Separating the conjoined red clump in the Galactic Bulge: Kinematics and Abundances
Roberto De Propris, R. Michael Rich, Andrea Kunder, Christian I. Johnson, Andreas Koch, Sarah Brough, Christopher J. Conselice, Madusha Gunawardhana, David Palamara, Kevin Pimbblet, Dinuka Wijesinghe
aa r X i v : . [ a s t r o - ph . GA ] A p r Preprint typeset using L A TEX style emulateapj v. 11/10/09
SEPARATING THE CONJOINED RED CLUMP IN THE GALACTIC BULGE: KINEMATICS AND ABUNDANCES. R OBERTO D E P ROPRIS , R. M ICHAEL R ICH , A NDREA K UNDER ,C HRISTIAN
I. J
OHNSON , A NDREAS K OCH AND S ARAH B ROUGH , C HRISTOPHER
J. C
ONSELICE , M ADUSHA G UNAWARDHANA , D AVID P ALAMARA ,K EVIN P IMBBLET , D INUKA W IJESINGHE ABSTRACTWe have used the AAOMEGA spectrograph to obtain R ∼ spectra of 714 stars that are members oftwo red clumps in the Plaut Window Galactic bulge field ( l, b ) = 0 ◦ , − ◦ . We discern no difference betweenthe clump populations based on radial velocities or abundances measured from the Mg b index. The velocitydispersion has a strong trend with Mg b -index metallicity, in the sense of a declining velocity dispersion athigher metallicity. We also find a strong trend in mean radial velocity with abundance. Our red clump sampleshows distinctly different kinematics for stars with [Fe/H] < − , which may plausibly be attributable to aminority classical bulge or inner halo population. The transition between the two groups is smooth. Thechemo-dynamical properties of our sample are reminiscent of those of the Milky Way globular cluster system.If correct, this argues for no bulge/halo dichotomy and a relatively rapid star formation history. Large surveysof the composition and kinematics of the bulge clump and red giant branch are needed to define further thesetrends. Subject headings:
Galaxy: formation — Galaxy: bulge — Galaxy: kinematics and dynamics INTRODUCTION
The existence of a prominent bar in the Galactic bulge isnow well established from multiple lines of evidence (e.g.,Liszt & Burton 1980; Blitz & Spergel 1991; Stanek et al.1994; Babusiaux & Gilmore 1995). The large-scale kinemat-ics of the bulge sampled by the BRAVA (Bulge Radial Veloc-ity Assay) survey (Rich et al. 2007; Howard et al. 2008, 2009)can be fitted with simple cylindrical rotation and little or noclassical spheroidal component (Shen et al. 2010), a conclu-sion also reached (albeit at lower confidence) from proper mo-tion surveys (Rattenbury et al. 2007). Our ‘bulge’ appears toconsist of a peanut-shaped bar, reminiscent of the ‘pseudob-ulges’ defined by Kormendy & Kennicutt (2004) and encoun-tered elsewhere (Kormendy & Barentine 2010).At the same time, this result is puzzling. The BRAVAdata imply that our Galaxy has not undergone any significantmerger since the epoch at which the disk formed, in contrastwith expectations from simulations within Λ CDM cosmo-logical models (e.g., Bullock & Johnston 2005; Cooper et al.2010). However, it is well known that stars in the bulge aremetal-rich and α -enhanced, indicating a rapid star formationhistory typical of classical bulges (McWilliam & Rich 1994;Ballero et al. 2007). The observation of an abundance gradi-ent for bulge stars (Zoccali et al. 2008) and a correlation be-tween abundance and kinematics (Babusiaux et al. 2010) mayalso indicate the presence of a classical spheroidal compo-nent, although dynamical data (Shen et al. 2010) seem to be Cerro Tololo Inter-American Observatory, La Serena, Chile Department of Physics and Astronomy, University of California, LosAngeles, CA, USA Department of Physics and Astronomy, University of Leicester, Uni-versity Road,Leicester LE1 7RH, UK Australian Astronomical Observatory, Epping, NSW, Australia Department of Physics and Astronomy, University of Nottingham, UK Sydney Institute for Astronomy, School of Physics A29, The Univer-sity of Sydney, NSW, Australia School of Physics, Monash University, Clayton, VIC, Australia inconsistent with this . On the other hand, detailed abun-dances for subgiants and lensed dwarfs, are more similar tothose of the thick disk, in accordance with a pseudobulge for-mation scenario (Melendez et al. 2008; Alves-Brito, A. et al.2010; Bensby et al. 2010; Ryde et al. 2010).Adding to the complexity, the red clump in thebulge color-magnitude diagram appears doubled at | l | > ◦ (McWilliam & Zoccali 2010; Nataf et al. 2010).McWilliam & Zoccali (2010) carry out a very careful analysisof the photometric properties of the double red clump andconclude that the splitting of the red clump is due to adistance effect and that the bulge may contain an X-shapedstructure, extending