The Rich Globular Cluster System of Abell 1689 and the Radial Dependence of the Globular Cluster Formation Efficiency

Home , 365 Corduba

arXiv:1308.1958v1 [astro-ph.CO] 8 Aug 2013 eLcae l 08 ontne l 2008). 2006 al. et al. Johnston et 2008; Font al. (see et later Lucia accreted De origina were likely and satellites densities dwarf lower surviving et kinem the Abadi and contrast, distributions 2006; In ea spatial al. their the including et of 2006), (Moore al. tracers substructures reliable star-forming provide (G est light clusters stellar globular diffuse metal-poor and old si that numerical indicate high-resolution envi- dy- ulations However, dense early the these histories. about in information namical longest accretion much the and obscured has for merging ronments assembling rapid self been The have largest time. and the formation earliest galaxies, the star saw of today, jor Luc universe clusters the De Massive in 2005; systems gravitating al. 2006). progres et al. form Springel et to (e.g., coalesce inst structures and larger gravitational sively regions through high-density in forming bility begin objects galactic H IHGOUA LSE YTMO BL 69ADTERADIAL THE AND 1689 ABELL OF SYSTEM CLUSTER GLOBULAR RICH THE .A A A. K. vne ai,C 51,USA 95616, CA Davis, Avenue, aaa itra CVE27 Canada 2E7, V9E BC Victoria, Canada, uóoad éio oei 89,Mxc;[email protected] México; 58090, Morelia México, de Autónoma aóiad hl,7246Mcl atao Chile Santiago, Macul, 7820436 Chile, de Católica atao Chile Santiago, hsc,Pkn nvriy ejn 081 China 100871, Beijing University, Peking physics, fWsenAsrla 5Siln iha,Caly A6009 WA Crawley, Highway, Stirling 35 Australia, Western of USA rpittpstuigL using typeset Preprint nteuulheacia tutr omto oe,pre- model, formation structure hierarchical usual the In D ATVERSION RAFT 4 3 2 1 5 8 7 6 eatmnod srnmayAtoíia otfii Unive Pontificia Astrofísica, y Astronomía de Departamento eateto hsc,Uiest fClfri,Dvs On Davis, California, of University Physics, of Department ezegIsiueo srpyis ainlRsac Cou Research National Astrophysics, of Institute Herzberg etod aiatooí srfsc,UiesddNaci Universidad Astrofísica, y Radioastronomía de Centro nentoa etefrRdoAtooyRsac,TeUni The Research, Astronomy Radio for Centre International ai icelOsraoy etlSre,Nnukt MA Nantucket, Street, Vestal 4 Observatory, Mitchell Maria uoenSuhr bevtr,Aos eCroa30,Vit 3107, Córdova de Alonso Observatory, Southern European eateto srnm al nttt o srnm and Astronomy for Institute Kavli & Astronomy of Department LAMO adas erahmagnitude reach we bandpass, ( ytm suigtewl-nw asinfr fteG lumi GC the of of form population Gaussian well-known the Assuming system. nomnfrcnrlglxe nmsiecutr,rsn f rising clusters, massive in galaxies central for uncommon asv aayfdn ol increase would fading galaxy Passive per iia o ifrn lses eseuaeo the on headings: speculate not Subject we is clusters; efficiency different formation for GC typic similar The appears are radii. syste efficiency, physical this formation same GC of the the at nature interme thus appears extreme and profile the masses, mass Despite total GC mass. the of cluster shape total The Way. Milky the ais h siae asi GCs, a in mass, mass estimated total The lensing-derived radius. and gas, intracluster stars, aaees hsi yfrtelretG ytmsuidt da to studied system GC largest the far by uncertain is sizable the this with parameters, Even component. intracluster an z esuyterc lblrcutr(C ytmi h etro center the in system (GC) cluster globular rich the study We 0 = -M . 8,oeo h otpwru rvttoa esskon Wi known. lenses gravitational powerful most the of one 18), J ARTÍNEZ UNE A T E tl mltajv 5/2/11 v. emulateapj style X 1. ,2018 8, N INTRODUCTION GC total 1,2 lblrcutr:gnrl—glx lses individual( clusters: galaxy — general clusters: globular .P B P. J. , 162 = , LAKESLEE 850 + − 75 51 I 814 .R M R. G. , , 450 310 LSE OMTO EFFICIENCY FORMATION CLUSTER 9with 29 = 2 C ihnapoetdrdu f40kc smn shl a co may half as many As kpc. 400 of radius projected a within GCs .J J J. M. , M S N GC total EURER by Australia , =3.9 rf eso ue8 2018 8, June version Draft nam.mx EE ∼ Shields e & clof ncil e at ted 02554, versity 3 atics. Astro- 5 0 at 20% rsidad acura, × .C P. , .W P W. E. , ma- ABSTRACT 0 opeeesadsml h brightest the sample and completeness 90% onal Cs) rli- m- 10 a- ia - - ; 10 ÔTÉ M z ⊙ 0 ecntutterda aspolso h GCs, the of profiles mass radial the construct We =0. ENG 2 dw opr h asfatosa ucinof function a as fractions mass the compare we nd esn o hssmlrt nprofile. in similarity this for reasons to comparable is , rae.Cnesl,frmr uiosellipticals, luminous more for Conversely, creases. ∼ rlyhv uhsasrG ytm.Frerytp dwarfs, 2006), early-type For Strader systems. & S GC gen sparser Brodie spirals much as have 2004; galaxies, erally early-type al. in et commonly most West Har- though 2001; by reviews 1991, (see ris environments and types morphological ati l ag aaisadms wrswt asso at of masses with dwarfs most and 10 galaxies least large all in dant C e ntglx uioiy sanraie esr of measure normalized a as richness. luminosity) population galaxy unit per GCs ages have generally they hi en eit feryacein ars&vndnBerg den van frequency & specific Harris wit the Puzia consistent accretion. introduced 2003; 2012), early (1981) 1998, 2011, of relicts al. al. being et et their Chies-Santos Cohen 2005; (e.g., al. solar et of percent few a h oa Cpplto fe xed 000 n h centra the have and may 10,000, galaxy exceeds cD often population GC wh total clusters, fo massive the is of exception centers The dynamical the 2012). near al. galaxies et Alamo-Martínez (e.g., 2012; environments al. isolated et and clusters both in galaxies from 95.Baelee l 19)fudta h number the that found (1997) al. et Blakeslee 1995). ubrprui oa hl)ms hw uhls variation less much the shows Thus, mass (halo) luminosity. than total galaxy unit host per with number than c of rather indicators mass, other ter and veloc luminosity, cluster X-ray with dispersion, approximately ity scales systems such in GCs h xetdbro rcin ueia oesas predic also wit models consistent Numerical was factors fraction. scale baryon the expected of the (199 ratio Blakeslee the mass; that baryonic showed the with ellipticals massive .F L. , N is hc anyrsl rmteucranGCLF uncertain the from result mainly which ties, rom ,terto fteG ast h ayncand baryonic the to mass GC the of ratios the m, 0,adsm ednyt nraea h uioiyde- luminosity the as increase to tendency some and 100, C r seilyueu rcr eas hyaeabun- are they because tracers useful especially are GCs iia rnsof trends Similar 6 lo hs nohrrc lseswe comparing when clusters rich other in those of al xiisalresatr ihvle agn rmzr to zero from ranging values with scatter, large a exhibits .J W J. M. , osat u aiswt ais namne that manner a in radius, with varies but constant, S e h pcfi frequency specific The te. ERRARESE ∼ N S cagln(99 on iia cln of scaling similar a found (1999) McLaughlin . h asv lse fglxe bl 1689 Abell galaxies of cluster massive the f to 2 8 N oiyfnto GL) eetmt total a estimate we (GCLF), function nosity M it ewe hs fteselrlgtand light stellar the of those between diate ≈ ⊙ ∼ ertecne to center the near 5 EST eg,Pn ta.20) si h ik Way, Milky the in As 2008). al. et Peng (e.g., h28 th 0 n ed oices ihluminosity. with increase to tends and 10, 2 7,8 .A G A. R. , bl 1689) Abell ∼ HST S 0 ftettlselrms of mass stellar total the of 80% S EEDNEO H GLOBULAR THE OF DEPENDENCE S N N N ihlmnst cu o early-type for occur luminosity with a enmaue o aaiso all of galaxies for measured been has > ASobt nteF814W the in orbits /ACS & ONZÁLEZ 0(arse l 95 ete al. et West 1995; al. et (Harris 