Mapping the Galactic Halo with blue horizontal branch stars from the 2dF quasar redshift survey
Roberto De Propris, Craig D. Harrison, Peter J. Mares, CTIO, Cornell University
aa r X i v : . [ a s t r o - ph . GA ] J un Preprint typeset using L A TEX style emulateapj v. 11/10/09
MAPPING THE GALACTIC HALO WITH BLUE HORIZONTAL BRANCH STARS FROM THE 2DF QUASARREDSHIFT SURVEY R OBERTO D E P ROPRIS , C RAIG
D. H
ARRISON P ETER
J. M
ARES ABSTRACTWe use 666 blue horizontal branch (BHB) stars from the 2Qz redshift survey to map the Galactic halo in fourdimensions (position, distance and velocity). We find that the halo extends to at least 100 kpc in Galactocentricdistance, and obeys a single power-law density profile of index ∼ − . in two different directions separatedby about 150 ◦ on the sky. This suggests that the halo is spherical. Our map shows no large kinematicallycoherent structures (streams, clouds or plumes) and appears homogeneous. However, we find that at least 20%of the stars in the halo reside in substructures and that these substructures are dynamically young. The velocitydispersion profile of the halo appears to increase towards large radii while the stellar velocity distribution is nonGaussian beyond 60 kpc. We argue that the outer halo consists of a multitude of low luminosity overlappingtidal streams from recently accreted objects. Subject headings:
Galaxy: formation — Galaxy: halo — stars: horizontal-branch INTRODUCTION
The motion of old stars preserves the fossil record of theearliest phases of formation in the Milky Way, as was realizedin the two seminal papers by Eggen, Lynden-Bell & Sandage(1962) and Searle & Zinn (1978). The picture that has sinceemerged is one where the Galaxy has been built via a seriesof accretions and mergers (e.g., Freeman & Bland-Hawthorn2002), in agreement with hierarchical structure formation sce-narios (Johnston et al. 2008; Cooper et al. 2009). Consistentwith these theoretical expectations, the halo is thickly popu-lated by debris from past and on-going accretion events (e.g.,Ibata, Gilmore & Irwin 1994; Belokurov et al. 2006). Promi-nent debris features have also been observed in the halosof M31 (Ibata et al. 2001, 2007; McConnachie et al. 2009),M33 (Ibata et al. 2007; McConnachie et al. 2009) and in othernearby spirals (Martinez-Delgado et al. 2010).At the same time, the inner halo (at galactocentric dis-tance R GC < kpc) of the Milky Way contains an old,dynamically smooth and metal-poor component, which wasprobably formed by rapid early merging or violent relax-ation (Carollo et al. 2007, 2009; Bell et al. 2008). While thebuildup of most of the inner halo is believed to take placerapidly, the outer regions of the Milky Way should grow moreslowly via the disruption of dwarf galaxies: the outer halo istherefore expected to be quite inhomogeneous and dominatedby large streams, clouds and plumes from infalling objects(Johnston et al. 2008; Cooper et al. 2009). The main questionwe need to address is the relative role of mergers and accre-tion vs. in situ formation, i.e., whether the observed struc-tures represent the main mechanism by which the Galaxy wasformed or are only a comparatively minor contribution overan ancient stellar component.In order to clarify these issues we explore the space dis-tribution and kinematics of the outer halo using Blue Hori-zontal Branch (BHB) stars. These objects have often beenused as tracers of the Milky Way halo (e.g., Yanny et al. 2000;Xue et al. 2008; Brown et al. 2010). BHB stars are compara-tively bright and reliable standard candles, are more commonthan other commonly used probes (such as Carbon stars or RR Cerro Tololo Inter-American Observatory, La Serena, Chile Department of Astronomy, Cornell University, Ithaca, NY, USA