from the ends of the bar. So far, theonly spectroscopic study of this population has been carriedout by Rangwala et al. (2009) using Fabry-Perot imagingspectroscopy in Baade’s Window and some adjacent fields,and their results suggest that there is a metallicity gradientwith galactic latitude (Rangwala & Williams 2009), as wellas dynamical differences, which would be more consistentwith the presence of a stream, although it is clear that moreaccurate kinematics and metallicities are needed to trulyunderstand the nature of this feature.Here we present the first low resolution survey of a field at l = 0 ◦ b = − ◦ (Plaut’s Window) specifically targeting thered clump stars. We analyze the kinematics and metallicityof stars belonging to each peak and derive the metallicity-kinematics trends. Although this is clearly a ‘first look’ ex-ploratory analysis, our data imply that the two populations aredynamically and chemically similar and favor the X-shapedbulge hypothesis proposed by McWilliam & Zoccali (2010). DATA REDUCTION AND ANALYSIS
The data for this project were kindly provided by theGalaxy and Mass Assembly Survey (GAMA – Driver et al.2010). This survey is observing three 12 deg fields at 9h,12h and 15h RA using the AA Ω multi-fiber spectrograph onthe Anglo-Australian Telescope (AAT). At the end of May,2010, it was found to be impossible to reach either GAMA De Propris et al. F IG . 1.— The color-magnitude diagram (from 2MASS data) for our bulgefield is shown in the left-hand panel of this figure. The blue box shows the(non-dereddened) selection limits used. The right-hand panel shows the his-togram of the distribution of stars in K and the two peaks identified by theGMM algorithm, marked by arrows (red for the brighter peak and green forthe fainter peak). field during the last two hours of the night, at sufficiently lowairmass. The survey kindly offered to observe one of the fieldscontaining a double red clump for us.Two 400-fiber configurations were observed for Plaut’sWindow (where the two red clumps are relatively well sep-arated and the extinction is lower). The target stars were se-lected from the 2MASS survey (Skrutskie et al. 2006), in a2 degree (diameter) field. Bulge stars were required to have K = 7 . J − K ) + 9 to exclude disk contamination and tolie between . < K < . and . < ( J − K ) < . to probe the double red clump. Magnitudes were dereddenedfrom the Schlegel et al. (1998) maps. Figure 1 shows thecolor-magnitude diagram of the target stars. The two redclumps are identified by applying a Gaussian Mixture Mod-elling algorithm to the data (Muratov & Gnedin 2010), whichalso assigns to each star a probability of belonging to eitherpeak. The magnitudes of the two red clumps peaks are ingood agreement with those reported by McWilliam & Zoccali(2010) for this region. By necessity, we used the same spec-troscopic setup as employed for the GAMA survey; this cov-ers the entire optical window, from about 3700 to 9000 ˚Aat a resolution of about 1500. Although this is somewhatless than optimal for determining stellar radial velocities andabundances, it suffices for our initial analysis of this field.The data were reduced using the automated pipeline sup-plied by AA Ω . A total of 714 stars were eventually observed.The typical signal to noise ratios of these spectra are about 10at Ca H & K, 20 in the Mg b region and 50 for the CalciumTriplet wavelength range. Radial velocities were measuredby cross-correlation against stellar templates using the runz program (Saunders et al. 2004) which is specifically writtenfor analysis of AA Ω data. Among the several templates avail-able in runz we imposed the choice of a K-giant template,if this was not automatically selected by the program. Of the714 targets, 631 returned a valid radial velocity ( > prob- ability that the measured velocity is correct from the heightof the correlation peak). Typical velocity errors are 1/10 of aresolution element, or 25 km s − for the R ∼ of GAMAdata, but no stars with a velocity error above 2 σ (50 km s − )were used for our analysis.We experimented with a number of techniques to measuremetal abundances, including the non-SEGUE stellar parame-ters pipeline (e.g., Li et al. 2010 and references therein) andthe calcium triplet (although red clump giants were outsidethe luminosity range of the calibrations). We eventually foundthat the most reliable measurements (when compared withthe high resolution abundances measured for giants in Plaut’swindow by Johnson et al. 2011) were given by using the Mg b Lick index (Worthey 1994; Ibata & Gilmore 1995). We mea-sured this index and its errors (based on the CCD noise pa-rameters and the radial velocity errors) using the