10 0Gradpa ealcte just metallicities peak and Gyr 10 ∼ S N -L 2a ag radii. large at 12 ∼ shg,btnot but high, is ÓPEZLIRA %o h GC the of 5% S N mprise tenme of number (the 1 .J A. , S ORDÁN N N N ranges GC GC Cho lus- ere of 4 9) in h h , - r - l t 2 Alamo-Martínez et al. an approximately constant frequency of old metal-poor GCs Teague et al. 1990; Girardi et al. 1997; Lemze et al. 2009). relative to total halo mass, depending on the degree of local It has one of the largest known Einstein radii, and its mass variation in the epoch of reionization (Moore et al. 2006). distribution has been extensively studied from both strong Peng et al. (2008) studied SN and the stellar mass frac- and weak lensing (Tyson & Fischer 1995; Taylor et al. 1998; tion contained in GCs for 100 early-type galaxies from the Broadhurst et al. 2005, Zekser et al. 2006; Limousin et al. Advanced Camera for Surveys (ACS) Virgo Cluster Survey 2007; Coe et al. 2010), as well as X-ray properties (An- (Côté et al. 2004). They found that the GC stellar mass frac- dersson & Madejski 2004; Lemze et al. 2008). A recent tion tends to be larger in giant and dwarf galaxies, but is uni- multi-wavelength analysis by Sereno et al. (2013), including versally low at intermediate masses. Such intermediate mass Sunyaev-Zeldovich,X-ray, and lensing data, finds a total mass 15 galaxies also appear to have been most efficient in convert- M200 = (1.3 0.2) 10 M⊙ within r200 =2.1 Mpc. ing baryons into stars (e.g., van den Bosch et al. 2007; Con- A1689 was± one× of the targets selected for early release ob- roy & Wechsler 2009; Guo et al. 2010). Thus, as previously servations with the ACS Wide Field Channel (ACS/WFC; found for central cluster galaxies, the SN and GC stellar mass Ford et al. 2002, 2003) after installation on the Hubble Space fraction for the ensemble of Virgo galaxies scale more closely Telescope (HST). In addition to the strong lensing analyses with halo mass (mainly dark matter) than with the stellar mass referenced above, the same observations were used to inves- or luminosity. The observed variations therefore indicate a tigate ultra-compact dwarf galaxies in an extremely rich en- variable field star formation efficiency following the GC for- vironment (Mieske et al. 2004). Further inspection of these mation epoch. Spitler & Forbes (2009) compiled a sample early ACS data revealed a concentration of faint point sources of galaxies with a wide range of masses in various environ- that were consistent with GCs at the distance of A1689. As- ments and also concluded that there was a direct proportion- suming the usual Gaussian GC luminosity function (GCLF) ality between the mass in GCs and the total halo mass. Fol- implied a huge population of & 105 GCs (Blakeslee 2005), lowing this idea, Georgiev et al. (2010) studied GC formation but the number was very uncertain, as it involved an extrapo- efficiencies within an analytical model that included mass- lation by two orders of magnitude. dependent feedback mechanisms. They found that the GC Here we report results from very deep HST/ACS obser- formation efficiency is roughly constant with halo mass, but vations of A1689 obtained in order to test the previous un- SN and stellar mass fraction vary with the field star formation certain results and study the GC population in the center of efficiency, which in turn depends on halo mass and is high- this extraordinary cluster. The following section summarizes est at intermediate values. In this model, the increased scat- the observations and image reductions. Section 3 describes ter among dwarf galaxies was attributed to stochastic effects the photometric analysis of the galaxies and GCs in detail, at low mass. However, Peng et al. had found that the Virgo while Section 4 presents our results on the GC number den- dwarfs with high GC mass fractions were nearly all within sity distribution, total population, and specific frequency. In 1 Mpc of M87, indicating that location plays a key role in Section 5, we compare our results on the GC and galaxy determining the relative fractions of baryons in GCs and field light distributions with the mass profiles of the X-ray emit- stars in these systems. This could also be a consequenceof the ting gas and dark matter. The final section discusses our formation epoch, with GC formation efficiency being higher conclusions. Throughout this work, we adopt the WMAP7 at early times (e.g., Kruijssen 2012). maximum likelihood cosmology (Komatsu et al. 2011) with In the Coma cluster, Peng et al. (2011) found a very large (h,Ωm,Ωλ)=(0.704,0.27,0.73),which yields a distance mod- population of intracluster globular clusters (IGCs), which are ulus for A1689 of (m−M)=39.74 mag and a physical scale of bound to the cluster potential, rather than to individual galax- 3.07 kpc arcsec−1 (These numbers change by < 0.5% for the ies. Limits on the amount of diffuse intracluster light (ICL) WMAP9 cosmology of Hinshaw et al. 2013). imply a high SN for the IGC population, similar to the values for high-SN cD galaxies and the centrally located Virgo dwarfs. IGCs appear to be a common feature of galaxy clus- 2. OBSERVATIONS AND IMAGE REDUCTION ters: a significant population likely resides in Abell 1185 As part of HST Program GO-11710, we imaged the central (Jordán et al. 2003; West et al. 2011), a cluster in which the field of A1689 for 28 orbits with the F814W bandpass (also cD galaxy is offset from the centroid of X-ray emission; Lee referred to as I814) of the ACS/WFC during seven visits from et al. (2010) also report evidence for IGCs in Virgo. The IGC 2010 May 29 to 2010 July 08. The charge transfer efficiency populations are overwhelmingly metal-poor, with colors typi- (CTE) of the ACS/WFC detectors has become significantly cal of GCs in dwarf galaxies and the outer regions of massive degraded in recent years. After standard Space Telescope Sci- galaxies. Because of their early formation and subsequent dis- ence Institute pipeline processing to the point of calibrated sipationless assembly, the spatial density profile of the IGCs “flt files,” we therefore used the stand-alone script version of is expected to trace the total cluster mass profile. However, the empirical pixel-based CTE correction algorithm of Ander- because of their low surface densities and contamination from son & Bedin (2010) on each of the 56 individual exposures in GCs boundto galaxies, it remains unclear whether they follow our program. This correction script did an excellent job of more closely the stellar light, the baryonic matter (including removing the CTE trails in the data. the X-ray gas), or the dark matter distribution. Observations The individual exposures were then processed with Apsis thus far have been limited to nearby (z . 0.05) moderate mass (Blakeslee et al. 2003) to produce a single geometrically cor- clusters. Further progress requires studying massive clusters rected, cosmic-ray cleaned, stacked image with a total ex- with many thousands of GCs and well characterized baryonic posure time of 75,172s. At nearly 21 hr, this is the single and total mass density profiles. deepest ACS/WFC image in the F814W bandpass. After ex- Abell 1689 (hereafter A1689) is an extremely massive perimenting with different Drizzle (Fruchter & Hook 2002) galaxy cluster at z =0.183, with a velocity dispersion σ & parameters within Apsis, we adopted the Gaussian interpola- 1400 km s−1, and complex kinematical substructure (e.g., tion kernel with a “pixfrac” of 0.5 and an output pixel scale of 0.′′033 pix−1. This set of parameters provides improved reso- Globular Clusters in A1689 3