Lyrae) and are representative of the old and metal poor stellarpopulation that constitutes the Galactic halo. These stars canbe easily identified spectroscopically, even at comparativelymoderate resolution, but are sufficiently rare that extensive ra-dial velocity surveys are needed to obtain a significant sample(Yanny et al. 2000; Xue et al. 2008; Brown et al. 2010).Here we exploit a large suite of archival spectroscopy fromthe 2dF Quasar Redshift Survey (2Qz – Croom et al. 2004) toidentify a sample of 666 BHB stars in the halo out to 100 kpcand explore its structure and kinematics. We use this datasetto set limits to the size of the Milky Way halo, its shape, pres-ence of streams, degree of substructure and the velocity dis-persion profile out to large radii. The next section describesthe 2Qz survey and how we identified BHB stars. Section 3discusses the 4D map of the Galactic halo we produce fromthese data. Section 4 examines the question of substructurein the halo and the final section analyzes the halo kinematicsand discusses the results in the context of galaxy formationmodels, especially as apply to the Milky Way. THE 2QZ DATA: IDENTIFICATION OF BHB STARS
The data for this project come from archival observationsof the 2Qz survey. The 2Qz obtained spectra for ∼ , A-colored point sources with < b J < . selected fromSupercosmos survey photometry (Hambly et al. 2001). Thetargets lie in two ◦ × ◦ strips on the sky, one on the celestialequator between 09h 30m < RA <
14h 30m and − . ◦ <δ < +2 . ◦ and the other centred in the Southern GalacticCap between 22h 00m < RA <
03h 00m and − . ◦ <δ < − . ◦ . We refer to these as the Northern and Southernsamples, respectively.The targets were selected on the basis of their u , b J and r F colors (photographic photometry from the SuperCosmos sur-vey – Hambly et al. 2001) to lie in the parameter space cov-ered by known quasars (see Croom et al. 2004 for further de-tails): targets were selected to have b J < . and − . ≤ u − b J ≤ . and − . ≤ b J − r F ≤ . , excluding a box withcolors . ≤ u − b J ≤ . and ≤ b J − r F ≤ . which con-tains main sequence stars. Spectroscopy for these objects wasobtained during 1998-2003 with the Anglo-Australian 3.9mTelescope in Coonabarabran, NSW, Australia, using the 2 de-gree field (2dF) multi-object spectrograph (Lewis et al. 2002).The spectra cover the entire optical range ( < λ < De Propris et al. -0.4 -0.2 0 0.2 0.4g-r00.20.40.60.811.21.4 u - g F IG . 1.— The color distribution of BHB stars identified in the 2Qz survey.The original Supercosmos colors have been transformed to the SDSS usingstars in common between the two surveys. ˚A) at a resolution of about 2000, with exposure times of atleast one hour per target, although a fraction of the objectswere exposed for considerably longer, if they happened to liein a region where the 2dF tiles used by the survey overlapped(to insure greater completeness) or the field was reobserved(because of lower than expected signal).As with all color-selected surveys, 2Qz suffers from signif-icant contamination from QSO-colored stellar objects, suchas white dwarfs, disk A stars, blue stragglers and BHB stars,although this is minimized by observing at high galactic lat-itude. The data were reduced and redshifted via a semi-automated technique and whenever a stellar redshift ( z = 0 )was obtained, the object was discarded from further analysisbut placed in the database. We proceeded to retrieve all stellarspectra from the database and classify them on the basis of theequivalent widths of the H γ and H δ Balmer lines, to identifya sample of 666 bona fide
BHB stars. Following Yanny et al.(2000); Xue et al. (2008); Brown et al. (2010) we first mea-sured radial velocities for all star by cross correlating withsynthetic templates from the library of Munari et al. (2005):the templates had temperatures and gravities typical of A-starsand field horizontal branch stars (Beers et al. 2001). We thenfit Gaussian curves to the H γ and H δ lines and measured thewidth of the Gaussian fit at 20% of the normalized contin-uum level: this indicator D . has been shown to be a gooddiscriminator between BHB stars, blue stragglers and othercontaminants (Pier 1983; Yanny et al. 2000). We only usedspectra that we deemed to be of sufficient quality to allow asecure classification. In order to be included in our sample ra-dial velocities had to be determined to within 50 km s − andthe H γ and H δ widths had to have errors of less than 20%. Inorder to classify stars as bona fide BHB stars we require thatthe mean D . , from both lines, lie between 17 and 31 ˚A (asin Pier 1983; Yanny et al. 2000, leaving a total of 666 BHBstars in our sample.Figure 1 shows the distribution of the stars in the u − g vs. g − z plane, where we transformed our Supercosmos u − b J and b J − r F colors to the