LECTOR software . The typical error in measuring the Mg b index isabout 5%.In order to derive metal abundances from the Mg b index, weused fitting functions for cool stars by Worthey et al. (1994).We assumed log g = 2 . which is appropriate for red clumpgiants based on the Padova isochrones (Marigo et al. 2008).Temperatures were estimated from the dereddened J − K color, using the T eff – color relations by Houdashelt et al.(2000). This approach was used by Cote et al. (1999), amongothers, to measure the abundance of red giants in M31, al-though we found that only Mg b can be reliably measured inour data. We derived metallicities for 545 stars in our data.The typical random error in metal abundance is ± . dex,based on the error in the index measurement, but of coursethere are systematic errors depending on the assumed valuesof log g and T eff . Altering log g by 0.1 at the same tem-perature yields a change in [Fe/H] of around 0.15 dex, whilealtering T eff by 100K yields abundance changes of 0.3 dex.In addition, our data are not on the Lick system as no stan-dards were observed, and this may introduce a systematic ef-fect of the order of 0.2 dex in the measurement of metal abun-dances. We used some Lick standards observed by ELODIEMoultaka et al. (2004) to measure Mg b and find that our abun-dances tend to be ∼ . dex too low, which would bring ourdata in better agreement with high resolution measurementsby Johnson et al. (2011). In addition, the known α enhance-ment of the bulge (e.g., McWilliam & Rich 1994) means thatthe Mg b abundance may overestimate the actual [Fe/H] of thestars, although it is probably a better proxy of the total metalabundance. THE NATURE OF THE RED CLUMP
Figure 2 shows the radial velocity distribution for red clumpstars derived from our data. The mean heliocentric radial ve-locity for the entire sample is − ± km s − (rms) , whichis equivalent to a velocity of − km s − (Galactocentric stan-dard of rest), consistent with data from the BRAVA survey.The radial velocity dispersion is ± km s − , which issomewhat larger than what is measured for this galactic lati-tude by BRAVA, although we are looking at a fainter sample.The intrinsic dispersion and its error were computed follow-ing Spaenhauer et al. (1992). The higher velocity dispersionseems to be due to a low metallicity population that may notbe present in the BRAVA data.Stars in the first (brighter) peak have a mean heliocen-tric velocity of − ± km s − and velocity dispersion of ouble red clump kinematics and abundances 3 F IG . 2.— Radial Velocity distribution of stars belonging to each red clumppeak (red for the brighter peak and green for the fainter peak as in Figure 1).F IG . 3.— [Fe/H] distribution of stars belonging to each red clump peak (redfor the brighter peak and green for the fainter peak as in Figure 2). The blackdashed histogram shows the distribution of metal abundances for red giantsstudied by Johnson et al. 2011. ± km s − , while stars belonging to the second (fainter)red clump have mean velocity of − ± km s − and veloc-ity dispersion of ± km s − . Within the errors, starsin both red clumps have the same mean velocity and veloc-ity dispersion and therefore appear to have the same kinemat-ics. This would appear to be at odds with claims for a differ-ence in the kinematics of the two red clumps (Rangwala et al.2009), although more sightlines are needed. Application ofthe Gaussian Mixture Modelling algorithm shows that thereis < chance that the velocity distribution is bimodal. AKolmogorov-Smirnov two-sample test yields an 88% chancethat the two distributions come from the same parent. Thet-test and F-test show that the two distributions do not havesignificantly different means or variances.Figure 3 shows the distribution in metal abundance forstars in both clumps. The distribution for all stars has amean [Fe/H] of − . ± . . The distribution is consis-tent with that observed for red giants in Baade’s Windowby McWilliam & Rich (1994), the high resolution metallic-ities in Plaut’s window (overplotted in the figure) measuredby Johnson et al. (2011) and the expectations from the metal-licity gradient measured by Zoccali et al. (2008), but with asystematic offset ( ∼ . dex) probably due to the lack of anabsolute calibration based on Lick standards.Stars in the first peak have mean [Fe/H] of − . ± . with a dispersion of . , while stars in the second peak have mean [Fe/H] of − . ± . with a dispersion of . . Again,the Gaussian mixture modeling algorithm