and not well represented by the SExtractor background map. To remove these smaller (but resolved) structures, a very smooth image was created with the IRAF task rmedian (a ring median filter with inner and outer radii of 5 and 9 pix, respectively) applied to the final_residual image. This ring- median image was then subtracted from the final_residual to obtain the “rmed_residual” image (Figure 2). Although the rmed_residual image is extremely flat and suitable for point source detection, it cannot be used to measure reliable source magnitudes, as some flux is removed by the rmedian process. The source detection was performed with SExtractor on the rmed_residual image. In order to make the detection more robust, we used an RMS error image for the SExtractor detection WEIGHT map, produced as described by Jordán et al. (2004). This RMS map includes detector and photometric noise, as well as the signal-to-noise variations from the cor- rected bad pixels and cosmic rays. Bright stars, diffraction spikes, areas of lower exposure time near the image edges, and regions with large model residuals (due to sharp or ir- regular features within the cluster galaxies) were masked during the detection (black regions in Figure 3, right panel). At a luminosity distance of 885 Mpc, the GCs appear as point FIG. 1.— Deep ACS/WFC F814W image of the galaxy cluster A1689 from sources, and the SExtractor parameters were chosen to opti- Program GO-11710, shown in the observed orientation. The field-of-view is mize point source detection, using a threshold of 5 or more ∼3.′3×3.′3. lution with good sampling of the point spread function (PSF), connected pixels at least 1.5 σ above the local background. without introducing excessive small-scale correlations in the Source magnitudes were measured on the final_residual im- pixel noise. age with PSF photometry, using the SExtractor output co- We calibrated the photometry on the AB system, using ordinates. The PSF photometry is more accurate than the the F814W zero point of 25.947 determined from the time- various aperture magnitudes measured by SExtractor. The dependent ACS Zeropoint Calculator.9 The I -band Galac- PSF was constructed using the standard DAOPHOT (Stetson 814 1987) procedure. The important parameters were the FWHM tic extinction in the direction of A1689 is 0.04 mag (Schlafly ′′ & Finkbeiner 2011). of the PSF fwhmpsf=2.6 pix (0. 086), aperture=4 pix, and varorder=2, which means that the PSF is quadratically vari- 3. ANALYSIS able over the image. The best fitting function to describe the 10 3.1. Galaxy Modeling PSF was penny1. Since the radius of the PSF model is fi- nite, an aperture correction, mapcor, was estimated. To this As one of the few richness class 4 clusters in the Abell end, first the magnitude difference between 4 and 15 pixels catalogue (Abell et al. 1989), A1689 is exceptionally rich in ( 0.′′5) was measured, then a correction from an aperture of galaxies, particularly in its central region (Figure 1), a fact 0∼.′′5 to infinity was obtained from Sirianni et al. (2005). The that hampers source detection and photometry. In order to final aperture correction mapcor = −0.36 mag was applied to all have an image with the flattest background possible, we used the measured DAOPHOT fit magnitudes. the ellipse (Jedrzejewski 1987) and bmodel tasks within IRAF (Tody 1986, 1993) to construct isophotal models for 59 of the brightest galaxies in the ACS/WFC field. Neighboring galax- 3.3. GC Candidate Selection ies were masked during the fitting, and in cases with close The DAOPHOT parameters χDAO and sharp DAO, which companions, it was necessary to perform several iterations. gauge the goodness of the PSF fit and profile sharpness for The isophotal models for the 59 galaxies were combined to each object, together provide a good indication of whether an produce a single bmodel image, which was then subtracted object is a point source. Based on input and output parameters from the original image to create a first-pass residual im- of artificial stars constructed from the PSF, we selected point − age. SExtractor (Bertin & Arnouts 1996) was then run on this sources as objects with χDAO <5 and 0.9 < sharpDAO < 0.9. residual image to generate a map of the background due to Assuming that they are similar to nearby globular clus- the imperfect subtraction of the galaxies, as well as approx- ters, the GCs in A1689 should appear at I814 & 27.0 AB. imate representations of other, unmodeled, cluster galaxies. We therefore selected GC candidates as point sources having The SExtractor background map was then combined with the 27.0 < I814 < 29.32; the faint limit is the magnitude where our ellipse/bmodel models to produce our final luminosity model, detection is 50% complete in the innermost region of the clus- and this was subtracted from the original image to obtain what ter (see below). From this combined selection, we obtained a we refer to as the “final_residual” image. sample of 8212 GC candidates; Figure 3 shows their x,y positions, which are strongly concentrated near the cD galaxy, 3.2. Object Detection but excesses of GC candidates are also visible around other The final_residual image includes many smaller galaxies cluster galaxies. and residuals from larger galaxies that were difficult to model 10 An elliptical Gaussian core, that can be tilted at an arbitrary angle, with 9 http://www.stsci.edu/hst/acs/analysis/zeropoints. Lorentzian wings. 4 Alamo-Martínez et al.