Sloan system by using stars in com-mon in the equatorial region shared by 2Qz and the SDSS.This is at least qualitatively similar to the color distributionof stars in previous work (e.g., compare Fig. 1 in Brown et al.2010) and suggests that our sample is comparable to thoseused in previous studies, in terms of selection criteria anddegree of contamination from blue stragglers and other A-colored objects.As our targets span a relatively narrow range in colors (as selected by the 2Qz survey) we assumed an absolute mag-nitude of M b J = 0 . ± . which is typical for BHB stars(Layden et al. 1996). We finally used these distances, theknown sky positions and the measured radial velocities toplace all our stars on a cylindrical coordinate system at restwith respect to the centre of the Galaxy, assuming Solarpositions as in Dehnen & Binney (1998). This yields a 4-dimensional (position, distance and radial velocity) map ofthe galactic halo in two widely separated ( ◦ ) lines of sight. THE 4D STRUCTURE OF THE MILKY WAY HALO
In Figure 2 we plot the 4D map of the Galactic halo we pro-duced, projected along the three most relevant dimensions. Itis clear from this figure that the halo of the Milky Way extendsto at least 100 kpc from the Galactic centre, and likely well be-yond (the edge of the map is set by the magnitude limit of 2Qzdata), in both directions we survey. This is considerably largerthan previously believed and comparable to the large metal-poor halo observed in the Andromeda galaxy (Chapman et al.2006; Kalirai et al. 2006; Koch et al. 2008); it would includeseveral of the dwarf galaxy satellites (including the Magel-lanic clouds) within the Galaxy’s stellar halo. As a matter offact the Sextans dwarf is visible in Fig. 2 at x ∼ − kpc and y ∼ +50 kpc. Such large metal-poor halos may be ubiqui-tous (e.g., in NGC 3379 – Harris et al. 2007) and may be acommon byproduct of early galaxy formation.In previous work, the SDSS has identified BHB stars out to60 kpc from the Galactic centre (Xue et al. 2008), while theHypervelocity Stars Survey (Brown et al. 2010) found a BHBstar sample to a distance of 75 kpc. The Spaghetti survey(Morrison et al. 2000; Starkenburg et al. 2009) has observedhalo red giants to a distance of 100 kpc, and star counts inthe COSMOS field identify a halo component to a distanceof 80 kpc (Robin et al. 2007). Our study reaches deeper thannearly all these and covers a larger field of view than all exceptthe SDSS (it is comparable to the Hypervelocity Stars Surveycoverage). However, we sample the BHB stars more denselyas all potential targets have been targeted by the 2Qz survey,although of course we are not able to classify all stars. Sincewe cover nearly diametrically opposite areas on the sky, weargue that the detection of the stellar halo in our data is not dueto possible diffuse structures on the sky that accidentally lie inour line of sight (as for pencil-beam studies such as COSMOSor the Spaghetti survey) and the Milky Way halo truly extendsto large radii.Figure 3 shows the radial density profile of the halo alongboth directions we survey. In both cases we obtain a good fitto a single power law of index R − . ± . . This is somewhatshallower than the ∼ R − found by Morrison et al. (2000)and predicted by theory, but is in good agreement with pre-vious measurements using BHB stars by Xue et al. (2008)and Brown et al. (2010). We are of course incomplete in thatwe cannot detect and identify all BHB stars in 2Qz. Thisincompleteness is a complex function of our ability to reli-ably classify stars as a function of spectroscopic signal-to-noise. Naively, we would preferentially miss the most dis-tant objects, that would tend to make the radial profile steeperthan it actually is, while contamination from blue stragglerswould tend to make the profile flatter. The similarity betweenour profiles and those derived by (e.g.) Yanny et al. (2000);Xue et al. (2008) and Brown et al. (2010) argues that our con-tamination fraction (from blue stragglers) and completenessare not too different from previous samples.In any case, this should not affect our differential measure-apping the Galactic Halo 3 -20 0 20 40 60Distance along Sun - Galaxy centre line X (kpc)-100-80-60-40-20020406080100 H e i gh t Z ov e r t h e G a l ac ti c p l a n e ( kp c ) v GSR < -300 km s -1 -300 < v GSR < -200 km s -1 -200 < v GSR < -100 km s -1 GSR < -100 km s -1 GSR < +100 km s -1 +100 < v GSR < +200 km s -1 +200 < v GSR < +300 km s -1 v GSR > +300 km s -1 F IG . 