returns a distribu-tion consistent with a unimodal distribution. Within the er-rors, stars belonging to each red clump have the same [Fe/H]abundances. A K-S test gives a 73% probability for the twoclumps to be drawn from the same distribution. Similarly, thet-test and F-test show that two samples do not have signifi-cantly different means or variances.One aim in this work is to compare the kinematics andmetallicities of stars belonging to the two red clumps iden-tified in the galactic bulge by McWilliam & Zoccali (2010);Nataf et al. (2010). Our data show that the two populationshave the same kinematics and metal abundance, which sug-gests that the difference in luminosity between the two redclumps is a distance effect. In other words, the two red clumpsare observed at the two ends of the ∼ kpc bar that consti-tutes the bulge of the Milky Way. These data are thereforeconsistent with the analysis by McWilliam & Zoccali (2010),attributing the double red clump to the existence of X-shapedprotrusions at the end of the Milky Way bar (i.e., an X-shapedbulge). If these stars lie at the end of a bar, then the metalabundance gradient across the structure is expected to be quitesmall.An alternate possibility is that the two red clumps are dif-ferent in helium content. Helium abundance variations arebelieved to exist (from indirect evidence) in massive globularclusters in our Galaxy (e.g., Carretta et al. 2009). Helium en-hancement would produce differences in the luminosity of thered clump stars (D’Antona et al. 2010), but any process capa-ble of producing the necessary helium enhancement wouldalso overproduce metals. We would presume that populationswith different chemical abundances might also have differentkinematics. ABUNDANCE TRENDS WITH KINEMATICS
Classical bulges present a number of abundance trends withkinematics, some of which are also observed in the Galac-tic bulge. Zoccali et al. (2008) have measured a metallicitygradient with radius. Babusiaux et al. (2010) find a trend be-tween metal abundance and velocity dispersion in Baade’sWindow and two lower ( b = − ◦ and − ◦ ) galactic latitudefields, arguing for a two component bulge, with a metal-richsystem comprising the bar and a metal-poor spheroid or thickdisk, the relative contribution from each of these varying withgalactic latitude.We plot the mean velocity (lower panel) and velocity dis-persion (upper panel) as a function of metallicity in Figure4. We plot all stars (black points), stars in the first (brighter)peak (red points) and stars in the second (fainter) peak (greenpoints) separately. Stars in each group are binned in bins con-taining the same number of stars after sorting by metal abun-dance. For each of the groups we consider (all, stars in thebrighter and fainter peaks) the bins span a non-overlappingrange in metal abundance. Stars in both red clumps appearto obey the same relations between metallicity and kinemat-ics. This confirms that the two populations do not differ inkinematic properties or abundances, which is more consistentwith a projection effect and an X-shaped bulge structure.The data shown in Figure 4 show a clear trend for increasingvelocity dispersion with decreasing metal abundance. Thisis similar to what is found by Johnson et al. (2011) in theirhigher resolution data. Babusiaux et al. (2010) uses two fieldsat b = − ◦ and b = − ◦ and although there are few starsat [Fe/H] < − in their data, there is a hint of an increase in De Propris et al. σ r ( k m s - ) All starsPeak 1Peak 2-3 -2.5 -2 -1.5 -1 -0.5 0 0.5[Fe/H] (from Mg b) -100-80-60-40-20020406080100 < V h e li o > ( k m s - ) F IG . 4.— Dependence of heliocentric mean velocity on metallicity (lowerpanel) and radial velocity dispersion (upper panel) for stars in our field. Seelegend in figure to identify the samples. The bins are chosen to contain thesame number of stars, sorted by metal abundance; the horizontal error barsrepresent the range of metallicities in each bin. The vertical error bars are theerrors on mean radial velocity and velocity dispersion, as appropriate. the velocity dispersion at lower metal abundances, in agree-ment with our observations. Vieira et al. (2007) find a flat dis-tribution of metallicity with velocity dispersion (from propermotion data) for stars with [Fe/H] > − , which is not in dis-agreement with our observations, where most of the increasein velocity dispersion takes place for lower metallicity stars.For stars with [Fe/H] > − the velocity dispersion is ingood agreement with that measured for BRAVA M giantsand does not depend strongly on metal abundance, which isbroad agreement with what measured by Vieira et al. (2007)in Plaut’s Window and the two lower galactic latitude fieldsin Babusiaux et al. (2010). However, at [Fe/H] < − thereappears to be a (at face value) smooth transition to a dynam-ically hot component. Similarly, in the upper panel of Figure4 we see that the metal-rich component appears to have sig-nificant mean heliocentric velocity, while at [Fe/H] < − oneobserves a smooth trend towards a relatively static velocitycomponent. One caveat in this is that the metallicity errorsare large, and we cannot rule our a bimodal distribution withthe ‘wings’ of the errors simulating a smoother transition be-tween the two behaviors, although this would require a cor-relation between metallicity errors, measured radial velocityand velocity dispersion.The metal rich stars are best interpreted as part of thebar/bulge structure. Their kinematics show evidence of ro-tational support and bulk rotation and are consistent with datafrom the BRAVA survey in this region. The behavior of the more metal poor component, showing high velocity disper-sion, and low to zero velocity relative to the Sun may be ex-plained by a classical bulge or by inner halo stars. With amean metal abundance of [Fe/H] ∼ − . these stars appearto be best interpreted (at least provisionally) as an inner halopopulation, although bulges can of course be metal poor aswell. This is consistent with the observations by Zoccali et al.(2008) and Babusiaux et al. (2010), albeit for a single sight-line. However, the BRAVA data show no classical bulge com-ponent fitting their dynamical model (Shen et al. 2010). Onepossibility is that by selecting M giants and using the CalciumTriplet as their main radial velocity indicator, BRAVA may bebiased against lower metallicity stars and therefore preferen-tially miss the high σ component.The properties of galactic globular clusters present an inter-esting analogy with what is observed here: metal rich clusters,with mean [Fe/H] of ∼ − . are believed to be associatedwith the bulge and are supported at least in part by rotation,whereas inner halo clusters have mean [Fe/H] of ∼ − . theirkinematics are dominated by random motions and at mostvery slow rotation. It is tempting to speculate that the twocomponents we see in our data are analogous to the metal-poor and metal-rich globular clusters, whose properties theyappear to share to some extent (cf., Babusiaux et al. 2010 fora similar two-component model for the bulge). Ortolani et al.(1995); Zoccali et al. (2003) and Clarkson et al. (2008) haveargued, on the basis of isochrone fits to bulge globular clustersand field stars, that the bulge formed nearly coevally with thehalo. Most globular clusters in the inner halo formed within ± Gyr of each other (Marin-Franch et al. 2009). If this ap-plies to bulge stars as well, it would imply a rapid star forma-tion process, at least for the inner regions ( < kpc) of theMilky WayIf this is the case, the smooth transition between the metal-rich and metal-poor subsystems, with a ‘turn-over’ point at[Fe/H] ∼ − may imply that the bulge and halo componentsare continuous and that there is no clear dichotomy betweenthe two (modulo the large errors in metal abundance). Thiswould be consistent with the BRAVA result that the bulge wasformed (in a dynamical sense) from secular evolution at highredshift. As long as the stars also formed rapidly, the ob-served α -element enhancements are not in disagreement withthis hypothesis.This publication makes use of data products from the TwoMicron All Sky Survey, which is a joint project of the Univer-sity of Massachusetts and the Infrared Processing and Analy-sis Center/ California Institute of Technology, funded by theNational Aeronautics and Space Administration and the Na-tional Science Foundation. R. Michael Rich acknowledgessupport from grant AST 0709479 from the National ScienceFoundation. We would like to thank the GAMA survey forallowing us to use two hours of AAT time to carry out thisprogram. We would also like to thank Scott Croom, DavidWilkerson ,Rob Sharp and Will Sutherland. Facilities:
AAT (2dF).
REFERENCESAlves-Brito, A., Melendez, J., Asplund, M., Ramirez, I. & Yong, D. 2010,A&A, 513, 35Babusiaux, C. & Gilmore, G. 1995, MNRAS, 358, 1309Babusiaux, C. et al. 2010, A&A, 519, 77 Ballero, S. K., Matteucci, F., Origlia, R. & Rich, R. M. 2007, A&A, 467, 123Bensby, T. et al. 2010, A&A, 521, L57Blitz, L. & Spergel, D. N. 1991, ApJ, 379, 631Bullock, J. S. & Johnston, K. V. 2005, ApJ, 635, 931 ouble red clump kinematics and abundances 5ouble red clump kinematics and abundances 5