x x

15"=46kpc

FIG. 2.— Zoom to the central region (∼ 160×160 kpc) of A1689. The red cross marks the center of the cD galaxy. Left panel: residual after subtraction of the luminosity model (bmodel plus SExtractor background map); because of its shallow light profile, the cD galaxy itself subtracts very well. Right panel: residual after subtraction of the luminosity model and the smooth rmedian image, used only for object detection. 3.4. Completeness relatively fewer bright sources to be scattered to fainter lev- To quantify the completeness, 250000 artificial stars were els. In this case, including both magnitude bias and incom- constructed from the PSF model and added 500 at a time with pleteness, the recovered fraction was not well described by random r,θ positions. The origin of the polar coordinates was Eq. (1). However, we found that it could be represented by the center of the cD galaxy, and the uniform random distri- the following modified version of the Fermi function: bution in r yielded a higher density of sources near the clus- + − F 1 C exp[b(m m0)] ter center, mimicking the actual sources. In adding the arti- f (m)= + − , (2) ficial sources, the masked areas were avoided, and the added 1 exp[a(m m0)] sources were not allowed to overlap with each other. The arti- where m0 is the magnitude at which the completeness would ficial stars were added to the rmed_residual and final_residual be 0.5 for a standard Fermi function (C 0). The other pa- images, and their fluxes measured with the same procedure rameters are linked, but in rough terms, a≡controls the steep- that was followed for the real objects, including the selection ness of the cutoff, b (which must be < a) affects where the based on the values of χDAO and sharp DAO. We carried out departure above unity begins, and C determines the amplitude this whole process twice, using two differentmagnitude distri- of the departure. butions for the artificial sources: (1) a uniform, or box-shaped, Although the Pritchet function and uniform magnitude dis- distribution with 25.0 < m < 30.5; (2) a Gaussian distribution tributions are widely used in the literature, our analysis indi- with mean µ =31.6, σ =1.5, and the constraint m < 30.5 (more cates that these assumptions require great caution: when the than a magnitude beyond the completeness limit). The latter actual counts follow a steeply rising luminosity function, in- case approximates the expected GCLF in A1689 (see Sec. 4); completeness corrections based on a uniform magnitude dis- both distributions are illustrated in Figure 4. tribution overestimate the number of real sources. We adopt In the case of the uniform magnitude distribution, the frac- the results from the artificial star tests with the Gaussian mag- tion of recovered stars as a function of magnitude is well de- nitude distribution, since this more closely approximates the scribed by a function of the form: expected GCLF. The magnitude distribution of background sources is also a rising function, more similar to the bright 1 α(m − m ) f P(m)= 1 − lim , (1) side of the Gaussian than to the uniform distribution. 2 2 2 " 1 + α (m − mlim) # The ratios of recovered to added artificial stars as a function of recovered magnitude were calculated, and a modi- where mlim is the magnitudep at which the completeness is fied Fermi function was fitted to the completeness fractions 0.5, and α determines the steepness of the curve. This func- in three annular regions: from 0 to 33′′, from 33′′to 1.′1, P ′ tion f is sometimes referred to as a “Pritchet function” (e.g., and beyond 1.1 (red circles in Figure 3). The values of m0 McLaughlin et al. 1994). found for these three regions are 29.30, 29.35, and 29.33 In the case of the Gaussian magnitude distribution, the frac- mag, respectively. After applying the completeness correction of recovered stars actually exceeds unity before the steep tions, the number of GC candidates brighter than our adopted drop in completeness sets in. This excess of detected sources limit I814 < 29.32 increases from 8212 to 8710 100, or 6%. is due to Eddington (1913) bias: as a result of the steeply ris- Had we used the completeness estimates based on± the uniform ing luminosity function, measurement errors cause more faint magnitude distribution, the increase would instead have been sources to be scattered to brighter detection magnitudes, and 14%, overestimating the final GC number by 7.5%. Globular Clusters in A1689 5

arcsec 100 50 0 50 100

100 6000

5000 50

4000

pix 0 3000 arcsec

2000 50

1000

100 0 0 1000 2000 3000 4000 5000 6000 pix

FIG. 3.— Left panel: spatial distribution of GC candidates; the red circles mark the boundaries of the three separate regions where the completeness function was ﬁtted, and the green dashed circle indicates a radius of 400 kpc (130′′). Right panel: black regions indicate the areas masked throughout the entire analysis; the gray regions show the masks applied for the galaxies, including the cD, used in estimating the number of background contaminants, bg (see Sect. 3.5).

6000 1.0

5000 0.8

4000

0.6 N 3000

recovered/added 0.4 2000

0.2 1000

0.0 0 25 26 27 28 29 30 26 27 28 29 30 31 32 I814 I814

FIG. 4.— Left panel: completeness functions for the two different input magnitude distributions. Histograms: magnitude distributions of the artificial stars, i.e., a Gaussian with mean µ =31.6 mag and σ =1.5 mag (blue), and uniform (pink). The blue points and pink open circles are the respective recovered fractions, and the lines show the respective best-fit completeness functions: Pritchet (dashed pink) and modified Fermi (solid blue). The rising Gaussian distribution, meant to mimic the actual GCLF, causes an excess of recovered point sources near the completeness limit because of Eddington bias. Right panel: observed luminosity function of GC candidates (histogram); fitted modified Fermi function × Gaussian model (solid blue line); and the derived Gaussian GCLF (dashed red line). The green arrow at I814 = 29.32 indicates the magnitude limit used for the fits. 6 Alamo-Martínez et al.

12000 bg=120 n=11.4 Re =587.3" r>15"

bg=191 n=3.9 Re =63.9" r>10"

bg=159 n=0.3 Re =69.9" r>48"

10000 bg=30 n=4(fix) Re =123.5" r>5" 1.00 800

8000 ] ] 600 2 2

0.10 6000 400 [arcsec [arcmin N N

200 4000

0 50 80 110 140 0.01 2000 no bg subtraction bg=160.0 80.0

bgD =160.0 *D 0 0 20 40 60 80 100 120 140 10 100 R [arcsec] R [arcsec]

FIG. 5.— Radial distribution of the surface number density of GC candidates. Left panel: number of GC candidates per arcmin2 (black dots). The dotted magenta, dashed green, and solid yellow lines show fitted Sérsic functions for different radial domains with the background (bg) as an additional free parameter. The blue dots are similar to the black ones, but obtained after all the bright galaxies have been masked; the blue line shows the corresponding Sérsic fit. The Sérsic parameters of the various fits are shown at top. The dashed black line and gray shaded region indicate the final adopted background and 1-σ error: bg = (160± 80) arcmin−2. Inset: zoom of the outer parts of all fits. Right panel: logarithmic plot of the number of GC candidates per arcsec2 without background subtraction (black dots), after subtracting a constant density of 160 arcmin−2 (magenta diamonds), and after subtracting 160×D arcmin−2 (blue dots), where D represents the expected profile of the background dilution over this magnitude range as a result of the lens magnification (see text). The dashed black line indicates a value of 160 arcmin−2, while the solid orange line indicates 160×D arcmin−2. 3.5. Background Contamination examples in left panel of Figure 5). We adopt a conserva- tive uncertainty of 80 arcmin−2, where the error bar encom- To estimate the number of background contaminants bg,we ± first calculate the number of GC candidates per unit area as passes the unlikely case that all objects in the outermost radial a function of radius (Figure 5). Then, we fit to the surface bins are background objects. number density profile, (r), a Sérsic function (Sérsic 1968) We note that after masking the bright galaxies, the remain- plus the background bg Nas a free parameter: ing objects (blue points in the left panel of Figure 5) may represent a smooth population of IGCs in A1689. If we integrate r 1/n the Sérsic function for these putative IGCs over the full areaof (r)= e exp −bn − 1 + bg, (3) the image (including masked regions, since IGCs would also N N R ( " e #) be projected onto the galaxies), we find that 50% of the GC candidates are in this smooth component. Of course, some where Re is the effective radius that encloses half of the GC of these objects will be associated with galaxies, since the sample; e is at Re; n is the Sérsic index, which controls spatial distributions of GCs are often more extended than the N N − the shape of the profile; and bn 1.9992n 0.3271 (Graham starlight. We therefore consider 50% to be an upper limit ≈ & Driver 2005). on the IGC fraction within this central region. More detailed We found that the fitted Sérsic parameters and background modeling of the GC distributions of individual galaxies, and were very sensitive to the radial range of the fit. This is partly wider coverage to trace the profile of the smooth component because of deviations from a smooth profile, but also because to larger radii, would help to refine this estimate. the number density of sources continuesto decrease as a func- Our estimate of the background is based on the outer parts tion of radius, without reaching a constant level. However, the of the ACS/WFC image, but in the case of A1689, it is nec- mean background within our adopted magnitude limits must essary to make a radial-dependent correction for the effect of − be in the range 0 < bg . 240 arcmin 2 (the high value being cluster lensing (Blakeslee 1999). The gravitational field of the observed density in the outermost bins). After trying vari- A1689 affects the spatial and magnitude distributions of the ous radial cuts, we found the most robust value of bg resulted background sources. Following the formalism of Broadhurst by masking the regions around all the bright galaxies (gray et al. (1995), in the absence of lensing we expect a power-law βm region in the right panel of Figure 3, based on our luminos- distribution of background sources Nbg(m) N010 , where ≈ ity model). This removes the concentrations of point sources N0 is a constant, and β is the logarithmic slope of the counts. associated with individual galaxies; the remaining objects fol- The lensing magnifies the brightness of the sources by the low a shallower, smoother radial density profile, shown in the position-dependent magnification factor A, and thus shifts the left panel of Figure 5. The fitted background in this case source magnitudes brighter by 2 5 log(A). It also increases the − . was bg 160 arcmin 2, which was within the broad range surface area by the same factor, and thus decreases the surface returned≈ by the various fits prior to the galaxy masking (see Globular Clusters in A1689 7 density. As a result, the effect of the lens magnification can be approximated as ′ −1 β(m+2.5 log A) −2.5(0.4−β) Nbg(m)= A N0 10 = Nbg(m)A , (4) ′ where Nbg is the observed (lensed) number density. Generally, β < 0.4, so the background counts over a fixed magnitude − −β range are decreased, or diluted, by a factor D = A 2.5(0.4 ). 105 The counts can also be amplified (D > 1) in regions where A < 1 (see Broadhurst et al. 1995 for a detailed discussion). The magnification A depends on the distance and mass distribution of the lens, as well as on the distance to the source plane. In general, it can be written as: 1 A = , (5) N(