2.— The 4D map of the Galactic halo, with stars projected on the Galactic plane (along a line connecting the Sun to the Galactic centre) on the x-axis andvs. their height above the plane on the z-axis. The stars are color-coded by their Galactocentric radial velocity, as indicated in the figure legend. ρ ( s t a r s kp c - ) Northern sampleSouthern sampleR -2.6 R -2.4 F IG . 3.— Radial density profiles for BHB stars in the halo and best power-law fits to the Northern and Southern samples separately. ment of the radial density profile for the two individual sight-lines we survey. The similarity in the profile slope then arguesthat the stellar halo of the Galaxy is spherical (Majewski et al.2003; Smith, Wyn Evans & Ahn 2009), although more sight-lines would be helpful to obtain a more precise estimate forthe halo’s axial ratio.In a recent survey of RR Lyrae in SDSS stripe 82,Watkins et al. (2009) and Sesar et al. (2010) find that the haloradial profile becomes very steep ( R − ) at galactocentricradii larger than 40 kpc and the the outer halo seems to bedominated by a large structure known as the Virgo-Aquilacloud, which is either an infalling dwarf galaxy or a tidalstream. However this is not observed in our data, or thoseof Brown et al. (2010). One possibility is that the RR Lyraedistribution is more concentrated towards the Galactic centre,as RR Lyrae tend to be more metal rich than BHB stars andthe halo appears to have an abundance gradient (Carollo et al.2007, 2009). This would create an apparent break in the radialprofile, reflecting the lower overall metallicity of the sampleat large distances. SUBSTRUCTURE IN THE HALO
Inspection of Fig. 2 also shows that there are no obviouslarge kinematically coherent features, such as streams, plumesor shells, in our 4D map of the Galactic halo. We identifythree possible structures: one is the Sextans dwarf, as pre-viously mentioned. A plume of infalling stars is present inthe top right-hand corner of Fig. 2 at x ∼ +40 kpc and y ∼ +80 kpc. This may be related to the Virgo-Aquila struc-ture. Finally, a small clump of objects is observed in the di-rection of Piscis Austrinus at x ∼ +40 kpc and y ∼ − kpcand may be an undiscovered dwarf. Nevertheless, we do notappear to observe the large streams expected in simulations(Johnston et al. 2008; Cooper et al. 2009) and the outer haloappears to be more homogeneous than predicted.We can quantify the presence of streams or oth-erwise using the Great Circle Stream method ofLynden-Bell & Lynden-Bell (1995) in Figure 4. Wecompute the excess radial energy of each star compared toa smoothed Galactic potential and compare the results to arandomized sample, where we scramble the velocities, butnot the positions, of stars, 1000 times. Streams would showup in this figure as long lines of stars. It is easy to see thatour sample contains no significant stellar streams. The largemetal poor outer halo appears to represent an extension of thesmooth metal poor inner halo structure (Carollo et al. 2007,2009; Bell et al. 2008).However, lack of large streams does not mean lack of sub-structure. Bell et al. (2008) find that about 40% of stars inthe inner halo are substructured, even if the only prominentstream in their data is the well known Sagittarius stream,while Starkenburg et al. (2009) find that around 25% of thestars in the Spaghetti survey are more clustered than expectedfrom a random distribution. We apply the 4-distance methodused by Starkenburg et al. (2009) to our data in Figure 5.This is essentially a version of the correlation function for De Propris et al. -10010 E r ( / k m s - ) -10010Data Random F IG . 4.— The Great Circle Stream method applied to stars in our sample.We plot the excess radial energy of each stars with respect to a smoothedGalactic potential vs. its radial distance from the centre of the Galaxy. Theleft hand panel shows the actual data, where the right hand panel is a random-ization. No streams are observed in these figures as long lines of stars havingthe same excess radial energy. N p a i r s / N r a ndo m F IG . 5.