globulars,resulting in bimodal(or even multi-modal)metallicity distributions that are strikingly similar to those see in Figure1.Our prescription for building a large elliptical galaxy with a bimodal glob-ular cluster metallicity distribution is remarkably simple:Galaxy Deconstruction4Figure2.The brightest elliptical galaxy in the cluster Abell3827. Several smaller cannibalized galaxies are clearly evident in the central regions.Globular clusters belonging to these galaxies are likely to survive the eventual disruption of their parent galaxies,and thus will become part of this giant elliptical.If most large elliptical galaxies have grown by cannibalizing smaller neighbors,then their globular cluster populations today are composite systems that can provide information about the progenitor galaxies.•We assume that galaxies obey a Schechter-like luminosity function,as is observed.This sets the relative numbers of galaxies of different luminosi-ties that are available for merging.•We assume that each galaxy is born with its own intrinsic globular cluster population,and that the number of globulars per unit galaxy luminosity is constant,which is consistent with observations(Harris1991).ThisGalaxy Deconstruction5 determines how many globular clusters each galaxy has available to donate during mergers.•We assume,again from observations,that the mean metallicity of a galaxy’s original globular cluster population increases monotonically with parent galaxy luminosity(Cˆo t´e et al.1998).Smaller galaxies have metal-poorer globulars on average than larger galaxies.Figure3.A tidally disrupted galaxy in the Coma cluster(from Gregg&West1998).The top panel shows the raw image,and the bottom panel has been cleaned of foreground objects to highlight the∼150kpc long plume of material.The partial,or in some cases complete, disruption of galaxies in dense environments will create a population of intergalactic stars and globular clusters.These freely roaming globulars may be accreted later by other galaxies.Galaxy Deconstruction6 Beginning with a medium-sized elliptical galaxy as a seed,we allow it to consume its smaller neighbors at random,stopping after enough mergers have occurred to yield a large elliptical.We assume that globular cluster numbers are conserved during mergers,so the larger galaxy gains the globulars from the smaller galaxies that it consumed.Figure4shows some results of Monte Carlo simulations based on the Cˆo t´e et al.(1998,2000)model.Our simulations indicate that80to90%of large ellip-tical galaxies formed in this way exhibit bimodal(or in some cases multimodal) globular cluster distributions.The locations of the metal-rich and metal-poor peaks also agree well with observations(compare Figures1and4).In our model,Figure4.Results from some simulations based on the dissipationlessmerger model of Cˆo t´e,Marzke&West(1998).These simulations showthat bimodal globular cluster metallicity distributions are easily pro-duced by dissipationless galaxy mergers and accretion,without needingto posit the formation of multiple bursts of globular cluster formation.Galaxy Deconstruction7 the metal-rich globular clisuters of large ellipticals belonged to the progenitorgalaxy seed,and the metal-poor globulars were inherited from the many smaller galaxies that it consumed during its growth,or by accretion of intergalacticglobulars that were torn from other galaxies.The globulars gained from merg-ers or accretion are predominantly metal-poor because they originate mostly in low-mass galaxies.It is noteworthy that unimodal globular cluster metallicity distributions also occur from time to time in our model.An example can be seen in Figure4.This is not surprising,given the stochastic nature of the merger process.Forinstance,a large elliptical could in principle be built by merging many small dwarf galaxies(resulting in the globular cluster metallicity distribution of thefinal merger remnant exhibiting a single metal-poor peak),or by merging twoor three medium-sized galaxies(which might lead to a single metal-rich peak), or by merging galaxies over a wide range of luminosities(which yields bimodalor multimodal globular cluster metallicity distributions).Unfortunately,our model seems to often be misunderstood or misrepre-sented in the literature.For example,claims that the Cˆo t´e et al.(1998,2000)model is ruled out by the discovery of bimodality in some low-luminosity ellipti-cals(e.g.,Forte et al.2001;Kundu&Whitmore2001)are completely unfounded;Cˆo t´e et al.(2000)demonstrated that it is quite straightforward to reproduce the bimodal globular cluster metallicity distribution of our own Galactic spheroid(which has M V≃−19.9)as a consequence of accretion and mergers,and hence the same should be true for intermediate-and low-luminosity elliptical galaxies. Similarly,recent assertions that the location of the metal-poor globular clusterpeak also correlates weakly with parent galaxy luminosity is not“hard to explain within the accretion/merger pictures”(Larsen et al.2001).A careful reading of the original Cˆo t´e,West&Marzke(1998)paper would show that in fact we predict this very result;see Section3.1.2of that paper,which states“there is a slight tendency for the brighter gE’s to capture globular cluster populations that are more metal-rich than those accreted by their fainter counterparts(since the brighter gE’s are able to accommodate the capture of more luminous intruder galaxies).”If our model is correct,then it offers the exciting possibility of placing some quantitative constraints on the number and types of mergers that galaxies have experienced over their lifetimes by comparing the relative numbers of metal-rich and metal-poor globulars that they possess.Cˆo t´e et al.(1998)used this rea-soning to conclude that M49,the most luminous elliptical galaxy in the Virgo cluster,must have gained roughly2/3of its present luminosity by consuming other smaller Virgo galaxies.More recently,we have applied these same tech-niques to understanding the formation of our own Milky Way galaxy.Cˆo t´e et al. (2000)showed that the present-day globular cluster system of the Milky Way strongly suggests that the Galaxy’s spheroid was assembled from a large number of metal-poor protogalactic fragments.Hence even a relatively low luminosity system like the Milky Way spheroid can and does possess a bimodal distribution of globular cluster metallicities.Galaxy Deconstruction8 3.Where do we go from here?Clearly the competing theories for the origin of bimodal globular cluster metallic-ity distributions make quite different predictions regarding the ages of globulars. If the metal-poor and metal-rich globulars surrounding large elliptical galaxies are the result of multiple bursts of cluster formation,then the two populations should have quite different ages.If,on the other hand,bimodal metallicity dis-tributions can be explained by dissipationless merging as described above,then the metal-poor and metal-rich globulars should all be old.Recently,Beasley et al.(2000)measured ages and metallicities of globu-lar clusters in M49,and concluded that(within the sizeable uncertainties)the metal-poor and metal-rich populations are coeval and old.This clearly seems to support the Cˆo t´e et al.(1998,2000)picture.However,more precise data are needed to reduce the uncertainties beforefirm conclusions can be drawn.With that goal,we have obtained Hubble Space Telescope observations of∼103globu-lar clusters associated with the giant elliptical galaxy M87in order to accurately determine their ages using a powerful narrow-band photometry technique.The hypothesis that galaxies might accrete substantial numbers of inter-galactic globular clusters also needs to be tested with direct observations of these objects to determine if they really exist and in what numbers.Unlike many theories for the origin of globular cluster populations,the intergalactic globulars hypothesis is easily falsifiable;if a significant population of intergalac-tic globulars is not detected in the cores of galaxy clusters,then this idea will have to be abandoned.My collaborators and I are currently analyzing HST, Keck,Subaru and CFHT images that we obtained to search for intergalactic globulars in the Virgo,Coma and Abell1185galaxy clusters.There is also work to be done on the theoretical front.The simple merger model described here is admittedly somewhat naive.For example,we assumed equal merger probabilities for all galaxies,when in reality there is likely to be some mass dependence of merging.As a step towards more realistic models, Frazer Pearce and I are collaborating in a study of the merger histories of galaxies that form in very high-resolution N-body cosmological simulations.By following the detailed merger histories of galaxies from high redshifts to the present,and inputting simple models of globular cluster formation at different epochs,we will be able to make quantitative predictions regarding the evolution of globular cluster metallicity distributions in galaxies as a function of time.ReferencesAshman,K.M.&Zepf,S.E.1992,ApJ,384,50Beasley,M.A.,Sharples,R.M.,Bridges,T.J.,Hanes,D.A.,Zepf,S.E.,Ashman, K.M.&Geisler,D.2000,MNRAS,318,1249Calc´a neo-Rold´a n,C.,Moore,B.,Bland-Hawthorn,J.,Malin,D.&Sadler,E.M.2000,MNRAS,314,324Cˆo t´e,P.,Marzke,R.O.,&West,M.J.1998,ApJ,501,554Cˆo t´e,P.,Marzke,R.O.,West,M.J.,&Minniti,D.2000,ApJ,533,869 Forbes,D.A.,Brodie,J.P.,&Grillmair,C.J.1997,AJ,113,1652Galaxy Deconstruction9 Forbes,D.A.,&Forte,J.C.2001,MNRAS,322,257Gebhardt,K.&Kissler-Patig,M.1999,AJ,118,1526Gregg,M.D.&West,M.J.1998,Nature,396,549Harris,W.E.1991,ARA&A,29,543Kundu,A.,&Whitmore,B.2001,AJ in press(astro-ph/0103021)Larsen,S.S.,Brodie,J.P.,Huchra,J.P.,Forbes,D.A.,&Grillmair,C.2001,AJ, in press(astro-ph/0102374)Muzzio,J.C.1987,PASP,99,245Schweizer,F.1986,in Nearly Normal Galaxies,ed.S.Faber(New York:Springer), 18van den Bergh,S.2000PASP,112,932West,M.J.,Cˆo t´e,P.,Forman,C.,Forman,W.&Marzke,R.O.1995,ApJ,453, L77Acknowledgments.I wish to thank my principal collaborators in this re-search,Pat Cˆo t´e,Michael Gregg,Ron Marzke,and Frazer Pearce.This work was supported by NSF grant AST00-71149.Galaxy Deconstruction10 Discussion。