annulusin the rightpanel of Figure 8. However, since the GCs are not perfect tracers of the stellar light, and their spatial concentrations are less sharply peaked than the galaxy proﬁles,an excess of galaxies at a given location tends to decrease the local value of SN, if the galaxieshave a “normal” SN 4. There- fore, at the same radius of 200 kpc where there is≈ a “bump” 15 in the number density of GCs, we actually ﬁnd a “dip” in the local value of SN, as shown by the green band in the left panel of Figure 8. A corresponding dip occurs near 200 kpc in the cumulative SN distribution in Figure 7. The scaling of NGC in brightest cluster galaxies with the total underlying mass within a common projected radius ) 10 has been interpreted as a consequence of early universal

( GC formation efﬁciency in dense regions. Using the value

N 9 S 0.71 0.22 GCs per 10 M⊙(Blakeslee 1999), our derived total± 14 NGC would predict a total mass of 2.3 10 M⊙ within 400 kpc in A1689. However, this estimate∼ × depends on the GCs following the same radial profile as the total matter dis- 5 tribution. While observations indicate that their spatial distribution is more extended than the starlight, until now it has not been possible to test how they relate to the dark matter distribution. We explore these issues in the following section. 5. COMPARISON OF MASS PROFILES 0 10 100 With the goal of testing the existence of a universal GC for- R [kpc] mation efficiency in an extreme system such as A1689, we compare the amounts of mass in this central field in the form

FIG. 7.— Cumulative specific frequency SN as a function of radius; the error of GCs, stars, hot intracluster gas, and total mass (includ- bars include statistical uncertainties in the number of GCs for the assumed ing dark matter). Figure 9 provides a 2-D visual comparison (fixed) GCLF and the uncertainty in the galaxy light profile. of the number density of GCs, the galaxy luminosity model, the lensing-derived mass distribution, and the X-ray emission −2 mag arcsec , causing the size of the SN error bars to increase map. The symmetry and smoothness of the X-ray gas stand strongly with radius. out; this is not a resolution effect (as evidenced by the com- For comparison to nearby galaxies, we need to consider pactness of the X-ray point sources in the image). To make how SN would evolve over a lookback time of 2.25 Gyr. There quantitative comparisons, we now derive the projected radial is no significant star formation in the A1689 early-type galax- mass distributions for each component. ies that dominate this central field (Balogh et al. 2002), and For the GCs, the total number within 400 kpc calcu- stellar populationmodels that match the colors of nearby giant lated in the preceding section can be converted to a mass ellipticals (e.g., 10 Gyr, solar metallicity models of Bruzual & by assuming a mean individual GC mass of GC = Charlot 2003) indicate that they have passively faded by about 5 hM i 2.4 10 M⊙ (McLaughlin 1999; Blakeslee 1999). The es- 0.20 mag since z =0.183. Assuming negligible destruction of timated× total mass in GCs within this radius is therefore GCs in this time, the passive galaxy evolution will cause SN total 10 GC =3.9 10 M⊙. For perspective, this is 60–80% of the to become 20% higher at z=0 than at the observed epoch of totalM stellar× mass of the Milky Way galaxy (Flynn et al. 2006; A1689. Thus, the global value of SN = 11.7 within 400 kpc McMillan 2011). would correspond to SN =14.0 at z=0, after correcting for pas- To calculate the stellar mass ⋆ we assume a stellar mass- sive evolution, similar to the SN values observed for the cen- M to-light ratio ⋆/LV =4, based on solar metallicity Bruzual tral cD galaxies in the nearby Virgo, Coma, and Hydra clus- & Charlot (2003)M models with a Salpeter initial mass function ters (Tamura et al. 2006; Peng et al. 2008, 2011; Harris et al. (IMF). Of course, the assumption of a constant /LV at all 2009; Wehner et al. 2008). Of course, for a more exact com- M⋆ radii is a first-order approximation. Moreover, the ⋆/LV , parison, the SN values should be estimated on similar physical and thus the derived mass, can vary by 30%, dependingM on scales. Blakeslee (1999) measured SN within apertures of 40 the IMF, but the choice of Salpeter is∼ reasonable for early- and 65kpc for the cores of six rich Abell clusters; he reported type galaxies (e.g., van Dokkum & Conroy 2010; Conroy & SN(40) =8.7 and SN(65) =9.2 (with rms scatters of 2 in h i h i ∼ van Dokkum 2012). Finally, using MV,⊙ =4.81, we obtain both cases). The corresponding values for A1689 at z=0 are total 12 =4.7 10 M⊙. SN(40)=10.6 0.4 and SN(65)=11.5 0.5,near the high end of M⋆ × the observed± range for nearby massive± clusters. The X-ray gas mass was estimated from the 3-D gas density profile ρg(r) constructed by Lemze et al. (2008) based on In addition to the cumulative SN, it is also worth consider- Chandra X-ray data. We found that the published ρg(r) is well ing the behavior of the local SN. We have noted that the surface density distribution of GC candidates is not completely fitted by a function of the form: smooth, since they are preferentially located around the bright ρ0 ′ ρg(r) = α β , (8) galaxies. This clustering causes the “bump” at 1.1 (200 [1 + (r/r0) ] kpc) in the GC radial density profile shown in Figure∼ 5. This −25 −3 feature in the GC density profile corresponds to a grouping of with best-fit values ρ0 =2.1 10 gcm , r0 = 321 kpc, α = bright galaxies at this radius, as highlighted within the green 0.58, and β =5.6 (giving r −×αβ r −3.2 at large r). These pa- ∼ Globular Clusters in A1689 9