— The excess fraction of stellar pairs over random (see text) in 4-dimensional distance for our sample of BHB stars. σ level we find that at least 20% of our stars are morepaired than random. This agrees, broadly, with the estimatesby Bell et al. (2008) for the inner halo and Starkenburg et al.(2009) for the outer halo. In addition we detect a decrease inthe correlation strength at small 4-distance. This is expected ifthe outer halo is dynamically young and suggests the presenceof numerous streamlets and a complex structure, too weak tobe resolved by our data. We present further evidence to thiseffect from an analysis of halo kinematics. KINEMATICS OF THE HALO
The kinematics of metal-poor stars in the halo yieldinformation on the earliest phases of galaxy formation(Eggen et al. 1962) and the shape of the Galactic potential.Until recently, known samples of halo stars were small, espe-cially at large galactocentric distances. Battaglia et al. (2005)used a heterogeneous sample of red giants, BHB stars, glob-ular clusters and dwarf galaxies to trace the velocity disper- -600-400-2000200400600 V G S R ( k m s - ) σ G S R ( k m s - ) Northern SampleSouthern Sample F IG . 6.— Top panel:
Run of radial velocities of BHB stars as a functionof Galactocentric distance.
Bottom panel:
The radial velocity dispersion pro-file of the Galactic halo in both Northern and Southern samples. Each bincontains equal numbers of stars and the velocity dispersion and its error arecalculated using a maximum likelihood method. The error bars on points forthe x-axis represent the size of the bin used. sion profile of the Galaxy out to ∼ kpc, finding a mildlydeclining profile. Using BHB stars in SDSS Xue et al. (2008)found a flat or mildly declining profile out to 60 kpc and there-fore inferred a large mass of the Milky Way. Brown et al.(2010) also derive a mildly declining profile out to 75 kpc us-ing BHB stars in the Hypervelocity Star Survey. In all thesecases, the fits depends strongly on the accuracy of the last(most distant) data points, which generally contain fewer ob-jects.We use our data to derive the velocity dispersion profile inboth Northern and Southern samples separately. We use binscontaining equal numbers of stars and calculate the velocitydispersion and its error following a maximum likelihood ap-proach (Walker et al. 2006). Figure 6(a,b) plots our results.While we are in reasonable agreement with previous workover the range where we overlap, we find evidence of a risingvelocity dispersion at large radii, in both fields we study. Thisis unlike the results of Battaglia et al. (2005) and Brown et al.(2010), although we reach farther into the distant halo thanthey do. In the former case, the heterogeneous sample andits relatively small size may cause part of the difference, as itis known that different tracers have different kinematics (e.g.,Kinman et al. 2007). In the latter case, we and Brown et al.(2010) use the same tracers, but while our stars are concen-trated in two narrow strips, Brown et al. (2010) uses a wideregion selected from the SDSS. We have more stars (by a fac-tor of about 3) than they do at large ( R > kpc) distances.One possibility is that we sample this regime more densely,especially if the halo is as inhomogeneous as theory suggestsand as the evidence from Fig. 5 also argues.Is it possible that our result may be affected by contamina-tion from blue straggler stars. Since these objects are 2 to 3magnitudes fainter than BHB stars, blue stragglers will con-taminate the bins at R GC > kpc with objects that truly lieat distances of 10 to 30 kpc. The selection procedures we de-scribe above should exclude most contaminants but even if weare as effective as previous studies, ∼ of our sample mayconsist of blue stragglers. In order to assess the effects of con-tamination, we have used a model distribution function for thevelocity dispersion profile of our Galaxy from van Hese, Baes& Dejonghe (2009), with Galactic mass of 1.1 × M ⊙ andhalo scale radius of 40.5 kpc from Dehnen & Binney (1998).We sampled 100 stars at distances of 10 to 100 kpc with aGaussian random deviate having σ ( r ) from van Hese et al.apping the Galactic Halo 5 σ ( k m s - )
0% Blue Stragglers10%20%30%40%0 50 100 150R (kpc)150155160165170175180 σ ( k m s - )