′′ FIG. 8.— Left panel: local SN within ∼ 5 annuli as a function of radius. Right panel: galaxy light model. In both panels, the blue, yellow, green and red regions indicate radial ranges of 0-70, 70-180, 180-230 and 230-300 kpc, respectively. The “dip” in the local SN at R≈200 kpc is caused by the grouping of bright galaxies within the green annulus. rameters are only used for interpolating the X-ray data points. ure 10. The stellar mass fraction in GCs ( GC/ ⋆, panel f ) We then integrate this function along the line of sight ℓ to ob- increases by a factor of 2.5 within theM centralM 80 kpc, then ∼ tain the projected gas mass surface density ζ as: levels off, consistent with the SN profile shown previously. 1Mpc However, compared to either the X-ray gas or total matter, the ζ = 2 ρ(√R2 + ℓ2)dℓ, (9) GC mass fraction decreases as a function of radius (panels b 0 and a, respectively); the mass fraction in stars (panels d and c) Z shows a similar, even steeper, decline. As found in previous where R is the projected radius. This gives an X-ray gas mass total 13 studies, the gas mass fraction (panel e) appears to increase within 400 kpc of − =3.6 10 M⊙. The baryonicmass MX ray × monotonically with radius. Interestingly, the baryonic mass comprises field stars, GCs, and intracluster gas. Adding all fraction (ratio of baryons to total mass, panel g) reaches a 13 the components, we obtain baryon 4.1 10 M⊙. minimum near 150 kpc, because the stellar mass is more con- M ≈ × The total mass, total, is estimated from the κ map used centrated than the dark matter, but the gas is more extended. previously in Sec.3.5.M It includes both baryonic and non- This quantity has been studied extensively in galaxy clusters baryonic mass, but is dominated by the non-baryonic compo- as an approximation to the baryonic mass fraction of the uni- nent (at least outside the central few kpc). The convergence κ verse ( b/ DM Ωb/ΩDM). For instance, Lin et al. (2012) is the mass surface density, normalized by the critical surface measuredM baryonM fractions∼ for 94 galaxy clusters over a wide density (i.e., κ=Σ/Σcrit ), where range redshift; our value for A1689 in Figure 11 agrees well 2 with the results of this recent study. c Ds McLaughlin (1999) derived an average efficiency of GC Σcrit = ; (10) 4πGDLDLs formation per baryonic mass, i.e., ǫb = GC/ baryon, of 0.0026 within r < 100 kpc. Comparing Mclaughlin’sM M value c is the speed of light; DL, Ds, and DLs are the distances to the lens (A1689), to a reference source (assumed z=3), and with our estimate for the same radial range (Figure 11h), we from the lens to the source, respectively. Integrating, we find find ǫb=0.0021, which is within the scatter of the values found 14 by McLaughlin. Meanwhile, Blakeslee (1999) reported a total=6.4 10 M⊙ within 400 kpc. −4 M × mean efficiency per total mass of ǫt = GC/ total =1.7 10 Figure 10 presents the radial mass density and cumulative M M × −4 within r < 50 kpc; within this radius, we obtain ǫt =1.8 10 mass profiles for each of the above components. The stars × and GCs are strongly concentrated in and around the galaxies; (Figure 11a), in close agreement with the expected value. the most prominent feature is the “bump” at R 200 kpc, However, although the mass ratios within these relatively discussed above. The hot X-ray emitting gas and≈ total mass small radii are remarkably consistent with the “universal” ef- exhibit much smoother profiles. The hot gas dominates the ficiencies previously proposed in the literature, the global val- A1689 baryonic mass beyond the central 30 kpc. ues are not. Within 400 kpc, we find ǫb =0.00095 and ∼ A1689 −5 To investigate the radial run of the GC mass fraction relative ǫt =6.1 10 , both a factor of 3 lower than the valuescited to the other mass components, we plot in Figure 11 the mass above. After× converting to our assumed mean GC mass, the ratios for all possible combinations of the components in Fig- 10 Alamo-Martínez et al.

(a) (b)

′′ FIG.9.— (a) Surface number density of GC candidates, smoothed with a Gaussian of FWHM = 10 (300 pix); (b) surface brightness distribution from 59 galaxies with isophotal models plus the SExtractor background map; (c) lensing-derived total surface mass density (includes baryonic and non-baryonic components); (d) X-ray surface brightness distribution (from Chandra archive). −5 results of Spitler & Forbes (2009) imply ǫt =4.2 10 , about the Virgo cluster. The large error bars are due to the uncer- 30% lower than our value within 400 kpc but consistent× within tainty in the GCLF parameters. Although the Gaussian form the errors; however, it not clear what radius to use in this case. of the GCLF is well calibrated for giant ellipticals in rich clus- Thus, we emphasize again the importance of comparing such ters, even with 20.9 hrs of integration, our data fall 1.6σ short ratios within the same physical radii. of the GCLF turnover; we therefore sample only 10% of the GCs brighter than the turnover (or 5% of the total population, 6. DISCUSSION AND CONCLUSIONS assuming a symmetric GCLF). Thus, the large extrapolation Deep broadband imaging with HST/ACS (&90% complete yields a sizable systematic uncertainty. Nevertheless, this re- to I814 = 29) has revealed an extremely rich GC system in the mains the largest GC system yet discovered, with at least a center of the massive lensing cluster A1689. The estimated factor of two more than the next most populous systems, in- +75,450 total population of 162,850−51,310 GCs within a projected ra- cluding Coma and A3558 (Peng et al. 2011; Barber DeGraaff dius of 400 kpc represents the largest system of GCs stud- 2011). ied to date, six times the number within the same radius in Our analysis accounts for the effects of Eddington bias, Globular Clusters in A1689 11

15 104 10

103 1014

102 1013 ]

2 1

10 ]

⊙ 12

/pc 10 ⊙ 100 [M [M 1011 10-1 total (baryonic + non-baryonic) 10 X-ray gas 10 -2 10 stars

GCs 9 -3 10 10 101 102 101 102 R [kpc] R [kpc]

FIG. 10.— Left: radial mass density profile for each component. Right: cumulative mass profiles versus radius. The thick black line indicates the cumulative profile for the baryonic mass, Mbaryon = MX−ray + M⋆. gravitational magnification of the background surface density, the numerator. This highlights the fact that cannibalization redshifting of the bandpass (K corrections), and passive evo- of normal cluster galaxies by the central cD will tend to de- lution of the GCLF. Although it is possible that there has been crease the SN, rather than increase it; thus the high SN value some evolution in the shape of the GCLF since z =0.18, this must have been imprinted in the cluster core at early times. would mainly occur by the destruction of low-mass GCs (e.g., We have also noted that passive evolution of the galaxy lumi- Jordán et al. 2007; McLaughlin & Fall 2008) with little effect nosity since z=0.18 would cause the observed SN to increase for the masses of objects near the GCLF peak or brighter. Our by20%at z=0, but even with this effect, the high SN in A1689 assumption of a symmetric GCLF would therefore likely un- would not be anomalous among cD galaxies in local clusters. derestimate the population at z =0.18, but the additional low- Remarkably, the mass in GCs within 400 kpc of the center mass GCs would have little effect on our total mass estimates. of A1689 is equivalent to 60-80% of the total stellar mass of The spatial distribution of GC candidates in the center of the Milky Way. Integrated out to the virial radius of 3 Mpc, A1689 is not completely smooth; there are obvious concen- the total mass in GCs is likely twice that of the stars∼ in our trations around the cluster galaxies. However, by masking the Galaxy. Of course, this is a small fraction of the total mass in bright and intermediate-luminosity galaxies, we can trace an A1689. We have examined the mass profile of the GCs as a apparently smooth component, which comprises half of the function of radius and compared it with the mass profiles of total population when integrated over the field (via the best- the stellar light, hot intracluster gas, and total lensing-derived fit Sérsic model). If we identify these objects as belonging to matter content within this central field. The mass profile of a possible intracluster population of GCs, then there may be the GCs is somewhat more extended than the stellar light, but as many as 80,000 such IGCs within the central 400 kpc in more concentrated than the hot gas or dark matter. If the mass A1689. We∼ consider this an upper limit, since some of these fraction is viewed as a GC formation efficiency, then the effi- IGC candidates are undoubtedly bound to individual galaxies. ciency (in terms of either baryonic or total mass) decreases as We plan to investigate this in more detail by modeling the GC a function of radius, and there is no “universal” value. distributions around individual galaxies. Imaging to a similar On the other hand, when compared within the same physi- depth at larger radii from the cluster center would also help in cal radii, the GC mass fractions with respect to the total and constraining the IGC population. baryonic masses agree with the values found in samples of The cumulative SN increases from a value near 5 within nearby clusters, all of which have masses lower than A1689. 10 kpc to 10 0.5 (not including systematic uncertainty from This suggests the possibility of a universal GC formation pro- ± the GCLF) within 70 kpc. Although the uncertainties in SN file within galaxy clusters. In contrast, Laganá et al. (2011) become large beyond 100 kpc, the profile appears to flatten, estimated the stellar, intracluster gas, and total masses within and the value within 150 kpc is SN = 11.1 2.0. There is a r500 for 19 galaxy clusters, and found a decrease in the stellar ± clear dip in the cumulative SN around 200 kpc, before it rises mass fraction with increasing total mass of the system. That again to 12 within 300 kpc. The dip at 200 kpc occurs is, more massive clusters had lower overall star formation effi- despite a∼ local increase in the GC number density at the same ciencies. Taken together, these results are consistent with the radius; it is caused by a subgroupingof several bright galaxies view that the high SN values in cD galaxies are a consequence which appear to have a more normal SN < 10 (as was shown of “missing” stellar light in more massive clusters (Blakeslee by comparing the green regions in Figure 8). Such galaxies 1997), rather than an excess in the number of globular clus- contribute relatively more to the denominator of SN than to ters. 12 Alamo-Martínez et al.

Finally, we note that in a recent study, Suárez-Madrigal et Support for program GO-11710 was provided in part al. (2012) modeled the influence of the dark matter halo on through a grant from the Space Telescope Science Institute, molecular clouds at different locations within it. They found which is operated by the Association of Universities for Re- that the star formation efficiency of the clouds depends on search in Astronomy, Inc., under NASA contract NAS5- the ambient density, and thus decreases with distance from 26555. We thankDan Mageefor help with the CTE correction the halo center, in qualitative agreement with our finding that software. JPB acknowledges helpful conversation with James GC formation efficiency decreases with radius. If the mass Bullock. K.A.A-M acknowledges the support of CONACyT density profiles in galaxy clusters are approximately universal (México). This research has made use of the NASA/IPAC (e.g., Navarro et al. 1997), then it would also make sense that Extragalactic Database (NED), which is operated by the Jet the GC formation efficiency profile may follow a universal Propulsion Laboratory, California Institute of Technology, un- form. Further studies of the GC, baryonic, and total mass der contract with NASA. profiles in galaxy clusters are needed to test this intriguing Facilities: HST (ACS/WFC). possibility.

REFERENCES Abadi, M. G., Navarro, J. F., & Steinmetz, M. 2006, MNRAS, 365, 747 Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, submitted Abell, G. O., Corwin, H. G., Jr., & Olowin, R. P. 1989, ApJS, 70, 1 Jedrzejewski, R. I. 1987, MNRAS, 226, 747 Alamo-Martínez, K. A., West, M. J., Blakeslee, J. P., et al. 2012, A&A, 546, Jee, M. J., Ford, H. C., Illingworth, G. D., et al. 2007, ApJ, 661, 728 A15 Johnston, K. V., Bullock, J. S., Sharma, S., et al. 2008, ApJ, 689, 936 Anderson, J., & Bedin, L. R. 2010, PASP, 122, 1035 Jordán, A., Blakeslee, J. P., Peng, E. W., et al. 2004, ApJS, 154, 509 Andersson, K. E., & Madejski, G. M. 2004, ApJ, 607, 190 Jordán, A., McLaughlin, D. E., Côté, P., et al. 2007, ApJS, 171, 101 Balogh, M. L., Couch, W. J., Smail, I., Bower, R. G., & Glazebrook, K. Jordán, A., West, M. J., Côté, P., & Marzke, R. O. 2003, AJ, 125, 1642 2002, MNRAS, 335, 10 Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192, 18 Barber DeGraaff, R. G. 2011, Ph.D. Thesis, Washington State University Kruijssen, J. M. D. 2012, MNRAS, 426, 3008 Benítez, N., Ford, H., Bouwens, R., et al. 2004, ApJS, 150, 1 Laganá, T. F., Zhang, Y.-Y., Reiprich, T. H., & Schneider, P. 2011, ApJ, 743, Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 13 Blakeslee, J. P. 1997, ApJ, 481, L59 Lee, M. G., Park, H. S., & Hwang, H. S. 2010, Science, 328, 334 Blakeslee, J. P. 1999, AJ, 118, 1506 Lemze, D., Barkana, R., Broadhurst, T. J., & Rephaeli, Y. 2008, MNRAS, Blakeslee, J. P. 2005, Highlights of Astronomy, 13, 171 386, 1092 Blakeslee, J. P., Anderson, K. R., Meurer, G. R., Benítez, N., & Magee, D. Lemze, D., Broadhurst, T., Rephaeli, Y., Barkana, R., & Umetsu, K. 2009, 2003, Astronomical Data Analysis Software and Systems XII, 295, 257 ApJ, 701, 1336 Blakeslee, J. P., Tonry, J. L., & Metzger, M. R. 1997, AJ, 114, 482 Limousin, M., Richard, J., Jullo, E., et al. 2007, ApJ, 668, 643 Broadhurst, T., Benítez, N., Coe, D., et al. 2005, ApJ, 621, 53 Lin, Y.-T., Stanford, S. A., Eisenhardt, P. R. M., et al. 2012, ApJ, 745, L3 Broadhurst, T. J., Taylor, A. N., & Peacock, J. A. 1995, ApJ, 438, 49 McLaughlin, D. E. 1999, AJ, 117, 2398 Brodie, J. P., & Strader, J. 2006, ARA&A, 44, 193 McLaughlin, D. E., & Fall, S. M. 2008, ApJ, 679, 1272 Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000 McLaughlin, D. E., Harris, W. E., & Hanes, D. A. 1994, ApJ, 422, 486 Chies-Santos, A. L., Larsen, S. S., Kuntschner, H., et al. 2011, A&A, 525, McMillan, P. J. 2011, MNRAS, 414, 2446 A20 Mieske, S., Infante, L., Benítez, N., et al. 2004, AJ, 128, 1529 Chies-Santos, A. L., Larsen, S. S., & Kissler-Patig, M. 2012, MNRAS, 427, Moore, B., Diemand, J., Madau, P., Zemp, M., & Stadel, J. 2006, MNRAS, 2349 368, 563 Cho, J., Sharples, R. M., Blakeslee, J. P., et al. 2012, MNRAS, 422, 3591 Navarro, J. F., Frenk, C. S., & White, S. D. M. 1997, ApJ, 490, 493 Coe, D., Benítez, N., Broadhurst, T., & Moustakas, L. A. 2010, ApJ, 723, Peng, E. W., Ferguson, H. C., Goudfrooij, P., et al. 2011, ApJ, 730, 23 1678 Peng, E. W., Jordán, A., Blakeslee, J. P., et al. 2009, ApJ, 703, 42 Cohen, J. G., Blakeslee, J. P., & Ryzhov, A. 1998, ApJ, 496, 808 Peng, E. W., Jordán, A., Côté, P., et al. 2008, ApJ, 681, 197 Cohen, J. G., Blakeslee, J. P., & Côté, P. 2003, ApJ, 592, 866 Puzia, T. H., Kissler-Patig, M., Thomas, D., Maraston, C., Saglia, R. P., Conroy, C., & van Dokkum, P. G. 2012, ApJ, 760, 71 Bender, R., Goudfrooij, P., & Hempel, M. 2005, A&A, 439, 997 Conroy, C., & Wechsler, R. H. 2009, ApJ, 696, 620 Schlaﬂy, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103 Côté, P., Marzke, R. O., & West, M. J. 1998, ApJ, 501, 554 Sereno, M., Ettori, S., Umetsu, K., & Baldi, A. 2013, MNRAS, 428, 2241 Côté, P., Blakeslee, J. P., Ferrarese, L., et al. 2004, ApJS, 153, 223 Sérsic, J. L. 1968, Atlas de Galaxias Australes (Cordoba: Observatorio De Lucia, G., & Helmi, A. 2008, MNRAS, 391, 14 Astronomico) De Lucia, G., Springel, V., White, S. D. M., Croton, D., & Kauffmann, G. Sirianni, M., Jee, M. J., Benítez, N., et al. 2005, PASP, 117, 1049 2006, MNRAS, 366, 499 Spitler, L. R., & Forbes, D. A. 2009, MNRAS, 392, L1 Eddington, A. S. 1913, MNRAS, 73, 359 Springel, V., White, S. D. M., Jenkins, A., et al. 2005, Nature, 435, 629 Ferrarese, L., Côté, P., Cuillandre, J.-C., et al. 2012, ApJS, 200, 4 Stetson, P. B. 1987, PASP, 99, 191 Flynn, C., Holmberg, J., Portinari, L., Fuchs, B., & Jahreiß, H. 2006, Suárez-Madrigal, A., Ballesteros-Paredes, J., Colín, P., & D’Alessio, P. MNRAS, 372, 1149 2012, ApJ, 748, 101 Font, A. S., Johnston, K. V., Bullock, J. S., & Robertson, B. E. 2006, ApJ, Tamura, N., Sharples, R. M., Arimoto, N., et al. 2006, MNRAS, 373, 588 638, 585 Taylor, A. N., Dye, S., Broadhurst, T. J., Benitez, N., & van Kampen, E. Ford, H. C., Clampin, M., Hartig, G. F., et al. 2003, Proc. SPIE, 4854, 81 1998, ApJ, 501, 539 Ford, H. C., Illingworth, G. D., Clampin, M., et al. 2002, BAAS, 34, 675 Teague, P. F., Carter, D., & Gray, P. M. 1990, ApJS, 72, 715 Fruchter, A. S., & Hook, R. N. 2002, PASP, 114, 144 Tody, D. 1986, Proc. SPIE, 627, 733 Georgiev, I. Y., Puzia, T. H., Goudfrooij, P., & Hilker, M. 2010, MNRAS, Tody, D. 1993, Astronomical Data Analysis Software and Systems II, 52, 406, 1967 173 Girardi, M., Fadda, D., Escalera, E., et al. 1997, ApJ, 490, 56 Tonry, J. L., Dressler, A., Blakeslee, J. P., et al. 2001, ApJ, 546, 681 Graham, A. W., & Driver, S. P. 2005, PASA, 22, 118 Tyson, J. A., & Fischer, P. 1995, ApJ, 446, L55 Guo, Q., White, S., Li, C., & Boylan-Kolchin, M. 2010, MNRAS, 404, 1111 van den Bosch, F. C., Yang, X., Mo, H. J., et al. 2007, MNRAS, 376, 841 Harris, W. E. 1991, ARA&A, 29, 543 van Dokkum, P. G., & Conroy, C. 2010, Nature, 468, 940 Harris, W. E. 2001, Star Clusters, Saas-Fee Advanced Courses, Volume 28, Villegas, D., Jordán, A., Peng, E. W., et al. 2010, ApJ, 717, 603 223 Wehner, E. M. H., Harris, W. E., Whitmore, B. C., Rothberg, B., & Woodley, Harris, W. E., Kavelaars, J. J., Hanes, D. A., Pritchet, C. J., & Baum, W. A. K. A. 2008, ApJ, 681, 1233 2009, AJ, 137, 3314 West, M. J., Cote, P., Jones, C., Forman, W., & Marzke, R. O. 1995, ApJ, Harris, W. E., Pritchet, C. J., & McClure, R. D. 1995, ApJ, 441, 120 453, L77 Harris, W. E., & van den Bergh, S. 1981, AJ, 86, 1627 West, M. J., Côté, P., Marzke, R. O., & Jordán, A. 2004, Nature, 427, 31 Globular Clusters in A1689 13

(a) 0.007 (b)

0.006 0.00020 0.005 total xray 0.00015 0.004 /M /M

GC GC 0.003 M M 0.00010 0.002

0.001

0.08 (c) (d) 2.5 0.07 0.06 2.0

total 0.05 xray 1.5 /M 0.04 /M ⋆ ⋆

M 0.03 M 1.0 0.02 0.5 0.01

(e) (f) 0.060 0.009 0.008 0.055

⋆ 0.007 total 0.050 /M

/M 0.006

0.045 GC

xray 0.005 M

M 0.040 0.004

0.035 0.003

0.12 (g) 0.0030 (h) 0.11 0.0025 0.10 total

baryon 0.0020 /M 0.09 /M 0.08 GC baryon 0.0015 M M 0.07

0.06 0.0010

0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400 R [kpc] R [kpc]

FIG. 11.— Ratios of the cumulative distributions for various mass components in A1689, based on the curves plotted in Fig. 10. The red stars in (a) and (h) −4 represent the values ǫt = 1.7×10 from Blakeslee (1999) and ǫb = 0.0026 from McLaughlin (1999), respectively. See text for discussion. Note that the GC stellar mass fraction MGC/M⋆ shown in panel ( f ) is equivalent to SM/100, where SM is the “speciﬁc mass” parameter used in some studies (e.g., Peng et al. 2008). West, M. J., Jordán, A., Blakeslee, J. P., et al. 2011, A&A, 528, A115 Zekser, K. C., White, R. L., Broadhurst, T. J., et al. 2006, ApJ, 640, 639