0% Blue Stragglers10%20%30%40% F IG . 7.— The effect of BHB star contamination on the velocity dispersionprofile of the Galaxy. The top panel is for a distribution with no velocityanisotropy while the bottom panel assumes that all velocities are tangentialat large radii. See text for details. The degree of contamination assumed isindicated in the legend. (2009) and zero mean velocity. We adopted two cases: onewith no velocity anisotropy β = β inf = 0 and one with fullytangential radial velocities at infinity ( β = 0 ; β inf = 1 ). Foreach of these we adopted a variable contamination, between10% and 40%, from blue straggler stars, which we assumedto be 2.5 mag. fainter than BHB stars. We replaced this frac-tion of stars in each bin with R GC > kpc, with blue strag-glers at the appropriate distance, with velocities sampled froma Gaussian random deviate with σ ( r ) from van Hese et al.(2009), at the appropriate r for a contaminating blue straggler.The results of this exercise are shown in Figure 7. Althoughthe blue stragglers increase the measured velocity dispersionat large radii (compared with the theoretical profile), they donot produce a flat or rising velocity dispersion profile (as inthe bottom panel of Fig. 6) even at 40% contamination, andeven for the relatively flat σ ( r ) profile produced by the fullyisotropic model.In Figure 8 we plot the histograms of radial velocities for allstars in 6 bins, containing equal numbers of objects and cov-ering a range of distances. We see that while the distributionsare acceptably Gaussian within the inner 60 kpc, we observeboth an increase in dispersion and a loss of Gaussianity in thetwo outer bins. This is consistent with what we observe in Fig. 5 and 6, where we find indications of an increasing ve-locity dispersion and a dynamically young and substructuredhalo. If the outer halo consists of a myriad of streamlets, ac-creted more or less recently from the disruption of satellitesor low luminosity galaxies, we would expect evidence of dy- N u m b e r o f s t a r s -400 -200 0 200 400V GSR (km s -1 )0102030 N u m b e r o f s t a r s -400 -200 0 200 400V GSR (km s -1 ) -400 -200 0 200 400V GSR (km s -1 )
10 < R < 25 kpc 25 < R < 36 kpc 36 < R < 46 kpc46 < R < 59 kpc 59 < R < 77 kpc 77 < R < 95 kpc F IG . 8.— Histograms of the velocity distribution for 6 bins containing equalnumbers of stars (the distance ranges sampled are indicated in figure legends).The two more distant bins appear to have a much less Gaussian distributionthan the four bins at R < kpc. namical youth and tidal heating, as we observe in Fig. 4 and5. A possible alternative explanation is that the dark halo isvery massive. Gnedin et al. (2010) has recently argued for avery massive halo based on the flatness of the velocity dis-persion profile. However, in this case the mass to light rationeeded would exceed several hundred. Tidal heating may alsoproduce an increasing velocity dispersion, as is observed insome dwarf galaxies (Munoz, Majewski & Johnston 2006).These explanations, while possible, would not probably pro-duce the non Gaussian velocity dispersion observed at largeradii.The combined evidence from the space distribution andkinematics of BHB stars points to a very large and metal poorhalo, whose outer regions appear to have been accreted rel-atively recently from low luminosity satellites, analogous tothe copious tidal debris observed in M31. Although largestreams are not observed the data appear to be in compara-tively good agreement with theoretical predictions, althoughit is possible that minor mergers are more important than ex-pected. More observations as well as models to truly ‘ob-serve’ the simulated halos in the same fashion as the data anddirectly compare theory and reality will be needed to refineour understanding of the formation of the Milky Way.We would like to thank the anonymous referee for a veryhelpful report which clarified many points in this paper. Wethank the 2dF Quasar Survey for making their data availableto the community and in particular Scott Croom for answeringmany questions on the data. We warmly thank all the presentand former staff of the Anglo-Australian Observatory for theirwork in building and operating the 2dF and 6dF facilities. The2QZ and 6QZ are based on observations made with the AATand the UKST. Facilities: